Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Source Temporal Attention Network for Precipitation Nowcasting

Rafael Pablos Sarabia, Joachim Nyborg, Morten Birk, Jeppe Liborius Sjørup, Anders Lillevang Vesterholt, Ira Assent

TL;DR

This work introduces an efficient deep learning model for precipitation nowcasting, capable of predicting rainfall up to 8 hours in advance with greater accuracy than existing operational physics-based and extrapolation-based models.

Abstract

Precipitation nowcasting is crucial across various industries and plays a significant role in mitigating and adapting to climate change. We introduce an efficient deep learning model for precipitation nowcasting, capable of predicting rainfall up to 8 hours in advance with greater accuracy than existing operational physics-based and extrapolation-based models. Our model leverages multi-source meteorological data and physics-based forecasts to deliver high-resolution predictions in both time and space. It captures complex spatio-temporal dynamics through temporal attention networks and is optimized using data quality maps and dynamic thresholds. Experiments demonstrate that our model outperforms state-of-the-art, and highlight its potential for fast reliable responses to evolving weather conditions.

Multi-Source Temporal Attention Network for Precipitation Nowcasting

TL;DR

This work introduces an efficient deep learning model for precipitation nowcasting, capable of predicting rainfall up to 8 hours in advance with greater accuracy than existing operational physics-based and extrapolation-based models.

Abstract

Precipitation nowcasting is crucial across various industries and plays a significant role in mitigating and adapting to climate change. We introduce an efficient deep learning model for precipitation nowcasting, capable of predicting rainfall up to 8 hours in advance with greater accuracy than existing operational physics-based and extrapolation-based models. Our model leverages multi-source meteorological data and physics-based forecasts to deliver high-resolution predictions in both time and space. It captures complex spatio-temporal dynamics through temporal attention networks and is optimized using data quality maps and dynamic thresholds. Experiments demonstrate that our model outperforms state-of-the-art, and highlight its potential for fast reliable responses to evolving weather conditions.

Paper Structure

This paper contains 8 sections, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Approach: multi-source temporal attention
  3. Experiments and results
  4. Discussion and conclusion
  5. Quality map
  6. Data preparation
  7. Inputs and outputs
  8. Splits

Figures (4)

  • Figure 1: Proposed architecture, capable of simultaneously processing multiple data sources.
  • Figure 2: Critical Success Index (CSI) for various models across lead times.
  • Figure 3: Sample ground truth, model prediction, GFS, Harmonie, and PySTEPS forecasts. Even though our model provides predictions at 10-minute intervals, hourly intervals are shown.
  • Figure A.1: Estimated quality map over Denmark, created by averaging measurements per pixel over a 3-year period. The locations of the three active radar stations, as indicated by DMI, are also shown.