EPA/600/R-04/103
September 2004
Comparison of Hydrologic Responses at Different Watershed Scales
By
Yusuf Mohamoud
Ecosystems Research Division
U.S. Environmental Protection Agency
960 College Station Road
Athens, Georgia 30605-2700
National Exposure Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
-------
Notice
The United States Environmental Protection Agency through its Office of Research and
Development funded the research described here. It has been subject to the Agency's peer and
administrative review and it has been approved for publication as an EPA document.
11
-------
Abstract
Land surface hydrology controls runoff production and the associated transport of sediments,
and a wide variety of anthropogenic organic chemicals, and nutrients from upland landscape areas
and hillslopes to streams and other water bodies. Based on interactions between landscape
characteristics and precipitation inputs, watersheds respond differently to different climatic inputs
(e.g. precipitation and solar radiation). This study compares the hydrologic responses of the Mid-
Atlantic watersheds, and identifies the landscape and climatic descriptors that control those
responses. Our approach was to select representative watersheds from the Mid-Atlantic region,
group the watersheds by physiographic province and ecoregion, and then collect landscape,
climate, and hydrologic response descriptor data for each selected watershed. For example, we
extracted extensive landscape descriptor data from soil, land use and land cover, and digital
elevation model geographic information system (GIS) databases. After sufficient data was
collected, we conducted a variety of studies to determine how different landscape and climatic
descriptors influence the hydrologic response of Mid-Atlantic watersheds.
This report is comprised of four main parts. Part I describes the selection of the representative
study watersheds and the determination of representative physical landscape descriptors for each
watershed using geographic information system analysis tools. Part II characterizes the climate and
associated hydrologic responses of the study watersheds. To select climate descriptors that are good
predictors of hydrologic response, we examined a large number of candidate descriptors. Based on
our examination, we selected dryness index and mean monthly rainfall as the best hydrologic
response predictors. In Part II, we also present the results of our study hydrologic response
comparisons of the study watersheds using a water balance approach. The water balance approach
was based on comparisons of precipitation, streamflow, and evapotranspiration at annual, monthly,
and daily time scales. These comparisons revealed that elevation and latitudinal position strongly
influence hydrologic response. The results also showed that mountainous watersheds of the
Appalachian Plateau, Ridge and Valley, and Blue Ridge Physiographic Provinces have more
streamflow and less evapotranspiration than watersheds located in the Piedmont Province, and that
snowmelt contributes a large portion of streamflow.
Part III presents relationships we derived between landscape-climatic descriptors and the
hydrologic response descriptors. Flow duration indices (Ql.. .Q95) were used to represent the
hydrologic responses of the study watersheds. In Part III, we also present comparisons of the
hydrologic responses of the study watersheds at high flow condition, represented by the Ql index,
medium flow condition represented by the Q50 index, and low flow condition represented by the
Q95 index. These comparisons revealed that: the Appalachian Plateau, ridge-dominated Ridge and
Valley, and Blue Ridge watersheds have the highest Ql and Q50 indices; the valley-dominated
Ridge and Valley watersheds have the lowest Q50 index, and the Piedmont watersheds have the
lowest Ql index and a relatively high Q95 index.
Finally, Part IV discusses some of the implications of the study results for watershed
management. We also present applications of the research for hydrologic modeling and watershed
assessment.
in
-------
Table of Contents
Notice ii
Abstract iii
Table of Contents iv
Figures vi
Tables vii
1 Introduction 1
2 Description of Study Area 4
Parti 6
3 Physical Landscape Descriptors Controlling Hydrologic Responses 6
3.1 Soil Descriptors 8
3.1.1 Dominant Soils of the Study Watersheds 9
3.1.2 Soil Descriptors as Potential Hydrologic Response Predictors 12
3.2 Characterization of Bedrock Geology 13
3.2.1 Bedrock Geology of the Study Watersheds 13
3.2.3 Bedrock Geology Descriptors as Potential Hydrologic Response Predictors 15
3.3 Land Use and Land Cover Descriptors 16
3.3.1 Dominant Land Use and Cover Types of the Study Watersheds 17
3.3.2 Land Use and Land Cover Descriptors as Potential Hydrologic Response Predictors 18
3.4 Geomorphologic Descriptors 19
3.4.1 Elevation Parameters 19
3.4.2 Slope Parameters 21
3.4.3 Channel Network and Other Parameters 22
3.4.4 Geomorphologic Descriptors as Potential Hydrologic Response Predictors 26
Part II 28
4 Climate Characterization 28
4.1 Precipitation 28
4.2 Temperature 28
4.3 Influence of Elevation on Climate and Hydrology 29
4.4 Potential Climate Descriptors 31
5 Hydrologic Response Characterization 32
5.1 Conceptual Approach: Water Balance as a Framework for Hydrologic Response Comparisons 32
5.2 Hydrologic Response Comparisons: Water Balance Approach 34
5.2.1 Precipitation Data 34
iv
-------
5.2.2StreamflowData 34
5.2.3 Evapotranspiration Data 34
5.2.4 Moisture Storage Data 35
5.3 Hydrologic Response Comparisons at Annual Time Scale 35
5.4 Hydrologic Response Comparisons at Monthly Time Scale 36
5.4.1 High Streamflow Category Comparisons 37
5.4.2 Medium Streamflow Category Comparisons 40
5.4.3 Low Streamflow Category Comparisons 42
5.5 Hydrologic Response Comparisons at Daily Time Scale 44
5.6 Hydrologic Response Comparisons at the Hourly Time Scale 48
Part III 50
6 Hydrologic Response Prediction 50
6.1 Approach 51
6.2 Comparisons of Hydrologic Responses at the Regional Scale 53
6.3 Prediction of Flow Duration Indices (FDIs) for Ungaged Watersheds at Regional Spatial Scale and
Multi-year Time Scale 56
6.3.1Q1 Model 59
6.3.2 Q5 and Q10 Models 59
6.3.3 Q20 and Q30 Models 60
6.3.4 Q40 and Q50 Models 60
6.3.5 Q70, Q90 and Q95 Models 60
6.4 Predicting FDIs for Ungaged Watersheds at the Physiographic Province Spatial Scale and Single-year
Time Scale 61
Part IV 65
7 Implications For Watershed Management 65
7.1 Applications to Hydrologic Modeling 66
7.2 Applications to Watershed Vulnerability Assessment 67
7.2.1 Climate Change Applications 67
7.2.2 Land Use Change Applications 68
7.2.3 Water Quality Applications 68
7.2.4 Water Resources Applications 68
8 Summary and Conclusions 70
9 References 72
-------
Figures
Figure 1. Location of Study Watersheds 5
Figure 2. Relationship Between Hydraulic Conductivity and Runoff Ratio of Study Watersheds.. 11
Figure 3. Study Watersheds Ranked in the Order of Increasing Baseflow Index 16
Figure 4. Study Watersheds Ranked in the Order of Increasing Deciduous Cover 18
Figure 5. A color-coded digital elevation contour of a headwater stream of Blockhouse Creek in
Pennsylvania showing geomorphologic features of the glaciated low plateau ecoregion (60a)
of the Appalachian Plateau Province 23
Figure 6. A color-coded digital elevation contour of a headwater stream of South Branch Potomac
River in West Virginia showing geomorphologic features of the northern dissected ridges and
valleys ecoregion (67d) of the Ridge and Valley Province 23
Figure 7. A color-coded digital elevation contour of a headwater stream of Owens Creek in
Maryland showing geomorphologic features of northern igneous ridges ecoregion (66a) of the
Blue Ridge Province 24
Figure 8. A color-coded digital elevation contour of a headwater stream of Deer Creek watershed in
Maryland showing geomorphologic features of the Piedmont upland ecoregion (64c) of the
Piedmont Province 24
Figure 9. Hypsometric curves of four study watersheds arranged in the order of decreasing HPC10
((a) having the highest HPC10 and (d) having the lowest HPC10) 27
Figure 10. Influence of Elevation on Seasonal Streamflow Patterns 30
Figure 11. A Schematic Diagram Showing the Components of the Water Balance Equation for a
Small Headwater Subwatershed 33
Figure 12. Monthly Water Balance Components of Two Appalachian Plateau Watersheds: (a)
reenbrier Watershed and (b) the Towanda Watershed 39
Figure 13. Monthly Water Balance Components of Two Ridge and Valley Watersheds: (a) Back
River Watersheds, (b) Little Juniata Watershed 41
Figure 14. Monthly Water Balance Components of Two Piedmont Watersheds: (a) Deer Creek
Watershed and (b) Slate Watershed 43
Figure 15. Comparisons of Water Balance Components of an Appalachian Watershed at Daily
Time Scale 45
Figure 16. Comparisons of Hydrologic Response of a Ridge and Valley Watershed at the Daily
Time Scale 46
Figure 17. Comparisons of Water Balance Components of a Piedmont Watershed at the Daily Time
Scale 47
Figure 18. Comparisons of Standardized Hourly Runoff Hydrographs 48
Figure 19. A One-year Flow Duration Curve Showing Lines that Represent the 10 Flow Duration
Indices (FDIs) and Normalized Daily Streamflow Data 52
Figure 20. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q5 Index. 54
Figure 21. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q50 Index.
55
Figure 22. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q95 Index.
56
VI
-------
Tables
Table 1. List of Study Watersheds, Their Location, Drainage Area, and Major River System 7
Table 2. List of landscape descriptors 8
Table 3. Soil Physical Properties Extracted from STATSGO Database for the Top Layers of the
Dominant Soils of each Watershed 10
Table 4. Study Watersheds Arranged by Physiographic Province and Ecoregion, and Dominant
bedrock Geology 14
Table 5. Geomorphologic Descriptors of the Study Watersheds by Physiographic Province 20
Table 6.Long-term Mean Annual Precipiation, Minimum Januray Temperature, Mean Elevation,
and Latitude and Longitude of Study Watersheds 29
Table 7. Climate Descriptors Examined as Potential Hydrologic Response Descriptors 31
Table 8. Long-term Mean Annual Precipitation, Mean Annual Streamflow, Mean Actual
Evapotranspiration Estimates, and Associated Ratios 36
Table 9. Correlation Coefficients Between Landscape, Climate, and Selected Hydrologic Response
Descriptors 58
Table 10. Flow Duration Indices Equations for Regional Spatial Scale and Multi-year Time Scale
59
Table 11. Flow Duration Indices Equations for Physiographic Province Spatial Scale and One-year
Time Scale 62
Table 12. Summary of Dominant Hydrologic Response Predictors at Different Spatial and
Temporal Scales 64
vn
-------
1 Introduction
The hydrologic response of a watershed is influenced by many soil property descriptors (e.g.,
infiltration capacity, soil depth, and porosity), geomorphologic descriptors (e.g., drainage area,
lake/pond areas, slope, channel length, drainage density, and relief ratio), geologic descriptors (e.g.,
lithologic and structural geologic properties), and land cover and land use descriptors (e.g., percent
forest, agricultural, and urban cover). Many investigators have developed regression equations to
relate landscape descriptors to hydrologic response variables, such as low flows or peak flow rates.
Some of the widely used landscape descriptors in hydrology include drainage area, channel
slope, channel length, forested area, drainage density, and relief ratio. Because landscape
descriptors influence the hydrologic response of a watershed, landscape descriptors have been the
foundation of many widely used empirical and statistical hydrologic response equations, such as
the rational method (Kuichling,1889) that relates the peak flow rate to the drainage area of a
watershed. Other important empirical equations that use landscape descriptors include the Soil
Conservation Service Method (SCS, 1972) that uses an arbitrary "curve number" to determine
direct runoff. The "curve number" represents the cumulative effects of landscape descriptors that
control initial abstractions or water losses that usually correspond to that fraction of precipitation
not translated into direct runoff. Such initial abstractions include surface depression storage
(controlled by land use and land cover, soil, and micro-topography), interception (controlled by
land use and land cover type), and infiltration losses, controlled by soil characteristics.
As high-speed computers became available and as more models were used for regulatory
purposes, the need for physically-based hydrologic models increased. In recent years, resource
managers and policy makers have demanded models that can be easily parameterized and that can
accurately simulate both hydrology and water quality processes at the watershed scale.
Unfortunately, these models need parameter values that reflect the effect of soil, geology,
topography, land use and land cover on the hydrologic response. The transition from simple
empirical models to physically based hydrologic models has to date met with limited success. One
limitation to developing and testing physically-based models is the lack of ways to represent the
relationships between landscape descriptors and hydrologic response. Without a clear
understanding of these relationships, it is difficult to identify how soil, vegetation, geology, and the
geomorphologic parameters influence hydrologic processes at different spatial and temporal scales.
Moreover, measured data is essential to process understanding, but it is not logistically feasible to
obtain measured landscape and hydrologic descriptor data for many large watersheds.
There are also a number of climate and hydrologic factors that influence the hydrologic response
of a watershed. These factors include precipitation input (e.g., rainfall and snow), including its
temporal and spatial distribution over the watershed (Singh 1997), antecedent soil moisture
conditions (Hawley et al. 1983; Montgomery and Dietrich, 2002), and available soil and
groundwater storage (Troch et el. 1993; Wittenberg and Sivapalan, 1999). A watershed's
hydrologic response is an indicator of how a watershed processes precipitation inputs, based on its
unique set of landscape descriptors. Many investigators have examined the relationship between
landscape descriptors and observed hydrologic response variables (Zecharias and Brutsaert, 1988;
-------
Nathan and McMahon, 1988; Lacey and Grayson, 1998). Other researchers have examined the
relationship between landscape descriptors and simulated hydrologic response variables (Sefton
and Howarth, 1998; Berger and Entekhabi, 2001). The focus of most of these cited studies was,
however, to examine relationships between landscape descriptors and low flow hydrologic
conditions.
Lacey and Grayson (1998) examined the relationship between baseflow index and landscape
descriptors that included a set of qualitative geology-vegetation parameters and dimensionless
topographic and climatic indices. They found no trends between plots of baseflow index against
any dimensionless topographic parameter within the geology-vegetation groups. Although most
landscape and hydrologic relationship studies have focused on baseflow or peak flow conditions,
some investigators examined long-term hydrologic responses. Berger and Entekhabi (2001)
demonstrated that long-term hydrologic response of a watershed could be determined from
physiographic and climatic descriptors. These investigators used the annual runoff ratio (ratio
between annual streamflow to annual precipitation) to represent the long-term hydrologic response
of a basin. They also have examined a number of other potential hydrologic response predictors,
but found that runoff ratio was also the most closely related to climate that they represented by the
wetness ratio (ratio between annual precipitation and annual potential evapotranspiration).
Although landscape descriptors and precipitation inputs influence the hydrologic response of a
watershed, scale also plays an important role because scale introduces heterogeneity in the
landscape descriptors. As the drainage area of a watershed increases, the soil, bedrock geology,
land use and land cover, and topographic features become more variable. As the variability in
landscape descriptors increases, different landscape characteristics can interact and possibly initiate
a different hydrologic response than could have been produced by any single set of descriptors
without the interaction. Unlike large watersheds, small headwater watersheds show a high degree
of homogeneity in landscape descriptors and, in some cases, can have: a single soil unit, and land
use and land cover type; homogeneous bedrock geology, and uniform topography. Small
watersheds also tend to receive a more evenly distributed rainfall, and thus their hydrologic
response reflects that uniformity in rainfall distribution (Singh, 1997). On the other hand, large
watersheds experience uneven rainfall distribution that often leads to an uneven runoff distribution.
Different processes become important at different spatial scales and processes that are important
at small scales may not be important at large scales (Meentemeyer and Box, 1987). Wood et al.
(1988) conducted an empirical study on the impact of scale on runoff. Wood (1994) later repeated
the same experiment on runoff ratio and Famiglietti and Wood (1995) repeated the same
experiment on evaporation. These studies found that both runoff and evaporation have large
variability, controlled by the variability in soils and topography.
The water balance equation is a fundamental hydrology equation that is valid for all temporal
and spatial scales. At large time scales, some terms of the water balance equation become
negligible while other terms transform themselves and become part of another water balance term.
For example, interception is a part of evaporation, while depending on the time scale, infiltration is
a part of soil moisture storage, subsurface runoff or groundwater recharge. For time scales that are
equal to or longer than one day, process aggregation occurs and infiltration and interception
processes that are important at short temporal scales are no longer important; other processes, such
-------
as evapotranspiration, that are negligible and are often ignored at short time scales become
important at large time scales.
Although considerable research has been conducted examining the relationship between
landscape descriptors and hydrologic responses, many past studies were based on field plot or
small experimental watersheds where landscape descriptors are nearly homogeneous and the
heterogeneities that exist in large watersheds are missing. Observations and insights gained from
these small-scale field studies are often used to build physically-based models. A critical limitation
of such single site studies is how to extrapolate knowledge gained from a small-scale area with
nearly homogeneous landscape descriptors to larger watersheds with variable size and variable
landscape descriptors.
In the past, to develop relationships between landscape descriptors and hydrologic response,
researchers used only a few landscape variables (e.g., drainage density, relief ratio, drainage area,
etc.) and only one hydrologic response variable at a time (e.g., baseflow index or peak flow rate).
The present effort differs from earlier studies in a number of ways. First, this study uses an
extensive array of landscape descriptors that incorporate soil, geology, topography, vegetation, and
climate data. The availability of geographic information system (GIS) analysis tools combined with
spatial databases, such as the digital elevation model (DEM) data, and soil databases, e.g., the
digital State Soil Geographic Database (STATSGO) provided the opportunity to easily and rapidly
obtain extensive soil and topographic parameters for our selected study watersheds. Second, unlike
other studies that focused on either low flow (baseflow) or peak flow conditions, this study uses
multiple hydrologic response representations that reflect a wide range of hydrologic conditions.
The objectives of this study were: (1) to compare hydrologic responses and identify the dominant
landscape descriptors that control the hydrologic responses of Mid-Atlantic watersheds; (2) to
develop relationships between landscape and climatic descriptors and hydrologic response
variables; and (3) to make recommendations on how to use the results of this study for resource
management and modeling purposes.
-------
Description of Study Area
The United States Geological Survey (USGS) has classified the nation's water resources into 21
hydrologic regions. According to their regional classification, Region 2 covers the Mid-Atlantic
States from Maine to Virginia. Our study area covers four physiographic provinces: Appalachian
Plateaus, Ridge and Valley, Blue Ridge, and Piedmont. A physiographic province is a landform
characterized by similar elevation, relief, geologic structure, and climate. In general, physiographic
provinces are subdivided into ecoregions (Woods et al. 1999). Woods et al. (1999) defined
ecoregions as areas of relative homogeneity in ecological systems (e.g., soils, vegetation, geology,
and physiography) and their components. They also stated that, because of their similar landscape
descriptors, ecoregions could be an effective framework for inventorying and assessing regional
environmental resources and setting regional resources management goals.
