PrecipFusionNet: A Unified Deep Learning Model for Improving Numerical Precipitation Prediction

Numerical weather prediction (NWP) often suffers from substantial biases when forecasting extreme rainfall. Traditional corrections tend to underuse spatial information, while deep learning approaches typically struggle with data imbalance for rare events. In this study, we propose PrecipFusionNet, a novel deep learning framework designed to better capture localized heavy rainfall, which combines an Att‐UKAN architecture, a U‐Net‐derived design integrating convolutional, Kolmogorov‐Arnold Network, and attention modules, with a fusion loss function. PrecipFusionNet improves the prediction skill for extreme events (≥50 mm/day) by 50.42%. For short‐ to medium‐range forecasts (1–4 days), it increases the average Equitable Threat Score (ETS) over China by 35.03% compared to the Global Forecast System (GFS). At longer forecast lead times (72 and 96 hr), it matches the ETS performance of GFS forecasts at 24 and 48 hr. In addition, our data augmentation strategy improves robustness across different lead times and regions, maintaining high accuracy in cross‐domain evaluations (e.g., over North America) and demonstrating strong spatiotemporal generalization.

• Publication year

  2026

• Venue

  Journal of Geophysical Research

• Publication date

  2026-01-30

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1029/2025jh000887
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Advanced streamflow forecasting for Central European Rivers: The Cutting-Edge Kolmogorov-Arnold networks compared to Transformers

  2024cited by this paper
• U-KAN Makes Strong Backbone for Medical Image Segmentation and Generation

  2024cited by this paper
• Assessing the impact of floods on vegetation worldwide from a spatiotemporal perspective

  2023cited by this paper
• Correcting the Bias of Daily Satellite Precipitation Estimates in Tropical Regions Using Deep Neural Network

  2022cited by this paper
• A Study on Loss Function Against Data Imbalance in Deep Learning Correction of Precipitation Forecasts

  2022cited by this paper
• A multiscale long short-term memory model with attention mechanism for improving monthly precipitation prediction

  2021cited by this paper
• Deep learning‐based precipitation bias correction approach for Yin–He global spectral model

  2021cited by this paper
• Meteorological aspects of heavy precipitation in relation to floods – An overview

  2020cited by this paper
• Correction model for rainfall forecasts using the LSTM with multiple meteorological factors

  2019cited by this paper
• Data Quality Considerations for Big Data and Machine Learning: Going Beyond Data Cleaning and Transformations

  2017cited by this paper
• Visual Analysis of Nonlinear Dynamical Systems: Chaos, Fractals, Self-Similarity and the Limits of Prediction

  2016cited by this paper
• Spatio‐temporal precipitation climatology over complex terrain using a censored additive regression model

  2016cited by this paper
• The quiet revolution of numerical weather prediction

  2015cited by this paper
• The socioeconomic vulnerability index: A pragmatic approach for assessing climate change led risks–A case study in the south-western coastal Bangladesh

  2014cited by this paper
• Climate Impacts on Agriculture: Implications for Crop Production

  2011cited by this paper
• Frequency of precipitation and temperature extremes over France in an anthropogenic scenario: Model results and statistical correction according to observed values

  2007cited by this paper
• Reflections on the Conception, Birth, and Childhood of Numerical Weather Prediction

  2006cited by this paper
• Modelling climate change: the role of unresolved processes

  2005cited by this paper
• Statistics of extremes and estimation of extreme rainfall: II. Empirical investigation of long rainfall records / Statistiques de valeurs extrêmes et estimation de précipitations extrêmes: II. Recherche empirique sur de longues séries de précipitations

  2004cited by this paper

• No citing papers are available for this paper.