United States
Environmental Protection
Agency
Region 5
Great Lakes National EPA-905/4 88-002
Program Office GLNPO Report No. 2
230 South Dearborn Street February 1988
Chicago, Illinois 60604
Great Lakes
Atmospheric Deposition
(GLAD) Network,
1982 and 1983
JMLJH*
-------
EPA-905/4-88-002
GLNPO Report No. 2
February 1988
Great Lakes Atmospheric Deposition
(GLAD) Network, 1982 and 1983:
Data Analysis and Interpretation
Donald F. Gatz
Van C. Bowersox
Jack Su
Gary J. Stensland
Atmospheric Chemistry Section
Illinois State Water Survey
Champaign, Illinois 61820
Edward Klappenbach
Project Officer
Great Lakes National Program Office
Environmental Protection Agency
230 South Dearborn Street
Chicago, Illinois 60604
- ;:nv;rr..., . ital Protection Agency
-'' " ..Election (PL-12J)
, / ,,-;,(. jdCK.son Boulevard,
Chicago, IL 60604-3590
-------
TABLE OF CONTENTS
Table of Contents i
List of Figures ii
List of Tables v
1. INTRODUCTION 1
1.1. Background 1
1.2. Purpose 1
1.3. Scope 2
2. METHODS 2
2.1. Sampling and Analysis 2
2.2. Data Screening Criteria 6
2.3. Data Quality Checks 7
2.4. Calculation of Integrated Ion Concentrations 8
2.5. Metal Concentrations 9
2.6. Paired Site Comparisons 9
2.7. Deposition Calculations 9
3. RESULTS AND DISCUSSION 13
3.1. Data Quality Checks 13
3.2. Concentration Spatial Distributions 16
3.2.1. Effect of Adding GLAD Sites to the NADP Data Set . 16
3.2.2. Additional Ions, Combined Data Set 24
3.2.3. Paired Site Comparisons: GLAD vs NADP 24
3.3. Precipitation Amount 34
3.4. Deposition 35
3.4.1. Spatial Patterns 35
3.4.2. Effects on Deposition Estimates of Closing GLAD Sites . . 44
3.4.3. Atmospheric Loadings to the Lakes and Comparison to
Previous Estimates 50
3.4.3.1. Results 50
3.4.3.2. Discussion 57
4. CONCLUSIONS 60
5. RECOMMENDATIONS FOR FURTHER RESEARCH 62
6. ACKNOWLEDGEMENTS 63
7. REFERENCES 64
APPENDIX A 68
-------
LIST OF FIGURES
Figure Page
number number
1. Network map for GLAD and NADP/NTN sampling sites 5
2. Grid used for over-lake deposition flux calculations 5
3. National Weather Service observer network used for 30-yr mean
annual precipitation distribution 12
4. Frequency distributions of ion percent difference (IPD)
for GLAD and NADP/NTN networks, 1982-1983 14
5. Frequency distributions of conductance percent difference (CPD)
for GLAD and NADP/NTN networks, 1982-1983 14
6(a). Scatterplot of measured pH and conductance for the GLAD network,
1982-1983 17
6(b). Scatterplot of measured pH and conductance for the NADP/NTN
network, 1982-1983 17
7(a). Scatterplot of calculated pH vs measured conductance for the
GLAD network, 1982-1983 18
7(b). Scatterplot of calculated pH vs measured conductance NADP/NTN
network, 1982-1983 18
8(a). Spatial distribution of volume-weighted S04 concentrations
in the Great Lakes region, using NADP data for 1982-83 20
8(b). Same as (a), but using the combined GLAD-NADP data set 20
9(a). Spatial distribution of volume-weighted Ca concentrations
in the Great Lakes region, using NADP data for 1982-83 21
9(b). Same as (a), but using the combined GLAD-NADP data set 21
10(a). Spatial distribution of volume-weighted N03 concentrations
in the Great Lakes region, using NADP data for 1982-83 22
10(b). Same as (a), but using the combined GLAD-NADP data set 22
ll(a). Spatial distribution of volume-weighted NH4 concentrations
in the Great Lakes region, using NADP data for 1982-83 23
ll(b). Same as (a), but using the combined GLAD-NADP data set 23
11
-------
Figure Page
number number
12. Spatial distribution of volume-weighted Na concentrations
in the Great Lakes region, using GLAD-NADP data, 1982-83 25
13. Spatial distribution of volume-weighted Cl concentrations
in the Great Lakes region, using GLAD-NADP data, 1982-83 25
14. Spatial distribution of volume-weighted Mg concentrations
in the Great Lakes region, using GLAD-NADP data, 1982-83 26
15. Spatial distribution of volume-weighted K concentrations
in the Great Lakes region, using GLAD-NADP data, 1982-83 26
16. Map showing locations of GLAD--NADP/NTN site pairs for
which ion concentrations were compared 27
17. Paired site comparisons of 864 concentration percentiles,
using box diagrams 28
18. Paired site comparisons of Ca concentration percentiles,
using box diagrams 29
19. Paired site comparisons of N03 concentration percentiles,
using box diagrams 30
20. Paired site comparisons of NH4 concentration percentiles,
using box diagrams 31
21. Distribution of 30-yr mean annual precipitation over the
Great Lakes 36
22. Spatial distribution of annual deposition fluxes of 864
over the Great Lakes, using GLAD-NADP data, 1982-83 37
23. Spatial distribution of annual deposition fluxes of Ca
over the Great Lakes, using GLAD-NADP data, 1982-83 37
24. Spatial distribution of annual deposition fluxes of NC>3
over the Great Lakes, using GLAD-NADP data, 1982-83 38
25. Spatial distribution of annual deposition fluxes of NH4
over the Great Lakes, using GLAD-NADP data, 1982-83 38
26. Spatial distribution of annual deposition fluxes of Na
over the Great Lakes, using GLAD-NADP data, 1982-83 39
27. Spatial distribution of annual deposition fluxes of Cl
over the Great Lakes, using GLAD-NADP data, 1982-83 39
iii
-------
Figure FaSe
number number
28. Spatial distribution of annual deposition fluxes of Mg
over the Great Lakes, using GLAD-NADP data, 1982-83 40
29. Spatial distribution of annual deposition fluxes of K
over the Great Lakes, using GLAD-NADP data, 1982-83 40
30. Spatial distribution of annual deposition fluxes of Cd
over the Great Lakes, using GLAD-NADP data, 1983 41
31. Spatial distribution of annual deposition fluxes of Pb
over the Great Lakes, using GLAD-NADP data, 1983 41
32. Spatial distribution of percent differences that result from
calculation of the 864 deposition flux from all the sites
compared to those remaining open after January 1986 46
33. Spatial distribution of percent differences that result from
calculation of the Ca deposition flux from all the sites
compared to those remaining open after January 1986 46
34. Spatial distribution of percent differences that result from
calculation of the NC>3 deposition flux from all the sites
compared to those remaining open after January 1986 47
35. Spatial distribution of percent differences that result from
calculation of the NH4 deposition flux from all the sites
compared to those remaining open after January 1986 47
iv
-------
LIST OF TABLES
Table Page
number number
1. List of GLAD Network Sites 3
2. List of NADP Network Sites in the Great Lakes Region 4
3. List of GLAD Sites Excluded from Spatial Analysis of
Concentration because of Data Completeness Criteria 7
4. Summary of GLAD Toxic Metal Measurements for 1982 and 1983 10
5. Comparison of Grid-Box Integrated Lake Areas with
Literature Values 11
6. Summary of Ion Percent Difference Results for GLAD and NADP
Sites (1982-1983) 15
7. Summary of Conductance Comparison Results for GLAD and NADP
Sites (1982-1983) 15
8. Key to GLAD-NADP Paired Site Comparisons in Figures 17-20 32
9. Summary of Precipitation FLuxes to the Great Lakes,
1951-1980 42
10. Comparisons of Wet-Only Atmospheric Loadings to the Great Lakes
from 1) All Valid NADP and GLAD Sites Operating in 1981-82,
with 2) All Valid NADP Sites Plus Valid GLAD Sites Remaining
Open after January 1986 48
11. Effect of Closing Additional GLAD Sites, Compared (Percent
Change) to the Network Remaining after January 1986 49
12. Atmospheric Loadings to the Great Lakes from Combined NADP
and GLAD Data, and Comparisons with Previous Estimates 51
13. Great Lakes Loadings Estimates for Pb and Cd, Based on GLAD 1983
Concentration Measurements and 30-yr Mean Precipitation, Compared
with Previous Estimates 55
14. Comparison of current estimates of wet-only lake loadings
with previous estimates based on modeling and bulk sampling 56
-------
GREAT LAKES ATMOSPHERIC DEPOSITION (GLAD) NETWORK, 1982 AND 1983:
DATA ANALYSIS AND INTERPRETATION
Donald F. Gatz, Van C. Bowersox, Jack Su,
and Gary J. Stensland
Atmospheric Chemistry Section
Illinois State Water Survey
Champaign 61820
1. INTRODUCTION
1.1. Background
Under the Great Lakes Water Quality Agreement of 1972, the United States
and Canada were provided a framework for the surveillance, monitoring,
research, protection, and reclamation of the physical and chemical quality of
the Great Lakes system. Within this framework, the monitoring of atmospheric
deposition in the U.S. is coordinated by the Great Lakes National Program
Office (GLNPO) of the U.S. Environmental Protection Agency (U.S. EPA, 1985).
Research in the 1970's had shown that atmospheric deposition was an important
source of certain organic and inorganic chemicals to lake watersheds. A
network of stations to measure and characterize this deposition was established
in 1976. In 1981 the GLNPO upgraded this earlier measurement network by
establishing the Great Lakes Atmospheric Deposition (GLAD) network. Its
purpose was to determine atmospheric loadings of metals, nutrients, and major
inorganic species to the Great Lakes and to evaluate annual trends in the
chemical loadings of these species to the Lakes. During 1981 and early 1982,
36 monitoring stations were installed along the U.S. shores of the 5 Lakes.
The GLAD network was designed to collect wet-only deposition samples at these
near-shore locations.
1.2. Purpose
The purpose of this study was to analyze and interpret atmospheric wet
deposition data collected by the GLAD network, including:
1) an assessment of data quality,
2) a comparison of specific pairs of GLAD and National Atmospheric
Deposition Program (NADP) sites,
3) estimation of atmospheric loadings of selected elements to the five
Great Lakes, and
4) an analysis of the potential change in loading estimates caused by
closing certain GLAD sampling sites.
-------
1.3. Scope
This report describes and interprets data collected in the GLAD network
during 1982 and 1983. At some sites, data from one of the two years were
ignored if that year's data failed to meet the screening criteria described
later in this report. The sampling sites in the GLAD network that provided data
analyzed in this report are listed in Table 1, which gives information about
site name and code, location, elevation, and start date. GLAD site QG, the
Copper Harbor site, Keweenaw Co., Michigan, did not meet selection criteria for
either year, and does not appear in Table 1.
Site information is given in Table 2 for a comparison subnetwork of
National Atmospheric Deposition Program (NADP) precipitation sampling sites in
the Great Lakes area. Figure 1 shows the locations of sampling sites in both
networks. The 16 GLAD sites closed in January 1986 are shown as solid circles
in Figure 1.
2. METHODS
2.1. Sampling and Analysis
Collection and analysis of samples and data reporting for the GLAD network
were the responsibility of the Great Lakes National Program Office. Details of
these respective procedures have been described in GLNPO documents (GLNPO,
1985; GLNPO, undated), but for completeness brief descriptions are provided
here.
The purpose of the GLAD network is to provide measurements of wet-only
atmospheric deposition that may be used to estimate chemical loadings to the
Great Lakes. To do this, some GLAD sites were sited in lakefront cities so as
to measure deposition from industrial, transportation, and residential sources
in and near urban areas. Other samplers were placed in lakeshore sites in rural
areas to measure deposition of area-wide sources. In addition, one site was
located in central Minnesota, about 200 km west of Lake Superior, to measure
regionally representative deposition from sources largely upwind of the Great
Lakes. This site satisfied the same siting criteria as NADP sites.
Precipitation samples were collected in AeroChem Metrics samplers, which
were designed to collect wet-only samples by uncovering a plastic bucket only
during precipitation. These buckets were lined with polyethylene bags. Field
observers were instructed to inspect the sampler every Tuesday at about 9 A.M.
local time. When a bag contained 500 mL or more of liquid precipitation (or at
least 1.25 in. of snow), a sample was collected and another bucket and liner
installed. Samples not meeting this minimum volume were left in the field for
one or more additional weeks, until sufficient sample had accumulated.
Beginning in 1984, rain gauges were installed at GLAD sites, so that an
independent measurement of the precipitation amount was available. Rain gauge
measurements could be used to assess the catch efficiency of the wet-only
collectors and to calculate chemical loadings by forming the product of
precipitation amount and chemical concentration. Prior to the installation of
-------
Table 1. GLAD Network Sites.
a
Site Name
Site Elevation Start
Code County/State Latitude Longitude MSL(m) Date
Lake Superior
Ontonagon
Grand Marais
Hovland
Gooseberry Falls
Duluth
Gull Lake
Cornucopia
Escanaba
Empire
Beaver Island
Green Bay
Manitowoc
South Water Plant
Jardine Plant
Evanston
Benton Harbor
Muskegon
Milwaukee
Bay City
Port Austin
Mount Clemens
Port Sanilac
Tawas Point
Dunkirk
Grand Island
Toledo
Put-In-Bay
Lorain
Fairport Harbor
Ashtabula
Erie
Olcott
Rochester
Cape Vincent
Fair Haven
QF
QJ
QAb
QBc
QD
RQ
QEb
QL
QV
QWb
QMb
QN
QR
QQ
QP
QT
QUb
QO
AQ
BQ
DQC
CQ
QZ
KQ
LQb
FQ
GQb
HO?5
IQ
JQ
SQ
MQb
NQ
PQ
TQC
Ontonagon, MI
Alger, MI
Cook, MN
Lake, MN
St. Louis, MN
Crow Wing, MN
Bayfield, WI
Lake Michigan
Delta, MI
Leelanau, MI
Charlevoix, MI
Brown, WI
Manitowoc, WI
Lake Michigan
Cook, IL
Cook, IL
Cook, IL
Berrien, MI
Muskegon, MI
Milwaukee, WI
Lake Huron
Bay, MI
Huron, MI
Macomb, MI
Sanilac, MI
losco, MI
Lake Erie
Chautauqua, NY
Erie, NY
Lucas , OH
Ottawa, OH
Lorain, OH
Lake, OH
Ashtabula, OH
Erie, PA
Lake Ontario
Niagara, NY
Monroe, NY
Jefferson, NY
Cayuga, NY
46
46
47
47
46
46
46
49
39
50
08
46
24
51
11
54
50
17
07
40
44
089
085
089
091
092
094
091
38
58
57
28
05
21
08
18
16
50
16
15
09
13
194
194
224
210
186
376
195
07/07/81
07/14/81
07/21/81
09/22/81
07/21/81
01/19/82
02/17/81
(North)
45
44
45
44
44
44
51
43
31
03
44
14
40
50
56
087
086
085
087
087
03
02
32
54
39
02
05
25
44
23
181
229
195
201
182
06/09/81
06/30/81
09/22/81
03/31/81
04/07/81
(South)
41
41
42
42
43
43
43
44
42
43
44
42
43
41
41
41
41
41
42
43
43
44
43
45
53
03
07
08
04
39
02
34
25
15
30
03
41
39
28
45
54
07
20
13
07
19
23
41
33
26
52
31
44
50
00
36
42
13
26
18
29
20
17
30
42
27
48
30
08
087
087
087
086
086
087
083
082
082
082
083
079
078
083
082
082
081
080
080
078
077
076
076
32
36
40
28
16
53
54
59
50
32
26
19
58
24
49
08
16
46
06
41
34
20
42
39
20
22
30
04
02
39
47
22
32
30
26
09
39
40
36
22
30
03
35
45
30
11
181
180
180
183
186
205
179
180
177
187
180
182
173
176
180
186
187
179
192
88
81
79
74
06/02/81
06/02/81
07/07/81
02/03/81
03/24/81
03/17/81
03/24/81
04/07/81
05/11/82
03/17/81
05/05/81
01/19/82
01/26/82
01/27/81
02/17/81
02/17/81
01/27/81
02/02/82
01/25/82
01/19/82
02/02/82
01/26/82
01/09/82
Except as noted, these sites all have valid wet-only data for at least 3/4 of the
precipitation and 3/4 of the 1/82 through 12/83 summary period.
bValid data for 1982 only.