Using physiographic provinces and ecoregions as a selection framework, we selected 25
watersheds from the Appalachian Plateaus, Ridge and Valley, Blue Ridge, and the Piedmont
Physiographic Provinces of Maryland, Pennsylvania, Virginia, and West Virginia (Figure 1). Note
that different physiographic provinces have different numbers of ecoregions. For instance, there are
12 ecoregions in the Appalachian Plateaus, nine ecoregions in the Ridge and Valley, five
ecoregions in the Blue Ridge, and eight ecoregions in the Piedmont. Our 25 study watersheds
represent about 80 percent of the Ridge and Valley ecoregions, 50 percent of the Appalachian
Plateau ecoregions, 40 percent of the Blue Ridge ecoregions, and 35 percent of the Piedmont
ecoregions. In other words, we selected six watersheds from the 12 ecoregions of the Appalachian
Plateau, seven watersheds from the nine ecoregions of the Ridge and Valley, two watersheds from
the five ecoregions of the Blue Ridge, and three watersheds from the eight ecoregions of the
Piedmont. While some Mid-Atlantic ecoregions are not represented in this study, others are
represented more than once. The use of ecoregions as a selection framework facilitates the
extrapolation of our results to all watersheds within Mid-Atlantic region.
-------
Legend
$ Apalachian Plateau
N Blue Ridge
% Piedmont
# Ridge and Valley <
$6
4
#
£L
\x:"
100
200 Miles
West Virginia
Virginia
12-South Branch Potomac
13-Patterson Creek
14-Cacapon River
15-Back Creek
16-Greenbier at Durban
17-Greenbrier River Buckeye
19-SF Shenandoah River
20-NF Shenandoah River
21-Bullpasture Creek
22-Johns Creek
23-Hardware Creek
24-Slate River
18-Greenbrier River at Alderson25-Holiday Creek
Pennsylvania
1-Lehigh River
2-Towanda River
3-Wapwallopen
4-Marsh Creek
5-Pine River
6-Blockhouse Creek
7-L. Juniata River
8-Sherman Creek
Maryland
9-Deer Creek
10-Owens Creek
11-Fishing Creek
Figure 1. Location of Study Watersheds.
-------
Parti
Physical Landscape Descriptors Controlling Hydrologic
Responses
When comparing hydrologic responses, we assumed that watersheds located in the same or
similar ecoregions within a physiographic province have closely related landscape descriptors and,
may therefore, have comparable hydrologic responses. A list of the selected watersheds, their
drainage areas, latitudinal and longitudinal positions, and the major river system to which each
belongs is shown in Table 1. The study watersheds have drainage areas that range from 15 to 4250
km2 (variable spatial scale) and all have a long record of climatic and hydrologic data. Most of the
watersheds have a high percentage of forest cover and limited impacts of human-induced watershed
disturbances such as urbanization, flow regulation, and/or agricultural land use. To ensure that the
study watersheds had long records of good-quality streamflow data, we selected all 25 watersheds
from the U.S. Geological Survey's Hydro-climatic Data Network (HCDN) (Slack and Landwehr,
1992). The HCDN data consists of a list of about 1659 gaging sites with good-quality streamflow
data.
As stated before, in order to examine the relationships between landscape descriptors and
hydrologic response variables, one must obtain coincident landscape, climate, and hydrologic
response descriptor data. In this study, we denoted landscape descriptor data as that extracted from
soils, geology, land use and land cover, and topography databases using geographic information
systems (GIS) tools. We denoted climate and hydrologic response descriptors or variables as
climate and streamflow time series data (See Table 2). We used surrogate descriptors when
quantitative measures of landscape descriptors were not available. A more detailed description of
some of the descriptors is given in the following sections of this report. All landscape descriptor
data extracted from the GIS databases were considered as potential predictors of hydrologic
response.
-------
Table 1. List of Study Watersheds, Their Location, Drainage Area, and Major River System.
Watershed
Lehigh River at Stoddarsville, PA
Towanda Creek, Monroetown, PA
Pine Creek at Cedar Run, PA
Blockhouse Creek Near English, PA
Greenbrier River at Durbin, WV
Greenbrier River at Buckeye, WV
Greenbrier River at Alderson, WV
Wapwallopen Creek Near Wap., PA
Marsh Creek at Blanchard, PA
Little Juniata River at Spruce, PA
Sherman Creek, Shermans Dale, PA
Patterson Creek, Headsville, WV
South Branch Potomac Sprinfield, WV
Cacapon River Great Capapon, WV
Back Creek Near Jones Spring, WV
S F Shenandoah River at Front, VA
N F Shenandoah River at Cootes, VA
Bullpasture River, Williamsville, VA
Johns Creek at New Castle, VA
Owens Creek at Lantz, MD
Fishing Creek Near Lewistown, MD
Deer Creek at Rocks, MD
Slate River Near Arvonia, VA
Hardware River Near Scottsville, VA
Holiday Creek, Andersonville, VA
Major Drainage
River Basin Area (km2)
Appalachian Plateaus
Delaware 237.40
Susquehanna 556.61
Susquehanna 563.68
Susquehanna 97.60
Ohio 344.32
Ohio 1398.00
Ohio 3531.23
Ridge and Valley
Susquehanna 113.39
Susquehanna 114.17
Susquehanna 569.55
Susquehanna 517.78
Potomac 566.96
Potomac 3808.24
Potomac 1752.67
Potomac 629.10
Potomac 4250.94
Potomac 543.67
James 284.78
James 269.24
Blue Ridge
Potomac 15.35
Potomac 18.87
Piedmont
Susquehanna 244.39
James 585.09
James 300.31
James 23.30
Latitude
41:07:49N
41:42:25N
41:31:18N
41:28:25N
38:32:37N
38:11:09N
37:43:27N
41:03:33N
41:03:34N
40:36:45N
40:19:24N
39:26:35N
39:26:49N
39:34:43N
39:30:43N
38:54:50N
38:38:13N
38:11:43N
37:30:22N
39:40:36N
39:31:35N
39:37:49N
37:42:10N
37:48:45N
37:24:55N
Longitude
075:37:33W
076:29:06W
077:26:52W
077:13:52W
079:50:OOW
080:07:51W
080:38:30W
076:05:38W
077:36:22W
078:08:27W
077:10:09W
078:49:20W
078:39:16W
078:18:34W
078:02: 15W
078:12:40W
078:51:11W
079:34: 14W
080:06:25W
077:27:50W
077:28:OOW
076:24:13W
078:22:40W
078:27:20W
078:38:10W
-------
Table 2. List of landscape descriptors
Symbol
Land use and land cover
AGRC
URBN
FRSD
FRST
FRSE
Geomorpho logic
HMIN
HMAX
HAVG
HMED
HSTD
BREL
RRAT
SAVG
SMED
SSTD
MCHL
TCHL
MCHS
DDEN
DARE
HPC10
Soil (top two layers)
SOLD
A WC 12
KSAT12
ORGC12
CLAY112
SILT 12
SAND12
ROCK12
STOR
DRATIO
KRATIO
Geology
DLIT
BFI
Variable description
Agriculture
Urban
Decidiuous forest
Mixed forest
Evergreen forest
Minimum elevation
Maximum elevation
Average elevation
Median elevation
Standard deviation of elevation
Basin relief
Relief ratio
Average slope
Median slope
Slope standard deviation
Main stream channel length
Total length of streams
Main channel slope
Drainage density
Drainage area
Hypsometric curve elevation corresponding to 10% of area
Total soil depth
Plant available water content
Saturated hydraulic conductivity
Soil organic carbon
Percent clay
Percent silt
Percent sand
Percent rock fragment
Available moisture storage
Depth ratio (layer 2/layer 1 )
Depth weighted ratio of saturated hydraulic conductivity for layers 1
Dominant lithology (qualitative)
Baseflow index (daily)
Units
(%)
(%)
(%)
(%)
(%)
(m)
(m)
(m)
(m)
(m)
(m)
(m)
(m/m)
(m/m)
(m/m)
(km)
(km)
(m/m)
(km/km2)
km2
(%)
(mm)
(mm)
(mm/hr)
(%)
(%)
(%)
(%)
(%)
mm
-
and 2
-
-
3.1 Soil Descriptors
The underlying bedrock geology (Lacey and Gray son, 1998) and topographic position in the
landscape of an area often influence the rate of formation and properties of its soils. The
topographic position determines important soil properties that influence the hydrologic response of
a watershed (England et al. 1968). For example, soil depth decreases with an increase in elevation,
and elevation has both direct and indirect influence on hydrologic response. Specifically,
watersheds with shallow soils and steep slopes retain less precipitation in the soil, that is they have
low moisture storage capacity. When precipitation fills the available storage capacity of a soil, the
soil becomes highly responsive and quickly releases a high fraction of the incident precipitation as
-------
quickflow runoff (Carey and Woo, 2001). The geographic information system analysis tools
available within the U.S. EPA BASINS3 Modeling System were used to extract soil characteristics
for the dominant soil of each study watershed. We extracted all the soil parameters from the digital
State Soil Geographic Database (STATSGO) (NRCS, 1991). The extracted soil parameters for the
25 study watersheds are shown in Table 3.
3.1.1 Dominant Soils of the Study Watersheds
The properties of the soils of the Ridge and Valley Province clearly vary with the topographic
position in the landscape and the underlying bedrock geology. For example, colluvial soils derived
from sandstone and shale bedrock are found on the ridge slopes, while the valley soils are formed
from limestone and shale. Limestone-derived soils are generally deep and fertile. In many valley-
dominated watersheds of the Ridge and Valley Province, these deep and fertile soils are under
intensive agricultural use. The dominant Ridge and Valley Province soils include Berks, Dekalb,
Lehew, and Wallen soil series (Table 3).
The Appalachian Plateau watersheds have deep soils with high rock fragments. These soils
formed in residuum or from glacial till deposits. The soils are classified as Ultisols, Inceptisols, and
Alfisols. Some Appalachian Plateau soils have a flow-impeding subsoil layer known as fragipan.
The fragipan occurs at a depth of 43 to 91 cm below the surface, where it restricts root penetration
and controls the flow of water and solutes. Norris and Volusia soils are two dominant soils that are
commonly found in our study watersheds located in the Appalachian Plateau Physiographic
Province (Table 3).
Only four out of our 25 study watersheds are found in the Piedmont Physiographic Province.
These watersheds are Deer Creek located in Hartford County, Maryland, and the Hardware, Slate,
and Holiday watersheds located in Albemarle, Buckingham, and Appomattox counties,
respectively, of Virginia. Chester soil series is the dominant soil type for the Deer Creek watershed
in Maryland while Georgeville is the dominant soil type for the Hardware, Slate, and Holiday
watersheds in Virginia (Table 3). In general, soils of the Piedmont Province are derived from
metamorphic rocks and are relatively deep with a low rock fragment content and a thick, saprolite
layer.
-------
Table 3. Soil Physical Properties Extracted from STATSGO Database for the Top Layers of the
Dominant Soils of each Watershed
Watershed
Soil
Series
Soil Soil AWC KS
Depth1 Depth2 mm/mm mm/hr
(mm) (mm)
soc
%
Clay
%
Silt
%
Sand
%
Appalachian Plateau
Lehigh (62a)
Towanda (60a/62c
Pine (62c/62d)
Blockhouse (60a)
Greenbrier (69a/69c)
Morris
Morris
Volusia
Morris
Berks
1524
1524
1778
1524
863
292
304
174
381
212
0.18
0.16
0.14
0.17
0.13
5.40
4.85
8.52
2.90
22.32
2.0
2.0
1.0
1.0
1.6
21
21
20
21
17
53
53
45
53
45
26
26
35
27
38
Ridge and Valley
Wapwallopen (67b)
Marsh (67 d)
L. Juniata (67c)
Sherman (67f)
South Branch (67 d)
Patterson (67b/67c)
Cacapon (67b/67c)
Back (67a/67b)
S.F. Shenan (67a/67b)
N.F. Shenand. (67 d)
Bullpasture (67b/67c)
Johns (67g/67h)
Owens (66a)
Fishing (66b)
Deer (64c)
Hardware (45e)
Slate (45e)
Holiday (45f)
Morris
Lechkill
Hazelton
Hazelton
Berks
Berks
Lehew
Berks
Frederick
Berks
Wallen
Wallen
Fauquier
Fauquier
Chester
Georgeville
Georgeville
1524
1295
1447
1447
863
863
838
863
1828
863
736
736
Blue Ridge
1524
1524
Piedmont
1574
1600
1600
Georgeville 1600
332
207
206
199
216
238
198
240
197
228
148
148
152
152
226
163
163
152
0.16
0.12
0.10
0.11
0.11
0.11
0.11
0.13
0.14
0.10
0.07
0.07
0.13
0.13
0.17
0.14
0.14
0.14
4.01
14.84
23.74
17.65
70.69
33.07
81.10
26.36
16.89
35.00
39.54
39.54
15.00
15.00
9.95
25.19
25.19
27.00
1.5
1.4
1.6
1.4
1.9
1.8
1.7
1.7
0.9
1.4
0.7
0.7
1.2
1.2
1.2
0.8
0.8
0.7
20
14
13
14
12
14
12
17
17
14
15
15
18
18
17
17
17
16
53
49
44
46
39
41
33
44
44
36
19
19
39
39
47
40
40
40
27
37
43
40
49
45
55
39
39
50
66
66
43
43
36
43
43
44
Numbers in parenthesis refer to the eco-region of each watershed
province. The superscripts 1 and 2 refer to total profile soil depth
respectively, in each watershed.
within each physiographic
and depth of the top soil layer,
The Blue Ridge Province consists of narrow, mountain ridges that run parallel to the Ridge and
Valley Province. Out of our 25 study watersheds, only Owens Creek and Fishing Creek watersheds
in Frederick County, Maryland are located in the Blue Ridge Physiographic Province (Table 3). In
both of those watersheds, Fauquier soil type covers the entire watershed. This soil is derived from
igneous and metamorphic rocks. It contains significant amounts of rock fragments and is often
found on areas with steep slopes. In some areas, the Blue Ridge Province soils are nearly identical
to the soils of the Piedmont Province. One difference between the soils of these two provinces is
that the Blue Ridge soils have a high percent of rock fragments whereas the Piedmont Province
soils have a low percent.
Figure 2 shows an inverse relationship between saturated hydraulic conductivity (KS) - a soil
descriptor - and the monthly runoff ratio- a hydrologic response descriptor. Soils of our study
10
-------
watersheds vary from province to province and from watershed to watershed. Figure 2 shows that,
as a group, the Ridge and Valley Province soils have the highest saturated hydraulic conductivity
while the Appalachian Plateau soils have the lowest. Saturated hydraulic conductivity of some
Ridge and Valley soils can be as high as 81 mm/hr (Table 3). Wilson and Luxmore (1988)
conducted infiltration experiments on forested watersheds in the Ridge and Valley Province. They
reported that infiltration rates could reach 72 mm/hr and could exceed the rainfall intensity during
storm events, thus eliminating the occurrence of infiltration-excess runoff.
0.60
0.50 -
0.40 -
o
1
ig 0.30
§.
0.20 -
0.10 -
Runoff Ratio-July *t
A - Soil Hydraulic Conductivity (mm/hr) A' i
•Linear (Soil Hydraulic Conductivity (mm/hr)) , •
A'
90
80
70 ~
60 I
50 >
40 g
o
30 %
£
20 £
10
0
1 .-. ; . v ~UJ>- -V ^OW^^ /^/^
&)«) •$p*'^Cb(S^ff'^.§-/T-$P$3-$P-5P
vM// &&$* p #** * * *
& £> ^ A?/e /?
" £ j? £ "
5* ^ £ £ £ ^
^
«?
^ ^ / / / / / g£ £
£*&£co£$gs*
$
Watershed
£
Figure 2. Relationship Between Hydraulic Conductivity and Runoff Ratio of Study Watersheds.
Figure 2 also shows that valley-dominated Ridge and Valley watersheds and Piedmont and
Blue Ridge watersheds have the lowest runoff and, therefore, have the highest hydraulic
conductivities. Note that the use of runoff ratio and the selection of the month of July improved
hydrologic response comparisons because the use of runoff ratio removed two factors that normally
complicate hydrologic response comparisons. First, by dividing the monthly runoff by monthly
precipitation, we removed the effect of different precipitation inputs on the hydrologic response of
different watersheds. Second, if all other factors remain the same, watersheds that have larger
drainage areas tend to have greater runoff. To address this scale factor problem, we normalized
runoff as depth of water instead of a flow rate. Third, because differences in elevation result in
11
-------
differences in hydrologic response, we excluded the months when elevation influence on
hydrologic response was high i.e., when elevation would likely confound the influences of other
landscape descriptors. In this study, we limited hydrologic comparisons to when elevation
influence was negligible. For this reason, for both the monthly and daily comparisons, we used
only the months of June, July, August, and September. By selecting July for the comparison shown
in Figure 2, we assume that most of the runoff in July is from rainfall since the snowmelt period
has ended and groundwater storage and baseflow are receding during the summer months.
3.1.2 Soil Descriptors as Potential Hydrologic Response Predictors
Soil characteristics control hydrologic processes such as rainfall infiltration, percolation, and
moisture storage. Some soil properties, e.g., saturated hydraulic conductivity, control entry of
precipitation inputs into the soil, and consequently the generation of infiltration-excess runoff.
Other soil properties control soil moisture storage. These latter soil descriptors include soil depth,
porosity, plant-available water content and depth to bedrock or the water table. Soils normally
retain water by capillary forces and release water through gravity, evaporation from the soil surface
and transpiration from plant stomata.
The importance of soil descriptors as hydrologic response predictors is well documented and is
evidenced by the inclusion of soil parameter values in many hydrologic models. Although the
importance of soil descriptors as hydrologic response predictors is widely recognized, what is not
known is how to identify and quantify the specific soil descriptors that control hydrologic
responses at different spatial and temporal scales. For instance, at short time scales, particularly
during storm events, parameters such as saturated hydraulic conductivity that control processes
such as infiltration capacity are important. At longer time scales, such as days or months, however,
processes that control moisture storage and internal soil drainage are more important and so are the
soil properties that influence these processes (e.g., soil depth and porosity).
Although soil-controlled hydrologic processes are represented in current hydrologic models,
there are some limitations in the identification and quantification of the relevant soil parameters at
the watershed scale. As an example, depending on its topography, watersheds can have different
soils at different topographic positions in the landscape. Some soils are deep and permeable, others
are shallow and may have a flow-impeding layer at some depth below the soil surface. Soil
characteristics clearly vary across the landscape and within the soil profile, and unfortunately
methodologies to combine the effects of all these diverse soil characteristics for reliable model use
are not currently available.
In this study, we used the dominant soil unit of each watershed, as if a single soil, to cover the
entire watershed. For small watersheds, the dominant soil may indeed be the only soil unit in the
watershed. However, for large watersheds, using only the dominant soil may not entirely reflect all
the combinations of soil influences on the hydrologic response of the watershed. In other words,
depending on a particular soil type's topographic position on the landscape and its relative
contribution to the overall hydrologic response of a watershed, the dominant soil type by area of
coverage may not sometimes be the hydrologically dominant soil in a large watershed.
12
-------
3.2 Characterization of Bedrock Geology
Two major rock types cover the study area, consolidated crystalline and consolidated to
unconsolidated sedimentary. The crystalline rocks, found in the Piedmont and the Blue Ridge
Physiographic Provinces, consist of igneous and metamorphic rocks. By contrast, the consolidated
to unconsolidated sedimentary rocks, found in the Appalachian Plateau and Ridge and Valley
Provinces, consist of sandstone, siltstone, and shale (United States Geological Survey, 1992-1997).
Both lithologic and structural geologic properties influence the watershed hydrologic response, but
no quantitative geologic descriptors were used in this study. We have, however, made some
inferences about the geologic descriptors of each watershed from qualitative measures such as
dominant rock type, and surrogate quantitative measures, such as baseflow index. The qualitative
bedrock geology data were extracted from digital United States Geological Survey maps (Shruben,
1998).
