Predicting London’s Precipitation: A Spatio-Temporal Neural Network Approach

Huma Zafar; Stelios Kapetanakis; Giacomo Nalli; Khuong An Nguyen

doi:10.1007/978-3-032-11442-6_13

Predicting London’s Precipitation: A Spatio-Temporal Neural Network Approach

Huma Zafar
, Stelios Kapetanakis
, Giacomo Nalli
, Khuong An Nguyen

Department of Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Chapter

Abstract

This study presents a data-driven approach to forecasting total precipitation in London using an Artificial Neural Network (ANN) within a spatio-temporal framework. Leveraging ERA5 data from 2010 to 2025, the methodology includes automated NetCDF extraction, feature engineering with lagged precipitation and cyclic time encodings, and dimensionality reduction via a trained Autoencoder. The ANN, designed in a GenCast-style architecture, was trained using the Adam optimiser over 50 epochs and achieved strong performance. SHAP analysis highlighted the importance of lag features and seasonal time variables, enhancing interpretability and supporting the model’s application in urban flood risk management and climate resilience.

Original language	English
Title of host publication	Artificial Intelligence XLII
Subtitle of host publication	45th SGAI International Conference on Artificial Intelligence
Publisher	Springer, [Cham]
Pages	179
Number of pages	189
ISBN (Electronic)	978-3-032-11442-6
ISBN (Print)	978-3-032-11441-9
DOIs	https://doi.org/10.1007/978-3-032-11442-6_13
Publication status	Published - 24 Nov 2025

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-032-11442-6_13

Predicting_London_PrecipitationAccepted author manuscript, 776 KB

Cite this

@inbook{5f4b2bb4e1ba481ab7aa1fcf3d8c025e,

title = "Predicting London{\textquoteright}s Precipitation: A Spatio-Temporal Neural Network Approach",

abstract = "This study presents a data-driven approach to forecasting total precipitation in London using an Artificial Neural Network (ANN) within a spatio-temporal framework. Leveraging ERA5 data from 2010 to 2025, the methodology includes automated NetCDF extraction, feature engineering with lagged precipitation and cyclic time encodings, and dimensionality reduction via a trained Autoencoder. The ANN, designed in a GenCast-style architecture, was trained using the Adam optimiser over 50 epochs and achieved strong performance. SHAP analysis highlighted the importance of lag features and seasonal time variables, enhancing interpretability and supporting the model{\textquoteright}s application in urban flood risk management and climate resilience.",

author = "Huma Zafar and Stelios Kapetanakis and Giacomo Nalli and Nguyen, \{Khuong An\}",

year = "2025",

month = nov,

day = "24",

doi = "10.1007/978-3-032-11442-6\_13",

language = "English",

isbn = "978-3-032-11441-9",

pages = "179",

booktitle = "Artificial Intelligence XLII",

publisher = "Springer, [Cham]",

}

TY - CHAP

T1 - Predicting London’s Precipitation: A Spatio-Temporal Neural Network Approach

AU - Zafar, Huma

AU - Kapetanakis, Stelios

AU - Nalli, Giacomo

AU - Nguyen, Khuong An

PY - 2025/11/24

Y1 - 2025/11/24

N2 - This study presents a data-driven approach to forecasting total precipitation in London using an Artificial Neural Network (ANN) within a spatio-temporal framework. Leveraging ERA5 data from 2010 to 2025, the methodology includes automated NetCDF extraction, feature engineering with lagged precipitation and cyclic time encodings, and dimensionality reduction via a trained Autoencoder. The ANN, designed in a GenCast-style architecture, was trained using the Adam optimiser over 50 epochs and achieved strong performance. SHAP analysis highlighted the importance of lag features and seasonal time variables, enhancing interpretability and supporting the model’s application in urban flood risk management and climate resilience.

AB - This study presents a data-driven approach to forecasting total precipitation in London using an Artificial Neural Network (ANN) within a spatio-temporal framework. Leveraging ERA5 data from 2010 to 2025, the methodology includes automated NetCDF extraction, feature engineering with lagged precipitation and cyclic time encodings, and dimensionality reduction via a trained Autoencoder. The ANN, designed in a GenCast-style architecture, was trained using the Adam optimiser over 50 epochs and achieved strong performance. SHAP analysis highlighted the importance of lag features and seasonal time variables, enhancing interpretability and supporting the model’s application in urban flood risk management and climate resilience.

U2 - 10.1007/978-3-032-11442-6_13

DO - 10.1007/978-3-032-11442-6_13

M3 - Chapter

SN - 978-3-032-11441-9

SP - 179

BT - Artificial Intelligence XLII

PB - Springer, [Cham]

ER -