Do Echo Top Heights Improve Deep Learning Nowcasts?

2025-07-01

AI Summary

Precipitation nowcasting is critical for disaster prevention and mitigation, yet most existing deep learning models rely solely on two-dimensional radar reflectivity, neglecting valuable three-dimensional structural information. To address this limitation, this work introduces echo-top height (ETH) as an auxiliary variable—the first such integration in radar-based nowcasting—and proposes a dual-channel 3D U-Net architecture that jointly encodes both radar reflectivity and its ETH-derived two-dimensional projection. Experiments demonstrate that incorporating ETH significantly improves forecasting skill for light precipitation (≤10 mm/h), notably enhancing the probability of detection and critical success index. However, the model exhibits reduced stability and systematic underestimation for heavy precipitation, with case studies indicating increased error variance. This study provides both a novel methodological framework and empirical evidence for leveraging vertical structural priors in multi-source radar data modeling.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: 3D Computer Vision
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Precipitation nowcasting -- the short-term prediction of rainfall using recent radar observations -- is critical for weather-sensitive sectors such as transportation, agriculture, and disaster mitigation. While recent deep learning models have shown promise in improving nowcasting skill, most approaches rely solely on 2D radar reflectivity fields, discarding valuable vertical information available in the full 3D radar volume. In this work, we explore the use of Echo Top Height (ETH), a 2D projection indicating the maximum altitude of radar reflectivity above a given threshold, as an auxiliary input variable for deep learning-based nowcasting. We examine the relationship between ETH and radar reflectivity, confirming its relevance for predicting rainfall intensity. We implement a single-pass 3D U-Net that processes both the radar reflectivity and ETH as separate input channels. While our models are able to leverage ETH to improve skill at low rain-rate thresholds, results are inconsistent at higher intensities and the models with ETH systematically underestimate precipitation intensity. Three case studies are used to illustrate how ETH can help in some cases, but also confuse the models and increase the error variance. Nonetheless, the study serves as a foundation for critically assessing the potential contribution of additional variables to nowcasting performance.

Problem

• Exploring Echo Top Height's impact on rainfall nowcasting accuracy
• Assessing vertical data integration in deep learning precipitation models
• Evaluating ETH's inconsistent benefits across varying rain intensities

Innovation

• Uses Echo Top Height as auxiliary input
• Implements 3D U-Net for dual-channel processing
• Explores vertical data for improved nowcasting