Spatial Flash Flood Susceptibility Prediction in Semi Arid Area (Case Study in Jigjiga Jerer Upstream Watershed, Ethiopia)
| dc.contributor.advisor | Fiseha, Behulu (PhD.) | |
| dc.contributor.author | Wondimu, Walie | |
| dc.date.accessioned | 2021-01-13T09:50:07Z | |
| dc.date.accessioned | 2023-11-11T08:32:24Z | |
| dc.date.available | 2021-01-13T09:50:07Z | |
| dc.date.available | 2023-11-11T08:32:24Z | |
| dc.date.issued | 2019-10 | |
| dc.description.abstract | The aim of this research was to identify areas that are most susceptible to spatial flash floods. The study applied an integration of ArcGIS based spatial data analysis and logistic regression model for producing areal flash flood susceptibility prediction map from flash flood conditioning factors and historical inventory maps. The flash flood inventory map consisting a total of equal 560 flood and non flood affected locations, were extracted from areal photographs and field observations. The flood inventory data were randomly divided into two groups where 70% (392) were used for training, and the remaining 30% (168) for validation respectively for Logistic Regression Model. Eleven flash flood conditional factors such as lithology, slope, topographic wetness index (TWI), stream power index (SPI), landuse-landcover (LULC), rainfall, normalized difference vegetation index (NDVI), distance from drainage, morphometric hazard index (MHI), drainage density, and soil type, were selected. These factors were mainly derived from Shuttle Radar Topographic Mission digital elevation model (SRTM DEM), and Landsat 8 OLI/TIRS imagery using spatial analysis tools (i.e. ArcGIS 10.4 and ERDAS 2014). Finally, the flash flood susceptibility index model was produed summing up all factors multiplying with their respective weights computed in logistic regression model. Furthermore, the spatial flash flood susceptibility probability map was produced. The accuracy of the final model for spatial flash flood sucesptibility prediction was evaluated by calculating the model relative operating characteristic (ROC) curve using R studio. For validation, success and prediction rate curves were produced using area under the curve (AUC) method. The predictive capability of the model was determined from the area of under the cuve relative operating characteristic (ROC) which is found to be 0.96 (i.e. highest prediction rate of 96%). Finally, the results of spatial flash flood susceptibility prediction map were classified into five susceptiblity classes such as very low, low, moderate, high and very high clsses. The spatial flash flood susceptibility map obtained from this study could therefore assist city planners and engineers for development, landuse planning and watershed management. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/12345678/24646 | |
| dc.language.iso | en_US | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Flash flood susceptibiliy | en_US |
| dc.subject | Logistic regression model | en_US |
| dc.subject | Flash flood conditioning factors | en_US |
| dc.subject | Receiver operating characteristic curve | en_US |
| dc.subject | GIS | en_US |
| dc.title | Spatial Flash Flood Susceptibility Prediction in Semi Arid Area (Case Study in Jigjiga Jerer Upstream Watershed, Ethiopia) | en_US |
| dc.type | Thesis | en_US |