Responsible AI for Sustainable Agriculture: Forecasting Rice Yield to Combat Global Food Insecurity
Climate change has emerged as a significant contributor to acute hunger, and its impact is projected to worsen with the exacerbation of key climatic drivers, including temperature rise, changing precipitation patterns, increasing frequency of extreme events, and rising sea levels. Along with man-made conflicts and economic downturns, climate change is among the primary causal factors of hunger, which necessitates comprehensive and innovative approaches for mitigation. In this context, EY Open Data Science Challenge was undertaken to develop models that enable scientists to better comprehend the implications of climate change on crop yields and forecast rice cultivation yield across Vietnam's identified areas. Microsoft's Planetary Computer was leveraged for the challenge, and it involved Sentinel-1 and Sentinel-2 data, along with temperature data from open-source packages. The study constructed new features that integrate radar, optical, and temperature data to investigate their influence on vegetation growth and employed machine learning to establish the interpretability of the rice yield prediction. The Light GBM model achieved the highest R2 score of 0.65 on the submission data among all the models tested.
Paper
Full text
Responsible AI for Sustainable Agriculture: Forecasting Rice Yield to Combat Global Food Insecurity
Semantic Scholar · Agricultural and Food Sciences · 2023
Abstract
Climate change has emerged as a significant contributor to acute hunger, and its impact is projected to worsen with the exacerbation of key climatic drivers, including temperature rise, changing precipitation patterns, increasing frequency of extreme events, and rising sea levels. Along with man-made conflicts and economic downturns, climate change is among the primary causal factors of hunger, which necessitates comprehensive and innovative approaches for mitigation. In this context, EY Open Data Science Challenge was undertaken to develop models that enable scientists to better comprehend the implications of climate change on crop yields and forecast rice cultivation yield across Vietnam's identified areas. Microsoft's Planetary Computer was leveraged for the challenge, and it involved Sentinel-1 and Sentinel-2 data, along with temperature data from open-source packages. The study constructed new features that integrate radar, optical, and temperature data to investigate their influence on vegetation growth and employed machine learning to establish the interpretability of the rice yield prediction. The Light GBM model achieved the highest R2 score of 0.65 on the submission data among all the models tested.