Haojun Zheng

Accurate phenological model predictions are needed for assessing grassland ecosystem dynamics under climate change. However, most phenological models for arid and semi-arid grasslands remain temperature-based and have rarely been evaluated across the full growing season. Using long-term phenological and meteorological observations from 26 sites in the Inner Mongolian grassland, we compared five candidate model formulations for four dominant grassland species across five phenological stages: green-up, heading, flowering, fruiting, and brown-down. These formulations included the traditional thermal-time model ( M _t) and four temperature–precipitation models differing in precipitation metric (cumulative amount or frequency) and accumulation structure (parallel or sequential). Compared with M _t, precipitation-inclusive formulations had the lowest AICc in 15 of the 19 species–stage combinations; absolute RMSE reductions ranged from 0.18 to 7.39 d, with a median of 1.05 d, and 7 of the 15 reductions were M _t was top-ranked for flowering in three species and sequential formulations were top-ranked for all fruiting combinations. Seven of the 19 combinations retained competing alternatives (runner-up ΔAICc < 2), whereas all four brown-down combinations showed clear model separation (runner-up ΔAICc > 10). Precipitation-frequency formulations were supported in selected combinations, particularly for Stipa krylovii and Artemisia frigida , but did not consistently outperform cumulative precipitation amount. These results support a stage-specific modelling strategy for semi-arid grassland phenology, in which precipitation information should generally be considered, but precipitation metric and temperature–precipitation structure should be selected according to phenological stage and species rather than applied uniformly across the growing season.

The Inner Mongolia grassland has been facing critical degradation. To combat this issue, China has launched a series of ecological restoration programs since the early 2000s to promote the recovery of degraded grasslands (e.g., the Beijing–Tianjin Wind/Sand Source Control Program, the Grazing Withdrawal Project, and the Ecological Compensation Policy). However, the long-term effects of these measures on the recovery of the Inner Mongolia grassland still need further evaluation. To quantitatively assess the grassland degradation status and dynamics, we utilized long-term remote sensing datasets (GIMMS NDVI and MODIS NDVI) from 1982 to 2020 to calculate the fractional vegetation cover (FVC, the ratio of vegetation canopy's vertical projection area to the unit area) and grassland degradation index (GDI, the ratio of average FVC during peak growing season to historical optimum FVC). Results showed that across the study region, the average FVC was 42% over the past four decades, with degraded grassland (GDI < 0.9) accounting for 84.6% of the total area. Noteworthy spatial differences were observed. Specifically, study region to the west of 115°E had an average FVC of 20.7% with 98.2% areas exhibiting a degraded state. By contrast, study region to the east had an average FVC of 57.7% with 68.5% areas experiencing degradation. Over time, the proportion of degraded grassland generally decreased, particularly in the past two decades. However, approximately 3/4 of the grassland remained degraded in the last decade (2011–2020). Although degraded grassland is recovering, grassland degradation remains serious, especially in the west of 115°E, where 80.8% of the grassland was moderately and severely degraded (GDI < 0.6) in the last decade. Our findings suggest that while the implementation of ecological protection measures in Inner Mongolia has slowed down grassland degradation, the persisting degradation underscores the necessity of region-specific grassland protection and restoration.