Benestad (2026) A simple weather generator that converts statistical information from downscaled global climate models to sequences of daily precipitation and temperature
Identification
- Journal: Geoscientific model development
- Year: 2026
- Date: 2026-11-13
- Authors: Rasmus Eduard Benestad
- DOI: 10.5194/gmd-19-10849-2026
Research Groups
The Norwegian Meteorological Institute
Short Summary
This paper introduces and evaluates a simple weather generator, implemented in the open-source R-package 'esd', designed to convert downscaled global climate model statistics (wet-day frequency and mean precipitation, and temperature mean and standard deviation) into daily sequences of precipitation and temperature for use in impact models.
Objective
- To develop and evaluate a simple, computationally efficient weather generator that links downscaled statistical information from global climate models (specifically wet-day frequency and mean precipitation, and temperature mean and standard deviation) to daily sequences of precipitation and temperature required by impact models, such as hydrological models.
Study Configuration
- Spatial Scale: Single-site evaluations across diverse local climates in Norway (Bjørnholt, Oslo, Tromsø), Ghana (Akosombo, Volta catchment), Romania (Cluj-Napoca), Tanzania (Arusha), and Italy (Naples).
- Temporal Scale: Daily time series generation and evaluation. Data records used for evaluation spanned various periods, e.g., 1990–2025 for Ghana, 1950–2020 for Romania.
Methodology and Data
- Models used:
- Simple weather generator for daily precipitation (based on wet-day frequency f_w and wet-day mean precipitation μ, exponential distribution for amounts, Fourier transform for scrambling).
- Simple weather generator for daily temperature (based on mean and standard deviation, normal distribution for anomalies, phase-scrambled Fourier series).
- Implemented in the open-source R-package "esd" (version 1.11.30 or higher).
- Designed to provide input for hydrological models like SWAT and HBV.
- Data sources:
- Rain gauge observations from Norway (Norwegian Meteorological Institute), Ghana (Ghana Meteorological Agency), Romania (ECA&D).
- Global Historical Climatology Network-Daily (GHCND) for Arusha (Tanzania) and Naples (Italy).
- Climate Hazards Infrared Precipitation with Stations (CHIRPS) for Volta catchment (Ghana).
Main Results
- The precipitation weather generator generally reproduced daily precipitation amounts and climatologies (total precipitation, wet-day frequency, wet-day mean precipitation) well across different climates.
- Quantile-quantile plots showed good agreement for daily precipitation amounts, with modest deviations for extreme values (> 60 mm).
- Root-mean-square errors (RMSE) for climatologies were low: 4.0–6.8 mm for total precipitation, 0.26–0.48 mm/d for wet-day mean precipitation, and 0.011–0.016 for wet-day frequency.
- A limitation was the underestimation of long dry and wet spell durations (e.g., Kolmogorov-Smirnov D = 0.13 for dry spells and D = 0.20 for wet spells in Bjørnholt, indicating a partial match).
- The temperature weather generator reproduced observed daily temperatures reasonably well, though with some bias for the lowest temperatures (e.g., < 28 °C in Akosombo).
- The generators successfully linked downscaled statistical parameters to daily sequences, demonstrating their utility for impact models.
Contributions
- Development of a novel, simple, and computationally efficient weather generator specifically designed to bridge the gap between empirical-statistical downscaling (ESD) of climate parameters (wet-day frequency, wet-day mean precipitation, temperature mean/standard deviation) and the daily input requirements of impact models (e.g., hydrological models).
- Implementation as an open-source R-package ("esd"), making it readily available and reproducible for the scientific community.
- Comprehensive evaluation across diverse climatic regions (Norway, Ghana, Romania) demonstrating its ability to reproduce salient daily weather characteristics and climatologies.
- Addresses a critical need for generating stochastic daily weather sequences from downscaled climate statistics, which was previously a gap in existing weather generators.
Funding
- EU-SPRINGS project.
- Horizon 2020 (grant no. 101137255).
Citation
@article{Benestad2026simple,
author = {Benestad, Rasmus Eduard},
title = {A simple weather generator that converts statistical information from downscaled global climate models to sequences of daily precipitation and temperature},
journal = {Geoscientific model development},
year = {2026},
doi = {10.5194/gmd-19-10849-2026},
url = {https://doi.org/10.5194/gmd-19-10849-2026}
}
Original Source: https://doi.org/10.5194/gmd-19-10849-2026