Research and experience have delivered a rich set of tools and techniques for evaluating and managing soil erosion from various sources within watersheds. While significant gains in the control of soil erosion have been made, there remain significant problems with sedimentation of water supply and navigations systems, transport of soil attached contaminants and the cost of some management techniques. Further, while much research exists for individual methods in individual locations little work has been devoted to integrated approaches which characterize interaction between erosion processes and management methods across scales and within a system of land management strategies
To better support the multilevel management we propose a methodology for watershed characterization and erosion modeling at multiple scales and levels of complexity. Because the spatial unit, modeled at the local level is a part of a larger watershed, the evaluation of the impact of numerous, locally implemented conservation practices on the entire watershed requires multiscale approach which links the high resolution, implementation level simulation with low resolution/regional simulation. Models and simulations of watershed behavior demonstrating how the lands within the watershed are connected, how they interact and influence each other help to better understand the impact of local actions within the watershed necessary for succesfull conservation and sustainable land management programs. Complex spatial interactions have to be considered in decisions about optimal locations of training and conservation areas.
Recent advances in Geographic Information Systems (GIS) technology, linkage of numerous models with GIS, (Moore et al. 1993, Maidment 1996, Saghafian et al. 1995, Srinivasan and Arnold 1995) create a potential to develop an environment for coordination of conservation efforts at a hierarchical system of management levels by providing the tools to design and evaluate the impact of land use alternatives at both local and watershed/regional levels for a given time horizon. They facilitate evaluation of prevention practices not just depending on the type of the prevention measure, but also on their location within the watershed. Spatial analysis and simulations can also provide supporting information for alocation of resources to those areas and those types of practices which will provide the most effective protection. Distributed models facilitating multiscale approach have been recently identified as promising for studies of landscapes with significant human impact, such as agricultural and urbanized watersheds (e.g., Quinn 1998, Mitas and Mitasova 1997, 1998). However advances in models, algorithms and GIS tools are needed to fully support this approach.
One area of special promise to address this problem is simulation modeling which can give decision makers interactive tools for both understanding the physical system and judging how management actions might affect that system. However, a report "New Strategies for America's Watersheds" (Committee on Watershed Management, National Research Council 1999) analyzes the current status in watershed modeling for decision making and concludes that the available models and methods are outdated and "a major modeling effort is needed to develop and implement state-of-the-art models for watershed evaluations"(pp.160-161).
In this report, we describe the new development and improvements of
GIS based methods for processing of input data for erosion models, implementation
of simple erosion models in GIS (GRASS and ArcView) and applications of
models at Ft. Hood and Ft. Polk.
Local level of management requires detailed spatial representation (10-1m resolution) and models capable to simulate effects of spatially variable land cover. This level of detail is necessary for implementation of conservation practices at the most effective locations, as well as for selecting the projects that save the most soil or benefit the most acres per dollar cost. The empirical models such as USLE/RUSLE, which have been used at this level for many years, are now being replaced by process-based models, such as WEPP (Flanagan and Nearing 1995). Both RUSLE and WEPP are based on routing of water and sediment over hillslope segments and pre-defined channels, a concept which requires substantial manual input and is not very practical for large areas typical for military installations. New generation of models based on raster representation of multivariate functions has been developed recently with the support from SERDP and ARO (SIMWE: Mitas and Mitasova 1998, CASC2d: Julien, Ogden, Saghafian, Johnson 1995-99, MIT models (Brass, Wilgoose, Moglen, ... CHILD)). These models have a potential, after undergoing a substantial experimental testing and calibration to be useful for both local and a regional scale modeling.
The watershed models based on homogeneous spatial units have been described in detail in literature (e.g., Arnold et al. 1993) and are widely available, including the possibility to run it online. Therefore, we focus on distributed modeling of erosion and deposition patterns for spatially variable conditions. Within the spatially continuous approach, model inputs and outputs are represented by multivariate functions discretized as grids. Flows of water and sediment are described as bivariate vector fields rather than commonly used systems of 1D flows (e.g., Moore et al., 1993). To support modeling at different levels of complexity we have developed a set of tools which range from modifications of relatively simple empirical models to a more complex, process-based approach. GIS technology is used to support the processing, analysis and visualization of the data and simulation results (Mitas et al., 1997). The models are described in detail in our previous report (Mitasova et al. 1998) therefore in the following sections we focus on the issues related to practical applications of the models and their use with GIS.