:Valid data for 1983 only.
-------
Table 2. NADP Network Sites in the Great Lakes Region.1
Latitude
Elev. Start
Longitude MSL fm) pate
County/State
Champaign, IL
DeKalb. IL
DuPage, IL
Marion, IL
Porter, IN
Tippecanoe, IN 40 28 17 86 59 18 215 07/13/82
MI09 Cheboygan, MI 45 33 40 84 40 42 233 07/03/79
40 03 12
41 50 29
41 42 04
38 38 36
41 37 57
88 22 19
88 51 04
87 59 43
88 58 01
87 05 16
212
265
229
173
208
02/27/79
05/26/81
03/11/80
04/15/80
07/15/80
Site
Site Name Code
Bondville IL11
Shabbona IL18
Argonne IL19
Salem IL47
Indiana Dunes IN34
National Lakeshore
Purdue University IN41b
Agricultural Farm
Douglas Lake-Univ. of
Michigan Blol. Sta.
Kellogg Biol. Sta.
Wellston
Chassell
Marcell Exper. Forest
Fernberg
Ashland Wildlife Area
Washington Crossing
Aurora Research Farm
ChauCauqua
Huntlngton Wildlife
Bennett Bridge
Jasper
Delaware
Caldwell
Wooster
Kane Exper. Forest
Leading Ridge
Lake Dubay
Trout Lake
Spooner
Parsons
Mount Forest0
aExcept as noted, these sites all have valid wet-only data for at least 3/4 of the
precipitation and 3/4 of the 1/82 through 12/83 summary period.
bValid data for 1983 only.
cThis site is located in lower Ontario Province, and It was operated for the purpose of
intercomparing Canadian and U.S. Data
MI26
MI53b
MI99b
MN16
MN18
MOO 3
NJ99
NY08
NY10
NY20
NY52b
NY65
OH17
OH49
OH71
PA29
PA42
WI28b
WI36
WI37
WV18
CAN2b'c
Kalamazoo, MI
Wexford, MI
Houghton, MI
Itasca, MN
Lake, MN
Boone, MO
Mercer, NJ
Cayuga, NY
Chautauqua , NY
Essex, NY
Oswego, NY
Steuben, NY
Delaware, OH
Noble, OH
Wayne , OH
Elk, PA
Huntingdon, PA
Portage, WI
Vilas, WI
Washburn, WI
Tucker, WV
Wellington, ONT
42 24 37
44 13 28
47 06 17
47 31 52
47 56 45
38 45 13
40 18 54
42 44 02
42 17 58
43 58 20
43 31 34
42 06 22
40 21 19
39 47 34
40 46 48
41 35 52
40 39 32
44 39 53
46 03 09
45 49 21
39 05 23
43 59 29
85 23 34
85 49 07
88 33 05
93 28 07
91 29 43
92 11 55
74 51 17
76 39 35
79 23 47
74 13 19
75 56 50
77 32 08
83 03 58
81 31 52
81 55 31
78 46 04
77 56 10
89 39 08
89 39 11
91 52 30
79 39 44
80 44 46
288
292
277
431
524
239
72
249
488
494
245
634
285
276
315
618
282
338
501
331
305
410
06/26/79
10/10/78
02/15/83
07/06/78
11/18/80
10/20/81
08/04/81
04/17/79
06/10/80
10/31/78
06/10/80
02/19/80
10/03/78
09/26/78
09/26/78
07/18/78
04/25/79
06/29/82
01/22/80
06/03/80
07/05/78
05/05/81
-------
O GLAD open after Jan 1986
GLAD closed after Jan 1986
D NADP
Figure 1. Network map for GLAD and NADP sampling sites. (GLAD
site QG, in Keweenaw Co., MI, did not meet selection
criteria, and does not appear on this map.)
Figure 2. Grid used for over-lake deposition flux calculations,
-------
the gauges, the only on-site measurement of precipitation amount was the
collected volume from the AeroChem Metrics samples.
After collection, a 20-40 mL aliquot of sample was decanted from the bag
for field pH and conductance measurements. Next, a half liter polyethylene
bottle was filled and shipped to the EPA analytical laboratory. Upon arrival at
the laboratory, samples were stored at 4°C prior to preparation for analysis.
Samples were split and preservatives were added to the splits, as appropriate
to the analysis (e.g., nitric acid was added to the split intended for metals
analyses).
Ca, Mg, and Na ions were measured using the inductively-coup led argon
plasma (ICAP) method. Flame atomic absorption spectrophotometry (AAS) was used
for K analyses, and flameless AAS for Pb and Cd analyses. Automated wet
chemical methods (Technicon) were used to determine Ntfy, N03+N02, 804, and Cl.
2.2. Data Screening Criteria
As indicated in footnotes of Tables 1 and 2, the criteria for including
data from a GLAD or NADP sampling site in this study were that the site have
valid data for at least 75% of the two-year period and valid data representing
at least 75% of the two-year precipitation total. Since GLAD sites did not have
rain gauge measurements during 1982 and 1983, it was necessary to convert
sample volumes to precipitation amounts in order to calculate the two-year
precipitation totals. (See Appendix A for a list of criteria for selecting
valid samples.) If a sampling site did not meet the 2-year criteria, it was
reevaluated for inclusion of one year's data only, using a one-year
precipitation total calculated from sample volumes.
Whereas internal NADP data screening criteria assure that only samples
with analyses for all ions measured are included in the data available to the
public, no similar requirement was placed on the GLAD data. Thus, samples
appear in the GLAD data for which one or more analyses may be missing. This
means that the checks for data completeness (75% of the time and 75% of the
precipitation during the two-year period) had to be applied ion by ion for the
GLAD data, instead of sample by sample, as for the NADP data. This resulted in
a varying number of valid sites from which to draw maps of concentration
distributions for the various ions. Sites excluded from the spatial analyses of
concentrations for the various ions are listed in Table 3.
-------
Table 3. GLAD Sites Excluded from Spatial Analyses of Concentration
because of Data Completeness Criteria.
Ion Sites excluded
Cl AQ, BQ, CQ, KQ, LQ, MQ,
PQ, QA, QD, QF, QJ, QL,
QM, QP, QT, QU, QV, QW,
QZ, RQ
Ca MQ, QU, QW
Mg QD, QU, QW
K QB, QD, QE, QU, QW
Na QU, QW
NH4 DQ, QM
S04 QA, QL, QQ, RQ
Pb GQ, HQ, LQ, MQ, QA, QB,
QD, QE, QG, QM, QU, QW
Cd GQ, HQ, LQ, MQ, QA, QE,
QG, QM, QU, QW
2.3. Data Quality Checks
Several checks of internal self-consistency of the sample analytical
results are possible. These include 1) ion balance calculations, 2) comparisons
of measured and calculated conductance, and 3) comparisons of measured pH and
conductance. Further details of each of these data quality check procedures
are given in the following paragraphs.
Based on the concept that the net charge of all ions in an aqueous
solution should be zero, a comparison of the total measured anion and cation
equivalents is useful in establishing the possibility of either 1) the lack of
analysis of one or more important ions, or 2) inaccurate analyses. In other
words, if all ions in solution have been measured, and measured accurately, the
difference between the sums of anion and cation equivalents should be zero. We
express any differences as the ion percent difference (IPD) , defined as
follows:
Sum A equiv - Sum C equiv
IPD - 100 x ,
Sum A equiv + Sum C equiv
where (Sum A equiv) is the sum of the chemical equivalent concentrations of the
measured anions plus the concentration of bicarbonate, HCC>3~, calculated from
the measured pH. (See Stensland and Bowersox (1984) for details of the exact
method used to calculate HC03~.) Also, (Sum C equiv) is the sum of the chemical
equivalent concentrations of the measured cations, including H+ from the
measured pH. The performance of the GLAD precipitation sampling and analysis
-------
program was examined by comparing plots of the frequency distribution of the
IPD for the GIAD and NADP data (see Section 3.1).
A further check on the completeness and accuracy of precipitation
composition analysis can be made by comparing measured conductance against
conductance calculated by summing the contributions from the individual
measured ions. Again, if all ions present were measured, and measured
accurately, the conductance percent difference (CPD), defined as:
Calc Cond - Meas Cond
CPD - 100 x
Meas Cond
would equal zero. (See Lockard (1987) for the formula used to calculate
conductance.) Also in this case, we evaluated the GLAD network data from a plot
of the frequency distribution of its CPD in comparison with that of the NADP
network (see Section 3.1).
As examination of the GLAD data set progressed, the possibility of biased
pH measurements arose. This suggested a test of the consistency of the measured
pH and conductance measurements. The test consisted of a plot of measured pH vs
measured conductance in view of the relationship between the two. Specifically,
for a given measured pH, the measured conductance must be at least that of the
corresponding H+ concentration. The GLAD data set was evaluated from such a
plot, noting the frequency with which samples occurred in the "forbidden zone"
of the plot, where measured pH and conductance values were inconsistent with
each other. The results were compared to a similar plot for the NADP data.
A slight variation of the previous test was also carried out using
calculated pH instead of measured pH, in a test of the hypothesis that there
was pH bias in the data.
2.4. Calculation of Integrated Ion Concentrations
Some measure of the overall concentration of each ion at each sampling
site is necessary both for showing spatial patterns on maps and for use in
estimating long-term (annual or longer) wet deposition fluxes. For this work we
used sample volume-weighted mean concentrations. These concentrations were
computed for each site using the combined 2-yr data set wherever possible, or
for individual years, when data for only one year were available. This method
was chosen as the best for providing overall spatial patterns of the
concentrations. It did not allow for comparison of yearly patterns. In any
case, such a comparison was not feasible, since the screening process produced
a different set of valid sites for any ion in each year. To reduce the set of
sites to only the ones that were valid in both years would have resulted in too
little data for a representative comparison of the year to year differences.
-------
2.5. Metal Concentrations
Data for the toxic metals Hg, Cd, and Pb were examined in some detail to
ascertain whether they were suitable for determining precipitation-only fluxes
to the lakes. Separate summaries of our findings appear in Table 4 for 1982 and
1983. In 1982, about 40-50% of the samples were analyzed for Hg and Cd, and
only about 8% were analyzed for Pb. Many of the samples analyzed had
concentrations less than detection limits, as indicated in the table. Close
examination of the data revealed that most of the 1982 Cd and Pb measurements
were made using the ICAP method, for which detection limits were higher than
for the furnace AA method. This appears to account for the high fraction of
less-than-detection- limit values for 1982. With such few data available, it
was not possible to provide reliable flux estimates for 1982.
The situation was much improved in 1983, when virtually all the Cd and Pb
analyses were done using the furnace AA method (see Table 4). Thus, fluxes of
Cd and Pb to the Great Lakes were computed for 1983.
2.6. Paired Site Comparisons
Measured ion concentrations at paired GLAD and NADP sites were compared in
two ways. First, box diagrams of percentile distributions (Cleveland, 1985)
were prepared for four species common to both the GLAD and NADP networks:
sulfate, nitrate, ammonium, and calcium. When plotted side by side, these
provide a convenient visual comparison between sites. Second, the measured
concentrations at the respective sites were subjected to the nonparametric
Wilcoxon rank sum test (SAS Institute, Inc., 1982) to estimate the probability
that they came from the same population.
2.7. Deposition Calculations
Our objective was to estimate climate-averaged loadings of atmospheric
pollutants to the five Great Lakes from precipitation alone. This was a two-
step process. First, deposition fluxes were computed as the product of ion
concentrations and precipitation amounts (depth). Then, to get annual
loadings, the calculated fluxes were integrated over the lake areas.
Ideally, the fluxes would-be measured over a suitably long period.
Unfortunately, the GLAD data set is not among the very few with such a record
of measurements. Further, rain gauge measurements were not available at the
GLAD sites for 1982 or 1983. Since year to year fluctuations in mean
precipitation are likely to be greater than those in volume-weighted mean
concentration, we chose to compute deposition fluxes as the product of the
available short-term mean concentrations and long-term (30-yr) mean
precipitation values. The 30-yr precipitation record available from the
National Weather Service (NWS) has the added advantage that the spatial density
of samplers is much greater than that of the combined GLAD and NADP networks.
Fluxes computed in this way should be better approximations of the climate-
averaged loadings than those measured for one or two years, although not as
good as those measured over, perhaps, a 20-yr period.
-------
Table 4. Summary of GLAD Toxic Metal Measurements
(a) 1982
Hg
Cd
Pb
Total samples
Total measurements
Measurements above
analytical detection limit
Typical detection limit
975
482
49.4% of 975
37
7.7% of 482
0.1 ug/L
975
403
41.3% of 975
45
11.2% of 403
Highly variable
975
81
8.3% of 975
48
59.3% of 81
2.0 ug/L
Sites with >3 values Hovland, MN Erie, PA Grand Island,
above detection limit Fairport Harbor, OH Grand Marais, MI NY
Lorain, OH Manitowoc, WI
Method
Cold vapor AA
ICAP,
Furnace AA
ICAP,
Furnace AA
(b) 1983
Hg
Cd
Pb
Total samples
966
966
966
Total measurements
506
(52.4% of 966)
834
(86.3% of 966)
822
(85.1% of 966)
Measurements above 352
analytical detection limit (69.6% Of 506)
689
(82.6% of 834)
807
(98.2% of 822)
Typical detection limit
0.1 ug/L
0.10 ug/L
0.6 ug/L
Method
Cold vapor AA
Furnace AA
Furnace AA
10
-------
Computations were carried out at the vertices of each grid box in Figure
2. Spatial distributions of annual wet deposition fluxes were computer-plotted
from these gridded values. Annual wet loadings to the lakes were estimated from
the values computed for each grid box having some portion of its area over one
of the lakes. A mean value for each of these boxes was computed as the
arithmetic mean of the values at the vertices. Where only a portion of the box
was over water, the grid-box deposition was decreased by multiplying by the
water-area/total-area ratio, so as to include only the over-water portion of
the deposition.
