How well can we predict vegetation growth through the coming growing season?

Abstract

The prediction of vegetation growth under climate change has become much more important in recent years, but is still a challenge. This study developed a machine learning method to predict vegetation growth, indicated by satellite-based normalized difference vegetation index, for the coming growing season based on the meteorological forecast dataset SEAS5. First, we evaluated the accuracy of the meteorological forecast data against site-based meteorological observations in China. Air temperature, surface net radiation, and relative humidity of the monthly forecast data from SEAS5 were found to have excellent accuracy and were selected as the forcing data for predicting the vegetation index. The results show that our method is stable and highly accurate for forecasting the temporal changes of a vegetation index through the entire growing season, even under extreme drought conditions. Our results highlight that machine learning is a powerful tool for predicting vegetation indices during the growing season at the start of growing season and has the potential for improving our ability to predict crop yields and vegetation growth.