Preparation of data for modeling and land use management still requires a substantial effort, in spite of the great progress in the measurement techniques and distribution of data. While the spatial data are currently much easier to exchange and import/export between different systems, with higher resolutions the data sets are substantially larger and often more complex. The growth of the size of data sets is "keeping up or even exceeding" the growth of computational power and efficient GIS tools for processing, analyzing and visualizing spatial data for land use management are as important as ever. When processing the spatial data, we are focusing on extensive use of standard GIS tools available in both commercial and public license systems, such as GRASS5.0 and ArcView/Spatial Analyst, however enhancements or modifications of existing tools are often needed to fulfill the needs of distributed, process-based modeling.
Spatial interpolation and smoothing is necessary for transformation of scattered data to regular grids, resampling of grid data and smoothing of data with noise. The selection of an adequate method and its parameters can have a profound impact on the predicted patterns, as we have shown e.g. in Mitas and Mitasova 1999. The current project has again demonstrated the crucial role of spatial interpolation for processing of DEMs from various sources (ISARE, contours, field measurements) as well as for the interpolation of Ceasium data. We have used and enhanced the bivariate regularized spline with tension and smoothing (RST, Mitas and Mitasova 1999) and its implementation in GRASS5.0 to perform spatial interpolation and topographic analysis for this project applications, in particular for smoothing of the IFSARE data, interpolating field measured elevations with fault, interpolating very large combined point and contour data set and deriving surface gradients for water and erosion models.
Interpolation of data with a fault by RST. Spline is by definition a smooth function and has always been regarded as a tool for representation of smooth surfaces. Therefore its application to surfaces with faults has been often described as problematic. However, RST combined with GIS masking tools can be very effectively used to create models of surfaces with complex pre-defined faults by using the multiple-pass approach similar to zonal kriging. First, the points defining the fault are extracted from the elevation data set based on their attributes and transformed from site through vector to raster masks representing continuous (smooth) subsets of the data. The masks are then used to create separate elevation data subsets for each smooth subregion which is then interpolated by RST. The masked surfaces are then merged using map algebra to create a single surface with faults (Figure 4a,b,c gully.) Each subregion can be interpolated at different resolution to create a multiscale representation of a surface (Figure 4d?).
Surface analysis provides basic topographic or surface geometry parameters needed for modeling of fluxes in landscape including water and sediment. Estimation of slope and aspect (direction of flow) are standard functions in GIS software and their computation using the RST methods is is described by Mitasova and Mitas 1993. For several hydrologic and erosion models which simulate water and sediment flow as a bivariate function, partial derivatives representing the surface gradient grad z, are needed:
We have developed two approaches to estimate partial derivatives. First is a commonly used local polynomial approximation of a surface at a grid point from the elevations in its 3x3 neighborhood (Horn 1984, Shapiro and Westervelt 1992). This approach can be applied only to raster data and has been implemented as an option for the command r.slope.aspect in GRASS5.0 The second approach uses the derivatives of the RST function (Mitasova and Hofierka 1993), it can be used for both scattered point data and raster data and is implemented as an option for the s.surf.rast command in GRASS 5.0. To compute the partial derivates within a GIS where the direct computation of derivatives is not available we use the known relationship between the slope and aspect and derivatives:
where b is slope and a is aspect in degrees, computed by standard GIS functions.
No improvements to flowtracing r.flow except for bug fixes fro GRASS5.0.
Terrain modification tools are useful for incorporation of data about streams and rivers into terrain model, which usually captures only the approximate water surface elevation. We have developed a methodology for modification of a DEM by "carving-in" streams and rivers using their vector representation and attributes such as width and depth. The method was implementated as r.enforce in GRASS5.0, an example of terrain model without and with a river is in Figure 5. (BILL describe in more detail).
