&EPA
EPA/600/R-10/009
December 2008
Evaluion of the Effectiveness of Coatings in Reducing
Dislodgeable Arsenic, Chromium, and Copper from
CCA Treated Wood
Prepared by:
Mark A. Mason
U.S. Environmental Protection Agency
Office of Research and Development
National Risk Management Research Laboratory
Research Triangle Park, NC
Victor D'Amato, PE
Arcadis, Geraghty, & Miller, Inc.
Durham, NC
Leonard A. Stefanski, Ph.D.
Department of Statistics
North Carolina State University
Raleigh, NC
-------
Notice
The U.S. Environmental Protection Agency through its Office of Research and
Development funded and managed in the research described here under
contract number EP-C-014-023 to Arcadis, Geraghty & Miller, Inc., and contract
number EP-C-00-133 to Dr. Leonard Stefanski. It has been subjected to the
Agency's review and has been approved for publication as an EPA document.
-------
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
List of Figures
List of Appendices
List of Acronyms
Executive Summary
1. Project Description
1.1 Background
1.2 Project Objectives
1.3 Experimental Design
1.4 Data Quality Objectives
2. Sampling and Analysis Protocol
2.1 Selection of Wood Sources
2.2 Preparation and Characterization of Wood Sources
2.2.1 Source Wood Harvesting and Preliminary Characterization
2.2.2 Baseline Sampling
2.2.3 Wood Core Sampling and Analysis
2.3 Minideck Construction
2.4 Selection of Coatings
2.5 Coating Application
2.6 Outdoor Weather Site Setup
2.7 Weather Data Collection
2.8 DCCA Measurement Methods
2.8.1 Sampling and Extraction Methods
2.8.1.1 Wipe Blank Study
2.8.1.2 Spiking Study
2.8.2 Wipe Sampling Events
2.8.2.1 Baseline Sampling of Boards and Minidecks
2.8.2.2 Subsequent Wipe Sampling Events
2.8.3 Wipe Sampling Methods
x
xii
xiii
1
9
9
10
10
13
15
15
16
16
20
22
23
27
28
30
32
35
35
35
36
37
38
39
39
IV
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
2.8.3.1 EPA Acid-Wash, Rinse, and Saturate with Dl Water Wipe
Preparation Technique (A2 Method) 39
2.8.3.2 EPA 2X Dl Water Wipe Preparation Technique (2X
Method) 40
2.8.3.3 Standard EPA Wipe Method (Adaptation of CPSC Staff
Method) 40
2.8.4 Wipe Extraction and Analysis Techniques 41
2.8.5 Differences between EPA and CPSC Procedures 42
2.8.6 Calculation of DCCA from Extraction Fluid Concentrations 43
2.9 Preparation and Analysis of Coating Samples 43
2.10 Archiving of ICP-MS Samples 44
2.11 Moisture Analysis of Wood Specimens 44
2.12 Miscellaneous Samples 44
2.13 Quality Control Samples 44
2.13.1 Positive (CCA-Treated, Uncoated) Controls 45
2.13.2 Negative (Untreated, Uncoated) Controls 45
2.13.3 Wipe Frequency (Rewipe, Abrasion) Controls 45
2.13.4 Analytical Laboratory Control Samples 46
2.14 Paint Chip Sampling 46
3. Study Results 49
3.1 Source Characterization and Sampling Events 49
3.1.1 Distribution of Baseline Data 49
3.1.2 Wood Core Sample Data 53
3.1.3 Coating Application 53
3.2 Coating Performance Data 54
3.2.1 DCCA vs. Time 54
3.2.2 Graphical Data Analysis 55
3.2.2.1 Single Minideck Data Plot 59
3.2.2.2 Data Plots for Other Minidecks 62
3.2.2.3 All Thirty-nine Minidecks Plotted Together 62
3.2.2.4 Averages from All Thirty-nine Minidecks 79
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
3.2.2.5 The Comprehensive Plot
3.2.3 Analysis of Variance of Coatings by Time
3.2.4 Analysis of Variance of Coating Pair Comparisons by Time
3.2.5 Analysis of Variance Modeling Results
3.2.5.1 Time Period Specific Modeling
3.2.5.2 Fixed Effects
3.2.5.3 Random Effects
3.2.5.4 Coating Comparisons
3.2.5.5 Statistical Model Sensitivity Analysis
3.2.5.6 Checking for Heteroscedasticity
3.2.5.7 Combined Data Modeling
3.2.6 Summary of Coating Comparison Results
3.3 Coating Appearance
3.4 Weather Data
4. Study Limitations, Exploratory Analyses and Future Studies
4.1 Substrate Characteristics
4.2 Geography/Climatic Effects
4.2.1 Exploratory Analysis of Potential Weather Effects
4.3 Coating Type and Selection
4.4 Surface Preparation, Coating Application and Reapplication
4.4.1 Exploratory Analysis of Coating and Paint Chip Samples
4.5 Test Specimen Dimensions
4.6 Abrasion Effects
4.7 CCA Transport from Treated Specimens
4.8 Sampling Methodology
4.8.1 Baseline (Pre-Coat) Measurement
4.8.2 Wipe Sampling Method Comparisons
4.9 Potential Future Studies
5. Quality Assurance and Quality Control
5.1 Assessing DQI Goals
81
81
85
89
89
90
91
91
93
94
96
98
101
101
107
107
108
109
115
117
118
120
120
122
124
125
127
127
129
129
VI
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
5.1.1 Precision
5.1.2 Accuracy and Bias
5.1.3 Completeness
5.2 Data Validation Summary
6. Conclusions
7. References
130
130
131
131
133
135
VII
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
List of Tables
Table 1-1. Data Quality Indicator Goals for Critical Measurements 13
Table 2-1. Wood Specimens Used to Construct Minidecks 26
Table 2-2. Selected Products for Evaluation 28
Table 2-3. Minideck Block Assignments (blocks correspond to those identified in
Appendix H) 31
Table 2-4. Vantage Pro Plus™ Weather Station Data 33
Table 2-5. NOAA-Generated Weather Data 34
Table 2-6. Wipe Blank Analyses 35
Table 2-7. Results from September 2003 Wipe Comparison Study 36
Table 2-8. Results of Spiking onto Glass 36
Table 2-9. Results of Spiking Wipes Directly 37
Table 2-10. Miscellaneous Samples Collected 44
Table 2-11. Composite Samples Taken for Paint Chip Sampling/Analysis 47
Table 3-1. Summary Statistics for Source Deck Specific Random Effects Analysis 49
Table 3-2. Comparison of Nominal, Source A, and Source C CCA Actives Composition 53
Table 3-3. Legend for Individual Minideck Data Plots 59
Table 3-4. DAs Data for Minideck 9-A, as used in Figure 3-13 62
Table 3-5. P-Values for Split-Plot Models with Homoscedastic Error Structures 90
Table 3-6. Coating Performance Rankings and P-Values of Tests Adjusted using
Dunnett's Multiple Comparison Procedure 92
Table 3-7. Sensitivity Analysis of 24-month Coatings Comparisons p-values 94
Table 3-8. Heteroceastic Analysis of Variance Models 95
Table 3-9 Mean Residual Variances from Table 3-8 95
Table 3-10. Summary of Results of Alternative Variance Modeling 97
Table 3-11. Coatings Comparisons from the Combined Data Modeling 97
Table 3-12. Summary of Estimated In(DAs) Difference and Associated Statistical
Significance between Each Coating and Control [#13 (uncoated)] Minidecks
at Each Time Interval 99
Table 3-13. Estimated Percent Reduction for Each Coating and Estimated 95 Percent
Confidence Intervals (color-coding same as in Table 3-12) 100
Table 3-14. Summary of Visual Observations of Minidecks 102
Table 3-15. Summary of Key Weather Parameters 103
Table 3-16. Summary of Temperature Measurements 103
Table 4-1. List of Weather Variables Considered in Statistical Analysis 110
VIM
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 4-2. Partial Correlations for Pairs with p-values < 0.10 111
Table 4-3. Pain/vise Comparisons of Coating Characteristics 116
Table 4-4. Raw Coating CCA Analyte Concentrations (raw/wet weight basis) 118
Table 4-5. Paint Chip CCA Analyte Concentrations 119
Table 4-6. Summary of Cross-Contamination and Blank Control Minideck Results 122
Table 5-1. Data Quality Indicator Goals for Critical Measurements 129
IX
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
List of Figures
Figure E-1. MinideckTest Site at EPA's Research Triangle Park, North Carolina,
Campus 2
Figure E-2. Sampling to Determine Dislodgeable Arsenic (DAs) Residues on CCA-
Treated Wood Using the CPSC Wipe Sampling Apparatus 3
Figure E-3. DAs vs. Time Determined on Four Specimens of Minideck 8C, Coated with
a Water-Based Stain 4
Figure E-4. DAs vs. Time Determined on Four Specimens of One of the Positive
Control Minidecks (no coating applied) 5
Figure 1-1. Wood Nomenclature 11
Figure 2-1. ERC Deck Map (Source A) 17
Figure 2-2. Views of ERC Deck 17
Figure 2-3. New Hill Deck Map (Source C) 18
Figure 2-4. Views of New Hill Deck 18
Figure 2-5. Specimen Identification and Baseline Sampling Scheme Example 20
Figure 2-6. Schematic of Minideck Construction 24
Figure 2-7. Photograph of Typical Minideck 25
Figure 2-8. Photograph of Minideck Site 32
Figure 2-9. CPSC Wipe Sampling Apparatus 38
Figure 3-1. Box Plot, Baseline DAs, by Coating, Source A 50
Figure 3-2. Box Plot, Baseline DCr, by Coating, Source A 50
Figure 3-3. Box Plot, Baseline DCu, by Coating, Source A 51
Figure 3-4. Box Plot, Baseline DAs, by Coating, Source C 51
Figure 3-5. Box Plot, Baseline DCr, by Coating, Source C 52
Figure 3-6. Box Plot, Baseline DCu, by Coating, Source C 52
Figure 3-7. Coating Application (total of triplicate minidecks on both A and C sources) 54
Figure 3-8. Average DAs vs. Time for All Coatings 56
Figure 3-9. Average DCr vs. Time for All Coatings 57
Figure 3-10. Average DCu vs. Time for All Coatings 58
Figure 3-11. Data Plot for Minideck 9A 61
Figure 3-12. Data from Minideck 1A, Data from Minideck 1B, Data from Minideck 1C 63
Figure 3-13. Data from Minideck 2A, Data from Minideck 2B, Data from Minideck 2C 64
Figure 3-14. Data from Minideck 3A, Data from Minideck 3B, Data from Minideck 3C 65
Figure 3-15. Data from Minideck 4A, Data from Minideck 4B, Data from Minideck 4C 66
-------
Figure 3-16. Data from Minideck 5A, Data from Minideck 5B, Data from Minideck 5C 67
Figure 3-17. Data from Minideck 6A, Data from Minideck 6B, Data from Minideck 6C 68
Figure 3-18. Data from Minideck 7A, Data from Minideck 7B, Data from Minideck 7C 69
Figure 3-19. Data from Minideck 8A, Data from Minideck 8B, Data from Minideck 8C 70
Figure 3-20. Data from Minideck 9A, Data from Minideck 9B, Data from Minideck 9C 71
Figure 3-21. Data from Minideck 10A, Data from Minideck 10B, Data from Minideck 10C 72
Figure 3-22. Data from Minideck 11 A, Data from Minideck 11B, Data from Minideck 11C 73
Figure 3-23. Data from Minideck 12A, Data from Minideck 12B, Data from Minideck 12C 74
Figure 3-24. Data from Minideck 13A, Data from Minideck 13B, Data from Minideck 13C
(all uncoated) 75
Figure 3-25. Composite Data Plot of All 39 Minidecks (note that scale and axis labels are
the same as for Figure 3-11) 77
Figure 3-26. Averages from All Thirty-nine Minidecks (See Section 3.2.2.4 for further
explanation) 78
Figure 3-27. Comprehensive Data Plot (note that scale and axis labels are the same as
for Figure 3-11) 83
Figure 3-28. Analysis of Variance Plot of Coatings by Time [each subplot has In(DAs) as
its y-axis and time, in years, as its x-axis] 84
Figure 3-29. Analysis of Variance Plot of Coating Pair Comparisons by Time [each
subplot has In(DAs) as its y-axis and time, in years, as its x-axis] 87
Figure 3-30. Solar Radiation Data Summary 104
Figure 3-31. Rainfall Data Summary 104
Figure 3-32. Temperature Data Summary 105
Figure 4-1. Partial Regression Plots of Coating x Weather Factor ([each subplot has
In(DAs) as its y-axis and the weather variable as its x-axis] 113
Figure 4-2. Average Spacer Board DAs versus Average CCA board DAs at Each of
Seven Time Intervals, plotted by Coating. 123
Figure 4-3. Average Spacer Board DAs versus Average CCA board DAs for Each of
Thirteen Coatings, plotted by Time Inteval (note that letter A = Coating 10,
B = Coating 11, C = Coating 12, and D = Coating 13). 124
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
XI
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
List of Appendices
Appendix A. Wipe Comparison Report
Appendix B. Specimen Characterization Data
Appendix C. Tabulated Data Summaries
Appendix D. Minideck Photo Log
Appendix E. Specimen Baseline Data
Appendix F. Wood Preparation Flow Sheets
Appendix G. Coating Application Data Form
Appendix H. Site Layout Plan
Appendix I. Weather Data Confirmation Details
Appendix J: Baseline Box Plots, by Board
Appendix K: Wood Core Sample Data
Appendix L. Cross-Contamination (Untreated) Sample Data
Appendix M. Negative Blank Control (BC) Sample Data
Appendix N. Baseline and PSA Wipe Sampling Data
Appendix O. Full Page Minideck Plots
Appendix P. Weather Data
Appendix Q. SEM Analysis of Paint Chips
Appendix R. Data Validation Reports
(These appendices are contained on CD-Rom)
XII
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
List of Acronyms
As
Av
AWPA
BL
CCA
CD-Rom
CL
CPSC
DCCA
DAs
DCr
DCu
Dl
DPD
DQI
EPA
ERC
FIFRA
H&S
HF
ICP
ICP-MS
ID
Ln
Mph
MS
MS/MSD
MSD
MSD
NCDC
NE
NFG
NOAA
NOPW
NW
Arsenic
Average
American Wood Preservers' Association
Baseline
Chromated Copper Arsenate
Compact Disk- Read Only Memory
Confidence Limits
Consumer Product Safety Commission
Dislodgeable CCA Wood Analytes
Dislodgeable Arsenic
Dislodgeable Chromium
Dislodgeable Copper
Deionized Water
Dew Point Depression
Data Quality Indicator
United States Environmental Protection Agency
Environmental Research Centerfold building)
Federal Insecticide Fungicide and Rodenticide Act
Health and Safety
Hydrofluoric Acid
Inductively Coupled Plasma
Inductively Coupled Plasma—Mass Spectrometry
Identification
Natural Logarithm
Miles Per Hour
Matrix Spikes
Matrix Spikes and Matrix Spike Duplicates
Matrix Spike Duplicates
Mean of the Squared Difference
National Climate Data Center
Northeast
National Functional Guidelines
National Oceanic and Atmospheric Administration
Number of Previous Wipes
Northwest
XIII
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
NWS National Weather Station
OLS On-Site Laboratory Support
OPP Office of Pesticide Programs
pcf Pounds per Cubic Foot
PEA Performance Evaluation Audit
PFA Perfluoroalkoxy
PM Project Manager
PSA Primary Sampling Area
PTFE Polytetrafluoroethylene
QA/QC Quality Assurance and Quality Control
QAM QA Manager
QAO Quality Assurance and Quality Control (QA/QC) Officer
QAPP Quality Assurance Project Plan
RDU Raleigh—Durham International Airport
RPD Relative Percent Difference
RSD Relative Standard Deviation
RSD/RPD Relative Standard Deviation or Relative Percent Deviation
RTP Research Triangle Park
SAP Scientific Advisory Panel
Std. Dev. Standard Deviation
STL Severn Trent Laboratory
SYP Southern Yellow Pine
TFE Tetrafluoroethylene
TSPW Time (Months) Since the Previous Wipe
U.S. United States
WA Work Assignment
WAL Work Assignment Leader
WAM Work Assignment Manager
UV Ultraviolet
XIV
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
xv
-------
Executive Summary
EPA conducted a study to evaluate the effect of coatings on dislodgeable arsenic,
chromium, and copper residues on the surfaces of chromated copper arsenate (CCA)
treated wood. Dislodgeable CCA (DCCA), determined by wipe sampling the wood
surfaces, was the primary evaluation criterion for the coatings tested in this study, due
to the potential for ingestion of CCA chemicals by hand-to-mouth activities of young
children that contact CCA treated wood. The study was conducted by EPA's Office of
Research and Development in Research Triangle Park, North Carolina, in support of
the Office of Pesticide Programs (OPP) and in collaboration with the staff of the
Consumer Product Safety Commission (CPSC), who used a similar protocol to
evaluate several of the same products at a site in Gaithersburg, Maryland.
EPA risk assessors, the coatings and wood treating industries, and citizens who may
be concerned about contact with CCA treated wood need sound information regarding
the impact of coatings on dislodgeable arsenic residues. Although no longer produced
or available for most consumer residential uses, CCA treated wood has been used
extensively for construction of decks and play structures that may have a long service
life and therefore, may pose a potential exposure route for CCA residues for many
years to come.
Locally purchased deck treatment products suitable for application by homeowners
were applied per manufacturers' instructions to purpose-built test decks. Application of
the coatings to CCA-treated decking initially reduced the amount of dislodgeable
arsenic, chromium, and copper residues, and the effectiveness decreased for all
products tested over the two-year period of the study. The significance of the results
reported in this study must be viewed in light of the limitations of the study design; most
notably that the study evaluated the effectiveness of a relatively small convenience
sample of the deck-coating products available for consumers at the time the study was
initiated and the study was conducted on small test decks in the relatively mild climate
of central North Carolina. Since the test decks were not subjected to the normal wear
and tear stresses that full-scale, in-use decks experience, the results of this study may
be indicative of the maximum period of effectiveness expected from the products that
were tested.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Two sources of weathered CCA-treated southern yellow pine (SYP) were harvested
from in-service decks and used to construct a series of miniature decks ("minidecks"),
onto which one of twelve coatings was applied per manufacturer's instructions.
Products included oil- and water-based sealants and stains, two porch and deck paints,
and two products that were advertised to encapsulate CCA treated wood. Instructions
for ten of the products recommended surface preparation with a cleaning agent
followed by a rinse prior to application of the coating, instructions for one product
recommended only a water rinse, and instructions for one product recommended
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
neither cleaning nor rinsing prior to application. Instructions for six of the products
recommended two applications of the product.
Each minideck contained fourCCA-treated specimens: two from a source deck which
had been in service for two years prior to the study and two from a source deck in
service for seven years. Each coating was applied to three minidecks. There were also
three positive control (CCA treated, uncoated) minidecks and one negative control
(untreated, uncoated) minideck for a total of 40 minidecks. After coating, the minidecks
were subjected to outdoor weathering at a controlled site in North Carolina where
climate measurements were recorded on a near-continuous basis (Figure E-1).
Figure E-1. Minideck Test Site at EPA's Research Triangle Park, North Carolina,
Campus
Dislodgeable arsenic (DAs), chromium (DCr), and copper (DCu) were determined at
specified intervals: baseline (pre-coat and pre-rinse or wash), and after 1, 3, 7,11, 15,
20 and 24 months. The DCCA residues were determined by wiping deck surfaces with
a polyester wipe attached to a 1.1 kg weight that slides between rails of a wipe
apparatus, as developed by the staff of the CPSC (Figure E-2). The arsenic, chromium
and copper were extracted from the wipe in a nitric acid solution and the total amount
of each metal in the extract was determined by inductively coupled plasma-mass
spectrometry (ICP-MS). Baseline samples on uncoated control minidecks were taken
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
before rinsing the minideck surfaces using a mild pressure wash. Thus, the baseline
DCCA results provide insight into the impact of rinsing CCA-treated decking.
Figure E-2. Sampling to Determine Dislodgeable Arsenic (DAs) Residues on CCA-
Treated Wood Using the CPSC Wipe Sampling Apparatus
The statistical analyses used to evaluate coating performance compare dislodgeable
CCA residues determined on the uncoated CCA controls with CCA residues
determined for the coated minidecks at each time period. For completeness and to
establish the robustness of the study's major conclusions, the data were analyzed
using a variety of alternative statistical models.
The amount of DAs versus time for each specimen on each minideck exhibited some
immediately recognizable trends. Each coating, as well as the positive controls
(uncoated CCA-treated minidecks), showed a significant decrease in DAs between
baseline (prerinse and precoat) and samples taken 1 month after coating. The rinsing
preparation step for the uncoated CCA controls initially reduced the DAs by about 71
percent. Thus, the rinse step provides some reduction in DAs; however, the reduction
is relatively small compared to the reduction attributable to most of the coatings. The
results indicate that coating (using any of the coatings tested) mitigates DAs to some
degree in excess of the reduction associated with rinsing alone, although not always at
3
-------
a statistically significant level. Overtime, DAs increased as specimens weathered
after they were coated. However, the time trend is generally uneven, because the
amount determined at any wipe sampling event is an imprecise snapshot in time of
the effects of processes that decrease and processes that increase the amount of
dislodgeable residues -wipe sampling, and presumably, precipitation remove
residues, whereas weathering and time are expected to adversely affect coating
performance and result in increased residues.
Example log-scale plots, Figures E-3 and E-4 below, illustrate changes in DAs with
time for two minidecks: deck 8C, coated with a water-based stain, and Deck 13A, an
uncoated CCA-treated positive control deck. Lines above the data in each plot indicate
the baseline DAs determined prior to rinsing and coating. The colors are keyed to
individual specimens, source deck, A (older) or C (newer), and end grain orientation,
up or down. The difference between the lines at the top of the graph and the first data
point at time = 1 month in Figure E-3 demonstrates the impact, after one month of
weathering, of rinsing and coating, whereas the difference in Figure E-4 is attributable
to the impact of rinsing the uncoated CCA-treated control deck.
Data from Mini Deck 8C
CM
E
^
CD
tfl
ec
Q
CM
CO
CD
O
CO
<- in
-p '
CO
z CD
Bl = A,Up AAGM4
Cy = A,Dn AZM2
Gr = C.Up CCAM2
Re = C.Dn CBXM2
4 7 10 13 16 19
Time [months]
22
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure E-3. DAs vs. Time Determined on Four Specimens of Minideck 8C, Coated with a
Water-Based Stain
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 13A
CXI
E
o
CD
w
Q
CM
I
rt!
L
CD
ra
o
CD
£_
-P
CD
CD
I
CO
I
Bl = A,Up AAGM1
Cy = A,Dn AYM3
Gr = C.Up CSM1
Re = C.Dn CEM1
7 10 13 16 19
Time [months]
22
Figure E-4. DAs vs. Time Determined on Four Specimens of One of the Positive Control
Minidecks (no coating applied)
-------
The key findings of the study are captured in the pair-wise analyses of differences
between each coating and the uncoated CCA-treated minidecks at each of the seven
sampling events over the two-year study. The pair-wise analyses demonstrate that (1)
all of the coatings tested reduced DCCA. However, the efficacy of each product
decreased over the period of the study and by 24 months only two of the products, the
paints, had DAs levels sufficiently low enough that their differences from the positive
control were found to be statistically significant (p-value1 < 0.05); and (2) the reductions
of DAs for one product, a water-based sealant, were not statistically different from the
positive control over the entire course of the study (p-value ranged from 0.0885 to
0.9635).
The results of these tests were consistent with the results of parallel tests conducted by
CPSC staff in Gaithersburg, MD, despite differences in the wood substrate tested
(CPSC staff used new CCA-treated wood) and some of the methods employed.
Several coatings were common to both the EPA and CPSC staff studies and the
ordering of performance of these coatings was similar between studies.
Additional research is needed to address key questions, including:
• What is the effect of periodic (e.g., monthly, quarterly) rinsing or cleaning as
compared to coating overtime?
• What is the effect of application of multiple coats and periodic re-coating?
• What products perform best in specific climates?
• What products perform best on boards of different ages and conditions?
• How do abrasion and wear and tear effect mitigation of DAs?
• What are the potential hazards and risks associated with preparation steps, e.g.,
sanding and scraping previously coated CCA-treated wood? Elevated levels of
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
1A p-value is a measure of the amount of statistical evidence supporting an alternative hypothesis relative to
a particular null hypothesis, with smaller p-values corresponding to greater evidence. The null hypothesis is
usually the hypothesis of "no difference" or "no effect," for example, no difference in mean DAs between
Coating 1 and the Coating 13 at time period 1. Whereas the alternative hypothesis asserts that a difference
(or effect) exists, for example, Coating 1 and Coating 13 have different mean DAs at time period 1. P-values
always lie between 0 and 1 and thus are probabilities. In fact, the p-value is the probability of finding a
difference (or an effect) as large as the one observed in the data, when the null hypothesis is true. A p-value
less than 0.05 is generally considered to be "statistically significant" although the choice of 0.05 is somewhat
arbitrary and sometimes other cutoff values - 0.1 or 0.01, are used to classify findings as statistically
significant or not.
-------
arsenic were measured in peeling paint chips (and possibly attached wood fibers)
recovered from the weathered painted test decks.
The results of this study provide a reasonable indication of the maximum ability of
typical existing deck coating products to reduce dislodgeable CCA residues and
suggest that there is an opportunity for development of products specifically targeted to
not only enhance the appearance of CCA-treated wood, but also to maximize the
ability of the treatments to reduce dislodgeable CCA residues. Ideally, products would
be formulated to maximize the period of time between treatments and minimize
preparation steps that may create the potential for human exposure and movement of
the CCA residues to the environment.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
In support of such efforts to develop coatings and other potential mitigation measures,
it may be useful to develop and utilize testing methods that help gain insight into other
characteristics of CCA residues. For example, greater insight into the nature and
behavior of the soluble and insoluble fractions of the CCA residues obtained by wipe
sampling may suggest that certain coating ingredients (e.g., water repellant) are
appropriate. Likewise an improved understanding of the relationships between CCA
residues and exposure routes may suggest other appropriate mitigative measures.
This report supersedes the previously issued "Interim Data Report." In addition to
covering the complete dataset through 24 months of testing, this final report also
utilizes more sophisticated data analysis models. The draft quality assurance project
plan (QAPP) was posted for public and stakeholder comment and the QAPP as well as
the Interim Data Report were the subjects of letter peer review. An open peer review of
the draft final report by OPPTS's Scientific Advisory Panel took place in November of
2006. Records of the letter peer reviews and the supporting documents, and minutes of
the peer review meeting (SAP Minutes No. 2007-02) can be found in EPA's Science
Inventory (www.epa.gov/si/) at Record 150970.
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
1. Project Description
1.1 Background
CCA is a wood preservative that is impregnated under pressure into wood to protect it
from decay and insect damage. CCA is registered under the Federal Insecticide,
Fungicide, and Rodenticide Act (FIFRA) by EPA's Office of Pesticide Programs (OPP).
In October 2001, EPA-OPP prepared a preliminary deterministic exposure assessment
for selective internal and external peer review comments as an interim report intended
to address child residential "playground" exposures. In addition, EPA requested
guidance from the FIFRA Scientific Advisory Panel (SAP) for risk mitigation measures
such as sealants and coating processes. The SAP Panel made "recommendations
regarding the need for additional studies in this area..." because the weight-of-evidence
from available studies indicates that certain coatings can substantially reduce
dislodgeable and leachable CCA chemicals." The Panel also recommended that "EPA
inform the public of the ability of certain coatings to substantially reduce leachable and
dislodgeable CCA chemicals..."
In March 2003, the registrants of CCA wood preservatives signed an agreement with
EPA for voluntary cancellation of CCA-treated wood for residential uses (such as
playsets and decks) effective beginning January 1, 2004. However, existing decks and
playsets made of CCA-treated wood will still be in use. Therefore, the potential remains
for dermal contact with arsenic, chromium, and copper residues on treated surfaces,
and this may be a concern for infants and small children, due to their close contact with
surfaces and hand-to-mouth activities (Kwon et al., 2004; Ursitti et al., 2004; Zartarian
etal.,2006).
In support of OPP's need for information for risk assessment development, the Office
of Research and Development (ORD) conducted a study that evaluated the ability of
selected coatings to reduce the amount of dislodgeable CCA (DCCA) chemicals on the
surfaces of CCA-treated wood. The Consumer Product Safety Commission (CPSC)
was a collaborator on this project via an interagency agreement (CPSC-l-03-1235)
between EPA and CPSC. The test data support EPA and CPSC efforts to inform the
public regarding the use and maintenance of existing CCA-treated wood products,
such as decks and playground equipment.
In previous studies, CPSC staff determined the relationship between the amount of
arsenic transferred to a hand by contact with CCA-treated wood and the amount
determined by the CPSC wipe technique (Thomas et al., 2004; Levenson et al., 2004;
Hatlelid et al., 2004). EPA modified the CPSC wipe technique in an effort to increase
sensitivity. Appendix A reports experiments conducted by EPA and CPSC staff to
determine the relationship between dislodged CCA residues determined using each
approach. Although calculation of human exposure is considered beyond the scope
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
this coatings study, the wipe comparison experiments established a relationship
between the two surrogate wipe methods. However, further analysis or
experimentation may be needed to reduce uncertainty and/or estimate potential intake
due to contact based upon the data presented in this report.
This report supersedes the previously issued "Interim Data Report." In addition to
covering the complete dataset through 24 months of testing, this final report also
utilizes a more sophisticated data analysis model. The draft quality assurance project
plan (QAPP) was posted for public and stakeholder comment and the QAPP as well as
the Interim Data Report were the subjects of letter peer review. An open peer review of
the draft final report by OPPTS's Scientific Advisory Panel took place in November of
2006. Records of the letter peer reviews and the supporting documents, and minutes of
the peer review meeting (SAP Minutes No. 2007-02) can be found in EPA's Science
Inventory (www.epa.gov/si/) at Record 150970.
1.2 Project Objectives
The objective of this project was to evaluate the ability of typical deck coating products
to reduce dislodgeable CCA chemicals (DCCA) on pressure treated wood.
Environmental variables that impact coating performance include ultraviolet (UV)
radiation, condensation, precipitation, and thermal shock. Efficacy of coatings may also
be impacted by the level and fixation of CCA treatment, age and condition of the wood
at the time of coating, and type and dimensions of the treated wood, among other
variables (Lewbow et al., 2003). Due to the large number of variables, and EPA's
desire to provide guidance quickly for in-service wood, the scope of this evaluation was
limited to selected coatings applied to aged CCA-treated southern yellow pine (SYP)
wood exposed to natural outdoor weathering at a site in North Carolina.
Due to the large number of variables that potentially impact coating performance and
the lack of a standardized protocol for evaluation of the efficacy of coatings to reduce
DCCA, secondary objectives were to gain insight into the usefulness of the CPSC wipe
sampling approach for evaluation of coating performance and identification of future
research needs.
1.3 Experimental Design
Two sources of weathered CCA-treated SYP were harvested from in-service decks
and used to construct a series of miniature decks (minidecks), onto which selected
coatings were applied. Dislodgeable arsenic (DAs), chromium (DCr), and copper (DCu)
were measured at specified intervals. Weathering tests were conducted by exposing
minidecks to natural outdoor conditions at a site in Research Triangle Park, North
Carolina. The ability of the coatings to reduce DCCA as the wood and coatings
10
-------
weather was evaluated by periodically determining the amount of DCCA removed from
the surface of the wood specimens using a specific wipe technique.
For the purposes of this study, DCCA was defined as the amount of CCA analyte
removed from the surface of the test specimen using a method based on a dermal
wipe procedure initially developed and demonstrated by the staff of the CPSC. The
CPSC staff method was slightly modified for the EPA study. The different wipe
sampling methods and all modifications implemented by EPA at each stage of the
project are fully described in Section 2.8.3 of this report. A side study was conducted to
determine the comparability of the two methods and the results of that study are
presented in Appendix A.
