Troubleshooting

• Slope map has contour-like pattern.
  An artificial, contour-like pattern in the slope map is usually caused by insufficient vertical resolution of a DEM (most often integer meters) or inadequate interpolation from contour data. The problem is typical for areas with relatively flat terrain and becomes more severe with higher resolution.
  

  A DEM with elevation values given as integer meters produces a slope map as shown in this image, where areas of zero slope are combined with steeper areas along 1m contours. The zero slope areas are caused by insufficient vertical resolution and will result in underestimation of erosion (see the tables shown for the "Large areas with zero erosion" section). The best solution is to find a more accurate DEM, preferably with centimeter vertical precision. If it is not available, the DEM should be reinterpolated either from contours generated from the DEM or random samples with sufficient smoothing. Note, that the contours generated from a DEM have very dense sampling compared to manually digitized ones. Reducing the density of points reduces the problems with the contour-like pattern in the slope maps.
  
  A DEM interpolated from dense points along contour lines may generate a slope map such as this, which overestimates flat areas and contains lines along the contours of artificial drops, forming terraces. This does not represent the natural landscape, and will usually result in underestimation of sheet erosion because flat areas are overestimated. The DEM should be reinterpolated using different parameters or a different method. When using splines, lower tension and higher smoothing will reduce waves along contours.
  
  Using smoothing techniques during interpolation of the DEM and thinning the sample points along the contour will reduce the terracing effect somewhat. The best DEMs are produced not from contour data, but from more evenly spaced sampling methods such as IFSARE (for areas without vegetation) or LIDAR, or from contours combined with breaklines, surface specific points (peaks, depressions) and mass points.
  

• Erosion rates are unreasonably high.
  First, check whether the computation was done with correct units. A common mistake is to compute slope from a DEM with horizontal coordinates in meters while elevation is in feet. If that is the case, multiply your elevation by 0.3048 using map algebra and recompute the slope and erosion with the new slope values.
  

  Lower exponents for m and n may be needed. The empirical constants used in the models were derived from experimental plots that were much shorter than the length of slopes commonly observed at landscape scale. While the exponents m=0.4 and n=1.3 should give reasonable results for RUSLE3D, it is useful to have a few sample areas with field observations of the erosion rates so that these exponents can be adjusted to the flow type typical for the study area.
  

  Finally, erosion models used in this tutorial were originally designed for sheet and rill erosion caused by shallow overland flow. The implementation of upslope contributing area enables application of these models in complex terrain where concentrated flow is quite common in hollows and valleys. Erosion rates in these areas is much higher and exceed the typical values obtained by RUSLE or USLE. These higher erosion rates may be realistic, however, extensive field testing similar to those done for USLE is not available to verify the values. Due to this uncertainty, especially if it is impossible to verify the values in the field, it is recommended that erosion rates predicted for concentrated flow areas are reclassified to values expected for gullies in the studied area. This reclass procedure affects only a very small number of grid cells and has little impact on erosion averages for the entire region. However, it is important to keep these areas in the maps and reports, as they can contribute substantial amount of sediment to the total loads and should be the primary targets for erosion prevention and conservation.

• Erosion map appears uniform although the values vary.
  Due to the non-linear distribution of the erosion rate values, with the lower values covering most of the area and higher values covering very small areas,  most default color tables available in GIS will be insufficient for visualizing the results. Customized color tables are needed to create meaningful maps. You may handle the colors as follows:

  When you compute RUSLE3D you may get a map like this if you use the default color table
  You may try to change to one of the color tables provided by GIS, for example, a red/yellow/green ramp, however, the result won't be much better because the colors are evenly spread between the minimum and maximum erosion, but most of the spatial variability is at the low range of values (e.g. 0-20).
  A better solution is to reclassify the map so that the large interval of high values is merged into a single class, for example, like in this table. Then you can apply the standard red/yellow/green ramp and areas with high erosion will pop-up in orange. However, your map now only has 6 values and some spatial variability has been lost.
  The best results are obtained when the color table is defined manually (see suggested color tables for RUSLE3D and USPED)
  The  image to the right illustrates the difference between a linear and non-linear color table, where the upper legend is the RYG ramp and the lower legend is the one used for our RUSLE3D results:

• Flow accumulation map has many straight lines.

  This problem is restricted to the use of ArcView or ArcGIS flowacc module for computation of flow accumulation and is a consequence of  the D8 flow routing algorithm that moves water only in 8 discrete directions.
  The problem can be reduced by increasing the grid cell size (decreasing resolution), however, some detail will be lost. A better, although more time consuming, solution is to use one of the ArcView flow tracing or hydrologic modeling extensions with D-infinite or multiple flow paths routing algorithm.
  Also the most recent version of GRASS5 with easy to install binaries for Linux, Mac OS X and Windows can be used. You can import the DEM into GRASS, run the command r.flow, and export the resulting flow accumulation map. The exchange of raster files between the two systems is quite easy.

• LS factor has values over 100.
  This problem is due to the use of upslope area and incorporation of concentrated flow. It should be kept as is, and resolved after the entire computation is complete as suggested above in the section about the "Erosion rates are unreasonably high".

• Large areas with zero erosion.
  This is caused by the zero slope in DEM with integer meter resolution as described above. To resolve this problem, the DEM should be reinterpolated. Important note: this problem affects not only the models covered by this tutorial but all models where slope is estimated from a DEM . Lowering the resolution reduces (but does not eliminate) the problem and leads to loss of detail (information).

  Compare the erosion estimates derived from RUSLE3d when using the DEM with integer meter resolution (67% of the area is stable) and with a reinterpolated and smoothed DEM (only about 22% of the area is stable):

  DEM given in integer meters
  |----------------------------------------------------------|
  |           Category Information         |   %  |          |
  |         #|description                  | cover|     acres|
  |----------------------------------------------------------|
  |-2010--100|extreme. . . . . . . . . . . |  0.38|   6.27117|
  |  -100--50|severe . . . . . . . . . . . |  1.18|  19.36020|
  |   -50--20|high . . . . . . . . . . . . |  2.65|  43.48044|
  |    -20--5|moderate . . . . . . . . . . | 10.04| 164.77722|
  |     -5--1|low. . . . . . . . . . . . . | 16.11| 264.42489|
  |      -1-0|stable . . . . . . . . . . . | 67.79|1112.87790|
  |         *|no data. . . . . . . . . . . |  1.86|  30.51364|
  |----------------------------------------------------------|
  |TOTAL                                   |100.00|1641.70546|
  +----------------------------------------------------------+

  Average soil loss is 3.7 tons/acre year which may be still acceptable.

  |----------------------------------------------------------|
  |           Category Information         |   %  |          |
  |         #|description                  | cover|     acres|
  |----------------------------------------------------------|
  |-2010--100|extreme. . . . . . . . . . . |  0.49|   8.03119|
  |  -100--50|severe . . . . . . . . . . . |  2.46|  40.38263|
  |   -50--20|high . . . . . . . . . . . . |  5.42|  88.98757|
  |    -20--5|moderate . . . . . . . . . . | 31.28| 513.58741|
  |     -5--1|low. . . . . . . . . . . . . | 36.35| 596.79047|
  |      -1-0|stable . . . . . . . . . . . | 21.74| 356.83027|
  |         *|no data. . . . . . . . . . . |  2.26|  37.09593|
  |----------------------------------------------------------|
  |TOTAL                                   |100.00|1641.70546|
  +----------------------------------------------------------+

  Average soil loss is 8.6 tons/acre per year indicating an unsustainable level of erosion.