Remote sensing for mapping wetland floods in Kafue Flats, Zambia

Abstract

Monitoring huge and dynamic floodplains such as the Kafue Flats in Zambia is critical to its sustainable use. This requires among other things accurate, past and current geo-referenced flood maps. The aim of this study was, therefore, to use remotely sensed data to generate flood maps for Kafue Flats. Flood maps were created by classifying seven cloud-free Landsat images of selected months in 2001 and the areas flooded computed. Further spatial analyses were performed to integrate the flood maps to generate a multi-temporal flood map and to compute flood durations. The results indicate that 2001 was an extremely wet hydrological year with a maximum flooded area of 31%. Also, approximately 2% of study area remained flooded for six months while permanent open water bodies covered an area of approximately 4%. Statistical analysis showed that the flood extent is significantly related to water levels and discharge when water levels are high. Therefore, prediction equations were generated to estimate flood areas with respect to water flows. However, it has been revealed that rainfall within the flats has little influence on flooding. Rather, the most critical factor in controlling the flooding is water releases from the ltezhi-tezhi dam located upstream. This study has demonstrated that remote sensing is suitable for operational mapping and monitoring of tropical floodplains.