Nowcasting

MS-nowcasting: Operational Precipitation Nowcasting with Convolutional LSTMs at Microsoft Weather

Sylwester Klocek, Haiyu Dong, Matthew Dixon, Panashe Kanengoni, Najeeb Kazmi, Pete Luferenko, Zhongjian Lv, Shikhar Sharma, Jonathan Weyn, Siqi Xiang

NeurIPS 2021 Workshop on Tackling Climate Change with Machine Learning, 2021. https://www.climatechange.ai/papers/neurips2021/19 · 18 November 2021 · Microsoft Research (AI for Science)

PaperPDF

In one sentence. Microsoft's operational ConvLSTM radar nowcasting model, conditioned on HRRR, beats optical flow and HRRR by 20–25% for lead times up to 6 hours.

Abstract

We present the encoder-forecaster convolutional long short-term memory (LSTM) deep-learning model that powers Microsoft Weather's operational precipitation nowcasting product. This model takes as input a sequence of weather radar mosaics and deterministically predicts future radar reflectivity at lead times up to 6 hours. By stacking a large input receptive field along the feature dimension and conditioning the model's forecaster with predictions from the physics-based High Resolution Rapid Refresh (HRRR) model, we are able to outperform optical flow and HRRR baselines by 20-25% on multiple metrics averaged over all lead times.

Labels

Topic

Nowcasting

Type

New model or method

Method

RNN / LSTM / ConvLSTM, CNN / U-Net

Time scale

0–6 h

Domain

Regional

Data

Radar, NWP analysis or forecasts (IFS, GFS, ...)

Status

Operational

Identifiers

arXiv:2111.09954 OpenAlex:W3217077725

Added

2026-10-22

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.