Yang et al. (2025) Topographical influence on kilometer-scale hourly precipitation prediction during the 2021 Zhengzhou flood

Identification

• Journal: Natural Hazards
• Year: 2025
• Date: 2025-12-26
• Authors: Linyun Yang, Haoming Chen
• DOI: 10.1007/s11069-025-07790-3

Research Groups

• State Key Laboratory of Severe Weather, The Chinese Academy of Meteorological Sciences, Beijing, China

Short Summary

This study investigates the influence of topographical factors on 3-kilometer hourly precipitation forecasts from the China Meteorological Administration’s Mesoscale Weather Numerical Forecast System (CMA-MESO) during the 2021 Zhengzhou flood using spatio-temporal geographically weighted regression (GTWR). It finds that CMA-MESO overestimates the topographical impact on the spatial distribution of precipitation while underestimating its influence on temporal variation, with near-surface temperature (ST) being a dominant factor for model bias.

Objective

• To investigate how topographical factors influence 3-kilometer hourly precipitation forecasts from the CMA-MESO model during the 2021 Zhengzhou flood.
• To quantify how the influences of topographic factors evolve over time during an extreme precipitation event.
• To determine the extent to which topographic factors contribute to the model bias of a kilometer-scale model for an extreme event.

Study Configuration

• Spatial Scale:
  • Study Area: Henan Province, China (31.38° N – 36.37° N, 110.35° E – 116.65° E).
  • Observational Stations: 963 high-quality stations within 31° N – 37° N, 110° E – 117° E.
  • Model Resolution: 3 kilometers (CMA-MESO).
  • Topographical Data Resolution: 30 meters (SRTM-V3.0).
  • Reanalysis Data Resolution: 25 kilometers (ERA5).
  • GTWR Minimum Spatial Bandwidth: 40000 meters.
• Temporal Scale:
  • Period: 00:00 BJT on July 17th to 23:00 BJT on July 22nd, 2021 (the “7.20” flood event).
  • Resolution: Hourly precipitation, 2-minute averaged wind data.
  • Forecast Period: 12–36 hours from 00:00 UTC initial forecast time.

Methodology and Data

• Models used:
  • China Meteorological Administration’s Mesoscale Weather Numerical Forecast System (CMA-MESO).
    • Parameterization schemes: WSM6 cloud microphysics, RRTM longwave radiation, Dudhia shortwave radiation, Monin-Obukhov near-surface layer, Noah land surface model, MRF boundary layer scheme.
  • Spatio-temporal Geographically Weighted Regression (GTWR) model.
• Data sources:
  • Observation: Hourly station dataset from the China Integrated Meteorological Information Sharing System (CIMISS). Missing values imputed using Kriging with External Drift (KED).
  • Topography: Shuttle Radar Topography Mission (SRTM-V3.0) product (30-meter resolution) for observed topography. CMA-MESO uses Global 30 Arc-Second Elevation data (GTOPO30) (~1-kilometer resolution).
  • Model Output: CMA-MESO (3-kilometer, hourly) for precipitation, 2-meter temperature, 10-meter wind predictions, and wind, specific humidity, air pressure, and temperature at 19 levels.
  • Reanalysis: ERA5 (25-kilometer, hourly) for air temperature, pressure, specific humidity, and wind components at 37 pressure levels.
  • Topographical factors analyzed: Topographical Elevation (TE), Topographical Slope (TS), Near-Surface Temperature (ST), Near-Surface Wind Speed (WDS), Prevailing Wind-Direction Effect Index (PWEI). Topographical Relief (TR) was excluded due to multicollinearity (VIF > 10).
  • Thermodynamic parameters: Convective Available Potential Energy (CAPE) and Convective Inhibition (CIN) calculated from ERA5 and CMA-MESO outputs.

Main Results

• CMA-MESO simulated precipitation center was shifted eastward compared to observations, resulting in a negative rainfall bias of -0.1 meters over most of Henan Province.
• CMA-MESO overestimates topographical elevation in western Henan and underestimates it in the east, leading to biases in other topographical factors (e.g., smoothed topographical slope with a negative bias of -0.017 to -0.087 radians).
• GTWR analysis showed that topographical forcing accounts for 28.6% of the variance in simulated precipitation, compared to 14.6% for observed rainfall.
• The CMA-MESO model suggests topographical forcing influences simulated precipitation for a more extended duration (spatial-temporal distance ratio (tau) of 1.5) compared to observations (tau of 3.6), indicating less pronounced temporal variability of topographical effects in the model.
• CMA-MESO overestimates the impact of topographical forcing on the spatial distribution of precipitation, with approximately 22% of the precipitation bias attributable to this overestimation.
• Near-surface temperature (ST) and near-surface wind speed (WDS) were identified as the dominant topographical factors for most observational stations, with ST being particularly important in areas experiencing heavy rainfall (accumulated precipitation exceeding 0.15 meters).
  • Observed ST coefficients ranged from -0.00078 to -0.00023 meters per Kelvin.
  • Observed WDS coefficients ranged from 0.00012 to 0.00040 seconds.
• CMA-MESO overestimates the effect of PWEI and TS in western Henan and simulates a spurious positive correlation between TS and precipitation in southern and western Henan, contributing to drier biases.
• The temporal variability of topographical effects is more evident in observational data than in CMA-MESO. The influence of ST on precipitation is significantly suppressed over time in observations, while CMA-MESO exhibits nearly stable effects.
• The negative ST bias in CMA-MESO, coupled with an overestimation of the Taihang Mountains' flow-blocking effect, leads to an underestimation of CAPE and CIN, ultimately causing the eastward shift of the rainfall center.
• CMA-MESO shows lower CAPE values and low CIN values across the entire study region compared to ERA5, which displays high CIN exceeding 200 J kg⁻¹ in some areas.
• Water vapor flux divergence in CMA-MESO is underestimated northwest of Henan Province compared to ERA5, consistent with an overestimation of the flow-blocking effect.

Contributions

• Expands the application of GTWR to quantitatively assess the temporally varying topographical influence on extreme precipitation events, addressing a significant gap in understanding non-stationary topographical forcing.
• Provides quantitative measures of the topographical influence on extreme precipitation, distinguishing it from previous qualitative or less detailed studies.
• Offers critical insights for improving planetary boundary layer (PBL) and orographic parameterization schemes in kilometer-scale numerical weather prediction systems. Specifically, it recommends balancing turbulent vertical mixing driven by buoyancy and mechanically driven mixing, and capturing the temporal variability of topographical forcing.
• Identifies thermal forcing associated with topography (e.g., near-surface temperature) as the dominant factor contributing to hourly precipitation prediction bias during the Zhengzhou flood, rather than mechanical forcing (e.g., terrain height and shape).

Funding

• National Natural Science Foundation of China (U2142214, 42105065, 42275015)
• Joint Funds of the National Key R&D Program of China (2021YFC3000904)

Citation

@article{Yang2025Topographical,
  author = {Yang, Linyun and Chen, Haoming},
  title = {Topographical influence on kilometer-scale hourly precipitation prediction during the 2021 Zhengzhou flood},
  journal = {Natural Hazards},
  year = {2025},
  doi = {10.1007/s11069-025-07790-3},
  url = {https://doi.org/10.1007/s11069-025-07790-3}
}