Evaluation of relations between DEM-Based USPED Model Output and Satellite-based spectral indicesAbbass Alimohammadi K.N. Toosi University of Technology Faculty of Geodesy and Geomatics Eng. alimoh_abb@yahoo.com  Sara Sheshangosht K.N. Toosi University of Technology Faculty of Geodesy and Geomatics Eng. sara_shesh@yahoo.com Mohammad-Jafar Soltani K.N. Toosi University of Technology Faculty of Geodesy and Geomatics Eng. Abstract: Climate, topography, geology and land use/cover are the most critical factors for acceleration or control of the soil erosion. Because of its key role in soil conservation and high sensitivity to human activities, land use/cover has been considered as the most important parameter. USPED (Unit Stream Power Erosion Deposition) is an erosion model which is mainly based on effective use of the capabilities of DEM together with other properties of the landscape. Understanding relations between physical properties of the landscape and land cover can be very useful for selection and formulation of the proper erosion model. In this research, relations between erosion sensitivity as defined by the DEM and geology and spectral indices extracted from remotely sensed data, have been examined. The study has been performed in Latyan Watershed located in Tehran, Iran. Erosion sensitivity has been modeled through the USPED model. Spectral indices have been calculated from multi temporal data of Landsat TM and ETM. Correlations between the USPED output and spectral indices have been examined through the multiple regression analysis. Results have shown that significant correlations exist between the spectral indices and landscape physical sensitivity to erosion as defined by the USPED. Integration of the geology layer with DEM has lead to stronger correlations. 1- Introduction: Soil erosion is considered as one of the critical environmental problems. Climate, topography, geology and land use/cover are the most critical factors for acceleration or control of the erosion. The first three factors can be classified as natural, whereas the land use factor is mainly human-based and well known as the most critical component of the erosion processes. Because of the complex interactions between the four mentioned factors and other components of the environment, the problem of erosion modeling is not an easy task. There are very limited possibilities to control the natural conditions of the environment, but proper use or misuse of the land can act as an efficient tool for sustainable use of the land resources or their degradation (Baker et al., 2005; Carlson and Arthur, 2000). In a given climatic conditions and in a watershed scale, topography and geology are the main natural factors for definition of the susceptibility of a land for erosion. In this scale, topography provides additional information about the microclimatic conditions of the landscape. It has been demonstrated in many studies that natural land covers provide an optimum situation for protection and sustainability of the Land resources. Because of the grazing, mining, development projects and human activities many natural land covers are degraded or replaced by the man-made or man-induced land cover types. Therefore change and degradation of the natural land covers are the most important drivers of the soil erosion, and research for modeling the effects of land use change on soil erosion is very important for sustainable development. Understanding the relations between topographic and geologic susceptibility for erosion and present land cover is the first step towards modeling the erosion. The main hypothesis here is that land cover, topography and geology are not independent but they are correlated. Remotely sensed data provide useful information and quantitative indices about the properties of the actual land covers. In fact remotely sensed data and digital elevation model (DEM) play very unique roles in many modeling tasks. The latter is a rich source of land covers data, whereas the former is capable of providing many physical parameters of the landscape by careful processing and analysis (e.g. Moore et al., 1991). USPED model provide straightforward techniques for evaluation and integration of the topographic and geologic susceptibility of a land in a unique value which is an indicator of the erosion (in erosion-prone areas) or deposition. DEM is a basic layer in this model and effective use and processing of the DEM data is an important advantage of the USPED model. Although the main objective of our research has been to model the effects of land use change on the soil erosion, in this paper part of our work for test of hypothesis for examination and formulation of relations between the USPED model output and Landsat data-driven indices has been presented. First a brief introduction of the USPED has been provided and then methods and results of the research are presented. 