China’s deserts greening and response to climate variability and human activities

Spatial differences in correlations. To further compare the differences in the impacts of climate variability on NDVI at different scales spatially, the Pearson correlation coefficients for the relationship between monthly and annual NDVI and temperature and precipitation from 2000 to 2017 were calculated for each pixel, and the proportion of the area of the correlation coefficient is shown in Table 5. Fig 6(a) illustrates the spatial distribution of the temperature effect on the NDVI at the monthly scale; 94.15% of the area in which the NDVI had a positive correlation with tempera- ture was distributed in areas other than the eastern part of the Taklimakan Desert and the central part of the Qaidam Basin Desert. The strong positive correlation was principally dis- tributed in the Gurbantunggut Desert, the western edge of the Taklimakan Desert, the western and eastern edges of the Qaidam Basin Desert, the Tengger Desert, and the Kubuqi Desert, indicating the area where the NDVI showed significant greening trends. Fig 6(b) illustrates the spatial distribution of the influence of precipitation on NDVI at the monthly scale, which was similar to that of temperature; however, the correlation in the Gur- bantunggut Desert was not as strong as that of temperature. Overall, 83.28% of the area showed a positive correlation, and 68.19% was significant (p<0.05), which was lower than that of temperature. In addition, 16.72% of the area had a negative correlation; this area was distrib- uted in the eastern part of the Taklimakan Desert and the central part of the Qaidam Basin Desert.

The correlations between the annual NDVI and climatic factors were not significant in the majority of pixels, which were much weaker than the correlations between the climatic factors and the monthly NDVI. In the very few pixels with significant correlation, tempera- ture mainly had a negative effect on the NDVI, while precipitation had a positive effect and had a larger area than temperature. Specifically, the percentages of the significant positive and negative correlations with temperature were only 0.50% and 3.98%, respectively, and those with precipitation were only 14.99% and 0.15%, respectively. In terms of spatial distri- bution, the positive correlation with temperature was mostly in the southeastern part of the Gurbantunggut Desert, the Qaidam Basin Desert, and the Ulan Buh Desert, and the negative correlation was mostly in the western part of the Gurbantunggut Desert, the midwestern part of the Taklimakan Desert, and the Badain Jaran Desert. In contrast, the positive corre- lation with precipitation was distributed in most areas, with a slight negative correlation in the mideastern part of the Taklimakan Desert and the central part of the Qaidam Basin Desert. Table 5. Proportion of the area of the correlation coefficients for the relationships between the monthly and annual NDVI and temperature and precipitation. Scale Correlation level Temperature Precipitation Monthly Significant positive (%) 71.04 68.19 Positive but not significant (%) 23.11 15.09 Significant negative (%) 2.22 5.93 Negative but not significant (%) 3.63 10.79

Annual Significant positive (%) 0.50 14.99 Positive but not significant (%) 38.35 69.97 Significant negative (%) 3.98 0.15 Negative but not significant (%) 57.17 14.89 “Significant positive” denotes pixels r>0 and p<0.05, “Positive but not significant” denotes pixels r>0 and p>0.05, “Significant negative” denotes pixels r<0 and p<0.05, and “Negative but not significant” denotes pixels r<0 and p>0.05. https://doi.org/10.1371/journal.pone.0256462.t005 PLOS ONE China’s deserts greening and response to climate variability and human activities PLOS ONE | https://doi.org/10.1371/journal.pone.0256462 August 30, 2021 11 / 20