The accuracy of the areal integration is shown in Table 5 by a comparison
of lake areas computed by integrating whole and partial grid boxes over the
lakes with literature values of lake area. The differences from literature
values assume that the literature values are accurate to four significant
figures and are expressed as percents of the literature values. The differences
are quite small, probably within measurement accuracy. Nevertheless, computed
fluxes were increased by the ratio of the literature area to the grid-box area
for the respective lakes.
Table 5. Comparison of Grid-Box Integrated Lake Areas with
Literature Values.
Superior
Michigan
Huron
Erie
Ontario
a Todd (1970).
Values of precipitation at the grid vertices were estimated using an
objective analysis procedure developed by Achtemeier (1987) and Achtemeier, et
al. (1977), based on earlier work by Barnes (1964, 1973). These estimates are
based on measured values at the 5 nearest sampling sites, weighted using a
negative exponential function of distance from the grid point. Each grid point
value is based on measured, not interpolated, values; thus grid values over the
lakes were based on on-shore measurements. A map of the precipitation
measurement network is shown in Figure 3. The data used in this analysis were
obtained from the NOAA National Climatic Center (NCC), Asheville NC. They are
30-yr (1951-80) precipitation normals, which are computed as the arithmetic
mean of 30 annual precipitation amounts. Station histories, documented at NCC,
were used to screen all precipitation data from 1) the NWS cooperative station
network, and 2) the NWS first order station network. This screening was
conducted by the NCC staff. The data used in our analysis were screened to
assure that instrument exposure and station location were "homogeneous" for the
30-yr period. Where it was determined that there were inhomogeneities, the
records at the two locations or for the two gauge exposures were compared to
11
Grid-box integrated
area (so km)
80,200
56,840
58,170
25,350
19,070
Literaturea
area (sq km)
82,410
58,020
59,600
25,740
19,530
Difference
(percent)
-2.7
-2.0
-2.4
-1.5
-2.3
-------
Figure 3. National Weather Service observer network used for 30-yr mean
annual precipitation distribution.
12
-------
nearby homogeneous data to ascertain that the records were unaffected by the
change of location or exposure. A brief description of this approach is
available (National Climatic Center, 1982).
Ion concentrations at the grid points were estimated similarly, also using
the nearest 5 measurements. The map of sampling sites is shown in Figure 1.
Separate concentration and associated deposition estimates were made for 1) the
NADP sites only, 2) the combined GLAD and NADP networks, and 3) the full
combined network less certain GLAD sites or combinations of GLAD sites. These
results were compared to show the effect on deposition of 1) adding the full
GLAD network to the NADP network, 2) removing GLAD sampling sites closed in
January 1986 from the full set of GLAD and NADP sites, and 3) various options
for closing GLAD sites in addition to the 16 closed in January 1986.
3. RESULTS AND DISCUSSION
3.1. Data Quality Checks
Results of the internal data consistency checks described earlier are
presented here, beginning with the ion balance test.
Frequency distributions of IPD for both the GLAD and NADP networks appear
in Figure 4, and the results of certain statistical tests on the data are given
in Table 6. The 798 GLAD samples have a relatively broad distribution and a
median IPD of -20.6% (mean = -21.3%), which is significantly different from
zero. The negative value indicates either an excess of cations or a deficit of
anions. The 2168 NADP samples have a relatively narrow distribution and a
median IPD of 2.59% (mean = 2.66%), which is much closer to, but still
significantly different from, zero and indicates a slight excess of anions or
deficit of cations. The sign of this difference is consistent with the absence
of trace metal measurements of NADP samples.
Frequency distributions of CPD for both networks appear in Figure 5, and
associated statistics in Table 7. The GLAD samples again have a relatively
broad distribution and a median CPD of 47.9% (mean = 70.0%), which is
significantly greater than zero. Combined with the ion balance results, this
indicates a broad tendency for excess cations in the GLAD data. The NADP
samples exhibit a narrow, highly peaked distribution with a median CPD of
-8.73% (mean = -9.34%), which is also significantly different from zero, and,
as before, is consistent with the absence of measurements of trace components.
Comparisons of the IPD and CPD distributions of the GLAD and NADP data
show large differences that were not expected. Both data sets were derived from
the same geographic area, thus one would expect the measured chemical
composition of precipitation from the two networks to be similar. Some GLAD
sites were in urban and industrial regions, which could lead to significant
inputs of substances that were not measured routinely. This could explain the
greater variability (larger spread) of the IPD and CPD distributions of the
GLAD data. Another source of the variability could result from measurement
problems and this was explored next.
13
-------
-100. -W. -"
-a. o. a. so.
IBM PERCENT DIFFERENCE
Figure 4. Frequency distributions of ion percent difference (IPD)
for GLAD and NADP networks, 1982-1983.
-150. -100. -SO. 0. SO. 109. 19.
C0NDUCTRNCE PERCENT DIFFERENCE
Figure 5. Frequency distributions of conductance percent difference (CPD)
for GLAD and NADP networks, 1982-1983.
14
-------
Table 6. Summary of Ion Percent Difference Results for GLAD and NADP Sites
(1982-1983). (Also see Figure 4.)
GLAD
798
-20.6
-21.3
15.7
NADP
2168
2.59
2.66
7.22
N
Median
Mean
Std dev
Tests: 1) GLAD median ^ 0 (P - 0.0001, sign rank test).
2) NADP/NTN median t 0 (P - 0.0001, sign rank test).
3) Reject H0 (P - 0.0001, Wilcoxon 2-sample test) that
GLAD = NADP.
A) Reject H0 (P
GLAD median
0.0001, Brown-Mood test) that
- NADP median.
Table 7. Summary of Conductance Comparison Results for GLAD and NADP Sites
(1982-1983). (Also see Figure 5.)
N
Median
Mean
Std dev
Tests: 1) GLAD median ± 0 (P - 0.0001, sign rank test).
2) NADP/NTN median J> 0 (P = 0.0001, sign rank test).
3) Reject H0 (P = 0.0001, Wilcoxon 2-sample test) that
GLAD = NADP.
4) Reject H0 (P - 0.0001, Brown^Mood test) that
GLAD median = NADP median.
GLAD
795
47.9
70.0
161
NADP
2168
-8.73
-9.34
9.38
15
-------
Since H+ is the major contributor to the conductance of precipitation
samples, we chose to investigate the possibility that a negative pH bias
(erroneously high H+ concentrations) was a cause of the cation excess indicated
by the combined results of the ion balance calculations and the comparison of
calculated and measured conductances. The plots of measured pH versus measured
conductance described in Section 2.2. should confirm or refute the suggestion
of a pH bias in the GLAD measurements.
Figure 6(a) shows results for the two-year GLAD data set. A considerable
number of the data points are in the "forbidden" zone below the sloping line
showing conductance of H"1" only, as a function of measured pH. In comparison,
the NADP data in Figure 6(b) behave as expected, with no measurements in the
forbidden zone. The results in Figure 6, by themselves, indicate that either
the pH or the conductance measurements are biased low, but we already know from
the results in Figures 4 and 5 that we have an excess of cations, which is
consistent only with a pH bias. This is further confirmed in Figure 7(a), where
the pH calculated from the other ion measurements shows a much closer
correspondence to the ideal that no points should occur in the forbidden zone
beneath the sloping line.
Strictly speaking, however, the prohibition against points below the line
is not absolute when all the other ions in solution are considered. In very
acidic samples, small imprecisions in the pH and conductance measurements can
produce points below the line. Note that a few points occur below the line for
both GLAD data in Figure 7 (a) and NADP data in Figure 7(b).
By comparison with results for comparable NADP samples, the distributions
of ion balance and conductance differences for the GLAD network clearly show an
anomalous cation excess. The occurrence of many points in the "forbidden zone"
in a plot of measured pH versus measured conductance for the GLAD data shows
that the cation anomaly was caused by biased pH measurements. This was further
confirmed when the anomaly largely disappeared in a comparable plot of
calculated pH vs measured conductance for the same samples.
Because of the strong evidence of a pH bias in the GLAD data, no further
use was made of the GLAD pH data.
3.2. Concentration Spatial Distributions
3.2.1. Effect of Adding GLAD Sites to the NADP Data Set
An important potential benefit of the GLAD network is the additional
spatial resolution in the Great Lakes region that it might provide to national
atmospheric deposition networks. We present here a series of figures comparing
objectively-analyzed spatial distributions of concentration. The objective
analysis scheme described in section 2.7 was used to develop these figures. In
each figure, the top (a) panel shows the concentration pattern resulting from
regional NADP data alone, and the bottom (b) panel shows the patterns obtained
when the GLAD data were added to the data set.
16
-------
8.0
3.5 -
3.0
0.
1. 10. 100.
MEASURED C0NOUCTHNCE (MICR0SIEMENS/CM)
1000.
Figure 6(a). Scatterplot of measured pH and conductance for the GLAD
network, 1982-1983. There is a "forbidden zone" beneath the
sloping line, which corresponds to the conductance due only to
the measured H ion in solution.
8.0
3.0
1000.
MEfiSUREO C0NDUCTONCE (MICR0SIEMENS/CMI
Figure 6(b). Scatterplot of meaured pH and conductance for the NADP/NTN
network, 1982-1983. The sloping line defines a "forbidden
zone," as described above.
17
-------
8.0
7.5
7.0
6.5
6.0
I
I
1 S.5
S.O
4.0
3.5
3.0
Forbidden Zone
a.
10.
100.
1000.
HEflSURED CBNDUCTRNCE (MICR0SIEMENS/CM)
Figure 7(a). Scatterplot of calculated pH vs measured conductance for the
GLAD network, 1982-1983. The sloping line defines a "forbidden
zone," as described in Figure 6(a).
8.0
7.5 -
3.5 -
3.0
0.
I. 10. 100.
MEflSUREO CBNDUCTRNCE IM1CRBSIEMENS/CM)
1000.
Figure 7(b). Scatterplot of calculated pH vs measured conductance for the
NADP/NTN network, 1982-1983. The sloping line defines a
"forbidden zone," as described in Figure 6(a).
18
-------
Figure 8 compares the respective concentration patterns for 804. The broad
patterns are quite similar, as one would expect. However, the GLAD data provide
additional detail in the pattern near and over the Great Lakes. Relatively
minor differences appear over Lakes Superior, Michigan, and Huron, but
considerable detail is added over Lakes Erie and Ontario. Concentrations at
most GLAD sites along the lower Lakes are 1.5 to 2 times higher than at NADP
sites in this area. To the extent that the sampling locations were
representative of their locales, and to the extent that there were no other
sampling, analytical, or data handling biases, the GLAD data add information
useful for computing lake loadings. However, there are large uncertainties in
the sizes and shapes of the areas represented by urban samplers. Further
research is needed to quantify the effect of these uncertainties on computed
lake loadings.
Figure 9 compares patterns with and without GLAD data for Ca. Judging from
the magnitudes of the additional urban peaks provided by the GLAD data, their
utility is even more striking than in the 804 case. This is to be expected,
since many of the GLAD sites were in or near urban areas having sources of Ca,
such as unpaved roads and parking lots, construction, and demolition. Since
aerosols carrying Ca are emitted at ground level and since they are emitted as
large particles, they tend to be transported over shorter distances than 864,
prior to being deposited in precipitation. Ca concentrations along the lower
Lakes are 2 to 3 times higher than at NADP sites in this area.
Figure 10 compares patterns with and without GLAD data for NC>3. Again, the
GLAD data add resolution. In this case the additional features in the
concentration pattern from the combined networks are somewhat broader in
spatial scale than we saw with Ca. This is consistent with the somewhat more
distributed nature of automotive exhaust, the major NOX source, as compared
with the more localized Ca sources. Though NC>3 concentrations at most GLAD
sites were higher than at nearest NADP sites, values were less than twice as
high. This feature is also consistent with the suggestion that the gaseous NOX
is more widely dispersed before it is deposited by precipitation.
Figure 11 compares patterns for NH4- Again, some increased detail is seen
near the lakes, but not the extreme local concentrations of Ca.
In general, the differences between the combined NADP-GLAD and NADP-alone
patterns were greater over Lakes Erie and Ontario than over the other Lakes.
Ion concentrations at (urban) GLAD sites along these lower Lakes were
unquestionably greater than those at NADP sites in the same geographic area. At
GLAD sites along the other Lakes, Chicago stands out as a major deposition
area. For N03, concentrations were anomalously high in the eastern Wisconsin
area.
In summary, there was an apparent influence from local (urban) sources on
GLAD data (if there were no important sampling biases). This urban effect
results in ion concentrations that were higher than at the regionally more
representative NADP sites. The spatial extent of this local influence has a
large uncertainty. More research is needed to improve the quantification of
this effect. This urban effect is most apparent at GLAD sites along Lakes Erie
19
-------
SO4 Concentration (mg/L)
NADP only
Figure 8(a). Spatial distribution of volume-weighted S
concentrations in the Great Lakes region,
using NADP data for 1982-83.
S04 Concentration (mg/L)
GLAD + NADP
S
CM'
Figure 8(b). Same as (a), but using the combined GLAD/NADP
data set.
20
-------
Ca Concentration (mg/L)
NADP only
Figure 9(a). Spatial distribution of volume-weighted Ca
concentrations in the Great Lakes region,
using NADP data for 1982-83.
0.20
0.20
.0.30
Ca Concentration (mg/L)
G LAD + NADP
Figure 9(b)
Same as (a), but using the combined GLAD/NADP
data set.
21
-------
N03 Concentration (mg/L)
NADP only
Figure 10(a). Spatial distribution of volume-weighted NC>3
concentrations in the Great Lakes region,
using NADP data for 1982-83.
1.25
1.25
NO-j Concentration (mg/L)
GLAD + NADP
1.50
Figure 10(b).
Same as (a), but using the combined GLAD/NADP
data set.
22
-------
NH4 Concentration (mg/L)
NADP only
Figure 11(a)
Spatial distribution of volume-weighted NH^
concentrations in the Great Lakes region, using
NADP data for 1982-83.
NH4 Concentration (mg/L)
G LAD + NADP
Figure ll(b)
Same as (a), but using the combined GLAD/NADP
data set.
23
-------
and Ontario. Many of the sites along Lakes Superior, Huron, and northern Lake
Michigan were rural and did not exhibit this urban effect.
3.2.2. Additional Ions. Combined Data Set
Beside 804, Ca, NC>3, and NH4, for which we examined the effect of adding
the GLAD sites to the regional NADP network, the spatial patterns of volume-
weighted concentrations of four additional ions are shown, using the combined
data set from both networks.
Figure 12 shows the spatial pattern of Na concentrations in the combined
data set. Aside from the sea-salt influence seen in the SE corner of the map
near the Atlantic coast, the major concentration peaks occur in urban or other
lakeshore areas. A very similar pattern occurred for Cl (Figure 13). The same
general patterns also occurred for Mg (Figure 14) and K (Figure 15).
3.2.3. Paired Site Comparisons: GLAD vs NADP
A map showing the 12 pairs of sites for which a comparison was requested
by GLNPO is given in Figure 16. Comparisons were made for 804, Ca, NC-3, and NIfy
in the form of box diagrams of concentration percentile distributions. These
comparisons were made partly as an indirect check on the accuracy of ion
concentration measurements, and partly for the purpose of suggesting where GLAD
measurements might be providing little additional information in addition to
that already available from the NADP network.
