Introduction

Ensuring access to affordable and clean energy is essential for achieving the Sustainable Development Goals1, necessitating a substantial increase in the share of renewable energy, such as hydro, solar, and wind. Hydropower has been the leading renewable energy source worldwide, with a total capacity of 1,268 GW by the end of 2023, accounting for 14% of global installed power capacity and 33% of global renewable power capacity2. In the transition to a net-zero emissions energy system, hydropower offers a cost-effective alternative to the high solar and wind installation costs at the current technological level3. Its flexibility and energy storage capacity can effectively balance the intermittent supply from solar and wind sources and meet peaks in energy demand on grids4,5,6. Consequently, hydropower has experienced renewed interest and rapid expansion in recent decades7, with an unprecedented 75% growth from 2000 to 20218, particularly in Asia-Pacific, South America, and Africa9. Nevertheless, vast untapped hydropower potential remains10, with considerable expansion planned or underway in the Global South11.

Despite being the most mature renewable technology, hydropower faces significant challenges, notably reservoir sedimentation. Sedimentation impairs reservoir’s storage capacity and abrades power turbines12, undermining the longevity of reservoirs and diminishing the operational flexibility of hydropower generation. Without adequate storage, hydropower facilities become reliant on seasonal flows, which might not occur when energy is needed, thus negating one of the primary advantages of hydropower over other renewable energy sources. Reservoirs worldwide lose approximately 0.5–1% of their initial storage capacity annually due to sedimentation13. A study of over 47,000 dams across 150 countries revealed that these dams had already lost 16% of their combined original storage capacity by 2022 due to trapped sediment, with this figure expected to rise to 26% by 205014. The higher erosion rates and increased sediment load caused by climate change and changes in watershed land use are likely to exacerbate reservoir sedimentation issues faster than previously anticipated15,16, amplifying the threat to the sustainability of hydropower.

Watershed land use management to control soil erosion can effectively mitigate reservoir sedimentation from upstream sources17. Vegetation measures, such as afforestation and grass planting, mitigate erosion by enhancing infiltration and reducing overland flows18,19, while engineering measures, such as terraces and check dams, decrease discharge velocity and increase surface storage, thereby reducing sediment transport19,20. These measures can extend the operational lifespan of hydropower facilities by reducing sediment flows into reservoirs18. However, large-scale afforestation often increases evapotranspiration (ET), leaving less water for runoff21—especially in semiarid and arid regions22—thereby diminishing hydropower potential, which depends directly on streamflow23. These opposing effects through changes in sediment and runoff raise an elusive question: what is the net impact of watershed land use management on hydropower potential? Existing studies typically focus on the separate effects through runoff or sediment change23,24 or seek future optimal management strategies by simultaneously considering these effects25,26. However, research integrating these opposing effects to analyze how real-world practices affect hydropower potential remains limited.

To address this knowledge gap, we select China’s Yellow River (YR) to explore how large-scale ecological restoration (ER) in the Yellow River Basin (YRB) affects the hydropower potential of the Xiaolangdi Reservoir, a strategic water conservancy project in the YR controlling 92.3% of the total basin area (Fig. 1a). The YR is an ideal case for this research, as it once carried the largest sediment load of any river in the world27,28. Severe sedimentation problems plagued its first dam, the Sanmenxia Reservoir, which lost more than 40% of its storage capacity within the first four years and underwent extensive reconstruction afterward29. To address the sediment issue, numerous soil and water conservation practices have been implemented since the 1970s, including terracing, check dam construction, and vegetation restoration20 (Supplementary Fig. 1). Among them, the Grain-for-Green Program (GFGP) implemented since 1999 is the largest and most successful30. Vegetation coverage of the YRB has significantly increased since 2000 (Fig. 1a), leading the greening process of China31. Along with these ER measures, the sediment load and streamflow of the YR have significantly decreased20 (Fig. 1b, c). The Xiaolangdi Reservoir, completed in 2001, has a total installed capacity of 1800 MW32. It has a total storage capacity of 12.65 billion m3, including a sediment storage capacity of 7.55 billion m3 for sediment deposition and a regulating storage capacity of 5.10 billion m3 for flow regulation and electricity generation32 (Supplementary Fig. 2). The operating period of the Xiaolangdi Reservoir coincides with the implementation of large-scale ER in the YRB, providing an excellent opportunity to analyze the overall impact of ER on its hydropower potential by affecting the sediment and runoff of the YR.

