Delineating Rwandan wetland using topographic and climatic parameters

In order to establish a sustainable use of wetland in Rwanda, an accurate and up-to-date inventory and mapping of wetlands was found to be the most important baseline information. The inventory of Rwandan wetlands have been done two times during last three decades with very huge difference in the area they are covering and their defining criteria. The last was found to be inaccurate when referred to during the systematic land registration and titling and necessitated to be improved using the recently produced aerial photography. The aim of this study was to replace as far as possible the costly and time consuming imagery and/or photos interpretation techniques by an automated wetland discrimination approach combining terrain derivatives and climatic factors. Using the most original (to avoid data redundancy) and high resolution grid maps of elevation, slope gradient, contributing area, temperature and rainfall, a logistic regression model approach fitted the dependent dichotomous variable (either wetland/no wetland) with multiples predictor variables at national scale and sub-national scales (agro-ecological zones and watershed level). The sensitivity analysis of fitted models were done to fix the threshold of probability value that allow to visualize a wetland occurrence probability map reaching 95% of confidence and more than 87% of their spatial occurrence accuracy. The automated approach discriminated wetland considering the terrain complexity, hydrological and climatic conditions at national and local level. On the basis of these results, future wetland mapping, planning and monitoring activities should be based at local particularity upon at a multi-scales approach to ensure their sustainable utilization.