Precipitation forecast post-processing - blending deterministic NWPs with machine learning - EGU General Assembly 2024

Aug 6, 2026

Upload by Eliana

Direct precipitation forecasts from Numerical Weather Prediction (NWP) models suffer from sensitivity to initial and boundary conditions and from imperfect parametrization schemes, leading to rapidly growing systematic and random errors in chaotic atmospheric dynamics. Despite decades of improvements, model uncertainty remains operationally relevant, especially for localized weather alerts in Italy. This work presents a machine learning multimodel post-processing approach that blends deterministic and probabilistic information using convolutional neural networks, benchmark regressions, and ensemble statistics over 24–48h.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages3

Reading Time8 min

AccessFree to download

Introduction

Sources of error in NWP precipitation forecasts

Motivation for uncertainty-aware post-processing

Method

Machine learning multimodel post-processing with CNNs

Deterministic and probabilistic benchmarks

POLITECNICO DI TORINO Repository ISTITUZIONALE Precipitation forecast post-processing: blending deterministic NWPs with machine learning Original Precipitation forecast post-processing: blending deterministic NWPs with machine learning / Monaco, Luca; Cremonini, Roberto; Laio, Francesco. - (2024) . (Intervento presentato al convegno European Geosciences Union Assembly 2024 tenutosi a Vienna) [10 .5194/egusphere-egu24-2361] . Availability: This version is available at: 11583/2991667 since: 2024-08-12T12:08:52Z Publisher: Copernicus Published DOI:10.5194/egusphere-egu24-2361 Terms of use: This article is made available under terms and conditions as specified in the corresponding bibliographic description in the repository Publisher copyright (Article begins on next page) EGU24-2361, updated on 05 Sep 2024 [https://doi.org/10.5194/egusphere-egu24-2361](https://doi.org/10.5194/egusphere-egu24-2361)[ ](https://doi.org/10.5194/egusphere-egu24-2361)EGU General Assembly 2024 © Author(s) 2024. This work is distributed under the Creative Commons Attribution 4.0 License. Precipitation forecast post-processing: blending deterministic NWPs with machine learning Luca Monaco1, Roberto Cremonini2, and Francesco Laio3 1 Politecnico di Torino, DIATI, Italy (luca. monaco@polito. it) 2ARPAP, Environmental and Natural Risks Department, Italy ([robecrem@arpa.piemonte.it](robecrem@arpa.piemonte.it)) 3 Politecnico di Torino, DIATI, Italy (francesco. laio@polito. it) Direct model output forecasts by Numerical Weather Prediction models (NWPs) present some limitations caused by errors mostly due to sensitivity to initial conditions, sensitivity to boundary conditions and deficiencies in parametrization schemes (i.e. orography) . These sources of error are unavoidable, and atmosphere chaotic dynamics makes prediction errors to spread rapidly in time in the course of the forecast, inducing both systematic and random errors. Nonetheless, in the last 50 years NWPs had a significant decrease in the impact of these source of errors, even in the long-term forecast, thanks for instance to an ever-increasing computational capability, but still their relevance is not neglectable. Moreover, different NWPs present specific different pros and cons which are findable empirically. For instance, in the case of precipitation forecast in the north-west Italy, low spatial resolution models (e.g. ECMWF-IFS) tend to be more reliable in terms of space and time in predicting the average precipitation, while high resolution models (e.g. COSMO-2I) tend to forecasts the maximum precipitation better. Research purposes apart, actual limitations must be seen in an operational context, where weather forecasts’ skillfulness and associated uncertainty are information of the utmost importance to the forecaster and in general to the user of a certain forecasts system. In order to tackle the limitations of NWPs and the need of an uncertainty-quantified meteorological forecast, we propose a machine learning based multimodel post-processing technique for precipitation forecast. We focus on precipitation since it is the most important variable in the issue of spatially localized weather alert notice by the Italian Civil Protection’ system and at the same time it is one of the most challenging variables to forecast. We use different Convolutional Neural Networks (CNNs) to obtain both deterministic and probabilistic forecast grids over 24h up to 48h focusing in the North-West Italy, using different high and low resolution deterministic NWPs as input and using high resolution rain-gauge corrected radar observations as ground truth for the training. We use constrainted linear regressions as a mean of deterministic benchmark, and ECMWF-EPS as a mean of probabilistic benchmark. The test phase show decent improvements in terms of RMSE for every season.

FAQs

What problems affect deterministic NWP precipitation forecasts?

They include sensitivity to initial and boundary conditions and deficiencies in parametrization schemes such as orography, which cause forecast errors to spread quickly over time.

Why is precipitation particularly important for operational forecasting in Italy?

Precipitation is central to spatially localized weather alert notices by the Italian Civil Protection system and is also one of the hardest variables to predict.

How does the proposed method improve precipitation forecasts?

It uses convolutional neural networks to produce deterministic and probabilistic grids over 24–48h by combining low- and high-resolution NWP outputs and training against rain-gauge corrected radar observations, with benchmark models for comparison.

Aug 6, 2026

Upload by Eliana