3.2.1 Bedrock Geology of the Study Watersheds
The bedrock geology of the Appalachian Plateau watersheds consists of shale, siltstone,
sandstone, and carbonates. Sandstone and some erosion-resistant carbonates are found in the
upland areas, while shale is usually found in the valleys. These watersheds have very limited
recharge and most of the precipitation that falls on the ground may run rapidly off the slopes.
Except for some springs that occur in areas where stream channels intersect the water table, there
are no major regional groundwater aquifers in the Appalachian Plateau Province (United States
Geological Survey, 1992-1997).
Unlike the Appalachian Plateau province, the Ridge and Valley Province has distinct landforms
characterized by sequences of ridges and valleys that reflect folded and faulted bedrock geologic
formations (Table 4). The valleys and the ridges are shaped by differential erosion of rocks with
different erosion resistance. For example, erosion-resistant sandstone bedrock forms the ridges
while less erosion resistant rocks, such as shale and limestone form the valleys. In some valley
areas, the bedrock is covered by a thick regolith. For most of the Ridge and Valley Province
watersheds, aquifer rocks are mainly low porosity and low permeability sandstone and shale
bedrock material. Secondary porosity and permeability caused by fracturing and dissolution often
create increased water storage and transmission properties yielding groundwater movement along
fractures and bedding planes.
In the Piedmont and Blue Ridge watersheds, the bedrock is covered by regolith and precipitation
enters the aquifer through this porous regolith. Unlike the overlying regolith, the bedrock has very
low porosity and limited capacity to store water. Most of the Piedmont and Blue Ridge
Physiographic Provinces are underlain by dense and nearly impermeable bedrock that yields water
primarily from secondary porosity and permeability provided by fractures. Water stored in the
porous regolith normally moves through the regolith laterally until it discharges into a nearby
stream.
13
-------
Table 4. Study Watersheds Arranged by Physiographic Province and Ecoregion, and Dominant
bedrock Geology
Watershed
Dominant eco-region
lithology
and dominant formations
Appalachian Plateaus
Lehigh (AP) Pocono High Plateau (62a)
Towanda (AP)
Pine (AP)
Blockhouse (AP)
Greenbrier (AP)
Glaciated high (62c) and low plateau (60a)
Glaciated high (62c)/unglaciated plateau (62d)
Glaciated low plateau (60a)
Forested hills (69a); Greenbrier Karst (69c)
Ridge and Valley
Wapwallopen (RV) Northern shale valleys (67b)
Marsh (RV) Northern dissected ridges (67d)
Little Juniata (RV) Northern shale valleys/sandstone ridges (67c)
Sherman (RV) Northern limestone/sandstone ridges (67f)
South B. Potomac (RV) Northern dissected ridges and valleys (67d)
Patterson (RV) Northern shale valleys/sandstone ridges (67bc)
Cacapon (RV) Northern shale valleys/sandstone ridges (67bc)
Back (RV) Northern shale/limestone valleys (67ba)
S.F. Shenandoah (RV) Northern shale/limestone valleys (67ba)
N.F. Shenandoah (RV) Northern dissected ridges (67d)
Bullpasture (RV) Northern sandstone ridges/shale valleys (67cb)
Johns (RV) Southern sandstone ridge/shale valleys (67hg)
Blue Ridge
Owens (BR) Northern igneous ridges (66a)
Fishing (BR) Northern sedimentary and Metased.,ridges (66b)
Piedmont
Deer (PD) Piedmont uplands (64c)
Hardware (PD)
Slate (PD)
Holiday (PD)
Northern inner piedmont (45e)
Northern inner piedmont (45e)
Northern outer piedmont (45f)
Sandstone-siltstone-mudstone
Duncannon member of Catskill
Mudstone-siltstone-sandstone
Lock haven/Burgoon
Sandstone-siltstone-shale
Catskill/Pottsville
Sandstone-siltstone-shale
Catskill/Huntley
Shale-Limestone
Chemung/Greenbrier
Shale-siltstone-sandstone
Pottsville
Sandstone-siltstone-shale
Catskill/Lock haven
Siltstone-shale-dolomite
Brallier/Harrel/Lock haven
Quarzite-Shale-limestone
Tuscarora/bloomsburg
Shale-sandstone-limestone
Chemung/Oriskony
Shale-sandstone
Brallier/Harrel/Chemung
Shale-sandstone
Chemung/Hampshire
Shale-limestone
Martinsburg
Sandstone-shale
Sandstone-shale
Sandstone
Sandstone-shale
Greenstone schist
Greenstone schist
Albite-Phylite-chlorite schists
Octoraro formation
Igneous and metamorphic
Igneous and metamorphic
Igneous and metamorphic
Source: Eco-region numbers were obtained from Woods et al.(1999).
14
-------
3.2.3 Bedrock Geology Descriptors as Potential Hydrologic Response Predictors
In the absence of quantitative measures of bedrock geologic descriptors, we used baseflow index
as a surrogate variable to represent the geologic properties that control groundwater storage and
discharge (e.g., porosity and permeability). Baseflow index is defined as the volume of baseflow
divided by the total volume of streamflow. To separate streamflow into baseflow and quickflow,
we used a baseflow index method developed by the Institute of Hydrology (Gustard et al. 1992).
Quickflow is defined as the rapid runoff component of a streamflow hydrograph that is observed
during or after a rainfall event. Baseflow is one of the most important low flow hydrologic
characteristics of a catchment (Lacey and Grayson, 1998; Smakhtin 2001). Theoretically, soil and
geologic properties control baseflow. Specifically, soil properties control initial entry of
precipitation into the soil, and the soil and the underlying aquifer properties combine to control
storage and release of that water to nearby streams. Farvolden (1963) stated that the streamflow
hydrograph during dry weather flows represents depletion of the groundwater reservoir, and that
baseflow can be used as an indirect indication of soil moisture deficiency.
To compare the baseflow indices of our study watersheds, we ranked the watersheds in the order
of increasing baseflow index (Figure 3). In addition to the baseflow index rankings, Figure 3 also
displays the fraction of the total annual streamflow that is baseflow and the fraction of total annual
streamflow that is quickflow. Note that some watersheds have low total streamflow (e.g., the South
Fork Shenandoah River) while other watersheds have high total streamflow (e.g., the Lehigh
River). An analysis of Figure 3 reveals that Appalachian Plateau watersheds generally have high
total streamflows, but their baseflow indices are not very high. In general, watersheds located in the
mountainous areas of the Appalachian Plateau and the ridge-dominated watersheds of the Ridge
and Valley Province show high total annual streamflow. This high annual total streamflow is a
result of generally high precipitation and low evapotranspiration rates (Hartley and Dingman,
1993). Among all study watersheds, those located in the Blue Ridge Physiographic Province had
the highest total annual streamflow and baseflow indices. By contrast, the Piedmont watersheds
had the lowest total annual streamflows, but had relatively high baseflow indices. The valley-
dominated Ridge and Valley watersheds, such as Back, Patterson, Cacapon, and South Branch
Potomac had both low total streamflows and low baseflow indices. The relative baseflow and
quickflow contributions to total annual streamflow have important implications for water resources
management.
15
-------
(D
CU
CO
800
700-
600-
500-
400
300
200
100
0
<<
Watershed
^$^$*$5$d<
Vy/c/^
A°
.&
v^ -
> *$•'
Figure 3. Study Watersheds Ranked in the Order of Increasing Baseflow Index.
3.3 Land Use and Land Cover Descriptors
The land use and land cover data used in this study were obtained from U.S. EPA's Better
Assessment Science Integrating Point and Non-point Source (BASINS) GIRAS databases. The data
is based on the Anderson land use and land cover classification system (Anderson et al. 1976) and
has a scale of 1:250,000. The data reflects land cover conditions from the mid 1970s to the early
1980s. More information about land use and land cover data can be found at the following website
http://www.epa.gov/waterscience/basins/metadata/giras.htm.
16
-------
3.3.1 Dominant Land Use and Cover Types of the Study Watersheds
In the Mid-Atlantic Highlands Region, vegetation type varies from one province to another and,
within each; vegetation type varies from one ecoregion to another. Much of the variability in
vegetation type depends on differences in soil and topography. In mountainous watersheds of the
Appalachian, Ridge and Valley, and Blue Ridge provinces, deciduous forest cover is dominant.
Figure 4 ranks the study watersheds in order of increasing deciduous forest cover. Some of the
mountainous watersheds with high deciduous forest cover are Bullpasture and Johns Creek in
Virginia, Fishing and Owens Creek in Maryland, South Branch Potomac in West Virginia, and
Pine River and Blockhouse Creek in Pennsylvania (Figure 4). Normally, the low-lying valley areas
of the Ridge and Valley Province are covered by mixed forest type, while the deciduous forest type
covers the ridge-dominated watersheds and high elevation areas, such as the ridge tops. Because
low valley areas are relatively flat and suitable for agriculture, in some low valley areas of the
Ridge and Valley Province forest cover has been converted to agriculture. On the contrary,
mountainous areas not suitable for agriculture remain covered by deciduous forest. The dominant
species in mountainous areas of the Appalachian Plateau and Ridge and Valley Provinces are
mainly oaks, maples, and other hardwood trees.
Vegetation in the Piedmont Province is significantly different than that found in the
mountainous areas of the Appalachian Plateau and Ridge and Valley Provinces. In the Piedmont
Province, evergreen forest cover is dominant. For example, the dominant vegetation types in the
Hardware, Holiday, and Slate watersheds located in the Northern Inner Piedmont ecoregion (45e)
of Virginia are hickory (Carya spp.), loblolly pine (Pinus taeda), and Shortleaf Pine (Pinus
echinata), respectively. Within the Piedmont province, vegetation type also varies; virgin chestnut
oak (Quercuspinus\ hemlock (Tsuga Canadensis), and Beech (Fagus grandifolia) dominate the
watersheds in the Piedmont Uplands ecoregion (64c) of Maryland.
17
-------
100
90
80
70
60
50
40
30
20
10
0
D Decidiuous Forest 5 Mixed Forest • Evergreen Forest
I
Watershed
Figure 4. Study Watersheds Ranked in the Order of Increasing Deciduous Cover.
3.3.2 Land Use and Land Cover Descriptors as Potential Hydrologic Response Predictors
Many investigators have studied the relationship between vegetative cover and hydrologic
response. Forest cover is critical to the hydrology of the Mid-Atlantic watersheds because the forest
stabilizes the thin residuum soils and prevents landslides and excessive soil erosion from occurring
on steep hillslopes. Forest cover also influences hydrologic response in a number of other ways.
For example, forest cover directly affects such hydrologic processes as interception (Swank et al.
1972), rainfall infiltration, evaporation from plant canopy, and transpiration through plant stomata
(Fujieda, 1997). For most of our study watersheds, the deciduous forests are leafless about half the
year and evapotranspiration is high only during the summer months (i.e., June, July, August, and
September) (Patric, 1973). Therefore, it is during the summer when the differential role of
vegetation on hydrologic response can be best evaluated because the snow season has ended and
the effect of elevation on hydrologic response is negligible.
Vegetation is widely recognized as an indicator of climatic conditions within a watershed
(Lacey and Grayson, 1998). In this study, the land use and land cover descriptors used are percent
of forest land, percent of agricultural land, and percent of urban land of each watershed. In addition
to land use type, we further classified forest cover into deciduous forest, mixed forest, and
evergreen forest. Because deciduous forest is closely related to elevation, we hypothesized that the
percent of deciduous forest cover of a watershed could be a useful predictor of watershed
hydrologic response.
18
-------
3.4 Geomorphologic Descriptors
Geomorphologic descriptors of each watershed were extracted from 30 m by 30 m grid
resolution Digital Elevation Model (DEM) data. Geographic Information System tools, such as
those available from the Better Assessment Science Integrating Point and Nonpoint Sources
(BASINS) (USEPA, 1998) Modeling System and MicroDEM (Guth, 1989), were used to extract
quantitative measures of geomorphologic descriptors from digital elevation model data obtained
from the United States Geological Survey Website (http://edc.usgs.gov/geodata). For large
watersheds, we merged several 1:24000 digital elevation model databases to achieve full coverage.
The extracted topographic parameters were elevation (minimum, maximum, average, median, and
standard deviation) and slope (maximum, average, median, and standard deviation) (Table 5).
Other important geomorphologic descriptors extracted include stream network parameters such as
total stream length, average channel slope, and main channel length. While most of the topographic
parameters were directly measured, there were also a number of parameters calculated from the
measured parameters. These calculated parameters included drainage density and relief ratio. A
brief description of each geomorphologic descriptor and symbol was listed in Table 2.
3.4.1 Elevation Parameters
Table 5 displays the elevation parameter values extracted for each watershed; a brief description
of each parameter is given in the following section. Note that some elevation parameters may not
have any direct influence on hydrologic response, but can be useful for estimating other
geomorphologic parameters.
Maximum elevation (HMAX). Maximum elevation of a basin is the highest watershed
elevation. A high maximum elevation indicates the presence of mountain summits in the
watershed. Although maximum elevation was used to estimate other parameters, such as basin
relief and relief ratio, maximum elevation had very limited hydrologic significance.
Minimum elevation (HMIN). Minimum elevation is the lowest elevation point of a watershed. A
low minimum elevation indicates the presence of low valley areas in a watershed. Although
minimum elevation was used for the determination of basin relief and relief ratio, minimum
elevation also had very limited hydrologic significance.
19
-------
Table 5. Geomorphologic Descriptors of the Study Watersheds by Physiographic Province
Watershed
HMIN HMAX HAVG HMED HSTD BREL SMAX SMED SAVG SSTD HPC10
Appalachian Plateau
Lehigh
Towanda
Pine
Blockhouse
GreenbrierD
GreenbrieB
GreenbrierA
443
235
244
314
826
633
470
692
745
773
722
1371
1461
1390
543
495
562
541
1067
957
830
537
470
557
540
1062
927
805
40
120
100
79
96
153
159
249
510
529
407
52
1973
118
110
545 3564
828 3909
920 3830
4
12
8
15
26
27
27
6
15
12
21
27
30
28
Ridge and Valley
5
16
11
16
33
86
79
60
88
85
83
88
80
78
Wapwallopen
Marsh
Little Juniata
Sherman
Patterson
South Branch
Cacapon
Back
NF Shenand.
SF Shenand.
Bullpasture
Johns
Fishing
Owens
Deer
Hardware
Slate
Holiday
157
195
231
115
190
170
170
135
327
150
510
430
222
302
88
95
72
145
632
669
813
694
1070
1490
1026
785
1228
1321
1335
1318
562
543
322
726
340
290
398
397
496
310
404
618
495
279
611
468
801
Blue Ridge
702
457
Piedmont
454
201
207
156
210
390
372
457
264
353
595
470
261
576
392
766
649
466
459
197
182
150
206
100
115
149
127
147
252
158
87
168
182
151
169
64
50
43
83
33
22
475
474
582
579
880
872
856
650
900
1171
825
888
340
241
234
531
274
145
79
60
206
103
140
430
1910
1143
264
8575
161
4087
70
55
45
820
80
40
13
12
15
16
19
23
17
13
29
12
24
23
13
11
9
10
6
10
11
14
17
18
20
27
21
16
31
21
26
25
15
13
10
14
9
11
8
10
12
15
13
16
36
35
19
115
16
20
10
8
6
18
6
6
86
78
85
75
58
70
70
48
68
55
75
73
90
90
78
45
56
65
* Normalized elevation that corresponds to 1% of the normalized area on the hypsometric curve.
Average elevation (HAVG). The average elevation is the arithmetic mean of all the digital
elevation model (DEM) data points within a watershed. The average elevation has important
hydrologic and climatic influence because elevation influences soil, geology, vegetation, and
microclimate of a watershed that, in turn, influence the hydrologic response. Average elevation is a
reasonable measure of the overall watershed elevation, but it can be indirectly influenced by the
presence of very low or very high elevation points.
Median elevation (HMED). The median elevation of a basin is that elevation where half of the
watershed elevation data is higher and the other half is lower. Unlike average elevation, median
elevation is more representative of the watershed elevation because it provides more information on
the elevation distribution within a watershed.
Standard deviation of elevation (HSTD). The elevation standard deviation is a measure of the
variability in watershed elevation. For example, Ridge and Valley watersheds usually have high
20
-------
standard deviations because the ridge areas are rugged with great elevation differences and the
valley areas are nearly uniform (Table 5).
Basin relief (BREL). Basin relief is measured as the difference between the maximum and
minimum watershed elevations. Normally, the lowest watershed elevation is found at the watershed
outlet and the highest elevation is found in the headwater area. Basin relief is an indicator of the
potential energy of the water being drained from the system (Bras, 1990). It is also highly
correlated to drainage area and is an indicator of the overall watershed gradient. High relief may
also indicate the presence of high elevation summits, thus high precipitation inputs and large
recharge and discharge areas within a watershed (Farvolden, 1963).
Relief ratio (RRAT). The relief ratio is defined as basin relief divided by a representative basin
length, usually selected as the distance between the furthest watershed boundary and the watershed
outlet. Among the geomorphologic descriptors, relief ratio is often considered as a good predictor
of the hydrologic response of a watershed.
3.4.2 Slope Parameters
The slope of a watershed determines the direction of flow and flow velocity, and controls soil
erosion from hillslopes and channel areas. Some slope indices used to represent the slope of a
watershed are given in Table 5. A brief description of each slope parameter is also given in the
following.
Maximum slope (SMAX). Maximum slope is a measure of the greatest watershed slope.
Watersheds with very high maximum slopes may indicate the presence of escarpments in the
watershed. For example, watersheds in the Appalachian Plateau and the Ridge and Valley
Provinces have very high maximum slopes (Table 5). Among our study watersheds, the South
Branch Potomac, South Fork Shenandoah, North Fork Shenandoah, Greenbrier, Bullpasture, and
Johns Creek all have very high maximum slopes. Because a maximum slope may correspond to a
single point (e.g., an outlier), this parameter has very limited utility in predicting the hydrologic
response of a watershed.
Average slope (SAVG). Average slope is the arithmetic average of the measured watershed
slopes. It contains very limited information on slope distribution over a watershed since the
presence of a few very high slopes or a few very low slopes can shift the average slope higher or
lower.
Median slope (SMED). The median watershed slope is that slope value where 50 percent of the
measured watershed slopes are higher and the other 50 percent are lower. Median slope is more
representative than the average slope because it is based on the distribution of watershed slope
values. Although it represents the overall slope of a watershed, it does not differentiate between
hillslope and stream channel slope.
21
-------
Standard deviation of slope (SSTD). The standard deviation of slope is the square root of the
variance of all the measured watershed slopes. In general, watersheds with high slope standard
deviation reflect rugged topography while watersheds with low standard deviation reflect
watersheds with more uniform slopes.
3.4.3 Channel Network and Other Parameters
The landscape of a watershed normally consists of upland, hillslope, and channel segments.
Channel parameter values extracted from the digital elevation model (DEM) databases include
total channel length, main channel length, and main channel slope. A brief discussion of each
channel parameter follows. Other geomorphologic descriptors, including drainage area, drainage
density and hypsometric curve are also described. In addition, Figures 5, 6, 7, and 8 present color-
coded digital elevation contours for four selected study watersheds. Figures 5 and 6 show
Appalachian and Ridge and Valley watersheds with deeply dissected valleys and steep slopes
whereas Figures 7 and 8 show Blue Ridge and Piedmont watersheds that have less dissected
valleys and gently sloping hills.
