Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Pays Bas → 2013 → Mapping of flash flood potential areas in the Western Kape, South Africa, using remote sensing and in Situ data

International Institute for Geo-Information Science and Earth Observation (ITC) 2013

Mapping of flash flood potential areas in the Western Kape, South Africa, using remote sensing and in Situ data

Bangira, T (Tsitsi)

Titre : Mapping of flash flood potential areas in the Western Kape, South Africa, using remote sensing and in Situ data.

Auteur : Bangira, T (Tsitsi)

Etablissement de soutenance : University of Twente International Institute for Geo-Information Science and Earth Observation (ITC)

Grade : Master of Science in Geo-Information Science and Earth Observation 2013

Résumé
Flash floods are caused by large amounts of runoff from short duration and high intensity rainfall. Besides high rainfall intensity being the major cause for flash flood, this study considers flash flood as a hydro meteorological problem. Although high rainfall intensity was evaluated as the major cause of flash flooding in different literature, the run off generation as a result of causative rainfall event for flash floods is contributed by hydro geomorphic characteristics of catchments. High spatial and temporal resolution remote sensing rainfall products tested in this study are MPE, CMORPH and TRMM 3B42.The aim of the study was to evaluate flash flood potential areas in the western part of South Africa by integrating remote sensing products of high rainfall intensity, antecedent soil moisture and topographic wetness index based on assigning different weights to each layer. Rainfall has high spatial and temporal variability, thus needs to be quantified at an area scale in real time from remote sensing unlike from sparsely distributed, point gauge network measurements. Western part of South Africa was found to have high spatial variation in topography which results in major differences in received rainfall within areas not far from each other. Satellite rainfall products were statistically compared with local gauge measurements at 3 and 24 hourly time scales. TRMM 3B42 was found to overestimate whereas CMORPH and MPE underestimate extreme rainfall events at both time scales. At 3 hourly time step, bias for TRMM 3B42, MPE and CMORPH was found to be 0.19, -0.96 and -0.97 respectively. However, the accuracy of all satellite rainfall estimates in form of bias, correlation coefficient and RMSE was found to improve with increasing time step of analysis. Satellite products are susceptible to systematic errors ; they have to be calibrated and validated with local in situ data before further applications. Cumulative satellite rainfall estimates for two flash flood events that occurred on 12 July 2009 and 11 August 2012 in Berg and Breede catchments respectively were calibrated for bias and systematic errors by using quadratic curve fitting method. Quadratic curve fitting method improves accuracy by lowering bias an d RMSE and increasing the correlation coefficient for areal rainfall quantification. RMSE for TRMM 3B42 and MPE for 12 July 2009 in Berg catchment improve from 29.54 mm/3 hours to 8.15 mm/3 hours and 34 to 15.73 mm/3 hours respectively. Corrected satellite rainfall estimates were used to calculate rain fall anomaly using daily climatological satellite rainfall product RFE ARC2 data . Areas of high flash flood potential were found to be associated with high rainfall, antecedent precipitation and topographic wetness index. Although, TRMM 3B42 was found to have better accuracy, the product is not available in near real time but rather at a rolling archive of 3 months therefore, MPE rainfall estimates available in near real time are opted for flash flood events. ASCAT soil moisture observations were found to have low RMSE and MAE of 4.18 m 3 /m 3 and 3.8 respectively when validated with in situ soil moisture measurements. Hydro geo-morphometric catchment properties associated with flash flood potential such as relief ratio, drainage density, basin shape, form factor and bifurcation ratio among others were discussed. Relief ratio was found to be 0.18 for Berg and 0.12 for Breede.

Mots clés : Topographic Wetness Index, satellite rainfall estimates, gauge comparison, antecedent soil moisture, hydro geomorphic, satellite rainfall estimates, flash flood, in situ and remote sensing

Version intégrale (ITC)

Page publiée le 2 avril 2018