Tan et al. (2025) Evaluating multi-source precipitation data for streamflow simulation using the SWAT model in the Alpine Manas River Basin, Northwest China

Identification

• Journal: Scientific Reports
• Year: 2025
• Date: 2025-12-15
• Authors: Li Tan, Chanjuan Zan, Tie Liu, Jie Xiong, Aoxiang Zhang
• DOI: 10.1038/s41598-025-27391-1

Research Groups

• School of Geophysics, Chengdu University of Technology, Chengdu, China
• School of Geographical Sciences, China West Normal University, Nanchong, Sichuan, China
• State Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi, China
• College of Geoinformatics, Zhejiang University of Technology, Hangzhou, China

Short Summary

This study evaluated the accuracy of four multi-source precipitation products (CMFD, MSWEP, ERA5-Land, IMERG) in Xinjiang's arid regions and their applicability for streamflow simulation in the alpine Manas River Basin using the SWAT model. While CMFD showed optimal precipitation performance, MSWEP and ERA5-Land achieved superior streamflow simulation accuracy, indicating that meteorological superiority does not guarantee hydrological efficacy.

Objective

• To compare and analyze the accuracy of four high-spatiotemporal-resolution precipitation products (CMFD, MSWEP, ERA5-Land, IMERG) in the arid regions of Xinjiang, Northwest China.
• To explore the applicability and adaptability of these multi-source precipitation products for streamflow simulation in the alpine Manas River Basin using the SWAT hydrological model.

Study Configuration

• Spatial Scale: Alpine Manas River Basin (MRB), Northwest China. Geographical extent: 43º04’N to 43º58’N and 85º00’E to 86º15’E, with a total area of 5109.3 km². Elevations range from 850 m to 5138 m.
• Temporal Scale:
  • Common analysis period: June 1, 2000, to December 31, 2018.
  • SWAT model warm-up period: 2000–2001.
  • SWAT model calibration period: 2002–2012.
  • SWAT model validation period: 2013–2018.

Methodology and Data

• Models used:
  • SWAT (Soil and Water Assessment Tool) hydrological model.
  • SWAT-CUP software with SUFI-2 algorithm (for sensitivity analysis and calibration).
  • SPAW program (for estimating soil properties).
  • Williams equation (for calculating the erodibility factor USLE_K).
• Data sources:
  • Precipitation Products:
    • Integrated Multi-satellite Retrievals for GPM (IMERG) V06 (0.1º spatial resolution, 1-day temporal resolution).
    • Multi-Source Weighted-Ensemble Precipitation (MSWEP) V2.8 (0.1º spatial resolution, 3-hour temporal resolution, daily aggregated).
    • China Meteorological Forcing Dataset (CMFD) 1.6 (0.1º spatial resolution, 3-hour temporal resolution, daily aggregated).
    • Fifth Generation European Centre for Medium-Range Weather Forecasts Reanalysis for Land (ERA5-Land) (0.1º spatial resolution, 1-hour temporal resolution, daily aggregated).
  • Station Data:
    • Daily precipitation data from 37 meteorological stations across Xinjiang (China Meteorological Administration).
    • Meteorological inputs for SWAT (precipitation, maximum/minimum temperature, wind speed, relative humidity, sunshine duration) from a single ground station (ID 51356) in the upper basin.
    • Daily streamflow data from the KSWT Hydrological Station (Xinjiang Uygur Autonomous Region Hydrological Bureau) for model calibration and validation.
  • Spatial Data:
    • Digital Elevation Model (DEM): ASTER GDEM V2, 30 m resolution (Geospatial Data Cloud).
    • Land Use and Land Cover (LULC): Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, 30 m resolution (2010 data).
    • Soil Database: Harmonized World Soil Database (HWSD), 1 km scale (Food and Agriculture Organization).

Main Results

• Precipitation Product Performance (Xinjiang arid region):
  • CMFD exhibited the best overall performance on a daily scale, with the highest correlation coefficient (CC = 0.54), smallest root mean square error (RMSE = 1.64 mm), smallest mean absolute error (MAE = 0.45 mm), and good precipitation detection ability (Probability of Detection, POD = 0.74). It also showed the most stable performance across different precipitation intensities.
  • MSWEP and ERA5-Land performed comparably, ranking second. MSWEP had a smaller daily-scale bias (0.10) than CMFD (0.15). ERA5-Land showed a substantial overall overestimation of precipitation (approximately 45.6%).
  • IMERG performed poorly on a daily scale, with the lowest CC (0.06), highest RMSE (2.32 mm), highest MAE (0.71 mm), lowest POD (0.28), and highest False Alarm Rate (FAR = 0.76). It underestimated monthly totals by 39% to 45.7% in spring and autumn, while overestimating by 6.43% to 35.2% in July and August.
• Streamflow Simulation Performance (Manas River Basin):
  • The SWAT model driven by ground station precipitation achieved excellent performance (calibration: Nash-Sutcliffe Efficiency, NSE = 0.77, Percent Bias, PBIAS = -2.16%; validation: PBIAS = 8.49%).
  • MSWEP-driven daily streamflow simulations yielded optimal results, achieving NSE > 0.65 and PBIAS < 15% during both calibration and validation periods. It significantly improved high flow (approximately 50%) and median/low flow (approximately 33.3%) simulations compared to station-based input.
  • CMFD and ERA5-Land-driven simulations consistently overestimated streamflow by 12.79% to 27.07% in both periods.
  • IMERG-driven simulations significantly underestimated streamflow (average 67.16%), with NSE less than 0.3 and Root Mean Square Error to Observation Standard Deviation Ratio (RSR) greater than 0.8.
  • Crucially, the study found that meteorological superiority (e.g., CMFD's precipitation accuracy) did not directly translate to superior hydrological simulation efficacy, as MSWEP and ERA5-Land achieved better streamflow simulation accuracy in the MRB.

Contributions

• Provided a comprehensive evaluation of four widely used high-spatiotemporal-resolution precipitation products in the data-scarce arid and alpine region of Xinjiang, Northwest China.
• Quantified the adaptability of these multi-source precipitation products for streamflow simulation in the alpine Manas River Basin using the SWAT model.
• Established fundamental benchmarks for hydro-meteorological research and offered valuable references for streamflow simulation and water resource management in similar data-scarce alpine basins.
• Highlighted the critical finding that a precipitation product's superior meteorological accuracy does not guarantee its superior performance in hydrological modeling, underscoring the necessity for product-specific calibration and evaluation for hydrological applications.

Funding

• Sichuan Science and Technology Program (2023YFN0022)
• Key R&D Program of Xinjiang Uygur Autonomous Region (Grant No. 2022B03021-2)

Citation

@article{Tan2025Evaluating,
  author = {Tan, Li and Zan, Chanjuan and Liu, Tie and Xiong, Jie and Zhang, Aoxiang},
  title = {Evaluating multi-source precipitation data for streamflow simulation using the SWAT model in the Alpine Manas River Basin, Northwest China},
  journal = {Scientific Reports},
  year = {2025},
  doi = {10.1038/s41598-025-27391-1},
  url = {https://doi.org/10.1038/s41598-025-27391-1}
}