22
-------
N 11.48429° W 77.09959°
N 41.47217"W 77.10523°
N 41.47641" W 77.12133"
Figure 5. A color-coded digital elevation contour of a headwater stream of Blockhouse Creek in
Pennsylvania showing geomorphologic features of the glaciated low plateau ecoregion (60a) of the
Appalachian Plateau Province.
N 38'41.232'W 79'3.125'
N 38-39.615'W 79'3.312'
Figure 6. A color-coded digital elevation contour of a headwater stream of South Branch Potomac
River in West Virginia showing geomorphologic features of the northern dissected ridges and
valleys ecoregion (67'd) of the Ridge and Valley Province.
23
-------
N 39.65254° W 77.51108°
39.65974° W 77.48228°
Figure 7. A color-coded digital elevation contour of a headwater stream of Owens Creek in
Maryland showing geomorphologic features of northern igneous ridges ecoregion (66a) of the Blue
Ridge Province.
N 39.70081° W 76.59380°
N 39.68705° W 76.58681°
N 39.68227° W 76.60160°
0
Figure 8. A color-coded digital elevation contour of a headwater stream of Deer Creek watershed in
Maryland showing geomorphologic features of the Piedmont upland ecoregion (64c) of the
Piedmont Province.
24
-------
Main channel length (MCHL). Main channel length is the centerline length of the main channel
of the watershed. It is the distance between where the channel begins at the headwater areas to the
watershed outlet where the channel arbitrarily ends. Main channel length is closely correlated to
drainage area and is an indicator of flow travel time. Channel length usually affects the shape of
runoff hydrographs because large watersheds tend to have longer times of concentration,
hydrographs with lower rising limb slope, and hydrographs with longer recession periods.
Main channel slope (MCHS). Main channel slope is a measure of the slope of the main stream
channel. It is determined by subtracting the elevation at the watershed outlet from the elevation at
the headwater end and then dividing the difference by the distance between the two points. Main
channel slope also controls flow velocity and travel time.
Total Channel Length (TCHL). Total length of all the channels of a watershed is determined by
adding the length of all the perennial channels within each watershed. In this study, we used
Geographical Information System (GIS) tools to determine total channel length from data obtained
from the National Hydrography Datasets at the website (http://nhd.usgs.gov/data.html).
Drainage area (DARE). Drainage area is the area of a watershed as determined from the
topography of the surrounding watershed divide. The drainage area is an indicator widely used to
develop relationships between watershed characteristics and hydrologic response variables (e.g.,
peak flow and/or low flow rates). Drainage area is a scale factor that introduces heterogeneity in
landscape descriptors when watersheds have different drainage areas. Variable landscape
descriptors would then introduce variability in hydrologic responses. When comparing hydrologic
responses of watersheds with different drainage areas, normalizing the hydrologic response
variable is often desirable. Normalization adjusts hydrologic response differences introduced by
variability in watershed size. In this study, we normalized the hydrologic response variables by
dividing the daily streamflow by the drainage area of each watershed (e.g., cfs/sq. mile).
Drainage density (DDEN). Drainage density is measured by dividing the total length of channels
of the watershed by its total drainage area. Harlin (1984) reported that drainage density is related to
time-to-hydrograph peak, but Berger and Enthekhabi (2001) and Dingman (1978) found no
significant relationship between drainage density and hydrologic response variables. A high
drainage density generally indicates a dense stream network throughout a watershed, whereas low
drainage density indicates a sparse stream network and a watershed with large upland and hillslope
areas relative to its channel areas. Drainage density has some influence on hydrologic response
because it reflects the distance that water has to travel along a hillslope before reaching a nearby
stream. A short hillslope length may not, however, lead to a short travel time because, in humid
regions, subsurface flow is usually controlled by soil and geologic properties of the hillslope
segments (Buttle and McDonald, 2002). Our hypothesis is that, over time, watershed systems have
evolved into efficient water and sediment delivery systems and that their hydrologic response is
well adjusted to the interactions between landscape descriptors (e.g., soil, bedrock geology,
topography, and vegetation) and precipitation input characteristics of each watershed (Troch,
1995). In other words, as an efficient system, the watershed creates only the channel network
needed to transfer the precipitation incident upon it to streamflow at the watershed outlet.
25
-------
Hypsometric curve (HPC10): The hypsometric is an area-elevation relationship curve that plots
normalized elevation against normalized area of a watershed (Langbein et al. 1947; Strahler, 1952).
In this study, the normalized elevation of the hypsometric curve that corresponds to 10 percent of
the normalized area was determined for each watershed and this point is shown in the watersheds
plotted in Figure 9. The hypsometric curves shown in Figure 9 represent four physiographic
provinces and watersheds with variable drainage areas. The hypsometric curves shown in Figure 9
illustrate watersheds with different levels of geomorphic maturity as influenced by various forcing
factors such as tectonics, climate and lithology. The Owens Creek and the Blockhouse watersheds
exhibit low geomorphic maturity level while the Deer Creek and South Branch Potomac
watersheds exhibit high geomorphic maturity level.
3.4.4 Geomorphologic Descriptors as Potential Hydrologic Response Predictors
Dingman (1981) reported that, in the mountainous areas of New Hampshire and Vermont, an
increase in elevation resulted in an increase in precipitation, and snow depth, snow water
equivalents, and a decrease in temperature. Boyer (1984) also reported that most of the spatial
variability of daily temperature means in the central Appalachian region can be accounted for by
differences in elevation and latitude. Elevation, therefore, influences precipitation and temperature.
Changes in precipitation and temperature then influence streamflow and evapotranspiration. As
Dingman (1981) concluded, in the mountainous areas of Vermont and New Hampshire, elevation
could be used as the single, independent variable for predicting streamflow.
Many researchers have reported the influence of the hypsometric curve on the hydrologic
response of a watershed. Harlin (1984) concluded that the hypsometric curve
characteristics might have some predictive power that can enhance rainfall-runoff modeling. Our
hypothesis is that the HPC10 may be related to headwater drainage development, and that a high
HPC10 value may be an indicator of a watershed that has a high headwater elevation. The shape of
the hypsometric curve may be related to the shape of the longitudinal profile of the main stream
channel of a watershed.
26
-------
Ecoregion 60a
Ecoregion 66a
Blockhouse Creek,
Pennsylvania
Appalachian Plateau
Province
Owens Creek, Maryland
Blue Ridge Province
0 0,1 0.2 0,3 0,4 0.5 0.6 0.7 0.8 0.9 0.1 0,2 0.3 0,4 0,5 0,6 0,7 0.8 0,9
Proportion of Area
Proportion of Area
00.5
50,3
a
0.1
a
Deer Creek, Maryland
Piedmont Province
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Proportion of Area
South Branch
Potomac
River, West Virginia
Ridge and Valley
Q 2 ^ M ^ ^ „ 7
Proportion of Area
Figure 9. Hypsometric curves of four study watersheds arranged in the order of decreasing HPC10
((a) having the highest HPC10 and (d) having the lowest HPC10).
27
-------
Part II
Climate Characterization
The climate of the Mid-Atlantic Region is classified as humid to semi-humid continental
because the region has relatively evenly distributed precipitation throughout the year and marked
temperature contrasts between summer and winter. However, within the study area, both
precipitation and temperature vary with elevation and latitudinal position. Elevation has a strong
influence on the climate of the Mid-Atlantic Region. Specifically, the climate of the mountainous
watersheds in the Appalachian Plateau, the Blue Ridge, and the ridge-dominated Ridge and Valley
watersheds differs significantly from the climate of the Piedmont and the valley-dominated Ridge
and Valley watersheds.
4.1 Precipitation
Table 6 presents mean annual precipitation, minimum January temperature, mean watershed
elevation, and latitude and longitude of each study watershed. Mean annual precipitation of the
study area ranges from 889 to 1207 mm (Table 6). In general, precipitation increases with elevation
and even within a physiographic province, some watersheds, depending on their elevation, receive
more precipitation than others. Among the four physiographic provinces of the study area, the
valley-dominated Ridge and Valley watersheds have the lowest mean annual precipitation.
Watersheds with high mean annual precipitation include the Greenbrier Watershed in West
Virginia, the Little Juniata Watershed in Pennsylvania, and the Owens Creek Watershed in
Maryland. These watersheds also receive more precipitation in the form of snow. Depending on
their elevation, some watersheds may receive up to 30 % of their annual precipitation as snow.
4.2 Temperature
The mean annual temperature of the study watersheds varies across physiographic provinces.
Within a physiographic province, temperature also varies with the elevation and latitudinal position
of a watershed. The mean annual temperature of the study area ranged from 39 to 64°F and the
minimum temperature in January from 16 to 27 °F. The Appalachian Plateau watersheds had the
lowest minimum January temperature while Piedmont watersheds had the highest (Table 6).
Watersheds located in mountainous areas of the Appalachian Plateau, Ridge and Valley, and Blue
Ridge Provinces usually have low temperatures. Where the temperature is relatively low,
evapotranspiration may also be low and streamflow may be high. To the contrary, watersheds with
relatively high temperature may exhibit high evapotranspiration and low streamflow. The Piedmont
watersheds belong to the high temperature, high evapotranspiration, and low streamflow category.
28
-------
Table 6.Long-term Mean Annual Precipiation, Minimum Januray Temperature, Mean Elevation,
and Latitude and Longitude of Study Watersheds
Watershed Mean annual Jan. minimum Mean Latitude Longitude
Precipitation (mm) Temp. Deg F. Elevation degrees degrees
meter a.s.l
Appalachian Plateaus
Lehigh(AP) 1143
Towanda (AP) 970
Pine(AP) 1021
Blockhouse (AP) 1036
Greenbrier(AP) 1067
Greenbrier (AP) 1067
Greenbrier (AP) 1041
Ridge and Valley
Wapwallopen (RV) 1057
Marsh (RV) 970
Little Juniata (RV) 1123
Sherman (RV) 1077
Bullpasture (RV) 1018
Johns Creek (RV) 970
South Branch (RV) 889
Cacapon(RV) 914
Patterson (RV) 914
Back (RV) 991
S.F. Shenandoah (RV) 1057
N.F. Shenandoah (RV) 909
Blue Ridge
Owens (BR) 1207
Fishing (BR) 1181
Piedmont
16
18
17
19
19
19
20
18
21
19
23
22
24
20
22
20
23
24
21
26
23
543
495
562
541
1103
957
830
398
397
496
310
801
702
618
495
404
279
468
611
454
457
41:07:49N
41:42:25N
41:31:18N
41:28:25N
38:32:37N
38:11:09N
37:43:27N
41:03:33N
41:03:34N
40:36:45N
40:19:24N
38:11:43N
37:30:22N
39:26:49N
39:34:43N
39:26:35N
39:30:43N
38:54:50N
38:38:13N
39:40:36N
39:31:35N
075:37:33W
076:29:06W
077:26:52W
077:13:52W
079:50:OOW
080:07:51W
080:38:30W
076:05:38W
077:36:22W
078:08:27W
077:10:09W
079:34:14W
080:06:25W
078:39:16W
078:18:34W
078:49:20W
078:02:15W
078:12:40W
078:51:11W
077:27: SOW
077:28:OOW
Deer (PD)
Hardware (PD)
Slate (PD)
Holiday (PD)
1130
1120
1072
1054
24
27
27
27
201
207
156
210
39:37:49N
37:48:45N
37:42:10N
37:24:55N
076:24: 13W
078:27:20W
078:22:40W
078:38: 10W
4.3 Influence of Elevation on Climate and Hydrology
The elevation trend lines shown in Figure 10 illustrate an appropriately linear relationship
between elevation and runoff ratio recorded in March and April. Since elevation influences
temperature and precipitation (Dingman, 1981; Boyer, 1984), it also influences evapotranspiration
and streamflow. For most of the study watersheds, March has the highest runoff ratio because, in
March, the snow that accumulates during the winter melts and generates high streamflow. Figure
10 ranks the study watersheds in the order of increasing runoff ratio for March. The watersheds
ranked toward the left side of Figure 10 had strong elevation influence, whereas the watersheds
ranked toward the right side of Figure 10 had only limited elevation influence. For example,
watersheds with high elevation or latitudinal position influence included the Towanda, Little
29
-------
Juniata, Sherman, and the Greenbrier watersheds. Conversely, watersheds with low elevation
influence were the four Piedmont watersheds shown on the right side of Figure 10. As the spring
and summer seasons progressed, the influence of elevation on runoff ratio decreased until it became
negligible in July, August, and September (Figure 10).
Because elevation has such a strong influence on hydrologic response, we hypothesize that
elevation influences may dominate for most of the year except during the summer months. For
some watersheds, the elevation influence may mask the influences of all the other landscape
descriptors. To moderate the dominance of elevation influences over other landscape descriptors,
i.e. soil descriptors, hydrologic response comparisons were also conducted during the summer
months when the elevation influence was negligible (Figure 10). During the summer period, soil,
vegetation, and geology influences on hydrologic response were less masked by the elevation
effects.
1.4
o
to
1.2 -
1 -
0.8 -
0.6 -
0.4 -
0.2 -
• Mar
-Jul
Sep
-Linear (Elevation (m))
Jun
Aug
—Elevation (m)
- 1000
- 800
1200
.-- 200
f&S^tffi&s^ffis/fyty
*-
Watershed
Figure 10. Influence of Elevation on Seasonal Streamflow Patterns.
30
-------
4.4 Potential Climate Descriptors
To determine useful climate descriptors as predictors of hydrologic response, our approach was
to evaluate a number of potential candidate climate descriptors that showed strong correlation with
the hydrologic responses of the Mid-Atlantic watersheds. Among the candidate climate descriptors,
dryness index seemed to have a strong correlation with hydrologic response descriptors. Indeed,
Berger and Entekhabi (2001) had reported that wetness index, the inverse of the dryness index, was
highly correlated to long-term hydrologic responses of selected basins. A large dryness index value
is indicative of a relatively dry watershed while a small dryness index value reflects a relatively wet
watershed. In other words, a watershed with high streamflow has a low dryness index value
because the water saved due to reduced evapotranspiration becomes available for streamflow.
Using stepwise regression analysis, the climate descriptors listed in Table 7 were tested and the
potentially useful climate descriptors selected. Among the climate descriptors listed in Table 7, the
dryness index (AET/PREC) and mean monthly rainfall depth showed the highest correlation with
the watershed hydrologic response descriptors.
Table 7. Climate Descriptors Examined as Potential Hydrologic Response Descriptors
Variable Description Units
MAP Mean annual precipitation (mm)
MMP (e.g.,JULYPREC) Mean monthly precipitation (mm)
PET Mean annual potential evapotranspiration (mm)
AET Mean annual Actual evapotranspiration (mm)
PET/PREC Mean annual dryness index (potential) (mm/mm)
AET/PREC Mean annual dryness Index (actual) (mm/mm)
-------
5 Hydrologic Response Characterization
To compare the hydrologic responses of the Mid-Atlantic watersheds, we employed a
conceptual approach that is based on the water balance equation (Thornthwaite and Mather, 1955).
This conceptual approach allowed comparison of streamflow, precipitation, and evapotranspiration
of the study watersheds. To examine how the terms of the water balance equation vary over
different time scales, we conducted comparisons of the watershed hydrologic responses at annual,
monthly, and daily time scales. At the annual comparisons, we compared precipitation, streamflow,
and actual evapotranspiration for all study watersheds. At the monthly and daily comparisons,
however, we only compared the hydrologic responses of representative watersheds. For the hourly
time scale comparisons, we did not use the water balance approach, rather we only did hydrograph
comparisons for three representative watersheds.
5.1 Conceptual Approach: Water Balance as a Framework for Hydrologic
Response Comparisons
The water balance equation is a fundamental hydrology equation that is valid across all spatial
and temporal scales (Eagleson, 1978) Theoretically, the water balance equation should serve as the
basis of all hydrologic models, but despite its sound theoretical basis, the water balance equation is
not widely used in hydrologic models. What limits the widespread application of the water balance
equation is the difficulty in measuring or estimating the terms of the water balance equation at
different time scales. Figure 11 shows the inflows, outflows, and storage terms of the water balance
equation for a small, headwater watershed. The water balance equation can be written in its
simplest form as:
R= P-AET-AS (1)
where,
R = Streamflow (controlled by climate, soil, geology, topography, and vegetation)
P = precipitation (climate),
AET = Actual evapotranspiration (climate, soil),
AS = Change in soil moisture storage (controlled by geology, climate, soil etc.).
The different terms of the water balance equation represent or are dominated by different
hydrologic processes. Note that some terms of the water balance equation are more complex and
are controlled by many descriptors while other terms are controlled by only one or two descriptors.
Because it is not always possible to estimate some terms of the water balance equation,
hydrologists often make assumptions that enable them to ignore some of the terms. Some of the
commonly used assumptions are those of no upstream surface and groundwater inflows, and no
groundwater outflows. These latter assumptions eliminate all inflows and all outflows except
streamflow. In other words, these assumptions are based on the hypothesis that watersheds are
hydrologically "isolated"from surrounding watersheds. Another commonly made assumption is
that, on a long-term basis, the average net change in soil moisture storage (AS) approaches zero.
32
-------
Based on these assumptions, the change in soil moisture storage can be eliminated and the water
balance equation can be written in a reduced form as:
R = P - AET
(2)
N 34,10586°W 83,26017'
N 34.10246" W 83.26023°
Evapo-transpiration (E)
0.1
(Go)
Groundwater recharge
0.2
0.3
1
N 34.10591" W 83.26425°
0.2
excMnge (AS)
0.1
foundwater inflow (Gi)
N 34.10251'W 83.26432
Inflows
Outflows
Figure 11. A Schematic Diagram Showing the Components of the Water Balance Equation for a
Small Headwater Subwatershed.
33
-------
5.2 Hydrologic Response Comparisons: Water Balance Approach
5.2.1 Precipitation Data
Daily precipitation was obtained from cooperative weather station databases archived by the
National Climate Data Center (NCDC) in Ashville, NC. The station data closest to each watershed
was used even if the station was located outside the watershed boundary. A principal term of the
water balance equation, precipitation is often measured using a network of rain gages over a
watershed. Although precipitation is a measured term, it is not measured at each point in a
watershed. Lack of spatially distributed precipitation data over a watershed, therefore, introduces
errors in the precipitation input term of the water balance equation.
5.2.2 Streamflow Data
Streamflow is another measured term of the water balance equation. It represents the watershed
outflow term and is mainly influenced by the interactions between climate (precipitation and solar
radiation) and landscape descriptors. Daily Streamflow data was obtained from the USGS website:
http://www.waterdata.usgs.gov. All study watersheds were included in the USGS's Hydroclimatic
Data Network (HCDN) database (Slack and Landwehr, 1992). The HCDN dataset consists of
good-quality hydrologic data compiled by the U.S. Geological Survey. These data had been
recommended for climate change studies being compiled from watersheds that had long records of
Streamflow and from watersheds with limited human-induced disturbances. Most of our study
watersheds have a high percent forest cover. To separate the measured total Streamflow into its
baseflow and stormflow components, computational baseflow separation techniques were used.
5.2.3 Evapotranspiration Data
Unlike precipitation and Streamflow, evapotranspiration is not a measured term of the water
balance equation. Evapotranspiration was estimated using evapotranspiration equations.
