Nguyen et al. (2026) Elevation dependent spatial interpolation of hourly rainfall for accurate flood inundation modelling

Identification

• Journal: Hydrology and earth system sciences
• Year: 2026
• Date: 2026-01-09
• Authors: Chi Nguyen, Jai Vaze, Cherry May R. Mateo, Michael F. Hutchinson, Jin Teng
• DOI: 10.5194/hess-30-45-2026

Research Groups

• CSIRO Environment, Australian Capital Territory, Canberra, Australia (Chi Nguyen, Jai Vaze, Cherry May R. Mateo, Jin Teng)
• Australian National University, Australian Capital Territory, Canberra, Australia (Michael F. Hutchinson)

Short Summary

This study develops and evaluates the Commonwealth Scientific and Industrial Research Organisation Hourly Rainfall (CHRain) dataset, providing hourly and 1 km gridded rainfall surfaces using elevation-dependent thin-plate spline interpolation. The CHRain dataset demonstrates superior performance in capturing sub-grid rainfall distribution and high rainfall values compared to existing Australian gridded datasets, making it suitable for accurate flood inundation modeling.

Objective

• To develop and demonstrate a method for generating a high-resolution (hourly, 1 km) gridded rainfall dataset (CHRain) using elevation-dependent thin-plate spline interpolation, suitable for accurate flood inundation modeling, and to evaluate its performance against existing datasets.

Study Configuration

• Spatial Scale: Richmond River catchment, New South Wales, Australia (approximately 7025 km²), with an analysis area of 30389 km². Gridded rainfall surfaces at 1 km resolution. Optimal Digital Elevation Model (DEM) horizontal resolution of approximately 5 km.
• Temporal Scale: Hourly rainfall surfaces, covering the period from 30 January 2007 to 31 December 2022.

Methodology and Data

• Models used: Thin-plate spline interpolation (ANUSPLIN Version 4.4).
• Data sources:
  • Hourly and daily point rainfall measurements from the Bureau of Meteorology (BoM) and Water New South Wales Corporation (WaterNSW).
  • Radar images (used for disaggregation guidance and comparison in occurrence analysis).
  • LiDAR Digital Elevation Model (DEM) from Geosciences Australia (5 m resampled to 1 km averaged).
  • Existing gridded datasets for comparison: BARRA-SY (1.5 km hourly), ANUClimate (1 km daily), and AGCD (5 km daily).

Main Results

• The CHRain dataset successfully generated hourly and 1 km gridded rainfall surfaces for the Richmond River catchment from 2007 to 2022.
• The performance of the spline interpolation significantly improved with the inclusion of elevation data, particularly for larger rainfall events.
• Optimal topographic parameters were identified: a DEM focal distance of approximately 5 km and an elevation scaling parameter of 4000-5000.
• A stable spatial occurrence analysis effectively removed 0.26% (42193 out of 15737817) of false zeros from the input data, with a true accuracy of up to 98% on high rainfall days.
• During the 2017 flood event, CHRain achieved a high correlation coefficient of 0.949 against hourly gauges, significantly outperforming BARRA-SY (0.234) and radar data (0.154).
• The CHRain dataset demonstrated superior capability in representing sub-grid rainfall distribution, daily and hourly rainfall variation, and high rainfall values compared to existing Australian gridded datasets (BARRA-SY, ANUClimate, AGCD).
• The 24-hour total CHRain showed strong correlation with daily measurements at independent gauges (correlation coefficients of 0.935 for 2017 and 0.938 for 2022 flood events).
• CHRain captured greater sub-grid variability, with a rainfall range of 55 mm/d within a 5 km window, compared to ANUClimate (7.4 mm/d) and AGCD (0 mm/d).

Contributions

• Developed the CHRain dataset, a novel high-resolution (hourly, 1 km) gridded rainfall product specifically tailored for detailed hydrological/hydrodynamic modeling in Australia.
• Introduced and validated a robust automated method for detecting and removing false zeros in hourly rainfall data using spatial occurrence analysis, a critical improvement for high-resolution interpolation.
• Quantified the elevation dependence of hourly rainfall patterns, identifying optimal DEM parameters (5 km focal distance, 4000-5000 elevation scaling) for improved interpolation accuracy, particularly for intense rainfall.
• Demonstrated that locally generated hourly splines capture sub-grid rainfall variability and high rainfall values more effectively than existing daily gridded products and reanalysis data.
• Provided open-source Python scripts and a sample for generating CHRain, enhancing reproducibility and facilitating wider application of the method.

Funding

• Northern Rivers Resilience Initiative project (led by CSIRO)
• National Emergency Management Agency (NEMA)
• Commonwealth Contract – Services with Reference ID DCD100396

Citation

@article{Nguyen2026Elevation,
  author = {Nguyen, Chi and Vaze, Jai and Mateo, Cherry May R. and Hutchinson, Michael F. and Teng, Jin},
  title = {Elevation dependent spatial interpolation of hourly rainfall for accurate flood inundation modelling},
  journal = {Hydrology and earth system sciences},
  year = {2026},
  doi = {10.5194/hess-30-45-2026},
  url = {https://doi.org/10.5194/hess-30-45-2026}
}