Table 8 lists the pairs of sites at which ion concentrations were compared
and provides a key to the numbers by which the sites are labeled on the
abscissas of Figures 17-20.
At each site in Figures 17-20, concentration percentiles are represented
by a box diagram (Cleveland, 1985) in which the diamond symbol in the middle of
the "box" represents the 50th percentile, and the top (a triangle) and bottom
(a plus sign) of the box are at the 75th and 25th percentiles, respectively.
The upper (an x) and lower (a square) extremes, connected by single lines,
represent the 90th and 10th percentiles, respectively. The site pairs are
ordered from left to right in the figure approximately in order from NW to SE
across the Great Lakes basin. Asterisks on site pairs indicate that the
distributions of ion concentrations were significantly different at the 1%
level. Parentheses on site pairs denote differences significant at the 5%
level.
Figure 17 shows paired site comparisons for 804. Note that generally lower
concentrations prevail in the NW portion of the basin (left side of the
figure), and higher concentrations occur in the SE (right side of the figure).
Detailed comparisons of individual site pairs require consideration of
possible local sources, differences in siting criteria (e.g., GLAD allows
rooftop sampling; NADP does not), distance between sites, and other issues.
However, we can gain a general impression of how the two networks compare in
24
-------
Na Concentration (mg/L)
GLAD + NADP
Figure 12. Spatial distribution of volume-weighted Na concentrations
in the Great Lakes region, GLAD/NADP data, 1982-83.
Cl Concentration (mg/L)
G LAD + NADP
Figure 13. Spatial distribution of volume-weighted Cl concentrations
in the Great Lakes region, GLAD/NADP data, 1982-83.
25
-------
Mg Concentration (mg/L)
GLAD + NADP
Figure 14. Spatial distribution of volume-weighted Mg
concentrations in the Great Lakes region, GLAD/
NADP data, 1982-83.
K Concentration (mg/L)
GLAD + NADP
.025
Figure 15. Spatial distribution of volume-weighted K
concentrations in the Great Lakes region, GLAD/
NADP data, 1982-83.
26
-------
Figure 16. Map showing locations of GLADNADP site pairs
for which ion concentrations were compared.
27
-------
a -
s -
7 -
6 -
5 -
4 -
3 -
2 -
1 -
n -
i
I
}
.
\
\ 1
S 1
:
U
3 U
^
1
1
3
5,
..
f
I
J
*
J
(
I
\
i
( *
i
^
y
i
3
C
1
i
:
t
,(
1
3
I I
3
1
J
i
lOth
25th
PERCENTILES
o 50th
A 75th
x QOth
Figure 17. Paired site comparisons of 864 concentration percentiles, using
box diagrams. Asterisks signify that distributions were
different at the 1% level, based on the Wilcoxon rank sum test.
28
-------
1.5 -
1.4 -
1.3 -
1.2 -
_J 1-1 -
O* 1
E
O.9 -
Z
<~i
£ 0.8 -
£ O.7 -
UJ O.6 -
O
0 0-5 -
O
O.4 -
0.3 -
O.2 -
0.1 -
₯ x t
t
1
n
A
i
. f
b I
i
i
;
v
1
b
^l
V
1
^
t
F
I
b^
i i r i i i i i i
12 34 56 78 91
X
C
*
I1
A
5
. >
:
X
5
4
'
5
'
H
1
V
1
i
1
O 1112 1314*1516^171^1926*2122*2324
lOth
25th
PERCENTILES
o 5Oth
75th
9Oth
Figure 18. Paired site comparisons of Ca concentration percentiles, using
box diagrams. Asterisks signify that distributions were
different at the 1% level, using the Wilcoxon rank sum test.
29
-------
TO
1
A
CONCENTRATION. mg/L
9
TJ o. cr *d
B H- o o>
i HI x H.
n HI 11
3 n a. n
n n H. o.
3* n «
m 3 TO n
o rt »» H-
n B rt
09 n B »
rr o
r* o
3 rt O
a 3* > B
H. n o> -o
n rt B
B i IB n
rt M n *
n H- CD
M M O
a. m rr s
H- < CD B)
H» (8
H» M W O
(9 M> Hi
1 TO
n c s z
3 CO H> O
O H- H»U>
m s *<
ca m n
rt o
m rt 3- 3
H* 3* B O
TO ft rt (D
3 3
M- C O. rt
HI (-« H- »1
H> M W B
O O rt rt
B O 1 H-
3 X H- O
rt O CT 3
3 C
» rt -a
rt n h* (D
B O 1
rt 3 3 O
3* 7? CO A
n 3
W rt
iji c n M-
X S H (-
a> n
CO
est.
evel
using
Oth
75th
9Oth
M -
"UT-
**
N)
w
to
i
Oi
t
0)
-K
D-
-X
-X
-K
n-
» t>-
-x
-^<
-X
-X
D-
-X
D-
-X
D-
O-
-------
o
I
UJ
u
z
o
o
1.1 -
1 -
o.g -
O.8 -
0.7 -
0.6 -
O.5 -
0.4 -
O.3 -
O.2 -
O.I -
O -
^
3
C
5
r
]
1
.
I
.
2
»
.
i
.
V
a
a
a
3
r
C
r
1
f
k
1
.
C
-
1 <
r
i i i i i iii ii iii iii i i i i i i ii ''.I iii ii iii
1 2* 3 4* 5 6 78 9ld*"l112 1314 (1516J 1718* 132O*" 2T22**2324
1Oth
25th
PERCENTILES
o 5Oth
75th
SOth
Figure 20. Paired site comparisons of NH4 concentration percentiles, using
box diagrams. Asterisks signify that distributions were
different at the 1% level, using the Wilcoxon rank sum test.
Parentheses indicate differences significant at the 5% level.
31
-------
Table 8. Key to GLAD/NADP Paired Site Comparisons in Figures 17-20.
Separation
Are concentration distributions
different at the 1% level?
No. on
graph
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Site
code3
MN-18
QA
MN-18
QB
WI-37
QE
WI-36
QF
MI-09
QV
IN-34
QT
MI-26
QT
NY-10
SQ
PA-29
SQ
NY-10
KQ
PA-29
KQ
NY-52
PQ
Site distance
typeb (km)
remote
rural
remote
rural
rural
rural
rural
remote
rural
rural
suburban
urban
suburban
urban
rural
urb 100
rural
urb 100
rural
suburban
rural
suburban
rural
rural
115
90
134
85
133
74
89
61
138
22
115
74
Y's
804
Site name
Fern berg
Hovland
Fernberg
Gooseberry Falls
Spooner
Cornucopia
Trout Lake
Ontonagan
Douglas Lake
Empire
Indiana Dunes
Benton Harbor
Kellogg Biol Sta
Benton Harbor
Chautauqua
Erie
Kane Exp Forest
Erie
Chautauqua
Dunkirk
Kane Exp Forest
Dunkirk
Bennett Bridge
Cape Vincent
Y's:
N's:
with GLAD higher:
Y
N
N
Y
N
N
N
Y
Y
Y
Y
N
6
6
5
Ca
N
N
N
N
N
N
Y
Y
Y
Y
Y
N
5
7
5
N03
N
(N)c
(N)
Y
N
N
N
N
(N)
(N)
Y
N
2
10
1
NH4
Y
Y
N
N
Y
N
N
(N)
Y
Y
Y
N
6
6
6
a In each pair, the NADP site is odd-numbered and the GLAD site is even.
b Definitions of site types are: remote no towns within 15 km; rural
no towns of population 10,000 within 15 km; suburb (suburban)
towns of population 10,000 to 100,000 within 15 km; urban site in a
town of population 10,000 to 100,000; urblOO site in a city of
population over 100,000.
c Parentheses signify that differences were not significant at the 1%
level, but were significant at the 5% level.
32
-------
terms of the concentrations of dissolved constituents. We do this by noting the
number of pairs, compared to the total, for which the GLAD sites had the higher
50th percentile values .
A more quantitative comparison of the ion concentration measurements at
paired GLAD and NADP sites was made using the nonparametric Wilcoxon rank sum
method to test for differences in the overall concentration distributions
between sites. Table 8 shows the results of testing for the significance (1%
and 5% levels) of differences between concentration distributions for SO^, Ca,
NC>3 , and Nlfy at each pair of sites.
For 804, the GLAD site had the higher 50th percentile value in eight of
the twelve pairs. The difference in distributions was significant at the 1%
level for five of these pairs, of which four involved either the Erie (SQ) or
Dunkirk (KQ) GLAD sites. Both the SQ and KQ sites had a much different physical
setting than the NADP sites with which they were compared. SQ was in Erie,
Pennsylvania, a city of over 200,000 people, and KQ was near Dunkirk, New York,
a town of over 10,000 people. NADP sites NY- 10 and PA- 29 both have a rural
setting. Thus, 804 differences at these pairs are not surprising, given the
likely presence of SOX sources near the GLAD sites.
For one pair of sites, the NADP site had a 50th percentile value greater
than that of its GLAD counterpart, significant at the 1% level. For this pair
the situation was reversed; the GLAD site, QF, was a "remote" site with no
towns within 15 km, whereas the NADP site, WI-36, at Trout Lake, is in a rural
area with pulp and paper industry about 50 km SSE and occasional forest cutting
in the area. Differences in sulfate concentrations are consistent with the
regional decrease in concentrations from S to N across northern Wisconsin.
Local sources from the few small towns in the vicinity are unlikely to be the
cause of the difference. There were no additional pairs where differences, in
either direction, were significant at the 5% level.
These results show some tendency for the GLAD sites to have higher 864
concentrations , but of course do not distinguish between true spatial
differences and possible analytical bias. Analytical bias can only be
determined from duplicate sampling and/or laboratory comparisons, and such
activities were not carried out for the GLAD network.
Figure 18 shows paired site comparisons for Ca. For the GLAD sites, there
appear to be higher concentrations at all percentiles in the SE portion of the
region, compared to the NW, but this does not appear to hold for the NADP
sites. As a result, the differences between the GLAD and NADP sites are larger
in magnitude in the SE than elsewhere. For Ca, the 50th percentile values were
higher at the GLAD site in nine of the twelve pairs, and higher at the NADP
site in three pairs. Where the GLAD site had the higher 50th percentile
concentration, the differences in distribution were significant (1%) in five
pairs, four of which again involved the urban or suburban sites at Erie and
Dunkirk, respectively. The fifth pair involved the "urban" GLAD site QT, at
Benton Harbor, Michigan, and the "suburban" NADP site at Kellogg Biological
Station, MI -26. Results from all five of these pairs are consistent with the
stronger influence of local traffic, construction, and other activities on Ca
concentrations at the more urban GLAD sites than at the regional NADP sites.
33
-------
There were no additional pairs where the GLAD distribution was
significantly greater at the 5% level, and no pairs where the NADP distribution
was significantly greater at either the 1% or 5% level.
Figure 19 shows paired site comparisons for N(>3. As in the case of 804,
there is a trend from lower concentrations at all percentiles in the NW to
higher concentrations in the SE, and this occurred for both networks. For NC>3,
the 50th percentile values were higher at the GLAD site in nine of the twelve
pairs. In only one (Dunkirk) of these nine cases was the difference in
distributions significant at the 1% level, but in three other cases (two
involving Erie or Dunkirk) the differences were significant at the 5% level. An
urban influence, similar in direction, but smaller in magnitude than for Ca and
804, is the explanation for these differences. Apparently the NOX sources are
not as localized as those of large-particle Ca and point-source SOX. There was
also one case where the NADP distribution exceeded GLAD at the 1% level, and
another where the difference (in the same direction) was significant at the 5%
level.
Figure 20 gives the paired site data for NH4. The evidence for overall
spatial trends is weak or lacking. The variability in the measurements, as
indicated by the ratio between the 90th and 10th percentile values, is the
highest of the four ions examined, and occurs over the whole region in both
networks. Also in this case, 50th percentile values are typically higher at the
GLAD sites, occurring that way in nine of the twelve pairs. Of these nine
pairs, six (three of these involving Erie or Dunkirk) were significant at the
1% level, and one more (Erie) at the 5% level. The NADP distributions were
never significantly greater than those of their GLAD counterparts at either 1%
or 5%.
To summarize the site comparisons for 804, Ca, N03, and Nlfy, we note that
the GLAD sites often had concentrations significantly higher than the paired
NADP sites. The GLAD concentration distributions exceeded those of their NADP
counterparts at the 1% level much more than expected by chance for 804, Ca, and
NH4, and somewhat more than expected by chance for N03. However, most of these
significant differences involved either the Erie or Dunkirk sites. These
differences could reflect true spatial gradients, or could alternatively be
related to differences in siting criteria (many GLAD sites are on roofs in
urban areas, while NADP sites are on the ground in regionally representative
locations) or local sources. It will take a careful intercomparison study with
co-located samplers to quantify the importance of these factors.
3.3. Precipitation Amount
Precipitation amount is the other factor, besides concentration, that goes
into the calculation of deposition, and the year-to-year variability of its
mean value is generally much greater than that of ion concentrations in
precipitation. To show ion fluxes to the lakes that represent long-term means,
we calculated deposition as the product of 2-yr volume-weighted mean
concentrations and 30-yr mean precipitation.
34
-------
Figure 21 shows the distribution of the 30-yr mean annual precipitation
over the lakes. Precipitation amounts generally range from about 90 cm in the
southern portions of the region to about 75 cm in the northern portions. As was
the case for concentrations, there were no routine over-water measurements of
precipitation from which to calculate ion deposition fluxes. Instead, overland
measurements from around the Lakes were used. These data were analyzed using
the same procedure applied to the ion concentrations. Isopleths were drawn from
the resulting grid point data set. Studies by Changnon (1972), Wilson (1977),
and Bolsenga (1979) have shown that precipitation data from onshore gauges near
the shore is a good estimate of what fell over the Lakes, because the excess
precipitation that falls over the Lakes in the winter is about counter-balanced
by the deficit over the Lakes in the summer. On an annual basis, the lake-land
differences are less than measurement errors.
Table 9 summarizes the annual precipitation fluxes to the individual
lakes, showing 30-yr means for the 1951-80 period, and comparative 30-yr means
and standard deviations, as well as maximum and minimum annual values from an
independent estimate based on the 1954-83 period. Except for Lake Huron, where
there is about a 10% difference, the agreement between estimates based on
slightly different but largely overlapping 30-yr periods is very close,
providing further evidence that the areal integration method is a reasonable
one. We have examined the data for an explanation of the 10% difference for
Lake Huron. The difference was not caused by unusual values during the non-
overlapping years of the two data sets. It appears to be related to differences
in the measurement networks (only the 1954-83 data set includes Canadian data),
and the respective methods used for areal integration.
3.4. Deposition
3.4.1. Spatial Patterns
Spatial patterns of annual wet-only deposition fluxes are given in Figures
22-31 for ten ions measured in precipitation. Except for the Cd and Pb results
(Figures 30 and 31) , the computed deposition patterns are based on a combined
data set from the NADP and GLAD networks, using data for 1982 and 1983. The
NADP does not analyze for metals, so the Cd and Pb results were computed from
GLAD data for 1983. This was the only year of the two examined that had
adequate metals data. As explained in detail in Section 2.6, the depositions
were computed from 2-yr (or 1-yr, as available) volume-weighted mean
concentrations and 30-yr mean precipitation values.
