CLIMATIC EFFECTS ON ARGENTINA’S TRADITIONAL DRINK – A MACHINE LEARNING BASED YERBA MATE PRODUCTIVITY PREDICTION
Keywords: agricultural production, machine learning, regression analysis, climate indices, climate change, El Niño Southern Oscillation
Abstract. Yerba mate belongs to the most important agricultural products of Argentina. Due to climate change, together with El Niño and droughts, yields are negatively affected. Drought propagation from precipitation deficits to plant water stress, its impacts and especially its predictability are becoming an emerging field of research. This paper explores how the future yerba mate production in the states of Misiones and Corrientes, north-eastern Argentina, can be projected by using climate variables and satellite data. The applied methodology focuses on a machine learning approach based on multiple linear regression and random forest regression. The results indicate a significant relationship between yerba mate and the NDVI as well as the SOI. The highest yerba mate productivity is expected during weak La Niña events. The methodology of this analysis can successfully predict mate productivity based on satellite and climate data and can also easily be used for further research areas.