Specifically, the daily potential evapotranspiration was estimated using a method developed by
Hamon (1961) and later described by Federer and Lash (1978) and Vorosmarty et al. (1998). The
method uses measured daily minimum and maximum air temperatures, daytime length, and
saturated vapor pressure. The WDMutil Software, part of USEPA's BASINS Modeling System,
was used to calculate the potential evapotranspiration of our study watersheds using Hamon's
Method. In addition to the potential evapotranspiration, the actual evapotranspiration (AET) was
also estimated as the difference between precipitation and Streamflow using Equation 2. Both
estimated actual and potential evapotranspiration data were used to derive climate descriptors, such
as the dryness index.
34
-------
5.2.4 Moisture Storage Data
Measured soil moisture storage data are rarely available. The moisture storage term in the water
balance equation is normally determined by solving the water balance equation after all the other
terms had been either measured or estimated. Moisture storage is an important term of the water
balance equation because it determines the fraction of precipitation input that returns to the
atmosphere as evapotranspiration and the fraction of the precipitation input that leaves the
watershed as both streamflow and groundwater. Landscape descriptors, particularly soil, geology,
and vegetation, have strong influence on moisture storage. Lack of measured soil moisture storage
data often limits the use of the water balance equation as a basis for hydrologic model development
and testing.
5.3 Hydrologic Response Comparisons at Annual Time Scale
For long-term hydrologic response analysis, a three-term water balance equation (Equation 2)
was used to estimate actual evapotranspiration as the difference between the measured values for
precipitation and streamflow. The measured annual precipitation was apportioned into streamflow
and actual evapotranspiration and, for each study watershed, dimensionless ratios (as percent) of
streamflow to precipitation (runoff ratio) and precipitation to evapotranspiration (dryness index)
were calculated (Table 8). For these calculations, the long-term mean annual precipitation and
streamflow data were used. In general, all our study watersheds had comparable long-term mean
annual precipitation values with an overall mean of about 1040 mm, with a range about that mean
of 318 mm. The overall mean annual streamflow and estimated overall mean annual actual
evapotranspiration of the study watersheds were 457 mm and 582 mm, respectively. This means
that, on average, little more than half of the mean annual precipitation was evapotranspirated while
little less than half of the precipitation became streamflow.
The dryness index is a climate descriptor that is highly correlated to hydrologic responses of a
watershed. Comparisons of the dryness indices calculated for our study watersheds show that
dryness index varies from province to province. For example, dryness indices of the Appalachian
Plateau, Blue Ridge, Ridge and Valley, and Piedmont watersheds were 0.47, 0.49, 0.60, and 0.66,
respectively. These comparisons indicate that 47 percent of the precipitation received by the
Appalachian Plateau watersheds left as evapotranspiration, while 66 percent left from the Piedmont
watersheds.
The average annual runoff ratios of the study watersheds also varied from province to province.
The runoff ratios of the Appalachian Plateau, Blue Ridge, Ridge and Valley, and Piedmont
watersheds were 0.53, 0.51, 0.40, and 0.34, respectively. These long-term hydrologic response
comparisons indicate that 53 percent of the precipitation received by the Appalachian Plateau
watersheds left as streamflow, while only 34 percent left from the Piedmont watersheds.
35
-------
Table 8. Long-term Mean Annual Precipitation, Mean Annual Streamflow, Mean Actual
Evapotranspiration Estimates, and Associated Ratios
Watershed Precipitation
(mm)
Streamflow
(mm)
AFT Runoff Ratio Dryness Index
(mm) % %
Appalachian Plateau
Lehigh River, PA
Towanda River, PA
Pine River , PA
Blockhouse Creek, PA
Greenbrier River at Durban, WV
Greenbrier River -Buckeye, WV
Greenbrier River- Alderson, WV
1143
970
1021
1036
1067
1067
1041
691
453
483
517
686
574
499
452
517
538
519
381
493
542
60
47
47
50
66
54
48
40
53
53
50
34
46
52
Ridge and Valley
Blue
Wapwallopen Creek, PA
Marsh Creek, PA
Little Juniata River, PA
Sherman Creek, PA
Bullpasture Creek, VA
Johns Creek, VA
S. Branch Potomac River, WV
Cacapon River, WV
Patterson Creek, WV
Back River, VA
S.Fork Shenandoah River , VA
N.Fork Shenandoah River, VA
Ridge
Owens Creek, MD
Fishing Creek, MD
1057
970
1123
1077
1018
970
889
914
914
991
1057
909
1207
1181
509
453
588
493
460
424
325
313
269
289
327
330
632
591
548
517
535
575
558
546
564
601
645
702
730
579
575
590
48
47
52
46
45
44
37
36
29
29
31
36
52
50
52
53
48
54
55
56
63
64
71
71
69
64
48
50
Piedmont
Deer Creek, MD
Hardware Creek, VA
Slate Creek, VA
Holiday Creek, VA
1130
1120
1072
1054
448
395
350
329
682
725
722
725
40
35
33
30
60
65
67
70
5.4 Hydrologic Response Comparisons at Monthly Time Scale
The annual hydrologic response comparisons presented in the preceding section revealed that
Streamflow varies across physiographic provinces (Table 8). To capture the variability in
Streamflow, i.e., hydrologic response across different physiographic provinces, we grouped the
study watersheds into three categories: high, medium, and low Streamflow. The high Streamflow
category mainly consisted of Appalachian Plateau watersheds whereas medium Streamflow
watersheds mainly consisted of Ridge and Valley watersheds and low Streamflow watersheds
mainly consisted of Piedmont watersheds. Note that some valley-dominated Ridge and Valley
watersheds had lower annual Streamflow than the Piedmont watersheds. To compare the hydrologic
responses of the study watersheds at monthly time scales, we selected two representative
watersheds from each Streamflow category.
The representative watersheds for the high Streamflow category were the Greenbrier Watershed
in West Virginia and the Towanda Watershed in Pennsylvania. Both watersheds are located in the
Appalachian Plateau, but within the high Streamflow category, the Greenbrier Watershed
represented a very high Streamflow subcategory while the Towanda Watershed represented a
relatively low Streamflow subcategory. The two watersheds selected to represent the medium
36
-------
streamflow category were the Back River Watershed in West Virginia and the Little Juniata
Watershed in Pennsylvania. Both these watersheds are located in the Ridge and Valley
Physiographic Province, but within the overall medium flow category, these two watersheds
represent a very high streamflow subcategory and a very low streamflow subcategory. The two
watersheds selected to represent the low streamflow category were the Deer Creek Watershed in
Maryland and the Slate Watershed in Virginia. Both these latter watersheds are located in the
Piedmont Physiographic Province, but within the overall low streamflow category, these two
watersheds represent a high streamflow subcategory and a low streamflow subcategory.
5.4.1 High Streamflow Category Comparisons
The Greenbrier Watershed represents the low evapotranspiration, high precipitation, and high
streamflow watersheds in the Appalachian Plateau. The comparisons of water balance components
at the monthly time scale illustrated that monthly precipitation exceeds monthly potential
evapotranspiration throughout the year (PREC»PET) (Figure 12a). Despite very high
precipitation and relatively low evapotranspiration, it seems that the months with the highest
precipitation did not usually result in the highest streamflow.
For most of the high streamflow watersheds, such as the Greenbrier Watershed, streamflow
showed a high degree of variability. The highest streamflow occurred when snow melts in March.
The lowest monthly streamflow occurred in September when snow influence no longer exists and
when increased evapotranspiration results in depletion of soil moisture storage. Note that
precipitation more than satisfies the evaporative demand of the atmosphere throughout the year and
moisture storage deficit does not occur because potential evapotranspiration is always less than or
equal to actual evapotranspiration.
Comparisons of water balance components of the two representative high flow category
watersheds showed some contrasting differences. The two watersheds had almost equal mean
monthly potential evapotranspiration of 53 mm, so most of the differences were due to differences
in precipitation and streamflow. For instance, the Towanda watershed had 40 percent less mean
monthly precipitation and 33 percent less mean monthly streamflow than the Greenbrier
Watershed. Precipitation exceeded potential evapotranspiration in the Towanda Watershed for most
the year, except in June, July, and August (Figure 12b). During the summer months, soil moisture
storage was depleted by evapotranspiration that far exceeded the moisture that precipitation could
replenish. As a result, unlike the Greenbrier Watershed where precipitation exceeded potential
evapotranspiration throughout the year, the Towanda Watershed experienced a period of moisture
storage deficit.
Differences in elevation may explain the differences between the precipitation inputs to the two
watersheds. Because of their elevation and latitudinal position, both watersheds receive a large
portion of their winter precipitation in the form of snow. The Greenbrier Watershed received 34
percent of its annual precipitation in February, March, and April with about 55 percent of its annual
streamflow occurring in February, March, and April. The Towanda Watershed received 22 percent
of its annual precipitation in February, March, and April, with about 47 percent of its annual
streamflow occurring during these months.
37
-------
For both watersheds, September was the month with the lowest streamflow. In general, moisture
storage builds-up from October to March corresponding to a period of low evaporative demand. As
the evaporative demand of the atmosphere increased, streamflow fell until it reached its lowest in
September. The rate at which streamflow declined in late spring was somewhat proportional to the
rate at which potential evapotranspiration increased during the same period. Streamflow recovered
in October when evapotranspiration started to decrease and when moisture storage for the next
season started to build-up.
38
-------
Greenbrier at Durban, WV
240.00
- 220.00
1 200.00
| 180.00
& 160.00
1 140.00
(a)
] Streamflow
•PREC
•PET
10 11 12
Towanda River near Monroeton, PA
240.00 -
- 220.00 -
£
= 200.00 H
2 3 4 5 6 7 8 9 10 11 12
Figure 12. Monthly Water Balance Components of Two Appalachian Plateau Watersheds: (a)
reenbrier Watershed and (b) the Towanda Watershed.
39
-------
5.4.2 Medium Streamflow Category Comparisons
Two Ridge and Valley watersheds, the Back River and the Little Juniata, were selected to
represent the medium streamflow hydrologic response category. The Back River and the Little
Juniata watersheds represent valley-dominated and ridge-dominated watersheds of the Ridge and
Valley Physiographic Province, respectively. Comparisons of the water balance components of
these two watersheds are shown in FigureslSa and 13b.
These two watersheds received comparable mean monthly precipitation and had nearly similar
mean monthly potential evapotranspiration, but the monthly mean streamflows for the ridge-
dominated watershed were about twice as high as those of the valley-dominated watershed. One
explanation for the differences in mean monthly streamflow may be due to differences in the soil
and groundwater storage characteristics of the two watersheds. It appears that the valley-dominated
watershed stored less water than the ridge-dominated watershed. The reduced moisture storage
capacity of the valley-dominated watershed may indicate that a large percentage of the monthly
precipitation may be was lost as deep groundwater. Note that deep groundwater losses are not often
measured as streamflow because deep groundwater losses may not resurface as baseflow at the
watershed outlet. The valley-dominated watershed also had low elevation and received less
precipitation in the form of snow than the ridge-dominated watershed. For both watersheds, the
highest streamflow was observed in March, and both showed moisture storage deficit in July,
August, and September when potential evapotranspiration exceeded precipitation (PET » PREC).
40
-------
Back River Near Jones Spring , West Virginia
240.00 -
220.00 -
200.00 -
Streamflow
PREC
PET
10 11 12
Little Juniata River, PA
10 11 12
Figure 13. Monthly Water Balance Components of Two Ridge and Valley Watersheds: (a) Back
River Watersheds, (b) Little Juniata Watershed.
41
-------
5.4.3 Low Streamflow Category Comparisons
Figures 14a and 14b show water balance components of two Piedmont watersheds that represent
the high evapotranspiration and low Streamflow hydrologic response category. Comparisons of the
monthly water balance components showed that the hydrologic responses of the Piedmont
watersheds were very different from the hydrologic responses of watersheds located in other
physiographic provinces. The reasons for these differences were elevation differences, proximity to
the Atlantic Coast, and reduced snow accumulation. In comparison to the watersheds in other
physiographic provinces, the Piedmont watersheds had less snow-generated Streamflow in March
and, had therefore, lower seasonal variability of Streamflow.
In addition, the Piedmont watersheds also had evenly distributed precipitation and the
availability of high soil moisture storage capacity. Among the physiographic provinces, watersheds
in the Piedmont Province are characterized by low mean monthly Streamflow, high potential
evapotranspiration (PET»PREC), and a long period of moisture deficit (May, June, July, and
August). Comparison of the water balance components of the two Piedmont watersheds revealed
that the Deer Creek Watershed in Maryland had 25 percent higher Streamflow, 13 percent higher
precipitation, and 69 percent higher potential evapotranspiration than the Slate Watershed.
42
-------
Deer Creek at Rocks, MD
2 3 4 5 6 7 8 9 10 11 12
Slate River Near Arvania, VA
240.00
220.00 1
11 12
Figure 14. Monthly Water Balance Components of Two Piedmont Watersheds: (a) Deer Creek
Watershed and (b) Slate Watershed.
43
-------
5.5 Hydrologic Response Comparisons at Daily Time Scale
Water balance components at the daily time scale were made using three representative
watersheds. The selected representative watersheds were the Blockhouse, the Patterson, and the
Hardware Watershed, representing the Appalachian Plateau, Ridge and Valley, and the Piedmont
Physiographic Provinces, respectively. Because of elevation influences on hydrologic responses
during winter and spring seasons, daily water balance data for the months of July, August, and
September was used to compare the hydrologic responses of the three representative watersheds.
This period was selected for hydrologic response comparisons at the daily time scale. Note that
hydrologic responses at the daily time scale had larger variability than the monthly and annual
hydrologic responses. The main source of variability at the daily time scale was the variability
associated with the temporal and the spatial distribution of daily rainfall and daily streamflow.
Because of the random nature of precipitation occurrence, temporal variability in precipitation
dominates the hydrologic response at the daily time scale.
The method used to analyze the components of the water balance equation at the daily time
scale is based on the determination of dimensionless indices from daily precipitation, potential
evapotranspiration, and streamflow. As shown in the hydrologic response analysis at the monthly
scale, this period was almost free of elevation influences on streamflow and coincided the period
when most of the watersheds experienced moisture deficit.
To eliminate any differences introduced by differences in drainage area, we converted daily
rainfall, potential evapotranspiration, and streamflow into equivalent depth (e.g., mm/day). The
Blockhouse, Patterson, and Hardware watersheds had 247, 295, 393 mm of rainfall during the
selected three month comparison period. To obtain an average daily rainfall depth (P) for the
three-month period, we divided the sum of the rainfall amount observed over the three-month
period by the number of days. The resulting average daily rainfall depth was then assigned to all
the days including the rainy days. Using the average daily rainfall depth, the following indices were
calculated for each representative watershed:
PET IPREC = Potential dryness index (3)
l-Q/PREC = Mean actual dryness index (4)
PREC - PET - Q = Change in storage (5)
44
-------
The period selected to compare the water balance components at the daily time scale
corresponds to the period when most of the study watersheds experience moisture deficits. The
daily time scale comparisons showed more variability in the water balance components and
provided more detailed hydrologic response comparisons of the three representative watersheds.
The water balance terms of the three representative watersheds compared were potential dryness
index, mean actual dryness index, and moisture storage as calculated by Equations 3, 4, and 5,
respectively.
Potential dryness index comparisons showed that the Blockhouse and Patterson Creek
watersheds had higher potential dryness index values than the Hardware Watershed (Figures 15,
16, and 17). Watersheds with high rainfall and low evapotranspiration are usually wetter and
exhibit a low dryness index. For all three watersheds, the potential dryness index decreased as the
season comparison progressed, the highest potential dryness index was observed in early July while
the lowest was observed in September. This decrease in potential dryness index was mainly due to
a decrease in temperature, resulting in a decrease in the potential evapotranspiration.
Blockhouse Creek, PA
-Actual dryness index
• Moisture storage (mrrVdy)
Potential dryness index
Streamflow
1.5
12
10
61 66 71 76 81
14
1
CO
Days from July 1 to September 21,1975
Figure 15. Comparisons of Water Balance Components of an Appalachian Watershed at Daily
Time Scale.
45
-------
As noted previously, the mean actual dryness index was calculated using Equation 4 that
assumes zero net soil moisture storage. As a function of actual soil moisture storage, the mean
actual dryness index was lower than the potential dryness index whenever a soil moisture storage
deficit existed, was equal to the potential dryness index when soil moisture storage was equal to
zero, and became higher than the potential dryness index as the soil moisture storage surplus
increased.
In summary, a soil storage surplus occurred whenever the potential dryness index was less than
the mean actual dryness index. The difference between potential dryness index and mean actual
dryness index reflects the amount of moisture deficit or moisture surplus in a watershed. Among
the three representative watersheds, the potential dryness index of the Hardware Watershed was
nearly equal to the mean actual dryness index throughout the comparison period. When the
potential dryness index is nearly equal to the mean actual dryness index line, soil moisture storage
is nearly zero and potential evapotranspiration is nearly equal to the actual evapotranspiration. For
the Blockhouse and Patterson watersheds, the potential dryness index was greater than the mean
actual dryness index during the early part of the comparison season, and was lower than the mean
actual dryness index at the end of the three-month season.
Patterson Creek, WV
2.5
2
1.5
-2.5 J=
• Potential dryness index
•Moisture storage (mm/day)
Actual dryness index
Streamflow
12
61 64 67 70 73 76 79 8:
14
Days from July 1 to September 22,1975
Figure 16. Comparisons of Hydrologic Response of a Ridge and Valley Watershed at the Daily
Time Scale
46
-------
Comparisons of the estimated soil moisture storages of the three representative watersheds
showed that the Hardware watershed of the Piedmont Physiographic Province generally exhibited
high moisture storage. Moisture storage comparisons also revealed some useful insights about the
storage characteristics of the three watersheds. For example, the Hardware Watershed not only had
high storage capacity but also a controlled storage release mechanism. As a result, the streamflow
of this Piedmont watershed rarely declined to low levels because streamflow was sustained by
release of moisture stored in the soil regolith. By comparison, the Blockhouse Watershed of the
Appalachian Plateau Province and the Patterson Creek of the valley-dominated Ridge and Valley
Province watersheds, had low soil moisture storage. In general, those watersheds with high soil
moisture storage capacity maintained sustained low flows during the summer when potential
evapotranspiration exceeded precipitation (PREC< PET).
2.5
Hardware Creek, VA
Potential dryness index
Moisture Storage (mm/day)
Actual Dryness index
Streamflow
Days from July 1 to September 22,1975
Figure 17. Comparisons of Water Balance Components of a Piedmont Watershed at the Daily
Time Scale
47
-------
5.6 Hydrologic Response Comparisons at the Hourly Time Scale
At the hourly time scale, the water balance components were highly variable and comparisons
made at the hourly scale may not be meaningful. As a result, hourly water balance comparisons
were not made, but hydrograph comparisons were made. The volume of the runoff hydrograph is
influenced by the drainage area of the watershed, initial moisture storage condition, and the
characteristics of the storm event. Watersheds with large drainage areas tend to have large runoff
volumes - assuming similar precipitation inputs and antecedent soil moisture conditions. To
compare hydrographs of watersheds with different drainage areas at the hourly time scale, the
hourly streamflow data was converted to equivalent depth (mm/day).
