Drought disaster monitoring and land use dynamics: identification of drought drivers using regression-based algorithms.

Droughts are particularly disastrous in South Africa and other arid regions that are water-scarce by nature due to low rainfall and water sources. According to some studies, droughts are not uncommon in Africa's drylands and have been rising in dry African terrain. Warm to hot summers and cool to cold winters describe the climate of the Free State Province, South Africa, a province that has been severely affected by drought events in recent times. Several studies have been carried out as regards drought prediction and mapping in arid and semi-arid areas using various models, tools and techniques. However, the use of machine learning algorithms is just emerging, especially in Sub-Saharan Africa. Studies have shown that machine learning and artificial intelligence methods have a high potential for assessment, prediction and identification of extreme events such as drought. Hence, this study aimed to evaluate drought dynamics in the Free State Province and identify drought drivers using regression-based algorithms. Results revealed that 2015 was severely affected by drought episodes as the study area observed extreme drought. More so, findings from this study showed that agricultural lands, cultivated grasslands, and barren surfaces were influenced or impacted by the drought disaster, especially in 2015, a drought year in the Free State Province. From the feature selection results, the influence of climate proxies and anthropogenic factors on VCI shows the ecological situation within the Free State Province. [ABSTRACT FROM AUTHOR]