Browsing by Author "Thapa, Pawan Kumar"

Now showing 1 - 1 of 1

Open Access
Physically based spatially distributed rainfall runoff modelling for soil erosion estimation
(2010) Thapa, Pawan Kumar; Bárdossy, András (Prof. Dr. rer. nat. Dr.-Ing.)
Addressing different environmental and geomorphologic issues needs prediction of erosion patterns and source areas within the catchment. Several modeling alternatives exist, all with certain potential and limitations. Physically-based distributed erosion models are very much data-hungry making them of limited use in data-poor countries where erosion problem is se-verer. In addition, owing to problems like, large spatial and temporal variability of soil ero-sion phenomena and uncertainty associated with input parameter it is clear that accurate erosion prediction is still difficult and problem will not be solved by constructing even more complex models. USLE is simple but still most widely used erosion model. Its adequate ca-pability for predicting gross erosion has been proved in innumerable cases. However, the pre-diction capability has, so far, been assessed based on their ability to correctly predict lumped results at watershed outlet. The first objective of work is to investigate reliability of predicting spatial patterns of catch-ment erosion using the simple USLE-based erosion model when fed with better hydrology us-ing a physically-based spatially-distributed rainfall-runoff model (WaSiM-ETH). A small agricultural catchment (Ganspoel), located in central Belgium is chosen for investigation. The runoff and sediment yield at catchment outlet and the spatially distributed erosion within the catchment for different events have been simulated. Several results, mainly from, SCS-CN and WaSiM-ETH for erosivity computation and different algorithms for topographical factors and sediment delivery ratio (SDR) computation have been compared. Besides the predictions at outlet, the simulated spatially distributed erosion patterns and source areas have agreed rea-sonably well with the observed ones and also with the results from another physically-based more complex and data-intensive erosion model (MEFIDIS). This improved capability of simple erosion model for predicting spatial patterns of catchment erosion is extended further to devise an approach for determining spatially and temporally varying erosion risk in a big-ger Rems catchment in southern Germany. Runoff distributions are estimated from long-term simulation with WaSiM-ETH, crop cover distribution is obtained from series of MODIS-NDVI. The soil and topographical features, obtained from soil map and DEM, are considered to be temporally constant. The spatial and temporal variability hence captured through the in-tersection of Hydrologically Sensitive Areas, HSAs (from runoff simulations) and Erosion Susceptible Areas, ESAs (from geomorphic factors) yields dynamics of the erosion risk areas categorized as Critical Source Areas (CSAs). Hence, in this research work, it is shown that the dynamic behavior in hydrological sensitivity and erosion risk, estimated in such a simple ap-proach, potentially lessens landuse restrictions on landowners as the arable and agricultural fields could be prioritized for management practices by their degree of hydrological and ero-sive sensitivity. On the other hand, this research work also reveals some unreasonable consequences that have been encountered while calibrating the distributed rainfall-runoff model. From the calibration of the events in Ganspoel catchment, using Gauss-Marcquardt-Levenberg algorithm, very nice results are obtained with closely matching hydrographs and quite high NS efficiency. But a very much unrealistic patterns are observed with almost all the runoff is coming from a small isolated patch in the catchment. In Rems catchment, the model is calibrated using more accepted Shuffled-Complex-Evolution (SCE-UA) algorithm where also it is seen that the very good model performance are not accompanied by reasonable runoff patterns. A new concept, based on a statistical depth function, has been investigated further which yields not a single best parameter set but several sets of good parameter. The model performs quite well and runoff patterns within the catchment are also reasonable. But the amount of surface runoff from the different good parameter sets, when separated by using a digital filter, are found to vary highly, thus giving unacceptably different results when they are used further. The high values of spatial correlation and the rank correlation among the surface runoff from different good parameter sets prove that the patterns are uniform and reasonable but high variation in the amount raise the question mark in their quantitative reliability. These results, thus, show the very good predictions by the rainfall-runoff model but for all wrong reasons. This indi-cates that simply the better hydrograph prediction by a physically-based distributed rainfall-runoff model does not guarantee better hydrology representation by it thus making its distrib-uted results in doubt to be accepted.