Assessment of Soil Erosion Risk within the Zumbro River by pharmphresh36

VIEWS: 109 PAGES: 22

									Assessment of Soil Erosion Risk within a Subwatershed using GIS and RUSLE with
a Comparative Analysis of the use of STATSGO and SSURGO Soil Databases

Todd Breiby
Department of Resource Analysis, Saint Mary’s University of Minnesota, Winona, MN

Keywords: GIS, RUSLE, Raster Analysis, Land Use, Soil Erosion Risk, STATSGO,
SSURGO, Soil Conservation Practices, Watershed Analysis, Nonparametric Regression


Land degradation and subsequent soil erosion and sedimentation play a significant role in
impairing water resources within subwatersheds, watersheds and basins. Using
conventional methods to assess soil erosion risk is expensive and time consuming.
Geographic Information Systems (GIS), coupled with the use of an empirical model to
assess risk, can identify and assess soil erosion potential and estimate the value of soil
loss. The objectives of this project are to: 1) assess soil erosion risk within a Zumbro
River subwatershed in southeastern Minnesota using GIS and the Revised Universal Soil
Loss Equation (RUSLE), 2) comparatively analyze the use and scaling effect of
STATSGO and SSURGO soil databases with RUSLE and 3) assess the sensitivity and
scaling effect of estimated soil loss to model variables. Soil, land use, digital elevation,
flow accumulation and climatic data are used to generate RUSLE variables. This
empirical soil erosion model estimates soil loss values by tons/acre/year and assesses the
spatial distribution of soil erosion risk within the entire subwatershed. By comparing soil
loss estimates, spatial distribution and variable sensitivity from the RUSLE model using
STATSGO soil data and SSURGO soil data, it is possible to compare the responses of
both soil databases. Nonparametric regression shows the level of relatedness between
STATSGO and SSURGO RUSLE model outputs at the subwatershed scale. Correlation
coefficients (R2) of 0.914, 0.928, and 0.922 for 10, 30, and 50 meter resolutions
respectively highlight the significance of the relationship. At high to very high levels of
estimated soil erosion loss the relatedness between STATSGO and SSURGO-based
RUSLE model outputs lessened. Of the LS, K, and C model variables investigated, the C
variable (cover management) exhibited a greater level of relatedness to RUSLE model
outputs than the other variables at 10, 30 and 50 meter resolutions but not enough to be

Introduction                                          detachment and movement of soil
                                                      particles. Soil erosion is one of the major
Agricultural land in the U.S. is losing               non-point pollution sources in many
invaluable soil faster than it can be                 watersheds (Wang and Cui, 2005).
replenished because of erosion and the                        Soil loss from agricultural lands

Breiby, Todd. 2006. Assessment of Soil Erosion Risk within a Subwatershed using GIS and RUSLE with a
Comparative Analysis of the use of STATSGO and SSURGO Soil Databases. Volume 8, Papers in
Resource Analysis. 22pp. Saint Mary’s University of Minnesota Central Services Press. Winona, MN.
Retrieved (date) from

is estimated to be in the billions of tons          State Soil Geographic (STATSGO), Soil
per year. It is also estimated that more            Survey Geographic (SSURGO),
than 6 million metric tons of nitrogen              National Resources Inventory (NRI),
fertilizer and over 100,000 metric tons of          Food and Agriculture Organization Soil
pesticides are applied to crop fields in            Map of the United Nations/World Soil
the Midwest alone (Porter et al, 2001).             Classification (FAO), and other local
Soil erosion, sedimentation, and the                and state soil databases developed by
subsequent conveyance of fertilizers,               local governments and state natural
pesticides, and herbicides play a                   resource agencies.
significant role in impairing water                         The two most commonly
resources within subwatersheds and                  available soil databases for soil erosion
watersheds.                                         risk modeling and watershed assessment
         Yet conventional methods to                are STATSGO and SSURGO (Gowda
survey land and assess soil erosion are             and Mulla, 2005). Both were developed
costly and time consuming. Mapping                  by the U.S. Department of Agriculture
soil erosion using GIS can easily identify          (USDA) Natural Resource Conservation
areas that are at potential risk of                 Service (NRCS). For both of these
extensive soil erosion and provide                  databases, maps are produced from
information on the estimated value of               different intensities and mapping scales,
soil loss at various locations (Yusof and           and each database is linked to attribute
Baban, 1999).                                       data for each soil and map unit.
         By effectively predicting soil                     The STATSGO soil database was
erosion, it is possible to: develop sound           developed primarily for regional,
land-use practices as they relate to earth          multistate, state, basin, and multicounty
disturbing activities, estimate the                 resource planning. STATSGO data are
efficiency of best management practices             not detailed enough for planning at the
required to prevent excess sediment                 county scale or smaller. STATSGO soil
loading, and identify target areas for              maps are compiled by generalizing more
conservation funds or research (Hickey              detailed SSURGO soil databases and
et al, 2005).                                       utilizing generalized county soil maps
         Several soil erosion and non-              (USDA NRCS, 1994).
point source pollution models have been                     The SSURGO soil database in
developed, modified, and combined with              contrast provides a more detailed level
GIS software to take advantage of these             of soils interpretation and resolution and
new capabilities and provide regional               was developed primarily for much
soil erosion and non-point water quality            smaller scale resource planning activities
assessments during the past decade                  including those at the county, township,
(Wilson, 2003). Among these models is               farm, ranch, and land parcel level. This
the Revised Universal Soil Loss                     soil database is an excellent source for
Equation (RUSLE).                                   determining erodible areas, assisting in
         An inherent variable in the                developing appropriate erosion control
RUSLE model, which will be described                practices and developing land use
in further detail later in this paper, is the       assessments (USDA NRCS, 1994).
use of soils data to generate erosion risk                  For those researchers and
estimates. Several soil databases are               resource managers that decide to utilize
available for use and they include the              the USDA NRCS soil databases for