Wood nomenclature used in this report is defined in Figure 1-1. A "board" is defined as
the unit of wood removed from an existing structure (the "source"), while "specimen"
refers to the pieces of each board cut for this project (specimens are sometimes called
coupons in weathering testing jargon). Furthermore, areas on specimens that were
wipe-sampled during each sampling interval are termed "primary sampling areas"
(PSA), in contrast to adjacent areas which were not sampled at each interval. Each
specimen used in this project contained one PSA and one adjacent area. The PSAs
were wipe-sampled during each sampling event (i.e., at 1, 3, 7,11,15, 20 and 24
months after coating). Areas on the minidecks adjacent to the PSAs were termed
"baseline areas" (BL) because they were used to calculate specimen-specific baseline
DCCA values (refer to Section 2.8.2.1 for further explanation).
6" (nominal)
/ /
J
/~
Top Face ~« —
t
fiU»fnnmirah BOttOITl F3C6
General Direction of Grain — +
i
t
Uncut Edge
/?
=f' cut End
Figure 1-1. Wood Nomenclature
All sampling was done on the top faces of the boards (those facing up on the original
source structure from which the boards were harvested). A "grain-up" or "bark side up"
board is defined as one where the tree growth rings, evident on the cut end of the
board, form a convex pattern (a "hill") when observed with the face of the board that
was exposed on the source deck facing up. Likewise, a "grain-down" or "bark side
down" board is defined as one where these rings form a concave (a "valley") pattern
when the exposed face is facing up. Because wood tends to deform along these ring
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
11
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
lines, grain orientation may be an important variable in the measurement and mitigation
of DCCA on surfaces of CCA-treated wood. Grain-down boards tend to deform in a
manner which "cups" and holds water and moisture while compressing the wood fibers
on the sampling surface, while grain-up boards tend to deform in a manner which
sheds water from the surface of the board but also creates tension in the wood fibers
which may exacerbate splitting and cracking on the sampling surface.
Each minideck had nine decking specimens: two specimens from each of the aged
wood sources (one specimen with bark side up grain orientation and one with bark side
down orientation), separated by specimens of new untreated wood (all positioned bark
side up) to prevent cross-contamination and to serve as blank controls to assess cross-
contamination potential as a result of splash-over of rain water, for example. The
minidecks were constructed with each of the aged wood specimens having the same
top face as the specimen had during its exposure on its source structure. Prior to
coating the minidecks, baseline DCCA concentrations were determined by averaging
the results of wipe samples from the two areas adjacent to the PSAs. A total of twelve
different coatings were applied to three minidecks each. Additionally, three identical,
but uncoated minidecks and one untreated, uncoated minideck were included as
controls. The minidecks were prepared for coating according to the specific coating
manufacturer's recommendations. Coating was then applied to each minideck per
coating manufacturer's instructions. After allowing the coatings to dry and cure, the
minidecks were subjected to natural weathering for a period of two years. During the
two-year weathering period, at specified intervals (1 month, 3 months, 7 months, 11
months, 15 months, 20 months and 24 months after coating application) each
specimen was wipe-sampled for DCCA and the results from each sampling event were
statistically compared with corresponding DCCA results from uncoated control
minidecks. Supporting samples included wood core samples, liquid samples of the
coatings applied and paint chip samples.
The experimental design is a split plot with repeated measurements. In the language of
split-plot designs, each minideck is a whole plot, and the whole-plot treatment is
coating. The split-plot treatment is defined by the two-by-two factorial arrangement of
the combinations of source decks (A, C) and grain orientation (up, down), each board
corresponding to a split-plot. Repeated measurements were taken on each board over
the seven time periods.
The coatings were randomized to minidecks with the only restriction being that
neighboring minidecks were not assigned the same coatings. In particular, the
replication was not blocked, and this is a difference between the experiment and a
typical split-plot design. Models for the analysis of the data include fixed effects for
coating, source deck, grain, and time; and random effects for minideck, boards, and
experimental error.
12
-------
Potential statistical models for the data all have the same fixed effect structure (main
effects and all interactions), but differ according to the assumed covariance structure of
the random effects, the scale of the response (e.g., log-transformed or not), and the
method of baseline adjustment (direct, covariate, none). In Section 3.2.5, the results
from a number of model variations are reported in order to establish the robustness of
the major findings to modeling assumptions.
1.4 Data Quality Objectives
The critical measurements for the natural weathering tests are total arsenic, and to a
lesser extent, total chromium, and total copper concentrations, which are subsequently
converted to dislodgeable arsenic, chromium, and copper, reported as mass per unit
area (ug/cm2), based on the area of the wiped surface. Data quality indicator (DQI)
goals for concentration in terms of accuracy, precision, and completeness, as
established in the QAPP for this project, are shown in Table 1-1. The method detection
limit (MDL) for the analytical method is also shown. The 0.1 ug/L MDL corresponds to
0.000032 ug/cm2 DAs. An assessment of accuracy, precision and completeness goals
is presented in Section 5.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 1-1. Data Quality Indicator Goals for Critical Measurements
Analyte
Arsenic (total)
Chromium
(total)
Copper (total)
Method
SW-846 Method 6020
(modified)
SW-846 Method 6020
(modified)
SW-846 Method 6020
(modified)
Accuracy
(Percent
Recovery)
90-110
90-110
90-110
Precision
(Percent
RSD/RPD)
10
10
10
Completeness
(Percent)
90
90
90
Method
Detection Limit
(ug/L)
0.1
0.1
0.1
13
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
14
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
2. Sampling and Analysis Protocol
Many elements of the experimental protocol were considered in detail because, at the
time of the tests, no standardized protocol existed to evaluate the ability of coatings to
reduce DCCA from CCA-treated wood. The discussion below is primarily limited to
what was actually done. Methodological alternatives that were considered and side
experiments to assess various aspects of the experimental procedures are described
in more detail in Section 4.
2.1 Selection of Wood Sources
Criteria established to select from candidate aged wood source structures included the
age and condition of the deck and the availability of sufficient amounts of grain up and
grain down boards treated to 0.40 pound per cubic foot (pcf) with Ground Contact
CCA-C to meet project needs. Additionally, sources needed to be of the SYP species,
as it is the most widely-used CCA treated wood species in the U.S. According to the
Southern Pine Council (SPC), SYP is preferred "because of its ease of treatability. The
unique cellular structure of Southern pine permits deep, uniform penetration of
preservatives, rendering the wood useless as a food source for fungi, termites and
micro-organisms. Some 85 percent of all pressure-treated wood is Southern pine"
(SPC, 2006). In 2004, the SPC estimated that 76 percent of all CCA treated lumber
and timbers were SYP with the next most common species group being the Hemlock-
Fir group at 9 percent (Lebow, 2006). Regional differences are likely, with SYP
overwhelmingly predominant in the Southeast U.S., a greater percentage of hemlock-
fir, Douglas-fir and spruce-pine-fir on the west coast, and red pine and ponderosa pine
also being used in the north and mid-west (Lebow, 2006).
Age, general condition and coating/treatment history were additional source deck
selection criteria as it was recognized that these variables might impact coating
performance. Given the almost infinite combinations of age, treatment history, and
condition, and the need for a relatively large amount of material, two locally available
in-service decks were selected for testing from approximately a dozen candidate decks
that were offered for use in the project. Details of the selected decks are provided
below.
Environmental Research Center (ERG) Deck: This structure was located outside at
EPA's original (leased) Research Triangle Park laboratory facility. It was a stand-alone
deck with generally full weather exposure and only moderate shading by adjacent
buildings during low sun positions. Given its open, stand-alone nature, abrasion
patterns appeared very consistent and the boards were visually similar to one another.
The deck was constructed of SYP, treated to 0.40 pound per cubic foot (pcf) with
Ground Contact CCA-C. This source was approximately 7 years old and was reported
to have received one application of a deck sealant near the beginning of its use (over 5
15
-------
years ago). The overall condition of the wood was considered fair: the coloration was
gray and there was slight-to-moderate splintering. Specific locations and orientations of
individual boards were documented during dismantling of the source structure; a map
of the structure showing the location of each specimen tested was prepared. This map
is shown in Figure 2-1. Photos are provided in Figure 2-2. This deck is referenced as
the "A" source.
New Hill Deck: This source was an outdoor deck on a private residence. It represents a
source of relatively new, good-condition, aged CCA-treated wood. The coloration of the
wood was light brown and relatively bright and there was minimal splintering. The New
Hill Deck was an exposed, attached structure. There was no noticeable biological
growth or other dampness-related defects. The deck was constructed of SYP, treated
to 0.40 pcf with Ground Contact CCA-C, had been in service for just over one year,
and had never been chemically cleaned or treated. Specific locations and orientations
of individual boards were documented during dismantling of the source structure; a
map of the structure showing the location of each specimen tested is shown in Figure
2-3. Photos are provided in Figure 2-4. This deck is referenced as the "C" source.
2.2 Preparation and Characterization of Wood Sources
2.2.1 Source Wood Harvesting and Preliminary Characterization
Wood specimens were prepared from boards with nominal 5/4" x 6" cross-sectional
dimensions, taken from the source structures described in Section 2.1. Care was taken
to minimize handling and abrasion of the primary (i.e., 6" width) faces of the treated
source boards, with the short edges of the board preferentially held during transport
and cutting. New 5/4" x 6" SYP that was not CCA treated was used for the blank
control specimens and the cross-contamination control specimens that were located at
the ends of each minideckand between each of the four CCA-treated boards on each
minideck.
For each aged CCA-treated board, the total board length was recorded along with
visually-observable source wood characteristics; the most important being predominant
grain orientation (up versus down). Each board was also characterized as to
predominant grain type (percent flat versus percent edge grain), predominant ring
spacing (tight, medium, wide), predominant wood season (percent early versus percent
late wood), and predominant wood type (percent heartwood versus percent sapwood).
The percentages of the various grain characteristics defined above were visually
observed, estimated and documented.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
16
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 2-1. ERG Deck Map (Source A)
Figure 2-2. Views of ERG Deck
Note that moisture stains were temporary and that boards under benches were not used to construct minidecks.
17
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 2-3. New Hill Deck Map (Source C)
Figure 2-4. Views of New Hill Deck
18
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Aged boards were cut into specimens of approximately 86 cm (34") lengths using a
circular table saw. These lengths were cut in such a manner as to capture three sets of
existing nail holes on each aged wood specimen, and required that the nail holes were
spaced on 16-inch centers as typical. Of utmost concern was that the PSAs were
segments of the specimen with a 38-cm (15-in) or more clear distance between
adjacent nail holes, to allow for the full wipe length without crossing nail holes. Nail
holes were not wiped during either the baseline or routine wipe sampling events. To
prevent cross-contamination of samples, the saw was decontaminated between cutting
the different types of wood and the untreated wood was cut separately (after
installation of a new blade). Decontamination of the saw was done by wiping the blade
with a cloth moistened using deionized water.
Where possible, the ends of each board were removed and archived and segments
between each 86-cm test specimen were removed and archived. Some of these
interior segments were used to characterize the source wood via moisture content
measurement and wood core sampling for total arsenic, chromium, and copper
analyses. The 86-cm wood specimens were visually inspected to ensure that they did
not have excessive amounts of deformities, presence of heartwood, knots, resin
pockets, or other defects (no specimens were disqualified). Each segment was
identified with a unique alphanumeric code as follows:
• Aged board codes were prefixed by the letter "A" for source A, the ERG Deck, and
"C" for source C, the New Hill Deck.
• Each aged board from the two sources was identified with a unique letter (A, B, C,
and so forth).
• Each space between adjacent nail holes was identified with an alphanumeric code,
where the prefix "BL" refers to segments used for establishing baseline
characteristics, while the prefix "M" refers to segments that were to be regularly
wiped; that is, the PSAs. These codes were suffixed with sequential numbering
(1, 2, 3, etc.) along the length of each source board.
• Unused, unwiped segments were designated with the prefix "X."
The specimen identification criteria presented above is illustrated by the example
schematic in Figure 2-5. In this example, BL1, BL2, BL3, BL4, and BL5 would be wipe-
sampled before cutting the board. These results would be used to establish baseline
DCCA values for the PSAs identified as M1, M2, and M3. After cutting the boards to
harvest 86-cm specimens (illustrated in the figure by the dashed boxes) for minideck
construction, BL2 and BL5 would be subsequently used for taking one core sample
each for total arsenic, chromium, and copper analyses, as well as moisture content.
M1, M2, and M3 would be wipe-sampled during routine sampling events to determine
19
-------
coating efficacy. BL1, BL3, and BL4 would be wipe-sampled only periodically in an
effort to determine the effects of abrasion and rewiping on coating efficacy and DCCA.
For example, if Figure 2-5 was Board A from Source C and cut as shown by the
dashed boxes, this board would generate wipe sampling areas C-A-BL1, C-A-BL3, C-
A-BL4, C-A-M1, C-A-M2 and C-A-M3.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
// X1 .-•'/ BL1 / M1 // BL2 ,••'/ M2 / BL3 // XI S/ BL4 / M3 // BL5 / X3
' •*
6-cm specimen
86-cm specimen
88-crri specirnen
Board "A"
Figure 2-5. Specimen Identification and Baseline Sampling Scheme Example
All cut specimens were identified on one cut end or uncut edge with its identification
code, as well as with its "top" side using permanent marker. All numbered specimens
were qualitatively and semi-quantitatively characterized for visually-observable wood
condition characteristics, with data recorded on a standardized wood characterization
data sheet. The characteristics recorded included number of knots for that specimen ,
splintering, cracking, and rotting (for these last three, a rating of 1 to 5, with 5 being like
new wood and 1 being complete failure, was assigned). All of this source
characterization data is included in full in Appendix B (a summary table can be found in
Appendix C, Table C-1). Additionally a photo record was made of all specimens which
includes photographs taken at the beginning of the test (i.e., precoating), as well as at
each sampling event after coating (Appendix D). Remaining segments of wood were
retained and archived.
2.2.2 Baseline Sampling
Each source board used in the construction of the minidecks had at least two time = 0,
baseline wipe samples taken from it prior to coating. The wipe method used to collect
these samples is described in Section 2.8.3.1. Samples were then digested and
analyzed for CCA content. The baseline wipe sample results were used to establish
baseline DCCA concentrations for the PSAs on the minidecks. Furthermore, the
baseline data were used to assess the variability of DCCA across each board
(intraboard) and between boards (interboard) and source decks. The results of this
sampling are presented in Section 3.1.1. The full dataset showing tables with
summarized data and the individual specimen baseline values is provided in
Appendix E. Summary tables are included in Appendix C, Tables C-2 and C-3. Data
20
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
summaries include averages, standard deviations, and RSD for each board and
include summary statistics at the end of each table for each source (A and C). The
RSDs are indicators of variability within and between boards.
The baseline measurements were made using acid-washed sampling wipes in a
method termed the "A2" sampling method, as described in Section 2.8.3.1. The "2X"
sampling method utilized unwashed wipes and was the method used for subsequent
sampling events; it is described in Section 2.8.3.2. A side study was conducted to
develop calibration equations to convert between results obtained using wipes
prepared by the baseline method (A2) and the main-study method (2X), taking the form
cX, as indicated below. However, these factors had no impact on the analysis of
variance because In(cX) = ln(c) + ln(X) and thus only the intercept in the ANOVA model
is affected by the use of the calibrated measurements; not comparisons among
coatings. The full wipe comparison report is included as Appendix D.
Additional variables that were considered in the aforementioned wipe comparison
study included: grain orientation (up, down), source deck (A, C), rinse (rinsed,
unrinsed), and sample preparation lab (EPA, CPSC). "Unrinsed" boards in this context
refers to boards that were taken directly from storage and wipe-sampled, while "rinsed"
boards were thoroughly hosed down with tap water and allowed to dry for several days
before wipe sampling. "Prep lab" refers to which laboratory digested or extracted the
wipes and subsequently either analyzed the samples in-house (CPSC) or sent them
out to an analytical laboratory for analysis (EPA).
Statistical model selection was used to identify calibration equations for predicting
method 2X DCCA measurements from method A2 DCCA measurements and the other
factors, including grain orientation (up, down), source deck (A, C), rinse (rinsed,
unrinsed), and prep lab (EPA, CPSC). Based on these analyses, separate calibration
equations are suggested for rinsed and unrinsed boards, but not for any of the other
factors. In other words, when models for predicting DCCA using 2X wipes from DCCA
using A2 wipes, grain, source deck, rinse, and prep lab were considered, the identified
prediction model depends only on DCCA using A2 wipes and rinse.
The wipe method correction factors are summarized as follows:
• For arsenic:
Rinsed Specimens: As-2X = 1.44 (As-A2), 95 percent Confidence Interval: (1.21, 1.66)
Unrinsed Specimens: As-2X = 0.81 (As-A2), 95 percent Confidence Interval: (0.72, 0.91)
The R-square value for the combined models is 0.75
• For chromium:
21
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Rinsed Specimens: Cr-2X = 1.32 (Cr-A2), 95 percent Confidence Interval: (1.08, 1.56)
Unrinsed Specimens: Cr-2X = 0.82 (Cr-A2), 95 percent Confidence Interval: (0.73, 0.92)
The R-square value for the combined models is 0.58
• For copper:
Rinsed Specimens: Cu-2X= 1.20(Cu-A2), 95 percent Confidence Interval: (0.99, 1.42)
Unrinsed Specimens: Cu-2X = 0.83 (Cu-A2), 95 percent Confidence Interval: (0.74, 0.91)
The R-square value for the combined models is 0.76
Because the baseline analyses for this study were done on unrinsed boards, the
unrinsed specimen equations were used to adjust the baseline results. In this report,
raw (uncorrected) baseline data are reported in the appendices, but only "corrected"
baseline DCCA data are presented in the body of the report. Likewise, reported values
which are calculated using baseline DCCA in the calculation (e.g., the percent
reduction values) always use the corrected baseline values. DCCA from all subsequent
sampling events are reported uncorrected, as they were conducted using the 2X wipe
method.
2.2.3 Wood Core Sampling and Analysis
Up to four wood core samples were taken from each CCA treated board used to
construct minidecks for this study. Professional judgment was used to select areas
from which to take core samples to ensure that their average would be representative
of the board in question. Individual core samples were taken by advancing a %-inch
diameter drill bit through the entire 1-inch (5/4" nominal) thickness of the board and
collecting the wood shavings.
The wood shavings were then dried to constant weight in a drying oven at
approximately 105°C. The dry weight of the sample was recorded. The wood shavings
were subsequently digested in concentrated nitric acid using a similar protocol to that
defined in Section 2.8.4 for the wipe samples. This procedure is consistent with
American Wood Preservers Association (AWPA) Standard A7-93 (microwave assisted
nitric acid digestion). Digestates were analyzed by inductively coupled plasma - mass
spectrometry (ICP-MS), consistent with AWPA Standard A21-00.
AWPA standards allow the actives composition of CCA-C - the formulation used to
treat the wood used in this study - to vary between 44.5 - 50.5 percent for CrO3, 17.0
- 21.0 percent for CuO, and 30.0 - 38.0 percent for As2O5 in a specific assay zone, the
outer 0.6 in (15 mm) (Lebow, 1996). Knowing that the source wood for this project was
treated to target retentions of 0.40 pounds per cubic foot (pcf), hypothetical, ideal
actives composition (analyte concentrations) can be calculated for each CCA analyte:
0.190 pcf (86.1 g/cf) CrO3, 0.074 pcf (33.6 g/cf) CuO, and 0.136 pcf (61.7 g/cf) As2O5.
22
-------
Furthermore, the average dry, pretreatment density of SYP is 32 pcf, or 14.5 kg/cf.
Using these values, predicted levels of CCA analytes in the study wood core samples
can be approximated as:
CrO3 (86.1 g/cf) / (14.5 kg/cf) x (1000 mg/g) = 5,938 mg/kg
CuO (33.6 g/cf) / (14.5 kg/cf) x (1000 mg/g) = 2,317 mg/kg
As2O5 (61.7 g/cf) / (14.5 kg/cf) x (1000 mg/g) = 4,255 mg/kg
For comparison, wood core sample data is presented in Section 3.1.2.
2.3 Minideck Construction
After cutting and marking specimens with their identification codes, source wood
specimens were transported to the minideck host site. The minidecks were constructed
on site in accordance with Figure 2-6. The minideck surfaces were initially constructed
without leaving spaces between the boards. When this mistake was discovered, the
three internal untreated boards per deck were removed and planed sufficiently on
either edge enough to be spaced approximately 1/8th inch (the thickness of a 16p nail
spacer) from adjacent boards. The untreated boards were then reattached to the tops
of the minidecks and the deck tops were rinsed with tap water and allowed to dry for at
least 48 hours before coating. Following construction, the minideck surfaces were
prepared (e.g., washed, rinsed, etc.) in strict accordance with the particular coating
manufacturer's recommendations for coating aged wood as described in Section 2.5.
Three minidecks were constructed for each of the 12 selected coatings and an
uncoated positive control. Minidecks were identified as 1-A, 1-B, 1-C (for coating #1),
2-A, 2-B, 2-C (for coating #2), and so on. Each minideck contained two 86-cm aged
source "A" specimens, two 86-cm aged source "C" specimens, and five 86-cm
untreated wood specimens. In addition to these 13 minideck series, an additional
uncoated minideck was constructed using 5 boards all of untreated SYP. The three
inside boards of this deck were wipe-sampled at the prespecified regular sampling
event intervals as negative control blanks.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
23
-------
34-
I r
•I 1
I 1
I L_
T 1
i 1
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
UNTREATED 5/4x6
AGED SOURCE A-up
UNTREATED
AGED SOURCE A-down
UNTREATED
AGED SOURCE C-up
UNTREATED
UNTREATED
36
Figure 2-6. Schematic of Minideck Construction
Note that untreated 34" specimens were planed so that 1/8" of space was provided between each pair of specimens
24
-------
New 4" x 4" CCA-C treated wood posts were used in the construction of the minidecks.
Specimens were screwed directly into a grid of 2" x4" untreated SYP supports. To
avoid contamination, these supports were slightly offset above the tops of the posts to
ensure that the treated posts did not have the opportunity to directly contact the wood
test specimens. Plastic-coated screws were advanced through existing nail holes on
the treated specimens to secure decking specimens to the minideck frames. The same
type of coated screws was used to attach the new, untreated specimens to the support
frames. The minidecks were free-standing (i.e., posts are not set into the ground) and
were leveled after setting them on the test site. A photograph of a typical minideck is
provided as Figure 2-7. A list of the treated wood specimen IDs that were used on each
minideck is shown in Table 2-1. Specimens were matched with minidecks randomly,
with the only qualifier being that the three replicate minidecks per coating utilize boards
cut from different source boards. In other words, for each set of three minidecks, a total
of 12 (4 boards x 3 minidecks) different source boards are represented.
Figure 2-7. Photograph of Typical Minideck
Note that white plastic containment system underneath minideck was for a separate bioavailability study,
otherwise not related to this project
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
25
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-1. Wood Specimens Used to Construct Minidecks
Coating #
1
2
3
4
5
6
7
8
9
10
11
12
13
(uncoated)
Deck ID
1-A
1-B
1-C
2-A
2-B
2-C
3-A
3-B
3-C
4-A
4-B
4-C
5-A
5-B
5-C
6-A
6-B
6-C
7-A
7-B
7-C
8-A
8-B
8-C
9-A
9-B
9-C
10-A
10-B
10-C
11-A
11-B
11-C
12-A
12-B
12-C
13-A
13-B
13-C
Source - Grain Orientation
A- up
A-AE-M1
A-V-M3
A-AJ-M1
A-O-M3
A-BC-M2
A-AR-M1
A-T-M1
A-AG-M3
A-AD-M2
A-T-M2
A-BC-M1
A-I-M3
A-U-M2
A-AD-M1
A-AR-M3
A-U-M1
A-AC-M2
A-BC-M3
A-O-M2
A-V-M1
A-AJ-M3
A-AR-M2
A-I-M1
A-AG-M4
A-T-M3
A-AC-M1
A-AG-M2
A-AD-M3
A-X-M1
A-AJ-M2
A-U-M3
A-X-M2
A-AJ-M4
A-O-M1
A-AC-M3
A-V-M2
A-AG-M1
A-I-M2
A-X-M3
A - down
A-Z-M1
A-AT-M3
A-BW-M4
A-BY-M2
A-AH-M4
A-P-M1
A-L-M3
A-AF-M1
A-BW-M2
A-BG-M4
A-AH-M1
A-Q-M2
A-L-M2
A-Z-M3
A-BG-M3
A-BY-M1
A-AN-M3
A-P-M2
A-Y-M2
A-AH-M3
A-BW-M1
A-BY-M3
A-AT-M1
A-Z-M2
A-P-M3
A-AE-M2
A-AN-M1
A-BG-M2
A-Y-M1
A-Q-M3
A-Q-M1
A-AH-M2
A-BW-M3
A-AN-M2
A-AE-M3
A-L-M1
A-Y-M3
A-AT-M2
A-BG-M1
C-up
C-N-M1
C-BE-M2
C-S-M2
C-BZ-M3
C-BI-M1
C-BY-M2
C-N-M3
C-BJ-M2
C-CD-M1
C-CD-M2
C-BM-M2
C-AC-M1
C-AC-M2
C-BM-M3
C-CA-M1
C-BZ-M2
C-AJ-M1
C-S-M3
C-N-M2
C-BY-M1
C-BZ-M4
C-BE-M1
C-AC-M3
C-CA-M2
C-AP-M1
C-BI-M2
C-BZ-M1
C-AP-M3
C-BJ-M1
C-BU-M2
C-AP-M2
C-BE-M3
C-BJ-M3
C-AJ-M2
C-BI-M3
C-BM-M1
C-S-M1
C-AJ-M3
C-BU-M1
C-down
C-BO-M2
C-CC-M1
C-AA-M2
C-E-M3
C-AN-M1
C-BX-M3
C-CE-M2
C-AN-M3
C-AA-M1
C-AD-M2
C-AM-M2
C-BT-M4
C-CE-M1
C-BO-M1
C-AD-M3
C-AA-M3
C-AI-M1
C-CC-M2
C-AM-M3
C-BX-M1
C-E-M2
C-AE-M3
C-AM-M1
C-BX-M2
C-BW-M1
C-AN-M2
C-AE-M2
C-AD-M1
C-AK-M4
C-BT-M2
C-AI-M3
C-BW-M2
C-AE-M1
C-AM-M4
C-AD-M4
C-BT-M1
C-E-M1
C-AI-M2
C-BT-M3
26
-------
2.4 Selection of Coatings
Basic formulation and application information for a large number of products was
obtained primarily by using Internet searches and visits to local retail hardware and
home improvement stores and compiled into a master list of 125 products. Products
ultimately selected for testing are listed generically in Table 2-2. The list includes two
paints, two products specifically marketed to prevent DCCA exposure, and eight
products marketed as stains and sealants (four oil-based products and four water-
based products, one of which was marketed as alkyd-based, one acrylic, one both
alkyd and acrylic, and one specifying neither).
For preserving the appearance of CCA-treated decks, experts generally recommend
the use of penetrating finishes such as oil and water-based sealants rather than film-
forming products such as paints due to the tendency of film-forming products to crack
and peel as the substrate shrinks and swells with uptake and release of moisture
(Williams and Feist, 1993). Forthis reason, eight of the twelve products selected were
oil or water-based sealants and stains that fall into the category of penetrating finishes.
The oil and water-based paints were selected because porch and deck paints are
widely available to consumers and may be applied to CCA decking. The two products
that were marketed to encapsulate CCA treated wood were selected because they
were marketed as products that would reduce exposure to CCA residues and could be
readily applied by consumers.
Products were classified by base (oil vs. water), cover (clear, semi-transparent,
opaque), and product type. Product type was broken out into the following: paints,
primers, sealants, stains, and other. Additional classification descriptors included
ingredients (primarily alkyd or acrylic), surface (penetrating vs. film-forming) and color.
Note that the classifications are derived from information provided on the product label
and from information provided on product MSDS sheets. In that regard, the
classifications are arbitrary and are not necessarily exclusive of one another. Forthis
reason, and due to the limited number of products tested that were attributed to each
category, analyses of the coatings by category are presented and discussed in Section
4, "Study Limitations, Exploratory Analyses, and Future Directions".
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
27
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-2. Selected Products for Evaluation
#
1
2
3
4
5
6
7
8
9
10
11
12
13
Product Type
Sealant
Sealant
Stain
Stain
Sealant
Sealant
Stain
Stain
Paint
Paint
Other
Other
No coating
Base
Oil
Oil
Oil
Oil
Water
Water
Water
Water
Water
Oil
Water
Water
N/A
Cover
Semi
Clear
Clear
Clear
Clear
Clear
Semi
Clear
Opaque
Opaque
Clear
Clear
N/A
Main Ingredients
Aliphatics, Napthalene
Acrylic, alkyd, urethane
Acrylic
Alkyd
Unknown1
Acrylic, alkyd
Alkyd
Acrylic
Acrylic
Alkyd, polyurethane
Elastic vinyl
Polymer
N/A
Comments
"Cedar" with UV blocker
"Clear"
"Deep tone base"
"Clear stain"
"Clear"
"Clear"
"Cedar" with UV blocker
"Tint base, solid" with no tint added2
"Gray". Latex, designed for porches and floors
"Gray". Designed for porches and floors
Designed for CCA encapsulation
Designed for CCA encapsulation
Uncoated control minidecks
1 MSDS for coating listed no ingredients >1 percent.
2 Note that the labeling for product #8 specifically states that it must be tinted before use.
2.5 Coating Application
After baseline characterization of source wood, construction of the minidecks, and
preparation of the minidecks for coating, all exposed surfaces of the decks were coated
in accordance with coating manufacturers' recommendations. Coatings were applied to
fully cover the top faces, exposed uncut edges, and cut ends of CCA-treated wood
specimens. For the paints, products #9 and 10, a common latex primer was first
applied in accordance with the paint and primer manufacturers' instructions. For all
coatings, top faces were coated first, followed by the exposed edges and the cut ends.
Because the coatings' manufacturers generally recommended that application not be
done during periods of direct sunlight, an open air shelter (canopy) was temporarily set
up on site to allow for coating minidecks in the shade. After 24-hours of initial coating
drying in the shade, minidecks were manually relocated to allow for additional drying in
exposed conditions.
Flow sheets generically detailing the wood preparation procedures employed for each
coating are provided as Appendix F. The surfaces of the minidecks for each coating
(except coating #7, which did not call for a rinse prior to coating) were rinsed with a
pressure washer at a 1,000-3,000 psi setting. Ten of the coatings also had a deck
cleaning product applied, as specified in their instructions. Note that specific products
28
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
used to prepare the minidecks for coating are not mentioned in order to maintain
coating confidentiality. The surfaces of the positive (#13, uncoated) and negative
control minidecks were rinsed with a pressure washer at a 1,000-3,000 psi setting.
Two types of 2" chip brushes were used to apply coatings to the minidecks: one brush
with natural bristles and the other with polyolefin bristles (to apply coatings whose
manufacturer recommended a synthetic bristled brush). Prior to coating application,
both brush types were analyzed in order to ensure that they did not contribute
significant amounts of arsenic, chromium, or copper to the wood surfaces. Each type of
brush used was prequalified for use per a set of control samples whereby two brushes
of each type were agitated in a 250 ml vessel containing 40 ml of deionized water for
30 seconds. The liquid samples were than transferred to Digitubes for digestion. Four
milliliters of nitric acid were added to each tube and the samples were then digested at
95°C for two hours. Digested samples were sent to the analytical laboratory used for
this project, for analysis. Arsenic levels were found to be below the reporting limit of 0.1
ug/L, chromium levels were <1 ug/L and copper levels were lower than levels seen on
blank, untreated boards. A summary of the data is tabulated in Appendix C, Table C-4.
