[Paper Review] Skillful Precipitation Nowcasting using Deep Generative Models of Radar

Suman Ravuri, Karel Lenc+18 more|arXiv (Cornell University)|Apr 2, 2021

Meteorological Phenomena and SimulationsEarth and Planetary Sciences46 references30 citations

TL;DR

The paper introduces a Deep Generative Model for probabilistic, high-resolution nowcasting of precipitation from radar, delivering realistic, non-blurred forecasts up to 90 minutes ahead over large regions and evaluated by expert forecasters.

ABSTRACT

Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socio-economic needs of many sectors reliant on weather-dependent decision-making. State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-based wind estimates, and struggle to capture important non-linear events such as convective initiations. Recently introduced deep learning methods use radar to directly predict future rain rates, free of physical constraints. While they accurately predict low-intensity rainfall, their operational utility is limited because their lack of constraints produces blurry nowcasts at longer lead times, yielding poor performance on more rare medium-to-heavy rain events. To address these challenges, we present a Deep Generative Model for the probabilistic nowcasting of precipitation from radar. Our model produces realistic and spatio-temporally consistent predictions over regions up to 1536 km x 1280 km and with lead times from 5-90 min ahead. In a systematic evaluation by more than fifty expert forecasters from the Met Office, our generative model ranked first for its accuracy and usefulness in 88% of cases against two competitive methods, demonstrating its decision-making value and ability to provide physical insight to real-world experts. When verified quantitatively, these nowcasts are skillful without resorting to blurring. We show that generative nowcasting can provide probabilistic predictions that improve forecast value and support operational utility, and at resolutions and lead times where alternative methods struggle.

Motivation & Objective

• Address the limitations of traditional radar-advection and unconstrained deep learning nowcasting methods.
• Develop a probabilistic, generative model that produces realistic, spatio-temporally consistent precipitation nowcasts.
• Demonstrate operational utility and skill in producing useful predictions for real-world decision-making.

Proposed method

• Develop a deep generative model for radar-based precipitation nowcasting.
• Provide probabilistic forecasts to capture uncertainty and rare events.
• Ensure spatio-temporal consistency over large regions (up to 1536 km × 1280 km) and lead times (5–90 minutes).
• Evaluate predictions against expert forecasters to assess accuracy, usefulness, and physical plausibility.

Experimental results

Research questions

• RQ1 Can a deep generative model produce probabilistic, non-blurry nowcasts that remain skillful at longer lead times?
• RQ2 Do generative nowcasts provide improved operational value and physical insight compared with conventional methods?
• RQ3 Are forecasts accurate and useful to real-world forecasters across large geographic scales and varying precipitation intensities?

Key findings

• The generative model produces realistic and spatio-temporally consistent predictions for lead times of 5–90 minutes over regions up to 1536 km × 1280 km.
• In a formal evaluation with more than fifty Met Office forecasters, the model ranked first for accuracy and usefulness in 88% of cases against two competitive methods.
• Quantitative verification shows skillful nowcasts without introducing excessive blurring, offering probabilistic forecasts that enhance forecast value.
• Generative nowcasting provides both improved probabilistic predictions and operational utility at scales where other methods struggle.