Vegetation Productivity Loss and Recovery Associated with the July 2023 Hot–Dry Event on the Huang–Huai–Hai Plain

Extreme climatic events can alter temperatures and water availability and thereby suppress vegetation productivity. On the Huang–Huai–Hai (HHH) Plain, China’s key agricultural belt, July 2023 featured widespread heat, spatially heterogeneous drying, and limited spatial co-occurrence of heat and root zone soil drought; however, its productivity response and within-season recovery remain unclear. Here, we combine satellite-derived solar-induced chlorophyll fluorescence (SIF), MODIS gross primary productivity (GPP), environmental variables, and vegetation and elevation stratification and use extreme gradient boosting (XGBoost) with spatial block validation and SHapley Additive exPlanations (SHAP) to quantify July productivity anomalies relative to the same months during 2018–2022, identify affected pixels using a standardized SIF anomaly based on the same reference period (zi ≤ −1.5), track recovery from August to October, and evaluate environmental associations with SIF. In July 2023, 59.6% of vegetated area exceeded its local July 90th-percentile temperature. Area-weighted anomalies were −0.0129 W m−2 μm−1 sr−1 for SIF, −0.4853 g C m−2 d−1 for GPP, and −0.0147 m3 m−3 for root zone soil moisture (SMrz). Negative anomalies occurred in 57.3% of vegetated pixels for SIF and 73.4% for GPP. Among deciduous broadleaf forest (DBF), grasslands (GRA), and croplands (CRO), GRA had the largest mean SIF loss, whereas CRO had the smallest despite widespread local declines. Losses above 1000 m exceeded those at 0–500 m. Of 2833 affected pixels, 85.6% returned to non-negative SIF anomalies by October, while recovery above 1000 m was 75.6%. Within the July models, SMrz contained the largest individual predictive contribution across classes, although the combined contribution of air temperature, vapor pressure deficit, and downward shortwave radiation was comparable. In the recovery model, recovery stage, initial July loss, and radiation and atmospheric demand contained substantial predictive information. Together, the results characterize spatially uneven productivity loss and recovery and the environmental conditions associated with these differences.