modeling and assessment, availability of                      Researchers and resource
the databases may drive which databases               managers who often use USDA NRCS
are used. Figure 1, a map from the                    soil databases have 1) begun using
USDA NRCS, depicts the status of                      SSURGO when available, 2) combine
SSURGO soil database development by                   SSURGO and STATSGO when
counties in Minnesota. As evident in                  conducting assessments at the regional
Figure 1, one can see SSURGO is only                  scale and SSURGO is not available over
available, as of August 2006, for 68 of               the entire project area, or 3) continue
Minnesota’s 88 counties. Minnesota’s                  utilizing STATSGO even when
current status is similar to what is found            SSURGO is available as an option.
throughout the eastern and Midwestern                         Researchers and resource
U.S. with many counties still in the                  managers, when considering which soil
planning, development or review phases                databases to utilize, must consider the
for SSURGO soil database                              benefits and drawbacks for each
establishment. The status of states and               database based on project objectives and
counties in the western U.S. is more                  scale of research. It is also important to
incomplete with many areas that have                  understand the effects of spatial scale at
not begun the planning and development                which soil databases are developed prior
phases of SSURGO establishment.                       to choosing a database with which to
National SSURGO coverage is planned                   work (Gowda and Mulla, 2005).
for completion by 2008.                                       If choice of database is not an
                                                      option, or different databases must be
                                                      stitched together (Hickey et al, 2005),
                                                      what kind of impact, if any, will result
                                                      based on model outputs? With respect to
                                                      comparing soil attributes of STATSGO
                                                      and NRI databases, it was demonstrated
                                                      that there was disagreement for selected
                                                      soil properties. This result implies risk
                                                      assessment and ecosystem modeling
                                                      outputs can be influenced by the
                                                      selection of data sources (Ding et al,
                                                              Additionally, differences in
                                                      runoff and soil properties can be
                                                      attributed to the differences in the spatial
                                                      resolution of the data sets (Levick et al,
                                                      2004). It was demonstrated that when
                                                      evaluating alternative agricultural
                                                      management practices that a STATSGO-
                                                      based model predicted annual nitrate
                                                      losses consistently higher than that for
                                                      SSURGO data and that a SSURGO-
Figure 1. Minnesota map depicting the status of       based model predicted annual
SSURGO soil database development by county
as of August 2006. Source: USDA NRCS.                 phosphorous losses consistently higher
                                                      than that for STATSGO data (Gowda

and Mulla, 2005). On the other hand, it           locations to identify the role of
was demonstrated that the integration of          individual variables in contributing to
a FAO soil database into a watershed              the observed erosion potential value (Shi
hydrologic model produced results                 et al, 2002).
comparable to the results produced when                    RUSLE computes average
calculated using both STATSGO and                 annual erosion from cover slopes as
SSURGO soils data (Levick et al, 2004).           (Renard et al, 1997):
        Because the determination of
potential soil erosion risk can differ            A=R*K*L*S*C*P
depending upon what data sources are
used, it is difficult for resource managers       Where:
to identify critical areas and apply
appropriate management techniques.                A = computed average annual soil loss
Consequently, a comparison of the most            in tons/acre/year
commonly available soil databases is              R = rainfall-runoff erosivity factor
needed. This project seeks to compare             K = soil erodibility factor
the STATSGO and SSURGO soil                       L = slope length factor
databases to determine their relatedness.         S = slope steepness factor
                                                  C = cover management factor
Methods                                           P = conservation practice factor

Empirical Model                                           In examining the RUSLE
                                                  variables the equation can be broken
The Universal Soil Loss Equation                  down into two parts: 1) environmental
(USLE), developed by Wischmeier and               variables and 2) management variables
Smith in 1978, is the most frequently             (Hickey et al, 2005). The environmental
used empirical soil erosion model                 variables include the R, L, S and K
worldwide and was later modified into a           factors. These variables remain
revised Universal Soil Loss Equation              relatively constant over time. The
model by including improved means of              management variables include the C and
computing soil erosion factors (Shi et al.,       P factors and may change over the
2002). These improved means for                   course of a year or less.
computing soil erosion factors generally
fit into two categories: incorporation of         Model Limitations
new/better data and consideration of
selected erosion processes. The inclusion         There are several limitations to the
of these factors into RUSLE has “the              RUSLE model and they appear to be in
potential for broader prediction                  three main categories: 1) the research
improvements” (Sonneveld and Nearing,             location in which RUSLE is applied, 2)
2003; Jones et al, 1996).                         limitations inherent in the mathematical
         The RUSLE model can predict              calculations and 3) limitations in scale.
erosion potential on a cell-by-cell basis,
which is effective when attempting to             Research Location Limitations
identify the spatial pattern of soil loss
present within a large region. GIS can            RUSLE was designed primarily for
then be used to isolate and query these           agricultural regions. Soil-erosion

potential as identified in non-agricultural       pattern of erosion, or vulnerability,
regions may be inconsistent (Hickey et            should be examined (Hickey et al.,
al., 2005). Further, RUSLE has had                2005).
limited application outside of the U.S. In
one study using data that was collected           Project Site
on natural runoff plots located primarily
in the eastern half of the U.S., the              The Zumbro River Watershed, (Figure
RUSLE model did not outperform the                2), is approximately 1,513 mi.2 in size
USLE in its prediction accuracy                   and is one of 12 watersheds that make up
(Sonneveld and Nearing, 2003).                    the Lower Mississippi River Basin in
                                                  southeastern Minnesota. The Zumbro
Limitations in Mathematical                       River Watershed lies in Dodge,
Calculations                                      Goodhue, Olmsted, Rice, Steele and
                                                  Wabasha Counties. The Watershed itself
The environmental variables used in               is made up of 91 subwatersheds that
RUSLE are relatively constant over the            range in size from 5.45 mi.2 to 51.21
timescale of tens of years (at a                  mi.2.
minimum), while the management
variables may change over the course of
a year or less. Consequently, it is
difficult to obtain current and accurate
management variable coverage (Hickey
et al., 2005).
         Several algorithms are required
when processing data for input into
RUSLE. Each of those algorithms may
accentuate existing errors in data.
Because RUSLE requires six input data
layers to be multiplied together, the
errors inherent in each layer are similarly
multiplied, contributing to an even
greater error in the derived soil loss            Figure 2. The Zumbro River Watershed and its
values (Shi et al., 2002) Consequently,           associated 91 subwatersheds in southeastern
results of calculations should only be            Minnesota.
used in a comparative sense, and not to
calculate sediment loads unless further                    Many of the 91 subwatersheds
validation or correction of the data              that make up the Zumbro River
occurs (Hickey et al., 2005).                     Watershed have not been named. The
                                                  subwatershed chosen as the project site
Limitations in Scale                              for this research has not been named and
                                                  for the purposes of this project and paper
The erosion processes which are                   is considered the Zumbro River
considered by RUSLE are often driven              subwatershed.
by relatively small features. Therefore,                   The Zumbro River subwatershed,
any output should be treated as                   (Figure 3), is approximately 17.80 mi.2
qualitative, not quantitative, and the            (11,391.57 ac., 4610.11 ha.) in size. The

