Mgbanyi, Liberty Lazarus Orapine
(2020)
Predicting concentrated flow erosion in data sparse regions.
PhD thesis, University of Nottingham.
Abstract
This PhD explores the possibility of using the Compound Topographic Index (CTI), which traditionally requires high-resolution data to predict concentrated flow erosion, to predict susceptibility to concentrated flow erosion in data-sparse regions. To achieve this, the PhD had to focus on a number of issues. The first is the accuracy of slopes derived from DEMs of low resolution over large topographic areas. Slope is vital input to many erosion models, including the CTI, but its accuracy is shown here to be dependent on the slope equation used and the terrain ruggedness. This was found to be problematic on high-resolution DEMS but, when using low resolution DEMs, the topography is generally much less rugged and so it is possible to use a single equation to calculate representative slopes. This slope calculation was used to determine CTI at a study site at Mount St Helens, USA, across a range of DEMs with different spatial resolutions. The results confirm earlier research in that they demonstrate that a terrain model’s ability to predict gully locations and centre-line coordinates declines with coarsening DEM resolution. In an attempt to improve model performance for low resolution data, the CTI was adapted using a fuzzy logic approach and the fuzzy-CTI model was compared to the original CTI first, at Mount St Helens, across a range of DEM resolutions, and second, using data already available from the www for a study area in South Africa. In all cases, the fuzzy logic approach showed a small improvement over the original CTI, but both still gave relatively poor predictions (~50% accuracy) of gully locations when using coarse, 30 m resolution data. It was then found that the accuracy of the fuzzy CTI could be improved in two ways. First, by setting a threshold value rather than taking any positive value as indicating that a gully would be present at a given node in the DEM. This reduced the number of false positives and improved the accuracy of fuzzy CTI to >70%. A ROC curve validation approach than allowed the most appropriate threshold value for a given study site to be determined, quantitatively. The second way that accuracy was improved was to apply the fuzzy CTI to specific environment units. This revealed that in areas with weak soil structures and/or cropland land-cover, the predictive performance of the fuzzy CTI improved further. In contrast, where soils were strong and/or forest cover was present, predictive performance was poor because gullies were often predicted where they were not observed, most likely because of the stabilising effects of current environmental conditions. It follows that areas such as these are topographically-susceptible to gully formation and could suffer serious gullying if land use were changed (e.g. from forest to cropland) and/or soil structure was weakened. Finally, the results are discussed, including key conclusions and areas for future analysis. Of particular note is the difficulty in validating a model of gully susceptibility using data on gully presence/absence because some areas may be susceptible to gullies even though gullies have not yet formed. The results are an important step toward developing a terrain model capable of accurately delineating areas susceptible using low-resolution DEMs that are available globally. However, it must be remembered that topography alone is insufficient to predict exactly where and when a gully will actually form.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
![[img]](/style/images/fileicons/application_pdf.png)