nbul 2004
|]
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
3.1.3 Topographic factor (LS):
Four 1:50,000 topographic map sheets were
digitized. A DEM was generated using TIN and
GRID Modules. The resulting layer was exported to
ERDAS Imagine to generate a slope layer. The slope
image was classified into six classes using Model
Maker. Slope length was calculated by combining the
slope angle thematic layer with a hydrological layer
containing the wadi network for the study area.
The LS factor was calculated using the tables
produced by Renard et al. (1994) for the RUSLE in
conjunction with the thematic slope layer. The
results of these operations are presented in Table 2.
Slope 0-3 3.1-6 6.1- 10.1- | 151-1 725.
class (9a) 10 15 25 |
Mean L 177. 71.6 50.0 51.6 56.7 59.0
(m) 16 2 | 7 5 3
Mean LS, | 0.29 0.78 1.28 2.30 4.39 10.0
rangeland 0
Mean LS, | 0.37 0.93 1.48 2.69 5.04 11.8
cropland 6
Table 2: The estimation of LS values for each class
present in the study area, for each type of land use.
3.1.4 Vegetation/Cropping and Management
Practices Factors (C and P): The
vegetation/cropping — C — factor was parameterized
from remotely sensed data. The fractional vegetation
cover map for the 1992 TM imagery produced by
linear spectral unmixing was used to estimate the
values of C for rangelands and croplands in the 1992
dry season. Calibrated NDVI images were used to
estimate C from similar areas in 1972.
In these estimations three cultivation systems (rain-
fed fields, irrigated fields and rangeland) were
considered. It was considered that the residual plant
remains (stubble) would be very low for rain-fed
fields due to low yields and stubble grazing. At 30%
cover of mulch the estimated C value is 0.4. For the
irrigated fields and rangeland the C values are
presented in Table 3.
Type of < < < <
land use 20% 40% 60% 80% 100%
Irrigated | 0.48 0.37 0.22 0.12 0.04
fields
Rangela 0.35 0.20 0.12 0.062 | 0.027
nd
Table 3: Estimation of C values for vegetated fields
and rangelands by vegetation cover percentage classes.’
A close inspection of the NDVI statistics revealed
that for the dry season 1972 the NDVI values never
exceeded 0.087. According to the calibration data,
this indicates that the vegetation cover percentages
never exceeded 20% (Edwards et al. 1996).
Moreover, as 20% is the lowest threshold value for
the estimation of C, a constant value for C (= 0.35)
was applied to the rangelands and bare fields in the
1972 image, and this value was used in the soil loss
model (Wischmeier 1978).
The management factor (P factor) was assigned a
value of one for the entire study area as no specific
507
management measures are used for either rangelands
or croplands in the study area.
3.2 Maps of Soil Losses
The GIS input layers discussed are listed in Table 4.
They were combined, as described by the RUSLE, to
estimate annual soil losses on a pixel-by-pixel basis.
A low pass (7 * 7 Kernel) filter was applied to all
input layers before running the model.
Layer Description
number
1 Rainfall erosivity layer (R).
2 Soil erodibility layer (K).
3 Topographic layer (LSc) for
cropland.
4 Topographic layer (LSg) for
rangeland.
3 Vegetation cover layer (Co?)
for bare fields, for 1992.
6 Vegetation cover layer (Cs)
for vegetated fields, for
1992.
7 Vegetation cover layer (Cop)
for rangeland, for 1992.
8 Vegetation cover layer (C7)
for bare fields and
rangeland, for 1972.
Table 4: Thematic input layers to the GIS-based
erosion model.
Change in vegetation cover proportions is the key
dynamic variable in predicting soil losses over time in
the study area. It was assumed that all other variables
were constant over the 20 year time period of the
study. To capture this change, the soil loss model was
run separately for the years (1972 and 1992). These
runs were based on the following combinations of
GIS layers (layers refer to Table 4):
Soil loss maps for 1972:
layer 1 x layer 2 x layer 3 x layer 8
layer 1 x layer 2 x layer 4 x layer 8
Soil loss map for 1992:
-. layer 1 x layer 2 x layer 3 x layer 5
layer 1 x layer 2 x layer 3 x layer 6
layer 1 x layer 2 x layer 4 x layer 7
The results of the model runs, as soil loss maps for
each year are shown in Figures 2 and 3. These maps
were classified using Model Maker, and 11 soil loss
classes were produced.