“Decoding Cotton Yield Sensitivity to Meteorological and Agricultural Droughts Using Integrated Droughts Indices in China-Pakistan Economic Corridor (Cpec)”

Abstract

Drought significantly threatens social stability and food security, with crop growth influenced by both meteorological and agricultural droughts. Moreover, most research has focused on individual sites or specific counties, leaving a gap in regional studies, leading to incomplete assessments of drought impacts for major cotton-producing areas like the China–Pakistan Economic Corridor (CPEC) region. This study uses four machine learning methods to investigate the effects of meteorological and agricultural drought indices on cotton yields. Drought severity was assessed at 1-, 3-, 6-, and 12-month timescales across 75 stations using the Standardized Precipitation–Evapotranspiration Index (SPEI) and the Self-Calibrated Palmer Drought Severity Index (SC-PDSI). The DSAAT-CROPGRO-Cotton model was calibrated to simulate cotton yield responses using Random Forest, Linear Regression, Ridge Regression, and Support Vector Regression models. The analysis showed that most drought events during the 2000-02s and 2009-10s occurred in southwestern Pakistan and northeastern Xinjiang, respectively. The DSSAT-CROPGRO-Cotton model demonstrated robust performance, with coefficient of determination (R2) values ranging from 0.70 to 0.74, root mean square error (RMSE) between 408.88 and 2,725.92 kg/ha, and an index of agreement (d) from 0.72 to 0.93 for cotton yields. For growth-related variables such as biomass, dry matter, canopy cover, and leaf area index, R2 ranged from 0.84 to 0.99, RMSE from 0.03 to 324, and d from 0.78–0.99, meeting key performance evaluation criteria. In addition, water use efficiency during cotton growth period was found greater in Xinjiang (China) 0.17 to 1.05 kg/m3 than Pakistan (0.12-0.80 kg/m3). Among the four machine learning models, Random Forest outperformed others across most of the CPEC region, except in two to three counties. Agricultural drought indices impacted cotton yield reduction (20.9%), more than meteorological drought indices (16.29%). This study provides valuable insights for policymakers to develop sustainable drought management strategies and climate adaptation plans to mitigate drought impacts in the CPEC region.

Keywords: Drought monitoring, integrated drought indices, DSSAT-CROPGRO-Cotton model, machine learning algorithms, yield reduction rate.