3ème conférence sur l’agriculture et le changement climatique
Dans un contexte où le climat change rapidement, une meilleure prise en compte des évolutions moyennes du climat (hausse de la température moyenne, hausse de la concentration en CO2 etc.) et des accidents climatiques (canicules, gelées tardives, sécheresses plus intenses et longues etc.) est nécessaire. ITK sera présent à la 3ème conférence sur l’agriculture et le changement climatique à Budapest (Hongrie) du 24 au 26 mars 2019. Ca sera l’occasion pour nous de vous présenter nos différentes solutions Open-Data pour fournir des données météorologiques à nos outils !
Using open weather data for crop modeling: it is possible
Assessing the impacts of climate change on crop production requires large datasets of weather scenarios. However, getting such data is generally a time-consuming process that is often poorly documented. Nonetheless, since the 2010s, research institutions have released weather data under an open data policy.
The aim of this work is to compare accessibility (format, automatization, server reliability), large-scale exploitability, period (past, short/long-term forecasts and climate change scenarios) and usefulness for crop modelling (inputs provided, data continuity) of more than 20 global and national forecast models (GFS, AREPEGE, ICON, IFS-ECMWF etc.) and worldwide data providers (Infoclimat.fr, NASA, NOAA etc.).
Results are encouraging: (1) Accessibility: all providers gave simple accessibilities, with the exception for AROME/ARPEGE models, which had few minor server overloads. FTP and APIs connections were reliable and could be fully automatized. (2) Period: Open weather data covered the whole period from observations of the 1880s (historic stations from ECA/NOAA databases) to projections of 2100s (climate change scenarios from IPCC/DRIAS). (3) Large-scale exploitability: As it stores a collection of self-contained records of 2D-data in the same file (.grib), gridded format is the most useful for large scale crop models. However, only short (5 days) and long-term (2 weeks) weather forecast models provided this format from high resolution for local model (2km²) to larger scales for global model (25km²). Historical weather data (1880-2019) and climate change projections (2020-2100) are only provided by one-location files (.json). (4) Usefulness: in most cases temperatures, humidity, wind and precipitation are provided, while global radiation is more challenging to obtain.
Open weather data can provide whole range of new research opportunities for crop modelling. For instance, simulation of past years for calibration issues, testing agricultural decision tools or assessing the impact of global warming at different spatial scales. These opportunities are still underestimated and must be highlighted.
Merci à Serge Zaka, Laurent Tisseyre, Samuel Clarenc et Hubert Vincent Varella pour leurs contributions.
Illustration de la NOAA