subwatershed lies entirely in Olmsted            provided the soil data needed to generate
County, north of Rochester, Minnesota.           the K factor (soil erodibility).
        The Zumbro River subwatershed                    Digital elevation models (DEM)
was chosen as the project site for this          with a 30 meter resolution, established
research due to: 1) the availability of          by the U.S. Geological Survey (USGS),
spatial and tabular STATSGO and                  provided the elevation data needed to
SSURGO soils data for Olmsted County,            generate the L and S factors (slope
2) diverse land cover that includes              length and slope steepness).
natural lowland and wetland, conifer and                 Land cover digital data for the
deciduous forests, pasture, hay, and row         subwatershed was obtained from the
crops and 3) topographic variation in            Minnesota Land Cover Classification
elevation and slope.                             System (MLCCS), established by the
                                                 Minnesota Department of Natural
                                                 Resources (MDNR). This data, along
                                                 with land cover and agriculture tabular
                                                 data obtained from the USDA NRCS,
                                                 was used to generate the C factor (cover
                                                         Rainfall/precipitation data was
                                                 obtained from the USDA Agriculture
                                                 Handbook Number 537 (Wischmeier
                                                 and Smith, 1978). This information
                                                 source provided the R factor (rainfall-
                                                 runoff erosivity) for the Zumbro River
                                                 subwatershed project site.
Figure 3. The Zumbro River subwatershed                  Data for the P factor
project site.
                                                 (conservation practices) within the
                                                 subwatershed were not available.
                                                 Potential sources for the data were
                                                 investigated and included county land
Data needed for this research project
                                                 conservation departments, Farm Service
                                                 Agency (FSA), MDNR, and USDA
                                                 NRCS. An established and routinely
    •   Soil data                                used protocol for addressing this lack of
    •   Elevation data                           data will be utilized to control P factor in
    •   Land cover data                          RUSLE model replications.
    •   Rainfall/precipitation data                      Data for the state, county,
    •   Conservation practices data              watershed, and subwatershed boundaries
    •   State, county, watershed and             were obtained from the MDNR and the
        subwatershed boundary data               U.S. Fish and Wildlife Service
                                                 (USFWS). The digital data provided the
        The spatial and tabular State Soil       necessary information to develop locator
Geographic (STATSGO) database and                and project site maps and additionally
the spatial and tabular Soil Survey              provided the boundary framework to
Geographic (SSURGO) database,                    develop, clip, and analyze subwatershed
established by the USDA NRCS,                    data.

Analysis                                         polygon shapefiles, were added to the
                                                 ArcMap environment. Prior to adding
To address the project’s objectives, a           these Minnesota shapefiles to ArcMap,
GIS was developed to generate two                ArcCatalog was used to reproject any
separate RUSLE models, each using                data layers not in NAD83, UTM, Zone
either the STATSGO or SSURGO soil                15N to this correct projection. These
databases and each model being                   shapefiles included the two most
calculated at a 10, 30, and 50 meter (cell       important data layers, the polygon
size) spatial resolution to investigate          shapefile for the Zumbro River
scaling effects. Environmental Systems           Watershed and the shapefile for the
Research Institute (ESRI) software was           Zumbro River subwatershed project site.
used for these purposes.                                 The spatial and tabular data
        ArcCatalog was used to manage,           layers for the Minnesota DEM and the
manipulate, reproject, create, and delete        MLCCS land cover were added to the
the data layers for this project. ArcMap         ArcMap environment. Additionally, the
was used to view, develop, edit, query,          data layers for the STATSGO and
and analyze the project’s data layers            SSURGO soil databases were added to
while ArcToolbox was used in the                 each respective model. Again, prior to
development of the spatial data because          adding these data layers, the Project
of its geoprocessing functionality.              feature from Data Management Tools
        The projection used for this study       was used to reproject any layers not
was NAD83, UTM, Zone 15N. To                     projected in NAD83, UTM, Zone 15N.
ensure all spatial data obtained from data               Spatial data for the SSURGO soil
sources was in the correct projection,           database was obtained as a polygon
and to better understand the data overall,       shapefile. The spatial data for the
metadata from each data source was               STATSGO soil database, on the other
examined carefully.                              hand, is an older soil database system
        In the ArcMap environment,               and was obtained as an interchange file
three data frames were created to better         (.e00). Import71, a stand alone utility
manage each RUSLE model and their                from ArcView GIS that converts an
subsequent outputs for analysis. The first       ArcInfo interchange file to a more
data frame housed the data layers and            current coverage, was used to convert
RUSLE model calculations that used the           the STATSGO .e00 file to a coverage.
STATSGO soil database. The second                The STATSGO coverage was then
data frame housed the same data layers           converted to a shapefile and added to the
and RUSLE model calculations that used           project.
the SSURGO soil database. The third                      The DEM, MLCCS, STATSGO,
data frame housed the RUSLE outputs              and SSURGO data layers for the state of
for each model including the necessary           Minnesota needed to be clipped to the
data for further analysis.                       Zumbro River subwatershed project site.
                                                 To accomplish this, the Clip feature tool
RUSLE Spatial Data                               was used to clip the DEM, MLCCS,
                                                 STATSGO, and SSURGO data layers to
For both RUSLE models, Minnesota                 the subwatershed polygon.
state, county, watershed, and                            Data layers for the R factor and P
subwatershed spatial data, in the form of        factor also needed to be created. Each

layer was created as a shapefile and
clipped to the subwatershed project site
and added to the ArcMap environment
for each RUSLE model. The shapefiles
were created so that each shapefile
represented a single polygon that would
be represented by a single value.
        To further prepare the spatial
data for modeling, the data layers would
need to be converted from features to
raster. The MLCCS, STATSGO,
SSURGO, R factor, and P factor vector                  Figure 6. SSURGO grid (10 meter cell size)
shapefiles were converted to raster                    representing K factor (soil erodibility) values.
format. This was completed using the
Convert Features to Raster function
within the Spatial Analyst extension.
The output grids, as seen in Figures 4-8
(10 meter spatial resolution), were
generated at cell sizes of 10, 30, and 50

                                                       Figure 7. R factor grid (10 meter cell size)
                                                       representing the rainfall-runoff erosivity value.

Figure 4. MLCCS grid (10 meter cell size)
representing C factor (cover management)

                                                       Figure 8. P factor grid (10 meter cell size)
                                                       representing the conservation practice value.