Because the three watersheds received rainfall with different characteristics, the watersheds
showed runoff hydrographs that had different hourly peak flows. To adjust for differences in
hydrograph peak flows, we divided all hourly streamflow data by the highest hourly peak flow
depth. Figure 18 plots the logarithm of standardized streamflows of the three hydrographs adjusted
to a unit peak flow depth versus time for the last 13 days of the comparison period.
Stormflow response comparisons
Q.
a
151 176 201 226 251 276 301
0.01
Blockhouse
Hardware
Patterson
Hours from August 31 00:00 to September 12, 2003 19:00
Figure 18. Comparisons of Standardized Hourly Runoff Hydrographs
Comparisons of the adjusted hydrographs (Figure 18) showed that the Hardware Watershed
generally had higher and relatively consistent adjusted streamflow than the other two watersheds.
When adjusted, the Blockhouse and the Patterson Creek Watersheds seemed to have similar
streamflows, both more variable than that of the Hardware Watershed. Figure 18 also shows the
recession curves of the three representative watersheds. For the Piedmont Watershed, the recession
curve remained steady with almost zero slope; for the Appalachian Plateau watershed, however, the
recession curve decreased sharply soon after the rainfall ended and then decreased slowly with
48
-------
relatively flat slope. Unlike the other watersheds, the Ridge and Valley watershed had a recession
curve that declined steadily during the recession period.
The shape of the three recession curves may provide some valuable information about the
storage and release properties of the soil and geology of the representative watersheds. As the
recession curve indicated, the Piedmont watershed had adequate soil moisture storage that could be
released to maintain streamflow during non-rainy days. By contrast, the recession curve of the
Patterson Creek Watershed indicated that available soil moisture storage could not sustain
streamflow during non-rainy days. That is to say the recession curve showed that storage releases
were constantly decreasing with time and could not, therefore, maintain sustained low flows. The
Blockhouse Watershed behaved as if it had two distinct storage compartments. The first
compartment may constitute a fast release of water stored in the area near the stream channels at
the end of a rainfall event. After releases from this first compartment have ended, releases from the
second soil moisture storage compartment are low and can only maintain a very low streamflow.
49
-------
Part III
6 Hydrologic Response Prediction
To develop relationships between landscape-climate descriptors and hydrologic response
descriptors, one must first identify and quantify the key descriptors that can represent landscape,
climate, and hydrologic response. A major challenge in hydrologic response comparison studies at
different watershed scales is how to identify hydrologic response variables that can represent a
wide range of hydrologic conditions. In an effort to identify hydrologic response descriptors that
can represent high, medium, and low flows, we evaluated several potential hydrologic response
variables, such as mean, minimum and maximum flow. We ultimately selected the flow duration
curves as the most suitable hydrologic response descriptor because the flow duration curve (FDC)
synthesizes a lot of information on hydrologic response of a watershed (Bonta and Cleland, 2003).
Flow duration curves cover a wide range of flow conditions and provide valuable information
about streamflow variability over time (Smakhtin, 2001). The FDC synthesizes complex and long-
term records of time series streamflow data as a graphical display that illustrates the relationship
between flow magnitude and the frequency that is associated with each magnitude. Flow duration
curves are continuous, but to represent them herein we selected only 10 discrete points on the
curve. These 10 points are the normalized streamflow values (cfs/mi2) that correspond to the 1, 5,
10, 20, 30, 40, 50, 70, 90, and 95 percent exceedance probabilities. We denoted these 10 points on
the flow duration curve as flow duration indices (FDIs). These points define the shape of the flow
duration curve and characterize a wide range of flow conditions that include both high frequency
low flows (Q95) and low frequency very high flows (Ql) (Figure 19). Throughout this report, we
denote these 10 points as flow duration indices (FDIs) that represent the hydrologic responses of a
watershed.
Unlike other hydrologic response descriptors such as mean, minimum and maximum flows, the
flow duration indices (FDIs) represent a wide range of hydrologic responses. For example, Ql
corresponds to a very high flow (flood condition), whereas Q5 and Q10 indices correspond to high
flows that are equaled or exceeded only 5 percent and 10 percent of the time, respectively. Flows
that correspond to Q20, Q30, Q40, and Q50 FDIs are considered as medium flows, equaled or
exceeded 20, 30, 40 and 50 percent of the time, respectively. Likewise, flows that correspond to
Q70, Q90 and Q95 indices are low flows, equaled or exceeded 70, 90 and 95 percent of the time,
respectively.
FDCs graphically illustrate the percent of time during which a streamflow value is equaled or
exceeded over a given period of observation. A number of other investigators have reviewed the
application of FDCs for water resources management and planning (Searcy, 1959; Vogel and
Fennessey, 1994, and Smakhtin, 2001). Searcy (1959) provided a summary of FDC applications
that includes an analysis of the relationship between flow duration curve shape and watershed
geology, hydropower, and water quality issues. Recently, Vogel and Fennessey (1994) reviewed
FDC use and presented a number of applications that included water resources management and
50
-------
wasteload allocations. Smakhtin (2001) also reviewed low flow hydrology and stated that flow
duration curves are one of the most informative methods of displaying the complete range of
streamflow conditions - from low flows to high flows. FDCs have also been used for the
determination of wastewater treatment plant capacity (Male and Ogawa, 1984), hydropower
feasibility studies (Warwick, 1984), estimation of optimal reservoir release schedules (Alouze
1991), design of flow diversions (Pitman, 1993), assessment of in-stream flow requirements for
river habitats, and calibration of rainfall-runoff models (Gustard and Wesselink, 1993).
When observed streamflow data is available, the FDC can be determined therefrom. The
problem is that observed streamflow is not always available because many small watersheds are
ungaged. Managing ungaged watersheds is a problem for environmental planners and resource
managers who often need these data to make sound resource management decisions. FDCs have
been used to predict streamflow for ungaged watersheds. Because FDCs have important water
resources management applications, in recent years, there has been a renewed interest in predicting
FDCs for ungaged watersheds (Croker et al. 2003). Yu et al. (2002) presented a new method to
predict FDCs from annual rainfall, altitude, and drainage area. Dingman and Lawlor (1995)
predicted annual minimum seven-day average flow (7Q) for ungaged and unregulated drainage
basins in New Hampshire and Vermont from drainage area, elevation, and percent of watershed
covered by sand and gravel deposits.
One of the key objectives of this study is to identify suitable hydrologic response descriptors
and then develop relationships between the landscape-climate descriptors and the selected
hydrologic response descriptors (flow duration indices). Once the relationships between the
landscape-climate descriptors and the flow duration indices are developed in the form of regression
equations, these regression equations can be used to predict flow duration curves and streamflow
for ungaged watersheds.
6.1 Approach
Spatial scale has strong influence on watershed hydrology because streamflow rate increases
with an increase in drainage area. When comparing watersheds with different drainage areas, it is
necessary to eliminate the effect of spatial scale (drainage area) on streamflow by dividing the daily
streamflow values by the drainage area of each watershed (cfs/sq.mile). After normalization, the
normalized daily streamflow data (cfs/sq. mile) was used to generate the flow duration curves.
Using the normalized streamflow, we conducted a multivariate analysis to discover relationships
between landscape-climate descriptors and the hydrologic response descriptors (flow duration
indices). Note that the normalized streamflow data that correspond to the 10 flow duration indices
represent the dependent variables in the multiple regression equation while the landscape and
climate descriptors represent the independent variables. A large number of quantitative landscape
and climate descriptors were examined. The landscape-climate descriptors represented climate
(e.g., precipitation and climate-related dimensionless indices), soil (e.g., saturated hydraulic
conductivity, soil depth, and texture, etc.), geology (e.g., baseflow index), land use and land cover
(percent of deciduous forest cover), and topography (e.g., elevation and slope parameters). Factor
analysis and correlation analysis were used to reduce the number of landscape and climate
descriptors. These analyses identified landscape and climate descriptors that were further evaluated
51
-------
to develop relationships between the landscape-climate descriptors and hydrologic response
descriptors using a stepwise regression analysis.
Two periods of record were selected to generate flow duration curves for all our study
watersheds. The two periods represented two time scales: a long period of record, based on 21
years (1980 to 2000) of daily streamflow data, and a short period of record, based on a single year
(October 1974 to September 1975) of streamflow data. Throughout this report, we refer to the 21-
year period as the long period of record or the composite flow duration curve period and the one-
year period as the single year period. The 21-year indices are also based on regional scale data
aggregation while the one-year indices are based on physiographic province spatial scale data
aggregation.
Flow duration curves based on a long period of record are suitable for the determination of low
flow indices that can be used for establishing ecological flow requirements as well as for
establishing high flow indices that can be used for assessing the risk of flooding. In summary,
FDCs that are based on a long period of record may be suitable for water resources management
while FDCs based on a single year record are more suitable for hydrologic modeling applications.
100
1 10
1 --
0.1
- Streamflow
Q10
--Q40
-Q90
Q1
Q20
— - -Q50
Q95
Q5
Q30
Q70
Flow duration curve
Days
Figure 19. A One-year Flow Duration Curve Showing Lines that Represent the 10 Flow Duration
Indices (FDIs) and Normalized Daily Streamflow Data.
52
-------
Figure 19 illustrates the relationship among the daily streamflow, the flow duration curve and
the 10 flow duration indices. Note that each flow duration index line intersects both the streamflow
time series and the flow duration curve. The intersection points represent the magnitude of each
flow duration index. These intersection points are useful when reconstructing a flow duration curve
from the 10 flow duration indices. The intersection points between the flow duration indices and
the streamflow time series curve link the flow duration indices to the actual streamflow data. For
example, the Ql index line crosses only the peaks of the two highest daily streamflow hydrographs
while the Q5 index line crosses a number of smaller peak flow hydrographs and both the rising and
recession limbs of some high flow hydrographs.
This approach determines 10 flow duration indices (Ql to Q95) as a function of the landscape-
climate descriptors. These predicted flow duration indices (Ql to Q95) can then be used to
reconstruct the entire flow duration curve. After the flow duration curve is reconstructed, then the
normalized flow duration curve is converted to streamflow rates by multiplying the values by the
drainage area of the watershed. This approach is proposed for the prediction of streamflow for
ungaged watersheds where lack of data currently restricts the use of physically-based hydrologic
models.
6.2 Comparisons of Hydrologic Responses at the Regional Scale
The hydrologic response of a watershed may not be adequately defined by a single response
descriptor such as mean flow, but can be defined by the 10 flow duration indices because these
indices cover a wide range of flow conditions that include very high, high, medium, low, and very
low flows (Figure 19). To compare the hydrologic responses across the Mid-Atlantic Region, we
grouped the 10 flow duration indices into three categories that represented high flow (Ql, Q5, Q10,
Q20), medium flow (Q30, Q40, Q50), and low flow conditions (Q70, Q70, Q95). The hydrologic
responses of the study watersheds were then compared by ranking the study watersheds in the order
of increasing high flow represented by the Q5 index, medium flow represented by the Q50 index,
and low flow represented by the Q95 index.
Some FDIs, such as the Q5 and Q10, were highly correlated to each other, and watersheds that
had high Q5 index also had high Q10 and Q20 indices. Figure 20 displays the study watersheds
ranked in the order of increasing Q5 index. Note that the Q5 index corresponds to the normalized
streamflow value that is equaled or exceeded only 5 percent of the time. The Q5 rankings showed
that the study watersheds seem to follow the physiographic province arrangement. Among the
study watersheds, the Appalachian Plateau and Blue Ridge watersheds had relatively high Q5
indices followed by the ridge-dominated Ridge and Valley watersheds. Figure 20 also shows that
the Piedmont and valley-dominated watersheds had the lowest Q5 indices. Because Q5 index does
not represent rare flood events that are highly dependent on rainfall characteristics, watersheds with
high Q5 indices may have soils with low infiltration capacity while, on the other hand, watersheds
that have low Q5 index have soils with high infiltration capacity (Cole et al. 2003). We
hypothesized that the Q5 index was a good indicator of the dominant runoff mechanisms of a
watershed and that watersheds, with high Q5 indices had higher surface runoff components than
watersheds with low Q5 indices.
53
-------
16
14
0)
I
&10
1
M—
I 8
+J
5 ej
"(5
2
•Q1
Q10 -»-Q20
•Q30
0:
Q
0.
ct:
CD
t Q
I
o
r^
(D
o
O)
(D
ce ce .2
T3
r^
(D
o
r^
(D
D)
r^
(D
CO
£
o
CO
CD 2
•D
ra
i
Q
D.
Q
D.
O
Watershed
Figure 20. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q5 Index.
Comparisons of the hydrologic responses of the study watersheds at medium flow conditions are
given in Figure 21. To see how different watersheds compare at medium flow conditions, we
ranked the study watersheds in the order of increasing median flow, i.e., Q50 index. The Q50 index
rankings do not follow the physiographic province arrangement. When compared to the Q5 index
rankings, the Q50 index rankings show a slight shift in the order of arrangement of the watersheds.
It appears that, as a group, the two Blue Ridge watersheds had the highest medium flows, followed
by the Appalachian Plateau, and the ridge-dominated Ridge and Valley watersheds. The valley-
dominated Ridge and Valley watersheds had the lowest Q50 indices. It is interesting to note that
most of the Piedmont watersheds also have low Q50 indices. This implies that, most of the time,
the Piedmont watersheds are more likely to have low flows than to have high flows or even high
medium flows.
54
-------
3.5
-5T 3
"55 Z5~
M—
^
T 9
5
E
re
£ 1.5
M
•o
0)
= 1
(0
0.5-
Q70
Q_
<
co
O
O>
CD
IS
05
CD
ro o o
CD CD CD
CC > Q
CO CC Q.
.a ro
h- O
CD CD,
> (D
3 >
T3
CM
CD
13
CM
CD,
cu
m
T3
i^
CD
D)
!c
CD
D) i_
C (D
CD
£
_
O
O
ro
c
CD
SI
co
(D
S
CL
Watershed
Figure 21. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q50 Index.
The Q95 index was used to compare the hydrologic responses of the study watersheds at low
flow conditions. When the study watersheds were arranged in the order of increasing Q95 index,
watersheds that had low risk for droughts were placed on the left side and watersheds that had high
risk for droughts were placed on the right side of Figure 22. The Q95 rankings did not group the
study watersheds according to physiographic province or within a physiographic province; some
watersheds had high Q95 indices while others had low Q95 index. In general, however, Q95 index
was highly correlated to baseflow index; watersheds that had high baseflow indices had also high
Q95 indices.
The Q95 rankings revealed that, as a group, the Piedmont watersheds had sustained low flows
and were less vulnerable to drought. The Piedmont watersheds have desirable hydrologic response
characteristics that are influenced by both the soil moisture storage and release characteristics of
the thick soil regolith. The Q95 index rankings also showed that the Appalachian Plateau and Ridge
55
-------
and Valley watersheds were generally vulnerable to droughts because these watersheds had low
Q95 indices.
JO
lo
o
5Z
03
0 v-W&
^
£ & &* J? z? £? &• <& £ & •& £ & ^ fr &• &
•'//*'"/'///'*////
Watershed
/
Figure 22. Comparisons of Hydrologic Responses Across Study Watersheds Using the Q95 Index.
6.3 Prediction of Flow Duration Indices (FDIs) for Ungaged Watersheds at
Regional Spatial Scale and Multi-year Time Scale
Table 9 presents a correlation matrix between landscape-climate descriptors and the hydrologic
response descriptors, i.e., the FDIs (Ql.. .Q95). The correlation matrix shown in Table 9 was based
on a long period of record (1980-2000). It shows that flood conditions characterized by the Ql
indices were negatively correlated to the dryness index (AET/PREC) and the saturated hydraulic
conductivity of the upper soil layer, and were positively correlated to the bulk density of the top
soil layer, percent of deciduous forest cover, and the normalized hypsometric elevation that
corresponds to 1 percent of the normalized area. Table 9 also shows that medium flows,
56
-------
characterized by the Q50 index, are negatively correlated to dryness index (AET/PREC) and
saturated hydraulic conductivity of the upper soil layer. However, the Q50 index was positively
correlated to soil depth of the upper soil layer and the normalized elevation that corresponds to 1
percent of the normalized area of the hypsometric curve. Unlike the high and medium flows that
were highly correlated to dryness index, low flows characterized by the Q95 index, were poorly
correlated to the dryness index, but were highly correlated to the baseflow index (BFI).
57
-------
Table 9. Correlation Coefficients Between Landscape, Climate, and Selected Hydrologic Response Descriptors
Ql
Q20
Q50
Q95
BFI
Rr
MdSL
AET/P
KSA1
BD2
AWC2
KSA2
CLAY2
SILT2
FRSD
SOLD
HPC10
Ql
1.00
0.64
0.44
-0.13
-0.23
0.35
-0.04
-0.79
-0.44
0.70
-0.04
-0.19
0.10
0.08
0.69
0.08
0.56
Q20
1.00
0.90
0.28
0.30
0.55
-0.17
-0.88
-0.50
0.60
0.19
-0.25
0.35
0.20
0.56
0.22
0.62
Q50
1.00
0.59
0.62
0.40
-0.37
-0.81
-0.53
0.38
0.39
-0.32
0.41
0.32
0.37
0.42
0.45
Q95
1.00
0.86
0.04
-0.28
-0.29
-0.19
-0.03
0.21
-0.15
0.10
0.21
0.04
0.34
0.09
BFI
1.00
0.27
-0.47
-0.19
-0.24
-0.07
0.57
-0.34
0.48
0.45
0.01
0.57
0.07
Rr
1.00
-0.09
-0.34
-0.18
0.50
0.35
-0.33
0.63
0.43
0.51
0.09
0.43
MdSL
1.00
0.21
0.51
0.06
-0.66
0.32
-0.39
-0.42
0.11
-0.73
0.01
AET/P KSA1 BD2 AWC2 KSA2 CLAY2 SILT2 FRSD SOLD HPC10
1.00
0.48 1.00
-0.63 -0.42 1.00
-0.01 -0.52 -0.08 1.00
0.12 0.65 -0.32 -0.68 1.00
-0.06 -0.36 0.09 0.85 -0.63 1.00
-0.07 -0.49 0.11 0.78 -0.85 0.65 1.00
-0.70 -0.21 0.69 -0.18 -0.17 0.05 0.06 1.00
-0.24 -0.69 0.04 0.74 -0.47 0.53 0.46 0.06 1.00
-0.69 -0.27 0.54 -0.04 -0.01 0.01 0.14 0.62 0.14 1.00
58
-------
6.3.1 Ql Model
The 10 flow duration indices (FDIs) correspond to flows with different magnitudes and
frequencies. For instance, the Ql index represents a very high flow condition that can be predicted
from climate, geology, and soil descriptors (Table 10). The stepwise multiple regression analysis
identified the best predictors of Ql index as dryness index (AET/PREC), a climate descriptor,
baseflow index, a geology descriptor, and a soil descriptor - percent clay in the second soil layer.
Soil descriptors control the infiltration and moisture storage processes; climate controls the
precipitation inputs (rainfall and snow), and geology controls groundwater discharge (i.e., BFI =
baseflow index). The results of the regression analysis also revealed a negative relationship
between BFI and the hydrologic response index Ql that may indicate that watersheds with very
high baseflow index do not usually generate very high flows. The high flows represented by the Ql
index are mainly influenced by climate particularly precipitation characteristics. The percent of
clay in the second soil layer (CLAY2) was also a good predictor of the Ql index and may indicate
the presence of flow impeding layers and possibly the occurrence of very high subsurface lateral
flows from hillslopes.
