Normalized difference vegetation index prediction based on the delta downscaling method and back-propagation artificial neural network under climate change in the Sanjiangyuan region, China

Abstract

Predicting the response of vegetation to climate change through mathematical methods is an important way to understand ecosystem condition changes in ecologically vulnerable regions. We took the Sanjiangyuan region, one of the most sensitive areas to climate change, as the study area to construct a simpler calculation and higher resolution (suitable for regional scale study) nonlinear method to predict the normalized difference vegetation index (NDVI) under climate change by combining the delta downscaling method and backpropagation artificial neural network. We first used the delta downscaling method to downscale the coarse-resolution climate element data of the Coupled Model Intercomparison Project (Phase 6) (CMIP6) to 0.08333° (regional scale). By analysing the relationship between NDVI and climate elements, we found that NDVI has the highest correlation with annual total precipitation, annual mean temperature, variation range of precipitation and temperature, etc. Then, we used these impact factors to train the back propagation artificial neural network (BP-ANN) and predict the NDVI in 2030 and 2060 under the SSP1–2.6 scenario and SSP5–8.5 scenario. The simulated results show that the BP-ANN can be used to construct the nonlinear relationship between NDVI and the impact factors on different scales. In the future, NDVI will increase under both the SSP1–2.6 scenario and the SSP5–8.5 scenario. The western part of the study area has the highest altitude, the ecosystem is more vulnerable, and the changes will be the most intense. This study is expected to provide a reference for understanding the impact of climate change on vegetation in national parks in plateaus and to provide a simpler NDVI prediction method for the evaluation of environmental quality under the impact of climate change with NDVI as one of the parameters.