Visualization is an integral part of input data processing as well as of communicating the results. AS a result of collaboration with a research group in Slovakia and Germany, the 3D surface visualization tool NVIZ2.2 is now available within GRASS5.0 for LINUX - enhancements??? New visualization tools were developed for tracking the computation progress during interpolation of very large DEMs (milions of cells from hundreds of thousands of points) using the RST methodi, which often takes several hours even on fast, state of the art computers. BILL add more details and Figure 6.
We have renewed our use of p.vrml which allows us to directly serve the results as 3D interactive models on the Internet, because the more and better viewers are available although they still are not well designed for exploration of geospatial data.
where E [ton/(acre.year)] is the average soil loss, R [hundreds of ft.tonsf.in/acre.hr.year] (in SI: [MJ.mm/ha.hr.year ], R[SI]=17.02R[EU]) is the rainfall intensity factor, K [tons per acre per unit R] = [tons.acre.hr/hundreds.acre.ft.tonsf.in] is the soil factor, LS [dimensionless] is the topographic (length-slope) factor, C [dimensionless] is the cover factor and P[dimensionless] is the prevention practices factor. Several approaches were developed to adapt this field based model to GIS use. While slope is realtively easy to compute, slope length was more problematic. Examples of how L-factor is impleneted include use of average slope length for the entire area, average slope length for each soil type and spatially variable slope length estimated for each cell from a DEM using a flowtracing program such as r.flow in GRASS (Mitasova and Hofierka 1993, Figure 9, Appendix). The LS factor is then computed as
where l is the slope length and b is the slope angle.
To incorporate the impact of flow convergence, a replacement of the slope length by the upslope contributing area per unit contour width was suggested, e.g., by Moore and Wilson 1992, Mitasova et al. 1996, Desmet and Govers 1996. The modified LS factor at a point on a hillslope is
Both the standard and modified equations can be properly applied only
to areas experiencing net erosion. Depositional areas should be excluded
from the study area because the model assumes that transport capacity
exceeds detachment capacity everywhere and erosion and sediment transport
is detachment capacity limited. Therefore, direct application of USLE/RUSLE
to complex terrain within GIS is rather restricted. The results can also
be interepreted as an extreme case with maximum spatial extent of erosion
possible.
2.2.2 Unit Stream Power based Erosion/Deposition model
USPED (Unit Stream Power - based Erosion Deposition) is a simple model
which predicts the spatial distribution of erosion and deposition rates
for a steady state overland flow with uniform rainfall excess conditions
for transport
capacity limited case of erosion process. The model is based on the
theory originally outlined by Moore and Burch
1986 with numerous improvements. For this case, we assume that the
sediment flow rate qs(r) is at
the sediment transport capacity T(r), (Julien and Simons
1985)
No experimental work was performed to develop parameters needed for USPED, therefore we use the USLE or RUSLE parameters to incorporate the impact of soil and cover and obtain at least a relative estimate of net erosion and deposition.We assume that we can estimate sediment flow at sediment transport capacity as
where R~im, KCP~Kt and LS=Am sin b n (add units here), and m=1.6, n=1.3 for prevailing rill erosion while m=n=1 for prevailing sheet erosion USPED - role of exponent - type of flow 1.5 rill, 1 sheet, 0.6 small event before steady state. Then the net erosion/deposition is estimated as
where a [deg] is aspect of the terrain surface.This equation is equivalent
to the relationship with curvatures presented
in the report by Mitasova et al. 1998, however the computational procedure
is simpler. Caution should be used when interpreting the results
because the USLE parameters were developed for simple plane fields and
detachment limited erosion therefore to obtain accurate quantitative predictions
for complex terrain conditions they need to be re-calibrated ( Foster 1990,
Mitasova et al 1997 reply).
