4.Spatial distribution of the dominant drivers on vegetation growth across China over the past three decades

4. Spatial distribution of the dominant drivers on vegetation growth across China over the past three decadesa. Spatial distribution of the partial correlation coefficient (Spearman) between the drivers and the Normalized Difference Vegetation Index (NDVI)/Leaf Area Index (LAI)/Gross Primary Productivity (GPP) across China over the 1982–2015 periodThe drivers include air temperature (Tmp), precipitation (Pre), vapor pressure deficit (VPD), downward surface shortwave radiation (Rad), soil moisture (SM), and atmospheric carbon dioxide (CO2). We used the 25 10-year moving windows over the 1982–2015 period to reduce the uncertainty associated with inter-annual anomalies. We calculated the partial correlation coefficient (Spearman) for each grid cell for the 25 10-year moving windows. The factor with the highest absolute value of the significant partial correlation coefficient (two-tailed t-test: p < 0.05) was identified as the dominant driver of changes in vegetation growth in that grid cell. You can use the "China_vegetation_cover_map.tif" in the dataset "1.Spatiotemporal patterns of vegetation growth across China over the past three decades" to extract the part of China.b. The MATLAB (R2023a) code for calculating the partial correlation coefficient (Spearman) for each grid cell for the 25 10-year moving windowsThis includes "Partial_Correlation_Spearman.m" for calculating the partial correlation coefficient (Spearman) for each grid cell for the 25 10-year moving windows and "Random forest.m" for building RF models to gauge the importance of climate and atmospheric CO2 in driving changes in NDVI/LAI/GPP. We trained RF models pixel by pixel and then extracted the dominant driver for each grid cell by ranking the feature importance.