Fig. 1: Location of the Yellow River Basin and changes in vegetation, sediment load, and streamflow.
figure 1

a Annual moderate resolution imaging spectroradiometer (MODIS) leaf area index (LAI) trend for the entire basin from 2000 to 2023. b, c Annual sediment load and streamflow at the Tongguan station from 1956 to 2023.

Full size image

In this study, we use the designed sediment storage capacity of the Xiaolangdi Reservoir as a benchmark to evaluate total energy generation. Once this storage is filled, regulating storage is invaded, compromising the flexibility and efficiency of hydropower. Total energy generation is determined by monthly energy production and the sediment storage lifespan. As previously noted, energy production depends positively on reservoir inflow, while sediment storage lifespan depends inversely on sediment load. These factors are influenced by upstream land use, climate, socioeconomic water use, and reservoir operations. To isolate the impact of ER on hydropower potential, we design an idealized experiment simulating streamflow and sediment load of the YR under two scenarios: with and without ER implementation (2000–2019), while holding other influencing factors consistent. Both simulations use observed meteorological data to reflect climate variability, with upstream socioeconomic water use and reservoir operations excluded (Methods, Supplementary Table 1). The Geomorphology-Based Eco-Hydrological Model with Soil and Water Conservation (GBEHM-SWC), a distributed model that integrates the complex eco-hydrological interactions and human interferences in the YRB (Supplementary Fig. 3), is employed to simulate eco-hydrological and sediment dynamics. We then simulate the regulation of the Xiaolangdi Reservoir and its energy generation and sedimentation processes, comparing the total generated energy under both scenarios. By revealing the impact of ER in the YRB on the hydropower potential of the Xiaolangdi Reservoir, this study advances understanding of the ecosystem-water-sediment-energy nexus and provides valuable insights for integrated watershed management.

Results

Hydrological and sediment responses to ecological restoration

According to the Soil and Water Conservation Bulletin in Yellow River Basin33, by 2020, extensive ER measures had been implemented in the YRB, including the establishment of 12.64 × 104 km2 of forests, 6.08 × 104 km2 of terraces, 2.34 × 104 km2 of grasses, and 5.81 × 104 check dams (Supplementary Figs. 4 and 5). These measures led to significant increases in leaf area index (LAI) and gross primary productivity (GPP) across the basin (Fig. 2). Under the scenario with ER, the basin-averaged LAI from 2000 to 2019 increased by 6.33 × 10−2 m2 m−2 (a relative change of +10.78%) compared to the scenario without ER, while GPP increased by 47.50 g C m−2 (+13.04%). Notably, the differences between the two scenarios expanded over time. ER also effectively mitigated soil erosion, with basin-averaged erosion from 2000 to 2019 declining by 706.74 t km−2 (−57.11%). However, the large-scale implementation of ER increased basin-averaged ET by 11.39 mm (+3.02%), resulting in a corresponding reduction in basin-averaged runoff by 8.32 mm (−11.02%).

Fig. 2: Changes in leaf area index (LAI), gross primary productivity (GPP), soil erosion, evapotranspiration (ET), and runoff (R) in the Yellow River Basin under scenarios with and without ecological restoration.
figure 2

a–e Maps of the average values of these variables from 2000 to 2019 under the scenario with ecological restoration. f–j Maps of the average values of these variables from 2000 to 2019 under the scenario without ecological restoration. k–o Changes in the basin-average values of these variables from 2000 to 2019 under the two scenarios.

Full size image

These eco-hydrological changes in the YRB caused by ER have led to significant reductions in both streamflow and sediment load of the YR (Fig. 3). Simulations indicate that, under the scenario with ER, the average annual streamflow into the Xiaolangdi Reservoir from 2000 to 2019 was 401.35 × 108 m3, a 7.9% decrease from the 435.82 × 108 m3 observed under the scenario without ER (Fig. 3a). The reduction in average annual sediment load was even more pronounced, falling from 7.54 × 108 t to 4.61 × 108 t, a reduction of 38.9% (Fig. 3c). With the implementation of ER measures, the disparity in streamflow and sediment load between the scenarios with and without ER widened significantly (Fig. 3b, d), showing an average reduction of 53.60 × 108 m3 (−11.7%) in streamflow and 3.28 × 108 t (−42.3%) in sediment load during the last five years (2015–2019).