Three new brushes were used for each minideck: a brush for coating each of the two
aged wood sources and a brush for the new untreated wood surfaces. Untreated
surfaces were coated first, followed by the aged CCA surfaces. Brushes were prepared
for initial coating application in accordance with brush manufacturer's
recommendations. After a particular coating was applied to a given group of triplicate
minidecks, used brushes were sealed individually in plastic bags and archived.
Separate aliquots of coating were used for each minideck in order to prevent cross-
contamination of coating by re-dipping the brush applicator. Three aliquots of coating
were used for each minideck: one for the "A" specimens, one for the "C" specimens,
and one for the untreated (identified for sampling as "N") specimens. Separate aliquots
of coating were poured into disposable plastic graduated volumetric beakers. The
disposable beakers were acid-washed using a procedure similar to that specified in
Section 2.8.4 prior to use. Poured but unused coating for each replicate minideck was
composited so that one sample was retained for each coating and wood type
(new/untreated and the two, aged, CCA-treated sources). These leftover coating
samples were stored in sealed, unused paint containers and archived for possible
future analyses. Application procedures and any notable observations were
documented for each coating using a coating application data form (a sample form is
provided as Appendix G).
The amount of coating, measured as both volume and mass, applied to each substrate
on each minideck was determined by transferring 200 to 300-mL of coating directly
29
-------
from the original coating container into a 400-mL graduated beaker, from which the
substrates were coated. The starting volume of coating in the beaker and the final
volume (after squeezing out excess coating from the brush used) were recorded and,
from these, a calculation of the volume of coating applied was made. Additionally, the
container of unused coating and the brush to be used for application was preweighed.
After the coating was applied, the final weight of the beaker and brush was measured
and recorded. The weight applied was calculated as the difference between the initial
and the final weights.
Unused aliquots of each coating tested were sampled directly from the coating
container in duplicate and prepared and analyzed in accordance with methods
specified in Section 2.9.
2.6 Outdoor Weather Site Setup
Minidecks were subjected to outdoor weathering at a field on EPA's RTP research
campus. Prior to placement of the minidecks, the site was cleared, tilled, leveled and,
except for the area directly underneath each minideck, covered with landscape fabric
and gravel to retard growth of vegetation. Minidecks were randomly assigned to
gridded blocks, with the qualifier that minidecks featuring the same coating were not
allowed in the same row, column, or diagonally immediately adjacent to one another.
The site layout showing sequentially numbered minideck locations is provided as
Appendix H. The location of each minideck is summarized in Table 2-3 - the blocks
listed in Table 2-3 correspond to those shown on the site plan in Appendix H. Each
minideck was leveled after placement at the test site.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
30
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-3. Minideck Block Assignments (blocks correspond to those identified in Appendix H)
Minideck
1-A
1-B
1-C
2-A
2-B
2-C
3-A
3-B
3-C
4-A
4-B
4-C
5-A
5-B
5-C
6-A
6-B
6-C
7-A
7-B
7-C
8-A
8-B
8-C
Block
6
29
22
39
7
33
36
10
14
37
23
5
25
43
9
13
28
45
2
41
34
18
32
48
Minideck
9-A
9-B
9-C
10-A
10-B
10-C
11-A
11-B
11-C
12-A
12-B
12-C
13-A
13-B
13-C
BC
NC
LH
SC1
SC2
SC3
LC1
LC2
LC3
Block
46
3
26
42
27
24
30
4
20
1
35
21
19
40
12
8
47
15
31
38
16
44
11
17
BC = the blank, negative control, minideck
NC = 1 minideck with no CCA wood used (for a related
bioavailability study being conducted)
LH = 1 uncoated CCA minideck for leachate collection (for
a related bioavailability study being conducted)
SC1, SC2, SC3 = 3 soil controls (for a related bioavailability
study, no minidecks are located in these blocks)
LC1, LC2, LC3 = 3 leachate controls (for a related
bioavailability study, no minidecks are located in these
blocks)
31
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
2.7 Weather Data Collection
Weather data were collected using a Davis Instruments Vantage Pro Plus™ weather
monitoring station located as shown in Figure 2-8. The weather data was compiled with
WeatherLink® for Vantage Pro™ software, and downloaded to a Microsoft Excel file,
although there were several periods during the second year of operation when the data
logger malfunctioned and on-site data was missing. Data that was routinely collected
via the Vantage Pro Plus™ are listed in Table 2-4.
Figure 2-8. Photograph of Minideck Site
Note weather monitoring station on the right.
32
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-4. Vantage Pro Plus™ Weather Station Data
Barometric Pressure
Inside Humidity
Outside Humidity
Dew Point
Rainfall
Rate of Rainfall
Solar Radiation
UV Index & Dose
Inside Temperature
Outside Temperature
Apparent Temperature
Wind Speed
Predominant Wind Direction
Wind Chill
Units
in Hg
percent
percent
°F
In
in/hr
W/m2
index
°F
°F
°F
Mph
Mm Hg
°C
Mm
Mm/hr
Meds
°C
°C
°C
m/s
hPa (Tor)
km/h
Mb
N, NNE, NE, ENE, E, ESE, SE, SSE, S,
SSW, SW, WSW, W, WNW, NW, NNW
°F
°C
The data can be archived at 1 min, 5 min, 10 min, 15 min, 30 min, 1 h, or 2 h (data were archived
at 30-minute intervals).
All data points are discrete except for Rate of Rainfall and UV Dose.
The National Oceanic and Atmospheric Administration (NOAA) in RTP, North Carolina,
collected data on wind speed and direction, temperature, precipitation amount, direct
solar radiation, and total solar radiation at the site used for minideck weathering
through the first year of this testing. Weather data measurements collected by NOAA at
the site are summarized in Table 2-5.
33
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-5. NOAA-Generated Weather Data
Parameter
Required
Irradiance (UV)
Temperature
Precipitation, Duration
Precipitation, Amount
Dew Point (Measure of dew formation)
Wind direction + speed
Unit
W/m2
°F
hours
inches
°F
Remarks
Direct and total radiation is available.
Can be determined from strip chart, although
certain losses may occur due to evaporation.
Automated rain gage.
Dew point could be used to calculate dew
point depression (diff. with temp.) If DPD is
small, there is likely to be dew overnight.
NOAA's test site metrology instrumentation was calibrated against working standards
traceable to standards at Eppley Laboratory (for solar radiation). This calibration was
conducted periodically based on the stability of the instrument and the judgment of the
instrument operators. The temperature system was checked against certified data from
NOAA's Raleigh-Durham International Airport (RDU) weatherstation on stable days
and relative humidity measurements were calibrated using a sling psychrometer. The
weighing rain gage was calibrated with weights and also against a manual rain gage
with each precipitation event. The Aerovane wind system recorded wind speed in miles
per hour (mph) and only begins to register at 3 mph. It was also checked against RDU
on stable windy days.
These and other parameters are additionally collected by the NOAA's National Climate
Data Center (NCDC), at their RDU weather station, and are available in monthly
summaries, detailing specified conditions on a daily and, for some parameters, an
hourly basis. The RDU station does not monitor solar radiation. Supplemental solar
radiation measurements were obtained from the State Climate Office of North Carolina
for their Reedy Creek Field Laboratory monitoring site, also in close proximity to the
EPA test site.
NOAA-generated data from the test site were compared to data from the Vantage Pro
Plus™ weather monitoring station dedicated for use during this project. Spot checks of
all parameters common to both the NOAA and Vantage Pro Plus™ unit were
conducted and determined to be well within reasonable tolerances (+ 5-10 percent
depending on parameter). Because the on-site NOAA weather data became
unavailable several months into the study, certified data from NOAA's RDU station was
also used to confirm the on-site weather monitoring station data. Details of these spot
check results are provided in Appendix I. Finally, the solar radiation data from the
34
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Reedy Creek site was compared with data from the weather station at the test site and
showed agreement within 1 percent on a monthly average basis. All sets of
comparisons were in agreement.
During the second year of the study, the performance of the on-site weatherstation's
data logger became unstable and for several periods, data was not properly logged, so
the weather data was supplemented by NOAA-RDU data and State Climate Office of
North Carolina, Reedy Creek Station data.
2.8 DCCA Measurement Methods
2.8.1 Sampling and Extraction Methods
Prior to beginning DCCA measurements for use in the study, pre-testing was
performed to determine characteristics of the wipes and their ability to retain levels of
target analytes. These tests were fully described in the Interim Report and are
summarized here.
2.8.1.1 Wipe Blank Study
A number of blank evaluations were performed to evaluate the amounts of target
compounds inherent in the extraction process and the wipes themselves. Initially,
unwashed wipes directly out of the bag were acid extracted. Blank results for arsenic
and chromium are shown in Table 2-6. This initial wipe blank analysis did not include
an analysis for copper.
Table 2-6. Wipe Blank Analyses
Sample Number
AQS-54
AQS-55
As (ug/l_)
1.2
1.3
Cr(ug/L)
0.8
1.1
To investigate the potential to reduce the background levels of As and Cr measured in
the wipe blanks, nine wipes were pretreated using an acid-wash, rinsing thoroughly
with de-ionized water, and allowing them to dry in a clean environment. Results
comparing the acid-washed wipe to the out-of-the-bag wipe blanks are shown in Table
2-7. Measurements from a control sample consisting only of nitric acid that had been
through the extraction process (with no wipe added) are also presented.
35
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-7. Results from September 2003 Wipe Comparison Study
Average
Wipe Blank (Out-of-Bag)
Digested Nitric Blank
Acid-Washed Wpe Blank
As (ug/L)
0.41
<0.10
0.2
Cr(ug/L)
0.93
<0.50
<0.50
Cu (ug/L)
2.1
0.3
1.3
It should be noted that the results for the initial out-of-bag wipe blanks experiments
(Table 2-6) were based on testing a whole 12" x 12" wipe while the data collected from
the September 2003 Wipe Comparison Study (Table 2-7) were based on using half of
a wipe.
2.8.12 Spiking Study
A number of spike studies were done to ensure that the analytes of interest could be
effectively captured by the wipes, extracted and analyzed. Initially, three samples were
prepared by spiking known amounts of stock solutions of arsenic and chromium
standards onto a clean glass plate and allowing the liquid to evaporate. Each glass
plate was then wiped using the CPSC technique and the wipes were extracted and
analyzed. Recovery results are shown in Table 2-8.
Table 2-8. Results of Spiking onto Glass
Spiked Amount
Sample AQS-56
Sample AQS-57
Sample AQS-58
Average
SD
Percent recovery (avg)
Percent RSD
As
50
33
41
41
38.3
4.6
77 percent
1 2 percent
Cr
49.75
38
46
46
43.3
4.6
87 percent
11 percent
Units
ug
ug
ug
ug
ug
ug
A small amount of residue was observed on the glass after sampling with the wiping
apparatus. This residue could be removed with further cleaning which indicated that
the stain was metal salts and not etching.
In a subsequent methods development experiment, pretreated wipes were directly
spiked with 1 ug/L, 50 ug/L and 1000 ug/L of arsenic, chromium and copper standard,
extracted, and analyzed. Results are shown in Table 2-9. The higher recoveries (as
36
-------
compared with the results in Table 2-8) indicate that extraction and analysis techniques
may be more robust than the efficacy of the wipe sampling method in picking up
materials from the surfaces of the test substrates. However, the fairly low RSDs in
Table 2-8 indicate that the sampling method results in reproducible measurements
under the conditions of the glass spiking experiment.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-9. Results of Spiking Wipes Directly
Sample
ID
SS-562
SS-563
SS-564
Arsenic
(ug/L)
1.0
47
1100
Recovery
(Percent)
100
94
110
Chromium
(ug/L)
1.0
51
1000
Recovery
(Percent)
100
102
100
Copper
(ug/L)
1.2
47
970
Recovery
(Percent)
120
94
97
In addition to liquid standards, CPSC also provided EPA with CCA dust material that
contained a known amount of arsenic. A pre-measured weight of this material was
placed directly into extraction vessels containing the acid-rinsed wipes and the
samples were extracted and analyzed. This spiking was done in duplicate and
recoveries for arsenic were 98 percent and 102 percent. As a result of these spike
studies, it was determined that arsenic, chromium and copper could be adequately
recovered from wipe samples.
2.8.2 Wipe Sampling Events
Wipe samples were taken directly from the top faces of the specimens on the
minidecks. Each specimen had three sets of nail holes and two possible sampling
areas: a BL area and a PSA area, as previously defined. The length of all wipe
samples was 15 inches to avoid contact with nail holes which were typically spaced on
16-inch centers.
The wipe sampling device (see Figure 2-9) used for the tests was designed and
constructed by CPSC staff, (CPSC staff 2003b). The wipe sampling device utilizes a
1.1 kg disc that is approximately 8.65 cm in diameter as the wiping block (note that the
actual width of 5/4" x 6" decking is approximately 5.5" or 14 cm). The sampling device
allows for the use of variable wipe lengths; for this study, 38 cm was used, yielding a
sampling area of approximately 314 cm2. The EPA- wipe method incorporated
relatively minor modifications from the CPSC method, as described in Section 2.8.5.
37
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 2-9. CPSC Wipe Sampling Apparatus
2.8.2. 1 Baseline Sampling of Boards and Minidecks
The baseline (BL) areas were wipe-sampled prior to coating using wipes that had been
pre-washed in nitric acid in an effort to reduce trace contaminants from the wipes. Prior
to beginning collecting routine wipe samples 1 month post-coating, it was determined
that Dl water rinsing steps were not sufficient in removing the nitric acid from the wipe
and the nitric acid pre-wash was discontinued due to concerns about the unnatural and
potentially detrimental effect of the residual acid on the coating. Accordingly, the
routine wipe samples taken on the PSAs at 1, 3, 7,11,15, 20 and 24 months after
coating utilized wipes that had not been acid-washed.
Individual baseline DCCA values were determined for each PSA on each specimen to
be coated and tested. The baseline DCCA of a PSA was determined by averaging the
precoating baseline DCCA from the two adjacent BL specimens on either side of the
PSA.
38
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
The PSA itself was not directly wipe-sampled prior to coating to avoid artificially
removing DCCA from the PSAs before coating them. These areas were wipe-sampled
at specified intervals using wipes that were straight out of the bag and wetted, as
described in Section 2.8.3.2.
2.8.2.2 Subsequent Wipe Sampling Events
Subsequent sampling was done with "out of the bag" wipes, simply wetted with Dl
water. The wipe methods employed are described in Section 2.8.3.3. Several more
qualitative measures were taken to qualify and document wipe-sampling events:
• Wipe sampling events were only conducted when specimens appeared dry and
when weather forecasts indicated that consistent, dry weather would prevail for the
entire sampling event. Actual climatic conditions were recorded and well-
documented throughout the entire study, including sampling events.
• During each sampling event, each minideckwas digitally photographed prior to
sampling, with BL and PSA areas identified in a running photo log.
Routine control samples that were taken included:
• Three negative control wipe samples taken; one from each of the three interior
untreated, uncoated specimens on the control minideckthat contained no treated
boards. These control measurements provided an indication of whether there was
significant atmospheric deposition of CCA analytes at the site.
• One untreated (but coated, for minidecks prefixed 1 through 12) specimen from
each minideckwas wipe-sampled during each regular sampling event. Because
there are five untreated specimens on each minideck, there are a total of 10 such
potential sampling areas. The specific areas sampled during each routine sampling
event were randomly selected for each minideck and were different for each
sampling event.
2.8.3 Wipe Sampling Methods
2.8.3.1 EPA Acid-Wash, Rinse, and Saturate with Dl Water Wipe Preparation Technique (A2
Method)
For the precoat and baseline sampling events, TexWipes TX1009 clean room wipes
(100 percent continuous filament polyester), were cut in half using a new razor blade
that had been cleaned using acetone and a lint-free wipe on an acetone-cleaned lab
bench. After cutting, the half-wipes were placed in a wide-mouth glass bottle and
soaked in a 10 percent solution of Trace Metals Grade Nitric Acid. The bottle was
39
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
placed in an oven at 85 °C overnight. When the bottle was removed from the oven, the
nitric acid solution was decanted and wipes were rinsed in the bottle five times with
deionized H2O. After the final rinse, each wipe was removed and squeezed by hand so
that they were damp but no more water could be removed. The damp wipes were
placed into individual Digitubes until used for wipe sampling. This technique was
determined to consistently yield moisture contents of 2.1 + 0.1 (1 standard deviation)
times the dry wipe weight.
2.8.3.2 EPA 2X Dl Water Wipe Preparation Technique (2X Method)
For subsequent sampling events (samples taken 1, 3, 7,11,15, 20 and 24 months
after coating), TexWipe TX1009 clean room wipes were cut in half using a new razor
blade or scissors cleaned using acetone and a lint-free wipe on an acetone-cleaned lab
bench. After cutting, the half-wipes were inserted into PTFE tubes, into which two times
the wipe weight in Dl water was added to be soaked up by the wipe. Therefore the wet
wipe was three times its dry weight. Wetted wipes were stored in their sealed PTFE
tubes until use. Sampling staff, while cutting, transferring, and wetting the wipes, wore
nitrile or latex gloves.
2.8.3.3 Standard EPA Wipe Method (Adaptation of CPSC Staff Method)
The adaptation of the CPSC staff wipe method employed by EPA for all sampling
events is described below. The differences between this method and the CPSC staff
method are listed in Section 2.8.5.
Prior to starting a new wipe sample, the sampling staff donned a new pair of
disposable nitrile or latex gloves over a second pair of gloves that were not changed
between wipes (i.e., double gloved). The rubber-coated side of the steel rubbing disk
was covered with plastic wrap. The wetted wipe was then removed from the PTFE
tube, folded in half, and placed over the plastic wrap and secured with a plastic tie-
wrap strap. The disk was lowered so that it was in contact with the wood and slid along
the tracks of the sampling apparatus forward and backward for five 38-cm (15-inch)
strokes, where each stroke consisted of one forward and back movement. The speed
of sampling was variable depending on the quality of the area being wiped, with
rougher wipe areas sometimes requiring longer sampling times (slower average
speeds). Splinters and sampling area imperfections could "hold up" the sampler
requiring adjustment of the horizontal force (the only vertical force was the weight of
the disk) exerted on the weight to continue moving it forward. However, the poly wipe
material is very resilient and no losses of wipe material due to snagging were
observed. The wipe was rotated 90° on the rubbing disk, which was then slid forward
and back for five more strokes, for a total of 10 front-and-back strokes. Sampling staff
then removed the wipe from the disk and placed it back into its PTFE extraction tube.
40
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Wood splinters larger than a grain of rice were removed from the wipe prior to placing it
in the extraction vessel.
After the sample was taken, the plastic wrap was discarded and the wiping apparatus
was decontaminated by wiping the rails of the apparatus with lint-free wipes wetted
with Dl water.
2.8.4 Wipe Extraction and Analysis Techniques
Wipe samples were prepared for analysis using techniques employed by CPSC staff
(2003b) and Stilwell, et al. (2003), with adaptations for use with laboratory equipment
available for this project. As such, a microwave- or heat-assisted extraction and
digestion procedure was employed comparable to that used in prior studies, and
similar to SW-846 Methods 3051 and 3052.
Precleaned disposable digestion vessels were used for sample collection and
digestion. All volumetric glassware was prepared by acid cleaning by leaching with hot
1:1 nitric acid for a minimum of two hours, then rinsed with deionized water and dried in
a clean environment. 30 ± 0.1 ml 10 percent nitric acid (trace metal grade HNO3, Dl
H2O) was added slowly to the digestion vessel containing the wipe sample to allow for
preextraction. Once any initial reaction had ceased, the sample was capped and
introduced into an Environmental Express HotBlock metals digestion system. The
vessels were placed into the digestion system and heated for 1 hour at 95°C. After
digestion system extraction, sample vessels were allowed to cool for a minimum of 5
minutes prior to removing them from the system. The liquid was poured off into a 100
ml volumetric flask. In addition, as much extraction liquid as possible was squeezed by
hand from each wipe. The gloves, funnels, and flask necks were rinsed with Dl H2O.
The extracted wipe was then placed back into the extraction flask with an additional 30
ml of 10 percent HNO3.
Again, the vessels were placed into the digestion system and heated for 1 hour at
95 °C. After extraction, the liquid was poured off into the aforementioned 100 ml
volumetric flask. As much extraction liquid as possible was squeezed by hand from
each wipe and the gloves, funnels, and flask necks were rinsed with Dl H2O. The wipe
was placed back into the extraction vessel and 20 ml of 10 percent HNO3 was added
to each extraction vessel before the digestion system cycle was repeated. The extract
was then poured into the 100 ml volumetric flask. Deionized water was used to rinse
the extraction vessel; rinsate was added to the 10OmL volumetric flask. If necessary,
deionized water was added to take the contents to the 100 ml level. The contents of
the 100 ml flasks were then transferred to and stored in two plastic tubes (duplicate or
split samples) with caps, manufactured by SCP Science of virgin polypropylene and
certified contaminant-free. One split sample was sent to an analytical laboratory for
analysis, while the other was archived, under refrigeration or freezer storage.
41
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Analyses for total arsenic, chromium, and copper were conducted by STL in
Savannah, Georgia, using a modification of SW-846 Method 6020 (ICP-MS). STL
utilizes ICP-MS for arsenic analysis, modifying the technique to utilize hydrogen
plasma, rather than argon as classically performed. This modification eliminates
concerns over the formation of Ar40CI35, which can create a positive bias when
measuring As. STL-Savannah's analytical method has reporting limits of 0.10 ug/L for
all three CCA analytes (this corresponds to a DAs of 0.000032 ug/cm2). Per the
specified analytical method, the hold time for all metals other than mercury is 6 months.
Each set of samples submitted included blind blanks and spiked samples for continued
monitoring of laboratory performance during the project.
2.8.5 Differences between EPA and CPSC Procedures
Differences between the CPSC staff and EPA 2X methods for collection and analysis
of surrogate wipes on CCA-treated wood include:
• EPA used plastic wrap to cover the rubber-coated side of the rubbing disk rather
than Parafilm.
• C-clamps were not used by EPA to secure the horizontal wiper (because the
boards being wiped are part of a deck structure). An assistant held the wiper in
place.
• In the EPA method, poly wipes were immediately placed directly into the vessels in
which extraction will take place.
• A three-step extraction and digestion procedure was used by EPA rather than the
CPSC staff one-step water bath extraction and digestion.
• EPA used a half of a polyester wipe folder over itself once before being affixed to
the block, while CPSC used a quarter wipe which did not need to be folder over.
As such, EPA's wipe pattern was slightly wider than CPSC's.
• EPA used a 2X Dl water spike (wetted wipe weight is three times the dry wipe
weight) to prewet the wipes while CPSC staff used a 1x, 0.9-percent saline solution
spike (wetted wipe weight is two times the dry wipe weight).
• EPA used a 38-cm (15-in) wipe length (nominal 314 cm2 sampling area) while
CPSC staff used a 50-cm (19.7-in) wipe length (nominal 405 cm2 sampling area).
42
-------
2.8.6 Calculation of DCCA from Extraction Fluid Concentrations
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Raw data from the analytical laboratory were reported in units of |jg/L and represent
the mass of analyte per unit volume of extraction solution sent to the laboratory. For
standard wipe sampling results, data were converted to units of mass of analyte per
unit surface area wipe-sampled (i.e., DCCA), in units of ug/cm2, using the following
equation:
V
C
1000
DCCA
A
(Equation 2-1)
where
CDCCA= DCCA of a sample (ug/cm ),
CDF = Concentration of analyte in extraction fluid (ug/L),
V = Total volume of extraction fluid (ml_) =100 ml_ (typically),
A = Area of wiped surface (cm2) = 314 cm2
Using the reporting limits of 0.1 ug/L from the analytical laboratory, this correlates to a
sampling detection limit of 0.000032 ug/cm2.
2.9 Preparation and Analysis of Coating Samples
Total arsenic, chromium, and copper in the bulk coatings were determined in a manner
similar to that used to analyze the wipe samples (acid digestion and extraction followed
by ICP-MS). The coating was thoroughly shaken to ensure homogeneity and then an
aliquot was transferred to a tared PTFE digestion vessel and allowed to dry. Following
loss of volatiles through drying, the residue was digested using concentrated nitric acid
as described in EPA SW-846 Method 3052. Additionally, hydrofluoric acid (HF) was
added as necessary to ensure complete digestion in accordance with the method.
Specifically, 9 ml of concentrated nitric acid and 3 ml of concentrated hydrofluoric acid
were added and the samples were microwaved for 25 minutes at 50 psi followed by 45
minutes at 80 psi. The HF was neutralized before analysis by addition of 30 ml of a 4-
percent boric acid solution prior to a second microwave digestion for 30 minutes at 50
psi. This process converts the HF to BF3, which is a gas that leaves the sample. The
digestate was quantitatively transferred to a volumetric flask and brought to 100 ml with
10 percent nitric acid before submission to the analytical laboratory for ICP-MS
analysis (SW-846 Method 6020).
These results are presented in Section 4.4.
43
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
2.10 Archiving of ICP-MS Samples
Analysis of the samples by ICP-MS consumed only a fraction of the submitted sample.
ARCADIS/EPA archived an aliquot of each digestate, to be held until the completion of
the project. Samples were archived by storing them in TFE or PFA containers under
refrigeration. Additionally, any remaining sample volume at the analytical laboratory
was archived until the analytical results were confirmed.
2.11 Moisture Analysis of Wood Specimens
Initial precoatwood moisture content was measured using a hand-held meter, after the
technique had been qualified via side-by-side testing with the drying oven technique,
ASTM D4442 (Primary Oven Drying). Per ASTM D4442, a small representative wood
sample was weighed prior to drying overnight at 103° C in a forced air oven. After 24
hours, the sample was cooled in a desiccator, weighed and returned to the oven. The
process was repeated until weight changes between weighings were within ± 5
percent.
The results of the moisture analyses are included in the Specimen Characterization
Data contained in Appendix B.
2.12 Miscellaneous Samples
Other samples that were collected during the study and archived or analyzed are
summarized in Table 2-10.
Table 2-10. Miscellaneous Samples Collected
Sample Description
Unaltered coating
Leftover brush-applied coating
Brush wash water
Wood
# Samples Analyzed
2 for each coating
N/A
2 for each brush type
Up to 4 cores per board
# Samples to be Archived
Leftover coating was stored
1 for each coating and wood type
Brushes were retained
Leftover wood was stored
2.13 Quality Control Samples
The following types of quality control samples were included: (1) positive (CCA-treated,
uncoated) controls, (2) negative (untreated, uncoated) controls, (3) cross-
contamination controls, (4) wipe frequency (rewipe, abrasion) controls. Each is
discussed briefly below.
44
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
2.13.1 Positive (CCA-Trested, Uncoated) Controls
The three minidecks prefixed by the number 13 were constructed in exactly the same
way as the minidecks for coatings #1 through #12, except that they were not coated.
The results from these minidecks were used to determine how the DCCA values vary
overtime without considering the effect of coating. As such, control deck DCCA values
could be determined at each sampling event for comparison with DCCA from the
coated decks. These positive control results were used to assess coating performance
in the statistical data analysis.
The advantage to using the positive control minidecks for this purpose was that other
potentially important factors were captured in the minideck 13 (uncoated) results and
thus factored into the statistical analyses. These included: the effect of rinsing the
boards prior to coating (assessed via comparison of precoat and samples taken 1
month after coating for the positive control minidecks), the effect of weathering
between subsequent sampling events, and the effect of climatic conditions during the
sampling itself. The use of the positive control minideck DCCA results thus allowed for
these and other potential sources of bias to be considered in the data analysis.
These results are represented by the data presented for Deck #13 (uncoated).
2.13.2 Negative (Untreated, Uncoated) Controls
The single uncoated minideck, labeled BC (for "blank control"), consisting of five
untreated specimens, was used to routinely take blank samples to measure the
background levels or atmospheric deposition of CCA analytes. Wipe samples were
taken from the same areas of the middle three boards on this deck during each
monitoring event, similar to samples taken from the other minidecks.
These results are presented in Section 4.7.
2.13.3 Wipe Frequency (Rewipe, Abrasion) Controls
Each CCA-treated test specimen on each minideck included two sampling areas: a
PSA (suffixed "M") and an adjacent baseline sampled area (suffixed "BL"). BL areas
were those that were initially wiped prior to coating to establish baseline DCCA. A
subset of the BL areas were resampled during each of sampling events 1 through 5. All
of the BL areas on one of the three minidecks per coating (as well as the positive
control minidecks) were sampled during each of the five sampling events. Because
there were three minidecks per coating, a given BL area was resampled every third
sampling event. As such, the coatings on these sections of lumber were not abraded
by wiping to the same extent as the coatings on the PSAs, so the effect that wiping has
on coating efficacy could be investigated.
45
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
These results are presented in Section 4.6.
2.13.4 Analytical Laboratory Control Samples
A series of laboratory control samples were sent with each batch of samples tested by
the analytical laboratory. Each set of digested wipe samples submitted to the analytical
laboratory typically included five percent additional blind field blanks (extracted unused
wet wipes), one blind blank (extraction fluid only), one set of three different
concentration-spiked samples, and duplicates (split samples) for five percent of the
wipe sample digestates that were analyzed to assess laboratory performance. Control
samples were not identified as such to the analytical laboratory performing the
analyses. For example, assuming that a total of 200 wipe samples were collected and
shipped to the analytical laboratory in a single batch, the following additional samples
were to be included:
• Ten (10) field blank samples prepared by taking unused wetted wipes and
extracting them in accordance with the procedures previously specified
• One (1) blank consisting of extraction fluid only
• One (1) digestion fluid sample spiked to 1.0 ug/L (0.015 ug in 15 ml digestion
fluid) with As, Cr, and Cu (depending on the range of DCCA in the particular batch
of test samples being collected, this spike could instead be 10,000 ug/l to more
appropriately capture a higher range)
• One (1) digestion fluid samples spiked to 50 ug/L (0.75 ug in 15 mL digestion fluid)
with As, Cr, and Cu
• One (1) digestion fluid samples spiked to 1000 ug/L (15 ug in 15 mL digestion
fluid) with As, Cr, and Cu
• Ten (10) duplicates (selected split samples of digested wipes from actual samples
generated)
Furthermore, the analytical laboratory analyzed project-specific post-digestion spiked
samples for each analyte, as well as equipment blanks run on each batch of samples.
Results of the laboratory control samples are discussed in Section 5.
2.14 Paint Chip Sampling
Fragments of cracked and peeling paint were collected from the painted minidecks at
the end of the study period and analyzed to determine if elevated CCA analyte
46
-------
concentrations were associated with paint chips. Testing involved collecting samples of
chipping/peeling paint from the minidecks coated with paint using cleaned tweezers.
These decks were identified with the prefixes 9- and 10-, as they correspond to
coatings #9, a water-based (latex) paint, and #10, an oil-based (enamel) paint. For
both paints, the wood surfaces were coated with a commonly-available primer prior to
coating with the paints themselves. Samples were taken from each of the CCA-treated
boards of each of the three minidecks coated with the given paint and composited into
a single sample, homogenized, subsampled, weighed, digested and analyzed. The
same general procedure was used for the untreated boards from the three minidecks
coated with the given paint. For each composite sample, the extraction fluid was split
so that duplicate analyses for total arsenic, total chromium, and total copper could be
run by the analytical laboratory. Table 2-11 lists the samples that were analyzed.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 2-11. Composite Samples Taken for Paint Chip Sampling/Analysis
Coating #
9
9
10
10
Boards to
Sample
CCA-Treated
Untreated
CCA-Treated
Untreated
#of
Minidecks
3
3
3
3
# of Boards
(total)
12
15
12
15
Type of
Sample
Composite
Composite
Composite
Composite
Number of
Samples
Duplicate
Duplicate
Duplicate
Duplicate
Plastic tweezers were decontaminated prior to sampling and in between each of the
four samples. Approximately equal amounts of paint fragments were removed from
each of the five untreated wood boards on each of the three coating #9 minidecks.
