Machine Learning-Based Modeling of Spatio-Temporally Varying Responses of Coffee Production to Climate Change: A Case Study of the Northern Region of Thailand

The Intergovernmental Panel on Climate Change (CC) reports indicate that CC will have detrimental effects on coffee production, leading to reduced global yields and a decrease in suitable land for coffee cultivation by 2050. Coffee holds economic significance as a cash crop in Thailand. Changes in rainfall patterns, rising temperatures, and other climatic variables can harm coffee plants. Therefore, it is essential to understand the relationship between climatic variables and coffee yield. Developing an in-depth grasp of the changes in coffee yield is crucial for evaluating the vulnerability and adaptability of coffee production. This study involved a comprehensive data-driven analysis of the five coffee-producing provinces in Northern Thailand. The objective was to examine and model the impact of climate variability on rainfed coffee yield. Machine learning can potentially develop and understand these relations; we employed artificial neural network (ANN), support vector regression (SVR), and regular Ordinary Least Squares (OLS) regression model to estimate the correlation. Results revealed the predominant effects of climate average and extreme conditions on coffee yield. The OLS regression model demonstrated a good fit (R2=0.81). The climatic variables exhibited statistically significant relationships with the coffee yield. These variables have a substantial impact on the prediction of coffee yield. The findings highlight the significant influence of various climatic and environmental factors on the prediction of coffee yield. These results provide valuable insights into the relationship between these climatic variables and the coffee yield, contributing to our understanding of the factors influencing yield. Lastly, two machine learning algorithms (SVR and ANN) were employed to predict coffee yield using climate variables as predictors. The ANN model demonstrated superior performance (R2= 0.84) to the SVR model regarding in-season prediction skills. This study explored the relationship between climate variability and rainfed coffee yield in the northern region of Thailand, considering the potential impacts of climate change. The analysis revealed significant associations between climatic variables and coffee yield by employing OLS regression. Moreover, SVR and ANN were utilized to predict coffee yields based on climate variables. Thailand’s geographical advantages and efficient coffee production processes establish it as a potential regional hub for coffee production, despite its lower output than neighboring ASEAN countries such as Vietnam and Indonesia.