Predicting future US crop phenology dynamics under climate change using an integrated remote sensing and machine learning approach
Tuesday, August 3, 2021
Link To Share This Presentation: https://cdmcd.co/A9P3xZ
Yanjun Yang and Wei Ren, University of Kentucky, Lexington, KY, Bo Tao and Yawen Huang, Department of Plant & Soil Sciences, College of Agriculture, Food and Environment, University of Kentucky, Lexington, KY
University of Kentucky Lexington, Kentucky, United States
Background/Question/Methods Crop phenology provides essential information for monitoring and modeling crop growth dynamics and predicting crop production. All currently available climate models predict that a near-surface warming trend under the influence of rising levels of greenhouse gases in the atmosphere and this changing trend will even accelerate in the future. However, it remains unclear how crop phenology would shift under future climate scenarios. This study combines remote sensing imagery, climate projections, and a machine learning approach to predict future crop phenology across the Contiguous United States. We estimated potential shifts in corn and soybean phenology and climate impacts during 2021-2099 under the RCP4.5 and RCP8.5 scenarios from ten CMIP5 GCMs (Coupled Model Intercomparison Project 5, General Climate Models). Results/Conclusions Our estimated historical changes in crop phenology were consistent with those from ground observations, remote sensing, and the USDA survey data. Trend analysis showed that the temperature would continue to increase across the US during crop growing season under two climate scenarios. Our results predicted a shift to earlier planting and harvesting dates and shortened growing season length for corn during the 2021–2099 period. Future crop phenology dynamics would be more sensitive to climate temperature than precipitation. The future crop phenology data derived from machine learning in combination with remote sensing is beneficial for agricultural assessments and earth-system modeling over large areas.