                                                               With all the subwatershed layers
                                                       in raster format, the last step was to
                                                       generate a slope grid and a flow
                                                       accumulation grid from the DEM. To
Figure 5. STATSGO grid (10 meter cell size)            create the slope grid, the Slope function
representing K factor (soil erodibility) values.       feature was used. The output slope grid,

as seen in Figure 9 (10 meter spatial                          With this last step, the raster
resolution), was generated at cell sizes of            grids for the subwatershed, MLCCS,
10, 30, and 50 meters. The flow                        STATSGO, SSURGO, R factor, P
accumulation grid was constructed using                factor, slope, and flow accumulation
the ArcGIS extension, ArcHydro Tools,                  were ready to be included in both
which was downloaded from the                          RUSLE models at each of the three
University of Texas at Austin Center for               scales.
Research in Water Resources website.
                                                       RUSLE Attribute Data

                                                       Prior to converting the MLCCS,
                                                       STATSGO, SSURGO, R factor, and P
                                                       factor vector data layers into raster grids,
                                                       a new field needed to be added to each
                                                       layer’s attribute table. The new field
                                                       added to the MLCCS attribute table
                                                       housed the C factor values for each land
                                                       cover polygon in the subwatershed.
                                                               Table 1 highlights land cover
Figure 9. Slope grid (10 meter cell size).             types and their associated C factor
                                                       values (soil erodibility based on cover
        The Fill Sink feature under
Terrain Preprocessing was used to fill in              Table 1. A sample of the 83 MLCCS land cover
sinks within the DEM and create an                     classifications and associated C factor values in
                                                       the Zumbro River subwatershed.
output grid. This output grid was then
used to determine flow direction using
the Flow Direction feature. The flow
direction output grid was then used to
determine flow accumulation using the
Flow Accumulation feature. The output
flow accumulation grid, as seen in
Figure 10 (10 meter spatial resolution),
was generated at cell sizes of 10, 30, and
50 meters.

Figure 10. Flow accumulation grid (10 meter cell

management) for a sampling of the 83                1 was added to the attribute table’s new
cover types found in the Zumbro River               field. As mentioned, this is a technique
subwatershed using MLCCS. The C                     used by researchers and resource
factor is a numerical value from 0 to 1 in          managers that lack conservation practice
which cover management values closer                information for their models and simply
to 0 are less prone to soil erodibility. The        remove this variable from having any
C factor values were derived from a                 impact on the model.
combination of data gathered from the
USDA NRCS Minnesota office and                      Table 2. USDA NRCS STATSGO soil units and
several other soil erosion studies                  associated K factor values in the Zumbro River
conducted in comparable climates and
environments (i.e. Minnesota,
Wisconsin, and New York).
        The new field added to both the
STATSGO and SSURGO attribute
tables housed the K factor values for
each soil unit in the subwatershed. Table
2 highlights STATSGO soil units and
their associated K factor values (soil
erodibility). For the Zumbro River
subwatershed there are 9 soil polygons.
Table 3 highlights SSURGO soil units
and their associated K factor values. For           Table 3. A sample of the 1,396 USDA NRCS
the Zumbro River subwatershed there                 SSURGO soil units and associated K factor
are 1,396 soil polygons.                            values in the Zumbro River subwatershed.
        The K factor is a numerical value
from 0 to 1 in which soil erodibility
values closer to 0 are less prone to soil
erosion. The K factor values, including a
diversity of other soil property
characteristics, are found in separate
tabular data that were added to the
ArcMap environment, queried and
joined to the spatial data layers attribute
tables based on common fields.
        Lastly, a new field was added to
each of the R and P factor attribute
tables in ArcMap. To reiterate, each
spatial data layer consists of a single
polygon that fits the entire extent of the
subwatershed. The R factor (rainfall-               RUSLE Modeling
runoff erosivity) value for the entire
Zumbro River watershed is 140. Due to               With the C, K, R, and P factor values
the lack of availability of conservation            now added to the attribute tables and the
practice (P factor) information for the             MLCCS, STATSGO, SSURGO, R
Zumbro River subwatershed, a value of               factor, P factor, slope, and flow

accumulation layers converted from
features to raster, the stage is now set to
begin calculating both RUSLE models at
each designated scale.
         The remaining factor of LS
(slope length and slope steepness) was
calculated using the slope and flow
accumulation grids generated earlier.
The longer the slope length the higher
amount of cumulative runoff and the
steeper the slope the higher the runoff
velocity which contributes to erosion.             Figure 11. Slope length and steepness grid (10
         The original equation to calculate        meter cell size).
the LS factor was an empirical equation
published in the USDA Agriculture                        The first iteration of the RUSLE
Handbook No. 537 (Wischmeier and                   model was:
Smith, 1978). The equation has
undergone some minor changes                       A = R * K * LS * C * P
including the equation published by
Moore and Burch in 1986.                           Where:
         The LS empirical equation used
for this project is:                               A = computed average annual soil loss
                                                   in tons/acre/year
LS = (Flow Accumulation grid * cell                R = rainfall-runoff erosivity grid
       size / 22.13)0.4 * (Sin(Slope grid          K = soil erodibility grid (STASGO)
       * 0.01745) / 0.0896)1.4 * 1.4               LS = slope length and steepness grid
                                                   C = cover management grid (MLCCS)
        The Raster Calculator in the               P = conservation practice grid
Spatial Analyst extension of ArcMap
was used to calculate the LS grid. The                  The second iteration of the
Raster Calculator expression of the                RUSLE model was:
equation above was:
                                                   A = R * K * LS * C * P
LS = Pow([Flow Accumulation grid] *
      10 / 22.13, 0.4) * Pow(Sin[Slope             Where:
      grid] * 0.01745) / 0.0896, 1.4) *
      1.4                                          A = computed average annual soil loss
                                                   in tons/acre/year
         The output LS grid, as seen in            R = rainfall-runoff erosivity grid
Figure 11 (10 meter spatial resolution),           K = soil erodibility grid (SSURGO)
was generated at cell sizes of 10, 30, and         LS = slope length and steepness grid
50 meters.                                         C = cover management grid (MLCCS)
         The Raster Calculator was again           P = conservation practice grid
used to calculate both RUSLE model
grids to determine potential soil erosion                When comparing models, both
risk in the Zumbro River subwatershed.             model variables except for the K factor
                                                   (STATSGO and SSURGO) are identical,

