Improved Precipitation Variability in the Community Atmosphere Model through Implementation of a Neural-Network Deep Convection Scheme
Abstract
Conventional deep convection schemes remain a primary source of uncertainty in global climate models, yet replacing them with data-driven alternatives while maintaining physical consistency and numerical stability is an open challenge. We present a hybrid configuration of the Community Atmosphere Model version 6 (CAM6), in which a fully connected deep neural network replaces the Zhang-McFarlane deep convection scheme (ZM), while radiation, turbulence, and other cloud processes are left unchanged. The network is trained on output from a storm-resolving global model - the global System for Atmospheric Modeling (gSAM) - using a coarse-graining approach to derive subgrid-scale convective fluxes and tendencies. This data-driven parameterization is ported to CAM6 and found to run stably for more than a decade while conserving energy and moisture. It produces a broadly realistic climatology without retuning the remaining parameterizations. Relative to the default Zhang-McFarlane scheme, the neural network parameterization substantially improves some long-standing model deficiencies. In particular, the biases in light precipitation and extreme precipitation are markedly reduced, and the diurnal cycle of tropical land precipitation shifts from a spurious midday peak over many regions to the observed late-afternoon peak. These results support targeted ML-based convection schemes as a viable path toward reducing systematic model errors in next-generation climate models.
Related
- Climatology and trends of extreme precipitation in France: evaluation of an explicit-convection regional climate model
- Climate models lack decadal variability in precipitation because of a weak atmospheric response to sea surface temperature anomalies
- Rainfall isotopic variability in northern Central America driven by regionally-organized deep atmospheric convection and airmass rainout
- Effect of Middle Atmosphere density variability on re-entering objects
- Design and trial implementation of a continental-scale, kilometre-resolution hourly precipitation analysis for Australia using satellite, radar and gauges
BibTeX
@misc{levine2026improved,
title = {{Improved Precipitation Variability in the Community Atmosphere Model through Implementation of a Neural-Network Deep Convection Scheme}},
author = {Xavier J. Levine and Griffin Mooers and Judith Berner and Jack W. Atkinson and Marat F. Khairoutdinov and Janni Yuval and Dominic Orchard and Paul A. O'Gorman},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.20405},
eprint = {2610.20405},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.20405},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Improved Precipitation Variability in the Community Atmosphere Model through Implementation of a Neural-Network Deep Convection Scheme
AU - Xavier J. Levine
AU - Griffin Mooers
AU - Judith Berner
AU - Jack W. Atkinson
AU - Marat F. Khairoutdinov
AU - Janni Yuval
AU - Dominic Orchard
AU - Paul A. O'Gorman
PY - 2026
DA - 2026/10/20
PB - arXiv
DO - 10.48550/arxiv.2610.20405
UR - https://arxiv.org/abs/2610.20405
AB - Conventional deep convection schemes remain a primary source of uncertainty in global climate models, yet replacing them with data-driven alternatives while maintaining physical consistency and numerical stability is an open challenge. We present a hybrid configuration of the Community Atmosphere Model version 6 (CAM6), in which a fully connected deep neural network replaces the Zhang-McFarlane deep convection scheme (ZM), while radiation, turbulence, and other cloud processes are left unchanged. The network is trained on output from a storm-resolving global model - the global System for Atmospheric Modeling (gSAM) - using a coarse-graining approach to derive subgrid-scale convective fluxes and tendencies. This data-driven parameterization is ported to CAM6 and found to run stably for more than a decade while conserving energy and moisture. It produces a broadly realistic climatology without retuning the remaining parameterizations. Relative to the default Zhang-McFarlane scheme, the neural network parameterization substantially improves some long-standing model deficiencies. In particular, the biases in light precipitation and extreme precipitation are markedly reduced, and the diurnal cycle of tropical land precipitation shifts from a spurious midday peak over many regions to the observed late-afternoon peak. These results support targeted ML-based convection schemes as a viable path toward reducing systematic model errors in next-generation climate models.
ER -