Leadsee-Precip: A Deep Learning Diagnostic Model for Precipitation

TL;DR

Leadsee-Precip, a global deep learning model to generate precipitation from meteorological circulation fields, utilizes an information balance scheme to tackle the challenges of predicting heavy precipitation caused by the long-tail distribution of precipitation data.

Abstract

Recently, deep-learning weather forecasting models have surpassed traditional numerical models in terms of the accuracy of meteorological variables. However, there is considerable potential for improvements in precipitation forecasts, especially for heavy precipitation events. To address this deficiency, we propose Leadsee-Precip, a global deep learning model to generate precipitation from meteorological circulation fields. The model utilizes an information balance scheme to tackle the challenges of predicting heavy precipitation caused by the long-tail distribution of precipitation data. Additionally, more accurate satellite and radar-based precipitation retrievals are used as training targets. Compared to artificial intelligence global weather models, the heavy precipitation from Leadsee-Precip is more consistent with observations and shows competitive performance against global numerical weather prediction models. Leadsee-Precip can be integrated with any global circulation model to generate precipitation forecasts. But the deviations between the predicted and the ground-truth circulation fields may lead to a weakened precipitation forecast, which could potentially be mitigated by further fine-tuning based on the predicted circulation fields.

Figures (6)

• Figure 1: Structure of Leadsee-Precip. The model consists of feature extraction, hidden translator, and precipitation upsampling.
• Figure 2: Distribution of precipitation. The precipitation data is categorized into 92 bins, with the precipitation intensity gradually increasing from zero to the maximum of 400 $mm$$6h^{-1}$.
• Figure 3: LoRA fine-tuning 5-km model structure. The feature extraction part is frozen, the MogaNet hidden part uses LoRA fine-tuning, and the upsampling part is retrained.
• Figure 4: An illustrative example of a global 6-hour accumulated precipitation prediction generated by Leadsee-Precip and the ground truth of NOAA CMORPH. The calendar time-stamp of the figure was 00:00 UTC on April 1, 2022.
• Figure 5: 24-hour accumulated precipitation at weather station locations: observed and different model results. Panel (a) shows the ground truth of weather station data. Panel (b) shows the diagnosed precipitation of Leadsee. Panel (c) shows the Leadsee-Precip forecast driven by FuXi. Panel (d) shows the precipitation forecast of ECMWF HRES. Panel (e) shows the native precipitation forecast of FuXi. Panle (f) shows the fine-tuning Leadsee-Precip diagnosis. The accumulated precipitation covers the period from 18:00 UTC on July 29, 2024, to 18:00 UTC on July 30, 2024.