Abstract:
Rivers are natural forces that continue to shape the surface of the earth and also play an important role in supporting life. They serve as part of the earth’s interconnected circulatory system. This system acts on the rivers to alter topography and shape of river basins and nearby landmasses which influence river currents. Earth observation tools have developed over the years with enhanced and advanced technological input to also observe such changes. The White Volta has been affected by various factors both natural and anthropogenic activities causing some changes within its basin. But methods to extract such features for observation are now numerous. In order to know the effective method in semi-automatic riverine feature extraction, the study compared the pixel based, object based and the decision tree image classification methods conducted on remotely sensed data of the White Volta. The Landsat 8 OLI image data was used for the classification in ENVI software. The Maximum Likelihood Classification (M LC) had a kappa coefficient of 0.9492, the Decision Tree Classification (DTC) had a kappa of 0.9491 and the Object Based Classification (OBC) had a kappa index of agreement of 0.9505. the total accuracies were what differentiated them all vastly with the MLC, DTC and the OBC attaining 98.48%, 99.9% and 97.37% respectfully. The study however found the OBC to be a preferred choice because it was able to overcome the deficiencies of the other two methods and still attain a competitive accuracy level. The study therefore recommended the OBC for the choicest in the event that there was the need for a choice but also acknowledged the efficiency of the remaining two methods.