2.2.3 Process-based simulation of water erosion
SIMulation of Water Erosion model (SIMWE) is based upon the description of water flow and sediment transport by first principles equations (e.g., Foster and Meyer 1972, Bennet 1974). The model is described by Mitas and Mitasova (1998a), here we briefly present only its principles
equations including Di and infiltration
The following enhancements were developed for this project:
To approximatelly simulate the pattern of water depth in areas with depressions and flat terrain we have introduced a variable diffusion rate which depends on the depth of water and "momentum" estimated fro velocity vectors from n-cells before water enters the flat area or depression. This model enhancement allows us to create maps representing natural, topographic conditions fro wetlands. It also allows us to create maps of relatively wet areas (distingtuish the "lowland and upland" areas which have different vegetation etc.
Flat areas or depresssions are often drained by natural or manmade channels
which are given as vector data and usually are not very well represented
in the DEMs. SIMWE was further enhanced to simulate the flow in flat areas
either as a shallow diffusive flow creating wetlands or through predefined
channels given by stream/channel network in vector format. While our previous
efforts focused on hillslope water flow and erosion these enhancements
are a step towards modeling the interaction between channels and overland
flow, an important and not very well understood phenomenon.
multiscale approach - este v sprave nebol - vezmi z viedne alebo UNESCO rep
Furthermore, solution through the Green's function given by equation (6) can be reformulated for accommodation of {\it spatially variable accuracy} and {\it resolution}. The function $W({\bf r})$ can change (abruptly or smoothly) between regions with unequal resolutions and in fact, can be optimally adapted to the quality of input data (terrain, soils, etc) so that the accurate solution is calculated only in the regions with correspondingly accurate inputs. The reweighted Green's function $G^*({\bf r},{\bf r'},p)$, in effect, introduces higher density of sampling points in the region with large $W({\bf r})$. The statistical noise will be spatially variable as $\approx 1/[W({\bf r})\sqrt{M}]$, where $M$ is the average number of samples resulting in the accuracy increase for the areas with $W({\bf r}) > 1$.
The necessity for spatially heterogeneous or segmented modeling of fluxes
can arise due to various reasons:
(a) study area is represented by data with spatially variable resolution
and accuracy
(study area has high resolution data however there is mass coming in
from an area for which we have only low resolution data);
(b) study area is large and reduction of resolution to reduce the computational
demands is not acceptable (segmented processing is needed, although this
can be solved by computations by subwatersheds, however delineations of
watersheds for large areas can be difficult too)
(c) study area is large, with spatially variable complexity for which
simulation with spatially variable precission/detail is the most effective
(study area is large, but a certain part of it is homogeneous so it is
unnecessary to run it all at high resolution,
only area of interest/highs patial variability will be simulated at
high resolution while the rest is low resolution.)
Large areas are often characterized by data with variable resolution
and the above {\it multi-scale} formulation is able to make the best use
of such heterogeneous data. Another reason why one would like to change
the resolution and accuracy is the desire of the user: very often the area
of interestis rather small, however, it requires inputs from much larger
region.The system of equations (1-4) describes the water and sediment flow
at a spatial scale equal or larger than an average distance between rills
(i.e., grid cell size $\ge$ 1m) and therefore the presented approach allows
us to perform landscape scale simulations at variable spatial resolutions
from one to hundreds of meters, depending on the complexity and importance
of studied subregions.
Floating point support and nonlinear categorization and colortables are crucial
We have tested implementation in 2 systems :
Easy to use interface for ArcView? - wait for ArcModel which will allow
easy modification without
knowledge of Avenue. Code is in appendix. problem with viewing of floating
point data - work with ESRI to solve the problem?
Because of profound changes in ESRI sowftware which involves merging
of ArcINFO and ARCVIEW,
release of ARcmodel and troubles withing Arcview 3.1,3.2 we propose
testing the model again under the new
release and implementation in ArcView GIS rather than Avenue and ArcView
3.1 which has some fundamental problems with handling the data typical
for erosion modeling (highly nonlinear floating point data).
Process-based simulation: research tool linked to GRASS and Arcinfo through import/export of raster data. Most of preprocessing and visualization/analysis is done within GIS only the simulation of the process is run outside.