Fig. 3: Streamflow and sediment load into the Xiaolangdi Reservoir from 2000 to 2019 under scenarios with and without ecological restoration, and the differences between them.
figure 3

a Streamflow from 2000 to 2019. b Differences in streamflow between the two scenarios. c Sediment load from 2000 to 2019. d Differences in sediment load between the two scenarios.

Full size image

Changes in energy generation and sedimentation of the Xiaolangdi Reservoir

Based on the simulated hydrological and sediment conditions of inflow into the Xiaolangdi Reservoir under different scenarios, we further simulated and compared the reservoir’s potential energy generation and sedimentation from 2002 onwards (Fig. 4), following the completion of construction at the end of 2001. The average annual energy generation from 2002 to 2019 under the scenario with ER was 121.00 × 108 kWh, a 6.9% decrease from the 130.01 × 108 kWh generated under the scenario without ER. The energy generation process and differences between the two scenarios exhibited intra-annual variation, with June to November contributing more than 60% of the annual energy generation under both scenarios. August and September showed the largest declines in monthly energy generation under the scenario with ER, with reductions of 17.8% and 11.5%, respectively. The average annual reservoir sedimentation also decreased with ER implementation, reducing from 5.44 × 108 m3 under the scenario without ER to 3.36 × 108 m3 under the scenario with ER, a decline of 38.3%. The flood season, spanning May to October, accounted for more than 90% of the annual sedimentation under both scenarios, with July and August contributing the most. The largest declines in monthly reservoir sedimentation under the scenario with ER also occurred in July and August, with reductions of 51.5% and 47.0%, respectively.

Fig. 4: Annual and monthly energy generation and sedimentation of the Xiaolangdi Reservoir from 2002 to 2019 under scenarios with and without ecological restoration, and comparison of their mean values.
figure 4

a Annual energy generation from 2002 to 2019. b Annual sedimentation from 2002 to 2019. c Monthly energy generation from 2002 to 2019. d Monthly sedimentation from 2002 to 2019. e Comparison of average monthly energy generation. f Comparison of average monthly sedimentation.

Full size image

Accumulated energy generation enhanced by ecological restoration

Although the Xiaolangdi Reservoir exhibited higher average annual energy generation under the scenario without ER, the accompanying higher reservoir sedimentation significantly shortened the reservoir’s sediment storage lifespan compared to the scenario with ER (Fig. 5). The accumulated reservoir sedimentation under the scenario without ER was 97.91 × 108 m3 by December 2019, the endpoint of our simulation. Using the designed sediment storage capacity of the Xiaolangdi Reservoir, 75.50 × 108 m3, as the threshold for the reservoir’s lifespan, the sedimentation reached 75.90 × 108 m3 in July 2015, thereby exhausting the sediment storage capacity. The total energy generation under this scenario was 1713.63 × 108 kWh. However, under the scenario with ER, the accumulated reservoir sedimentation was only 60.43 × 108 m3 by December 2019, utilizing just 80% of the sediment storage capacity. Despite lower average annual energy generation, the accumulated energy generation under the scenario with ER exceeded that of the scenario without ER in May 2016, reaching 2177.96 × 108 kWh by December 2019. According to the trend derived from our simulation, the sediment storage capacity of the Xiaolangdi Reservoir under the scenario with ER would be exhausted in 2024, with a total energy generation of 2694.89 × 108 kWh, 57.3% higher than the scenario without ER, providing approximately 100 billion kWh of additional energy.

Fig. 5: Accumulated sedimentation and energy generation of the Xiaolangdi Reservoir under scenarios with and without ecological restoration.
figure 5

a Accumulated sedimentation. b Accumulated energy generation.

Full size image

Discussion

Integrating the opposing effects through sediment and runoff changes to understand the overall impact of ER on hydropower potential is crucial for effective watershed management. Utilizing eco-hydrological and reservoir regulation models, our study revealed that ER in the YRB enhances the hydropower potential of the Xiaolangdi Reservoir. Compared to the scenario without ER, the scenario with ER exemplifies the proverb “slow and steady wins the race”: despite a 6.9% reduction in average annual energy generation due to decreased streamflow, the reduced sediment load prevents the premature loss of regulating storage capacity, preserving the flexibility and efficiency of energy generation and ultimately providing an additional ~100 billion kWh of energy.