Figure 22 shows the spatial pattern of 804 wet deposition over the Great
Lakes. Annual fluxes ranged from about 10 kg/ha in the NW portion of the basin
to about 45 kg/ha in the SE. Maximum values occurred at GLAD sites in urban
and/or industrial areas such as the south end of Lake Michigan and the south
shore of Lake Erie. A similar overall pattern was observed for the wet
deposition of Ca (Figure 23). Minimum annual fluxes of about 2 kg/ha occurred
in western sections of Lake Superior and northern sections of Lakes Michigan
and Huron. Maximum values of 6 kg/ha or more occurred near Chicago in southern
Lake Michigan and over the southern shore of Lake Erie. The spatial patterns of
deposition fluxes of SO^ and Ca resemble their concentration patterns.
35
-------
Figure 21. Distribution of 30-yr mean annual precipitation (centimeters) over the
Great Lakes. Data source: National Climatic Center, Asheville, NC.
-------
SO^Deposition (kg/ha)
Figure 22. Spatial distribution of annual deposition fluxes of 804
over the Great Lakes, using GLAD-NADP data, 1982-83.
Figure 23. Spatial distribution of annual deposition fluxes of Ca
over the Great Lakes, using GLAD-NADP data, 1982-83.
37
-------
10.0
NOj Oepotilion (kg/ha)
Figure 24. Spatial distribution of annual deposition fluxes of N03
over the Great Lakes, using GLAD-NADP data, 1982-83.
Figure 25. Spatial distribution of annual deposition fluxes of NH4
over the Great Lakes, using GLAD-NADP data, 1982-83.
38
-------
Figure 26. Spatial distribution of annual deposition fluxes of Na
over the Great Lakes, using GLAD-NADP data, 1982-83.
Cl Deposition (kg/ha)
Figure 27. Spatial distribution of annual deposition fluxes of Cl
over the Great Lakes, using GLAD-NADP data, 1982-83.
39
-------
0 60
Mg Deposition (kg/ha)
Figure 28. Spatial distribution of annual deposition fluxes of Mg
over the Great Lakes, using GLAD-NADP data, 1982-83.
0.40
K Deposition (kg/ha)
Figure 29. Spatial distribution of annual deposition fluxes of K
over the Great Lakes, using GLAD-NADP data, 1982-83.
40
-------
Cd Deposition (g/ha)
GLAD only
Figure 30. Spatial distribution of annual deposition fluxes of Cd
over the Great Lakes, using GLAD data, 1983.
Figure 31. Spatial distribution of annual deposition fluxes of Pb
over the Great Lakes, using GLAD data, 1983.
41
-------
Table 9. Summary of Precipitation Fluxes to the
Great Lakes, 1951-1980.
1951-80a
mean
feu km)
Superior
Michigan
Huron
Erie
Ontario
65.
46.
44.
23.
16.
6
0
7
2
3
1954-83b
mean (+ S.D.)
feu km)
64.
46.
50.
23.
16.
3
3
1
3
5
± 8
± 5
± 4
± 2
+ 1
.0
.4
.5
.8
.6
1954-83
minimum
feu km)
52
36
39
15
13
.3
.4
.8
.5
.4
1954-83
maximum
feu km)
80
55
57
29
20
.5
.7
.3
.5
.0
a Data source: 1951-80 30-yr normal precipitation data, National Oceanic and
Atmospheric Administration, Environmental Data and Information Service,
National Climatic Center, Asheville, NC, September, 1982.
Data source: National Oceanic and Atmospheric Administration, Great Lakes
Environmental Research Laboratory, Ann Arbor, MI, July, 1986. (H. Hartman,
personal communication.)
42
-------
Figure 24 shows the spatial distribution of N03 wet deposition. Annual
fluxes ranged from about 10 kg/ha over W Lake Superior to 20 kg/ha or more over
E sections of Lakes Erie and Ontario. The overall pattern is similar to those
of 804 and Ca, but the locations of the maxima over the lower lakes are quite
different. From maximum to minimum, Ca and 804 fluxes vary by factors of about
3.5 and 4.5, respectively, whereas NC>3 has a smoother pattern and varies by a
factor of only about 2.2. The pattern for NH4 (Figure 25) also shows
similarities and differences compared to those already presented. Maxima
occurred in the Chicago area, and near Erie, Pennsylvania, as seen for ions
already discussed. New areas with relatively high wet fluxes of NH4 included
the Duluth, Minnesota, area at the W tip of Lake Superior, and central portions
of the W shore of Lake Michigan (Milwaukee to Green Bay) . Southern Lake Huron
also showed a stronger maximum than seen there for the ions presented
previously. The minimum NH4 wet fluxes occurred over N Lake Huron. Again these
patterns bear a strong resemblance to the respective concentration patterns.
Wet deposition patterns for Na and Cl appear in Figures 26 and 27,
respectively. Na fluxes ranged from about 0.25 to 2.5 kg/ha, a factor of 10
between the minimum and maximum. The range of the Cl fluxes was also rather
large, with minimum values of less than 1 kg/ha and a maximum of about 7. The
Cl pattern was similar to the Na pattern, with low values over large areas of
Lakes Huron and Superior and local maxima near Chicago and Cleveland. Cl also
had additional maxima over Lake Ontario and W portions of Lakes Erie and
Superior.
Annual patterns of wet deposition fluxes for Mg and K are shown in Figures
28 and 29, respectively. The Mg pattern was a very close copy of the Ca pattern
in Figure 23, except that the Mg fluxes were a factor of 5 lower, ranging from
about 0.2 to 1.2 kg/ha. The primary maxima occurred at Chicago and over the S
shore of Lake Erie. The overall pattern for K was similar, except for a strong
maximum over W Lake Erie and additional maxima over the SW shore of Lake
Superior, Green Bay, S Lake Huron, and E Lake Ontario.
Annual wet deposition fluxes of Cd and Pb, based only on 1983 GLAD data,
are shown in Figures 30 and 31, respectively. GLAD network sites used in the
analyses are shown in both figures. As for other spatial analyses, only sites
meeting the screening criteria were used. This explains why the sites that
appear in these two figures are not necessarily the same as in others. Minimum
fluxes of both metals occurred in the N areas of the Great Lakes basin. Lowest
measured annual Cd fluxes were about 1 g/ha, with maximum values in excess of 6
g/ha. Annual wet fluxes of Pb ranged from less than 20 to more than 100 g/ha.
Maximum Cd fluxes occurred over the middle and the S tip of Lake Michigan, much
of Lake Ontario, and both the W and E extremes of Lake Erie. Pb flux maxima
occurred in the Chicago area, E Lake Erie, and much of Lake Ontario.
For most ions, minimum wet fluxes occurred over Lake Superior and the N
portions of Lakes Michigan and Huron. The locations of flux maxima varied
somewhat from ion to ion, but the Chicago area and various locations along the
S shore of Lake Erie were often included. In general, the wet deposition
patterns may be viewed as the product of the regional patterns of precipitation
depth and ion concentrations in precipitation, modified locally by strong
43
-------
sources. Since precipitation increases somewhat from N to S, and concentrations
in general also increased from N to S or NW to SE, a rather strong increase in
deposition also occurred from N or NW to S or SE.
The additional spatial resolution provided by the GLAD sites makes it
possible to observe deposition maxima near strong source regions. These local
maxima are important in estimating lake loadings. Thus, the combined NADP and
GLAD data should provide better estimates of lake loadings of the major ions
than those computed from NADP data alone. Nevertheless, uncertainty still
exists regarding the dimensions of the areas represented by individual sites in
urban or industrial locations, as well as in extrapolating measurements made on
land to off-shore locations.
The deposition patterns produced by objective analysis are determined by
the number and location of the sampling sites, as well as the objective rules
used for interpolation between observations. Over southern Lake Michigan, for
example, the patterns were largely determined by three observations along the
Chicago lakefront and one at Benton Harbor, across the lake. The analysis
method ignores the fact that the lake is there, i.e., that sources are
effectively absent, and that lake-related meteorological processes may affect
deposition patterns over the lake.
Observations of deposition patterns east of St. Louis, based on bulk
precipitation sampling (Gatz, 1980a, 1980b), showed that pollutants with urban
sources were deposited in concentrations more than 5 times the regional mean
within about 20 km of their suspected sources. At greater distances,
deposition decreased rapidly. In contrast, Figure 31 shows a 100 g/ha isoline
of Pb deposition extending more than half of the roughly 110 km distance
between Chicago and Benton Harbor. The references above sugge'st that
deposition much higher than that shown in Figure 31 may have occurred over a
small area close to the Chicago shore, along with a broad area farther out in
the lake where deposition was lower than shown in Figure 31. However, the
magnitude of the error in the objective analysis, and even its sign, is not
obvious. Similar uncertainty exists regarding deposition gradients near source
areas on the other lakes. Additional research on the variation of wet
deposition with distance from sources will be needed to improve methods of
estimating lake loadings from land-based measurements.
3.4.2. Effects on Deposition Estimates of Closing GLAD Sites
The improved resolution provided by the GLAD network data for 1982-1985
may have diminished with the closing of 16 GLAD sites in January 1986. This
section examines the effects on deposition estimates of closing these sites.
To simulate the effects of reducing the number of sites in the GLAD
network, we computed deposition with and without selected sites. Differences
were characterized by 1) the resulting changes in deposition flux patterns or
2) the changes in lake loadings, or both. These differences were used to
estimate the effects of 1) the closing of 16 GLAD sites in January 1986, and 2)
the possible closing of selected additional GLAD sites in the future.
44
-------
Spatial distributions of the differences in wet deposition fluxes for
1982-1983, computed with and without the sites closed in 1986, are shown in
Figures 32-35 as percentages of the values computed for the full (pre-1986)
network. Differences were calculated by subtracting the grid point data field
resulting from the objective analysis of the reduced network from the grid
point data field from the full network, and converting these differences to
percentages. For all four ions examined, both positive (higher deposition in
the reduced network) and negative (lower deposition in the reduced network)
changes occurred locally near the sites removed. The increases ranged as high
as 35%, and the decreases as high as 45% in limited areas. Negative changes
occurred where the grid point field of deposition fluxes from the combined GLAD
and NADP networks was reduced locally by the removal of a data point (i.e., a
site) from the objective analysis. Positive changes occurred where the grid
point field was increased locally when a site was removed. The magnitude of
the percent differences was largest for Ca, which is consistent with the
observation that Ca had very large local maxima at urban GLAD sites.
The overall net changes in lake loadings estimates for each of the five
lakes are shown in Table 10. These loading estimates resulted from integrating
the over-lake deposition fluxes for each Lake. The changes were predominantly
negative; that is, the loadings computed from the reduced network were smaller
than those from the full network. This again is consistent with the fact that
many of the GLAD sites were in urban and suburban areas, where local effects
were apparent in the concentration and deposition patterns discussed earlier.
For 804, the changes were less than +5%, except for a -12.6% change in Lake
Ontario. The same was true for NC>3, except for a -5.2% change, also in Lake
Ontario. For NH4, the changes ranged from -1.4% to -14.2%, with the largest
change occurring in Lake Michigan and the next largest (-10.8%) in Lake Erie.
For Ca, the changes ranged from +4.3% (Lake Erie) to -26.8% (Lake Superior),
the largest change observed for any of the four ions. Changes of more than 10%
(both negative) also occurred in Lakes Huron and Ontario.
Table 11 shows the simulated effects on computed lake loadings of closing
additional existing sites. The table gives values of percent change in the
loadings of 804, N03, NH4, and Ca to Lakes Superior, Michigan, Erie, and
Ontario, based on loadings computed with and without certain sites (or
combinations of sites) in the data set. The reference data set was the combined
GLAD and NADP networks in operation after January 1986.
The upper portion of Table 11 shows percent differences in loadings to
Lake Superior on the assumption that the Hovland (QA), Cornucopia (QE), and
Ontonagan (QF) sites were closed--either individually, or in all possible
combinations of two and three sites. For S04, removal of sites from the data
set in any possible combination resulted in increases in the computed loadings
to Lake Superior, although the maximum increase was only 2.5% for the case
where sites QE and QF were removed.
For N03, NH4, and Ca, the computed changes were either increases or
decreases, depending on which sites or combinations of sites were removed from
the data set. For N03, the largest change was a decrease of 5.1%, for the case
where all three sites were removed. Removal of all three sites also had the
45
-------
Differences in SO4 Deposition
(percent)
GLAD sites closed In January
1986
Figure 32. Spatial distribution of percent differences that
result from calculation of the SO, deposition flux
from all the sites compared to those remaining open
after January 1986.
Differences in Ca Deposition
(percent)
GLAD sites closed in January
1986
Figure 33. Spatial distribution of percent differences that result
from calculation of the Ca deposition flux from all the
sites compared to those remaining open after January 1986.
46
-------
5
Differences in NOj Deposition
(percent)
GLAD sites clowd in Janutry
1986
Figure 34. Spatial distribution of percent differences that result
from calculation of the NO-j deposition flux from all the
sites compared to those remaining open after January 1986.
Differences In NH4 Deposition
(percent)
GLAD sites closed in Jenuiry
1986
Figure 35. Spatial distribution of percent differences that result
from calculation of the NH^ deposition flux from all the
sites compared to those remaining open after January 1986,
47
-------
Table 10. Comparisons of Wet-Only Atmospheric Loadings (1000s of Tonnes/yr)
to the Great Lakes from 1) All Valid NADP and GLAD Sites
Operating in 1982-83 with 2) All Valid NADP Sites Plus the Valid
GLAD Sites that Remained Open after January 1986.
Lake
Superior
Michigan
Huron
Erie
Ontario
Ion
504
NO 3
NH4
Ca
S04
N03
NH4
Ca
S04
N03
NH4
Ca
S04
NO 3
NH4
Ca
S04
NO 3
NH4
Ca
NADP + GLAD
(sites before 1-86)
104
81.2
22.9
18.6
118
90.3
22.1
16.7
116
83.4
20.6
13.8
90.3
47.0
11.1
12.1
60.7
37.3
8.07
5.13
NADP + GLAD
(sites after 1-86)
109
78.6
22.6
13.6
115
90.3
18.9
15.7
114
83.5
19.9
12.1
88.9
45.4
9.93
12.7
53.1
35.3
7.40
4.31
Percent
difference
+4.8
-3.2
-1.4
-26.8
-2.6
0.
-14.2
-6.1
-2.1
+0.1
-3.2
-12.4
-1.6
-3.5
-10.8
+4.3
-12.6
-5.2
-8.4
-16.0
48
-------
Table 11. Effect of Closing Additional GLAD Sites, Compared (Percent
Change) to the Network Still Open after January 1986.
S04 N03 NH^ Ca
Site(s) closed
Lake Superior
Hovland (QA)
Cornucopia (QE)
Ontonagan (QF)
QA + QE
QA + QF
QE + QF
QA + QE + QF
Lake Michigan
Benton Harbor (QT)
Lake Erie
Erie (SQ)
Lake Ontario
0.