Individual paint fragment samples were taken from various places on each board to
ensure a representative sample. All paint fragments for a given sample were added to
a single digitube which had been preweighed in the laboratory prior to sampling. A total
paint mass of 1.5 grams was targeted for collection. Sampling was repeated
accordingly for the five untreated wood boards on each of the three coating #10
minidecks, then the four CCA-treated wood boards on each of the three coating #9
minidecks and lastly, the four CCA-treated wood boards on each of the three coating
#10 minidecks. From each digitube, a representative 0.5-gram subsample of paint
fragments was removed and placed in another digitube. The balance of the paint
fragments (~1.0 gram) was left in each of the original digitubes. The Digitubes were
weighed and reweighed to determine the mass of paint in each sample. The contents
of each of the eight digitubes were then digested in their entirety, in accordance with
the procedures specified in Section 2.9. Each of the 1.0-gram composite sample
digestates were split into two equal duplicate samples and shipped to the analytical
laboratory for determination of total arsenic, chromium, and copper in accordance with
the procedures in Section 2.9.
47
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Comparison of the 0.5 and 1.0 gram paint fragment digestate analysis results were
used to help determine if there was any impact on analysis by the paint matrix itself.
These results are presented in Section 4.4.
48
-------
3. Study Results
The discussion of results focuses on DAs, rather than the other DCCA measurements,
because it is considered the analyte with the most critical exposure risks. However,
DCr results may become important if cleaning products change the chromium
speciation from (III) to (VI), for example. In general the trends for DCr and DCu are
similar to those presented for DAs. Additionally, all of the raw and reduced DCCA is
provided in appropriate sections of the appendix.
3.1 Source Characterization and Sampling Events
3.1.1 Distribution of Baseline Data
Source deck specific random effects analysis of the log-transformed, pre-averaging
baseline measurements resulted in the summary statistics presented in Table 3-1,
which shows that although the mean measured DAs on the newer source deck C
boards is considerably lower than on the older source deck A boards, the within- and
between-board variations associated with each source are comparable.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-1. Summary Statistics for Source Deck Specific Random Effects Analysis
Statistic
Mean In(DAs)
Within Parent Board Variance
Between Parent Board Variance
Source A
0.68
0.095
0.169
Source C
0.11
0,070
0.156
Figures 3-1 through 3-6 provide simple box plots of the distribution of baseline data.
These figures are grouped by analyte and source, and plot coating on the x-axis versus
baseline DCCA on the y-axis. Appendix J includes similar plots grouped by board
instead of by coating.
In each of the box plots presented, the box stretches from the lower hinge (defined as
the 25th percentile) to the upper hinge (defined as the 75th percentile) and therefore
contains the middle half of the scores in the distribution. The median is shown as a line
across the box. Therefore, % of the distribution is between this line and the top of the
box and % of the distribution is between this line and the bottom of the box. The plus
symbol (+) shows the mean. In these plots, the bars on either side of the box define the
minimum and maximum.
49
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
6-
5 -
CM (
E 4
o
1
.0
c
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Coating, Precoat,
1.75 -
1.50 -
0 1.25-
S
0 1-00 -
Q.
Q.
O
0 0.75-
0.50 -
0.25 -
T
+
1
1
:
T
i
2
+
T
-
3 4
j
-f
I
5
!_
6
T
4-
T
Source = A
-^i
—
T1
T
1
—
~r
T-
+
r
i
•*•
j
L
]
^ i J¥
7
Coating
8
ID
9
10
1 1 12
13
Figure 3-3. Box Plot, Baseline DCu, by Coating, Source A
3.0 -
2.5 -
CM
£ 2.0 -
JJ
U)
9)
(0
< 1.0-
0.5 -
o -
T
1
-J
1
Coating,
T E
+
3
T L
r 1
2 3 4
Precoat
- 1
T
6
Source = C
T
\
jj
J
?
T
8
+
i
9
I
10
-3L
^_
r!
T t
11 i
i
J
2
z
-t-
T1
1
13
Coating ID
Figure 3-4. Box Plot, Baseline DAs, by Coating, Source C
51
-------
Figure 3-5. Box Plot, Baseline DCr, by Coating, Source C
Coating, Precoat, Source = C
1.75 -
1.50 -
CM
E 1.25 -
O
0)
^ 1.00-
Q.
o
O 0.75-
0.50 -
0.25 -
- 1
] i
T
]
j
q $ a Q
j" ^ T >_!_' y i
2345678
Coating ID
T
J
9
ygi
i i .
10 1 !
t-
1
2
T
-t"
I
13
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Coating,
3.5 '
3.0-
*£
JJ 2.5-
D)
E 2.0-
o
£ 1.5 -
O
i.o -
0.5-
-t-
T
q
3 r
T x :
2
--
4.
I
3 4
]
Precoat, Source = C
~i
J
I
-
T
6
I
-
J
7
_
i
8
H
T ^
T
T
j
J_^
4-
--
9 10 11 12 13
Coating ID
Figure 3-6. Box Plot, Baseline DCu, by Coating, Source C
52
-------
3.1.2 Wood Core Sample Data
The wood core data for source A and source C are summarized in Tables C-5 and C-6
in Appendix C. The tables show averages, standard deviations, and RSDs for each
board and include summary statistics at the end of each table for each source (A and
C). RSDs should not be interpreted as indicators of data quality, but rather as
indicators of natural variability in CCA retention within and between boards. Note that
some of the boards listed in these tables were not actually used in the construction of
the minidecks; however, they are included here for completeness. Complete data
showing the results of each individual wood core sample measurement are provided in
Appendix K.
The overall average results, reported as solid-phase concentrations in Table 3-2,
generally compare favorably with the expected concentrations and ratios of
concentrations of CCA analytes calculated in Section 2.2.3. There are some wood core
sample data points that are clear outliers and overall variability is relatively high at
about 50 percent RSD. Additionally, because of the way that boards are cut from the
tree, taking core samples from the wide face - as done in this study - increased the
likelihood that heartwood (which does not accept and retain CCA as readily as
sapwood) was sampled. If the narrow faces had been sampled, sapwood would have
more consistently been sampled and the values would have likely been more
consistently high and less variable (Lebow, 2006).
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-2. Comparison of Nominal, Source A, and Source C CCA Actives Composition
As (mg/kg)
As as As2Os (mg/kg)
As2O5 (percent)
Cr (mg/kg)
Cr as CrO3 (mg/kg)
CrO3 (percent)
Cu (mg/kg)
Cu as CuO (mg/kg)
CuO (percent)
Nominal CCA
4,255
34.0(30.0-38.0)
5,938
47.5 (44.5-50.5)
2,317
18.5(17.0-21.0)
Source A
1,645
2,522
31.7
2,045
3,933
49.4
1,203
1,506
18.9
Source C
2,075
3,182
35.2
2,095
4,029
44.5
1,465
1,834
20.3
3.1.3 Coating Application
The total volume and mass of coating applied to the A, C, and N boards (the untreated
boards) on each minideck were determined as illustrated in Figure 3-7. No coating
53
-------
mass data were recorded for coating #1, because the decision to measure mass was
made after the coating #1 decks were coated. Measured volumes were rounded to the
nearest increment of 5 ml. Mass measurements are reported to the nearest 0.1 g.
A complete tabulated summary of the coating application data is included in Appendix
C, Table C-8. The summary includes average, standard deviation, and RSD for each
combination of board type and coating. RSDs should not be interpreted as indicators of
data quality, but rather as indicators of the variability inherent in the coating methods
utilized. Any variability in coating volume or mass among decks is part of the natural
variability one would see in practice, on real decks being coated by homeowners. Film
thickness proved to be quite difficult to measure on wood substrates and was not
appropriate for non-film forming coatings; as such, film thicknesses were only
measured on some minidecks and these data are not presented.
|
I
Mfcnc
Jl
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
1234-5
ID 11 Q
Figure 3-7. Coating Application (total of triplicate minidecks on both A and C sources)
3.2 Coating Performance Data
Tabulated summaries of the wipe sampling data for DAs, DCr and DCu is included in
Appendix C (Tables C-9, C-10 and C-11, respectively). A complete set of wipe
sampling data, including data for both the PSA (M) samples and baseline (BL) samples
at each sampling event, is provided in Appendix N.
3.2.1 DCCAvs. Time
Baseline and time series DCCA values for each CCA analyte, sorted by coating, and
averaged over the combined A and C sources are shown graphically using a natural
log-transformed DCCA scale on the y-axis in Figures 3-8 through 3-10. The tabulated
54
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
data used to create these graphs is included Appendix C (Table C-9). Several
generalized observations can be made from these data. First, each coating, as well as
the positive controls (uncoated minidecks), generally show a significant decrease in
DCCA between baseline (prewash, precoat) and samples taken 1 month after coating.
This suggests an impact on DCCA from rinsing or washing the minidecks. Second, the
coated minidecks all have lower DCCA than the positive controls, which indicates that
coating (using any of the coatings tested) initially mitigates DCCA to some degree.
Third, over the course of this study, DCCA increases with time after coating, most likely
due to the effects of weathering and possibly abrasion on the coating, although it
should be noted that the uncoated positive controls show similar, though generally less
pronounced, trends.
Note that confidence intervals are not shown on Figures 3-8 through 3-10 to avoid
obscuring the general trends. Refer to Figures 3-28 and 3-29 for graphical
representations that show statistically significant differences between coatings by time
and also Tables 3-11 and 3-12 that present p-values for differences between means at
each time point and present estimated 95% confidence intervals for estimated percent
reductions between coatings and the positive controls at each time period.
3.2.2 Graphical Data Analysis
The experimental design and associated statistical methods used in this section were
introduced in Section 1.3.
The total number of valid DCCA observations in the dataset is 1,090. The effects of
interest are coating, grain orientation, source deck, and time, and thus there are up to
four-way interactions that need to be considered.
The analysis of variance models are very useful for identifying significant factor effects,
but are less informative with regard to understanding the nature of the factor effects,
especially those involving higher-order interactions. Therefore, graphical displays are
extensively used to supplement the standard statistical analyses presented in the
following sections. The intent is to provide a graphical representation of the patterns
and trends in the data that are indicated by the numerical analysis of variance results.
The statistical analysis begins with a comprehensive graphical presentation and
summary of the data, in order to provide an objective and transparent "look" at the
experimental data unadulterated by any statistical model imposed on the data, and to
aid the interpretation of the formal statistical analysis of variance results presented later
in this discussion.
55
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
—•—Coating 1
-•-Coating 2
Coating 3
X Coating 4
—*—Coating 5
—•—Coating 6
—I—Coating 7
Coating 8
Coating 9
Coating 10
— —Coating 11
Coating 12
Positive control (13)
10 15
Time (months since coating)
20
25
Figure 3-8. Average DAs vs. Time for All Coatings
56
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
—•—Coating 1
—•—Coating 2
— —Coating 3
—X— Coating 4
-*- Coating 5
* Coating 6
—I—Coating 7
- Coating 8
- Coating 9
—•—Coating 10
Coating 11
Coating 12
•— Positive Control (13)
10 15
Time (months since coating)
20
25
Figure 3-9. Average DCr vs. Time for All Coatings
57
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
—•—Coating 1
-•-Coating 2
Coating 3
—X— Coating 4
-*- Coating 5
* Coating 6
—I—Coating 7
- Coating 8
Coating 9
Coating 10
Coating 11
Coating 12
—x— Positive control (13)
10 15
Time (months since coating)
20
25
Figure 3-10. Average DCu vs. Time for All Coatings
58
-------
The experiment included several factors and a comprehensive graphical summary of
the data, while very enlightening, requires an understanding of why and how certain
graphical elements were constructed. First the data and the graphical display for a
single minideck are described. A composite data display in which all of the data from
the thirty-nine minidecks, displayed on a single page, is then presented. This graphic
and other composite data displays facilitate comparison among coatings. Next, plots of
various averages designed to illustrate the effects of source deck, grain orientation,
time and coating type are presented. Finally, a grand composite display that includes
all of the relevant graphical elements on a single page is presented.
3.2.2.1 Single Minideck Data Plot
The data from Minideck 9-A is used to illustrate below how similar plots were
generated for each minideck. Minideck 9-A was chosen for illustration because
individual data points are well spread out and easily identified; and because it is
"typical" in the sense of manifesting certain patterns and trends that are common to
other minidecks. Figure 3-11 shows all of its PSA DAs data graphically and Table 3-4
presents the same data in tabular form.
Each minideck contains four CCA-treated specimens, with each specimen having
different source and grain characteristics: Source Deck A, Grain Up; Source Deck A,
Grain Down; Source Deck C, Grain Up; and Source Deck C, Grain Down. Figure 3-11
and subsequent minideck data plots are color coded according to board source deck
and grain orientation as shown in Table 3-3.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-3. Legend for Individual Minideck Data Plots
Color
Dark Blue
Green
Cyan (Light Blue)
Red
Source
A
C
A
C
Grain Orientation
Up
Up
Down
Down
Each specimen contributes eight measurements of DAs to the study: one baseline
measurement and seven post-baseline measurements at successive time points after
treatment with one of the thirteen coatings (actually twelve coatings plus the uncoated
control). Thus, each minideck plot displays thirty-two DAs measurements. However,
the manner in which the measurements are plotted is different for the baseline and
post baseline measurements.
59
-------
Baseline (time = 0) measurements are plotted as horizontal lines extending the full
width of the plot. Note that even though the baseline measurement is plotted over the
entire length of the time axis, there is only one baseline measurement per specimen (at
time = 0). One reason that the baseline values are plotted in this manner is because
the baseline measurements may serve as the "before treatment" measurement for
each of the "after treatment" measurements.
The post baseline measurements are plotted in time-order sequence with filled-circles
of the color matching the specimen characteristics (source, grain orientation), as
indicated in Table 3-3. The lines connecting the points are for clarity, and should not be
misconstrued as to infer that DAs values necessarily changed linearly (or even
smoothly) between successive time periods. Note, from Table 3-4, that the log-
transformed baseline values (column Time = 0) are 1.090, 0.402, -0.291, and -0.740
for the A-Up, A-Down, C-Up, and C-Down boards respectively. These values are
plotted in Figure 3-11 as the dark blue (Bl), cyan (Cy), green (Gr), and red (Re)
horizontal lines respectively. The same color coding is used in plotting the post
baseline measurements versus time. The log-transformed DAs values plotted in the
graph are from the data table columns with headings "1 mo." through "24 mos.". For
example, for Board A-T-M3 (Source Deck A, Grain Up), the seven time-ordered, post
baseline, natural log-transformed As values (-4.065, -3.976, -3.043, -2.917, -2.067, -
1.967, -1.407) are plotted versus time with dark blue circles and connecting lines.
One purpose of presenting the baseline measurements is to explain some of the
variability in the post-baseline measurements. A specimen with higher than average
initial DAs measurements could reasonably be expected to have higher-than-average
post baseline DAs measurements as well. Thus, in Figure 3-11, for Minideck 9-A, it can
be seen that the Source Deck A specimens had both higher baseline, and post-
baseline, measurements than did the Source Deck C specimens, regardless of grain
orientation or time period.
Note that the dark blue and cyan time plots (Source Deck A) overlap considerably and
that the red and green time plots (Source Deck C) also overlap to a lesser extent.
However, the blue and cyan curves lie entirely above the red and green curves. Thus,
for this minideck, the observed differences in DAs measurements between grain
orientation from the same source deck appear to be less pronounced than the
differences between specimens from different source decks.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
60
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 9A
CM
E
o
CD
13
CO
Q
1—I
E
_c
-p
CD
-P
CD
C\J
C\J
I
_
CD I
CD
^ r
ID
I
CD
I
CO
i
Bl = A,Up ATM3
Dy = A,Dn APM3
Gr = C.Up CAPM1
Re = C,Dn CBWM1
7 10 13 16
Time [months]
19 22
Figure 3-11. Data Plot for Minideck 9A
61
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-4. DAs Data for Minideck 9-A, as used in Figure 3-13
Color
Spec ID
Source
Grain
Baseline
1 mo.
3 mos.
7 mos.
11 mos.
15 mos.
20 mos.
24 mos.
Raw (Not Natural Log Transformed) Data
Blue
Cyan
Green
Red
A-T-M3
A-P-M3
C-AP-M1
C-BW-M1
A
A
C
C
Up
Down
Up
Down
2.974
1.495
0.747
0.477
0.017
0.035
0.003
0.004
0.019
0.019
0.003
0.008
0.048
0.032
0.012
0.007
0.054
0.108
0.032
0.009
0.127
0.156
0.046
0.028
0.140
0.162
0.051
0.015
0.245
0.350
0.115
0.045
Natural Log Transformed Data
Blue
Cyan
Green
Red
A-T-M3
A-P-M3
C-AP-M1
C-BW-M1
A
A
C
C
Up
Down
Up
Down
1.090
0.402
-0.291
-0.740
-4.065
-3.353
-5.987
-5.568
-3.976
-3.976
-5.823
-4.876
-3.043
-3.448
-4.470
-4.919
-2.917
-2.224
-3.448
-4.721
-2.067
-1.861
-3.090
-3.583
-1.967
-1.819
-2.978
-4.182
-1 .407
-1 .050
-2.167
-3.112
Finally, a noticeable feature of this graph is the increasing trend in DAs measurements
over time, regardless of board type. With the exception of one point from the "A-Down"
board (Cy = cyan-colored line), the time plots are strictly increasing with time. DAs
measurements are lowest at the first post-baseline measurement and highest at the
final post-baseline measurement for each board type, yet the final post-baseline
measurement is still less than the corresponding baseline DAs measurement.
3.2.2.2 Data Plots for Other Minidecks
Plots were constructed for all thirty-nine minidecks in the same fashion as the Minideck
9-A example in Figure 3-11. These plots (and corresponding data tables, similar to
Table 3-4 for the minideck 9-A example) are shown as Figures 3-12 through 3-24.
These figures enable detailed visual inspection of the data for individual minidecks.
These figures can be found in full-page size in Appendix O.
3.2.2.3 All Thirty-nine Minidecks Plotted Together
A composite graphic displaying all of the data is provided as Figure 3-25. It facilitates
comparisons among coatings and also illustrates the repeatability of the experimental
method which bears on the experimental method's validity.
62
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
CM
i-^ *~
1
Lagarithm[DAs ug/
-4 -3 -2 -1 0
ST
*1
l
r^
iM
(\T *~~
LogarithmCDfts ug/
-4 -3 -2 -1
f
h.
no
0
g r
j:
*
¥ c.
0
Dl
0 •<
1 »
"
0
Data from Mini Deck 1A
Z^
Bl — A UD AAEM1
gl _ A'jtfn Tm
Gr - C.Up CNM1
Re = C.Dn CBOM2
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 1B
^^:
Bl ~ A Up AVM3
Cy = A'.Dn AATM3
Gr = C.Up CBEM2 .
Re = C.Dn CCCM1
1 4 7 10 13 16 19 22
Tims [months]
Data from Mini Deck 1C
i
' Jfd
• \S
) y
Bl = A,Up AAJM,
Re = C.Dn CAAM2
i ,,,,,,,,
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cnT2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAEM1 A Up 1.9559 0.0827 0.0700 0.0604 0.0827 0.0821 0.2258 0.2926
Gr CNM1 C Up 1.3358 0.0413 0.0477 0.0604 0.1527 0.1183 0.1558 0.4134
Re CBOM2 C Down 1.0018 0.0258 0.0299 0.0312 0.1113 0.0973 0.1050 0.2481
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAEH1 A Up 0.671 -2.493 -2.660 -2.806 -2.493 -2.500 -1.488 -1.229
Cy AZM1 A Down 1.100 -2.455 -2.493 -2.615 -1.990 -1.821 -1.673 -0,615
Gr CNM1 C Up 0.289 -3.186 -3.043 -2.806 -1.880 -2.134 -1.859 -0.883
Re CBCM2 C Down 0.002 -3.659 -3.510 -3.468 -2.195 -2.330 -2.254 -1.394
Raw Data tug/cm"2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AVM3 A Up 2.8305 0.0604 0.0318 0.0572 0.0954 0.1285 0.1050 0.2194
Cy AATM3 A Down 1 4153 0 0264 0 0251 0 0541 0 1081 0 1399 0 1558 0 3816
Gr CBEM2 C Up 1.7651 0.1845 0.0509 0.1717 0.2608 0.1698 0.2417 0.5089
Re CCCM1 C Down 1.0177 0.0572 0.0350 0.0732 0.1113 0.1272 0.0922 0.4452
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AVM3 A Up 1.040 -2.806 -3.448 -2.860 -2.350 -2.052 -2.254 -1.517
Cy AATM3 A Down 0.347 -3.634 -3.684 -2.917 -2.224 -1.967 -1.859 -0.963
Gr CBEM2 C Up 0.568 -1.690 -2.978 -1.762 -1.344 -1.773 -1.420 -0.676
Re CCCM1 C Down 0.018 -2.860 -3.353 -2.615 -2.195 -2.062 -2.383 -0.809
Raw Data (ug/cm"2 DAs): Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAJM1 A Up 2.3058 0.0509 0.0118 0.0315 0.1113 0.0684 0.1050 0.2194
Cy ABWH4 A Down 1.0813 0.0283 0.0044 0.0245 0.0572 0.0255 0.0445 0.1463
Gr CSH2 C Up 2.0672 0.0382 0.0095 0.0254 0.1018 0.1447 0.1368 0.6997
1 Re CAAH2 C Down 0.7951 0.0382 0.0067 0.0350 0.0732 0.0744 0.0636 0.2512
.
Log Transformed Data: Months
Col Brd Src Gra PRE 1 37 11 15 20 24
Bl AAJH1 A Up 0.835 -2.978 -4.442 -3.458 -2.195 -2.683 -2.254 -1.517
Cy ABWM4 A Down 0.078 -3.565 -5.415 -3.709 -2.860 -3.669 -3.112 -1.922
Gr CSM2 C Up 0 726 -3 266 -4.652 -3 671 -2 285 -1 933 -1 990 -0 357
Re CAAM2 C Down -0.229 -3.266 -5.009 -3.353 -2.615 -2.598 -2.755 -1.381
Figure 3-12. Data from Minideck 1A, Data from Minideck 1B, Data from Minideck 1C
63
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
CM
l_^ r-
E
S°
Lagarithm[DAs u
-4 -3 -2 -1
IT
$ m
[S.
03
OJ
*-
1
il LogarithmCDAs ug/
-4 -3 -2 -1 0
ft
B
==• CO
T
CD
CM
Logarrthm[DAs ug
-4 -3 -2 -1
:
Z U,
Data from Mini Deck 2A
W^^
Re = C,DnCEM3
1 4 7 1D 13 16 19 22
Time [months]
Data from Mini Deck 2B
Hf^
Vv/^^
Bl = A.Up ABCM2 .
Re = C.Dn CANMI
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 2C
y^
Rl — A 1 In AAPLH
Bl A.Up AARM1
Re = C,Dn CBXM3
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/an"2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ADH3 A Up 2.1945 0.0413 0.0350 0.0795 0.2576 0.2223 0.2290 0.3053
Cy ABYM2 A Down 1.8605 0.1686 0.0509 0.1208 0.2417 0.1504 0.1272 0.2703
Gr CBZM3 C Up 0.7156 0.1527 0.0382 0.0636 0.1399 0.1279 0.1272 0.2989
Re CEM3 C Down 1.6697 0.1463 0.0350 0.0445 0.1940 0.3880 0.1940 0.4770
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AOM3 A Up 0.786 -3.186 -3.353 -2.532 -1.356 -1.504 -1.474 -1.186
Cy ABYM2 A Down 0.621 -1.780 -2.978 -2.113 -1.420 -1.894 -2.062 -1.308
Gr CBZM3 C Up -0.335 -1.880 -3.266 -2.755 -1.967 -2.057 -2.062 -1.207
Re CEM3 C Down 0.513 -1.922 -3.353 -3.112 -1.640 -0.947 -1.640 -0.740
Raw Data (ug/cm"2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ABCM2 A Up 2.1467 0.2322 0.1208 0.1177 0.5089 0.3944 0.2099 0.6679
Cy AAHH4 A Down 1.0813 0.0143 0.0105 0.0124 0.0636 0.0515 0.0254 0.0700
Gr CBIH1 C Up 1.0177 0.1336 0.0986 0.0668 0.2640 0.2980 0.0891 0.2417
Re CANMI C Down 1.3199 0.0732 0.0258 0.0289 0.1304 0.0855 0.0668 0.0445
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ABCM2 A Up 0.764 -1.460 -2.113 -2.140 -0.676 -0.930 -1.561 -0.404
Cy AAHM4 A Down 0.078 -4.247 -4.556 -4.390 -2.755 -2.966 -3.671 -2.660
Gr CBIM1 C Up 0.018 -2.013 -2.317 -2.706 -1.332 -1.211 -2.419 -1.420
Re CANMI C Down 0.278 -2.615 -3.659 -3.543 -2.037 -2.459 -2.706 -3.112
Raw Data Cug/cm'2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AARM1 A Up 3.2768 0.0668 0.0318 0.0302 0.2035 0.1622 0.1304 0.2799
Cy APM1 A Down 3.3871 0.1527 0.0763 0.1368 0.3498 0.1832 0.4770 0.7633
Gr CBYM2 C Up 0.8428 0.2926 0.0668 0.0668 0.4134 0.3371 0.2258 0.3816
Re CBXM3 C Down 0.9064 0.1527 0.0413 0.0350 0.1686 0.1584 0.1940 0.4134
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AARM1 A Up 1.187 -2.706 -3.448 -3.500 -1.592 -1.819 -2.037 -1.273
Cy APM1 A Down 1.220 -1.880 -2.573 -1.990 -1.060 -1.697 -0.740 -0.270
Re CBXM3 C Down -0.098 -1.880 -3.186 -3.353 -1.780 -1.843 -1.640 -0.883
Figure 3-13. Data from Minideck 2A, Data from Minideck 2B, Data from Minideck 20
64
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 3A
Bl = A.Up Al
Cy - A.Dn Al
5r = C.Up CNM3
Re = C.Dn CCEM2
1 4 7 1D 13 16 13 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
1
Months
7 11 15
20
24
Bl ATM1 A Up 2,3694 0.0445 0.0194 0.0954 0.1081 0.2112 0.1876 0.9541
Cy ALMS A Down 0.7951 0.0086 0.0073 0.0541 0.0795 0.1975 0.0986 0.4134
Gr CNM3 C Up 1.6697 0.0032 0.0028 0.0172 0.0509 0.1275 0.1177 0.2831
Re CCEM2 C Down 1.5584 0.0095 0.0030 0.0350 0.1081 0.2544 0.3498 0.4134
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl ATM1 A Up 0.863 -3.112 -3.942 -2.350 -2.224 -1.555 -1.673 -0.047
Cy ALMS A Down -0.229 -4.757 -4.919 -2.917 -2.532 -1.622 -2.317 -0.883
Gr CNM3 C Up 0.513 -5.751 -5.878 -4.065 -2.978 -2.059 -2.140 -1.262
Re CCEM2 C Down 0.444 -4.652 -5.812 -3.353 -2.224 -1.369 -1.050 -0.883
Data from Mini Deck 3B
\
1 4 7 10 13 18 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AAGM3 A Up 2.4012 0.0477 0.0248 0.0541 0.1304 0.1050 0.1272 0.4134
Cy AAFM1 A Down 1.9718 0.0668 0.0248 0.0795 0.1240 0.0868 0.1654 0.2099
Gr CBJH2 C Up 3.1963 0.0086 0.0060 0.0223 0.0763 0.1279 0.1431 0.5407
Re CANM3 C Down 1.5425 0.0029 0.0017 0.0238 0.0668 0.0735 0.2926 0.5089
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AAGM3 A Up 0.876 -3.043 -3.697 -2.917 -2.037 -2.254 -2.062 -0.883
Cy AAFM1 A Down 0.679 -2.706 -3.697 -2.532 -2.087 -2.444 -1.800 -1.561
Gr CBJM2 C Up 1.162 -4.757 -5.109 -3.805 -2.573 -2.057 -1.944 -0.615
Re CAHM3 C Down 0.433 -5.833 -6.407 -3.736 -2.706 -2.611 -1.229 -0.676
Data from Mini Deck 3C
A,Up AADM2
A.Dn ABWM2
Re = C.Dn CAAM1
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cm'2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AADH2 A Up 3.3076 0.0572 0.0305 0.0604 0.1145 0.0929 0.0636 0.3180
Cy ABWH2 A Down 1.1449 0.0445 0.0286 0.0509 0.1018 0.0483 0.1304 0.3498
Gr CCDH1 C Up 1.7174 0.0041 0.0038 0.0153 0.0477 0.1323 0.0859 0.3816
Re CAAH1 C Down 1.0018 0.0099 0.0108 0.0169 0.0636 0.0394 0.0922 0.2481
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AADH2 A Up 1.196 -2.860 -3.489 -2.806 -2.167 -2.377 -2.755 -1.146
Cy ABWH2 A Down 0.135 -3.112 -3.554 -2.978 -2.285 -3.029 -2.037 -1.050
Gr CCDM1 C Up 0.541 -5.489 -5.568 -4.182 -3.043 -2.023 -2.455 -0.963
Re CAAH1 C Down 0.002 -4.619 -4.527 -4.083 -2.755 -3.233 -2.383 -1.394
Figure 3-14. Data from Minideck 3A, Data from Minideck 3B, Data from Minideck 3C
65
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 4A Raw Data (ug/cm-2 DAs) : Months
LagarithmtDAs us/cm^]
-4-3-2-10 1 2 :
| in
Z to
^
OJ
LogarithmtDAS ug/cm^l
-4-3-2-10123
Natural LogarcthmCDAs us/txift Natural
-8 -7 -6 -5 -4-3-2-10 1 2 3 -8 -7 -6 -5
^
Bl = A.Up AT.M2
Cy ~ A.un ABGM4
Re = C.Dn CAOM2
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ATM2 A Up 2.4648 0.0572 0.0572 0.0604 0.1845 0.1600 0.4452 0.5089
Cy ABGH4 A Down 1.6379 0.0382 0.0245 0.0668 0.0223 0.1730 0.2354 0.7315
Gr CCDM2 C Up 2.3694 0.1558 0.0572 0.0763 0.1368 0.1349 0.0572 0.5089
Re CADH2 C Down 0.8428 0.1717 0.0795 0.0477 0.1749 0.1119 0.1336 0.3085
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ATM2 A Up 0.902 -2.860 -2.860 -2.806 -1.690 -1.833 -0.809 -0.676
Cy ABGM4 A Down 0.493 -3.266 -3.709 -2.706 -3.805 -1.754 -1.447 -0.313
r.r- rrnMo r Tin o RR^ -1 ffc>Q -o «Kn -o R7^ -1 qon -o nn-a -9 ftfin -ri fi7fi
1 4 7 10 13 16 19 22 Re CADH2 c Doffn _0.171 -1.762 -2.532 -3.043 -1.743 -2.190 -2.013 -1.176
Time [months)
Data from Mini Deck 4B Raw Data Cug/citr2 DAs) : Months
•^?