The observed enhancement of hydropower potential due to ER deepens our understanding of the ecosystem-water-sediment-energy nexus. The GFGP and other ER measures in the YRB have significantly increased vegetation coverage and improved multiple ecosystem services, such as carbon sequestration and soil conservation20,34, while reducing sediment load and improving water quality of the YR35. However, vegetation restoration has also led to soil moisture decline and soil desiccation in certain areas and streamflow reduction of the YR due to the conflict between plant growth and water consumption20,30,34. The net positive impact of ER on hydropower potential is mainly attributed to the distinct sources of water and sediment of the YR. The water primarily originates upstream of Lanzhou, contributing about 60% of the annual runoff from less than 30% of the basin area36, while nearly 90% of the sediment comes from the middle reaches, the Loess Plateau, which suffered severe soil erosion due to historical deforestation and agricultural development37. Since large-scale ER was mainly implemented in the middle YRB (Fig. 1, Supplementary Figs. 4 and 5), the reduction in sediment load into the Xiaolangdi Reservoir is disproportionately higher than the reduction in streamflow. Consequently, the negative impact of ER on hydropower through reduced annual energy generation is offset by the significant reduction in reservoir sedimentation. Beyond the focus on energy production in this study, the Xiaolangdi Reservoir provides other critical socioeconomic benefits, including flood and ice jam control, agricultural irrigation, and the regulation of seasonal water supply to meet the demands of cities and villages along the lower YR32. Since these functions depend on effective storage capacity, their reliability is also affected by reservoir sedimentation16. The sediment load reduction achieved through ER mitigates sedimentation, enhancing the security of energy, food, and environment17. However, the broader impacts of ER on these benefits require further exploration, integrating hydrological and sediment changes as demonstrated in this study.

Sediment is a fundamental challenge for hydropower development17, with reservoirs worldwide losing storage capacity much faster than expected due to climate change and human activities16. For example, reservoirs in the upper Yangtze River, China’s largest hydropower production region, have been affected by increasing sediment flux from the headwaters in recent decades16; the Tarbela Reservoir in the Indus River, Pakistan’s largest hydropower and irrigation project, lost ~30% of its original storage capacity between 1974 and 2006, far exceeding initial expectations38; the Manwan and Dachaoshan Dams in the Mekong River have lost over 50% of their storage capacity because of sediment accumulation9; the Jirau and Santo Antonio Dams in the Madeira River have had to use dredgers to remove sediment accumulating at unexpected rates within five years of their completion15. As reservoirs are usually constructed at the most viable sites, their storage capacity loss to sedimentation is challenging to offset by building new ones14. The experiences from the YR can be extended to other rivers worldwide to mitigate reservoir sedimentation and enhance hydropower potential through rational watershed land use management. While controlling erosion at its source is widely recommended, it remains poorly implemented in reservoir sediment management due to the lack of direct benefits for land users39. In the YRB, ER was primarily funded by China’s central and local governments31 and separated from the construction of the Xiaolangdi Reservoir, yet it yielded the unintended but beneficial outcome of enhanced hydropower potential. This bonus benefit highlights the feasibility of payment for ecosystem services (PES) in other rivers with hydropower and clarifies who should bear the cost. If simulations of the ecosystem-water-sediment-energy nexus for other rivers indicate that effective upstream land use management can increase energy generation from certain existing or planned reservoirs, hydropower companies could allocate a portion of their additional profits to compensate upstream communities for adopting sediment-reducing land use measures. Drawing on real-world practices in the YRB, these measures include converting cropland to terraces, forests, or grasslands, planting vegetation in landslide-prone areas, constructing check dams, and removing excess sediment from these dams20. By achieving targeted sediment load reductions through such measures, upstream farmers could receive compensation in the form of cash payments or rewards such as reduced energy costs from hydropower companies. Such PES mechanisms have proven successful in some small watersheds like Indonesia’s Way Besai watershed40. As our study demonstrates their significant potential for large-scale integrated watershed management, they merit further expansion worldwide.