The proposed multiscale approach can be implemented within any
standard raster GIS such as ARCGRID, ARCvie spatial analyst, GRASS,
IDRISI (REF) without the need for developing new data structures. for each
resolution a separate set of inputs is created with overlapping region,
simulation algorithm combines the data from different files
by going through several passes of computation.
Algorithm produces a useful byproduct - identification of borderline
segments with inflow of matter and segments with outflow - important for
indetification of areas which require protection from e.g. inflow of harmful
chemicals or reas where outflow of matter should be controlled.
To perform the modeling efficiently, customized interface can be created for particular applications...
Visualization is supported by Nviz, using multiple masked surfaces
indicate future research. Include use of r.sun, ecosystem protection etc.
Topographic potential for wetlands, wet areas
While SIMWE was developed to model erosion processes, its hydrologic
submodel supports simulation of water depth in depressions
and flat areas allowing us to identify those locations which have
natural topographic conditions for wetlands. We have simulated water depth
from shallow overland flow (without the impact of channels/streams) to
find the areas which can accumulate and hold enough water from surface
flow (Figure 3 a/b). According to the wetland inventory data from
ILGIS, these high risk subwatersheds have only 0.32% area covered by wetlands,
however the model indicates that terrain configuration has natural conditions
for 2.5% of wetlands, especially in flatter locations and depressions along
the streams.
Demosntrate at Fort Polk - water flow pattern with different levels
of channelization (are wetland inventory data available?)
Eroding grassways
Hedges
Interpolation and smoothing of IFSARE data. IFSARE data contain speckle noise which is relatively homogeneous and appears as "bumps" about 1-3m high/wide on top of the measured terrain surface with vegetation (Figure 3a,b). New types of radar measurement techniques are being developed to eliminate the noise, however the speckle noise was substantial in the data that we have used and it had to be removed. We have found that the RST method performed very well as a smoothing tool, in spite of the fact that it has not been designed specifically to deal with the type of noise occuring in the radar data (Figure 3c). ADD Equation here? Difference between IFSARE and field data and accuracy of smoothing ?
Models USLE, hUSLE, USPED only for the field data. Figure XX USLE, FIGURE XXX USPED p=0.6, p=1.6, p=aver. Color maps of erosion, ersoion/deposition, erosion status, gully potential (from 2-3 types of data, evaluate the data suitability)
Processing of Cs data (see Warren) - data, erosion estimate.
Comparison - area of erosion/deposition, difference in points. No correlation.
TABLES
calibration of erosion models based on Cs problematic - demonstrate
on Exeter,
show typical distribution of Cs (Fig from Tomlin
The Cs measurements (these and many others) have pattern of very local
movement and tillage is often used as explanation, but we don't have any
tillage here. The pattern is close to the one which we are geting from
simulations of small and/or very short events (p=0.6, or SIMWE for short
time see also Exeter, but I will do it for this field too). Then the question
is what is happening with Cs during the large events? Cs is bound to
clay which hardly deposits on hillslopes and gets carried over long distances
unless it is in aggregates.
Comaprison is difficult also due to the different sampling density - 100 points versus 1400(from our experince from precision farming it makes a big difference in correlation) New methodology is needed for comparing spatiual distributions (surfaces, biv. functions) sampled/modeled with different level of detail(sampling density, resolution,scale)
The concavities were lower and bigger and they are filling up
We have tried to use simulations not just to predict, but also to understand
and explainwhat the processes.The simulations indicate that Cs pattern
is closer to patternwhich result fro "short range" processes that to "long
range" ones (which contradicts the fact that CS modevs with water and clay).
In our case the pattern is closer to the pattern produced by many short/small
rainfall events (before the water reaches steady state) when erosion/depostion
is more evenly distributed and maxima are on the ridges, than the big events
when erosion on ridges can be susbstantial but it is still low relative
to erosion in the center of the valleys and lower part of hillslopes.
The effect of small rainfall is similar to the effect of tillage which
is being used as explanation of Cs distribution in tilled areas.
