Schuddeboom et al. (2026) EGGS-WG: an open source global gridded stochastic weather generator derived from ERA5-Land
Identification
- Journal: Geoscientific model development
- Year: 2026
- Date: 2026-11-16
- Authors: Alex J. Schuddeboom, Christian Zammit, David R. Plew, Piet Verburg, Aidin Jabbari
- DOI: 10.5194/gmd-19-10863-2026
Research Groups
- School of Physical and Chemical Sciences, University of Canterbury, Christchurch, New Zealand
- Earth Sciences New Zealand, Christchurch, New Zealand
- School of Geography, Environment and Earth Sciences, Victoria University of Wellington, Wellington, New Zealand
Short Summary
This paper introduces EGGS-WG, a new open-source global gridded stochastic weather generator derived from ERA5-Land data, capable of simulating precipitation, air temperature, and dew point temperature at hourly resolution for any terrestrial region. Validation across diverse climates demonstrates its ability to accurately reproduce observed weather patterns, though with some limitations in simulating extreme precipitation and dew point temperature variability, particularly in monsoon-affected areas.
Objective
- To develop and validate EGGS-WG, an open-source, low-resource, and portable global gridded stochastic weather generator.
- To simulate precipitation, air temperature, and dew point temperature (and derived relative humidity) at hourly resolution for any terrestrial region of the planet.
Study Configuration
- Spatial Scale: Global terrestrial coverage at 0.1° × 0.1° grid resolution. Validation performed over five climatically distinct regions: West Coast (New Zealand), Canterbury (New Zealand), Taupō (New Zealand), Australia, and New Guinea. Additional tests on North America and the Indian Subcontinent.
- Temporal Scale: Hourly resolution for temperature, daily for precipitation occurrence and amount. Simulations of 110 years duration, based on ERA5-Land data from 1950 to 2022.
Methodology and Data
- Models used:
- ERA5-land Global Gridded Stochastic Weather Generator (EGGS-WG), an open-source model.
- Based on Richardson-style stochastic weather generators.
- Precipitation occurrence: First-order two-state Markov chain with spatially correlated random numbers.
- Precipitation amount: Three-parameter mixed exponential distribution (with future plans for gamma distribution), fitted using SciPy optimization (Nelder-Mead and Powell algorithms).
- Air and dew point temperature: Modified residual approach using conditioned Fourier series (daily and hourly) and a first-order autoregressive model (AR(1)) with lag-1 auto-correlation and cross-correlation.
- Spatial coherence of random numbers: Iterative approach based on Brissette et al. (2007).
- Relative humidity: Calculated from simulated air temperature and dew point temperature using the Tetens equation.
- Data sources:
- ERA5-Land reanalysis dataset (2 m air temperature, 2 m dew point temperature, precipitation, geopotential) at 0.1° × 0.1° spatial resolution and hourly temporal resolution.
- Copernicus Climate Change Service (C3S) Climate Data Store.
Main Results
- EGGS-WG accurately reproduces monthly average precipitation occurrence rates and amounts, capturing seasonal cycles and spatial correlations across all tested regions.
- Precipitation spell lengths (wet and dry days) are generally well represented, though the model shows some issues with very long streaks in the New Guinea region.
- The simulation of extreme precipitation values (above the 99th percentile) shows deviations: generally overestimating in most regions but underestimating in New Guinea.
- Seasonal cycles and diurnal variability of 2 m air temperature and 2 m dew point temperature are very well captured, with strong agreement between simulations and observations.
- Spatial correlations for 2 m air temperature are accurately reproduced. Spatial correlations for 2 m dew point temperature are slightly underestimated in simulations compared to observations.
- Cross-correlation between 2 m air temperature and 2 m dew point temperature shows consistent underestimation of the observed relationship in most regions, with complex behavior in New Guinea.
- The model ensemble generates realistic time series with appropriate variability, although the New Guinea region exhibits reduced variability in monthly anomalies for all variables compared to observations.
- Model performance is notably weaker in regions strongly impacted by monsoons (e.g., New Guinea, some Indian Subcontinent regions), suggesting climate-specific limitations.
Contributions
- Development and public release of EGGS-WG, the first open-source global gridded stochastic weather generator.
- Enables simulation for any terrestrial region of the planet, moving beyond single-site or regionally constrained models.
- Provides hourly temporal resolution for temperature simulations, an advancement over many existing daily models.
- Includes simulation of dew point temperature and derived relative humidity, expanding the range of output variables.
- Designed for low computational resource requirements and portability, enhancing accessibility for a wide range of end-users.
- Utilizes the high-resolution ERA5-Land reanalysis dataset as its basis.
Funding
- New Zealand Ministry of Business, Innovation and Employment (MBIE) via the 2022 Endeavour Fund – Smart Ideas (contract C01X2205).
- Aotearoa New Zealand Tāwhia te Mana Research Fellowships, administered by the Royal Society Te Apārangi (contract MTP-UOC2401).
Citation
@article{Schuddeboom2026EGGSWG,
author = {Schuddeboom, Alex J. and Zammit, Christian and Plew, David R. and Verburg, Piet and Jabbari, Aidin},
title = {EGGS-WG: an open source global gridded stochastic weather generator derived from ERA5-Land},
journal = {Geoscientific model development},
year = {2026},
doi = {10.5194/gmd-19-10863-2026},
url = {https://doi.org/10.5194/gmd-19-10863-2026}
}
Original Source: https://doi.org/10.5194/gmd-19-10863-2026