Besides the positive impact of ER on hydropower potential, our results also highlight the significant contributions of other sediment management measures. By the end of 2019, the accumulated sedimentation in the Xiaolangdi Reservoir was 97.91 × 108 m3 under the scenario without ER and 60.43 × 108 m3 with ER, compared to the observed value—representing the actual scenario with all influencing factors, including ER—of just 32.97 × 108 m3. This indicates that ER accounted for 57.7% of the actual sedimentation reduction of 64.94 × 108 m3, with the remaining 42.3% attributed to other measures not included in our simulations. These measures likely involve reservoir operations such as sediment bypassing, sluicing, dredging, and drawdown flushing41, as well as the complex water and sediment regulation scheme (WSRS) of the YR28,32. The WSRS, a coordinated regulation of three major reservoirs (Wanjiazhai, Sanmenxia, and Xiaolangdi) along the YR mainstream42, is generally operated every summer to enhance sediment delivery by unleashing artificial flood waves. Together with ER in the YRB, the WSRS has greatly extended the life expectancy of the Xiaolangdi Reservoir’s sediment storage capacity, which was initially projected to be exhausted by 2020. Such coordinated reservoir regulation provides a promising complement to watershed PES and could be applied to other rivers with cascade reservoirs, such as the Yangtze, Nile, and Mekong. Effective operation of cascade reservoirs requires two regulation phases: in the first phase, clear water is released from the downstream reservoir to lower its water level; in the second phase, floodwaters from the coordinated operation of upstream reservoirs scour previously deposited sediment in the downstream reservoir, generating an artificial hyperpycnal flow with high sediment concentration and fluid density42. In this way, sedimentation of downstream reservoirs can be effectively mitigated.

As this study primarily focuses on determining whether the net impact of ER on hydropower potential is positive or negative, rather than accurately simulating real hydropower production, several simplifications and compromises were made in the experimental design. First, the estimated hydropower represents the potential of the Xiaolangdi Reservoir in the absence of upstream reservoirs’ operations and socioeconomic water use. Simulating the long-term operations of upstream reservoirs and socioeconomic water use is challenging. These processes depend on upstream streamflow and sediment load. If included in the scenarios, these processes would vary in tandem with changes in upstream streamflow and sediment load driven by ER, rather than remaining consistent across different scenarios, making it complex and difficult to isolate the impact of ER. Our simplification ensures that differences between scenarios reflect the hydrological and sediment responses attributable solely to ER, although it limits direct comparability with real-world gauge observations. In practice, upstream reservoirs’ operations and socioeconomic water use reduce the actual streamflow and sediment load entering the Xiaolangdi Reservoir compared to our ER simulation (Supplementary Fig. 6). Although this results in lower average annual energy generation and sedimentation in reality, the relationship between accumulated sedimentation and energy generation suggests that the actual accumulated hydropower production may exceed our simulations (Supplementary Fig. 7). Second, large-scale land use change may indirectly influence hydropower potential by affecting precipitation, partially offsetting its impact on local water availability23,43. While this complex feedback between land use change and climate was not incorporated into our simulations—given prior findings that precipitation increase from vegetation greening was statistically insignificant in North China44—it remains an important consideration for future simulations in other regions. Third, sedimentation distributions, which affect reservoir storage, are shaped by complex hydrological and sediment processes, such as the movement of fine-grained sediment deposits45. Accurately modeling these processes over the long term remains challenging. Instead, we assumed that sediment initially accumulates at the reservoir’s lowest elevation and gradually increases over time. This assumption implies that sedimentation occurs first in the sediment storage of the Xiaolangdi Reservoir (mainly below the water level of 245 m, Supplementary Fig. 2) and enables us to simulate changes in the relationship between storage and forebay water level as sedimentation accumulates. Although idealized, this assumption aligns reasonably well with daily storage and water level data spanning 2006 to 2023 (“Methods”, Supplementary Fig. 8, Table 3). Fourth, sediment storage exhaustion was used as a benchmark to evaluate and compare total energy generation across scenarios; while it reduces reservoir flexibility and efficiency, power generation remains possible. Analysis of the relationship between accumulated sedimentation and energy generation (Supplementary Fig. 7) showed that, regardless of the storage threshold chosen as the benchmark, the scenario with ER consistently yielded higher energy output—generally exceeding 50% more—underscoring the robustness of our findings. These simplifications, while necessary for the scope of this study, may affect the precision of our results and should be addressed in future research to enhance accuracy and applicability.