Table 10. Flow Duration Indices Equations for Regional Spatial Scale and Multi-year Time Scale
FDIs Regression equations Coefficient of Determination
Ql =27.38-22.95 (AET/PREC)-12.95(BFI)+0.075 (CLAY2) R2 = 0.85
Q5 = 11.83-10.82 (AET/PREC)-0.025(Rrat)-2.389(BFI)-3.976(AWC2) R2 = 0.95
Q10 = 7.61-6.63 (AET/PREC)+0.019(Rrat)-0.029(SILT2)-0.187(KSATl) R2 = 0.92
Q20 = 6.38-7.67 (AET/PREC)+0.028(CLAY2)-0.011 (SOLD)-0.485 (FRSD) R2 = 0.92
Q30 = 4.59-6.44 (AET/PREC)+0.02(CLAY2)-0.673 (FRSD)+0.016(MDSL) R2 = 0.92
Q40 = 3.06-4.787 (AET/PREC)+0.851(BFI)+0.01 (CLAY2)-0.399 (FRSD) R2 = 0.93
Q50 = 2.263-3.66 (AET/PREC)+1.027(BFI)-0.379(FRSD)-0.007(CLAY2) R2 = 0.93
Q70 = 0.736+1.419(BFI)-1.42(AET/PREC)-0.437(HPC10)+0.015(SOLD) R2 = 0.95
Q90 = -0.206+1.19 (BFI)-0.005(CLAY2)-0.031(Rrat) R2 = 0.88
Q95 = -0.15+1.02(BFI)-0.004(CLAY2)-0.602(AWC2) R2 = 0.88
Symbol descriptions are given in Tables 2 and 7.
6.3.2 Q5 and Q10 Models
Q5 and Q10 indices represent high flow conditions that correspond to peak daily flows
generated by medium to low intensity rainfall events or flows observed at the rising or receding
limbs of high flow hydrographs (Figure 19). Both Q5 and Q10 indices are highly correlated to
dryness index and relief ratio. Moreover, Q5 is also correlated to baseflow index and available soil
water content in the second soil layer (AWC2). The landscape descriptors that are correlated to the
59
-------
Q10 index include the percent of silt in the second soil profile layer (SILT2) and the saturated
hydraulic conductivity of the upper soil profile layer (KSAT1). Q10 was also highly correlated to
Q5 and similar watershed characteristics may influence both Q5 and Q10 flow conditions (Figure
20). We hypothesized that the Q5 and Q10 indices represent a period of high soil moisture
availability, when the hydrologic responses of a watershed are controlled by topographic
descriptors that control the flow conveyance system, i.e., relief ratio, stream channel length, and
drainage density.
6.3.3 Q20 and Q30 Models
Streamflow magnitudes that correspond to Q20 and Q30 flow duration indices were classified
as high to medium flows. As shown in Figure 19, Q20 and Q30 indices corresponded to the rising
or the recession limbs of the Streamflow hydrographs. As shown in Table 10, the dryness index
(AET/PREC) was the best predictor of Q20 and Q30 indices, followed by the percent of clay in the
second soil profile layer (CLAY2). Other landscape descriptors that were highly correlated to the
Q20 and Q30 indices included percent of deciduous forest cover (FRSD), depth of the top two soil
profile layers (SOLD), and median watershed slope (MDSL). We hypothesize that the Ql, Q5, and
Q10 indices reflect conditions so wet that the watersheds are energy limited for maximum
evapotranspiration to occur, while the Q40, Q50, Q70, Q90, and Q95 indices reflect conditions too
dry for maximum evapotranspiration to occur. Based on this hypothesis, Q20 and Q30 indices
correspond to relatively wet hydrologic conditions that are characterized by the availability of
sufficient water and energy to support maximum evapotranspiration.
6.3.4 Q40 and Q50 Models
Q40 and Q50 indices correspond to the low end of the medium flow conditions and to the high
end of the low flow conditions. Note that the Q50 index corresponds to the flow that is equaled or
exceeded 50 percent of the time. Among the hydrologic response descriptors, climate had a strong
influence on the Q40 and Q50 indices and dryness index was the best predictor of the Q40 and Q50
indices (Table 10). In addition to dryness index, baseflow index (BFI), which was used as a
surrogate descriptor for geologic properties, emerged as the second best predictor for the Q40 and
Q50 indices. The inclusion of baseflow index as a predictor for the Q40 and Q50 indices indicated
the beginning of a shift in hydrologic response. This shift implies that as watersheds become dry
soil and geology have more influence on hydrologic response than climate. Other landscape
descriptors that are correlated to medium flow conditions (i.e., the Q40 and Q50 indices) include
the percent of clay in the second soil profile layer (CLAY2) and the percent of deciduous forest
cover (FRSD).
6.3.5 Q70, Q90 and Q95 Models
The climate descriptors that correlated best to the high flow conditions had no strong correlation
to the low flow conditions. For example, dryness index, the best predictor for medium and high
flow FDIs, was the second best predictor for the Q70 index and had no significant correlation with
Q90 and Q95. Unlike the high flow conditions where climate was the dominant hydrologic
60
-------
response predictor, it appeared that low flow conditions were mainly influenced by soil and
geologic descriptors. The regression equations shown in Table 10 indicated that base flow index
was the best predictor of the Q70, Q90, and Q95 indices. Among the landscape and climate
descriptors, baseflow index explained a high percentage of the variance associated with low to very
low FDIs (Q70, Q90, and Q95). The percent of clay in the second soil profile layer (CLAY2) was
the second best predictor for the Q90 and Q95 indices. Other descriptors that influenced low flow
conditions included soil depth of the top two soil profile layers (SOLD), relief ratio (RRAT), and
soil water content in the second soil profile layer (AWC2).
6.4 Predicting FDIs for Ungaged Watersheds at the Physiographic Province
Spatial Scale and Single-year Time Scale
FDCs vary with the period of record. For example, a 30-year based FDC would differ from a 10-
year based FDC or a one-year based FDC. Table 11 presents regression equations that relate
landscape-climate descriptors to hydrologic response descriptors for a single year. Unlike the
equations developed for the long-term regional flow duration indices (Table 10), the single-year
based FDI regression equations are not applicable to the entire region, but are only applicable to
specific physiographic provinces within the Mid-Atlantic Region (Table 11). The regression
equations presented in Table 11 indicate that different landscape-climate descriptors dominated the
hydrologic responses among the different physiographic provinces.
The landscape-climate descriptors, particularly the dryness index, were the best FDI predictors
for both the single-year and multi-year flow duration indices for the Ridge and Valley watersheds.
The similarities between the regional regression equations (Table 10) and physiographic-based,
single-year equations (Table 11) for Ridge and Valley watersheds might be explained by the fact
that almost half of the study watersheds are located in the Ridge and Valley Province and,
therefore, these watersheds may mask the influence of watersheds from other provinces. Another
explanation is that, the Ridge and Valley watersheds show the highest climate variability and
therefore the dryness index, a climate descriptor, is a logical FDI predictor.
A climate predictor that corresponds to the mean monthly precipitation for December
(DECPREC) was the best FDI or hydrologic response predictor for the Appalachian Plateau
watersheds (Table 11). Other important hydrologic response predictors included soil, topography,
and geology descriptors. Among the soil descriptors, saturated hydraulic conductivity (KSAT1 and
KSAT2), percent clay (CLAY1 and CLAY2), silt (SILT1 and SILT2), and rock fragments (percent
ROCK2) were correlated to the hydrologic responses of the Appalachian plateau watersheds. For
the Piedmont province watersheds, however, monthly rainfall was the best hydrologic response
predictor for the highest three indices (Ql, Q5, Q10). At medium flow conditions, soil moisture
storage of the upper soil profile layer (STOR1) was highly correlated to medium flow conditions
(Q20, Q30, Q40, and Q50). For low to very low flows (Q70, Q90, Q95), monthly precipitation had
the highest influence on the hydrologic response for both the Appalachian and Piedmont
watersheds. For these latter watersheds, soil descriptors were also important hydrologic response
predictors, followed by topography, vegetation, and geology.
61
-------
Table 11. Flow Duration Indices Equations for Physiographic Province Spatial Scale and One-year
Time Scale
Regression equation Coefficient of Determination
Appalachian Plateau
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
22.169-1.968(DECPREC)H).376(KSAT2)
-2.97+2.076(JUNPREC)+0.055(CLAY2)
7.121+0.321(DECPREC)-0.13(SILT2)
-0.078+0.447(DECPREC)+.010(MCHS)
-1 .39+0.562(DECPREC)+0.057(SMED)
1. 1392+1. 117(DECPREC)-1 1.894 (BFI)
-0.313+0.338(DECPREC)-0.046(KRATIO)
-0.903 +0.279(DECPREC)H).014(CLAY2)
0.288+0.161(DECPREC)-0.038(SAND2)
-0.265+0.145(DECPREC)-0.005(ROCK2)
R2
R2
R2
R2
R2
R2
R2
R2
R2
R2
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Piedmont
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
Ridge and Valley
-19.856-10.146 (APRPREC)+1.723(SILT1)-20.537(AET/P)
5.94-1.154(JULPREC)+0.155(STOR TOTAL)
8.414-2.041(JUNPREC)H).126(DRATIO)
-3.694+5.12(STOR1)-0.0010(SOLD2)
-0.372+3.44(STOR1)-0.125(CLAY1)
-2.621+3.318(STOR1)H).003 (MCHS)
-1.754+3.193(STOR1)-0.047(CLAY1)
3.789 -0.326 (AUGPREC)-0.192(STOR2)
0.997-0.077(AUGPREC)+0.002(CLAY2)
0.986-0.079(AUGPREC)+0.002(FRSD)
R2
R2
R2
R2
R2
R2
R2
R2
R2
R2
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
-0.310+8.495 (JULPREC)H).056(KSAT1)-0.003(SOLD)
5.126-7.847(AET/P)-0.010(SLENG)+4.313(DRI)+0.058(CLAY2)
5.482-6.038(AET/P)+0.872(DRI)-0.031(RRAT)
4.507-4.308(AET/P)+0.858(DRI)-5.427(AWC1)
5.905-6.971(AET/P)-0.013(FRSD)H).226(SEPPREC)
5.818-5.744(AET/P)-0.009(FRSD)-0.072(OCTPREC)
1.821-2.028(AET/P)-0.189(SOLD1)^2.424(BFI)
-1.312+2.727(BFI)-0.005(SOLD1)-0.002(STLEN)
-0.561+2.024(BFI)-0.001(TCHL)+0.033(FEBPREC)
-.514+1.659(BFI)-0.001(TCHL)+0.021(FEBPREC)
R2
R2
R2
R2
R2
R2
R2
R2
R2
R2
0.89
0.96
0.95
0.98
0.96
0.93
0.94
0.90
0.99
0.94
Symbol descriptions are given in Tables 2 and 7
Table 12 presents a summary of the factors that control the hydrologic responses of the Mid-
Atlantic watersheds. The results of this study showed that climate controlled the long-term high
and medium flow conditions of the Mid-Atlantic watersheds, while geology and soil descriptors
controlled the low flow hydrologic responses. For most of the flow conditions, in addition to
climate and geology, soil descriptors were also good predictors of the hydrologic responses,
followed by topography and vegetation.
62
-------
The single-year based regression equations showed that different landscape and climate
descriptors were important in the various physiographic provinces. Specifically, the one-year based
relationships showed that climate was the best predictor of hydrologic response for the Ridge and
Valley Province, whereas monthly rainfall was the best hydrologic response predictor for the
Appalachian Plateau and Piedmont watersheds, followed by various soil descriptors. The difference
between the regional scale and physiographic province scale regression equations were due to the
spatial scale and temporal scale influences on both inputs and hydrologic response. Note that
spatial scale is a major source of variability in watershed hydrologic response. As the spatial scale
was reduced from region to physiographic province, the variability in watershed hydrologic
response also reduced - thus improving the predictive capability of the regression equations.
63
-------
Table 12. Summary of Dominant Hydrologic Response Predictors at Different Spatial and
Temporal Scales
Flow condition
Multi-year time
Very High flow
High flow
High flow
High to medium flow
Medium flow
Medium flow
Medium to low flow
Low flow
Very low flow
Very low flow
Hydrologic Response Variables
scale and regional spatial scale
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
Single year time scale and physiographic province
Very High Flow
High flow
High flow
High to medium flow
Medium flow
Medium flow
Medium to low flow
Low flow
Very low flow
Very low flow
Very high flow
High flow
High flow
High to medium flow
Medium flow
Medium flow
Medium to low flow
Low flow
Very low flow
Very low flow
Very high flow
High flow
High flow
High to medium flow
Medium flow
Medium flow
Medium to low flow
Low flow
Very low flow
Very low flow
Appalachian Plateau
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
Ridge and Valley
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
Piedmont
Ql
Q5
Q10
Q20
Q30
Q40
Q50
Q70
Q90
Q95
Controlling Variable(s)
climate and soil
climate, topography, and soil
climate and topography
climate, soil, and vegetation
climate, soil, and vegetation
climate, soil, vegetation
climate and soil
geology and climate
geology and soil
geology and soil
spatial scale
climate and soil
climate and soil
climate and soil
climate and soil
climate and topography
climate and topography
climate and geology
climate and soil
climate and soil
climate and soil
climate and soil
climate and topography
climate and topography
climate and topography
climate and vegetation
climate and vegetation
climate and soil
geology and soil
geology and topography
geology and topography
climate and soil
climate and soil
climate and soil
soil and soil
soil and soil
soil and topography
soil and soil
climate geology and climate
climate and soil
climate and vegetation
64
-------
Part IV
Implications For Watershed Management
Streamflow measured at a watershed outlet is the result of complex watershed processes that
operate at the watershed scale. These watershed processes are normally influenced by landscape
and climate descriptors. Some of the dominant hydrologic processes are recharge, controlled
mainly by precipitation, moisture storage controlled mainly by soil, groundwater storage
characteristics, and deep groundwater and evapotranspiration losses.
Many investigators have reviewed the current state of the art in hydrologic modeling and made
recommendations for future research (Hornberger and Boyer, 1993; Beven, 1996). Hornberger and
Boyer (1993), in a review paper, stressed the need to develop procedures to measure model input
parameter values independently of the model output values. They pointed-out that a major
limitation in hydrololgic modeling is lack of methods to easily and reliably specify model input
parameters. In their review, they concluded that progress in hydrologic modeling is linked to
acquisition of new data and to new experimental work. Troch et al. (2003) acknowledged that key
input parameter identification and specification is a major constraint to the use of distributed
hydrologic modeling. They recognized the current mismatch between model complexity and the
level of data available to parameterize, initialize, and calibrate models.
In addition to parameter specification limitations, general lack of high quality data has also been
recognized as one of the constraints to the development of hydrologic models. Beven (1996) stated
that, "Essentially hydrological science suffers from very severe data constraints. What happens at
the point scale is reasonably well understood (at least well enough to understand that our
'physically-based' descriptions are inadequate due to the effects of both surface and subsurface
preferential flows), but data and ideas are lacking to know how to extend that knowledge to larger
scales". NAS (1993) and Hornberger (1993) concluded that hydrologic science is in greater need of
more and better experimentation than of more and better models (NAS, 1991; Hornberger, 1993).
Many of the currently used hydrologic models are based on relationships developed from data
and observations obtained from field plot or small watershed studies where soils, geology, and
vegetation are nearly homogeneous and where the complex interactions of watershed
characteristics and climate inputs are absent. The small field plots provide a more controlled
environment where researchers can understand processes and their interactions, identify
parameters, and develop models based on data and observations at the point scale. Hornberger and
Boyer (1993) also recognized data limitations and associated scaling problems that occur when
incorporating a relatively small scale heterogeneity into models applied to relatively larger scales.
As a result, models developed from data and observations obtained from small areas are often
applied to large watersheds without undergoing any upward scaling procedures. Because reliable
methods to identify the key parameters and estimate parameter values for large watersheds are not
65
-------
currently available, the inaccuracies associated with upward scaling will also remain unknown until
such methods are developed.
In traditional hydrology, streamflow measured at the watershed outlet is often used to
parameterize hydrologic models using various calibration procedures. Use of model calibration
procedures as a model parameter estimation tool assumes that hydrologic models have the correct
model structure, that processes are represented correctly in the model, and that all parameters that
influence the hydrologic processes are included in the model. Such assumptions falsely assume that
the current state of hydrologic modeling is satisfactory and, therefore, reduce the incentive to look
for new methods to identify model parameters. Nevertheless, the approach presented in Part III of
this report can be used to identify the key parameters for large watershed response. Hornberger and
Boyer (1993) also underscore the need for "empirical studies" to develop relationships among
measurable catchment characteristics and the estimated parameters of watershed models.
The use of empirical relationships between landscape-climate descriptors and hydrologic
response descriptors also provides insight into the dominant hydrologic processes, and the key
parameters representing those processes, in hydrologic models. An understanding of the dominant
hydrologic processes would enhance our ability to select the most appropriate models, or build new
hydrologic models when necessary.
7.1 Applications to Hydrologic Modeling
This work has presented an alternative approach to the identification of key watershed
hydrologic model parameters. The approach is based on the development of empirical relationships
between landscape-climate descriptors and hydrologic response descriptors using multi-variate
regression analysis. In Tables 10 and 11, we presented two sets of regression equations that can be
used to predict the 10 flow duration indices (e.g., Ql... Q95) that represent a wide range of
hydrologic responses. The regional flow duration indices equations presented in Table 10 are
suitable for development of long-term watershed hydrologic responses and are useful for water
resources management applications. The single-year based regression equations presented in Table
11 may be suitable for general watershed hydrologic modeling applications.
The regression equations in Table 10 and 11 are also suitable for predicting flow duration
indices for ungaged watersheds. The procedure to follow for this is to develop the regression
equations using landscape-climate descriptors and hydrologic response descriptors (Ql... Q95) for
gaged watersheds located near the ungaged watersheds of interest. Once the regression equations
are developed, users can then determine the specific landscape and climate descriptors for the
ungaged watersheds to be modeled, and then substitute these parameters into the regression
equations to predict the flow duration indices for the ungaged watersheds. After the flow duration
indices have been determined for the ungaged watersheds, the entire flow duration curve can be
constructed. The constructed flow duration curves can then be converted to streamflows by
multiplying the normalized FDCs by the drainage area of each watershed.
66
-------
7.2 Applications to Watershed Vulnerability Assessment
Watersheds are widely recognized as important resource management units. Today, in the
United States, many watersheds are continuously undergoing land use change, and watershed
management is becoming an essential tool for sustainable development and protection of natural
resources. Such land use change usually results in altered hydrologic response of the watershed. An
analysis of the hydrologic response of a watershed, therefore, can provide useful information about
the dominant watershed hydrologic controls. This study attempts to link landscape descriptors that
represent watershed characteristics and watershed climate descriptors to those hydrologic response
descriptors that represent the dominant watershed hydrologic regimes. Specific knowledge of the
likely hydrologic responses of a watershed would enable environmental planners and resource
managers to assess the risks that are associated with different management decisions.
Environmental planners and water resources managers clearly need tools to better manage natural
resources under changing conditions. A reliable method that predicts FDIs from watershed
characteristics under natural and changing conditions has, therefore, important implications to
water resource management.