Bl = A,Up ABCM1
Re = C.Dn CAMM2
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 4C
•
^/*
. XC^
y^=^s
Bl = A Up AIM3
Cy - A'.Dn AQM2
Re = C.Dn CBTM4
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ABCH1 A Up 1.8446 0.1145 0.0604 0.1050 0.2067 0.1622 0.2004 0.4134
Cy AAHH1 A Down 0.8269 0.1304 0.0318 0.0636 0.1908 0.1857 0.2576 0.3180
Gr CBMH2 C Up 0.9382 0.0445 0.0204 0.0267 0.0859 0.1107 0.0732 0.2958
Re CAMM2 C Down 1.4948 0.1336 0.0445 0.0636 0.2226 0.0935 0.1495 0.3816
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl ABCH1 A Up 0.612 -2.167 -2.806 -2.254 -1.576 -1.819 -1.608 -0.883
Cy AAHM1 A Down -0.190 -2.037 -3.448 -2.755 -1.656 -1.683 -1.356 -1.146
Gr CBMM2 C Up -0.064 -3.112 -3.895 -3.622 -2.455 -2.201 -2.615 -1.218
Re CAMH2 C Down 0.402 -2.013 -3.112 -2.755 -1.502 -2.370 -1.901 -0.963
Raw Data (ug/cnT2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AIMS A Up 2.5920 0.1622 0.0668 0.2258 0.4134 0.4961 0.3149 1.2085
Cy AQH2 A Down 3.6733 0.0732 0.0636 0.0652 0.3149 0.5407 0.6043 0.8905
Gr CACH1 C Up 1.0495 0.1018 0.0604 0.0541 0.1781 0.1453 0.1813 0.4770
Re CBTM4 C Down 0.6361 0.0197 0.0477 0.0245 0.0859 0.0620 0.0700 0.2417
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AIMS A Up 0.952 -1.819 -2.706 -1.488 -0.883 -0.701 -1.156 0.189
Cy AQH2 A Down 1.301 -2.615 -2.755 -2.730 -1.156 -0.615 -0.504 -0.116
1 4 7 10 13 16 13 22 Ur ^'ACMl C Up O.O48 -2.28b -2.806 -i. 91 C -1 . C2b -1.929 -l./'UK -0./40
Tims [months) Re CBTH4 C Down -0.452 -3.926 -3.043 -3.709 -2.455 -2.780 -2.660 -1.420
Figure 3-15. Data from Minideck 4A, Data from Minideck 4B, Data from Minideck 4C
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
CM
„ r-
LogahthmfDAs ug
-4 -3 -2 -1
? in
¥
Data from Mini Deck 5A
:^5>;
Bl — A IJp AUM2
Gr " CXlp CACM2 .
Re = C.Dn CCEM1
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cuT2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AUM2 A Up 0.7633 0.0238 0.0130 0.0185 0.0700 0.0401 0.0413 0.0922
Cy ALM2 A Down 0.9541 0.2131 0.2004 0.2258 0.6361 0.3212 0.4452 0.9223
Gr CACM2 C Up 1.1926 0.2322 0.1495 0.1781 0.6679 0.2598 0.2799 0.7315
Re CCEM1 C Down 1.8764 0.2481 0.2194 0.1686 0.7633 0.2808 0.3498 1.0177
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AUM2 A Up -0.270 -3.736 -4.340 -3.993 -2.660 -3.217 -3.186 -2.383
Cy ALM2 A Down -0.047 -1.546 -1.608 -1.488 -0.452 -1.136 -0.809 -0.081
Gr CACM2 C Up 0.176 -1.460 -1.901 -1.725 -0.404 -1.348 -1.273 -0.313
Re CCEM1 C Down 0.629 -1.394 -1.517 -1.780 -0.270 -1.270 -1.050 0.018
Data from Mini Deck 5B
1 4 7 1D 13 IB 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AADM1 A Up 3.1485 0.2194 0.1686 0.2449 0.8269 0.7060 0.5407 1.1131
Cy AZH3 A Down 2.8782 0.2322 0.1972 0.2067 0.7633 1.0591 0.4452 1.0813
Gr CBMH3 C Up 0.9064 0.2576 0.1208 0.1876 0.4770 0.2363 0.1495 0.2417
Re CBOH1 C Down 0.9223 0.0922 0.1113 0.1272 0.2385 0.0671 0.0891 0.4452
Log Transformed Data:
Col Brd Src Gra PRE
1
Months
7 11 15
20
24
Bl AADH1 A Up 1.147 -1.517 -1.780 -1.407 -0.190 -0.348 -0.615 0.107
Cy AZH3 A Down 1.057 -1.460 -1.624 -1.576 -0.270 0.057 -0.809 0.078
Gr CBMM3 C Up -0.098 -1.356 -2.113 -1.673 -0.740 -1.443 -1.901 -1.420
Re CBOM1 C Down -0.081 -2.383 -2.195 -2,062 -1.433 -2.701 -2.419 -0.809
$ T
a
I "
! ?
01
s r
Data from Mini Deck 5C
Re = C.Dn CADM3
1 4 7 10 13 16 13 22
Time Cmonths]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AARM3 A Up 0.9223 0.0700 0.0382 0.0891 0.2608 0.0522 0.1304 0.1177
Cy ABGM3 A Down 2.0195 0.1845 0.1368 0.1527 0.3816 0.3880 0.3816 0.5089
Gr CCAM1 C Up 0.6997 0.1940 0.0954 0.0763 0.4134 0.5788 0.3498 0.9223
Re CADM3 C Down 1.0972 0.1463 0.0668 0.0572 0.3180 0.1781 0.2226 0.2926
Log Transformed Data:
Col Brd Src Gra PRE
Months
11 15
20
24
Bl AARM3 A Up -0.081 -2.660 -3.266 -2.419 -1.344 -2.953 -2.037 -2.140
Cy ABGM3 A Down 0.703 -1.690 -1.990 -1.880 -0.963 -0.947 -0.963 -0.676
Gr CCAM1 C Up -0.357 -1.640 -2.350 -2.573 -0.883 -0.547 -1.050 -0.081
Re CADM3 C Down 0.093 -1.922 -2.706 -2.860 -1.146 -1.725 -1.502 -1.229
Figure 3-16. Data from Minideck 5A, Data from Minideck 5B, Data from Minideck 5C
67
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck; 6A
gr = C.Up CB2M2
Re = C,Dn CAAM3
7 10 13 16
Time [months]
Raw Data (ug/cm~2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AUM1 A Up 0.8746 0.0350 0.0150 0.0286 0.1081 0.0732 0.1050 0.2512
Cy ABYH1 A Down 1.3835 0.0827 0.0318 0.0827 0.3053 0.1294 0.1749 0.5089
Gr CBZM2 C Up 1.2244 0.0668 0.0293 0.0541 0.1050 0.1069 0.1813 0.3498
Re CAAM3 C Down 0.6361 0.0350 0.0121 0.0207 0.0477 0.0452 0.0445 0.3180
Log Transformed Data:
Col Brd Src Gra PRE
Months
11
15
20
24
Bl AUH1 A Up -0.134 -3.353 -4.203 -3.554 -2.224 -2.615 -2.254 -1.381
Cy ABYM1 A Down 0.325 -2.493 -3.448 -2.493 -1.186 -2.045 -1.743 -0.676
Gr CBZM2 C Up 0.202 -2.706 -3.532 -2.917 -2.254 -2.236 -1.708 -1.050
Re CAAM3 C Down -0.452 -3.353 -4.415 -3.879 -3.043 -3.098 -3.112 -1.146
Data from Mini Deck 6B
1 4 7 10 13 18 19 22
Time [months]
Raw Data Cug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AACH2 A Up 1.6697 0.0350 0.0299 0.0382 0.1717 0.1787 0.1781 0.5089
Cy AANM3 A Down 3.8800 0.4452 0.2831 0.7315 0.2354 0.8015 1.3039 1.5902
Gr CAJH1 C Up 1.6856 0.0604 0.0350 0.0445 0.1018 0.1078 0.1590 0.3053
Re CAIM1 C Down 1.3517 0.0477 0.0242 0.0572 0.0922 0.0684 0.1081 0.1622
Log Transformed Data:
Col Brd Src Gra PRE 1
Months
3 7 11 15 20 24
Bl AACH2 A Up 0.513 -3.353 -3.510 -3.266 -1.762 -1.722 -1.725 -0.676
Cy AANH3 A Down 1.356 -0.809 -1.262 -0.313 -1.447 -0.221 0.265 0.464
Gr CAJM1 C Up 0.522 -2.806 -3.353 -3.112 -2.285 -2.227 -1.839 -1.186
Re CAIH1 C Down 0.301 -3.043 -3.723 -2.860 -2.383 -2.683 -2.224 -1.819
IT
01
5 T
Data from Mini Deck 6C
Re = C.Dn CCCM2
7 10 13 16 19 22
Time [months]
Raw Data Cug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11
15
20
24
Bl ABCH3 A Up 1.7651 0.0891 0.0280 0.0572 0.1876 0.0776 0.2067 0.2672
Cy APM2 A Down 2.9100 0.1113 0.0382 0.0922 0.1399 0.2042 0.2576 0.6997
Gr CSH3 C Up 1.9241 0.1272 0.0509 0.0986 0.3498 0.3114 0.3498 1.1131
Re CCCM2 C Down 0.8428 0.0859 0.0289 0.0350 0.1050 0.1279 0.0859 0.1717
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl ABCH3 A Up 0.568 -2.419 -3.576 -2.860 -1.673 -2.556 -1.576 -1.320
Cy APH2 A Down 1.068 -2.195 -3.266 -2.383 -1.967 -1.589 -1.356 -0.357
Gr CSH3 C Up 0.654 -2.062 -2.978 -2.317 -1.050 -1.167 -1.050 0.107
Re CCCH2 C Down -0.171 -2.455 -3.543 -3.353 -2.254 -2.057 -2.455 -1.762
Figure 3-17. Data from Minideck 6A, Data from Minideck 6B, Data from Minideck 6C
68
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
CJ
KT "~
Logarithm [Dfl s ug/c
-4 -3 -2 -1 0
1 V
r*
CM
£ ""
Logarithm[DAs uf
-4 -3 -2 -1
I
f*.
QD
CM
,_
•£? D
LogarrthmCDAs ug
-4 -3 -2 -1
1:
PH.
t»
Data from Mini Deck 7A
^^
S$
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck SA
Bl = A,l.r
Cy - A,tm A
Or = C UP C
Re = C,Dn CAEM3
7 10 13 16 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AARM2 A Up 2.2422 0.0178 0.0099 0.0121 0.0188 0.0572 0.1050 0.3180
Cy ABYM3 A Down 2.0831 0.1081 0.0318 0.0248 0.0859 0.2481 0.2258 0.2735
Gr CBEM1 C Up 1.7810 0.0604 0.0413 0.0318 0.0541 0.3880 0.3498 0.8269
Re CAEH3 C Down 1.9400 0.0382 0.0156 0.0213 0.0572 0.0649 0.1081 0.2194
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AARH2 A Up 0.807 -4.028 -4.619 -4.415 -3.976 -2.860 -2.254 -1.146
Cy ABYM3 A Down 0.734 -2.224 -3.448 -3.697 -2.455 -1.394 -1.488 -1.296
Gr CBEM1 C Up 0.577 -2.806 -3.186 -3.448 -2.917 -0.947 -1.050 -0.190
Re CAEM3 C Down 0.663 -3.266 -4.162 -3.849 -2.860 -2.735 -2.224 -1.517
Data from Mini Deck BB
Bl = A,Up Aim
Re = C.Dn CAMM1
1 4 7 10 13 16 19 22
Tims [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11
15
20
24
Bl AIM1 A Up 1.6220 0.0827 0.0143 0.0305 0.1272 0.3403 0.5089 1.6220
Cy AATH1 A Down 1.9877 0.0732 0.0108 0.0210 0.0413 0.1279 0.2640 0.5407
Gr CACH3 C Up 1.0177 0.0382 0.0105 0.0130 0.0413 0.0766 0.1399 0.2767
Re CAMH1 C Down 1.5266 0.0350 0.0153 0.0178 0.0382 0.0840 0.1368 0.3816
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AIH1 A Up 0.484 -2.493 -4.247 -3.489 -2.062 -1.078 -0.676 0.484
Cy AATH1 A Down 0.687 -2.615 -4.527 -3.864 -3.186 -2.057 -1.332 -0.615
Gr CACM3 C Up 0.018 -3.266 -4.556 -4.340 -3.186 -2.569 -1.967 -1.285
Re CAMH1 C Down 0.423 -3.353 -4.182 -4.028 -3.266 -2.477 -1.990 -0.963
Data from Mini Deck 8C
E f
Gr = C UD CCAM2
Re = C.Dn CBXM2
1 4 7 1D 13 16 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AAGM4 A Up 1.6697 0.0318 0.0086 0.0185 0.0302 0.0617 0.1240 0.3021
Cy AZM2 A Down 2.6874 0.0763 0.0140 0.0318 0.1113 0.1565 0.3816 0.7633
Gr CCAH2 C Up 0.9541 0.0541 0.0140 0.0261 0.0795 0.1584 0.1813 0.5407
Re CBXM2 C Down 1.3039 0.0242 0.0095 0.0219 0.0445 0.0630 0.0922 0.2163
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AAGM4 A Up 0.513 -3.448 -4.757 -3.993 -3.500 -2.785 -2.087 -1.197
Cy AZM2 A Down 0.989 -2.573 -4.269 -3.448 -2.195 -1.855 -0.963 -0.270
Gr CCAM2 C Up -0.047 -2.917 -4.269 -3.647 -2.532 -1.843 -1.708 -0.615
Re CBXM2 C Down 0.265 -3.723 -4.652 -3.819 -3.112 -2.765 -2.383 -1.531
Figure 3-19. Data from Minideck 8A, Data from Minideck SB, Data from Minideck 8C
70
-------
Lo
-4
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 9A
Raw Data (ug/cm~2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl ATM3 A Up 2.9736 0.0172 0.0188 0.0477 0.0541 0.1266 0.1399 0.2449
Cy APM3 A Down 1.4948 0.0350 0.0188 0.0318 0.1081 0.1555 0.1622 0.3498
Gr CAPH1 C Up 0.7474 0.0025 0.0030 0.0115 0.0318 0.0455 0.0509 0.1145
Re CBWM1 C Down 0.4770 0.0038 0.0076 0.0073 0.0089 0.0278 0.0153 0.0445
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
7 1D 13 16 19 22
Time [manthsj
Bl ATMS A Up 1.090 -4.065 -3.976 -3.043 -2.917 -2.067 -1.967 -1.407
Cy APM3 A Down 0.402 -3.353 -3.976 -3.448 -2.224 -1.861 -1.819 -1.050
Gr CAPH1 C Up -0.291 -5.987 -5.823 -4.470 -3.448 -3.090 -2.978 -2.167
Re CBWM1 C Down -0.740 -5.568 -4.876 -4.919 -4.721 -3.583 -4.182 -3.112
Data from Mini Deck 9B
•
*-• Re = C.Dn CANM2
1 4 7 10 13 18 19 22
Time [months]
Raw
Col
Bl
Cy
Gr
Re
Log
Col
Bl
Cy
Gr
Re
Data (ug/cm~2 DAs) :
Brd Src Gra PRE
AACH1 A Up
AAEH2 A Down
CBIM2 C Up
CANM2 C Down
Transformed Data
Brd Src Gra
AACH1
AAEH2
CBIH2
CAHM2
A Up
A Down
C Up
C Down
1 . 5902
1.9559
0.8269
1.7015
PRE
0.464
0.671
-0.190
0.532
Months
1
0.0134
0.0038
0.0032
0 . 0009
1
-4.315
-5 . 568
-5.751
-7.059
3
0.0156
0.0099
0 . 0048
0.0007
3
-4. 162
-4.619
-5.345
-7.264
7
0.0289
0.0146
0.0124
0.0124
7
-3.543
-4.225
-4 . 390
-4.390
11
0 . 0509
0.0232
0 . 0226
0.0178
Months
11
-2.978
-3.763
-3.791
-4.028
15
0.0461
0.0201
0.0274
0 . 0206
15
-3.077
-3.907
-3.597
-3.883
20
0.1272
0.0700
0.0382
0.0350
20
-2.062
-2.660
-3.266
-3.353
24
0 . 2004
0 . 1081
0.0827
0.0795
24
-1.608
-2.224
-2.493
-2.532
CM
.._
*£
^ °
I Logai*fthm[DAa ug
-4 -3 -2 -1
:
i
r^
I
CO
Data from Mini Deck 9C
* -
^^
Jf Bl = A Up AAGM2
•^ Cy = A,Dn A;
Gr = C Up CBZM1
Re = C.Dn CAEM2
1 4 7 1D 13 16 19 22
Time [montha]
Raw Data (ug/cm'2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Cy AAMM1 A Down 4.7705 0.0115 0.0130 0.0445 0.0242 0.1234 0.3021 0.9223
Gr CBZM1 C Up 0.8110 0.0015 0.0044 0.0076 0.0111 0.0252 0.0273 0.0668
Re CAEM2 C Down 2.3853 0.0019 0.0031 0.0089 0.0102 0.0156 0.0312 0.0350
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAGM2 A Up 0.522 -4.876 -5.655 -4.182 -3.976 -3.599 -3.458 -2.978
Cy AAHM1 A Down 1.562 -4.470 -4.340 -3.112 -3.723 -2.092 -1.197 -0.081
Gr CBZM1 C Up -0.209 -6.482 -5.415 -4.876 -4.498 -3.679 -3.599 -2.706
Re CAEM2 C Down 0.869 -6.261 -5.770 -4.721 -4.587 -4.157 -3.468 -3.353
Figure 3-20. Data from Minideck 9A, Data from Minideck 9B, Data from Minideck 9C
71
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
CM
^ «-
E
)garithm[DAs u
-3 -2 -1
-J i
i ^
p>.
03
CM
^ T-
1°
LogarithmCDAs u$
-4 -3 -2 -1
| ?
r-.
(D
C\l
E
ft0
LogarrthmtOAs j
-4 -3 -2 -1
E in
1 '
Z [D
r-.
Data from Mini Deck 10A
^*
^"* — "• -s^"^
/ ~J&\ — A Up AADM''
i*^-- -
tr
Re = C.Dn CADM1
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 10B
x^
v£^
y
f Bl = A.Up AXM1
•T Cy = A,Dn AYM1
. /
R« = C.Dn CAKM4
1 4 7 10 13 16 19 22
Tims [months]
Data from Mini Deck 1QC
/'
J^f
Jf Bl = A,Up AAJM2
/ Cy = A.Dn AQM3
/ Qp = £• itf. CRUM2
Re = C.Dn CBTM2
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cm~2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AADM3 A Up 3.8323 0.0312 0.0185 0.0153 0.0700 0.0350 0.0283 0.0668
Cy ABGM2 A Down 2.0831 0.0051 0.0089 0.0111 0.0226 0.0531 0.0509 0.0636
Gr CAPM3 C Up 0.6838 0.0019 0.0060 0.0041 0.0248 0.0250 0.0219 0.0827
Re CADM1 C Down 0.7633 0.0011 0.0017 0.0027 0.0083 0.0127 0.0108 0.0238
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AADH3 A Up 1.343 -3.468 -3.993 -4.182 -2.660 -3.353 -3.565 -2.706
Cy ABGM2 A Down 0.734 -5.280 -4.721 -4.498 -3.791 -2.935 -2.978 -2.755
Gr CAPM3 C Up -0.380 -6.276 -5.109 -5.489 -3.697 -3.689 -3.819 -2.493
Re CADM1 C Down -0.270 -6.777 -6.383 -5.926 -4.795 -4.365 -4.527 -3.736
Raw Data (ug/cm"2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Cy AYM1 A Down 3.4507 0.0235 0.0191 0.0763 0.0763 0.0770 0.1208 0.1081
Gr CBJM1 C Up 3.3553 0.0020 0.0024 0.0115 0.0165 0.0372 0.0859 0.0700
Re CAKH4 C Down 0.7315 0.0010 0.0031 0.0137 0.0185 0.0196 0.0280 0.0382
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AXM1 A Up 0.703 -3.926 -4.835 -2.755 -2.860 -3.098 -1.474 -2.013
Cy AYM1 A Down 1.239 -3.749 -3.959 -2.573 -2.573 -2.564 -2.113 -2.224
Gr CBJM1 C Up 1.211 -6.195 -6.036 -4.470 -4.102 -3.291 -2.455 -2.660
Re CAKM4 C Down -0.313 -6.918 -5.793 -4.292 -3.993 -3.930 -3.576 -3.266
Raw Data (ug/cm~2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAJH2 A Up 2.2740 0.0219 0.0162 0.0296 0.0509 0.0817 0.0309 0.1463
Re CBTM2 C Down 0.6520 0.0009 0.0032 0.0067 0.0350 0.0350 0.0127 0.0277
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AAJH2 A Up 0.822 -3.819 -4.122 -3.521 -2.978 -2.504 -3.479 -1.922
Cy AQM3 A Down 1.106 -6.092 -6.036 -3.266 -3,043 -2.660 -3.565 -2.573
Re CBTM2 C Down -0.428 -7.024 -5.751 -5.009 -3.353 -3.353 -4.365 -3.587
Figure 3-21. Data from Minideck 10A, Data from Minideck 10B, Data from Minideck 10C
72
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Data from Mini Deck 11A
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11
15
20
24
Bl AUM3 A Up 1.6379 0.0064 0.0267 0.0445 0.1113 0.1053 0.2067 0.2290
Cy AQM1 A Down 2.7192 0.0092 0.0115 0.0541 0.0763 0.1466 0.3498 0.4452
Gr CAPM2 C Up 0.8587 0.0020 0.0054 0.0079 0.0289 0.0372 0.0954 0.2131
Re CAIM3 C Down 1.1290 0.0016 0.0076 0.0102 0.0509 0.0992 0.1622 0.4770
Log Transformed Data:
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AUM3 A Up 0.493 -5.058 -3.622 -3.112 -2.195 -2.251 -1.576 -1.474
Cy AQM1 A Down 1.000 -4.686 -4.470 -2.917 -2.573 -1.920 -1.050 -0.809
Gr CAPM2 C Up -0.152 -6.230 -5.220 -4.835 -3.543 -3.291 -2.350 -1.546
Re CAIM3 C Down 0.121 -6.425 -4.876 -4.587 -2.978 -2.310 -1.819 -0.740
Data from Mini Deck 11B
V
Bl = A.Up AXM2
Cy = A.Dn AAHH2
Gr = C.Up CBEM3
Re = C.Dn CBWM2
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 11C
Gr = C.Up CBJM2
Re = C.Dn CAEM1
1 4 7 10 13 16 19 22
Time [months]
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
1
Months
7 11
15
20
24
Bl AXM2 A Up 1.8287 0.0121 0.0064 0.0445 0.0763 0,1508 0.2894 0.5725
Cy AAHM2 A Down 0.6838 0.0008 0.0051 0.0541 0.0445 0.0703 0.1845 0.2989
Gr CBEM3 C Up 1.8128 0.0026 0.0060 0.0223 0.0541 0.2477 0.3180 0.2481
Re CBUM2 C Down 0.7315 0.0011 0.0019 0.0169 0.0350 0.1018 0.1399 0.3816
Log Transformed Data:
Col Brd Src Gra PRE
1
Months
7 11 15
20
24
Bl AXH2 A Up 0.604 -4.415 -5.058 -3.112 -2.573 -1.892 -1.240 -0.558
Cy AAHM2 A Down -0.380 -7.094 -5.280 -2.917 -3.112 -2.655 -1.690 -1.207
Gr CBEM3 C Up 0.595 -5.960 -5.109 -3.805 -2.917 -1.395 -1.146 -1.394
Re CBWH2 C Down -0.313 -6.831 -6.261 -4.083 -3.353 -2.285 -1.967 -0.963
Raw Data (ug/cm"2 DAs):
Col Brd Src Gra PRE
Months
7 11 15
20
24
Bl AAJM4 A Up 2.2104 0.0095 0.0089 0.0509 0.0732 0.2821 0.2194 2.1626
Cy ABWM3 A Down 1.1926 0.0153 0.0172 0.0795 0.0572 0.1708 0.1368 0.1940
Gr CBJH3 C Up 2.1467 0.0008 0.0035 0.0121 0.0137 0.1666 0.2194 0.6361
Re CAEM1 C Down 2.1945 0.0016 0.0025 0.0121 0.0509 0.1409 0.1527 0.3498
Log Transformed Data:
Col Brd Src Gra PRE
1
Months
7 11 15
20
24
Bl AAJM4 A Up 0.793 -4.652 -4.721 -2.978 -2.615 -1.265 -1.517 0.771
Cy ABWM3 A Down 0.176 -4.182 -4.065 -2.532 -2.860 -1.767 -1.990 -1.640
Gr CBJM3 C Up 0.764 -7.182 -5.655 -4.415 -4.292 -1.792 -1.517 -0.452
Re CAEM1 C Down 0.786 -6.425 -5.987 -4.415 -2.978 -1.960 -1.880 -1.050
Figure 3-22. Data from Minideck 11 A, Data from Minideck 11B, Data from Minideck 11C
73
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
cu
KT
Logarithm [DAs ug/
-4 -3 -2 -1 C
? n
^
CM
ET
E
r4°
LogarithmtDAs u
-4 -3 -2 -1
1 V
I
7
CXI
?"
f» °
l_ogar*rthm(DAs u
-4 -3 -2 -1
CD
t_ in
i
1^1
1
ID
Data from Mini Deck 12A
.
/
^^
el In AOM 1
v _ A.Dn AANM2
Gr = C.Up CAJM2 .
Re = C.Dn CAMM4
1 4 7 10 13 16 19 22
Time [months]
Data from Mini Deck 12B
«S*&\
Bl - A Up AACM3
Cy = A'.Dn AAEM3
R« = C.Dn CADM4
1 4 7 10 13 16 13 22
Time [months]
Data from Mini Deck 12C
^j^
Rl A 1 lr» AWUO
Bl A.Up AVM2
Re = C.Dn CBTM1
1 4 7 1D 13 16 19 22
Time [months]
Raw Data (ug/cm'2 DAs) : Months
Col Brd Src Gra PRE i 3 7 11 15 20 24
Cy AANM2 A Down 6.3607 0.3816 0.3498 0.6997 1.5584 1.4216 1.0813 3.8164
Gr CAJM2 C Up 1.7810 0.0572 0.0413 0.0763 0.2703 0.2048 0.2544 0.7633
Re CAMM4 C Down 1.0177 0.0922 0.0231 0.0891 0.0509 0.2900 0.3117 1.0177
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AOM1 A Up 0.928 -2.419 -2.978 -1.990 -0.558 -0.740 -1.050 -0.014
Cy AANM2 A Down 1.850 -0.963 -1.050 -0.357 0.444 0.352 0.078 1.339
Gr CAJM2 C Up 0.577 -2.860 -3.186 -2.573 -1.308 -1.586 -1.369 -0.270
Re CAMM4 C Down 0.018 -2.383 -3.770 -2.419 -2.978 -1.238 -1.166 0,018
Raw Data (ug/cm~2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AACM3 A Up 2.5602 0.0477 0.0445 0.0986 0.2226 0.1743 0.1749 0.2989
Cy AAEM3 A Down 1.4471 0.0350 0.0207 0.0954 0.1781 0.1279 0.1781 0.2099
Gr CBIM3 C Up 1.1608 0.0541 0.0350 0.0509 0.1654 0.2341 0.0986 0.3498
Re CADM4 C Down 0.6520 0.0382 0.0350 0.0413 0.1749 0.0773 0.0413 0.1749
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AACH3 A Up 0.940 -3.043 -3.112 -2.317 -1.502 -1.747 -1.743 -1.207
Cy AAEM3 A Down 0.370 -3.353 -3.879 -2.350 -1.725 -2.057 -1.725 -1.561
Gr CBIM3 C Up 0.149 -2.917 -3.353 -2.978 -1.800 -1.452 -2.317 -1.050
Re CADM4 C Down -0.428 -3.266 -3.353 -3.186 -1.743 -2.560 -3.186 -1.743
Raw Data (ug/cm~2 DAs) : Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AVM2 A Up 1.5584 0.0413 0.0382 0.0572 0.1208 0.1104 0.3498 0.4452
Cy ALM1 A Down 0.9541 0.0413 0.0445 0.0891 0.4134 0.2509 0.2481 0.5089
Gr CBMM1 C Up 0.7315 0.0318 0.0273 0.0232 0.2449 0.0811 0.1781 0.2481
Re CBTH1 C Down 0.6361 0.0172 0.0111 0.0143 0.0541 0.0604 0.0668 0.1686
Log Transformed Data: Months
Col Brd Src Gra PRE 1 3 7 11 15 20 24
Bl AVM2 A Up 0.444 -3.186 -3.266 -2.860 -2.113 -2.204 -1.050 -0.809
Cy ALM1 A Down -0.047 -3.186 -3.112 -2.419 -0.883 -1.383 -1.394 -0.676
Re CBTM1 C Down -0.452 -4.065 -4.498 -4.247 -2.917 -2.806 -2.706 -1.780
Figure 3-23. Data from Minideck 12A, Data from Minideck 12B, Data from Minideck 12C
74
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
There is one minor difference between the minideck plots in Figure 3-25 and the plot
depicted in Figure 3-11 for Minideck 9-A,. In Figure 3-11, the baseline values were
plotted as straight horizontal lines. With the decreased resolution inherent in Figure 3-
25, many of the straight horizontal baseline lines would overlay and obscure one
another. So in the plots in Figure 3-25, the baseline measurements horizontal lines
have been replaced by lines that have been intentionally "wiggled" or "jittered" to avoid
complete over plotting. This is an established graphical technique for increasing the
resolution of crowded graphs.
Each separate minideck plot can be examined for the effects of source deck, grain
orientation and time as was done for Minideck 9-A. However, this is not the intended
purpose of the composite plot (rather, the larger data plots in Figures 3-12 through
3-24 should be used to examine the data for specific minidecks). Here the primary
interest is in comparing patterns in DAs measurements among the thirteen coatings,
and, secondarily, in visually assessing repeatability of the experimental method by
easily comparing results among the three replicate minidecks treated with the same
coating.
Repeatability is indicated by the similarity in patterns and trends within coatings. For
example, it is evident that Coating 13 (uncoated) exhibited the least trend with time
(although Coating 5 is a close second), and that this relative lesser trend is manifested
in all three minidecks. This is even more noticeable upon consideration of other
coatings that exhibit more pronounced trends with time; for example, Coatings 9,10,
and 11 and to a lesser extent, Coatings 3, 8, and 12. It is evident that certain coatings
yielded time-series DAs measurements with different patterns than did Coating 13 (the
uncoated control), and furthermore that these patterns were generally consistent
among minidecks treated with the same coating.
As stated earlier, the DCCA observed at any point in time is a function of the time
elapsed since the previous wipe event and weathering, among other potential factors.
The two paints (products 9 and 10) and the vinyl elastic product (product 11) had the
lowest initial postcoat (1 month) values and, generally, DCCA showed a continuously
increasing trend with time for these products. Most of the other plots in Figure 3-27
exhibit an upward slanting down/up/down/up pattern, suggesting that, for some periods
between samples, other factors which lower the DCCA measurement had a greater
effect than the time/weathering effect on DCCA, which is assumed to always yield an
upward trend, or greater DCCA over time. These other factors may include a surface
cleaning effect caused by the wipe sampling. Additionally, it is possible that climatic
conditions, surface moisture, or other variables which vary more inconsistently over
time may affect sampling and DCCA measurements.
76
-------
Coating
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
1 2 3 4 5 6 7 8 9 10 11 12 13
o
CD
a
f+s
m
^
7
o
\f*
.y<^
v^
r
f
W
Figure 3-25. Composite Data Plot of All 39 Minidecks (note that scale and axis labels are the same as for Figure 3-11)
77
-------
Coating
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
1
8
10 11 12 13
i
o
0>
CO
•h
QL
O
•h
O
v
Grand mean (control deck)
90% less than grand mean
MD Minideck
GR Grain
SD Source deck
Figure 3-26. Averages from All Thirty-nine Minidecks (See Section 3.2.2.4 for further explanation)
78
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Before proceeding to the next graphical display, one more observation should be made
that bears on the validity of the experimental method, and the role and utility of the
baseline measurements in particular. A critical scan of all thirty-nine minideck plots
reveals two minidecks where the DAs measurements from a single board are
noticeably lower, for all time periods, than the DAs measurements from the other
boards on the same minideck: minidecks 5-A and 13-C (uncoated). For minideck 5-A,
the board with noticeably low post baseline DAs values also had the lowest baseline
DAs measurement of all boards, even though it was from Source Deck A, which
generally yielded higher-than-average DAs measurements. In other words, this is a
board for which the baseline measurement was noticeably low and also resulted in
noticeably-low post baseline values. For Minideck 13-C (uncoated), the board with
noticeably-low DAs measurements came from Source Deck C and had the lowest
baseline DAs measurements of all boards on that minideck.