In conclusion, this study revealed the positive impact of ER in the YRB on hydropower potential by integrating the opposing effects through sediment and runoff changes. It advances our understanding of the ecosystem-water-sediment-energy nexus and underscores the vital role of integrated watershed management in sustainable development. The insights from this study can be applied to other reservoirs worldwide, guiding PES practices in different watersheds.

Methods

Data sources

This study utilized various datasets to analyze and simulate the vegetation dynamics in the YRB, the streamflow and sediment concentration of the YR, and the regulation of the Xiaolangdi Reservoir (Supplementary Table 2). Meteorological variables, such as daily minimum, maximum, and mean temperatures, relative humidity, wind speed, solar radiation, and sunshine duration, were interpolated from data at 113 national meteorological stations distributed across or near the YRB, provided by the China Meteorological Administration (http://data.cma.cn). Precipitation data were sourced from the China Gauge-based Daily Precipitation Analysis (CGDPA), a 0.25° gridded daily dataset integrated from observations of 113 national meteorological stations and other undisclosed meteorological stations using the climatology-based optimal method detailed by Shen and Xiong46. Geographic and soil texture data for the YRB were obtained from the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) by Jarvis et al.47. and the soil information database by Shangguan et al.48. The border and DEM data for the Xiaolangdi Reservoir were sourced from Li et al.49. Soil hydraulic parameters were obtained from Dai et al.50. Vegetation type data were collected from the investigative vegetation maps of China (1:1,000,000)51,52 and the 30-m resolution Global Land Cover dataset53,54. ER construction records, including terracing, check dams, afforestation, and grass planting, were primarily collected from the Chinese National Water Resources Census (http://www.mwr.gov.cn/) and the Yellow River Conservancy Commission of the Ministry of Water Resources (http://www.yrcc.gov.cn/).

Comprehensive hydrological, ecological, and sediment datasets, such as Leaf Area Index (LAI), Gross Primary Productivity, evapotranspiration, etc., were employed for model calibration and validation. Specifically, this study aimed to simulate streamflow and sediment load without the influence of other reservoirs’ operations and socio-economic water use upstream of the Xiaolangdi Reservoir. To this end, hydrological processes were first calibrated and validated using natural streamflow sequences from 1982 to 2000, obtained from the third national water resources survey and evaluation conducted by the Ministry of Water Resources (http://www.mwr.gov.cn/). This dataset restored surface water consumption, reservoir operations, and evaporation and leakage caused by the reservoirs, based on gauge observations. After establishing reliable hydrological modeling, sediment load was then calibrated and validated using observed sediment load from 1960 to 1980, during which most reservoirs on the YR had not yet been constructed, to approximate sediment processes without the influence of reservoir sedimentation. Observed sediment load data, along with data on the regulation and storage of the Xiaolangdi Reservoir, were provided by the Yellow River Conservancy Commission of the Ministry of Water Resources (http://www.yrcc.gov.cn/). More details on these datasets and their processing for simulating and validating vegetation, runoff, and sediment can be found in Yan et al.19,55 and Yang et al.56.

Simulation of streamflow and sediment load of the YR

The Geomorphology-Based Eco-hydrological Model with Soil and Water Conservation (GBEHM-SWC), a distributed physically-based eco-hydrological model, was used to simulate hydrological and sediment dynamics under varying vegetation cover and engineering conditions. The Geomorphology-Based Eco-hydrological Model (GBEHM)57, integrates a comprehensive set of processes, including hydrological dynamics governed by multi-layer soil water dynamics and energy balance, and erosion processes on hillslopes influenced by underlying surface characteristics and the hydraulic properties of overland flow. Furthermore, the model simulates kinematic wave flow routing and one-dimensional non-equilibrium sediment transport in river channels, offering a holistic representation of eco-hydrological interactions across both hillslope and channel systems (Supplementary Fig. 3). Developed from GBEHM, the GBEHM-SWC has been further refined to capture the impacts of soil and water conservation measures (ER measures in this study) on hydrological and sediment processes19. These measures are explicitly represented to depict their hydrological and sediment influences in the GBEHM-SWC. Terracing, afforestation, and grass planting on hillslopes first alter vegetation cover—a change observable through remote-sensed LAI data. They are also parameterized to enhance surface water storage capacity and reduce erosion capacity, integrated into hillslope hydro-sediment processes. Check dams in the river network system are conceptualized as unregulated reservoirs within each flow interval and incorporated into the flow routing process. These dams slow down discharge velocity or obstruct streamflow to facilitate sediment deposition, with stored water further infiltrating or evaporating. The outflow discharge and sediment concentration are calculated based on the simulated inflow and effective storage capacity of check dams, while dynamically tracking changes in water storage and sediment deposition within the dams. The parameterization scheme is detailed in Yan et al.19 and Yang et al.56. The GBEHM-SWC was extensively calibrated and validated in the YRB using monthly natural streamflow data from 38 hydrological stations and monthly observed sediment discharge data from 22 hydrological stations, demonstrating advanced performance with Nash-Sutcliffe efficiency coefficient (NSE) values for streamflow exceeding 0.8 (Supplementary Fig. 9) and NSE values for sediment discharge exceeding 0.5 at all mainstream stations (Supplementary Fig. 10).

