Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

REE-TTT: Highly Adaptive Radar Echo Extrapolation Based on Test-Time Training

Xin Di, Xinglin Piao, Fei Wang, Guodong Jing, Yong Zhang

TL;DR

This work tackles the challenge of cross-regional generalization in high-resolution radar echo extrapolation for precipitation nowcasting. It introduces REE-TTT, which embeds an adaptive test-time training mechanism within a spatio-temporal translator (ST-TTT) that uses attention-based projections to capture dynamic meteorological evolution. The method combines outer-loop global parameter updates with inner-loop self-supervised adaptation, and employs a hybrid loss L = L_MAE + \\lambda L_FFL to emphasize strong echoes and recover high-frequency structure. Empirical results on Beijing and zero-shot Hangzhou scenarios show improved generalization and event-detection performance, with ablation and extension studies validating the effectiveness and adaptability of the approach for cross-regional precipitation forecasting.

Abstract

Precipitation nowcasting is critically important for meteorological forecasting. Deep learning-based Radar Echo Extrapolation (REE) has become a predominant nowcasting approach, yet it suffers from poor generalization due to its reliance on high-quality local training data and static model parameters, limiting its applicability across diverse regions and extreme events. To overcome this, we propose REE-TTT, a novel model that incorporates an adaptive Test-Time Training (TTT) mechanism. The core of our model lies in the newly designed Spatio-temporal Test-Time Training (ST-TTT) block, which replaces the standard linear projections in TTT layers with task-specific attention mechanisms, enabling robust adaptation to non-stationary meteorological distributions and thereby significantly enhancing the feature representation of precipitation. Experiments under cross-regional extreme precipitation scenarios demonstrate that REE-TTT substantially outperforms state-of-the-art baseline models in prediction accuracy and generalization, exhibiting remarkable adaptability to data distribution shifts.

REE-TTT: Highly Adaptive Radar Echo Extrapolation Based on Test-Time Training

TL;DR

This work tackles the challenge of cross-regional generalization in high-resolution radar echo extrapolation for precipitation nowcasting. It introduces REE-TTT, which embeds an adaptive test-time training mechanism within a spatio-temporal translator (ST-TTT) that uses attention-based projections to capture dynamic meteorological evolution. The method combines outer-loop global parameter updates with inner-loop self-supervised adaptation, and employs a hybrid loss L = L_MAE + \\lambda L_FFL to emphasize strong echoes and recover high-frequency structure. Empirical results on Beijing and zero-shot Hangzhou scenarios show improved generalization and event-detection performance, with ablation and extension studies validating the effectiveness and adaptability of the approach for cross-regional precipitation forecasting.

Abstract

Precipitation nowcasting is critically important for meteorological forecasting. Deep learning-based Radar Echo Extrapolation (REE) has become a predominant nowcasting approach, yet it suffers from poor generalization due to its reliance on high-quality local training data and static model parameters, limiting its applicability across diverse regions and extreme events. To overcome this, we propose REE-TTT, a novel model that incorporates an adaptive Test-Time Training (TTT) mechanism. The core of our model lies in the newly designed Spatio-temporal Test-Time Training (ST-TTT) block, which replaces the standard linear projections in TTT layers with task-specific attention mechanisms, enabling robust adaptation to non-stationary meteorological distributions and thereby significantly enhancing the feature representation of precipitation. Experiments under cross-regional extreme precipitation scenarios demonstrate that REE-TTT substantially outperforms state-of-the-art baseline models in prediction accuracy and generalization, exhibiting remarkable adaptability to data distribution shifts.
Paper Structure (18 sections, 13 equations, 7 figures, 4 tables)

This paper contains 18 sections, 13 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Spatio-temporal Prediction
  4. Test-Time Training
  5. METHODOLOGY
  6. Problem Formulation
  7. Overview
  8. Spatio-Temporal TTT block
  9. Loss Functions
  10. Experiments
  11. Experiment Setting
  12. Implementation Details
  13. Baselines
  14. Evaluation Metrics
  15. Comparative Experiments
  16. ...and 3 more sections

Figures (7)

  • Figure 1: Cluster analysis results of radar composite reflectivity samples from the Beijing and Hangzhou datasets reveals distinct distributions, where each point corresponds to a radar composite reflectivity image. Most samples in both datasets reside near cluster centers, representing low-intensity precipitation patterns. However, the Hangzhou dataset contains a greater number of outlier samples corresponding to intense precipitation events. In contrast, while the Beijing dataset is dominated by low-intensity precipitation, it also includes several outlier processes that deviate from these clusters, particularly samples captured during two historic torrential rain events (marked in the figure).
  • Figure 2: Overview of REE-TTT Model.
  • Figure 3: Proposed ST-TTT block.
  • Figure 4: Beijing radar echo extrapolation visualization.
  • Figure 5: Demonstrate the CSI$_{25}$ metric comparison across prediction timesteps, with REE-TTT serving as the baseline. Positive values indicate superior performance relative to the REE-TTT method. Subfigure (a) displays Beijing dataset results, and (b) displays Beijing-pretrained models on Hangzhou data.
  • ...and 2 more figures