Forecasting urban temperatures through crowdsourced data from Citizen Weather Stations

Abstract

This work explores the potential of using crowdsourced data from Citizen Weather Stations (CWS) to forecast urban temperatures. Five case studies were selected (Madrid, London, Rome, Paris and Berlin), using data from the manufacturer Netatmo to gather the hourly temperatures of the years 2021 and 2022 for 776 CWS. A quality-control process was implemented to enhance data accuracy, and heat maps were generated. Multiple linear regression models were created for each CWS and month using only reference weather data as input, resulting in the development of accurate models (RMSE lower than 1.5 °C) for an average of 374 CWS. Then, reference weather predictions were used on the 2nd of April 2023 to forecast the hourly temperatures of those CWS for the period from the 3rd to the 9th of April (168 h). The outcomes showed that accurate CWS data and precise prediction models for the reference weather are crucial to improve the accuracy of the forecasts using crowdsourced data. The study demonstrates the potential benefits of using CWS data and simple models to forecast urban temperatures in a cost-efficient way even 168 h ahead, with implications for various sectors such as urban planning, energy consumptions, health impacts, or climate change adaptation strategies.