2- USPED model The first universal model for estimating erosion and sedimentation for agriculture lands was USLE (Wischmaier and Smith, 1941). Thereafter this model has been modified in 1972 and has been widely used all over the world. The Unit Stream Power Erosion/Deposition (USPED) model (Mitasova et al., 1996 and Mitasova and Mitas, 1999) is simple and predicts the spatial distribution of erosion /deposition as the divergence of sediment flow under the steady state conditions (Sreenivas Kandarika and R.S. Dwivedi, 2003). According to the popular RUSLE model the annual soil loss (E) is given by equation(1): E=R*K*C*P*LS equation (1) Where R is the rainfall erosivity index, K is the soil erodibility index, C and P are the soil cover and management factors respectively, and LS is the slope and slope length factor. Many corrections have been suggested to account for landscape scale effects. The most important one has been to use the flow accumulation (drainage area) instead of the slope length to compute the topographic factor LS (Mitasova and Mitas 1996). In this way, the index E of equation (1) is reinterpreted as a “transport capacity function” T: (Pistocchi A., Cassani G.and Zani O., 2002) T = R K C P Am (sin b)n equation (2) Where A is the drained area and b is slope angle, m and n are empirical coefficients whose values depend on the kind of erosion consideration and m=1.6, n=1.3 for prevailing rill erosion while m=n=1 for prevailing sheet erosion. Then the net erosion/deposition is estimated by computing the divergence of transport capacity as (Sreenivas Kandarika and R.S. Dwivedi, 2003): ED = div (T) = d (T*cos a )/dx + d (T*sin a )/dy equation (3) Where a is aspect angle of terrain surface. 3- Materials and Methods The study area is part of Latyan Watershed located in Tehran province, between the longitude 35’35 ° till 35’52° north, and latitude 51’26° till 51’ 40° east and covers an area of 9005.579 hectares.  Figure 1- Position of the study area in Latyan watershed (right) and IRAN (left). The following data were used to develop data base for the study:
 Figure2: DEM-based estimated erosion and deposition by USPED model. In the second test, geologic erosivity layer as calculated from the geologic units of the geology map was incorporated in the USPED model. Because of the lack of a reliable soil map of the area, erosivity of the geologic units was used as a replacement for the K (soil erosivity) factor (Feizneia, S. 1995).  In both figures of 2 and 3, high erosion areas are closer to ridges and deposition zones are mostly concentrated in stream sides.  Figure3: DEM and Geology-based estimated erosion and deposition by USPED model. NDVI, normalized difference bands 5 and 4 termed as soil index and also normalized difference of bands 6 and 7 termed as ND67 were calculated for each of ETM and TM data and NDVI of MSS images were also calculated. NDVI, soil and ND67 layers were classified according to erosion and deposition classes. Correlations of both DEM-based and DEM and Geology-based erosion/deposition with those of image-based indices (NDVI, soil and ND67) and bands of Landsat-MSS were examined by single and multiple-regression analysis. Image bands of Landsat MSS, and the soil index as derived from Landsat ETM and TM did not show a significant relation with the DEM-based estimation of erosion/deposition. Therefore, they were not considered in the multiple regression analysis. The fitted equations and coefficients of determination resulting from this analysis are summarized in table 2.  In the case of erosion zones, usefulness and important role of the geology layer is confirmed by the significant increase in the R values of the regression analysis as compared to those of DEM-based models. Examination of the relations between predicted and actual USPED outputs by using the regression equations shows reasonable correlations between the two (Figure 3 and table 3).  Figure4- Correlations between the USPED output and estimated values by spectral indices as in regression equations in table 2).  The observed results in the case of two independent data sets of landscape characteristics (DEM and geology) and satellite data can be used as a useful two-way tool for validation of the deposition and erosion zones and values produced by the USPED. Simultaneously it can be used for examination and proper weighting of the image-driven spectral indices by their checking against the USPED output values. The high correlations between the output of USPED and spectral indices can be used as valuable indicators for examination of the usefulness of the erosion model, the input parameters and also tests of hypothesis. By consideration of the fact that the erosion sensitivity of physical properties of the landscape is mostly related to theoretic terms and potentials for erosion. Whereas spectral indices of soil, vegetation and temperature are mostly affected by the actual erosion. Therefore, observation of close correlations between the erosion sensitivity of physical properties of the landscape and spectral indices demonstrates the fact that most parts of the study area has been highly degraded and eroded and minimal conservation measures have been performed. References:
|