thereby controlling model calculations             River subwatershed. This technique was
and allowing for a comparative analysis            employed so that every cell in the
of the STATSGO and SSURGO soil                     subwatershed grid had an equal chance
databases.                                         of being selected. Simple random
                                                   sampling is probably the best method to
Comparative Analysis                               ensure a bias-free sample for self
                                                   contained units when data is available
With both RUSLE models calculated                  for all grid cells. It has several
using STATSGO and SSURGO soil                      drawbacks, including high variance,
databases at 10, 30, and 50 meter                  sampled data not spatially balanced, and
resolutions (cell sizes), the resulting            the potential for an increased probability
output grids are ready to be sampled for           that as the number of sampled data
comparison. Once sampled, XLSTAT                   increases the greater the chance the
and SPSS software were used to                     sampled data does not provide a good
statistically analyze the data.                    representation of the entire population of
                                                   grid cells (Theobald et al, 2005).
STATSGO vs. SSURGO: Estimated Soil                          Hawth’s Analysis Tools for
Loss (A) and Scaling Effect                        ArcGIS was used to create a point
                                                   shapefile of randomly selected points for
In comparing the degree of similarity              the subwatershed. The Generate Random
and relatedness between STATSGO and                Points feature under Sampling Tools was
SSURGO RUSLE models, the area and                  used to create the point shapefile.
cell counts for each reclassified attribute                 To determine the sampling size
class were compared between soil                   needed to effectively sample the
databases. In addition, the resulting              subwatershed, the following equation
RUSLE cell values for both models at               from PennState Cooperative Extension
each scale, A (tons/acre/year), were               was employed:
sampled within the subwatershed and
compared using regression analysis.                               ___P[1-P]____
                                                                  A2 + P[1-P]
Sensitivity and Scaling Effect of                      n =      __Z2        N____
Estimated Soil Loss (A) to Model                                       R
In comparing estimated soil losses (A) to
variables for both RUSLE models, the               n = sample size required
cell values for the C, K, and LS grids             N = population size (number of cells)
were separately sampled at each scale              P = estimated degree of variance
and compared, using regression analysis,                   (i.e., 0.5 for 50-50, 0.3 for 70-30)
to their respective RUSLE cell output              A = precision desired, margin of error
values.                                                    (i.e., 0.03, 0.05, 0.1 for 3%, 5%,
Sampling                                           Z = based on confidence level: 1.96 for
                                                           95% confidence, 1.6449 for 90%,
Simple random sampling was the                             and 2.5758 for 99%
technique used to sample the Zumbro                R = estimated response rate

For the purposes of this study, the                Results
variables include:
                                                   Assessment of Soil Erosion Risk within
n = sample size required                           the Zumbro River Subwatershed
N = 461,017 raster cells
P = 60-40 = 0.4                                    Raster maps of the R, K, LS, C, and P
A = 5% = 0.05                                      grid layers were integrated within the
Z = 95% confidence level = 1.96                    ArcGIS environment to generate
R=1                                                composite maps of estimated erosion
                                                   loss within the subwatershed project site.
So the sampling size equation for this                      In all, six RUSLE empirical
study looks like:                                  models were generated. Three models
                                                   were run using the STATSGO soil
                       0.4[1-0.4]_ __ __           database and associated K values at 10,
                (0.05)     + 0.4[1-0.4]            30, and 50 meter resolutions
    n =       __(1.96)2        461017___           respectively. The remaining three
                           1                       models were run using the SSURGO soil
                                                   database and associated K values also at
where n = 368.499 = 369 for sample size            10, 30, and 50 meter resolutions
required to adequately sample the                  respectively. The resulting six RUSLE
subwatershed.                                      subwatershed maps, Figures 19-24, can
        The sample size of 369 was used            be found in Appendix A. The RUSLE
in the Generate Random Points feature              maps were each overlaid onto a hillshade
in Hawth’s Analysis Tools extension to             raster layer, created using the Spatial
create a shapefile containing 369                  Analyst extension in ArcMap, to better
randomly placed points in the                      visualize subwatershed topography.
subwatershed. This point shapefile                          Each RUSLE map was then
(Figure 12) was used to overlay with the           reclassified into six categories of
RUSLE and model variable grids at each             estimated erosion loss. The erosion loss
scale to collect cell values for                   categories were developed using
comparative analysis. A total of 8,856             previous RUSLE model reclassifications
cell values were sampled.                          from temperate U.S. regions as a guide.
                                                            Table 4 provides an example of
                                                   the estimated erosion loss categories
                                                   used (and their soil loss values) for
                                                   reclassification and the resulting cell
                                                   count, proportion, and acreage for each
                                                   erosion category. The resulting six
                                                   reclassified RUSLE models at 10, 30,
                                                   and 50 meter resolutions can be found in
                                                   Appendix B (Figures 25-30).
                                                            Table 4 shows two-thirds of the
                                                   cells that make up each raster layer fall
                                                   within the Very Low Erosion category
Figure 12. Random sampling points shapefile        where estimated soil loss is less than 3
created by Hawth’s Analysis Tools.                 tons/acre/year. Within the U.S., 3

tons/acre/year is considered an                                     The inherent benefit of natural
acceptable loss. An evaluation of the                       lowland and upland cover types and
maps reveals a significant proportion of                    hay/forage practices becomes very
these cells occur in the north, central and                 evident if one examines the north,
southern regions of the subwatershed,                       central, and southern regions of the
where more open water, wetlands,                            subwatershed and reveals that regardless
natural uplands, forests, and hay/forage                    of significant slope and topography,
cover types occur.                                          minimal erosion is estimated.