Because we know (do we?) that most of the soil is being moved during
the large events it seems that the linkage between Cs and soil erosion
by water is rather complex and there is something else going on besides
Cs moving with soil.
COMBINED/averaged many small (200), few middle (20) and one big event
are starting to get closer - WEPP and other continuous simulations models
do it so it seems that it is really necessary to get the right pattern
but setting the weights correctly (based on temporal changes in rainfall
and c) is really tricky most of the continuous simulations are lumped and
they did not
look at what impact it has on spatial pattern
In the WRR paper the p=0.6 gave the closest results to observed colluvial
deposits. Distributed field measurements indicate
that the long term effects produce different patterns than the typical
rill erosion model ...
adding interril/detachment by raindrops which is idependent from water flow (I tried it once it was very small, I will look at it again) or diffusive term reflecting combined impact of various processes (zvetravanie, wind, raindrops, ...) should be bigger or dynamics in terrain (however subtle) in long term is important (e.g. as in hedges) - see MIT models???
Cs moves without water much more than with water (weathering, wind, raindrop splash, ...)
sheet erosion, small event erosion and rill/large event erosion, also soil type has impact. discuss tillage
compare typical and extreme events task 4, show that small event can
be approximated by USPED when p<1, demonstrate small events in SIMWE
(exeter, new Ft.Hood)
Figure XX
We have compute basic topographic paramneters and provided the results to MECA as a part of their contract . As a part of this project we computed topographic potential for hillslope erosion represented by the LS-factor using equation (x). The results show that .........The average LS factor is .... (if we use average C from LCTA and K=0.3... the average soil loss would be.... When we recieve the land use map we can compare this with the rates in intensive use areas...
...Comparison with the LCTA data, any BMPs in place or planned - for
future research
get the data and do it
We ahve used SIMWE to simulate water flow erosion/deposition in a subarea. Potential for wetlands in this area...
To illustrate the capability of the presented approach to simulate erosion/deposition
patterns at spatially variable accuracy/resolution we have applied the
model to a $36 km^2$ mountainous area at $20m$ resolution, with more detailed
predictions ($10m$ resolution) within a potential high intensity use $7
km^2$ subarea targeted for
rehabilitation. Detailed solution for the entire area would require
about 2 million samples applied to 360 000 grid cells. With spatially variable
resolution significant savings in processing time can be achieved by using
only 20 000 samples for 75 000 cells covering the entire study area at
$20m$ resolution, and a higher density sampling (2 mil.) used only
for the targeted area represented by 50 000 cells ($10m$ resolution).
The predicted sediment flow and erosion/deposition patterns have
a detectable level of noise in low resolution area, while the distributions
in the targeted subarea are modeled with high level of detail (Fig. 4).
%dispersal flow, split streams, alluvial cones in
% the high resolution area (Figure 1). Further investigation will be
performed into the error propagation from low to high resolution areas
and its dependence on the location of subarea within the basin structure.
SIMWE:Infiltration, dynamics
For both models more general equation for detachmenta and transport
capacity is needed - this is area of active research (e.g. Nearing new
slope-factor for RUSLE represented by 1 equation rather than previous 3
different ones).
There is a strong need for ca.ibartion of models both quantitatively
and spatially.
Bennet, J. P., 1974, Concepts of Mathematical Modeling of Sediment Yield, Water Resources Research, 10, 485-496.
Desmet, P. J. J, and Govers, G., 1995, GIS-based simulatin of erosion and deposition patterns in an agricultural landscape: a comparison of model results with soil map information. Catena 25, 389-401
Desmet, P. J. J., and G. Govers, 1996, A GIS procedure for automatically calculating the USLE LS factor on topographically complex landscape units, J. Soil and Water Cons., 51(5), 427-433.
Flanagan, D. C., and M. A. Nearing (eds.), 1995, USDA-Water Erosion Prediction Project, NSERL, report no. 10, pp. 1.1- A.1, National Soil Erosion Lab., USDA ARS, Laffayette, IN.
