REMOTE SENSING AND GIS-BASED FLOOD PREDICTION IN RIVER JAMA’ARE CATCHMENT, NORTHEASTERN NIGERIA

Abstract

Flood is one of the common and frequent natural disaster which causes loss of life and damage to property. River Jama’are which originates from Jos plateau frequently floods its basin such as in 1986, 2011, 2012, 2013, and 2015. Remote sensing and GIS was used in characterising Land Use/Land Cover of the catchment and in delineating the 2015 flood extent. Using Fuzzy logic approach, water level for Foggo station was predicted. Secondary data such as rainfall, runoff, soil, were collected from relevant agencies. LandSat imageries (1986, 2012, and 2015), DEM, and MODIS were downloaded and analysed in this work. The extent of River Jama’are catchment was delineated using DEM which serves as the areal extent of the study area. LandSat imageries were pre-processed using the Top of Atmosphere method to convert the DN value to the true measure of their reflectance and the imageries were subsequently classified using support vector machine algorithm. Error matrix technique was used to assess the classification accuracy. The analysis involve calculating the area and percentage of each LULC class for each year and subsequent comparison of the result. It is found out that built-up that accounted for 4% in 1986 increased drastically to 42% in 2015. Vegetation decreased from 80% in 1986 to 50% in 2015 while agricultural land decreased from 10% to 5% from 1986- 2015. MODIS data of the study area before and during the flood was downloaded and analysed using Land Surface Water Index (LSWI) method in order to delineate the spatial extent of the 2015 flood. It was found that the river flooded the northern part of the catchment in August with vegetation and built-up being the most affected LULC. In September, the river flooded both the northern and southern part of the catchment with vegetation and built-up being the most affected LULC category accounting for 60% and 33% of the inundated LULC. For the prediction of flood, input parameters were generated using GIS and converted to ASCII format. They were assigned a triangular membership function ranging from 0 to 1. The fifteen rules generated were combined to predict the water level at Foggo station. It is discovered that Mamdani fuzzy inference system is more efficient in the prediction of water level than Takagi- Sugeno fuzzy inference system in River Jama’are catchment with an average RMSE and R2 of 0.05 and 0.92 while Takagi-Sugeno fuzzy inference system shows an average RMSE and R2 of 0.08 and 0.97 respectively. Therefore fuzzy logic model can be used as a method for predicting water level in River Jama’are catchment. This is because it is simple and user friendly and has the ability to model non-linear dynamics present in the data.