Predicting the Start of the Growing Season in Boreal Forest Under High and Low Emission Scenarios

Abstract

The impact of global climate change on ecosystems has become increasingly pronounced, particularly with global warming leading to the earlier of the Start of the Growing Season (SOS). However, changes in SOS under future climate scenarios remain unclear. Therefore, this study uses remote sensing-based SOS data sets and bio-climatic variables to develop pixel-level SOS simulation models through machine learning methods. Future SOS predictions for boreal forest regions are made using climate data from four emission scenarios: SSP126, SSP245, SSP370, and SSP585. The results show that two machine learning models exhibit good simulation performance across the study area, with the RMSE for most pixels controlled within 9 days. Furthermore, predictions of future SOS based on these two models suggest that under all four emission scenarios, the SOS in boreal forest regions shows a significant advancing trend. Notably, as emission levels increase, the advancing trend in SOS becomes more pronounced. However, there are variations in the trends observed for different vegetation types. Our findings emphasize that the advancing trend in SOS differs under various emission scenarios and exhibits distinct vegetation type-specific and spatial distribution patterns. These changes will have profound implications for biodiversity and ecosystem stability.