Assessing the probability of land submergence for lowland rice cultivation in Africa using satellite imagery and geospatial data

Abstract

Sub-Saharan African countries are being strongly urged to enhance their rice production, because their rice consumption and importation rates have been rapidly increasing in recent years. Areas planted to rice in Africa are classified agro-ecologically into rainfed upland, rainfed lowland, and irrigated. Rainfed lowland includes extensive areas of unexploited land that has great potential for the promotion of rice growing. For the unexploited rainfed lowlands of Ghana, we have been studying the development of low-cost rice-farming systems that require no large-scale irrigation or land reclamation. For such systems, it is important to select suitable areas where water for rice farming can be obtained naturally; floodwaters offer promise for this purpose. Delineation and mapping of floodwater prone areas suitable for rice production is important for successful utilization of this land resource. Here, we propose a method of assessing flood probability from submergence frequency, as estimated from satellite imagery and geospatial data. ALOS/PALSAR images acquired in May, June, August, and September 2010 were used to classify land and water, and then a submerged-area map was produced. From the results, we were able to accurately detect non-submerged areas and submerged areas with water depths of at least 3 cm. The number of times classified into submerged area was defined as submergence frequency, and it was approximated by distance from reservoirs representing White Volta River, ponds, and swamps. In addition, flood extent derived from reservoirs was simulated using digital elevation model (DEM). Finally, a flood probability assessment map was produced by integration of the estimated submergence frequency and flood extent simulation. The results of a comparison of soil moisture data measured at 69 points in the field and the NDVIs computed by ALOS/AVNIR showed that areas with high potential for flooding retained high levels of soil moisture and were more likely to show less deterioration of vegetation in the dry season. The validation of these results confirmed the adequacy of the flood probability assessment method.