Threshold and nonlinear responses of vegetation productivity to extreme precipitation in Tajikistan.

Extreme precipitation strongly influences vegetation productivity in water-limited ecosystems, yet how different precipitation extremes regulate productivity, and whether these relationships vary across mountainous environments, remain poorly understood. Here, we combined the CASA model with an XGBoost-SHAP interpretability framework to quantify the associations between extreme precipitation characteristics and net primary productivity (NPP) across the arid mountain ecosystems of Tajikistan. Extreme precipitation exhibited pronounced spatial heterogeneity, with wet-extreme hotspots concentrated in the Vakhsh Valley, Gissar Valley, and central mountains, while NPP was strongly concentrated in the growing season, with 86.72% of annual NPP occurring during April-October and 64% during April-July. Wet-day frequency, dry-spell duration, and multi-day precipitation intensity emerged as the dominant statistical dimensions linking precipitation extremes to NPP. R10mm, CDD, R20mm, R5mm, and Rx5day collectively accounted for 71.3% of the total mean absolute SHAP importance, highlighting the stronger contribution of precipitation frequency, dry-spell persistence, and multi-day intensity than of precipitation totals alone. The dominant controls also shifted systematically along the elevation gradient. Low-elevation productivity was primarily associated with the frequency of moderate-to-heavy precipitation, whereas productivity at middle and high elevations increasingly reflected precipitation amount, event frequency, and dry-wet spell persistence. Moreover, Rx5day exhibited a distinct nonlinear, threshold-like relationship with NPP, with productivity peaking at approximately 30 mm before declining as multi-day precipitation extremes intensified further. These findings reveal elevation-dependent pathways and a regional nonlinear threshold in vegetation responses to extreme precipitation, providing new insights into how precipitation extremes regulate productivity in water-limited mountain ecosystems and informing their assessment under increasing hydroclimatic variability.