3.2.2.4 Averages from All Thirty-nine Minidecks
Figure 3-26 displays a series of plots of average DAs, which are constructed to isolate
and illustrate the effect of different factors in the experiment. The variables of interest
are: source deck, grain orientation and time. In order to reveal the effects of these
factors, Figure 3-26 presents a series of averages over coating or minideck (MD),
source deck (SD), grain orientation (GR), and combinations thereof.
3.2.2.4.1 Averages over Minidecks
The first row of plots in Figure 3-26 displays coating-specific averages over the three
replicate minidecks. The color coding is the same as the plots of the data in the
previous data displays, as indicated in Table 3-3. Statistically, these plots are the least
variable manifestations of the patterns and trends in DAs measurements as they relate
to source deck, grain orientation and time period (least variable because they are
averaged over all replicate minidecks).
In this first row of plots, two horizontal lines have been added to facilitate comparison
with Coating 13, the uncoated control. The upper, solid black line is plotted at the grand
mean of all post baseline Coating 13 (uncoated) DAs measurements. Recall that
Coating 13 (uncoated) exhibited little trend overtime and thus the grand mean is an
appropriate indicator of overall average Coating 13 (uncoated) DAs value. The upper,
black line makes it easy to see how much less the other coating measurements are
than the Coating 13 (uncoated) grand mean, although keep in mind that these
differences are on a logarithmic scale. The lower, dashed black line is plotted at a
value equal to ln(10) less than the upper black line. In order to understand its
relevance, let DAs13 denote the Coating 13 (uncoated) grand mean. Then, DAs13/10 is
the value equal to 90 percent less than the Coating 13 (uncoated) grand mean, and on
the log scale this is
79
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
ln(DAs-i3/10) = ln(DAs13) -
Thus the lower black line appears at a level corresponding to a 90 percent reduction
from the Coating 1 3 (uncoated) grand mean.
3.2.2.4.2 Averages over Minidecks and Grain
The second row of plots in Figure 3-26 displays coating-specific DAs values averaged
over minideck (MD) and grain orientation (GR). The purpose is to illustrate more clearly
the effects of source deck. The color coding for this row of plots necessarily differs from
the previous plots. In this case, cyan (Cy) denotes averages from Source Deck A, and
red (Re) denotes averages from Source Deck C. In effect, the cyan curves are the
averages of the cyan and dark blue curves from the plots in row one; and the red
curves are the averages of the red and green curves from the plots in row one. The
feature to note is that when the averages differ noticeably, it is generally the case that
the red curve lies below the cyan curve, indicating that DAs measurements from
Source Deck A boards were generally greater than those from Source Deck C boards.
3.2.2.4.3 Averages over Minidecks and Source Deck
The third row of plots in Figure 3-26 displays coating-specific DAs values averaged
over minideck (MD) and source deck (SD). Its purpose is to illustrate more clearly the
effects of grain orientation. The color code for this is row of plots necessarily differs
from the previous plots. In this case green (Gr) denotes averages for Grain Up boards,
whereas red (Re) denotes averages from Grain Down boards. In effect, the green
curves are the averages of the Green and Dark Blue curves from the plots in row one;
and the red curves are the averages of the Red and Cyan curves from the plots in row
one. The primary feature to note is that the separation between the red and green
curves in the third row is generally less than that between the different colored curves
in the second row, meaning that the observed effects of grain orientation on measured
DAs is noticeably less than that of source deck.
3.2.2.4.4 Averages over Minidecks, Grain, and Source Deck
The fourth and bottom row of Figure 3-26 displays averages of measured DAs
averaged over minideck (MD), grain orientation (GR) and source deck (SD). The
purpose is to illustrate the effect of time for the thirteen different coatings averaged
over the other factors. Each coating is now plotted with its own unique color. The black
lines surrounding the time plots are pointwise error bands and are included to give a
visual assessment of the variability in the averages. Each colored curve is the average
of the four curves in the top row of Figure 3-26. At each time point, the error-bands
were calculated as twice the standard error of the average of the three numbers
80
-------
obtained by first averaging In DAs values over the four boards on each of the three
replicate minidecks.
3.2.2.5 The Comprehensive Plot
Figure 3-27, is a composite of all the previous plots presented in Figures 3-25 and 3-
26. Its primary advantage is that all of the graphical information presented thus far is
compactly displayed on a single page. This necessarily entails some loss of resolution,
but has the advantage that the larger, and thus practically more significant, relevant
patterns and trends are still apparent, and can be seen entirely in one graphic. The
subplots are displayed in a seven-by-thirteen display with columns corresponding to
coatings. The first three rows are the same as in Figure 3-25; and the bottom four rows
are the same as in Figure 3-26.
3.2.3 Analysis of Variance of Coatings by Time
The top row of plots in Figure 3-28 is similar to the seventh row in Figure 3-27, but
there are important differences. As in the former figure, average In(DAs) values
(averaged over minidecks and board types) are plotted versus time. However, in these
new plots the thickness of the plotted line corresponds approximately to estimation
variability. The widths of the lines are approximately two standard errors of the
difference between two means. The standard error used to determine line thickness
was determined from the analysis of variance modeling. It is approximate because the
analysis of variance standard errors differ from one mean to the next due to imbalance
(two missing values) and the covariate adjustment. However, the approximation is
sufficient for visual comparisons.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
81
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
82
-------
MD1 MD2 MD3 MD 4 MD 5 MD 6 MD 7 MD 8 MD 9 MD 10 MD11 MD 12 MD 13
Minideck A
Minideck B
Minideck C
MD
MD, GR
MD, SD
MD, GR, SD
^
-r
J
Figure 3-27. Comprehensive Data Plot (note that scale and axis labels are the same as for Figure 3-11)
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
83
-------
10 11
13
LO
CN
O
LO
C\J
LO
h-
c\i
uO
CN
LO
h-
LO
0.075
0.625
.175
.725
Figure 3-28. Analysis of Variance Plot of Coatings by Time [each subplot has In(DAs) as its y-axis and time, in years, as its x-axis]
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
84
-------
The widths of the lines plotted in Figure 3-28 were determined to provide a rough visual
assessment of statistical significance. The widths are approximately equal to twice the
standard error of the difference between two means. It follows that when two of these
curves are overlaid, points where the lines touch are not significantly different; whereas
points where they do not touch are significantly different (at the 0.05 level of
significance with no adjustment for multiple tests).
The black horizontal lines are the same as those in the fourth row of plots in Figure 3-
27. The top row of plots serves as a legend for the large graph displaying the overlaid
coatings-by-time plots. Except for the overplotting, the larger plot makes it is easy to
compare means at various time points and to get a rough assessment whether they
are significantly different (not touching) or not (touching). In light of the approximations
involved and the lack of adjustment for multiple tests we caution against over
interpretation when inspecting this figure.
3.2.4 Analysis of Variance of Coating Pair Comparisons by Time
Figure 3-29 overcomes the overplotting deficiencies of the large plot in Figure 3-28 by
plotting all possible pairs of coating-by-time plots. There are 13 coatings and thus there
are "13 choose 2" or 78 possible pairwise comparisons among them. Figure 3-29
displays all 78 possible comparisons of the coatings-by-time plots. Running down the
diagonal (top left to bottom right plot) are the 13 individual time plots (1 to 13, in order).
In row j, the time plot of coating j is overlaid with the time plot of coating k (for k =
j+1,...,13). Color coding is established by the diagonal plots and is maintained in the
off-diagonal plots. Figure 3-29 makes visual comparisons easy and relevant.
The results presented above pertain only to DAs on the true time scale measured in
years from start date.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
85
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
86
-------
MD1 MD2 MD3 MD 4 MD 5 MD 6 MD 7 MD 8 MD 9 MD 10 MD11 MD 12 MD 13 VS.
NOTE: Running down the diagonal are the 13
individual time plots (1 to 13, in order). In row j, the
time plot of coating j is overlaid with the time plot of
coating k(for k = j+1,...,13). Color coding is
established by the diagonal plots and is maintained
in the off diagonal plots.
MD 1 (Sealant, Oil-based )
MD 2 (Sealant, Oil-based)
MD 3 (Stain, Oil-based)
MD 4 (Stain, Oil-based)
MD 5 (Sealant, Water-based)
MD 6 (Sealant, Water-based)
MD 7 (Stain, Water-based)
MD 8 (Stain, Water-based)
MD 9 (Paint, Water-based)
MD 10 (Paint, Oil-based)
MD 11 (Other, Water-based)
MD 12 (Other, Water-based)
MD13(Uncoated)
Figure 3-29. Analysis of Variance Plot of Coating Pair Comparisons by Time [each subplot has In(DAs) as its y-axis and time, in years, as its x-axis]
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
87
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
88
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
3.2.5 Analysis of Variance Modeling Results
We now present results from statistical analyses designed to investigate the ability of
study coatings to attenuate dislodgeable arsenic residues on the surfaces of CCA
treated wood compared to washing alone. The statistical models used to analyze the
data were chosen because of their simplicity and ability to complement the graphical
analyses in Figures 3-25 through 3-29, establish the robustness (lack of sensitivity) of
major conclusions to statistical modeling choices, and guide future experimentation by
identifying the major sources of variation and dependencies in the data.
The salient feature of Figures 3-25, 3-26, and 3-27 is the increasing time trend in DA
for several of the coatings. Given that the statistical significance of the effect of time is
not in doubt, we focus on separate analyses of DAs at each time period. These time
period-specific analyses avoid modeling and inference issues related to changes in the
magnitude of measured DAs overtime and temporal dependence among repeated
measurements.
The analysis variable (dependent variable) in the time period-specific analyses is the
natural logarithm of DAs adjusted by the corresponding baseline measurement:
Y= ln(DAs/baseline)
= In(DAs) - In(baseline).
An alternative to direct adjustment of the DAs measurements by baseline values is to
use the baseline values as covariates in the statistical models. In Section 3.2.5.5 we
show that the latter approach to baseline adjustment results in nearly identical
conclusions about coating performance.
3.2.5. 1 Time Period Specific Modeling
Fora given time period the design of the experiment is a that of a standard split-plot
and the basic analysis is a split-plot analysis with Coating as the whole plot treatment
and Source Deck and Board Bark Face as split plot factors. Table 3-5 summarizes
select results from time-period specific analyses using split-plot models with
homoscedastic error structures. Main table entries are p-values for the whole-plot
factor Coating (coatl) and split-plot factors Board Bark Face (bbf), Source Deck
(sdeck), and the their interactions. Note that p-values in Table 3-5 have not been
adjusted for multiple testing across time periods. Below the main table entries are the
whole-plot variance (mdeck(coatl)) and the residual error variance (residual) estimates
(with 95 percent lower and upper confidence limits). Column DF in Table 3-5 displays
the degrees of freedom for each effect tested.
89
-------
Table 3-5. P-Values for Split-Plot Models with Homoscedastic Error Structures
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Effect
coatl
Bbf
coatl *bbf
sdeck
coatl *sdeck
Bbfsdeck
coatl *bbf sdeck
DF
12
1
12
1
12
1
12
mdeck( coatl )
residual
lower
upper
Time (months after coating)
1
<0001
0.5169
0.5881
0.5354
<0001
0.5930
0.2818
0
0.4144
0.3214
0.5549
3
<0001
0.6251
0.6121
0.5653
0.0008
0.1757
0.6278
0.04947
0.4284
0.3199
0.6034
7
<0001
0.1571
0.1223
0.1548
<0001
0.0528
0.5397
0.03824
0.2512
0.1875
0.3540
11
<0001
0.5836
0.3363
0.0853
0.1579
0.3496
0.2241
0.01411
0.3226
0.2412
0.4537
15
<0001
0.5452
0.3802
0.0132
0.2557
0.0033
0.4108
0.05267
0.3189
0.2384
0.4484
20
<0001
0.8661
0.1000
0.4518
0.2203
0.0215
0.7013
0.05371
0.3046
0.2278
0.4283
24
<0001
0.4392
0.4684
0.0228
0.3149
0.0655
0.5332
0
0.3635
0.2819
0.4868
3.2.5.2 Fixed Effects
The most noteworthy feature of the ANOVA summaries in Table 3-5 is the persistently
strong statistical significance of Coating (coatl) across time periods, thus confirming
the visual assessment of differences among coatings in Figure 3-25. With regard to the
other factors, Board Bark Face (bbf) and its two- and three-way interactions are, with
the exception of its interaction with Source Deck (sdeck) in samples taken at 15, 20
and 24 months after coating, not statistically significant; whereas the statistical
evidence for Source Deck and its interactions is stronger.
The pattern of statistical significance suggests that much of the variability in DAs is
explained by coatl and sdeck; that is, bbf contributes least to the fit of the models.
Likelihood ratio tests comparing the full models in Tables 3-5 to corresponding models
without any bbf terms have p-values, 0.0399, 0.0673, 0.0779, 0.0544, 0.0223, 0.0381,
and 0.0619, for samples taken at 1, 3, 7, 11,15, 20, and 24 months, respectively.
Thus, the statistical evidence for bbf is marginal (all p-values are close to the 0.05 level
of significance). Examination of Akaike Information Criteria (AIC) for the two sets of
models reveals that for every time period, the model without bbf terms has the lower
AIC value. In other words, model selection based on AIC identifies the models without
bbf terms. However, these conclusions are based on the separate analyses and thus
do not account for the accumulation of evidence across time periods. A complementary
assessment of models with and without sdeck terms in them (but keeping bbf terms in
the model) was also conducted. In this case, the p-values for testing reduced models
(no sdeck terms) versus full models, are <0.0001, <0.0001, 0.0001, 0.0133, 0.0035,
90
-------
0.0717and 0.0130. Thus the statistical significance of sdeck is more firmly established
than for bbf, especially in the early time periods. The AIC values for these models are
smaller for the full models for the 1-, 3- and 7-month data, and smaller for the reduced
model in the subsequent four time periods.
3.2.5.3 Random Effects
Table 3-5 also displays variance components for each of the time period-specific
models. In all time periods, the whole-plot error variance (between-deck variance) is
small relative to the residual variance (within-deck variance). Although boards on the
same minideck are neighboring and thus might be expected to vary in unison in
response to shared external effects, the scale of the experiment is such that boards
were exposed to essentially identical environmental influences whether they were on
the same minideck or not. In this experiment, all boards on a minideck were prepared
and coated more or less simultaneously, and thus whole plot variation is likely largely
due to variation in the preparation and coating processes. That the whole plot error
variance is small suggests that the protocol for building, preparing, and coating decks
is well-replicated.
Finally, we note that the residual variance estimates in Table 3-5 do not differ in
magnitude overtime, nor do their standard errors which range from 0.04 to 0.07 (not
reported in Table 3-5).
3.2.5.4 Coating Comparisons
Our primary interest lies in comparing the performances of Coatings 1-12 to that of the
control Coating 13 (washing only). For each of the time-period specific analysis of
variance models, we calculated statistical tests comparing the main effects of Coatings
1-12 to Coating 13.
Table 3-6 summarizes the coating performance rankings and p-values of the tests,
adjusted using Dunnett's multiple comparison procedure within each time period. No
adjustments to p-values were made for the multiple time periods. In Table 3-6, for each
time period, the left hand column lists the coatings in order of performance relative to
the control coating, and the p-value associated with the test. For example, after 1
month: Coating 11 performed best followed by Coatings 10, 9, 3, etc.; also, with the
exception of Coating 5, all coatings performed significantly better than Coating 13 at
the 0.05 level of significance.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
91
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-6. Coating Performance Rankings and P-Values of Tests Adjusted using Dunnett's
Multiple Comparison Procedure
1 month
11
10
9
3
8
1
12
6
7
4
2
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0002
0.0885
3 months
10
9
11
3
8
1
6
12
2
4
7
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0002
0.0002
0.0003
0.0027
0.2471
7 months
10
9
8
11
3
1
2
6
7
4
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0002
0.0004
0.0973
11 months
9
10
8
11
3
1
6
4
2
7
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0010
0.0011
0.0019
0.7416
15 months
10
9
3
1
8
11
6
7
4
2
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0005
0.0006
0.0024
0.0036
0.0089
0.0213
0.1746
20 months
10
9
1
3
2
8
6
7
4
11
12
5
<0.0001
0.0001
0.0165
0.0200
0.0785
0.2435
0.3396
0.3678
0.3707
0.5729
0.7404
0.9994
24 months
10
9
2
1
3
8
7
6
11
4
12
5
<0.0001
<0.0001
0.0024
0.0133
0.0273
0.1203
0.1830
0.2001
0.2005
0.5434
0.7101
0.9635
The preponderance of highly significant p-values (< .0001) in Table 3-6 indicates that
several coatings attenuate dislodgeable arsenic residues compared to washing alone
for various lengths of time after application. With the exception of Coating 5, the effects
of the coatings (relative to initial washing only) remain statistically significant through 15
months and, for some coatings, throughout the length of the study.
A number of patterns in Table 3-6, suggest possible hypotheses for testing in future
studies. Coatings 9 and 10 exhibited the most consistently good relative performance,
ranking 1, 2 or 3 in all time periods and clearly outperforming washing alone throughout
the duration of the experiment. The relative performance of Coating 2 was low initially,
but improved in the last two time periods, whereas the relative performances of
Coatings 1, 3, 4, 6, 7, and 8 were relatively stable overtime. Finally, the relative
performances of Coatings 11 and 12 deteriorated with time, although they were
significantly better than washing through 15 months.
In summary there is strong evidence that coatings (with the exception of Coating 5)
have detectable positive effects relative to washing alone through 15 months after
coating, but the ability to detect differences after 15 months is much diminished, due
primarily to diminished performance (higher DAs) and not increased variation (in light of
the nearly constant experimental error variance estimates in Table 3-5). This does not
mean that positive effects of coatings were absent in the later time periods, but rather
that the effect magnitudes were not statistically significant.
92
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
3.2.5.5 Statistical Model Sensitivity Analysis
We now present comparisons like those in Table 3-6 obtained from select alternative
statistical models fit to the data. Our intent is to establish the lack of sensitivity of our
primary conclusions to variations in the statistical models used to analyze the data. We
consider only comparisons at 24 months after coating, because variations in the
coatings comparisons due to the choice of statistical model will be most evident in the
later time periods where the power to detect differences is smallest (alternatively, the
effects of coatings in the earlier time periods are so large that almost any analysis will
reveal them).
The alternative models (AM) we chose to investigate are:
AM1 - The basic split-plot model used for the results in Table 3-6 is augmented with a
random effect for "parent board." The specimens used to construct the minidecks were
cut from larger "parent" boards. Thus it would not be unexpected for minideck
specimens from the same parent board to respond to treatment similarly.
AM2 - The model in AM1 with In(baseline) as a linear covariate. This shows that the
method of adjusting for baseline values is not critical.
AM3 - The model in AM2 with residual variances depending on bbf and sdeck. This
model is explained in more detail in Section 3.2.5.6.
AM4 - The model in AM3 with board dominant wood type (heartwood/sapwood) as a
fixed factor. Minideck boards were classified into three categories: predominatly
sapwood; predominantly heartwood; and nearly equally mixture of heartwood and
sapwood.
The first two columns of Table 3-7 are identical to the last two columns of Table 3-6.
The columns labeled AM1-AM4 are p-values for the comparison of the coating
identified in column 1 to the control coating for the four alternative models. Thus p-
values in the same row are comparable in the sense thay are testing the same
comparison. The differences among the p-values are due to the fact they are derived
from different statistical models fit to the data. That these differences in p-values are
small, and do not result in any substantial qualitatively different conclusions about
coating performance relative to the control coating, indicate the robustness of our
primary conclusions to choice of statistical model. In other words, accounting for parent
board effect, using baseline values for covariate adjustment (as opposed to direct
adjustment), allowing for non-constant error variation, and adjusting for board dominant
wood type, do not substantially alter the primary coatings comparisons results reported
in Table 3-6.
93
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-7. Sensitivity Analysis of 24-month Coatings Comparisons p-values
From Table
3-6
10
9
2
1
3
8
7
6
11
4
12
5
<0.0001
<0.0001
0.0024
0.0133
0.0273
0.1203
0.1830
0.2001
.2005
0.5434
0.7101
0.9635
AM1
<0.0001
<0.0001
0.0036
0.0156
0.0578
0.1274
0.1677
0.2030
0.3444
0.5847
0.8110
0.9952
AM2
<0.0001
<0.0001
0.0085
0.0282
0.0956
0.1551
0.1540
0.1723
0.3102
0.5137
0.6713
0.9333
AM3
<0.0001
<0.0001
0.0079
0.0254
0.0962
0.1470
0.1561
0.1607
0.2802
0.5801
0.7331
0.9091
AM4
<0.0001
<0.0001
0.0175
0.0318
0.2458
0.1733
0.2903
0.1343
0.4607
0.6835
0.8325
0.9249
3.2.5.6 Ch ecking for Heteroscedasticity
In the previous section, we used a model allowing variances to depend on bbf and
sdeck to demonstrate the robustness of our main conclusions. We now consider a
similar model to shed light on the performance of the wipe sampling method.
The lower portion of Table 3-5 shows that the error variances do not change much over
time, but does not shed light on possible residual error variance dependence on the
experimental factors. We fit several heteroscedastic variance models for the twin
purposes of exploratory analysis and to check the robustness of the conclusions based
on the homoscedastic error models used in Table 3-5. Variance model assessment via
the Bayesian Information Criteria (BIG) indicated that homoscedastic variance models
were best forthree of the seven time periods and, in the other fourtime periods,
variance models depending on source deck or bark face were marginally preferable to
homoscedastic error models. Thus, the case for using models with heteroscedastic
error variances is not strong.
However, as both sdeck and bbf relate to the physical characteristics of the board, and
the sampling equipment depends critically on board-to-sampling wipe contact, there is
reason to conjecture that residual error variation might be dependent on the factors bbf
and sdeck. Thus we present one final set of time period specific analyses derived from
94
-------
fitting statistical models that allow the residual error variance to differ for different levels
of the factors bbf and sdeck. A summary of the analysis of variance results appears in
Table 3-8.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-8. Heteroscedastic Analysis of Variance Models
Effect
coatl
bbf
coatl *bbf
sdeck
coatl *sdeck
bbfsdeck
coatl *bbf*sdeck
Mdeck (coatl)
Residual Down A
Residual Down C
Residual Up A
Residual Up C
DF
12
1
12
1
12
1
12
1 mo.
<0001
0.5171
0.5884
0.5356
<0001
0.5932
0.2843
0
0.6792
0.4043
0.3260
0.2482
3 mos.
<0001
0.6198
0.6004
0.5652
0.0007
0.1715
0.6116
0.06786
0.5206
0.6244
0.2723
0.2633
7 mos.
<0001
0.1576
0.1235
0.1553
<0001
0.0533
0.5401
0.04111
0.3588
0.1760
0.3308
0.1372
11 mos.
<0001
0.5839
0.3386
0.0858
0.1606
0.3500
0.2267
0.01366
0.5188
0.1935
0.2760
0.3032
15 mos.
<0001
0.5487
0.3961
0.0141
0.2705
0.0037
0.4268
0.04519
0.4781
0.2460
0.3317
0.2412
20 mos.
<0001
0.8641
0.0896
0.4451
0.2017
0.0199
0.6782
0.06841
0.4117
0.1503
0.4727
0.1470
24 mos.
<0001
0.4393
0.4689
0.0229
0.3158
0.0657
0.5334
0
0.3659
0.3781
0.4796
0.2307
A comparison of the analysis of variance portions of Tables 3-5 and 3-8 shows very
little difference between corresponding p-values, suggesting the robustness of the
former to possible violations of the assumption of homoscedastic error variation. (A
comparison of the corresponding p-values for comparings coatings to the controls in
the manner of the previous section also shows only minor differences and thus is not
presented here.) However, although the tests of factor effects differ little between the
two sets of analyses, there are interesting and potentially useful insights suggested by
the latter. The means overtime (with simple standard errors) of the residual variances
overtime are reported in Table 3-9.
Table 3-9 Mean Residual Variances from Table 3-8
BBF
Down
Down
Up
Up
SDECK
A
C
A
C
Mean
0.476
0.310
0.356
0.224
Std Err
(.042)
(.064)
(.033)
(.023)
95
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Because of the longitudinal aspect of the study, there is no reason to assume a priori
that the estimated variances are independent overtime; thus, the reported standard
errors likely underestimate variability. However, the magnitude of the standard errors is
such that even this rough analysis of the residual variances suggests that variation in
DAs measurements differs in a manner related to source deck and board bark face. A
detailed confirmatory analysis of this conjecture is beyond the scope (and data) of the
current report. It is presented here primarily for the benefit of those designing future
experiments that make use of the wipe sampling procedure used in this study.
The suggested conclusions are that measurements from the older sdeck-A boards are
more variable; and to a lesser extent so too are boards with a bbf=Down orientation.
3.2.5.7 Combined Data Modeling
We now present select results from models fit to the combined data. Six variance
models were used from the collection of options in the SAS Proc Mixed procedure:
three that incorporate different levels of departure from compound symmetry (CS,
CSH, and CS, with parameters dependent on bbfsdeck) and three that incorporate
time dependence [AR(1), ARH(1), and SP(POW), with time measured in months] and
variance heterogeneity in the case of the ARH(1) model. Table 3-10 displays a
summary of the results of fitting the alternative variance models to the combined data
set. The dependent variable is In(DAs) adjusted by ln(baseline DAs).
The six alternative variance models result in very similar ANOVA tables with a few
exceptions (marked by * in the column to the right of DF). The temporal dependence
models (AR(1), ARH(1), SP(POW)) suggest effects of coatTbbf and coatTbbfsdeck
whereas the others do not. Also coat1*sdeck*time is highly significant in the non-AR
models, but borderline significant in the AR models.
Note that the exceptions partition the AR models from the non-AR models. An
evaluation using BIG indicates greater suitability of the non-AR models compared to
the AR models. Also recall from Table 3-5 that neither coatTbbf nor coatTbbfsdeck
were found to be statistically significant in the time-period specific analyses.
Table 3-11 is comparable to Table 3-6 in that each coating is compared with Coating
13 at each time period based on the analyses of the combined data (all time periods)
using the CS variance model. In Table 3-11, p-values were adjusted using a Bonferroni
adjustment within each time period. The key finding is that there are no practical
differences between Table 3-6 and Table 3-11. Tables like Table 3-11 for the other five
variance models are comparably similar to Table 3-6 and not shown.
96
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-10. Summary of Results of Alternative Variance Modeling
Effect
coatl
bbf
sdeck
bbfsdeck
coatl *bbf
coatl *sdeck
coatl *bbf*sdeck
time
coatl *ti me
bbftime
sdeck*time
bbf*sdeck*time
coatl *bbf*time
coatl *sdeck*time
coatl *bbf*sdeck*time
Variance Models
DF
12
1
1
1
12
12
12
6
72
6
6
6
72
72
72
*
*
*
-2LL
BIG
CS
p-value
<0001
0.7122
0.2835
0.0417
0.2025
0.0027
0.3534
<0001
<0001
0.4921
0.0007
0.2990
0.9987
<0001
0.9913
1421.0
1428.3
CSH
p-value
<0001
0.7084
0.2704
0.0369
0.1675
0.0019
0.3096
<0001
<0001
0.3335
0.0002
0.3030
0.9984
<0001
0.9834
1403.2
1436.2
CS(B*S)
p-value
<0001
0.7116
0.2839
0.0423
0.2077
0.0037
0.3557
<0001
<0001
0.4939
0.0007
0.2999
0.9987
<0001
0.9914
1406.8
1439.8
AR(1)
p-value
<0001
0.6598
0.1893
0.0124
0.0259
<0001
0.0794
<0001
<0001
0.3764
0.0011
0.3692
0.9997
0.0578
0.9967
1455.3
1466.3
ARH(1)
p-value
<0001
0.6624
0.1875
0.0124
0.0254
<0001
0.0789
<0001
<0001
0.2556
0.0004
0.3792
0.9994
0.0603
0.9941
1442.5
1475.5
SP(POW)
p-value
<0001
0.6439
0.1673
0.0086
0.0125
<0001
0.0459
<0001
<0001
0.4670
0.0019
0.3917
0.9999
0.0269
0.9984
1473.2
1484.2
Table 3-11. Coatings Comparisons from the Combined Data Modeling
Model CS
1 month
11
10
9
3
8
1
12
6
7
4
2
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0810
3 months
10
9
11
3
8
1
6
2
12
4
7
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0621
7 months
10
9
8
11
3
1
2
6
7
4
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0708
11 months
9
10
8
11
3
1
6
4
2
7
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0001
0.0001
0.0002
0.9999
15 months
10
9
3
1
8
11
6
7
4
2
12
5
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0002
0.0010
0.0618
20 months
10
9
1
3
2
8
6
7
4
11
12
5
<0.0001
<0.0001
0.0007
0.0010
0.0138
0.1262
0.2393
0.2798
0.2848
0.6930
0.9999
0.9999
24 months
10
9
2
1
3
8
7
6
11
4
12
5
<0.0001
<0.0001
0.0024
0.0157
0.0351
0.1907
0.3141
0.3500
0.3507
0.9999
0.9999
0.9999
97
-------
3.2.6 Summary of Coating Comparison Results
Inspection of the 13th column in Figure 3-29 reveals that each coating mitigated DAs
somewhat when compared to the positive control. However, recall that the visual
comparisons made possible in Figure 3-29 are not precise and not adjusted for the
effects of multiple hypothesis testing.
Table 3-12 summarizes the differences between the performance of each coating
versus the positive control [minideck 13 (uncoated)] as well as the statistical
significance of the differences at each time interval. The color-coding of cells indicates
level of significance: blue indicates a p-value of less than 0.01; green indicates a p-
value between 0.01 and 0.1; and beige indicates a p-value of greaterthan 0.1. The p-
values and estimated differences under the columns labeled "1 Mo." - "24 Mo." were
calculated form the time period specific analyses reported in Tables 3-5 and 3-6. The
p-values and estimated differences under the column labeled "Combined" were
calculated from the CS analysis of variance model reported in Tables 3.10 and 3.11. In
this case the estimated difference is the difference in estimated main effects over all
time periods. When justified, combining data overall time periods results in more
power for detecting differences in main effects of coatings. This is seen, for example,
with Coating 5, for which the time-period specific differences are mostly far from
significant; whereas the difference from the combined data analysis is borderline
nonsignificant (p-value = .0846).
From inspection of the pain/vise tests summarized in Table 3-12, the adjusted p-value
for comparing the main effects of coatings 5 and 13 (uncoated) is only marginally
significant at the .05 level of significance for two time periods. Thus, there is little
evidence of a difference in performance between coatings 5 and 13 (uncoated). All of
the other coatings exhibited statistically significant differences in performance versus
the control through the 11-month sampling event, or after approximately one year of
weathering. The paints were the only products that continued to show statistically
significant performance differences versus the control throughout the two-year study
period. All of these results are borne out visually in the graphical presentation in
Sections 3.2.2 through 3.2.4.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
98
-------
Table 3-12 reports estimated differences of the means of the natural logarithms of DAs.
This is the best scale for the statistical analysis, but not the best scale for interpretation.
