Shi Hu

Accurately representing crop growth in land surface models (LSMs) is crucial for capturing cropland–atmosphere interactions, but in practice this remains challenging because many crop-related parameters are poorly constrained. In this study, we focus on reducing parameter uncertainty in the Noah-MP-Crop model by combining global sensitivity analysis with a targeted parameter optimization approach. We first assess the sensitivity of simulated crop yield and leaf area index (LAI) using the Morris screening method and Sobol' variance decomposition at nine agricultural experimental sites across China. Across sites and crops, the analysis consistently points to parameters associated with CO2 assimilation and carbon use efficiency as the primary controls on crop growth simulations. In contrast, parameters describing crop structure and phenology tend to matter most at specific growth stages rather than throughout the entire season. Guided by these sensitivity results, we then apply a targeted optimization strategy, in which the ranges of key parameters are progressively refined using an interval-based search framework. Comparison with field observations shows clear improvements after optimization: yield RMSE is reduced by 25–86 %, LAI-related objective function values decrease by 49–90 %, and mean absolute errors drop by 40–80 % across major crop types. Overall, our results suggest that sensitivity-guided parameter optimization provides a practical and efficient way to reduce parameter uncertainty in Noah-MP-Crop. The optimized model better reproduces observed LAI and yield dynamics across different climatic regions, offering useful insights for improving crop representation in land surface models.

Accurately forecasting daily streamflow remains challenging as existing models struggle to balance flexibility for rapid process shifts with physical consistency. Here we present HydroMoE, a dynamic-gated Mixture-of-Experts framework pairing process-based and neural experts for each hydrological subprocess. Meteorology-responsive gates learn to allocate weight between physical and neural experts, advancing hybrid modelling toward interpretable process understanding. Across 550 CAMELS-United States basins, HydroMoE substantially improves daily streamflow prediction. In the independent test period, HydroMoE achieved a median NSE of 0.663 and a median KGE of 0.638, compared with 0.127 and 0.210 for a differentiable process-based baseline and single-module neural ablations. The learned gating weights exhibit physically interpretable patterns: runoff experts are activated during storm events, seasonal cycles align with snowmelt and evapotranspiration dynamics, and spatial transitions correspond to Köppen-Geiger climate zones. Future work may extend this framework to integrate multiple competing physical hypotheses and probabilistic forecasting.

Surface water extent (SWE) is a key indicator of the regional water balance in dryland environments. However, the hydrological processes regulating SWE responses remain poorly constrained. Focusing on the Mu Us Sandy Land (MUSL), this study integrates multi-source remote sensing and hydrological datasets to investigate the long-term evolution of SWE and, critically, to distinguish the hydrological linkages between SWE dynamics and water storage variability in endorheic and exorheic regions during 1987–2024. An improved water extraction method was implemented on the Google Earth Engine platform, and SWE dynamics were interpreted within a water-balance framework supported by attribution analysis using machine learning. The results show that total SWE exhibited a significant increasing trend (7.95 km² yr⁻¹, p < 0.05) during 1987–2024, primarily driven by permanent SWE, while fundamentally different hydrological regimes governed SWE evolution. In the endorheic basin, SWE exhibited strong co-variation with subsurface water storage, with soil moisture and groundwater storage changes occurring concurrently with SWE changes. In contrast, no synchronous increase in SWE with groundwater storage was observed in the exorheic region. Instead, SWE expansion was mainly associated with accelerated groundwater storage depletion and reservoir construction. These contrasting patterns indicated that SWE dynamics in the endorheic basin were primarily controlled by subsurface water storage, whereas in exorheic regions they were largely driven by human-induced water redistribution rather than increases in total water storage. These findings highlight the importance of integrated surface–subsurface water management for sustaining long-term water security under climate change and increasing human water regulation.

Reliable simulation of land–atmosphere interactions in land surface models (LSMs) requires realistic root distribution modeling, yet conventional static root parameterizations often fail to capture seasonal root dynamics in cropland ecosystems. In this study, we developed a soil–environment–responsive dynamic root distribution scheme (SE_root) within the Noah‐MP‐Crop framework that explicitly accounts for soil moisture (SM), temperature, aeration, bulk density, and soil texture. Evaluations against in situ observations in the North China Plain demonstrated that SE_root substantially outperformed the default static (fixed_root) and exponential dynamic (Exp_root) parameterizations. Site‐scale simulations exhibited improved accuracy in capturing SM dynamics, leaf area index (LAI), and latent heat flux (LHF), yielding R ² values consistently above 0.56. The simulated vertical root biomass distribution, with approximately 70% of root biomass concentrated in the upper 40 cm and declining with depth, closely matched field observations. Relative to the fixed_root and Exp_root schemes, the site‐scale mean absolute errors were reduced by 10%–12% for SM and 4%–10% for LAI. Regional simulations further revealed that by capturing the dynamic feedback between root growth and local soil constraints, SE_root better represented the spatial heterogeneity of SM and LAI, alongside modest LHF improvements. Overall, incorporating this soil‐responsive root parameterization improves the representation of SM dynamics, root allocation, crop growth, and land–atmosphere exchanges. These findings underscore the importance of explicitly representing root–soil interactions in agricultural LSMs, offering a robust pathway for coupling with climate models to capture crop–climate feedbacks and support sustainable management.

