There has been an increase in death as a result of drought worldwide. A good drought early warning system cannot actually stop the drought, but can help in reducing the adverse impacts of drought on human life, livestock, and nature. Most of the consequences that are faced due to drought particularly for Kyankwanzi district are as a result of food shortage and water. However, a good drought early warning system could guide farmers and policy makers in preparation for this hazard for example by changing the types of crops grown particularly throughout the drought period to enable them sustain themselves throughout the drought. The main objective of this study was therefore to assess the performance of using the Vegetation Health Index (VHI) in forecasting drought with Kyankwanzi district being the case study. In this study, Landsat images (Landsat8 OLI) from 2015 to 2018 were used in carrying out the analysis. The time period was divided into four equal quarters each year and used to generate actual VHI (Vegetation Health Index) values which were classified according to drought classes for all quarters. This actual data was used in a time series model to generate predicted VHI values using the seasonal and trend components for all quarters but in addition the first quarter of 2018 (January to March). The predicted VHI values were critically analysed using graphs to generate a correlation that was the basis of conclusions and recommendations. Results show that the first quarter of 2018 had a VHI value of 38.04% which according to drought classes developed by (Kogan, 2002) indicate a mild drought for Kyankwanzi district. The true value computed from the satellite images was 37.81%. A correlation factor was however generated to make the relationship between predicted values over the years and actual values more understandable generating a value of +0.763. This correlation factor showed a strong positive correlation between predicted and actual VHI values which implied that the method was a very accurate and reliable one.

show more