0.9
0.8
0.9
0.5
2.5
2.3
1.9
-6.2
-3.9
-0.1
0.6
-4.3
-4.1
0.8
-5.1
-1.5
-4.5
-5.5
-2.0
1.3
-8.2
-4.8
-0.8
-9.0
-0.5
-7.7
1.8
-1.0
-2.3
0.9
-1.3
-4.5
-4.7
-0.7
-11.7
Erie (SQ) 0. 0.1 0.2 0.5
49
-------
maximum impact on computed NH4 loadings, a decrease of 9.0%, and computed Ca
loadings, a decrease of 4.7%.
Table 11 also shows the effect on computed loadings to Lake Michigan of
removing the Benton Harbor (QT) site. Only minor changes occurred, ranging from
a decrease of 1.5% in the N03 loading to an increase of 1.9% in the 804
loading.
The biggest effect on any lake from removing sites occurred on Lake Erie,
when the Erie (SQ) site was removed. Smaller loadings were computed for all
four ions without the urban Erie site in the data set. The decreases ranged
from 4.5% for N03 to 11.7% for Ca.
The effects on Lake Ontario loadings from removing the Erie site are also
shown. All changes were either zero or very small increases, the maximum change
being an increase of 0.5% in the computed Ca loading. The change in the loading
estimates for Lake Ontario from the closing of SQ was the result of the
objective analysis procedures used to calculate deposition fluxes. The value
calculated at each grid point was based on the values at the five nearest
measurement sites. Calculation of one or more grid box fluxes over Lake Ontario
was thus affected by measurements at the SQ site. Removal of SQ eliminated its
use in the calculations and this resulted in a slight change in the loadings
estimates, which were calculated by adding all the over-lake grid box fluxes.
The results in Table 11 appear to vary with the observed variability of
deposition over the various lakes. Removal of a single site from Lake Erie,
which shows highly variable deposition because of its highly urbanized
shoreline, resulted in changes of about 5-12% in computed lake loadings. On the
other hand, three sites could be removed from Lake Superior, where deposition
patterns are relatively uniform, before comparable changes were seen.
3.4.3. Atmospheric Loadings to the Lakes and Comparison to Previous Estimates
3.4.3.1. Results
A summary of atmospheric wet-only deposition loadings to the Great Lakes
from the combined GLAD-NADP networks is given in Table 12. Previous results
based on modeling (Acres Consulting Services, Ltd., 1975, 1977) and on bulk
precipitation measurements (Acres Consulting Services, Ltd., 1975, 1977;
Eisenreich et al., 1977) are given in Table 12 for comparison. Also shown for
comparison are non-atmospheric loadings to each lake. Most of those shown were
computed from the total lake loadings of Upchurch (1976) by removing his
estimated contributions from precipitation. However, additional values from
other sources are also provided for some of the lakes.
For sulfate, the present (GLAD-NADP) wet-only loading estimate was about
35% of that of the Acres model, which includes both wet and dry deposition, in
four of the five lakes, and about 50% of it in Lake Superior. The present wet-
only loadings estimates ranged from about 50% to 90% of those based on previous
bulk precipitation measurements. Thus, the present estimates of wet-only
sulfate deposition were smaller than: 1) previous modeling estimates, which
50
-------
Table 12. Atmospheric Loadings (1000s of Tonnes/yr) to the Lakes from
GLAD-NADP Data, and Comparisons with Previous Estimates and
Non-Atmospheric Loadings
Lake
Superior
Michigan
Huron
Ion
so.
Cl
Ca
Mg
Na
K
N03-N
NH4-N
Total
N
so4
Cl
Ca
Mg
Na
K
N03-N
NH4-N
Total
N
so4
Cl
Ca
Mg
Na
K
N03-N
NH4-N
Total
NADP4GLAD
(wet only)
104
7.5
18.6
3.4
3.5
3.1
18.4
17.8
36.2
118
10
16.7
3.3
3.4
2.8
20.4
17.2
37.6
116
5.6
13.8
2.6
2.4
2.7
18.8
16.0
34.8
Acres model3
(wet + dry)
210
0.19
1.3
0.60
0.37
1.1
17
330
NAC
1.8
0.81
0.50
1.5
42
380
0.20
1.6
0.73
0.47
1.4
Precipitation
chemistry3
(bulk)
220
55
33
5.6
15
13
56
135d
82d
103d
22d
16d
9.6d
NA
230
49
280
8.2
45
32
Estimated"
non atmospheric
loadings
98
670
10
81
1650
294d
259d
6.6
170
NA
5.9
N
31
52
51
-------
Table 12. (continued)
Lake Ion
Erie S04
Cl
Ca
Mg
Na
K
N03-N
NH4-N
Total
N
Ontario SO^
Cl
Ca
Mg
Na
K
NADP+GLAD
(wet only)
90
11
12
2.2
3.3
2.2
10.6
8.6
19.2
61
4.9
5.1
1.0
1.5
1.1
Acres model3
(wet + dry)
270
NA
1.2
0.55
0.37
1.1
29
120
NA
0.62
0.29
0.19
0.55
Precipitation
chemistry3
(bulk)
120
NA
23
6.6
13
22
19
88
102f
15f
32
51f
4.1
8.1f
19
25f
3.3
6.1f
Estimated0
non-atmospheric
loadings
400
2300
27
45008
240
43008
3200
650Q8
14508
2000g
2108
N03-N
NH4-N
Total
N
8.4
6.3
14.7
14 21
10
a Source: Acres Consulting Services, Ltd. (1975) for Lakes Superior and
Huron; Acres Consulting Services, Ltd. (1977) for the other lakes,
except as noted.
Source: Upchurch (1976), except as noted. Does not include wet deposition
or inputs from upstream Great Lakes.
c NA = not available.
d Eisenreich et al., (1977).
e Sum of tributary and erosion inputs from Eisenreich (1980), based on
Robbins et al.. (1972), Schmidt (1977), Monteith and Sonzogni (1976),
and Fitchko and Hutchinson (1975).
f Shiorai and Kuntz, (1973).
8 Niagara River loadings to Lake Ontario, from Shiomi and Kuntz (1973).
52
-------
included both wet and dry deposition, and 2) previous bulk (open-bucket)
precipitation measurements, which included wet deposition as well as dry
materials deposited with generally unknown efficiencies.
Only one estimate of non-atmospheric lake loading was available for 864,
that being for Lake Ontario. Comparison of the atmospheric and non-atmospheric
loadings indicates that atmospheric deposition of 804, at least for Lake
Ontario, was negligible compared to other sources.
For chloride, the GLAD-NADP wet-only loading estimates (Table 12) exceeded
the Acres modeling estimates by factors of about 30 to 40 for Lakes Superior
and Huron, the only two lakes for which Acres estimates for Cl were available.
On the other hand, the wet-only loadings estimates were only about 1/8 of the
bulk precipitation loadings estimates for Lakes Superior, Michigan, and Huron,
and about 1/3 the reported value for Lake Ontario. This seems to indicate that
the Cl loadings estimated by the Acres model were much too low, and that the
bulk precipitation measurements were strongly affected by either dry deposition
or perhaps contamination.
The wet-only Cl loadings represent about 2-12% of the non-atmospheric
loadings of Upchurch (1976) for the various lakes. However, it should be noted
that an independent estimate (Shiomi and Kuntz, 1973) of non-atmospheric Cl
loading of Lake Ontario is far higher than that of Upchurch.
For calcium, the present wet-only loadings estimates were consistently
higher than the model results by a factor of about 10 for all the lakes.
Conversely, the present estimates ranged from 5-56% of the previous estimates
based on bulk precipitation measurements. Differences between wet-only and
bulk precipitation loadings were much as expected, because atmospheric Ca is
primarily on large particles, for which dry deposition can account for a large
fraction (probably more than half) of the total deposition. It is worth noting
that dry deposition of large particles to open buckets may be a reasonable
approximation of ambient dry deposition to natural surfaces, so that the bulk
deposition measurements may be a reasonable estimate of combined wet and dry
deposition for Ca.
The order-of-magnitude differences between the present estimates and the
model suggest major deficiencies in the model results for Ca. Inadequate
emissions data was a likely cause. In any case, as we shall see next,
atmospheric loadings of Ca were a relatively minor source, compared to those
from non-atmospheric sources.
Comparison of atmospheric and non-atmospheric inputs of Ca to the lakes
shows that the atmosphere was not a major source of Ca to the lakes (unless dry
deposition made a large contribution not measured in the bulk collectors used
previously). The present wet-only loadings represented only 0.2% (Lake
Ontario) to 2.8% (Lake Superior) of the estimated non-atmospheric inputs shown
in Table 12. However, notice again the wide disparity in non-atmospheric
loadings estimates by various authors for Lakes Michigan and Ontario.
For magnesium, the new wet-only loadings again exceeded those from the
Acres model by factors of 3-6 for the several lakes and were exceeded by those
53
-------
of the previous bulk measurements by about the same range of factors. From the
available estimates of the non-atmospheric Mg loadings for Lakes Michigan and
Ontario, it appears that wet deposition accounted for only about 1% of the
total.
For sodium, as for Cl, Ca, and Mg, the loadings from the model were much
too small, and the bulk measurements must have included either a large dry
deposition component or suffered from contamination, or both. There was a non-
atmospheric loadings estimate only for Lake Ontario, but it indicated that wet
deposition contributed very little to the total lake loadings of Na.
For potassium, the wet-only loadings again exceeded those of the Acres
model, but in this case by a factor of only about 2. As before, the loadings
estimates from the bulk precipitation measurements were much higher, in this
case by factors of 3-10. Again, the one estimate of non-atmospheric loadings
(for Lake Ontario) suggests that the wet-only contribution to total lake
loadings was very small.
Comparison of nitrogen loadings estimates required summing the nitrate N
(N03-N) and ammonium N (Nlfy-N) (see Table 12) to get a total N comparable to
the total N reported for the modeling and bulk precipitation measurements
(Acres Consulting Services, Ltd., 1975). Model estimates and bulk measurements
differed by a factor of only about 2, the present wet-only estimates falling
somewhere in between. The only measure of nitrogen available from non-
atmospheric sources was N03-N. The present estimates of wet-only input
exceeded Upchurch's non-atmospheric inputs in four of the five lakes, which
suggests that atmospheric sources were relatively important for nitrogen.
Atmospheric loadings of metals, particularly the potentially toxic ones
such as Hg, Pb, and Cd, are of much interest. These metals were not measured
by the NADP network, and, as indicated earlier, GLAD for 1982 were insufficient
for loadings estimates. The same was also true for Hg in 1983. However, the
measurements of Cd and Pb for 1983 were considerably improved over those for
1982, and concentrations above detection limits were reported for most samples.
Thus, we have computed loadings to the five lakes for these two metals from
GLAD measurements (only). The results are given in Table 13, along with
previous atmospheric loadings estimates from bulk precipitation measurements
and from models and a few non-atmospheric loadings estimates.
As we saw in the case of the major ions, the new loadings estimates for
wet-only deposition were in most cases much lower than previous estimates of
wet and dry input, based on both modeling results and bulk precipitation
measurements. For Lake Superior, the wet-only loadings estimates for Pb and Cd
were only a quarter to a third of the previous estimates.
For Lake Michigan, the wet-only loadings estimates for Pb were about a
third to a fifth of the bulk measurements of Eisenreich (1980) and the IJC
(1977) (Acres) model results. However, the new estimate slightly exceeded the
early model estimates of combined wet and dry deposition of Winchester and
Nifong (1971) and Gatz (1975). It also exceeded the tributary inputs
calculated by Winchester and Nifong (1971), but was less than the combined
tributary and shoreline erosion loadings of Eisenreich (1980). For Cd in Lake
54
-------
Table 13. Loadings Estimates (Tonnes/Yr) to the Great Lakes for Cd
and Pb, Based on GLAD 1983 Concentration Measurements and 30-yr
Mean Precipitation, Compared with Previous Estimates from the
Literature Based on Bulk Precipitation Measurements and Modeling
and with Estimates of Non-Atmospheric Inputs.
Superior
Pb Cd
Michigan
Pb Cd
GLAD wet-only measurements (1983)
This study 170 12 240 15
Huron
Pb Cd
Erie
Pb Cd
Ontario
Pb Cd
214
8.3 142
8.7 125
9.5
Bulk precipitation measurements
IJC (1977) 650
Eisenreich (1980)
Shiomi and Kuntz
(1973)
55
NA
640
NA
11
780 79 2200 150
280 45
255 17
Model estimates (wet + dry)
IJC (1977)
Winchester and
Nifong (1971)
Gatz (1975)
780 34 1100 48
220
200
1.2
3.2
960 39
650 25
440 18
Non-atmospheric loading estimates
Eisenreich (1980)
Winchester and
Nifong (1971)
Shiomi and Kuntz (1973)
420a 87a
100'
700(
a Tributaries, plus shoreline erosion.
b Tributary input.
c Niagara River input to Lake Ontario.
55
-------
Michigan, the new wet-only estimate slightly exceeded that based on
Eisenreich's (1980) bulk measurements, was about a third of the IJC (1977)
model estimate, and greatly exceeded the two earlier model estimates. The new
wet-only estimate is about 17% of Eisenreich's (1980) estimate of Cd loading
from tributaries and shoreline erosion.
For Lakes Huron, Erie, and Ontario, the new wet-only estimates for Pb were
only about 25-50% of most of the previous estimates, and only 5% of the 2200
metric tonnes/yr estimate for Lake Erie based on bulk precipitation
measurements. The estimated wet-only loading of Pb to Lake Ontario was about
17% of the Niagara River loading. The new estimates for wet-only Cd loading of
the same three lakes ranged from 5-50% of the previous estimates.
Finally, wet-only loadings estimates from this study, presented in Tables
12 and 13, are further summarized and compared in Table 14 with previous
estimates of loadings from bulk sampling and modeling. Comparisons are made of
the ranges of ratios of the previous estimates for the various lakes to the
current wet-only estimates. For example, for Cd (Table 13) the lowest IJC
model/wet ratio, about 2, occurred for Lake Ontario and the highest, about 5,
for Lake Huron. Similarly, the lowest bulk/wet ratio, about 0.8, occurred for
Lake Michigan, and the highest, about 17, occurred for Lake Erie.
Table 14. Comparison of current estimates of wet-only lake loadings with
previous estimates based on modeling (IJC, 1977) and bulk sampling.
Species
Pb
Cd
Model/Wet
4 - 5
2 - 5
Bulk/Wet
2 - 15
0.8 - 17
Ca
Mg
K
Na
Cl
S04
N
0.
0.
0.
0.
0.
07
2
3
1
0
1
5
- 0.
- 0.
- 0.
- 0.
.03
- 3
- 1.
1
3
5
2
5
2 -
2 -
3 -
4 -
3 -
0.7 -
1 -
6
7
10
20
8
2
1.5
The species measured in precipitation have been grouped according to
patterns of comparison with previous results. For example, the results for Pb
and Cd were similar in that in general the loadings estimated by both previous
methods exceeded the currently estimated loadings by factors of two or more.
The same was true for the elements grouped in the middle of Table 14 in
comparison with the bulk-sample estimates, but for this group the opposite was
true in comparison to the previous model estimates. For 864 and N, at the
bottom of the table, the current loadings estimates were mostly within a factor
of two of both previous estimates.
56
-------
3.4.3.2. Discussion
There are at least two very striking results here: 1) for many elements
the current wet-only loadings estimates were much smaller than those made
earlier based on bulk sampling, and 2) these large differences were not seen
for SO^ and N. Another is that the wet-only loadings for some elements were
substantially higher than the corresponding estimates from the Acres model.