A preferable scale for interpretation is percent reduction relative to the control (#13)
minidecks. The estimated differences of logs in Table 3-12 were converted to
estimated percent reductions relative to the control minidecks by exponentiating the
differences, subtracting from 1, and multiplying by 100. These estimated percent
reductions are reported in Table 3-13 along with lower and upper 95 percent
confidence limits (LCL /UCL) for all coatings and time periods. The confidence limits
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-12. Summary of Estimated In(DAs) Difference and Associated Statistical
Significance between Each Coating and Control [#13 (uncoated)] Minidecks at
Each Time Interval
Coating
1
2
3
4
5
6
7
8
9
10
11
12
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
p-value
est. diff. (In)
Time After Coating
1 Mo.
<.0001
-1 .9863
0.0002
-1.1823
<.0001
-3.3198
<.0001
-1.4124
0.0885
-0.6806
<.0001
-1 .5275
<.0001
-1.4183
<.0001
-2.1129
<.0001
-4.2475
<.0001
-4.5213
<.0001
-4.6796
<.0001
-1 .8045
3 Mo.
<.0001
-2.3589
0.0002
-1.7118
<.0001
-3.4487
0.0003
-1.6170
0.2471
-0.7028
<.0001
-1 .9779
0.0027
-1 .3527
<.0001
-2.9301
<.0001
-3.6731
<.0001
-3.8432
<.0001
-3.5821
0.0002
-1 .7526
7 Mo.
<.0001
-1 .7905
<.0001
-1.6411
<.0001
-2.1945
0.0002
-1 .3988
0.0973
-0.6919
<.0001
-1 .5046
<.0001
-1 .4666
<.0001
-2.6787
<.0001
-2.8327
<.0001
-3.0658
<.0001
-2.3494
0.0004
-1 .2906
11 Mo.
<0001
-1 .8829
0.0010
-1.1558
<0001
-2.2290
<0001
-1 .4880
0.7416
-0.3457
<0001
-1 .5669
0.001 1
-1.1477
<.0001
-2.6567
<.0001
-3.3215
<0001
-3.1612
<.0001
-2.5833
0.0019
-1 .0874
15 Mo.
<.0001
-1 .7465
0.0089
-1.1052
<.0001
-1 .7697
0.0036
-1.2137
0.1746
-0.7032
0.0006
-1.4131
0.0024
-1.2618
0.0001
-1 .6222
<.0001
-2.6053
<.0001
-2.6961
0.0005
-1 .4385
0.0213
-0.9965
20 Mo.
0.0165
-1.0185
0.0758
-0.8194
0.0200
-0.9944
0.3707
-0.5672
0.9994
-0.1633
0.3396
-0.5837
0.3678
-0.5687
0.2435
-0.6423
<.0001
-1 .6803
<.0001
-2.2269
0.5729
-0.4751
0.7404
-0.4052
24 Mo.
0.0133
-0.8097
0.0024
-0.941 1
0.0273
-0.7487
0.5434
-0.4034
0.9635
-0.2338
0.2001
-0.5467
0.1830
-0.5574
0.1203
-0.6050
<.0001
-1 .7832
<.0001
-2.4219
0.2005
-0.5465
0.7101
-0.3492
Combined
<.0001
-1 .6562
<.0001
-1 .2224
<.0001
-2.1007
<.0001
-1.1618
0.0846
-0.5030
<.0001
-1 .3029
<.0001
-1.1105
<.0001
-1.18926
<.0001
-2.8776
<.0001
-3.1338
<.0001
-2.2364
<.0001
-1.0911
99
-------
were calculated by first constructing 95 percent confidence intervals for the estimated
differences of the logs and then transforming those lower and upper bounds using the
same formula for converting the estimated differences of logs to percent reductions.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-13. Estimated Percent Reduction for Each Coating and Estimated 95 Percent
Confidence Intervals (color-coding same as in Table 3-12)
Coating
1
2
3
4
5
6
7
8
9
10
11
12
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
%Reduction
LCL/UCL (%)
%Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
% Reduction
LCL/UCL (%)
Time After Coating
1 Month
86
71/93
69
36/85
96
92/98
76
49/88
49
-6/76
78
54/90
76
49/88
88
75/94
99
97//99
99
98/99
99
98/100
84
65/92
3 Month
91
75/96
82
53/93
97
92/99
80
48/92
50
-30/81
86
64/95
74
32/90
95
86/98
97
93/99
98
94/99
97
93/99
83
54/94
7 Month
83
64/92
81
58/91
89
76/95
75
46/89
50
-8/77
78
52/90
77
50/89
93
85/97
94
87/97
95
90/98
90
79/96
72
40/87
11 Month
85
68/93
69
33/85
89
77/95
77
52/89
29
-50/67
79
56/90
68
33/85
93
85/97
96
92/98
96
91/98
92
84/96
66
29/84
15 Month
83
58/93
67
20/86
83
59/93
70
28/88
51
-20/80
76
41/90
72
31/88
80
52/92
93
82/97
93
84/97
76
43/90
63
11/85
20 Month
64
13/85
56
-6/82
63
11/85
43
-36/76
15
-104/65
44
-34/77
43
-36/76
47
-26/78
81
55/92
89
74/96
38
-49/74
33
-60/72
24 Month
56
11/78
61
22/80
53
5/76
33
-34/67
21
-58/60
42
-16/71
43
-15/71
45
-9/73
83
66/92
91
82/96
42
-16/71
29
-41/65
Table 3-13 further illustrates what has been demonstrated graphically in Figure 3-29;
that is, product performance for the reduction of dislodgeable arsenic residues
decreases with time. For the products tested, the paints ranked highest, followed by
multi-coat stains and sealants and the vinyl elastic encapsulant, followed by the single-
100
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
coat products. However, confidence intervals for estimated percent reduction are
relatively large, challenging any attempts to rank and compare product performance
based upon estimates of percent reduction. Nevertheless, the information in Table 3-13
provides an easily interpreted picture of the range of product performance and, if used
in conjunction with the information in Figure 3-29, makes is possible to easily
differentiate between individual products and compare the relative performance of
products at any point in time.
The study design stipulated the use of deck surface preparation protocols
recommended by the coating manufacturer. Accordingly, minidecks forten of the
products were cleaned using a specific recommended cleaning agent followed by a
water rinse, whereas minidecks coated with Product 3, an oil-based stain, received
only a water rinse, and the minidecks coated with Product 7, a water-based stain, were
neither cleaned nor rinsed. The positive control minidecks were rinsed with water but
not treated with a cleaning agent. As seen in Table 3-12, DAs for minidecks coated
with Product 7 were statistically different than the positive controls through the first 15
months of weathering. This evidence suggests that coating may be more effective than
rinsing with water, at least through a one-year period. The decks coated with Product 3
received the same treatment as the positive controls and DAs was significantly less
than the positive controls through 15 months. Again, these limited observations
suggest that coating (with most products) is more effective than washing alone in
reducing DAs. However, the experimental design does not evaluate the impact of
periodic washing or periodic re-treatment with a coating product.
3.3 Coating Appearance
Coating appearance throughout the study was qualitatively assessed in person and by
recording and reviewing photographs of each minideck at each sampling event
(precoat and after 1 month, 3 months, 7 months, 11 months, 15 months, 20 months
and 24 months of weathering). The photos are provided in Appendix D. Notable
observations are summarized in Table 3-14.
3.4 Weather Data
A number of meteorological measurements were made during this study, as described
in Section 2.7. The most relevant parameters in terms of coating performance and
sampling are Solar Radiation, Rainfall, and Temperature. Tables 3-15 and 3-16
summarize key weather parameters by sampling interval, where the two numbers in
the first column refer to the sequentially numbered sampling events defining the period
over which the weather parameters are summarized. Figures 3-30 through 3-32 are
plots of key weather-related parameters versus time. In each of these plots, vertical
bars indicating the dates of the sampling events are superimposed for reference.
Complete weather data collected are provided in Appendix P.
101
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-14. Summary of Visual Observations of Minidecks
Coating #
Coating 1
Coating 2
Coating 3
Coating 4
Coating 5
Coating 6
Coating 7
Coating 8
Coating 9
Coating 10
Coating 1 1
Coating 12
Coating 13
Product Type
Sealant
Sealant
Stain
Stain
Sealant
Sealant
Stain
Stain
Paint
Paint
Other
Other
No coating
Base
Oil
Oil
Oil
Oil
Water
Water
Water
Water
Water
Oil
Cover
Semi
Clear
Clear
Clear
Clear
Clear
Semi
Clear
Opaque
Opaque
Clear
Clear
Pigmentation
"Cedar"
"Clear"
"Deep Tone Base"
(no pigment added)
"Clear Stain"
"Clear"
"Clear"
"Cedar"
"Tint base, solid" (no
pigment added)
Gray
Gray
Clear
Clear
Main Ingredients
Aliphatic,
Napthalene
Acrylic, alkyd,
urethane
Acrylic
Alkyd
Unknown*
Acrylic, alkyd
Alkyd
Acrylic
Acrylic
Alkyd,
polyurethane
Elastic vinyl
Polymer
Summary of Visual Observations
Deep red coloration. Some wear-through where wiped.
No visible signs of coating. Relatively light and bright wood appearance.
Extensive black mold or mildew on untreated boards, with varying
amounts, from slight to extensive, on treated boards. Growth appeared
between 7 and 1 1 months after coating. Otherwise, no visible signs of
coating, although when freshly coated the wood appearance was
significantly darkened.
No visible signs of coating. Some mold or mildew on untreated boards.
No visible signs of coating. Relatively dark (gray) wood appearance.
Slightly yellow tint to treated boards. Some wear-through where wiped.
Light red coloration. Wear-through where wiped.
Very slight tint on treated boards, but generally no visible signs of
coating. Relatively light and bright wood appearance.
Retained gray paint coloration, but moderate-to-extensive chipping,
especially at cracks, starting around 7 months.
Retained gray paint coloration, slight-to-moderate chipping at cracks.
Some black mold or mildew on untreated boards.
Extensive black mold or mildew on untreated boards with varying
amounts, from none to moderate, on treated boards. Growth appeared
between 7 and 1 1 months after coating. Seems to visually perform
better on A source than C source. Some limited chipping and peeling at
large cracks. General appearance is slick and waxy, with an amber
coloration that held well on treated boards.
No visible signs of coating.
No visible signs of coating. Some mold and mildew on untreated
boards.
' MSDS does not include any ingredients >1 percent by weight
102
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 3-15. Summary of Key Weather Parameters
Sampling
Interval
0-1
1-2
2-3
3-4
4-5
5-6
6-7
Total Solar
Radiation
(W/m2 x day)
6075
8567
11953
26801
23980
15145
27860
Avg. Solar
Radiation
(W/m2)
196.0
142.8
105.8
216.1
179.0
116.5
222.9
Total Rainfall
(inches)
4.71
4.63
8.54
11.14
24.53
9.02
13.36
Precip Events
>0.25" (#)
5
4
13
13
22
11
13
Precip Events
>0.10"(#)
7
10
17
18
24
16
18
Table 3-16. Summary of Temperature Measurements
Sampling
Interval
0-1
1-2
2-3
3-4
4-5
5-6
6-7
Avg. Temp.
(degrees F)
74.5
61.1
41.3
65.3
69.8
44.3
70.8
Avg. High
Temp
(degrees F)
86.0
73.5
54.0
77.3
78.8
56.3
82.3
Avg. Low
Temp
(degrees F)
68.1
51.7
32.3
55.7
59.4
34.5
59.3
Max. High Temp
(degrees F)
95.4
86.3
82.2
95.1
95.9
78.0
103.3
Min. Low Temp
(degrees F)
52.9
31.8
9.5
21.3
28.0
11.0
35.0
103
-------
Weekly Avg. Solar Radiatioi
Sampling Events
Cumulative Solar Radiation
Figure 3-30. Solar Radiation Data Summary
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 3-31. Rainfall Data Summary
104
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 3-32. Temperature Data Summary
105
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
106
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
4. Study Limitations, Exploratory Analyses and Future Studies
The primary objective of the testing described in this report was to evaluate coatings for
their efficacy in reducing DCCA from the surfaces of pressure-treated wood subjected
to outdoor weathering because of concern of potential health consequences due to
ingestion of dislodgeable CCA residues through hand-to-mouth behaviors of young
children that contact CCA-treated wood. Thus, the study was designed to provide
insight into the utility of a potential risk management strategy that is readily available
and commonly used by consumers to maintain the appearance and durability of decks.
EPA used a modification of a wipe-sampling protocol developed by the staff of the
CPSC. CPSC staff had previously developed linkages between the amounts of DCCA
determined by the wipe-sampling protocol and amounts of DCCA transferred to hands
by contact with CCA-treated wood. Due to uncertainties regarding the relationship
between dose and risk and regarding the sensitivity of the wipe-sampling method to
detect changes in DCCA on coated surfaces, EPA modified the CPSC wipe-sampling
method to improve the sensitivity of the method to detect changes in DCCA.
Additionally, the project was focused in a way that precluded the ability to answer all of
the myriad questions raised in the development of the test protocol employed,
execution of the project, and evaluation of the results. The objective of the following
discussion, therefore, is to better define study limitations, present exploratory data
analyses that may identify unanswered questions, and suggest future research work
and considerations for subsequent studies.
4.1 Substrate Characteristics
Only two sources of aged CCA-treated wood were tested, which is not likely to be
representative of the universe of CCA wood structures currently in service. In
particular, the two sources represented two different ages and conditions (one-year
and seven year) of the same species wood (SYP).
Outdoor structures like residential decks may have service lives of 20 years or more.
As CCA treated wood is no longer widely used to construct such structures, the
average age of those currently in use is continually increasing. As such, testing of
CCA-treated wood from sources in service for longer than seven years may be
warranted in future studies.
In addition to southern yellow pine, other species such as hemlock fir, Douglas fir,
spruce-pine fir, red pine, ponderosa pine and radiata pine may be regionally popular
and thus could be tested in future studies.
Nailholes, knots, and other surface irregularities can be expected to have an impact on
wipe sampling and DCCA measurements. As such, these surface features were
107
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
avoided to the extent possible in this study. In particular, nailholes were completely
avoided during wipe sampling events (wipe samples were taken between adjacent sets
of nailholes). Furthermore, existing nailholes were reused when assembling minidecks.
Other surface irregularities were avoided to the extent practical when selecting
specimens to be used for assembling minidecks.
To the extent that such irregularities could not be avoided, each specimen and
sampling area were characterized visually in two ways: by filling out a specimen
characterization form (described further in Section 2.2 and reported in Appendix B) and
via a photo record of each specimen, which included pre- and post-coating
photographs of each minideck and photographs of each minideck prior to each regular
sampling event.
Defects recorded included number of knots and the degree of splintering, cracking, and
rotting for each specimen. The characteristics recorded included predominant grain
type (flat versus edge grain), predominant ring spacing (tight, medium, wide),
predominant wood season (early versus late wood), and predominant wood type
(heartwood versus sapwood). The only wood characteristic explicitly controlled during
the study was predominant grain orientation (up versus down). The effect of grain
orientation was minor compared to other factors in the experiment.
Heartwood/sapwood classification (predominantly HW, predominantly SW, even mix)
of the boards was not significant when as added as fixed main effect to the analysis of
variance models. However, this factor was not controlled in the experiment, and the
determination of heartwood/sapwood was subject to misclassification, thus the power
to detect such an effect was likely compromised.
Additional work could be conducted to determine the specific impacts of various wood
surface characteristics on DCCA and coating performance. It is known that certain
wood characteristics affect CCA retention. In particular, heartwood -the nonliving
center of the tree-typically contains little CCA after treatment (Lebowet al., 2003).
Additionally, large differences in wood erosion rates (where it is speculated that DCCA
correlates positively with wood erosion) between earlywood and latewood were
observed in vertically-oriented test specimens weathered at a site in Wisconsin over an
exposure period of 7 years or longer (Williams et al., 2001 a, 2001 b, 2001 c).
4.2 Geography/Climatic Effects
Experiments were conducted at an outdoor site in Research Triangle Park, NC, which
is representative of southeastern or mid-Atlantic US climatology and features summer
weather conditions characterized by intense solar radiation, high humidity, and
frequent thermal shock from rain showers, but mild average conditions throughout the
remainder of the year, relative to more northerly US geographic regions. CPSC staff
108
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
conducted their companion study in Gaithersburg, MD, which features similar climatic
conditions.
The wood and coating industries typically conduct product testing in extreme climates
represented by two geographical regions in the US, the deep south for moisture and
photodegradation and the Snow Belt for cyclic freeze-thaw stresses. Future studies
should consider testing in more extreme and/or representative locations in the US. In
addition to the two extremes mentioned, other major climatic regions could be
considered, including Northwestern, High Plains, Midwest/Ohio Valley, upper New
England/Mid-Atlantic, lower Southeast, Southern and Southwestern.
In addition to geographic location, directional exposure effects (e.g., south- versus
north-facing decks), particularly of the source structures, could be important variables
in studies similar to that which was conducted.
4.2.1 Exploratory Analysis of Potential Weather Effects
It is possible and perhaps even likely that the effects of weather on coating
performance will differ from one coating to another—film formers might be affected by
weathering differently than penetrating coatings, for example. Thus, in searching for
possible weather effects, it is necessary to consider coatings individually. For this
project, the effects of weather and time were jointly examined using regression
methods for each of the 13 coatings. The response variables in these analyses are the
In(DAs) values averaged over grain and source deck—the In(DAs) averages plotted in
the bottom row of Figure 3-26. Weather effects were investigated using multiple linear
regression models with time as one predictor and a second predictor corresponding to
one of the climatological variables taken in succession. For these analyses it is
assumed that the averaged In(DAs) values vary linearly with time and the included
weather variable. That is, with time and one weather variable, we assumed the
following statistical model:
In(DAs) = bO + b1*time + b2*weather+ (random error)
The limitation of this analysis is that there are only 7 data points to fit such a model.
Nevertheless, it can be done and the question becomes whether b2, the coefficient of
the weather variable, is significantly different from 0.
A total of 15 weather variables, as listed in Table 4-1, were selected for analysis. Since
there were 13 coatings by 15 weather variables, 13 x 15 = 195 such regressions were
conducted. The results are summarized in a 13-by-15 array and one figure.
109
-------
Table 4-1. List of Weather Variables Considered in Statistical Analysis
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Weather Variable
TSRad
Cum_TSR
Radjnt
TRain
#HRain
#LRain
AT
AHT
MHT
ALT
MLT
0-#Frz
ATDiff
#LTDiff
#STDiff
Description
Total solar radiation
Cumulative solar radiation
Average solar radiation
Total rainfall
Number of rain events greater than 0.25"
Number of rain events greater than 0.10"
Average temperature
Average high temperature
Maximum high temperature
Average low temperature
Minimum low temperature
Number of freezing events
Average daily temperature difference
Number of days with a temperature difference greater than 26 F
Number of days with a temperature difference greater than 32 F
The array provides the partial correlations between In(DAs) and the weather variables
after adjusting for time effects. Table 4-2 includes the correlations for all cases in this
array where the p-value < .10. These are the only correlations worthy of further
consideration.
110
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 4-2. Partial Correlations for Pairs with p-values < 0.10
Coating
1
2
3
4
5
6
7
8
9
10
11
12
13
(uncoated)
1
—
—
—
—
0.76
—
—
—
—
0.82
—
0.83
0.88
2
—
—
—
—
—
—
0.74
—
—
—
—
—
—
3
0.79
0.87
—
0.93
0.87
0.78
0.93
—
—
—
—
0.78
0.9
4
—
—
—
—
—
—
—
—
—
—
—
—
—
5
—
—
—
—
—
—
—
—
—
0.77
0.78
—
—
6
—
—
—
—
—
—
—
—
—
0.86
0.91
—
—
7
—
0.75
—
0.82
—
—
0.8
—
—
—
—
—
0.74
8
—
0.75
—
0.84
—
—
0.82
—
—
—
—
—
0.74
9
0.73
0.74
—
0.86
—
—
0.74
—
—
—
—
—
0.85
10
—
0.77
—
0.84
—
—
0.81
—
—
—
—
—
0.73
11
—
—
—
0.79
—
—
—
0.83
—
—
-0.74
—
—
12
—
—
—
—
—
—
0.77
—
—
—
—
—
—
13
—
—
—
—
—
—
—
-0.77
—
—
—
—
—
14
—
-0.8
—
-0.77
—
—
-0.73
-0.8
—
—
—
—
—
15
—
—
—
—
—
—
—
-0.75
—
—
—
—
—
Figure 4-1 shows the 195 partial regression plots resulting from the 13 x 15 matrix of
coating x weather factor. Columns correspond to the 15 weather variables shown in
Table 4-1, while rows correspond to coatings 1, 2,..., 13. Partial regression plots are
constructed to show how In(DAs) relates to weather after the linear effects of time are
removed. Each plot contains 7 points and also the partial regression line. A black solid
line means that the correlation is statistically significant at .10 significance level (.10 is
used because of the exploratory nature of the evaluation—so as to not miss
relationships that might exist but are weak).
The black lines in columns 7, 8, 9,10 and 11 of Figure 4-1 are all temperature
variables and are essentially the same for the purposes of this analysis. So if one of
them is significant then it is highly likely that the others will also be significant (or close
to significant).
111
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank.
112
-------
MD 1
MD2
MD3
MD4
MD5
MD6
MD7
MD8
MD9
MD 10
MD 11
MD 12
MD 13
Solar Radiation
Y
Precipitation
Temperature Factors
_*_.__
* *
\
*.
8
Weather Variable
10 11 12 13 14 15
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Weather Variable
TSRad
Cum_TSR
Radjnt
TRain
#HRain
#LRain
AT
AHT
MHT
ALT
MLT
0-#Frz
ATDiff
#LTDiff
#STDiff
Description
Total solar radiation
CumulativegS^plar radiation
Average solar^diation
Total rainfall
Number of rain events greaterth^ngO^"
Number of rain events greaterth^nrj):|1rj}"
Average temperature
Average high temperature
Maximumh^igh temperature
Average low temperature
Minimum low^mperature
Number of freezing events
Average daily^mperature^ifference
Number of^ys with a temperature
difference greater than 26 °F
Number of^ys with a temperature
difference greater than 32 °F
= Partial correlation significant at 0.1 level
Figure 4-1. Partial Regression Plots of Coating x Weather Factor ([each subplot has In(DAs) as its y-axis and the weather variable as its x-axis]
113
-------
This page intentionally left blank.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
114
-------
The largest correlations (0.93) are associated with Radjnt; nine of the thirteen
treatments correlated significantly with at least one of the radiation measures. Two of
the products - both film-formers - had a significant correlation with number of
precipitation events greater than 0.1 and 0.25 inches, consistent with the expectation
that shrink and swell due to moisture and thermal shock may impact the performance
of film-forming products. Significant temperature correlations were observed for about
one-half the treatments, and in nearly all cases, those treatments also responded
positively to a solar radiation measure. No significant correlations were observed for
products #3 and #9, products that performed relatively well in terms of reduction of
DCCA.
4.3 Coating Type and Selection
The coatings tested in this study represented a convenience sample readily available
to the project team with limited numbers of representatives from the major classes of
consumer products. It was not intended to be fully representative of the universe of
available coatings. Additionally, products within a given classification can vary
considerably in formulation and effectiveness and the products selected for this study
were not intended to be representative of their respective categories. Nevertheless,
exploratory analyses were conducted to assess differences among subgroups of
coatings as identified by information on the product label or MSDS.
Previous efforts in coating formulation, as reported in the literature, have focused on
the efficacy of coatings and surface preparation in resisting weathering as measured
by erosion rate and leaching of CCA into water from simulated or actual precipitation.
Lebow, et al. (2003) in a thorough review of published literature suggest that water
repellent content, pigment concentration, binder system and additives may affect the
ability of coatings to reduce DCCA.
In addition to the pairwise tests reported in Section 3.2.5, differences among certain
identifiable subgroups of coatings were also tested. The results of these exploratory
analyses are reported in Table 4-3 (the p-values are not adjusted for multiplicity). Note
that all of the comparisons presented in Table 4-3 were done without coatings 9 and 10
(the paints) in the analytical model with the exception of the "Multi-Coat versus Single
Coat" comparison.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
115
-------
Table 4-3. Pairwise Comparisons of Coating Characteristics
Comparison
Sealant vs. Stain
Penetrating vs. Film-Former
No Acrylic vs. Acrylic
No Alkyd vs. Alkyd
Oil-Based vs. Water-based
Multi-coat vs. Single Coat
Clear Cover vs. Semi-Transparent
t-statistic
1.5810
8.0861
8.0752
-9.1344
-4.3004
-7.5055
-0.4443
p-value
<.0001
<.0001
0.0002
<.0001
0.0444
<.0001
0.6076
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
The "sealant vs. stain" comparison tested the average difference in In(DAs) between
sealants and stains. Sealants had higher average In(DAs) than stains (p-value <
0.0001), meaning that the sealants performed worse than the stains in these
experiments. Penetrating coatings had higher average In(DAs) than the film formers (p-
value < 0.0001), non-acrylics had higher average In(DAs) than the acrylics (p-value =
0.0002), and non-alkyds had lower average In(DAs) than the alkyds (p-value < 0.0001).
The difference in performance between oil- and water-based products was not
significant at the 0.05 level when paints were excluded from the analysis (p-value =
0.0444); a potentially useful observation in light of the phase-out of oil-based products
in response to phased implementation of the Clean Air Act amendments restricting the
atmospheric release of volatile organic compounds. The products for which more than
one coat was recommended (and applied) performed significantly better than the
products for which a single coat was recommended (p-value <0.0001). However, the
number of coats is confounded with both the recommended application procedure and
with the product. As such, one cannot infer from this observation that multiple coats of
a product provide improved reduction of DCCA because the study design did not
examine this question directly by testing multiple application levels for any product.
However, inspection of Tables 3-6 and 3-9 reveals that all six of the single-coat
products ranked lowest of the twelve pair-wise comparisons through the first five
sampling intervals and five ranked in the lowest six at sampling periods six and seven.
Finally, no significant difference in performance was detected between products with
clear versus semi-transparent covers.
In this study, a convenience sample of products was tested for a purpose for which
most were not advertised or designed. Future studies should investigate the impact of
coatings specifically formulated for their ability to mitigate DCCA (note that two
products advertised to reduce exposure to CCA were tested in this study). As such,
this test protocol, or a modification of this protocol, may be useful in efforts to evaluate
and demonstrate the efficacy of products specifically formulated for their ability to
116
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
reduce DCCA on appropriate substrates in various climatic regions and for specific use
conditions.
4.4 Surface Preparation, Coating Application and Reapplication
In this study, manufacturers' instructions for surface preparation prior to coating were
strictly followed. As such, minideck surfaces for ten of the products were cleaned using
a specific recommended cleaning agent followed by a water rinse, thus confounding
surface preparation method with coating for these products. Minidecks coated with
Product 3, an oil-based stain, received only a water rinse, and the minidecks coated
with Product 7, a water-based stain, were neither cleaned nor rinsed, per strict
adherence to the manufacturers instructions (or lack thereof) on the product label. The
positive control decks were rinsed with water but not treated with a cleaning agent. The
results associated with these products allowed for a preliminary assessment of the
impacts of surface preparation on DCCA.
As seen in Table 3-12, DAs for minidecks coated with Product 7 were statistically
different than the positive controls through the first 15 months of weathering. This
evidence suggests that coating may be more effective than rinsing with water, at least
through approximately a one-year period. The decks coated with Product 3 received
the same treatment as the positive controls and DAs was significantly less than the
positive controls through 15 months. Again, these limited observations suggest that
coating (with most products) is more effective than washing alone in reducing DAs.
However, the project did not investigate the comparative performance of coatings
versus periodic (e.g., monthly, quarterly) cleaning of the surface of the deck overtime.
Periodic cleaning may mitigate dermal CCA exposure and its effects could be
compared with the impacts of coatings on DCCA in future studies.
The preparation of the wood surface prior to coating may be important in several ways.
First, it is possible that the surface preparation method itself may constitute a
significant exposure activity. For example, sanding CCA wood surfaces without proper
respiratory equipment may facilitate exposure via inhalation. The surface preparation
technique may also impact the pre-coat DA levels as well as future migration of DCCA
to the surface of the wood. In this study, the coating manufacturer's printed instructions
with regards to surface preparation were followed strictly. Baseline measurements,
used in some calculations of efficacy, were taken prior to any surface preparation or
rinse step so we did not obtain data to systematically evaluate the impact of
preparation steps on DCCA. Future studies could consider this issue more directly and,
compare the efficacy of periodic cleaning procedures with the efficacy of periodic
treatment with coating products.
It is possible that the method of applying coatings may also affect DCCA
measurements. Wood was prepared and coatings were applied per manufacturer's
117
-------
instructions. A brush application coating technique was used for each coating based
upon the assumption that brush application represents the most common method
employed by residential users of deck coatings. Additional studies could consider the
effect of coating application technique on coating performance and DCCA
measurements.
4.4.1 Exploratory Analysis of Coating and Paint Chip Samples
Preparing CCA surfaces for the reapplication of a coating may represent a potential for
exposure. For example, the removal of film-forming products in preparation for
recoating may subject users to increased exposure to CCA chemicals associated with
associated particulates, including dusts. During the minideck study, it was noted that
the painted surfaces were peeling (more notably for the water-based paint than the oil-
based paint) and, the peeling materials were sampled for an exploratory analysis of
CCA in the paint chips. The methodology associated with this testing is described in
Section 2-14.
The results of raw coating analyses are presented in Table 4-4 expressed in ug of total
arsenic, chromium and copper per g of raw (wet) coating and perg of dry coating.
While the paints, coatings #9 and 10, had higher concentrations than did the other
coatings, all of the concentrations were relatively low. A "U" entry in Table 4-4 is a
qualifier indicating that the result was below reporting limits, meaning that the of
concentration in the fluid used to digest the raw coating was less than 2 ug/L.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 4-4. Raw Coating CCA Analyte Concentrations (raw/wet weight basis)
Coating
1
2
3
4
5
6
7
8
9
10
11
12
Primer
Dry Solids
25.7 percent
28.3 percent
44.2 percent
35.6 percent
8.8 percent
13.6 percent
10.3 percent
28.8 percent
55.4 percent
67.2 percent
48.1 percent
17.9 percent
34.1 percent
raw concentration (|jg/g)
As
0.62
U
0.41
U
U
U
U
U
1.64
3.99
U
U
0.90
Cr
2.09
0.32
1.14
0.46
0.38
0.74
0.47
1.23
1.89
20.48
3.07
1.08
1.53
Cu
2.09
0.59
1.21
0.78
0.71
1.57
1.3
3.21
18.87
6.86
1.25
0.42
0.77
dry concentration (|jg/g)
As
2.42
U
0.93
U
U
U
U
U
2.96
5.94
U
U
2.65
Cr
8.15
1.13
2.58
1.29
4.32
5.46
4.54
4.28
3.41
30.49
6.38
6.03
4.50
Cu
8.15
2.08
2.74
2.19
8.07
11.58
12.57
11.16
34.07
10.21
2.60
2.34
2.25
118
-------
By contrast, Table 4-5 summarizes the results of the paint chip sampling and analysis.
There are several key points:
1. The concentrations of total arsenic, chromium and copper in the paint chips taken
from the treated wood are significantly higher than in the paint chips from the
untreated wood.
2. The concentrations of total arsenic, chromium and copper in the paint chips taken
from the treated wood are significantly higher than in the raw paint, even after
accounting for weight reductions due to volatile losses upon drying of the paints.
3. No matrix effects appear to be confounding the analyses. This is evidenced by the
close agreement in concentration between the split samples of paint chips with
different masses.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 4-5. Paint Chip CCA Analyte Concentrations
Sample
Paint 9 Treated
Paint 9 Treated
Paint 9 Untreated
Paint 9 Untreated
Paint 10 Treated
Paint 10 Treated
Paint 10 Untreated
Paint 10 Untreated
Mass of paint chips
(g)
1.00
0.50
0.70
0.34
0.25
0.10
1.00
0.50
Concentration (ug/g)
As
225
220
13
15
249
310
12
12
Cr
100
118
8
22
247
330
32
38
Cu
110
108
29
33
165
190
16
16
Additionally, archived subsamples of paint chips from Paint 9 Treated and Paint 10
Untreated were manually examined using a LEO/Zeiss Model 440 SEM equipped with
a PGT IMIX energy-dispersive X-ray spectrometer (EDX) to generate elemental
composition information on individual particles. This examination confirmed that
arsenic was associated with the top and bottom surfaces of the paint chips recovered
from the CCA-treated boards of deck 9. The full analysis report is included in Appendix
Q.