Table 4. Examples of two RUSLE models                       Comparative Analysis of the Use and
reclassified into six estimated erosion loss                Scaling Effect of STATSGO and
categories and subsequent count, proportion, and
acreage results. Soil loss, A, is in tons/acre/year.        SSURGO

                                                            In examining the level of agreement or
                                                            disagreement between a STATSGO-
                                                            based RUSLE model and a SSURGO-
                                                            based RUSLE model at 10, 30, and 50
                                                            meter resolutions, the cell counts,
                                                            proportions, and acreages of the
                                                            reclassified maps are first considered.
                                                            Three histograms, Figures 13-15,
                                                            compare acreages and soil databases at
                                                            each resolution.
                                                                     The histograms reveal that the
                                                            cell count, proportion of each erosion
        High to very high estimated soil                    category from the total, and acreage are
loss tends to occur more in the western                     very similar between the RUSLE models
and eastern regions of the subwatershed.                    that utilized STATSGO and SSURGO at
Within this landscape mosaic a greater                      each resolution. In addition to the
proportion of the subwatershed’s row                        similarity so far observed between the
crops are found.                                            soil databases, there is also a trend at
        What is interesting is that the
                                                                                            Erosion Risk and Total Area:
central and eastern regions of the                                                         STATSGO vs. SSURGO (10 meter)
subwatershed have greater slope and
topography. Parts of the eastern region                                                                                                          STATSGO

exhibit moderate to very high estimated                                         6000                                                             SSURGO
                                                                 Area (acres)

erosion loss, possibly due to the density                                       4000

of agricultural lands like row crops and                                        2000

areas with moderately exposed soils,
combined with topography. When parts                                                   No Erosion   Very Low
                                                                                                               Low Erosion   Moderate
                                                                                                                                        High Erosion   Very High

of the western region that have moderate                                                                          Erosion Risk

to very high estimated erosion loss are
examined, you have a greater                                Figure 13. STATSGO vs. SSURGO: total area of
agricultural presence but significantly                     estimated soil loss, by erosion category, at 10
reduced topography.                                         meters resolution.

                                Erosion Risk and Total Area:                                                        Lastly, a regression analysis was
                               STATSGO vs. SSURGO (30 meter)
                                                                                                            conducted to better understand the
                                                                                                            relatedness between STATSGO and
                    6000                                                             SSURGO                 SSURGO soil databases with respect to
     Area (acres)

                                                                                                            their estimation of soil erosion loss.
                                                                                                            Regression analysis is often used to
                                                                                                            model relationships between variables,
                           No Erosion   Very Low   Low Erosion   Moderate   High Erosion   Very High        determine the degree of the relationship,
                                         Erosion                  Erosion                   Erosion

                                                      Erosion Risk                                          and can be used to make predictions
                                                                                                            based on the models.
Figure 14. STATSGO vs. SSURGO: total area of                                                                        Before using regression analysis,
estimated soil loss, by erosion category, at 30                                                             it must first be determined what type of
meters resolution.                                                                                          regression is needed based on the
                                                                                                            available data. Linear (parametric)
                                Erosion Risk and Total Area:                                                regression assumes the data are
                               STATSGO vs. SSURGO (50 meter)
                                                                                                            continuous, independent, normally
                                                                                                            distributed, and the variance is equal
                    6000                                                             SSURGO                 (homoskedastic). Semiparametric
     Area (acres)

                                                                                                            regression assumes the data are not
                                                                                                            normally distributed and preserves the
                                                                                                            simplicity of parametric regression while
                           No Erosion   Very Low   Low Erosion   Moderate   High Erosion   Very High        employing the flexibility of
                                         Erosion                  Erosion                   Erosion

                                                      Erosion Risk                                          nonparametric regression. If the data are
                                                                                                            known not to be normally distributed,
Figure 15. STATSGO vs. SSURGO: total area of                                                                nonparametric regression would be
estimated soil loss, by erosion category, at 50                                                             better suited for analysis because it does
meters resolution.                                                                                          not make assumptions about the
                                                                                                            frequency distribution of the variables
each resolution in which calculated                                                                         and is much more flexible so as to more
erosion values (A) from SSURGO-based                                                                        likely detect the relatedness between the
RUSLE models score slightly lower, on                                                                       data.
average, in their estimations of soil loss.                                                                         The data is known to be
This can be seen from the basic                                                                             continuous and independent so the data
statistics, specifically the means, for                                                                     must be tested for normality to help
each RUSLE model in Table 5.                                                                                make the final determination on which
                                                                                                            type of regression to run. The lack of
Table 5. Basic statistics for STATSGO and                                                                   normality in data, including the presence
SSURGO-based RUSLE models at 10, 30 and 50
                                                                                                            of outliers, can falsely impact the
meter resolutions.
                                                                                                            correlation coefficient, R2, if normality
                                                                                                            is assumed incorrectly and linear
                                                                                                            (parametric) regression is used.
                                                                                                                    Using the Shapiro-Wilk,
                                                                                                            Anderson-Darling, Lilliefors and Jarque-
                                                                                                            Bera tests for the data sampled from
                                                                                                            each grid at 10, 30, and 50 meter

resolutions, the presence or absence of                                                                                          Robust Lowess Nonparametric Regression of Soil
normality in the data could be examined.                                                                                         Erosion Risk (A): STATSGO vs. SSURGO (30 meter)

        Scientific data from many                                                                                              130

disciplines exhibit strong nonconformity
to parametric models (Yang, 2006). So it

                                                                                                STATSGO [A (tons/acre/year)]
came as little surprise that each test                                                                                         90

calculated that the sampled data was not                                                                                       70

normally distributed, therefore strongly                                                                                       50

suggesting that for the sampled data, a
nonparametric regression technique is
the most suitable to detect the degree of                                                                                       10

relatedness.                                                                                                                   -1 0
                                                                                                                                 0                50                100            150

        The Robust Lowess                                                                                                                      SSURGO [A (tons/acre/year)]

nonparametric regression technique was
used to determine the relatedness                                                             Figure 17. STATSGO vs. SSURGO: Robust
between STATSGO-based RUSLE                                                                   Lowess nonparametric regression for soil erosion
samples and SSURGO-based RUSLE                                                                risk (A) at 30 meters resolution. R2 = 0.928
samples. Regression results have shown                                                        (blue: soil grid data, red: nonparametric
that estimated erosion loss values for
                                                                                                                                 Robust Lowess Nonparametric Regression of Soil
models at 10, 30, and 50 meter                                                                                                   Erosion Risk (A): STATSGO vs. SSURGO (50 meter)
resolutions are related. The correlation
coefficient (R2) is 0.914, 0.928, and                                                                                          150

0.922 for 10, 30, and 50 meter                                                                                                 130
                                                                                                STATSGO [A (tons/acre/year)]

resolutions respectively. The Robust                                                                                            1

Lowess regression for each scale can be                                                                                        90

seen in Figures 16-18.                                                                                                         70


                                  Robust Lowess Nonparametric Regression of Soil                                               30
                                  Erosion Risk (A): STATSGO vs. SSURGO (10 meter)

                                                                                                                               -1 0               50                100            150

                                                                                                                                               SSURGO [A (tons/acre/year)]
  STATSGO [A (tons/acre/year)]