Desertification threatens 24% of the world’s land area across 126 countries and affects 35% of the global population. However, mainstream global efforts to combat desertification prioritize short-term vegetation greening over addressing resource constraints and local livelihoods, posing hidden challenges to long-term Sustainable Development Goals (SDGs) attainment. Here, we evaluated the socioeconomic trade-offs of China’s long-standing desertification reversal strategies and explored the potential benefits of innovative agricultural practices using high-resolution suitability models for major crops and cultivated pasture. Compared with a no-intervention baseline, continuing the Grain-for-Green Program and grazing exclusion under projected 2023–2050 warming and wetting trends would lead to reductions of up to 54 to 55% in grain output, 81% in livestock production, and 61% in agricultural income. Conversely, strategies emphasizing pasture cultivation and crop switching could substantially lower national expenditures on combating desertification while restoring ecosystems, enhancing agricultural productivity and incomes, and conserving water resources. These benefits would contribute 0.7%, 9.8%, and 3.8% toward global progress on SDG 1 (no poverty), SDG 2 (zero hunger), and SDG 6 (clean water and sanitation), respectively, while exhibiting trade-offs across multiple SDGs. This study provides a reimagined strategy framework for combating global desertification and promoting sustainable development in vulnerable arid regions worldwide.

In the context of global climate warming, compound drought and heatwave events (CDHEs) have exhibited a pronounced escalation in frequency since the Second Industrial Revolution, incurring substantial socioeconomic losses. This study investigates the spatiotemporal variations of CDHEs in semi-arid and semi-humid regions of northern China based on daily Standardized Precipitation Index (SPI) and maximum temperature (Tmax) datasets. The results show that compared to the 1980s, the occurrence frequency of CDHEs during the 2010s exhibited an increasing trend increase by 20–50 times in the southern region and 10–30 times in the northern region, while some watersheds in the central part of the study area show a decreasing trend. From the 1980s to the 2010s, the percentage of area affected by CDHE with a duration exceeding 11 days/year has risen from 28.3% to 56.7%, reflecting a pronounced upward trend in CDHE duration. Spatiotemporal patterns revealed significant interdecadal disparities in both the frequency and duration of CDHEs, which are primarily determined by heatwave events pattern and the synchronicity of heatwave and drought events. However, drought intensity exhibits comparatively weaker influence. Due to the decrease in the proportion of short–duration heatwaves, the short–duration CDHEs (1–2 days) in all levels exhibited a declining trend in their proportions. Furthermore, the delayed occurrence of drought events resulted in the peak occurrence of CDHEs has gradually shifted June to July–August.

There has been increasing attention paid to the effects of drought, especially extreme drought, on vegetation productivity under climate change. However, there are still challenges in quantifying the variations and the adverse effect of extreme drought on vegetation productivity at a regional scale within the context of historical climate change. This study quantified changes in the characteristics of extreme droughts and their effects on vegetation productivity in the Hai River Basin (HRB) of China, using factual climate (with climate trends) and counterfactual climate (detrended) data from the ISIMIP3a project. Standardised Precipitation Evapotranspiration Index (SPEI) and Run theory were utilised to determine characteristic factors (drought frequency, duration, severity, intensity and peak) under climate change. By comparing factual and counterfactual forcing simulations, the detected changes in characteristic factors were attributed to climatic trends. The negative effects of extreme droughts on gross primary productivity (GPP) were quantified. Results showed that there were more serious extreme drought events that occurred in HRB from 1901 to 2019 under factual climate than those under counterfactual climate. The difference was exacerbated in the late stages (1980–2019) over most of the basin. A deceleration was found in the rising pattern of GPP over the last four decades, exacerbating the adverse effects of extreme droughts on GPP under climate change. Compared to those during 1982–2000, GPP experienced further losses related to extreme droughts during 2000–2018 at a rate of 2°gC°m⁻²°day⁻¹. Furthermore, extreme drought‐related GPP losses were most pronounced in summer, indicating that vegetation was more sensitive to extreme droughts during this season. These findings enhance our understanding of historically observed trends of extreme drought and suggest that more strategies should be implemented to protect vegetation from the increasing trends of extreme drought.

Intensive and frequent flash droughts pose a threat to the stability of terrestrial ecosystems and affect land carbon sinks, particularly in vulnerable water-limited regions. Here, we evaluated variations in flash drought events in water-limited regions of China from 2001 to 2022 and quantified vegetation resistance (its ability to maintain predrought level) and resilience (its capacity to bounce back from droughts) to flash drought based on solar-induced chlorophyll fluorescence and gross primary productivity. The average frequency of flash droughts was approximately 3.5 events per decade and the duration increased significantly by 3 d during 2001–2022. We found that vegetation resistance to flash drought did not show significant temporal changes, while resilience decreased significantly. Spatially, as aridity increased, vegetation resistance gradually decreased while resilience rose. The strong negative correlation between resistance and resilience demonstrated a trade-off in individual climate regions, which varied among different vegetation types. Among eight vegetation types, evergreen needleleaf forest, evergreen broadleaf forest and mixed forest displayed high-resistance and low-resilience, while deciduous needleleaf forest exhibited both high-resistance and high-resilience. Furthermore, we revealed that aridity index, vapor pressure deficit, and temperature were the most relevant environmental variables for both vegetation resistance and resilience, but had opposite effects and different magnitudes. Our findings highlight that the vulnerability and stability of terrestrial ecosystems are decreasing and are affected by water availability.