Abstract: The effect of climate change on vegetation cover is time-scale dependent. Based on the MODIS NDVI data from 2000 to 2020, the pixel dichotomy model was used to obtain fractional vegetation coverage in Hunan Province. The correlations between vegetation coverage and climate factors at different temporal scales were quantified by using multivariate empirical mode decomposition (MEMD). The vegetation coverage in Hunan Province was predicted. The MEMD could decompose the fractional vegetation coverage into seven intrinsic mode functions (IMFs) and the residuals. The dominant scale was IMF3 (12 months), followed by IMF1 (3 months) and IMF2 (6 months). The total contribution rate of those three scales to the variance of fractional vegetation coverage was 94.5%. Fractional vegetation coverage was significantly correlated with precipitation and potential evapotranspiration at each scale (P<0.05). There was no relationship between air temperature and fractional vegetation cover at 3-month time scale. However, air temperature, precipitation and potential evapotranspiration all had significant effects on vegetation growth when the time scales were greater than 12 months (P<0.05). The accuracy of fractional vegetation coverage prediction based on MEMD (R²=0.73) was better than that of the stepwise multivariate regression directly using the original data (R²=0.67). Temperature and potential evapotranspiration were the predominant factors in predicting fractional vegetation coverage. This study can provide a scientific basis for regional vegetation restoration and sustainable development of ecological environment.

Key words: fractional vegetation coverage, pixel dichotomy model, multivariate empirical mode decomposition, multiscale, climate change