AtmosMJ: Revisiting Gating Mechanism for AI Weather Forecasting Beyond the Year Scale

TL;DR

AtmosMJ introduces a deep convolutional network with a novel gating mechanism that achieves stable, long-range weather forecasts on standard grids, challenging the need for complex data transformations used in prior models.

Contribution

The paper presents AtmosMJ, a new model that operates directly on latitude-longitude data and uses a Gated Residual Fusion mechanism to enable stable, long-term weather forecasting without non-standard data representations.

Findings

Achieves stable 500-day forecasts with physically plausible results.

Performs competitively on 10-day forecasts compared to state-of-the-art models.

Requires only 5.7 days of training on a single GPU.

Abstract

The advent of Large Weather Models (LWMs) has marked a turning point in data-driven forecasting, with many models now outperforming traditional numerical systems in the medium range. However, achieving stable, long-range autoregressive forecasts beyond a few weeks remains a significant challenge. Prevailing state-of-the-art models that achieve year-long stability, such as SFNO and DLWP-HPX, have relied on transforming input data onto non-standard spatial domains like spherical harmonics or HEALPix meshes. This has led to the prevailing assumption that such representations are necessary to enforce physical consistency and long-term stability. This paper challenges that assumption by investigating whether comparable long-range performance can be achieved on the standard latitude-longitude grid. We introduce AtmosMJ, a deep convolutional network that operates directly on ERA5 data without any spherical remapping. The model's stability is enabled by a novel Gated Residual Fusion (GRF) mechanism, which adaptively moderates feature updates to prevent error accumulation over long recursive simulations. Our results demonstrate that AtmosMJ produces stable and physically plausible forecasts for about 500 days. In quantitative evaluations, it achieves competitive 10-day forecast accuracy against models like Pangu-Weather and GraphCast, all while requiring a remarkably low training budget of 5.7 days on a V100 GPU. Our findings suggest that efficient architectural design, rather than non-standard data representation, can be the key to unlocking stable and computationally efficient long-range weather prediction.