                                                                                              Figure 18. STATSGO vs. SSURGO: Robust
                                 50                                                           Lowess nonparametric regression for soil erosion
                                                                                              risk (A) at 50 meters resolution. R2 = 0.922
                                 30                                                           (blue: soil grid data, red: nonparametric

                                 -1 0
                                   0        20          40        60           80   10
                                                                                    0                 For every RUSLE value
                                                 SSURGO [A (tons/acre/year)]
                                                                                              calculated using the STATSGO soil
                                                                                              database, there is a high degree of
Figure 16. STATSGO vs. SSURGO: Robust                                                         confidence that the estimated erosion
Lowess nonparametric regression for soil erosion                                              loss value will be similar to the value
risk (A) at 10 meters resolution. R2 = 0.914                                                  calculated using the SSURGO soil
(blue: soil grid data, red: nonparametric                                                     database and vice versa. An important
                                                                                              note to make, however, is that when

examining Figures 16-18, the relatedness           and direct impact on RUSLE model
of low to moderate estimated erosion               outputs.
values between RUSLE models is
greater but for high to very high                  Table 6. The correlation coefficients (R2) of
estimated erosion loss values the                  model variables to estimated soil erosion loss
                                                   (A) for each RUSLE model at 10, 30, and 50
relatedness is less. These high erosion            meter resolutions using Lowess nonparametric
values are what drive the correlation              regression.
coefficient down from 1 to 0.918, 0.928,
and 0.922 for 10, 30, and 50 meter
resolutions respectively.

Assessment of Sensitivity and Scaling
Effect of Estimated Soil Loss to Model

In examining the level of sensitivity
between the model variables, LS, K, and
C to estimated soil loss (A) at 10, 30,
and 50 meter resolutions, the same
principle behind using nonparametric
regression analysis to determine
relatedness is used as the section above.                  The model variable C (cover
This includes lack of normality in                 management) has a greater relatedness to
conjunction with data being continuous             RUSLE model outputs at each scale than
and independent. Here, it is assumed that          LS and K, but not what would be
the greater the variable sensitivity to            deemed a significant relationship. For
estimated soil erosion loss the greater the        STATSGO-based RUSLE models the R2
relatedness.                                       is 0.493, 0.426, and 0.514 at 10, 30, and
        Table 6 below provides the                 50 meter resolutions respectively. For
coefficients of correlation (R2) for each          SSURGO-based RUSLE models the R2
sampled model variable regressed                   is 0.489, 0.411, and 0.516 at 10, 30, and
against its corresponding RUSLE model              50 meter resolutions respectively. It
at 10, 30, and 50 meter resolutions. The           would seem that the C factor, which
Lowess nonparametric regression                    includes land cover types and associated
technique was used for this analysis of            soil exposure, may play a slightly greater
relatedness. In all, 18 nonparametric              role in determining estimated soil
regressions were run.                              erosion loss for this project but not at a
        Table 6 suggests the model                 significant level.
variables LS (slope length and
steepness) and K (soil erodibility) are            Conclusion
not related to their corresponding A
values for either models that use                  The RUSLE empirical model was
STATSGO and SSURGO at any scale.                   applied six times to the Zumbro River
For this project, LS and K variables do            subwatershed during this study. The
not seem to provide a highly significant           variables, R, LS, C, and P were identical
                                                   for each model except for K (soil

erodbility). Three models used the                 River subwatershed but it may be that
STATSGO soil database at 10, 30, and               the sampling method used here did not
50 meter resolutions and the remaining             recognize any potentially existing
three models used the SSURGO soil                  relationship between LS and A.
database also at 10, 30, and 50 meter                      This study demonstrates that GIS
resolutions.                                       is a valuable tool in assessing soil
         The spatial distribution and              erosion modeling and in assisting the
estimated erosion loss values within the           estimation of erosion loss at the
subwatershed were significantly related            subwatershed scale. But there are
when comparing STATSGO and                         limitations that must be taken into
SSURGO-based RUSLE models at each                  account prior to modeling including the
resolution. Relatedness of estimated               quality of data and the spatial resolution
erosion loss values (A) between the soil           used.
databases at each resolution, however,                     The RUSLE model exemplifies
was greater for very low to moderate soil          that spatial resolution is sensitive to the
losses and lessened dramatically for high          estimations of erosion so caution must
to very high soil losses. The mean A               be taken when selecting grid size. When
(tons/acre/year) for the STATSGO-                  considering soil erosion modeling at
based RUSLE models were 4.23, 5.49,                scales much smaller than the
and 6.71 for 10, 30, and 50 meter                  subwatershed level (i.e. townships,
resolutions respectively. The mean A for           parcels, etc.), it is recommended that soil
the SSURGO-based RUSLE models                      databases chosen be more complex than
were 3.96, 5.15, and 6.28 for 10, 30, and          STATSGO. Lastly, caution must also be
50 meter resolutions respectively.                 practiced with data since minor errors
         For this study, the C model               can exponentially increase and skew
variable was more related to each                  results thereby compromising the
corresponding A than the other variables           implementation of conservation
but not at a significant level. This infers        practices, education, and funds to
that in the subwatershed, the C variable,          address soil erosion issues.
cover management, is a better indicator
for resulting RUSLE outputs, A. It is              Acknowledgements
generally accepted that ground cover is
the most important factor in the soil              I would like to take this opportunity to
erosion process, especially when                   thank Dr. David McConville, Mr. John
considering surface cover, canopy cover,           Ebert, and Mr. Patrick Thorsell in the
surface roughness and prior land use               Department of Resource Analysis for
(Yazidhi, 2003).                                   their assistance and ongoing support for
         Based on literature searches,             this project and during my tenure at
additional assumptions would have led              Saint Mary’s University.
to the LS variable, slope length and
steepness, as another good indicator for           References
estimations of soil loss (Lee and Lee,
2006; Liu et al, 2000). In these two cited         Ding, T., Zhang, M., and P. Hendley.
research projects in Korea and China,               1999. Comparison of the Soil
areas under study occurred on steeper               Properties in State Geographic
slopes than what is found in the Zumbro             (STATSGO) and National Resources

