South Africa - Using machine learning to map bioclimatic zones and crop yields in

Team The team included researchers at the University of KwaZulu-Natal, Pietermaritzburg, South Africa; the International Maize and Wheat Improvement Centre, Harare, Zim- babwe; the University of the Western Cape, Bellville, South Africa; the Water Research Commission of South Africa; the Council for Scientific and Industrial Research, Pre- toria, South Africa; and the International Water Managemen. [...] After that, a machine learning algorithm, the Classification and Regression Tree (CART) (Breiman, 2001) was used to produce a new dataset and Climate and Health: Science-based policy solutions 119 USING MACHINE LEARNING TO MAP BIOCLIMATIC ZONES AND CROP YIELDS IN WATER-SCARCE CONDITIONS Vegetation Drought Index (VegDRI) Biophysical Data Climate Data Satellite Data MODIS TERRA- Data input variable. [...] The five scales of intensity were classified as very severe drought (16%), severe drought (34%), moderate drought (38%), slight drought (11%) and no Figure 2: Average seasonal vegetation drought response index Climate and Health: Science-based policy solutions 121 USING MACHINE LEARNING TO MAP BIOCLIMATIC ZONES AND CROP YIELDS IN WATER-SCARCE CONDITIONS drought (1% of South African agricultural la. [...] Figure 3: Showing the correlation between district sorghum yields and the drought indices used in the study for the period 2010 to 2019: (a) Vegetation Drought Response Index (VegDRI), (b) Standardized precipitation Index (SPI), (c) Temperature Condition Index (TCI), and (d) Vegetation Condition Index (VCI) 122 Climate and Health: Science-based policy solutions USING MACHINE LEARNING TO MAP BIOCLI. [...] Acknowledgements The authors would like to thank the Water Research Commission of South Africa, the uMngeni Resilience Project and the Sustainable and Healthy Food Systems (SHEFS) Programme, supported by the Wellcome Trust.