This very likely resulted from a model input of these elements that was
inadequate or incomplete.
Several possible explanations for the differences between bulk and wet-
only loadings estimates come to mind, and depending on which is true (or which
combination is true), there could be important implications regarding the
direction of future research and policy for the Great Lakes. It seems possible
that the differences may have been caused, at least in part, by one or more of
the following:
1) Analytical bias, or loss or gain of ions from container walls,
2) Differences in precipitation amount,
3) Contamination of the bulk collections by debris (bird droppings,
etc.) and/or resuspended local surface dust,
4) Valid (i.e., non-contaminant) dry deposition to the bulk collectors,
or
5) Reductions in pollutant emissions between observation periods.
Since several analytical laboratories analyzed samples from the bulk
precipitation sampling network (IJC, 1977) and an entirely different laboratory
analyzed samples from the GLAD wet-only network, one possible explanation for
differences is analytical bias. Without a direct comparison of laboratory
performance on standard samples or split samples, such bias is difficult to
quantify. In any case, analytical bias is not likely to be large enough to
explain the differences of factors of two or more observed here.
A related possible problem is the adsorption or desorption of metal ions
by or from container walls. For adsorption on walls to have caused the observed
differences, the losses would had to have occurred in the polyethylene bags or
bottles used respectively to collect and ship the samples. However, Chan et
al^, (1983) have reported negligible losses of Pb and Cd to polyethylene on
one-day contact, and, on 29-day contact, minimal loss of Pb and about 10% loss
of Cd. Sample contamination by desorption of metal impurities from container
walls is also possible, but it is quite unlikely that the massive contamination
required to cause the observed differences would have gone undetected by normal
laboratory quality control procedures or blank measurements. Thus, it appears
unlikely that interactions between precipitation samples and container walls
could have caused the observed differences.
A procedural difference in the computation of loadings from the respective
bulk and wet-only measurements is another possible reason for the observed
differences in lake loadings. The bulk network was in operation during 1973 and
1974 (IJC, 1977), and loadings calculations were based on measured
precipitation during the period of operation. In contrast, the observed
57
-------
(weighted mean) wet-only concentrations were converted to lake loadings using
the 30-yr mean annual precipitation for each lake, as described earlier. If
actual precipitation during the bulk network operation was larger than the 30-
yr means, this would cause the loadings estimated from the bulk sampling data
to be higher than those from the wet-only data, even if the concentrations were
identical.
To evaluate this potential effect, annual precipitation fluxes to each of
the five lakes for 1973 and 1974 were computed from precipitation data supplied
by H. Hartman (see footnote to Table 9). The largest departure from the 30-yr
mean for any lake in either year was +6.0% for Lake Michigan in 1973. Thus,
this difference in methods of estimating lake loadings cannot account for the
observed large (factor of 2 and greater) differences in loadings.
Another possible explanation is that the bulk sampler collections included
sizeable contributions from local surface dust. This dust would include wind-
blown emissions from soils and both paved and unpaved roads. Thus, it would
naturally contain most of the major cations, and at times also Na and Cl from
winter road salting. In urban and industrial locations, it might also contain
high concentrations of Pb, Cd, and other pollutant metals, which are known to
occur in high abundance sorbed on surface dust after previous wet or dry
deposition (Hopke et al.. 1980; Harrison et al.. 1981). On the other hand, any
previously deposited soluble anions in surface dust would be leached into
deeper soil layers by subsequent rainfall, so deposition of resuspended local
surface dust in bulk samplers would not cause significant differences between
bulk and wet-only collections for 804 and N. This is consistent with the
observations in Table 14.
Special observations were conducted as part of the IJC bulk sampler
network to determine whether the bulk samplers collected resuspended surface
dust (J.R. Kramer, personal communication, 1987). The observations consisted of
monthly bulk precipitation samples collected in samplers 1) on the roof of an 8
ft high field shelter, 2) on top of an adjacent 30-75 ft tower, and 3) on a
buoy anchored 6 mi offshore. These observations were made in clearings at three
lakeshore sites--two relatively isolated, and one in a town with a population
of about 7500. One of the isolated sites was near Red Rock, Ontario, on the
northern shore of Lake Superior. The other isolated site was on Duck Island,
near Manitoulin Island in northern Lake Huron. The third site was at the town
of Goderich, Ontario, on the eastern shore of Lake Huron. A statistical
comparison of constituent concentrations measured in the various types of
samplers indicated that small amounts of reentrained surface dust were entering
the bulk samplers at times (J. R. Kramer, personal communication, 1987),
although an independent analysis of the data (Acres Consulting Services, Ltd,
1975) pointed out that the higher loadings often measured by the samplers on
towers compared to those near the ground was inconsistent with this
interpretation.
The special observations of the IJC network also included simultaneous
wet-only and bulk sampling at the three sites described above. Data on
elemental and ionic concentrations, both total and filtered, are available
(J.R. Kramer, personal communication, 1987) for both types of samplers. These
58
-------
data may be suitable for clarifying the role of local resuspended dust
deposited in the bulk samplers.
If contributions of local surface dust to the previous bulk collections
are a major cause of the observed differences between bulk and wet-only
collections, some may argue that such deposits should be included when tallying
atmospheric deposition inputs to the lakes. There are several reasons why such
actions would be risky. It is true, of course, that some surface dust
resuspended over land areas will travel over, and fall into the lakes, but the
affected areas of the lakes are not known with any accuracy. The flux to the
lakes will be greatest at the shoreline and decrease as some unknown function
of distance away from shore. The second reason is the absence of a proven
method for estimating the "locally resuspended" portion of the deposition
measured in bulk samples. A third reason is the lack of a procedure to estimate
deposition to a natural surface from that measured in a bulk collector.
If indeed local surface dust accounts for much of the difference between
the lake loadings estimated from bulk and wet-only collections, then it is
likely that the contribution of atmospheric deposition to the total input of
metals like Pb and Cd has been over-estimated, although it is not possible to
say how large the overestimate is.
The possibility of this explanation should be explored further through
detailed examination and comparison of bulk/wet-only differences in metal
deposition at urban and remote sites. These differences should be greatest at
urban locations where the surface dust would be contaminated with these metals,
and probably negligible at remote sites. Side-by-side comparison of the
respective types of samplers should also be carried out in both urban and
remote locations.
Another possible explanation is that the observed differences between the
lake loadings estimated previously from bulk sampling and the current ones from
wet-only sampling represent true dry deposition (i.e., other than locally
resuspended surface dust) to the bulk samplers. If this is the case, then dry
deposition to the lakes may be a much larger fraction of total deposition than
previously thought. However, from the measurements available, it is impossible
to say whether that is true, since there is no proven method for using
surrogate surface dry deposition measurements to predict deposition to the
surfaces of the Great Lakes.
A few independent estimates of dry deposition of Pb and Cd are available
in the literature for comparison. Sievering et al. (1984) estimated a dry
deposition input of 200-500 tonnes/yr of Pb to the S basin of Lake Michigan;
this compares reasonably well to the difference (400 tonnes/yr) between bulk
and wet-only inputs to all of Lake Michigan in Table 13. However, the
comparison is much less consistent in the case of Cd. The Cd dry deposition
loadings available in the literature are also for the S basin of Lake Michigan.
The estimates are 1.7 tonnes/yr (Gatz, 1975) and 2.2 tonnes/yr (Tisue and
Fingleton, 1984). Comparison of these estimates to the differences between
loadings estimated from bulk and wet-only sampling are not possible, however,
for reasons explained next.
59
-------
Lake Michigan presents an anomaly with respect to the comparison of
loadings estimates from wet-only and bulk precipitation measurements in that it
is the only lake where the wet-only estimate actually exceeded the previous
bulk-sampling estimate (15 vs 11 tonnes/yr, Table 13). It may be significant
that the bulk precipitation measurements for Lake Michigan were carried out by
a different group (Eisenreich, 1980) than those for the other four lakes (IJC,
1977), for which the loadings based on bulk sampling exceeded those from wet-
only sampling by amounts ranging from 35-141 tonnes/yr (Table 13).
It is clear that there are only a few estimates of dry deposition of
metals to the Great Lakes. It is equally clear to workers in this field that,
although such measurements are badly needed to understand all the processes of
atmospheric deposition to the lakes, the methods currently available are
inadequate. Further research is needed to develop the needed methods.
Another possible explanation of the differences between current and
previous estimates of lake loadings is that the deposition data merely reflect
the reductions in emissions that have taken place during the time between the
respective observations. For Pb, Eisenreich et al. (1986) reported marked
decreases in concentrations and deposition in precipitation in Minnesota
between 1979 and 1983, a period when the amounts of Pb used in gasoline also
dropped substantially. No similar reduction in a major Cd source is known, but
gradual reductions in Cd emissions may also have occurred between the mid-1970s
and the early 1980s due to economic factors and the gradual implementation of
improved emission control technology.
On the other hand, this explanation is not likely to apply to the alkali
and alkaline earth elements Na, Mg, K, and Ca, for which previous bulk
precipitation loadings also greatly exceeded recent wet-only loadings. It is
possible, of course, and we suggest that it is quite likely, that emission
reductions explain the most of the differences in Pb (and perhaps Cd) loadings,
while the exclusion of local resuspended dust from the wet-only samples (but
not the bulk samples) explains most of the observed differences for the alkali
and alkaline earth elements.
Uncertainties in computed loadings near strong source areas, arising from
the objective analysis method used, were discussed earlier in this report. The
sign and magnitude of the possible errors cannot now be estimated. Judging
from the rather limited lake areas involved, however, it appears quite unlikely
that they could account for the observed many-fold differences between our wet-
only loadings estimates and those based on previous bulk sampling.
4. CONCLUSIONS
The Great Lakes precipitation chemistry data analyzed in this report
represent the first two full calendar years (1982 and 1983) of data from an
ongoing network sampling operation. A number of important conclusions can be
drawn at this point:
1. GLAD pH measurements were biased low during 1982-1983, and should not
be used.
60
-------
2. Constituent concentrations at GLAD sites exceeded their counterparts at
the nearest NADP sites (at the 1% significance level) much more than expected
by chance for 804, Ca, and NH4, and somewhat more than expected by chance for
N03- Most of these differences involved comparisons of an urban GLAD site in
Erie, Pennsylvania, and a suburban GLAD site in Dunkirk, New York. These
differences could reflect true spatial gradients, or be related to differences
in siting criteria or local sources. In the absence of a side-by-side
comparison of NADP and GLAD sites, it is not possible to quantify sampling or
analytical biases between the two networks for the four ions tested, 804, Ca,
N03, and NH4.
3. GLAD data add useful spatial resolution to other available
precipitation chemistry data sets. Based on valid GLAD and NADP data from 1982-
83, the closing of 16 GLAD sites (actually closed in January, 1986) generally
resulted in a decrease in the loadings estimates to each of the 5 Lakes for
804, Ca, NC>3, and Nlfy. Decreases exceeded 10% for Ca (26.8%) in Lake Superior,
NH4 (14.2%) in Lake Michigan, Ca (12.4%) in Lake Huron, NH4 (10.8%) in Lake
Erie, and 804 (12.6%) and Ca (16.0%) in Lake Ontario. Closing of additional
sites would result in further reductions in the spatial resolution of
deposition and in the ability to quantify the loadings.
4. Spatial distributions of wet deposition fluxes show that annual values
were 2-10 times higher in the S or SE portions of the network than in the N or
NW portions. This results from both an increase in average precipitation amount
and an increase in constituent concentrations from N or NW to 8 or SE. This is
a general feature of all the deposition flux patterns, though there are
important (but not consistent) exceptions at one or more sites for nearly all
ions.
5. Annual precipitation-only loadings of Pb to the five Great Lakes, based
on the 30-yr mean annual precipitation and 1983 GLAD concentrations, ranged
from 125 to 240 tonnes. These values are about 20% of previous model loadings
estimates of wet and dry deposition to the respective lakes, and 7-50% of
previous estimates made from bulk precipitation sampling data.
For Cd the range of the same loadings was 8 to 15 tonnes per lake per
year. This is about 20-30% of previous wet and dry loadings estimates from
models and 5-120% of loadings estimates from bulk precipitation data.
For Cl and the major cations Mg, K, Na, and Ca, the precipitation-only
loadings greatly exceed those from the model. They are much smaller than those
based on earlier bulk sampling, but they are relatively insignificant compared
to non-atmospheric loadings.
For 804 and N, the precipitation-only loadings were within a factor of
about 2 of those from the model and from earlier bulk sampling.
6. The marked differences between current (precipitation-only) loadings
estimates and earlier (bulk precipitation) estimates are not likely to be
accounted for by analytical bias, interactions with container walls, or
differences in precipitation amount.
61
-------
The observed large differences between present and earlier loadings
estimates for the cations and Cl and the small differences observed for 804 and
N are consistent with contamination of the bulk samples by resuspended surface
dust and (for Cl) episodic road salt spray. The soluble anions are not
constituents of the surface dust and would not accumulate in it from previous
wet or dry deposition.
The same explanation may account for some or all of the differences found
for Cd, as well. For Pb, however, at least a part of the differences are likely
to have been caused by recent reductions in Pb emissions from automobile
exhaust, which are due to the current limitations on the Pb content of
gasoline.
7. To the extent that these explanations cannot account for the observed
differences, we must attribute them to dry deposition (from non-local sources)
in the bulk collectors. If contaminant dry deposition were truly a minor
portion of the observed differences, then dry deposition would have to be a
major portion. In that case, one might begin to suspect that dry deposition to
the lakes is more important, relative to wet deposition, than has been thought
up to now. However, that could only be a very tentative suspicion, because we
do not know how to infer dry deposition to natural surfaces from dry deposits
in bucket collectors.
5. RECOMMENDATIONS FOR FURTHER RESEARCH
1. Additional GLAD data now available for 1984-1986 should be analyzed, to
help to verify or refute the conclusions based on the 1982-1983 data. However,
data reporting, editing, clean-up, and archiving procedures should be improved,
so that users can be provided with error-free data sets.
2. The GLAD network should be continued, with as many sites as practical,
to provide a spatial resolution of wet deposition consistent with that of
relevant major sources of toxic contaminants. In addition, research should be
conducted to improve estimates of over-water wet deposition fluxes from land-
based measurements, especially near major sources of important airborne
pollutants.
3. Additional research should be conducted to explain the differences
between lake loadings calculated from the earlier bulk precipitation
measurements and the current wet-only measurements. Analyses should be carried
out on any existing, but unanalyzed, data sets from co-located bulk and wet-
only collectors. Additional field comparisons of these samplers should be
conducted in both urban and remote locations.
4. Where possible, Canadian data on precipitation amount and chemical
composition should be combined with the GLAD data base as additional years of
data are added, so that subsequent analyses and loadings calculations are based
on data from both sides of all the Great Lakes.
62
-------
5. Additional research is needed to provide valid methods of measuring or
estimating dry deposition of both gaseous and particulate forms of toxic
pollutants to the Great Lakes.
6. To verify that the differences between GLAD and NADP data are not
related to sampling, analytical, or data handling procedures, or analytical
biases, a co-located GLAD and NADP site should be operated for at least one
year at an urban location. Differences, if any, at such a site are more likely
to be manifested than in a clean rural environment.