Scenario analysis of ER impacts on energy generation of the Xiaolangdi Reservoir

We employed a single-factor fixed approach to assess the impacts of ER in the YRB on water and sediment changes of the YR and subsequently on the energy generation of the Xiaolangdi Reservoir. The scenario with ER represented current conditions with all restoration practices in place, while the scenario without ER reflected the pre-restoration landscape of the YRB (Supplementary Table 1). ER measures in the YRB, aimed at controlling severe soil erosion and reducing sediment transport, include terracing, afforestation, grass planting on hillslopes, and check dams in the river network. For the scenario with ER, actual ER measures from 2000 to 2019 and dynamic land use and LAI were used as inputs. The streamflow and sediment discharge into the Xiaolangdi Reservoir simulated by the GBEHM-SWC under this scenario showed high consistency with observed data from 2000 to 2019, with R2 values of 0.70 and 0.53 and NSE values of 0.67 and 0.45 for monthly natural streamflow and sediment discharge, respectively (Supplementary Figs. 11 and 12). In the scenario without ER, no ER measures were applied, and only the land use in 2000 and simulated natural vegetation dynamics based on the Biome-BGC model (Supplementary Fig. 13) were used as inputs.

The Biome-BGC model, a widely recognized tool for simulating vegetation ecophysiology, models the processes of energy, water, and carbon fluxes within terrestrial ecosystems and provides key variables like LAI. We used version 4.2 of the Biome-BGC model, which has been extensively validated and widely employed in previous studies58,59, to simulate the vegetation condition of the YRB from 2000 to 2019, accounting only for climatic influences and excluding human interference. The model was adapted to simulate vegetation dynamics on a 10-km grid cell scale, matching the spatial resolution of the eco-hydrological model. The NSE values for simulated monthly LAI for all vegetation types at five sample grids exceeded 0.85, averaging 0.90, demonstrating satisfactory performance. After calibrating the model parameters at the sample points, the Biome-BGC model was run for each grid. More details on the simulation process can be found in Yan et al.55.

Simulation of energy generation and sedimentation of the Xiaolangdi Reservoir

A random forest model was used to simulate the regulation rules of the Xiaolangdi Reservoir. Reservoir operation is typically based on its storage, purpose, and inflow60, and shows monthly variations attributed to seasonal inflow patterns, downstream irrigation water demands, and hydropower generation necessities61. Therefore, the monthly outflow released from the reservoir (Ot, m3/s) was predicted using the storage at the beginning of the month (St, 108 m3), the inflow of the month (It, m3/s) and the month before (It-1, m3/s), and the month itself (t) as explanatory variables. We used monthly regulation data of the Xiaolangdi Reservoir from June 2006 to March 2024 to build the model, with 80% of the data randomly selected for training and the remaining 20% for validation. The model showed high accuracy with NSE values of 0.88, 0.77, and 0.86 and R2 values of 0.94, 0.79, and 0.91 for training, validation, and all data, respectively (Supplementary Figs. 14 and 15).

Using the regulation rules model and hydrological and sediment output from the GBEHM-SWC, we simulated the regulation of the Xiaolangdi Reservoir under scenarios with and without ER (Supplementary Fig. 16). The simulation began in January 2002, following the completion of the reservoir construction at the end of 2001, with an initial recorded storage of 4.5 billion m3. The simulated storage was calculated using the water balance of the reservoir as Eq. (1):

$${S}_{t+1}={S}_{t}+({I}_{t}-{O}_{t}){{\cdot }}\varDelta t/{10}^{8}$$
(1)

where Δt is the total seconds of the corresponding month for each step size. To better reflect real-world reservoir regulation, we added two constraints: the maximum simulated storage should be less than the storage capacity at the simulated time, calculated as the initial storage capacity (12.65 billion m3) minus the accumulated sedimentation; the minimum simulated storage should be higher than the lowest values observed in the actual regulation of the reservoir (0.05 billion m3).