7.2.1 Climate Change Applications
Fisher et al. (2001) presented a detailed analysis of the potential consequences of climate
variability and change in the Mid-Atlantic Region. They reported climate model projections
indicating drier conditions in the summer and winter months. These drier conditions might have
serious impacts on the water resources of the Mid-Atlantic Region. For example, reduced
streamflow and increased drought frequency and severity would affect both water quantity and
water quality. In addition, agricultural water use, particularly the need for irrigation may increase
along with the need for water supply management during prolonged droughts.
As the frequency and intensity of droughts increase, rivers and streams may become more
polluted due to reduced assimilative capacity for point source wastewater treatment plant
discharges. During extremely wet years, serious floods may occur and rivers and streams become
polluted by non-point source pollution, particularly - high levels of sediments, nutrients, and
pathogens from agricultural lands. In the Mid-Atlantic Region, climate change might also influence
snow accumulation and the timing of snowmelt runoff. Among the study watersheds, high-
elevation watersheds in the Appalachian, ridge-dominated Ridge and Valley, and Blue Ridge
provinces are more vulnerable to climate change because these watersheds have high snow
accumulations. On the other hand, the Piedmont watersheds are the least vulnerable to climate
change because streamflow there is less dependent on snow accumulation and more dependent on
soil and groundwater storage reservoirs. Environmental planners can use flow duration indicates
such as Ql and Q5 for flood risk assessment and Q90 and Q95 for drought management.
67
-------
7.2.2 Land Use Change Applications
Non-point source pollution (NFS) is mainly produced from land surfaces and is transported by
overland flow, interflow, and baseflow from groundwater. The hydrologic response of a watershed
has a strong influence on the production of sediments and the ultimate transport of sediments and
nutrients from land surfaces to streams, rivers, and lakes. Production of non-point source pollution
is land use related and the specific pollutants generated by a watershed often depend upon the type
of land use in the watershed (Mueller et. al. 1995). The characteristics of the soil, geology, land use
and land cover, and topography of a watershed can all be used to assess vulnerability to land use
change. To evaluate the landscape characteristics of a watershed and their interaction with land use
change, one can use the results of this study. For example, land use change, by urbanization, would
result in changes in the landscape descriptors such as percent of forest cover that would, in turn,
result in changes in predicted hydrologic response of the watershed using the FDIs developed using
the methods presented herein. Environmental planners and water resources managers can therefore
link changes in land use to changes in hydrologic response by monitoring the changes in predicted
flow duration indices, particularly the high flow indices (Ql and Q5) and low flow indices (Q90
and Q95).
7.2.3 Water Quality Applications
In general, there is a relationship between the quantity and quality of the water in a stream.
During drought periods, when streamflow falls below a certain threshold value such as the Q95
index, the ability of a stream to assimilate pollutants decreases with decrease in streamflow. Low
flow indices are often used for the determination of instream flows to meet ecological flow
requirement and to allocate water withdrawal levels among multiple water users within a
watershed. Low flow indices can be easily determined from observed streamflow data. However,
for ungaged watersheds, these low flow indices have to be estimated. In this study, we presented a
set of regression equations that can be used to estimate the Q70, Q90, and Q95 low flow indices
from landscape and climate descriptors.
7.2.4 Water Resources Applications
Some of the scale dependent issues that water resources managers often encounter include water
withdrawal allocations between the upstream and downstream segments of large watersheds,
assessment of the risk of downstream floods and its dependence on upstream land use change, and
the assessment of the sustainability and reliability of water supply systems during prolonged
droughts. Watershed scale influences the availability of water resources in a number of ways. For
example, large watersheds have higher streamflow rates than small watersheds. In addition, unlike
small watersheds, large watersheds have longer and deeper channels that intersect the water table.
As a result, large watersheds have higher sustained low flows and are, therefore, more suitable for
water supply development.
Small headwater watersheds located in the Appalachian Plateau, Ridge and Valley, and Blue
Ridge Provinces have steep slopes and shallow soils. These watersheds also have limited soil
moisture storage capacity and may not have sustained low flows. Because small headwater
68
-------
watersheds are often located in mountainous areas with erosion resistant geologic formations, small
watersheds usually have low yield and may not be suitable for water supply development unless
supplemented by groundwater or water supply reservoirs.
In general, watersheds that have high Q5 indices have high surface flow runoff components and
low subsurface runoff components. On the other hand, watersheds with high Q90 and Q95 indices
have sustained low flows and are, therefore, suitable sources for water supply. Among the study
watersheds, the Appalachian Plateau and Ridge and Valley watersheds had the lowest sustained
low flows and thus the highest vulnerability to drought whereas the Piedmont watersheds had the
highest sustained low flows and therefore the lowest vulnerability to drought. To determine a given
watershed's vulnerability, resource managers can use combinations of landscape descriptors such
as slope, soil depth, baseflow index, and hydrologic response descriptors, such as flow duration
indices Q5 andQ95.
Environmental planners and water resource managers need methods to estimate low flow
indices, i.e., Q70, Q90, and Q95 so that they can assess the likelihood of exceeding a particular
index, say Q95, as well as the risks associated with exceeding that particular index. One of the
attractive features of flow duration indices application is that managers can use sequences of
indices, say Q70 to Q95, as trigger points for different management decisions, such as water use
restrictions.
69
-------
8 Summary and Conclusions
The physiographic provinces and ecoregions of the Mid-Atlantic Region were used as a
selection framework for the study watersheds. We selected 25 watersheds from the different
physiographic provinces. This study links landscape descriptors of watershed characteristics and
descriptors of watershed climate to hydrologic response descriptors of the dominant hydrologic
regimes of Mid-Atlantic watersheds. We presented two approaches to hydrologic response
comparisons: a conceptual water balance based approach and an empirical flow duration curve or
statistical based approach.
The water balance approach was used to compare hydrologic responses at different time scales,
i.e., annual, monthly, daily, and hourly. Based on long-term water balance comparisons, watersheds
with high elevation and latitudinal influences had high streamflow and low evapotranspiration,
while watersheds with low elevation and latitudinal influence had low streamflow and high
evapotranspiration. At the monthly time scale comparisons, watersheds with high elevation and
latitudinal influences (i.e., the Appalachian Plateau and ridge-dominated Ridge and Valley
watersheds) had greater seasonal variability in hydrologic response than watersheds with low
elevation and latitudinal influences (i.e., the Piedmont and valley-dominated Ridge and Valley
watersheds). At daily time scale comparisons, all the water balance components had high temporal
variability driven mainly by the daily precipitation inputs. The water balance comparisons revealed
that mountainous watersheds located in the Appalachian Plateau, Ridge and Valley, and Blue
Ridge watersheds had higher streamflow, lower evapotranspiration, and lower soil moisture storage
capacity than the Piedmont watersheds.
The empirical approach represented the hydrologic responses of the study watersheds as flow
duration indices (Ql.. .Q95). Using flow duration indices, we ranked the study watersheds
according to increasing Q5, Q50, and Q95, and identified those watersheds that have a high risk of
flooding or risk of drought. We developed two sets of regression equations that had different
temporal and spatial scales. Among the potential hydrologic response predictors, dryness index, a
climate descriptor, was the best predictor of long-term hydrologic response or flow duration
indices. Soil descriptors that were good predictors of hydrologic response included soil texture,
bulk density, soil depth, and saturated hydraulic conductivity of the top two soil profile layers. The
characteristics of the lower of the top soil layers seemed to have more influence on hydrologic
response than the characteristics of the upper soil layer. Percent of deciduous forest, a land cover
descriptor, was not highly correlated to all hydrologic response descriptors, but had some influence
on medium hydrologic responses (Q30 and Q40). Among the geomorphologic descriptors, drainage
density and relief ratio had some influence on hydrologic responses. Another important hydrologic
response predictor was the baseflow index, a surrogate geologic descriptor.
Among the study watersheds, the Appalachian Plateau and the Ridge and Valley watersheds had
the lowest sustained low flows and, thus, the highest vulnerability to drought. The Piedmont
watersheds had the highest sustained low flows and, therefore, the lowest vulnerability to droughts.
The Piedmont watersheds had desirable hydrologic response characteristics characterized by
relatively high Q95 indices and relatively low Q5 indices. Based on our general hydrologic
response comparisons of the Mid-Atlantic watersheds, the Piedmont watersheds were the least
70
-------
vulnerable to drought, flooding, land use change, and climate change. To determine a watershed's
vulnerability, resource managers can use combinations of landscape descriptors, such as slope, soil
depth, and baseflow index, and hydrologic response descriptors such as flow duration indices, .i.e.,
Q5andQ95.
We conclude that the methods presented in this report have important implications to hydrologic
modeling particularly the prediction of streamflow for ungaged watersheds. More research is
needed to further develop both the water balance approach and the empirical approach towards the
development of a hydrologic model. There is a need for hydrologic models that have a strong
physical basis and yet have less parameters than those currently available. The resultant model
should be tested with observed data and compared with existing hydrologic models such as the
Stanford Watershed Model - HSPF.
71
-------
9 References
Anderson, J.R., E.E. Hardy., J.T., Roach, and R. E. Witmar. 1976. A land use and land cover
classification system for use with sensor data, U.S. Geological Survey, Professional Paper 964
28p. Reston, VA.
Berger, K. P., and D. Entekhabi. 2001. Basin hydrologic response relations to distributed physiographic
descriptors and climate. Journal of Hydrology 247: 169-182.
Beven, K. 1996. The limits of splitting: Hydrology. The Science of the Total Environment 183: 89-97.
Bonta, J.V., and B. Cleland. 2003. Incorporating natural variability, uncertainty, and risk into water
quality evaluations using flow duration curves. Journal of Water Resources Association. 39:1481-
1496.
Boyer, D.G. 1984. Estimation of daily temperature means using elevation and latitude in mountainous
terrain. Water Resources Bulletin Vol. 20(4):583-588.
Bras, R. L. 1990. Hydrology: An Introduction to Hydrologic Science, Addison-Wesley Publishing
Company, Reading Massachusetts.
Buttle J.M., and D. J. McDonald. 2002. Coupled vertical and lateral preferential flow on a forested
slope. Water Resources Research 38(5), Art. No. 1060
Cole, R.A.J., Johnston, H.T. and D.J. Robinson. 2003. The Use of flow duration curves as a data quality
tool. Hydrologic Sciences-Journal-des-sciences Hydrologiques 48:939-951.
Carey S.K., and M.K Woo. 2001. Slope runoff processes and flow generation in a sub-artic, sub-alpine
catchment. Journal of Hydrology, 253: 111-129.
Croker K. M., Young, A.R., Zaidman, M.D., and H.G. Rees. 2003. Flow duration curve estimation in
ephemeral catchments in Portugal. Hydrological Sciences Journal-Journal Des Sciences
Hydrologiques 48: 427-439.
Dingman, S.L. 1978. Drainage density and streamflow: A Closer Look. Water Resources Research 30:
1183-1187.
Dingman, S.L. 1981. Elevation - a major influence on the hydrology of New Hampshire and Vermont,
USA. Hydrological Sciences Bulletin-Bulletin des Sciences Hydrologiques, volume: 26 , number: 4
, page: 399-413
Dingman, S.L., and S.C. Lawlor. 1995. Estimating low-flow quantiles from drainage-basin
characteristics in New Hampshire and Vermont. Water Resources Bulletin 31:243-256.
Eagleson, P.S. 1978. Climate, soil, and vegetation, 6. Dynamics of the annual water balance. Water
Resources Research. 14:749-764.
England, C,. and H. N. Holtan. 1968. Geomorphic grouping of soils in watershed engineering, Journal
of Hydrology 6: 13-21.
Famiglietti, J.S. and E.F. Wood. 1995. Effects of spatial scale on areally averaged evapotranspiration
Water Resources Research 31:699-712.
Farvolden, R.N. 1963. Geologic controls of groundwater storage and baseflow. Journal of Hydrology 1:
219-249.
Fisher A, Barren E, Yarnal B, Knight C.G, and J.Shortle. 1999. Climate change impacts in the Mid-
Atlantic Region - a workshop report, September 9-11, 1997. The Pennsylvania State University,
University Park. Preparing for a Changing Climate: The potential consequences of climate
variability and change - Mid-Atlantic Foundations. January 2001.
Frederer, C.A. and D. Lash. 1978a. BROOK: A hydrologic simulation tool for eastern forests. Durham,
NH: University of New Hampshire Water Resources Research Center Research Report 19.
72
-------
Fujieda, M., Kudoh, T., Decicco, V., and J. L. De Calvarich. 1997. Hydrological processes at two
subtropical forest catchments: The Sierra do Mar, Sao Paulo, Brazil. Journal of Hydrology 196:
24-46.
Gustard A., A. Bullock, and J.M. Dixon. 1992. Low flow estimation in the United Kingdom. Institute
of Hydrology Report 108. Wallingford, UK,
Guth, P.L.1989. MICRODEM+: comparing digital elevation models on a personal computer:
Proceedings of Defense Mapping Agency Systems Center Symposium 89, p.389-394
Harlin, J.M. 1984. Watershed morphometry and time to hydrograph peak. Journal of Hydrology Vol.
67, No. 1-4, p 141-154.
Hartley S., and S.L. Dingman. 1993. Effects of climatic variability on winter-spring runoff in New-
England River Basins. Physical Geography 14: 379-393.
Hawley, M. E., Jackson T.J., and R.H. McCuen. 1983. Surface soil moisture variation on small
agricultural watersheds. Journal of Hydrology 62: 179-200.
Hornberger, G.M. 1993. Hydrologic Science: Keeping pace with changing values and perceptions, in
Sustaining our water resources pp. 43-58, National Academy of Sciences Press.
Jakeman, A.J., and G.M. Hornberger. 1993. How much complexity is warranted in rainfall-runoff?
Water Resources Research 29: 2637-2649.
Kuichling, E. 1889. The relationship between the rainfall and discharge of sewers in populous districts.
Trans. ASCE 20.
Lacey G.C. and R.B. Grayson. 1998. Relating baseflow to catchment properties in south-eastern
Australia. Journal of Hydrology 204: 231-250.
Langbein, W.B., and others. 1947. Topographic characterization of drainage basins. U.S. Geological
Survey paper no. 968-c. Washington, D.C.
Meentemeyer, V., and E.O. Box. 1987. Scale effects in landscape studies. In Landscape Heterogeneity
and Disturbance, M.G. Turner, ed. Springier-Verlag, 15-34.
Montgomery, D. R., and W.E. Dietrich. 2002. Runoff generation in steep, soil-mantled landscape.
Water Resources Research 38:1168
Mueller, O.K., Hamilton, P.A., Helsel, D.R., Hitt, K.J., and B.C. Ruddy. 1995. Nutrients in
groundwater and surface water of the United States- An analysis of data through 1992: U.S.
Geological Survey, Water Resources Investigations Report 95 4031, 74p.
National Academy of Sciences (NAS). 1993. Opportunities in the hydrologic sciences, 348 pp. National
Academy Press, Washington, D.C.
National Soil Survey Center. 1991. State soil geographic (STATSGO) database. Technical Report 1492.
USDA. Natural Resources Conservation Survey
Patric, J.H. 1972. Deforestation effects on soil moisture, streamflow, and water balance in central
Appalachians. Research paper NE-259 :1-12
Searcy, J.K. 1959. Flow-duration curves, manual of hydrology-part 2, Low-flow techniques: U.S.
Geological Survey Water-Supply Paper 1542-A, 33 p.
Sefton, C.E.M., and S.M. Howarth. 1998. Relationships between dynamic response characteristics and
physical descriptors of catchments in England and Wales. Journal of Hydrology 211: 1-6.
Shruben, P.G., Arndt, R.E., and W.J. Bawiec. 1998. Geology of the conterminous United States at 1,
250,000 scale, a digital representation of the 1974 P.B. King and H.M. Beikman Map: U.S.
Geological Survey data series 001. Release 2.
Singh, V.P. 1997. Effect of spatial and temporal variability in rainfall and watershed characteristics on
stream flow hydrograph. Hydrologic Processes. Vol. 11, no. 12, pp. 1649-1669.
73
-------
Slack, J.R., and J.M. Landwehr. 1992. Hydro-Climatic Data Network (HCDN): A U.S. Geological
Survey streamflow data set for the United States for the study of climate variations, 1874-1988:
U.S. Geological Survey Open-File Report 91-129, 193 p.
Smakhtin, V.U. 2001. Low flow hydrology. Journal of Hydrology 240: 147-186.
Soil Conservation Service. 1972. National Engineering Handbook, Section 4. Hydrology, U.S.
Department of Agriculture, Washington, D.C.
Strahler, A. N. 1952. Hypsometric (area-altitude) analysis of erosional topography. Geological Society
of America Bulletin 63, 117
Swank, W.T., Norbert, E.G., and J.D. Helvey. 1972. Interception loss in loblolly pine stands of the
South Carolina Piedmont. Journal of Soil and Water Conservation 27:160-164.
Thornthwaite, C.W. and J.W. Mather. 1955. The water balance . Publication Climatol. Drexel Inst.
Technol. Vol. No.l pp. 1-104.
Troch, P.A., De Troch, P.P., and W. Brutsaert. 1993. Effective water table depth to describe initial
conditions prior to storm rainfall in humid regions. Water Resources Research 29: 427-434,
Troch, P. A., Paniconi, C., and D. McLaughlin. 2003. Catchment-scale hydrological modeling and data
assimilation. Advances in Water Resources 26: 131-135.
Troch, P.A., Detroch, P.P., M. Mancini, and E.F. Wood. 1995. Stream network morphology and storm
response in humid catchments. Hydrologic Processes 9: 575-587.
United States Geological Survey. 1992-1997. Groundwater Atlas of the United States, Reston, VA.
USEPA. 1998. Users Manual: Better Assessment Science Integrating Point and Nonpoint Sources:
BASINS Version 2.0. EPA-823-B-98-006. U.S. Environmental Protection Agency. Washington,
DC.
Vogel R.M., and N.M. Fennessey. 1995. Flow duration curves .2. A review of applications in water-
resources planning. Water Resources Bulletin 31: 1029-1039.
Vorosmarty, C.V., Federer, C.A., and A.L. Schloss. 1998. Potential Evaporation functions compared on
US watersheds: Possible implications for global-scale water balance and terrestrial ecosystem
modeling. Journal of Hydrology. 207:147-169.
Wilson, G.V., and R.J. Luxmore. 1988. Infiltration, macroporosity, and mesoporosity distributions on
two forested watersheds. Soil Science Society of America 52: 329-335
Wittenberg, H., and M. Sivapalan. 1999. Watershed groundwater balance estimation using streamflow
recession analysis and baseflow separation. Journal of Hydrology 219: 20-33
Woods, A.J., J.O. Omernik, and D. D. Brown. 1999. Level III and IV ecoregions of Delaware,
Maryland, Pennsylvania, Virginia, and West Virginia. National Health and Environmental Effects
Research Lab, US EPA, Corvallis, OR.
Yu, P.S., T.C. Yang, and Y.C. Wang. 2002. Uncertainty analysis of regional flow duration curves.
Journal of Water Resources Planning and Management 128:424-430.
Zecharias Y.B., and W. Brutsaert. 1988. Recession characteristics of groundwater outflow and base-
flow from mountainous watersheds. Water Resources Research 24: 1651-1658
74
------- |