6. ACKNOWLEDGEMENTS
This work was funded by the Great Lakes National Program Office of the
U.S. EPA. Edward Klappenbach was the project manager. The work was
facilitated by many discussions with Mr. Klappenbach and with Dr. Jacob Snyder
of Bionetics, Inc., EPA's contractor for analysis of the samples. Discussions
with, and information provided by, Dr. Tom Murphy of DePaul University, and
Holly Hartman, of the NOAA Great Lakes Environmental Research Laboratory, were
also important to this work. Many of the data processing operations were
carried out using computer software at the Illinois State Water Survey provided
by previous grants and contracts, in particular those sponsored by the U.S.
Department of Energy and the National Atmospheric Deposition Program. Support
was also contributed by the State of Illinois.
63
-------
7. REFERENCES
Achtemeier, G.L., 1987: On the concept of varying influence radii for a
successive corrections objective analysis. Monthly Weather Review. 115.
1760-1771.
Achtemeier, G.L., P.H. Hildebrand, P.T. Schickedanz, B. Ackerman, S.A.
Changnon, and R.G. Semonin, 1977: Illinois Precipitation Enhancement
Program and Evaluation Techniques for the High Plains Cooperative Program.
Final Report, Contract 14-06-D-7191, Illinois State Water Survey,
Champaign, 6-12, Appendix B.
Acres Consulting Services, Ltd., 1975: Atmospheric loading of the upper Great
Lakes, Volume 2, Main Text, Report to Canada Centre for Inland Waters,
Acres Consulting Services, Ltd., Niagara Falls, Ontario, Table 4.5.
Acres Consulting Services, Ltd., 1977: Atmospheric loading of the lower Great
Lakes and the Great Lakes drainage basin, Report to Canada Centre for
Inland Waters, Acres Consulting Services, Ltd., Niagara Falls, Ontario, p.
43.
Barnes, S.L, 1964: A technique for maximizing details in numerical weather map
analysis. J. Appl. Meteorol.. 3, 396-409.
Barnes, S.L., 1973: Mesoscale objective map analysis using weighted time-
series observations. NOAA Tech. Memo. ERLNSSL-62. [NTIS COM-73-10781].
Bolsenga, S.J., 1979: Determining overwater precipitation from overland data:
the meteorological controversy analyzed. J. Great Lakes Res. . 5_, 301-311.
Bowersox, V.C., 1984: Data validation procedures for wet deposition samples at
the Central Analytical Laboratory of the NADP. In: Johnson, T.R. and
S.J. Penkala, Editors, Quality Assurance in Air Pollution Measurements.
Air Pollution Control Association, Pittsburgh, PA, pp 500-524.
Chan, W.H., F. Thomassini, and B. Loescher, 1983: An evaluation of sorption
properties of precipitation constituents on polyethylene surfaces.
Atmos.Environ.. 17(9), 1779-1785.
Changnon, S.A., and D.M.A. Jones, 1972: Review of the influences of the Great
Lakes on weather. Water Resour. Res.. 8, 360-371.
Cleveland, W.S., 1985: The Elements of Graphing Data. Wadsworth Advanced Book
Program, Monterey, CA, 323 pp.
Eisenreich, S.J., P.J. Emmling, and A.M. Beeton, 1977: Atmospheric loadings of
phosphorus and other chemicals to Lake Michigan. J. Great Lakes Research.
3(3-4), 291-304.
Eisenreich, S.J., 1980: Atmospheric input of trace elements to Lake Michigan.
Wat.. Air. Soil Pollut.. 13, 287-301.
-------
Eisenreich, S.J., N.A. Metzer, and N.R. Urban, 1986: Response of atmospheric
lead to decreased use of lead in gasoline. Environ. Sci. Technol.. 20,
171-174.
Fitchco, J., and T.C. Hutchinson, 1975: A comparative study of heavy metal
concentrations in river mouth sediments around the Great Lakes. J. Great
Lakes Research. 1(1), 46-78.
Gatz, D.F., 1975: Pollutant aerosol deposition into southern Lake Michigan.
Wat.. Air. Soil Pollut.. 5, 239-251.
Gatz, D.F., 1980a: An urban influence on deposition of sulfate and soluble
metals in summer rains. Chapter 27 In: Shriner, D.S., C.R. Richmond, and
S.E. Lindberg, Eds., Atmospheric Sulfur Deposition: Atmospheric Impact
and Health Effects. Ann Arbor Science Publishers/The Butterworth Group,
Ann Arbor, Michigan, pp 245-261.
Gatz, D.F., 1980b: Associations and mesoscale spatial relationships among
rainwater constituents, J. Geophys. Res. . 8_5(C10) , 5588-5598.
Great Lakes National Program Office, 1985: Atmospheric deposition and
precipitation sampling network station operator's manual. U.S. EPA, Great
Lakes National Program Office, 536 S. Clark St., Chicago, Illinois 60605
Great Lakes National Program Office, (undated): Standard operating procedure
for sample analysis and data reporting, atmospheric monitoring program,
EPA Contract #68-04-5038, Task 5; inorganic parameters. (Working Draft)
U.S. EPA, Great Lakes National Program Office, 536 S. Clark St., Chicago,
Illinois 60605
Harrison, R.M., D.P.H. Laxen, and S. J. Wilson, 1981: Chemical associations of
lead, cadmium, copper, and zinc in street dusts and roadside soils.
Environ. Sci. Technol.. 15, 1378-1383.
Hopke, P.K., R.E. Lamb, and D.F.S. Natusch, 1980: Multi-elemental
characterization of urban roadway dust. Environ. Sci. Technol.. 14(2),
164-172.
International Joint Commission, 1977: Atmospheric loading of the lower Great
Lakes and the Great Lakes drainage basin, International Joint Commission,
International Reference Group on Great Lakes Pollution from Land Use
Activities, Great Lakes Regional Office, Windsor, Ontario.
Lockard, J.M., 1987: Quality Assurance Report. NADP/NTN Deposition Monitoring.
Laboratory Operations. July 1978 through December 1983. Natural Resource
Ecology Laboratory, Fort Collins, CO 80523, p. 31.
Monteith, T.J., and W.G. Sonzogni, 1976: U.S. Great Lakes shoreline erosion
loadings. U.S. Task D Report on Great Lakes Pollution from Land Use
Activities, International Joint Commission.
65
-------
National Atmospheric Deposition Program, 1982: NADP Instruction Manual for Site
Operation. Natural Resource Ecology Laboratory, Colorado State University,
Fort Collins, CO 80523.
NADP Subcommittee Number 3, 1985: NADP Annual Data Summary: Precipitation
Chemistry in the United States. 1982. Natural Resource Ecology
Laboratory, Colorado State University, Fort Collins, CO, p 3.
National Climatic Center, 1982: Monthly Normals of Temperature, Precipitation,
and Heating and Cooling Degree Days, In: Climatography of the United
States. No. 81. National Climatic Center, NOAA, Asheville, NC.
Peden, M.E., S.R. Bachman, C.J. Brennan, B. Demir, K.O. James, B.W. Kaiser,
J.M. Lockard, J.E. Rothert, J. Sauer, L.M. Skowron, and M.J. Slater, 1986:
Methods for Collection and Analysis of Precipitation. Illinois State Water
Survey Contract Report 381, Champaign, IL 61820.
Robbins, J.A., E. Landstrom, and M. Wahlgren, 1972: Tributary inputs of soluble
trace metals to Lake Michigan. Proceedings of the 15th International
Conference on Great Lakes Research, pp 270-290.
SAS Institute, Inc., 1982: SAS User's Guide: Statistics. 1982 Edition, SAS
Institute, Inc., Gary, NC.
Schmidt, J., 1977: Selected metals in air particulates over Lake Michigan. M.S.
Thesis, University of Wisconsin-Madison.
Shiomi, M.T., and K.W. Kuntz, 1973: Great Lakes precipitation chemistry, Part
1. Lake Ontario basin. Proceedings of the 16th International Conference on
Great Lakes Research, pp 581-602.
Sievering, H., D.A. Dolske, V. Jensen, and R.L. Hughes, 1984: An experimental
study of lake loading by aerosol transport and dry deposition in the Lake
Erie Basin. Report EPA-905/2-84-002, U.S. EPA, Great Lakes National
Program Office, 536 S. Clark St., Chicago, Illinois 60605
Stensland, G.J., and V.C. Bowersox, 1984: A comparison of methods of computing
precipitation pH averages. Proceedings, 77th APCA Annual Meeting, Paper
No. 84-19.1, Air Pollution Control Association, Pittsburgh, PA.
Tisue, T., and D. Fingleton, 1984: Atmospheric inputs and the dynamics of
trace elements in Lake Michigan. In: Nriagn, J.O., and M.S. Simmons,
Eds., Toxic Contaminants in the Great Lakes. Vol. 14, Advances in
Environmental Science and Technology Series, John Wiley, New York, pp.
105-125.
Todd, O.K., Ed., 1970: The Water Encyclopedia. Water Information Center, Port
Washington, NY, p 127.
66
-------
Unified Deposition Data Base Committee (undated): A unified wet deposition
data base for eastern North America: data screening, calculation
procedures, and results for sulphates and nitrates (1980). Coordinated by
M. Lusis, Ontario Ministry of the Environment, Toronto.
Upchurch, S.B., 1976: Chemical characteristics of Great Lakes, In: Report of
the Great Lakes Basin Framework Study, Appendix 4, Limnology of Lakes and
Embayments, Section 7, pp 151-238, Great Lakes Basin Commission, 3475
Plymouth Road, Ann Arbor, Michigan 48106.
U.S. EPA, 1985: A 5-year Program Strategy for the Great Lakes National Program
Office, 1986-1990, U.S. Environmental Protection Agency, Great Lakes
National Program Office, Chicago, IL, pp 1-5, 19.
Wilson, J.W., 1977: Effect of Lake Ontario on precipitation. Mon. Wea. Rev. .
105. 207-414.
Winchester, J.W., and G. Nifong, 1971: Water pollution in Lake Michigan by
trace elements from pollution aerosol fallout. Wat.. Air. Soil Pollut. I,
50-64.
67
-------
APPENDIX A
CRITERIA FOR SELECTION OF VALID SAMPLES
1. NADP Network (Bowersox, 1984)
a) The sample must have been a wet-only deposition sample. Samples exposed
continuously during both precipitation and dry weather (i.e., bulk samples)
were considered invalid.
b) Standard procedures, as specified in NADP field (NADP, 1982) and
laboratory (Peden et al. . 1986) manuals, must have been followed. When
followed, these procedures assure that the sample will not have been exposed to
uncontrolled ambient conditions in the field or laboratory and that the samples
will not have come in contact with surfaces that were not cleaned according to
specified procedures.
c) A complete set of ion concentration measurements must be present.
d) The sample must not have been grossly contaminated by foreign matter
(e.g., leaves, seeds, bird feces, or insects).
e) There must have been a measurement of precipitation amount from the
rain gauge or sample volume. (Where precipitation was not reported, sample
volume converted to precipitation amount was substituted; sample volume was a
required measurement.)
f) Note: any sample that was of insufficient volume to measure all of the
ions and for which the total precipitation reported from the rain gauge was <
0.01 in. was considered a valid "zero" measurement by default.
With two exceptions these are the same criteria used to select "valid"
samples for the annual NADP data summaries (e.g., NADP Subcommittee Number 3,
1985). The exceptions are 1) that the 6-8 day limit on the duration of the
sampling period was not used to exclude samples here, and 2) that samples of
insufficient volume for a complete analysis and "T" (trace) rainfall amounts
were considered valid "zero" measurements here, but not in the cited data
summaries. Very similar criteria are also being used by the U.S.-Canadian
Unified Deposition Data Base Committee (undated) to select valid samples to
verify an atmospheric deposition model that has been applied to data for
eastern North America.
2. GLAD
a) Samples must have a measured sample volume.
b) All other samples in the data set provided by EPA/GLNPO were considered
valid. We assumed that gaps in the computer data record represented samples
that were invalidated at EPA/GLNPO or in their analytical laboratory.
Based on these lists of criteria for NADP and GLAD data, it is evident
that there are differences in the degree or extent to which these two data sets
68
-------
could be matched in terms of their quality. The NADP data provide a more
complete set of codes and information from which to apply screening criteria.
Some screening is implicit in the standard operating procedures employed by the
GLAD monitoring program (GLNPO, undated). To some extent these may equate to
the procedures for identifying samples that fail NADP criteria (b) and (d). The
most notable differences relate to NADP criteria (a) and (e), verification that
the sample is precipitation-only and the presence of an independent measurement
of precipitation amount. It was not until 1984/85 that rain gauges were
installed at GLAD sites. In addition, there is no provision for continuously
monitoring the operation of the AeroChem Metrics sampler at GLAD sites, as is
the case with the open/close recorder in use at NADP sites.
69
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
REPORT NO.
EPA-905/4-88-002
I. RECIPIENT'S ACCESSION NO.
.TITLE AND SUBTITLE
Great Lakes Atmospheric Deposition(GLAD) Network,
1982 and 1983
5. REPORT DATE
February 1988
6. PERFORMING ORGANIZATION CODE
5GL
. AUTHORIS)
Donald F. Gatz, Van C. Bowershox, Jack Su and
Gary J. Stensland
8. PERFORMING ORGANIZATION REPORT NO.
GLNPO Report No. 2
'9. PERFORMING ORGANIZATION NAME AND ADDRESS
10. PROGRAM ELEMENT NO.
Illinois Department of Energy and Natural Resources
Atmospheric Chemistry Section
Illinois State Water Survey
Champaign, Illinois 61820
11. CONTRACT/GRANT NO.
R005882-01-1
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Great Lakes National Program Office
230 South Dearborn Street
Chicago, Illinois 60604
13. TYPE OF REPORT AND PERIOD COVERED
Final 1982-1983
14. SPONSORING AGENCY CODE
Great Lakes National Program
Office, U.S. EPA, Region V
15. SUPPLEMENTARY NOTES
Edward Klappenbach, Project Officer
16. ABSTRACT
In 1981 the Great Lakes National Program Office(GLNPO) of the U.S. Environmental
Protection Agency(U.S. EPA) established a network of Great Lakes Atmospheric
Deposition(GLAD) sites to determine atmospheric loadings of metals, nutrients,
and major inorganic species to the Great Lakes and to evaluate annual trends in
the chemical loadings of these species to the lakes. This network was designed
to collect wet-only deposition samples at near-shore locations.
This study contains an analysis and interpretation of atmospheric wet deposition
data collected by the GLAD network. Included in this study are: as assessment
of data quality; a comparison of specific pairs of GLAD and National Atmospheric
elements to the five Great Lakes; and an analysis of the potential changes in
loading estimates caused by closing certain GLAD sampling sites.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI Field/Group
Atmospheric Deposition
Precipitation fluxes to the Great Lakes
Loadings
Data quality
Sampling and analysis
18. DISTRIBUTION STATEMENT
Document is available to the public through
the National Technical Information Service
(NTIS). Springfield. VA 22161
19. SECURITY CLASS (ThisReport)
21. NO. OF PAGES
76
20. SECURITY CLASS (Thispage)
22. PRICE
EPA Fotm 2220-1 19-73)
U.S. GOVERNMENT PRINTING OFFICE: 543-859/62135
------- |