Based on simulated sediment concentration of the inflow into the Xiaolangdi Reservoir and outflow and storage of the reservoir, monthly energy generation and sedimentation of the Xiaolangdi Reservoir were calculated. According to a previous study62, the electricity generated by the Xiaolangdi Reservoir was calculated as Eq. (2):

$${E}_{t}=\, K{{\cdot }}\,{O{{\hbox{'}}}}_{t}\,{{\cdot }}\,({H}_{t}-{H}_{0})\,{{\cdot }}\, \varDelta t/3600$$
(2)

where Et is the generated electricity (kWh); K is the power generation coefficient (8.5 for large hydropower stations)62; O’t is the water released for power generation, generally equal to Ot but capped at 1800 m3/s, the maximum flow rate that the turbine can handle62; Ht is the average fore-bay water level during month t, and H0 is the turbine layer elevation (129 m)62. Ht was calculated based on the average storage during month t and a quadratic polynomial function fitted to the relationship curve between storage and fore-bay water level of the Xiaolangdi Reservoir. The initial relationship was derived using DEM data (Supplementary Figs. 8 and 17). This relationship evolves with sedimentation in the Xiaolangdi Reservoir. Given the challenges in accurately predicting sedimentation distributions, we assumed that sediment initially accumulates at the reservoir’s lowest elevation and gradually increases over time. Consequently, as the reservoir silts up, changes in the storage–water level relationship curve are primarily reflected in shifts in the intercept. The relationship function was recalculated annually during the simulation. We validated our assumption using daily storage and water level data of the Xiaolangdi Reservoir spanning 2006–2023. Data points from different years demonstrated similar curvature to the initial relationship (Supplementary Fig. 8a). Additionally, the fitting functions for the storage–water level relationship, derived from the initial function with adjusted intercepts for all years, showed high R2 values exceeding 0.99 (Supplementary Table 3). The changes in the fitting function intercept and the changes in actual accumulated reservoir sedimentation over different years exhibit an inverse relationship (Supplementary Fig. 8b), indicating that the reduction in storage at any given water level is nearly equal in magnitude to the sedimentation during those years. This consistency supports the robustness of our assumption that sediment initially accumulates at the reservoir’s lowest elevation.

To calculate reservoir sedimentation, the sediment concentration of the outflow was derived using an empirical formula for the sediment delivery ratio of the Xiaolangdi Reservoir established in a previous study63. This formula characterizes the relationship between the sediment delivery ratio and its influencing factors—including average storage, outflow, inflow, and inflow sediment concentration of the Xiaolangdi Reservoir—as Eq. (3) and exhibits a high fitting accuracy (R2 = 0.86) with observed data from 2000 to 2015 (Supplementary Fig. 18):

$${\eta }_{t}=1.493\,{{\cdot }}\,{({\bar{S}}_{t}/{O}_{t})}^{-1.008}\,{{\cdot }}\,{({I}_{t}/{O}_{t})}^{-0.278}\,{{\cdot }}\,{{{ScI}}_{t}}^{-0.404}$$
(3)

therefore, the sediment concentration of the outflow can be expressed as Eq. (4):

$${{ScO}}_{t}={\eta }_{t}\,{{\cdot }}\,{{ScI}}_{t}\,{{\cdot }}\,{I}_{t}/{O}_{t}$$
(4)

where ηt is the sediment delivery ratio; \({\bar{S}}_{t}\) is the average storage during month t (108 m3); ScIt and ScOt are the sediment concentrations of the inflow and outflow, respectively (kg/m3). Using the mass conservation equation, the reservoir sedimentation volume can be calculated as Eq. (5):

$${\varDelta V}_{t}=({I}_{t}\,{{\cdot }}\,{{ScI}}_{t}-{O}_{t}{{\cdot }}\,{{ScO}}_{t})\,{{\cdot }}\varDelta t/\rho$$
(5)

where ΔVt is the sedimentation volume of the reservoir in month t (m3); ρ is the dry density of the bed sediment, taken as 1200 kg/m3.