  Inventory (NRI) Databases. Retrieved:  
  June 11, 2006 from http://agis.ucdavis.          content/full/64/5/1759
  edu/research/ASA99_poster.pdf                   Moore, I., and G. Burch. 1986. Physical
Gowda, P.H., and D.J. Mulla. 2005.                 Basis of the Length-Slope Factor in the
  Scale Effects of STATSGO vs.                     Universal Soil Loss Equation. Soil
  SSURGO Soil Databases on Water                   Science Society of America Journal.
  Quality Predictions. Retrieved: June 8           50:1294-1298
  2006 from              Porter, S.D., Harris, M.A., and S.J.
  research/publications/Publications.htm           Kalkhoff.2001. Influence of Natural
  ?seq_no_115=178707                               Factors on the Quality of Midwestern
Hickey, R., Burns, E., Bolte, J., and D.           Streams and Rivers. Retrieved: March
  Walker. 2005. Development of a                   29, 2006 from
  Statewide Erosion Vulnerability                  wri/wri004288/pdf/section01.pdf
  Screening Tool for Oregon. Retrieved:           Renard, K.G., Foster, G.R., Weesies,
  March 29, 2006 from http://www.siue.             G.A., McCool, D.K., and D.C.
  edu/geography/online/Hickey05.pdf                Yoder.1997. Predicting Soil Erosion
Jones, D.S., Kowalski, D.G., and R.B.              by Water: A Guide to Conservation
  Shaw. 1996. Calculating Revised                  Planning with the Revised Universal
  Universal Soil Loss Equation                     Soil Loss Equation (RUSLE).
  (RUSLE) Estimates on Department of               Retrieved: April 3, 2006 from
  Defense Lands: A Review of RUSLE                 htpp://
  Factors and U.S. Army Land                       library/hbmanual/rusle/ah_703.pdf
  Condition-Trend Analysis (LCTA)                 Shi, Z.H., Cai, C.F., Ding, S.W., Li,
  Data Gaps. Retrieved: May 19, 2006               Z.X., Wang, T.W., and Z.C. Sun.
  from             2002. Assessment of Erosion Risk
  files/tps-96-8.pdf                               with the RUSLE and GIS in the
Lee, G.S., and K.H. Lee. 2006. Scaling             Middle and Lower Reaches of
  Effect for Estimating Soil Loss in the           Hanjiang River. Retrieved: March 28,
  RUSLE Model using Remotely Sensed                2006 from
  Geospatial Data in Korea. Retrieved:             gov/isco/isco12/VolumeIV/
  May 19, 2006 from http://www.                    AssessmentofErsoionRisk.pdf            Sonneveld, B.G.J.S., and M.A. Nearing.
  hessd-3-135_p.pdf                                2003. A Nonparametric/Parametric
Levick, L.R., Semmens, D.J., Guertin,              Analysis of the Universal Soil Loss
  D.P., Burns, I.S., Scott, S.N., Unkrich,         Equation. Retrieved: April 3, 2006
  C.L., and D.C. Goodrich. 2004.                   from
  Adding Global Soils Data to the                  unitPublications/PDFfiles/1571.pdf
  Automated Geospatial Watershed                  Theobald, D.M., Stevens, D.L., White,
  Assessment Tool (AGWA). Retrieved:               D., Urquhart, N.S., and A.R. Olsen.
  May 3, 2006 from                 2005. Using GIS to Generate
  nerlesd1/land-sci/agwa/pdf/pubs                  Spatially-Balanced Random Survey
  global-soil.pdf                                  Designs for Natural Resource
Liu, B.Y., Nearing, M.A., Shi, P.J., and           Applications. Retrieved: August 12,
  Z.W. Jia. 2000. Slope Length Effects             2006 from http://www.nrel.colostate.
  on Soil Loss for Steep Slopes.                   edu/projects/starmap/download/rrqrr/
  Retrieved: June 11, 2006 from                    theobald_etal_RRQRR20050706.pdf

U.S. Department of Agriculture, Natural
 Resource Conservation Service. 1994.
 State Soil Geographic (STATSGO)
 Data Base: Data Use Information.
 Misc. Publication No. 1492. National
 Soil Survey Center. U.S. Department
 of Agriculture. Washington, DC.
Wang, X., and P. Cui. 2005. Support
 Soil Conservation Practices by
 Identifying Critical Erosion Areas
 within an American Watershed Using
 the GIS-AGNPS Model. Retrieved:
 March 28, 2006 from http://www.
Wilson, J.P. 2003. GIS-based Land
 Surface/Subsurface Modeling: New
 Potential for New Models. Retrieved:
 March 28, 2006 from http://www.
Wischmeier, W.H., and D.D. Smith.
 1978. Predicitng Rainfall Erosion
 Losses: A Guide to Conservation
 Planning. Agriculture Handbook No.
 537. U.S. Department of Agriculture,
 Washington, DC.
Yang, L. 2006. Have you Considered
 Nonparametric Regression? Retrieved:
 June 10, 2006 from http://www.msu.
Yazidhi, B. 2003. A Comparative Study
 of Soil Erosion Modelling in Lom
 Kao-Phetchabun, Thailand. Retrieved:
 June 20, 2006 from
Yusof, K.W., and S.M. Baban. 1999. A
 Preliminary Attempt to Develop An
 Erosion Risk Map for Langkawi
 Island, Malaysia Using the USLE,
 Remote Sensing and Geographic
 Information System. Retrieved:
 March 28, 2006 from http://www.

Appendix A. RUSLE models and estimated soil erosion loss (tons/acre/year) using STATSGO and
SSURGO soil databases at 10, 30 and 50 meter cell sizes. Hillshading added for topographic visualization.

Figure 19. Estimated soil erosion loss at                     Figure 22. Estimated soil erosion loss at
10 meters resolution using STATSGO.                           10 meters resolution using SSURGO.

Figure 20. Estimated soil erosion loss at                     Figure 23. Estimated soil erosion loss at
30 meters resolution using STATSGO.                           30 meters resolution using SSURGO.

Figure 21. Estimated soil erosion loss at                     Figure 24. Estimated soil erosion loss at
50 meters resolution using STATSGO.                           50 meters resolution using SSURGO.

Appendix B. Reclassified RUSLE models and categorized soil erosion loss using STATSGO and SSURGO
soil databases at 10, 30 and 50 meter cell sizes.

Figure 25. Reclassified RUSLE model at                  Figure 28. Reclassified RUSLE model at
10 meters resolution using STATSGO.                     10 meters resolution using SSURGO.

Figure 26. Reclassified RUSLE model at                  Figure 29. Reclassified RUSLE model at
30 meters resolution using STATSGO.                     30 meters resolution using SSURGO.

Figure 27. Reclassified RUSLE model at                  Figure 30. Reclassified RUSLE model at
50 meters resolution using STATSGO.                     50 meters resolution using SSURGO.


To top