119
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
While these results suggest that higher levels of CCA chemicals are associated with
paint chips from the surfaces of CCA-treated wood surfaces, it is unclear whether the
increases can be attributed to wood fibers associated with the peeling paints, migration
and uptake of CCA chemicals into the coating products, or some other association.
4.5 Test Specimen Dimensions
Due to the relatively small size of the minidecks, the stress factors associated with
attached specimens during weathering are not likely to be representative of those of
full-sized, in-use structures.
Future studies may also evaluate the impact of attachment points (nail and screw) on
DCCA. For example, it is possible that the surface penetration associated with the
fasteners facilitates mobilization of CCA chemicals to the substrate surface. This may
require modifications to the sampling methodology due to the potential for hang-up of
the polyester sampling material on wood splinters associated with penetrations.
The performance of coatings on wood of different dimensions and on wood members
with original vertical or angled orientations could also be investigated. Horizontally-
oriented surfaces, as utilized in the study, receive more overall direct solar exposure
and more indirect UV radiation than vertically-oriented surfaces, like fences. In
Williams, et al. (2001 c), the wood erosion rate for horizontally-oriented boards was
measured to be two to three times more rapid than for vertically-oriented boards.
4.6 Abrasion Effects
The effects of abrasion (e.g., by repeated contact and walking) have not been directly
tested in this project. However, an attempt was made to derive some indication of its
impacts using the data collected under the assumption that repeated wipe sampling
was a source of abrasion.
Although the experiment was not designed to provide information about the effects of
wipe frequency and the number of previous wipe samples taken from a given sampling
area, the data collected provided a limited opportunity to investigate these factors. Our
analyses were inconclusive and thus we do not report them here. However, we do
discuss the rational underlying our attempts at understanding the abrasion issue as it
may prove useful should future studies be designed to address abrasion directly.
It was postulated the wipe sampling method itself could have one or two (or both)
unintended effects on DCCA. The first is a cleansing effect of removing DCCA from the
surface, leaving less DCCA available for the next sampling event. Presumably this
effect would be more noticeable when the time between wiping and subsequent wipe-
sampling measurement was short. The second is an abrasion effect whereby the
rubbing of the wipe against the sampling surface might have a substantial abrading
120
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
effect, thereby adversely affecting the coating surface and its ability to block DCCA
migration to the surface. Presumably such an effect might increase the amount of
DCCA available by compromising the coating and exposing fresh CCA-treated wood
fibers.
We attempted to tease these possible effects of rewiping from the data by constructing
two predictive variables defined for the purpose of quantifying the amount of previous
wiping and the elapsed time since the previous wipe:
NOPW = number of previous wipes
TSPW = time (months) since the previous wipe
NOPW is a surrogate for total prior post-coat abrasion; and TSPW measures wipe
frequency, or more specifically, the time interval between wipe samples on a given
sampling area. Possible effects of these two factors were investigated using the
combined dataset [i.e., both the primary sampling area (M) and baseline sampling area
(BL) DCCA results], though it must be noted that resampling of the BL areas were only
done through 11 months post-coating. The analysis strategy was to include these
variables as linear predictors in statistical models that incorporated all of the main-
study factors, with statistically significant upward or downward trends for these
variables shedding light the issues of abrasion and cleansing.
This exercise revealed no evidence of trends with TSPW, but weak evidence of a
downward trend with NOPW. That is, the greater the number of previous wipes, the
lower the DCCA level. The direction of the trend is opposite what would be expected if
wiping had abraded coatings and reduced their effectiveness, but consistent with the
hypothesis of a cleansing effect. In light of the weak effect, and possible confounding,
we view this finding with a healthy dose of skepticism. Because the experiment was not
designed for assessing abrasion and wipe frequency, there is not much relevant
information in the data, and it is difficult to claim with any certainty that the method of
analysis we used is the best way to mine the available data to teasing out abrasion and
rewipe information (from each other and from the other effects).
The effect of abrasion on DCCA as well as on coating efficacy for reducing DCCA does
not appear to have been directly addressed in previously reported studies but should
be addressed in future study efforts. It is important to note that there are many
variables that may potentially confound the effects of abrasion, including UV
degradation, surface preparation, moisture content of the wood surface, history of
coatings on the surface and wood characteristics.
Since purposeful abrasion (e.g., simulated walking) was not a factor in our study, the
results likely represent the upper limits of DCCA reduction performance of the tested
coating projects. As such, the coatings industry should adapt the test protocol that has
been developed to include an abrasion component. Related to abrasion, the transfer of
121
-------
CCA analytes via feet, pets, and other potential contact routes may be important but
has not been addressed by this study.
4.7 CCA Transport from Treated Specimens
In addition to providing some buffer space between test specimens, the untreated
boards separating the CCA-treated test specimens on each deck served as control
surfaces for assessing cross-contamination potential. During each wipe sampling
event, one untreated (but coated, for minidecks 1 -12) specimen from each minideck
was wipe-sampled. Because there were five untreated specimens on each minideck,
there were a total often potential sampling areas. The specific areas sampled during
each routine sampling event were randomly selected for each minideck but were kept
different for each sampling event. The results of these samples were used to assess
the potential for cross-contamination between adjacent sampling areas. Cross-
contamination could hypothetically be the result of splash-over of rainwater from one
specimen to the next or possibly other transfer mechanisms.
Cross-contamination potential was assessed by comparing the cross-contamination
results for the thirteen minidecks (the uncoated control boards) versus the results from
the non-CCA control minideck. These results are summarized in Table 4-6, while the
full non-CCA minideck dataset is provided in Appendix L and the full non-CCA "spacer"
board contamination control dataset is provided in Appendix M. The cross-
contamination data for all of the coatings are summarized in Appendix C (Table C-7).
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Table 4-6. Summary of Cross-Contamination and Blank Control Minideck Results
Non-CCA Minideck
Average
Std. Dev.
DAs (ug/cm2)
0.001
0.0006
DCr (ug/cm2)
0.003
0.004
DCu (ug/cm2)
0.034
0.011
Non-CCA Boards on Minidecks
Average
Std. Dev.
0.007
0.021
0.009
0.028
0.056
0.042
As summarized in Table 4-6, DCCA on the control boards was insignificant in relation
to the DCCA measurements used in the coating efficacy data analysis. While the DAs
results of all of the untreated wood wipe samples are, with a few exceptions early in the
study, very low in comparison with those from the treated wood specimens, on average
there appears to be more DCCA from the cross-contamination control surfaces versus
the non-CCA (negative control) minidecks.
To further analyze the cross-contamination control results, two series of plots were
developed: Figure 4-2 and 4-3.
122
-------
In Figure 4-2, thirteen plots and presented, one for each coating, with Coatings 1, 2
and 3 on the top row, 4, 5 and 6 on the next row, and so forth. The horizontal and
vertical axes are In(DAs) in all plots, where the x-axis is the average of DAs over all
CCA boards for the minidecks associated with that coating at each of the seven time
periods (denoted by the number next to the data point, where time periods 1 ,...,7
correspond to sampling events at 1, 3, 7, 11,15, 20, and 24 months after coating
respectively) and the y-axis is DAs averaged over all non-CCA boards for that coating
at the corresponding time period. The superimposed line is from the ordinary least fit.
This plot shows that as DAs increased overtime on the CCA boards, so too the DAs
increased on non-CCA "spacer" boards.
-'.T
.7
ff^'
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
Figure 4-2. Average Spacer Board DAs versus Average CCA board DAs at Each of Seven
Time Intervals, plotted by Coating.
In Figure 4-3, seven plots and presented, one for each time period, with sampling
events at month 1, 3 and 7 represented on the top row, 11,15 and 20 on the next row,
and 24 on the bottom. The horizontal and vertical axes are In(DAs) in all plots, where
the x-axis is the average of DAs over all CCA boards for the minidecks associated with
each coating (denoted by the number next to the data point, where coatings 1 through
9 are identified directly and coatings 10 through 13 are denoted by the letters A, B, C,
and D respectively) at the indicated time period and the y-axis is DAs averaged over all
123
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
non-CCA boards for the corresponding coating at the given time period. The
superimposed line is from the ordinary least fit. This plot shows that DAs on non-CCA
boards was higher when those boards neighbored CCA with higher DAs.
Figure 4-3. Average Spacer Board DAs versus Average CCA board DAs for Each of
Thirteen Coatings, plotted by Time Inteval (note that letter A = Coating 10, B
= Coating 11, C = Coating 12, and D = Coating 13).
4.8 Sampling Methodology
Various aspects of the wipe sampling method could be studied to improve
understanding of sampling factors that may impact DCCA measurements and for
refining of the method. These include the following:
• Surface moisture: a measure of surface moisture would ideally be made prior to
taking each wipe sample, although robust, reliable and non-destructive moisture
measurement protocols would need to be developed.
• Wipe stroke and snagging: the impacts of variable sampling stroke durations and
temporary pauses in sampling due to the wipe material catching on surface
irregularities could be studied in more detail to ascertain whether these issues
affect DCCA and, if so, how those effects might be mitigated.
124
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
• Splinter removal: The impacts of removing splinters larger than a grain of rice (as
specified in the sampling protocol) on DCCA measurements could be
investigated, particularly from a risk perspective. That is, do larger particles get
ingested and if so, are they biologically available and/or active.
• Soluble and particulate DCCA: The sampling method does not differentiate
between particulate and soluble DCCA fractions unless collected particles are the
size of a grain of rice, in which case they are removed from the wipe prior to
extraction/digestion and analysis. Although no protocols currently exist fordoing
so, soluble and particulate DCCA could be separately quantified yielding data to
inform the risk assessment and potentially also the manner in which exposure
risks can be best mitigated.
• CCA analytes and speciation: The speciation of CCA analytes could be an
important determinant of exposure risks. Only total arsenic, total chromium, and
total copper were routinely measured in this study, due to resource limitations, as
speciating CCA analytes is significantly more complex and costly.
• Cross-contamination controls: The results of analyses of untreated specimens
between each treated CCA specimen on the minidecks showed that DAs from the
untreated boards generally increased overtime, but still represented only
fractions of the measured DAs on treated (coated and uncoated) boards. As
such, it may not be necessary to rigorously sample these cross-contamination
controls, but they should still be provided as buffers between boards in future
studies.
4.8.1 Baseline (Pre-Coat) Measurement
A significant logistical issue arises as a result of the sampling process itself and the fact
that the wipe sampling technique may affect the surface of the wood in at least two
ways: by removing the CCA on the surface of the wood, and by potentially abrading the
wood or its coating (as described in Section 4.6).
Ideally, given the relative variability in DCCA between samples taken from areas along
the length of treated wood boards, initial surface wipe samples would be taken from
each sampling area to be further tested. However, several studies suggest that, as
could reasonably be expected, the act of wipe sampling the surfaces of CCA-treated
wood removes a considerable amount of the DCCA from a test specimen (CPSC staff
2003a; Stilwell et al., 2003). Furthermore, wipe sampling may cause some abrasion
which is suspected to be a significant variable in determining both uncoated DCCA as
well as durability and efficacy of tested coatings.
Because of these potential complications and their implication for affecting the results
of subsequent samples, the PSAs (i.e., those areas that were wipe-sampled during
125
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
each post-coat sampling event) were not sampled prior to coating. Instead, sampling
areas on the minidecks adjacent to the PSAs (termed "baseline areas" because they
were used to calculate specimen-specific baseline DCCA values) were sampled prior
to coating. Individual specimen-specific baseline values of DCCA were then
determined for each PSA to be coated and tested by averaging the DCCA from the two
adjacent wipe sampling areas on either side of the PSA.
This method of baseline determination was selected in part because existing data
suggested that intraboard (within-board) variability was relatively low. Research by
Stilwell et al. (2003) showed an intraboard variability (RSD) of 17% versus an
interboard (between-board) variability of 39%. It was further assumed that the
variability would be even less for sampling areas that were closer to one another
(intraboard, interspecimen variability between adjacent sampling areas) than for
sampling areas that were further apart, although this assumption was not supported by
the data gathered for this project.
Consideration was given to alternate methods of assessing pre-coat DCCA
concentrations of the PSAs and it is worth mentioning them here.
• Wipe sampling the PSAs prior to coating and then waiting or even exposing the
test specimen to weathering to induce more migration of DCCA to the surface of
the specimen priorto coating. While this concept may be sound, the existing data
supporting the design of such a method is fairly limited (Stilwell et al., 2003). That
is, it had not yet been well-established how much time must elapse or under
what conditions specimens must be maintained to ensure that surficial CCA
analyte concentrations have rebounded to pre-wipe conditions. This may vary
with climate, amount of abrasion, and other factors.
• Taking an average of a number of initial DCCA measurements taken from
sampling areas that would then be discarded and not used for the study in any
other way. In such an approach, the baseline values used would not be unique to
a specific PSA. Instead, a single baseline DCCA value might be used for
numerous PSAs or even for all of the PSAs in the study. However, it was thought
that this option might not provide the level of data resolution and statistical power
required to adequately assess coating efficacy for this project.
• Wipe sampling the undersides of the test specimens to establish the baseline
DCCA for each specimen. This was seen as a potentially good option for new
CCA-wood specimens, but not for aged specimens, as their top faces are well
defined and of much greater interest (given that these faces are the ones to
which users are most likely to be exposed) than their bottom faces. The top faces
of aged CCA wood specimens would be expected to have considerably different
characteristics than their bottom faces, because, for example, they may have
been exposed to direct sunlight, abrasion, and so forth, while their bottom faces
126
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
have not. While the same is not necessarily true of new CCA wood, CPSC staff
data suggests that sample variability along the length of a given board is less
than the variability between the top and bottom faces of a specimen, even for
new CCA-treated lumber (CPSC staff 2003a). As such, and as previously
indicated, the weathering test employed a method whereby the DCCA of
adjacent sampling areas were averaged in order to establish unique baseline
DCCA values for each individual PSA.
4.8.2 Wipe Sampling Method Comparisons
As described in Section 2.2.2, baseline measurements were made using acid-washed
sampling wipes and termed the "A2" sampling method, as described in Section 2.8.3.1.
The "2X" sampling method utilized unwashed wipes and was the method used for
subsequent sampling events; it is described in Section 2.8.3.2. Side experiments were
conducted to determine factors for converting results between the two methods. The
results are presented in Section 2.2.2 and the complete report on this series of
experiments is provided as Appendix A.
4.9 Potential Future Studies
Future studies could be directed to address any of the issues highlighted in this
section; certainly any future studies should at least consider these issues.
While there are many outstanding methodological issues, the results of the primary
experiment described in this report indicate that method is fairly robust (the results are
reproducible across many variables that were addressed in the protocol). With this in
mind, several avenues for immediate future studies are suggested:
• An abrasion element should be developed for use in future coating DCCA
mitigation efficacy studies.
• The study scope should be expanded to improve representation across: type and
condition of wood used, geographic and climatic regions tested, and coating
selection (preferably in coordination with manufacturers to specifically develop
products for reducing DCCA).
• Surface preparation, including the effects of periodic cleaning or other mitigation
strategies versus coating, and coating reapplication issues should be studied
specifically. These issues are important both with respect to the performance of
coatings overtime as well as to address potential secondary issues that may
impact mitigation strategies.
Given the expectation that coating performance may vary significantly with climatic
region, surface preparation, and use pattern, the cost-effectiveness of future studies
127
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
will be enhanced by testing products specifically developed by the coatings industry to
reduce DCCA with consideration given to these critical factors.
128
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
5. Quality Assurance and Quality Control
5.1 Assessing DQI Goals
This section summarizes the assessment of data quality goals. Full QA/QC summary
reports for each dataset are included in Appendix R. The critical measurements for the
outdoor weathering tests were total arsenic, total chromium, and total copper
concentrations (in wipe sample extraction fluid). Data quality indicator (DQI) goals for
concentration in terms of accuracy, precision, and completeness, as established in the
QAPP for this project, are shown in Table 5-1.
Table 5-1. Data Quality Indicator Goals for Critical Measurements
Analyte
Arsenic (total)
Chromium (total)
Copper (total)
Method
SW-846 Method 6020 (modified)
SW-846 Method 6020 (modified)
SW-846 Method 6020 (modified)
Accuracy
(Percent
Recovery)
90-110
90-110
90-110
Precision
(Percent
RSD/RPD)
10
10
10
Completeness
(Percent)
90
90
90
Note: EPA Method 200.8 substituted for Method 6020. See Section 1.5.1 below for
explanation.
5.1.1 Summary of Accuracy and Precision Measurement Results
Over the course of the outdoor weathering study, fifty spike samples at concentrations
of 1, 50,1,000 and 10,000 ug/L were submitted to the analytical laboratory. One
sample was disqualified due to spiking error that occurred prior to submission.
Completion rates for precision and accuracy were less than the 90% completion
targets for measures of precision and accuracy. However, as discussed below, review
of the data quality indicators clearly demonstrates that the data is of sufficient quality to
meet the purposes of the study. Ninety-three percent (93%) of the data meet a
precision criteria of ±15% RPD for duplicate sample extracts with reported
concentrations of 1 ug/L and above, and for spike levels >1 ug/L, ninety-five percent
(95%) of the spiked extracts met accuracy recovery targets of 100±15%. Completion
rates of fifty-eight percent (58%) for an accuracy target of 100±15% at the 1 ug/L spike
level with average recovery of 114%, indicates that the method is less reliable in
quantification of very small levels of DCCA. The QA results indicate a need for
inclusion of spikes at additional levels between 1 and 50 ug/L in future studies to better
define the ability of the method to quantify very low levels of DCCA.
Wipe extracts concentrations of <10 ug/L As, Cr, or Cu correspond to DCCA levels of <
0.0032 ug/cm2. For this study, the wipe extracts from 30 primary sampling area (PSA)
samples (about 2.7% of the PSA datapoints), all in the first three time intervals, through
7 months of weathering fell into the concentration range of <10 ug/L. Other samples
129
-------
that fell into the region of increased analytical uncertainty included samples collected
from the untreated control minideck (minidecks BC and NC) samples, samples from
the untreated "buffer" or "cross-contamination control" boards that were placed
between treated boards on the minidecks, and several BL samples (samples used to
assess effects of frequency and number of previous wiping events). The increased
measurement uncertainty observed at very low levels of DCCA do not impact
interpretation of coatings performance.
5.1.2 Precision
Precision of analysis of split wipe extract samples, is expressed as the RPD of the
duplicated measurement and is calculated using Equation 5.1, where Y1 is the
concentration of the first sample and Y2 is the concentration of the duplicate sample.
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
7
(Equation 5.1)
One hundred forty-five blind field duplicates (wipe samples split following extraction) or
7% of the total number of samples collected, were delivered to the laboratory for
analysis. Results for three of the split samples were reported by the analytical
laboratory as <0.5 or <1.0 ug/L As and Cr. Therefore, numbers of duplicate pairs
available for evaluation of precision are 142,143, and 145 for As, Cr, and Cu,
respectively. For the majority of samples, agreement between field duplicates was
very good. In only one set of duplicate samples was RPD >50% for each analyte; this
was the only sample set where the data was qualified as estimated "J" due to the RPD.
Achieved completeness was >80% for all three analytes, which did not meet the DQI
goal of 90% established in the QAPP. However, >90% of the duplicate samples met a
precision goal of ±15% RPD and this may be more realistic and adequate expectation
of precision for these types of samples,
5.1.3 Accuracy and Bias
The accuracy of the measurement is expressed in terms of recovery of a known spike.
Recovery is calculated by Equation 5.2, where R is the measured concentration and C
is the spiked (i.e., target) concentration.
r> /">
Recovery = x 100
(Equation 5.2)
Spiked samples at a minimum of three of four possible concentration levels (10,000
ug/L, 1,000 ug/L, 50 ug/L and 1 ug/L) were analyzed by the laboratory along with the
130
-------
samples from each sampling event. Overall mean recovery of As was 104% for forty-
nine reported spike samples averaged across all spike concentration levels.
Completion rates at 100±10% were below the 90% target completion rate, primarily
due to low completion rates at the lowest spike level of 1 |jg/L. Completion rates of
fifty-eight percent (58%) for an accuracy target of 100±15% at the 1 ug/L spike level
with average recovery of 114%, indicates decreased reliability of the overall method
in quantification of very small levels of DCCA. The QA data indicate a need for
inclusion of spikes at additional levels between 1 and 50 ug/L in future studies to
better define the ability of the method to quantify very low levels of DCCA.
5.2 Completeness
The ratio of the number of valid data points taken to the total number of data points
planned is defined as data completeness. Achieved completeness was >80 percent for
all three analytes, but did not meet the DQI goal of 90 percent originally established in
the QAPP. Results suggest that the DQI goal of ±10 percent for precision between
duplicates may be too ambitious. There was no acceptance criteria given in the
analytical method for agreement between duplicate samples and, based on the results
of this project, it appears that a DQI goal of ±15 percent RPD may be more realistic for
these types of samples. If DQI goals of ±15 percent RPD, completeness would have
been greater than 90 percent for all three analytes. Also, the analytical method cites
acceptance criteria for recovery of spiked blanks as 85-115 percent which is slightly
higher than the DQI goal of 90-110 percent. Using the analytical method criteria,
completeness of accuracy results would improve. DQI goals should be reviewed and
revised as appropriate for future studies.
5.2 Data Validation Summary
The analytical laboratory was required to submit calibration and QC data along with
each data package. All data packages received by ARCADIS/EPAwere internally
validated by a qualified staff scientist according to the QA/QC criteria set forth in the
U.S. EPA Contract Laboratory Program National Functional Guidelines for Inorganic
Data Review, July 2002 (NFG). When parameters called out in the NFG were different
from those established in the QAPP or the analytical method (EPA Method 200.8), the
more stringent criteria were used. Validation reports were prepared for each sample
delivery group and the reported data were qualified as appropriate. These reports are
also included in Appendix R. All of the data collected were used in the analysis of
results (i.e., no datapoints were disqualified).
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
131
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
This page intentionally left blank
132
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
6. Conclusions
This study evaluated the effect of coatings on DCCA on the surfaces of CCA treated
Southern Yellow Pine.
Each coating, as well as the positive (uncoated) controls, exhibited a significant
decrease in DCCA between baseline (prerinse and precoat) measurements and
samples taken 1 month after coating, demonstrating that a simple rinsing step reduces
DCCA, although it should be noted that evaluating the impact of rinsing or surface
preparation prior to coating was not an objective of the project and not explicitly
factored into the experimental design. The coated minidecks all had lower DCCA than
the positive controls, which indicates that coating (using any of the coatings tested)
reduces DCCA to some degree, although not always at a statistically significant level,
and that the reduction due to coating is greater than that attributable to rinsing. Over
the course of the two-year study, DCCA increased with time after coating, although
relatively small decreases in DCCA were sometimes observed between sampling
intervals, presumably where the sum of removal processes (previous wipe sampling
events, rinsing due to rain, etc.) had a greater impact on DCCA than weathering and
other factors that may increase DCCA.
The coatings that were tested were compared against one another and uncoated CCA-
treated positive controls based on their performance in reducing DAs. Several coatings
exhibited DAs reductions estimated to be between 75 and 100 percent (lower and
upper 95 percent Cl) compared with the positive control minidecks for through at least
the first year of weathering. These relatively high-performing products all utilized
multiple coats and included two penetrating products: product #3, an oil-based stain
with acrylic and product #8, a water-based stain with acrylic; and three film-forming
products: the two paints (products #9 and 10) and product #11, a vinyl elastic coating
marketed to reduce DCCA.
The efficacy of each product decreased over the study period and by 24 months after
coating, only two of the products (products #9 and 10, the paints) had DAs levels that
were significantly different than the positive controls. Additionally, the reductions of DAs
for one product (#5), a water-based sealant, were generally not statistically different
from the positive controls over the course of the study.
While exploratory statistical tests indicated that film-forming products performed
significantly better than non-film-formers in these tests, elevated arsenic concentrations
were measured in paint chips recovered from the painted CCA-treated minidecks (note
that CCA treated wood fibers may be associated with the tested paint chips). This
finding is also considered exploratory and warrants further investigation. The
occupational exposures due to sanding and milling CCA-treated wood, as reported by
Decker et al., (2002), and Nygren et al., (1992), underscore the need for research that
evaluates potential exposures to CCA chemicals due to preparation and maintenance
133
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
activities; for instance, particle generation due to scraping and sanding of CCA-treated
wood to remove cracked, chipped, or weathered coatings.
The protocol utilized in this investigation yielded results that provide a coherent picture
of the changes in DCCA as coated CCA wood weathers and, with appropriate
development or modification, may serve as a useful tool for the coatings industry or
product testing laboratories in efforts to develop new coatings and to measure coating
performance for reducing dislodgeable CCA. Specific findings that may inform future
use of the protocol include:
• There were significant differences in DCCA (pre- and post-coat) between the two
different source decks. That is, source deck is an important variable. Grain
orientation, however, was only a marginally significant variable in this study.
• The effects of abrasion resulting from the wipe sampling method used for this
study appear to be negligible, thus avoiding potential complications, or false
positive interferences, as a result of the sampling methodology. However, rewipe
effect - that is, the reduction in DCCA by "cleaning" the surface by sampling - may
be significant.
• As with any pilot experiment, there were many potentially important variables that
this study did not address, including: impact of stresses associated with full-scale
decks, impact of climatic regions, abrasion, multiple coats of the same product,
surface preparation procedures, and periodic recoating. Abrasion (e.g., resulting
from foot traffic on in-service deck surfaces), in particular, is considered likely to be
an important coating performance factor.
Taken as a whole, the results of this study suggest that typical deck coating products
(sealants and stains) need to be re-applied periodically in order to maintain significant
levels of DAs mitigation. However, this study did not examine the effect on DAs of
recoating after a period of weathering, and the results of this study may not be
indicative of the reduction in DAs that may be achieved with periodic re-treatment.
However, minor adaptation of the methodology should enable evaluation of this and
other important variables.
The differences in efficacy observed among products in reducing DCCA suggests that
products could be tailored to reduce DCCA. Broad-based testing of products may be
needed to empower consumers to make informed choices. The paint chip results also
indicate a need to better understand and characterize the potential impacts of recoat
preparation steps, particularly for film forming products.
134
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
7. References
CPSC staff 2003(a). "Statistical Analyses of CCA-Treated Wood Study Phases I and
II." Memorandum from Mark S. Levenson to Susan Ahmed and Russell Roegner,
dated January 10, 2003. Consumer Products Safety Commission.
CPSC staff. 2003(b). Memorandum from David Cobb to Patricia Bittner, "CCA-
Pressure Treated Wood Analysis - Exploratory Studies Phase I and Laboratory
Studies Phase II," in the Briefing Package, "Petition to Ban Chromated Copper
Arsenate (CCA)-Treated Wood in Playground Equipment (Petition HP 01-3), U.S.
Consumer Product Safety Commission, Washington, D.C., February 4, 2003. pp 229.
http://www.cpsc.gov/LIBRARY/FOIA/FOIA03/brief/briefing.html. April 27, 2005.
Decker, P., Cohen, B., Butala, J.H., and Gordon, T. Exposure to wood dust and heavy
metals in workers using CCA pressure-treated wood. AIHA J, 2002, No. 63 (2): 166-
171.
Hatlelid, K. et al., 2004. Exposure & Risk Assessment for Arsenic from CCA-Treated
Wood Playground Equipment. Journal of Children's Health, Vol 2, Nos. 3-4, pp. 215-
241.
Kwon, E., Zhang, H., Wang, Z., Jhangri, G.S., Lu, X,. Fok, N., Gabos, S., Li, X., and Le,
X.C. Arsenic on the Hands of Children after Playing in Playgrounds. Environmental
Health Perspectives On-line, Online 17 June, 2004, doi:10.1289/ehp.7197 (available at
http://dx.doi.org/).
Lebow, S. 1996. "Leaching of Wood Preservative Components and Their Mobility in
the Environment. Summary of Pertinent Literature." Forest Products Laboratory, U.S.
Department of Agriculture. General Technical Report FPL-GTR-93. August 1996.
Lebow, S., R.S. Williams, P. Lebow. 2003. "Effect of Simulated Rainfall and
Weathering on the Release of Preservative Elements from CCA Treated Wood."
Environ. Sci. Technol. Vol. 37, No. 18, p. 4077-4082
Lebow, Stan. USDA Forest Service, Forest Products Laboratory, Madison, Wisconsin.
Personal communication, 2006.
Levenson, M. et al., 2004. A Field Study of Dislodgeable Arsenic from CCA-Treated
Wood Using Human-Hand and Surrogate Wipes. Journal of Children's Health, Vol 2,
Nos. 3-4, pp. 197-213.
Nygren, O., Nilsson, C. A., and Lindahl, R. Occupational exposure to chromium,
copper and arsenic during work with impregnated wood in joinery shops. Ann. Occup.
Hyg. 1992, No. 36 (5): 509-517.
135
-------
Evaluation of the Effectiveness of
Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper
from CCA Treated Wood
December 2008
SAP Minutes No. 2007-02. EPA's Science Inventory: www.epa.gov/si/, Record 150970
SPC. 2006. "What is Pressure-Treated Lumber?"
http://www.southernpine.com/whatisptlumber.shtml
Stilwell, D., M. Toner, B. Sawhney. 2003. "Dislodgeable Copper, Chromium and
Arsenic from CCA-Treated Wood Surfaces." The Science of the Total Environment.
312(2003). 123-131.
Thomas, T. et al., 2004. The Development of a Standard Hand Method & Correlated
Surrogate Method for Sampling CCA-Treated Wood Surfaces for Chemical Residue.
Journal of Children's Health, Vol2, Nos. 3-4, pp. 181-196.
U.S. EPA. 2003. "Evaluation of the Effectiveness of Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper from CCA Treated Wood. Revision 6." U.S. EPA,
Research Triangle Park, NC, September 24, 2003.
U.S. EPA. 2005. "Evaluation of the Effectiveness of Coatings in Reducing Dislodgeable
Arsenic, Chromium, and Copper from CCA Treated Wood - Interim Data Report." EPA
Report: EPA/600/R-05/050; U.S. EPA, Research Triangle Park, NC, May 9, 2005.
Ursitti, F., Vanderlinden, L, Watson, R., Campbell, M. Assessing and managing
exposure from arsenic in CCA-treated wood play structures. Canadian Journal of
Public Health, 2004, 95, 429-433.
Williams, R.S. and Feist, W.C. Finishing Wood Decks. Wood Design Focus, Volume 4,
Numbers. U.S.D.A. Forest Products Laboratory, 1993.
Williams, R.S., Knaebe, M.T., Evans, J.W., Feist, W.C. Erosion rates of wood during
natural weathering. Part III. Effect of exposure angle on erosion rate. Wood Fiber Sci.
2001 (c), 33(1), 50-57.
Williams, R.S., Knaebe, M.T., Feist, W.C. Erosion rates of wood during natural
weathering. Part II. Earlywood and latewood erosion rates. Wood Fiber Sci. 2001 (b),
33(1), 43-49.
Williams, R.S., Knaebe, M.T., Sotos, P.G., Feist, W.C. Erosion rates of wood during
natural weathering. Part I. Effects of species, grain angle and surface roughness.
Wood Fiber Sci. 2001 (a), 33(1), 31 -42.
Zartarian, V.G., Xue, J., Ozkaynak, H., Dang, W., Graham, G, Smith L., and Stallings,
C. A Probabilistic Arsenic Exposure Assessment for Children Who Contact CCA-
Treated Playsets and Decks, Part I: Model Methodology, Variability Results, and Model
Evaluation. Risk Analysis, 2006, 26, No. 2, 515-531.
136
------- |