Foster, G. R., and L. D. Meyer, 1972, A closed-form erosion equation for upland areas, in Sedimentation: Symposium to Honor Prof. H.A.Einstein, edited by H. W. Shen, pp. 12.1-12.19, Colorado State University, Ft. Collins, CO
Foster, G. R., 1990, Process-based modelling of soil erosion by water on agricultural land, in Soil Erosion on Agricultural Land}, edited by J. Boardman, I. D. L. Foster and J. A. Dearing, John Wiley & Sons Ltd, pp. 429-445
Haan, C. T., B. J. Barfield, and J. C. Hayes, 1994, Design Hydrology and Sedimentology for Small Catchments, pp. 242-243, Academic Press.
Julien, P.Y., and Simons, D.B., 1985, Sediment transport capacity of overland flow. Transactions of the ASAE, 28, 755-762.
Julien, P. Y., B. Saghafian, and F. L. Ogden, 1995, Raster-based hydrologic modeling of spatially varied surface runoff, Water Resources Bulletin, 31(3), 523-536.
Mitas, L., and Mitasova, H., 1998a, Distributed soil erosion simulation for effective erosion prevention. Water Resources Research, 34(3), 505-516.
Mitas, L., Mitasova, H., 1998, Multi-scale Green's function Monte Carlo approach to erosion modelling and its application to land-use optimization In: Modelling Soil Erosion, Sediment Transport and Closely Related Hydrological Processes, (Eds: W.Summer, E. Klaghofer and W.Zhang), IAHS Publication no. 249, pp. 81-90.
Mitas L., Brown W. M., Mitasova H., 1997, Role of dynamic cartography in simulations of landscape processes based on multi-variate fields. Computers and Geosciences, Vol.23, No. 4, pp. 437-446
Moore, I.D., Turner, A.K., Wilson, J.P., Jensen, S.K., Band, L.E., 1993, GIS and land surface-subsurface process modeling. In GIS and Environmental Modeling, edited by Goodchild, M.F., Parks, B., Steyaert, L.T., Oxford University Press, New York, 196-203.
Moore, I.D., and Wilson, J.P., 1992, Length-slope factors for the Revised Universal Soil Loss Equation: Simplified method of estimation. J. Soil and Water Cons., 47, 423-428.
Nearing M.A., Norton L.D., Bulgakov D.A., Larionov G.A., West L.T., Dontsova K.M., 1997, Hydraulics and Erosion in eroding rills. Water Resources Research, 33, 865-876.
Srinivasan, R., and Arnold, J.G., 1994, Integration of a basin scale
water quality model with GIS, Water ResourcesBulletin, 30(3), 453-462.
Figure 2. Heterogeneos elevation data for Hohenfels
Figure 3. Processing of IFSARE data
Figure 4. Interpolation of a surface with fault
Figure 5. Carved-in river
Figure 6. Tracking computation of RST
Figure 7. Sharing results via Internet as VRML models
Figure 8. Set of distributed erosion models with increasing complexity a) USLE based on slope-length, b) modified USLE using upslope area, c) USPED erosion and deposition, d) SIMWE erosion and deposition.
Figure 9. Soil loss predcited by USLE using different methods for estimation of the LS-factor
Figure 10. Sediment flow and Erosion/deposition pattern predicted by USPED using different water term exponents
Figure 3. Multiscale water depth simulation: a) 10m resolution with linked in 2m resolution water depth pattern, b) c) comparison of results from 10m and 2m resolution for the linked-in subarea.
Figure 4. Spatial pattern of potential soil loss computed by the modified USLE at 30m resolution a) entire 91 square miles watershed, b) cover, c) modfied LS factor and d) USLE estimatefor zoomed in detail.
Figure 5. Erosion and deposition pattern for watershed with severe
disturbances a) C-factor, b) erosion/deposition pattern for 2 selected
subwatersheds and the respective histograms of spatial extent of erosion.
Ft. Polk
elevation: 5m from points+contours
topo analysis: slope, aspect, upslope area,
erosion: LS, topo erdep, SIMWE test
data will be available through these links upon request from the sponsor
due to their
large size (and security)