Skillful precipitation nowcasting using deep generative models of radar - Dr. Piotr Mirowski
Precipitation nowcasting, the high-resolution forecasting of precipitation up to two hours ahead, supports the real-world socioeconomic 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 nonlinear 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 rarer medium-to-heavy rain events. Here we present a deep generative model for the probabilistic nowcasting of precipitation from radar that addresses these challenges. Using statistical, economic and cognitive measures, we show that our method provides improved forecast quality, forecast consistency and forecast value. Our model produces realistic and spatiotemporally consistent predictions over regions up to 1,536 km × 1,280 km and with lead times from 5–90 min ahead. Using a systematic evaluation by more than 50 expert meteorologists, we show that our generative model ranked first for its accuracy and usefulness in 89% of cases against two competitive methods. 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.
00:00 Intro
05:28 Precipitation Nowcasting
09:20 Multi-resolution Multi-scale Data (US)
15:54 Additional data sources
20:54 Baseline: PYSTEPS
22:42 Baseline: Extending Optical Flow with PWC-Nets
24:41 Baseline: UNet (regression and classification)
26:19 Baseline: Axial Attention (MetNet) classification
30:57 (Crude and toy) analogy to autoregressive language model
32:48 Beware of simplistic modeling assumptions in the loss!
35:21 A Thought Experiment
39:23 Generative Adversarial Networks
41:17 Nowcasting model: conceptual diagram
45:22 Met Office RadarNet4 Data
48:39 Precipitation Data (UK)
51:39 Quantitative Verification and its Limits
56:52 Evaluating GAN results
58:47 ТР Critical Success Index (CSI)
01:05:36 Power Spectral Density (PSD)
01:10:05 Continuous Ranked Probability Score (CRPS)
01:12:12 Intercomparison case study
01:14:11 Ablation analysis
01:15:42 Experimental psychology for operational meteorology
01:17:13 Protocol (phase one): preference ranking
01:18:05 Protocol (phase two): retrospective recall interview
01:21:01 Final thoughts
01:28:45 Discussion | https://github.com/deepmind/deepmind-...
[Chapters were auto-generated using our proprietary software - contact us if you are interested in access to the software]
The talk is based on the speaker's and deepmind's paper published in nature:
Skilful precipitation nowcasting using deep generative models of radar
https://www.nature.com/articles/s4158...
https://deepmind.com/blog/article/now...
https://github.com/deepmind/deepmind-...
Presenter Bio:
Dr. Piotr Mirowski is a Staff Research Scientist at DeepMind, where he has been working within Dr. Raia Hadsell’s and Dr. Shakir Mohamed's teams. His work focuses on weather and climate forecasting as well as on navigation-related research, and scaling up autonomous agents to real world environments. He obtained his Ph.D. in Computer Science at New York University in 2011 with Prof. Yann LeCun (Outstanding Dissertation Award). His previous work experience encompasses epileptic seizure prediction from EEG, the inference of gene regulation networks, WiFi-based geo-localization, simultaneous localization and mapping on a smartphone, robotics, natural language processing, and search query auto-completion. He is also conducting independent research in human-machine co-creation for improvised theatrical performances.
Piotr's homepage: https://piotrmirowski.com/
A related talk we had last year on the same subject: • HydroNet: leverage River Structure for Hyd...
-------------------------
Find us at:
Newsletter for updates about more events ➜ http://eepurl.com/gJ1t-D
Sub-reddit for discussions ➜ / 2d3dai
Discord server for, well, discord ➜ / discord
Blog ➜ https://2d3d.ai