Exploring the combined effects of drought and drought-flood abrupt alternation on vegetation using interpretable machine learning model and r-vine copula function

Global warming has significantly increased the frequency and intensity of extreme events, such as drought and drought-flood abrupt alternation (DFAA). Vegetation, a crucial component of terrestrial ecosystems, contributes greatly to agricultural production and economic development. Assessing the impacts of droughts and DFAA on vegetation is essential for ecological environment protection, food security, and economic development. This study combines the random forest model with Shapley Additive Explanation (SHAP) values to create an interpretable machine learning model, which is then coupled with the R-Vine Copula function to explore the combined effects of drought (using nonstationary drought indexes) and DFAA (using the SDFAI: short-cycle drought flood abrupt alteration index) on vegetation in the China-Pakistan Economic Corridor (CPEC) from 1981 to 2019. The key findings are as follows: (1) Drought events intensified, the risk of flood-to-drought decreased, the risk of drought-to-flood increased, and net primary productivity (NPP) showed an upward trend; (2) The relative contributions to NPP were NSPEI (21.0 %), SDFAI (31.4 %), and SSMI (47.6 %); (3) A strong upper tail dependence occurred between SSMI and NPP, and a strong lower tail dependence occurred between SDFAI and SSMI. When SSMI acted as an intermediary variable, the indirect correlation between SDFAI and NPP was strong; (4) In flood-to-drought events, the proportions of SHAP<0 and SHAP>0 were 24 % and 76 %, respectively, indicating an antagonistic role of flood-to-drought in promoting vegetation growth in the CPEC. In drought-to-flood events, the corresponding proportions were 73 % and 27 %, respectively, indicating a synergistic effect of drought-to-flood in inhibiting vegetation growth. This study enhances the understanding of the mechanisms by which DFAA impacts vegetation and provides a novel approach for exploring the combined effects of multiple extreme events on vegetation.