agrometeoR

Geostatistics module and R package for agrometeorological forecasting.

Logo

Walloon Agricultural Research Centre R · mlr · gstat · sf · open-source package

agrometeoR is an open-source R package that spatialises connected weather-station data for Walloon agrometeorology. View the project →

Context

Deliver the spatialisation algorithm at the heart of the AGROMET platform. From the connected station measurements, generate a virtual weather station for every km² of Wallonia (temperature, rainfall, relative humidity, leaf wetness), so that crop disease-risk models can be run on the grid.

What I did

  • Developed the R package agrometeoR implementing the spatialisation algorithm, with supervised machine learning under mlr (sf, gstat).
  • Selected the geographic explanatory variables (elevation, latitude, longitude…) and benchmarked interpolation algorithms, compared by cross-validation.
  • Integrated gstat into mlr to extend its geospatial capabilities.
  • Intercalibration of the PAMESEB and IRM station networks.

Outcome

The algorithm is in production at the heart of the AGROMET platform: an hourly weather grid at 1 km² across Wallonia and a reusable open-source R package, feeding the agronomic risk models (WalDigiFarm).

FOSS4G Belgium 2018 talk

Why and How to use R as an open source GIS: The AGROMET project use case. An interactive presentation built with the R package learnr.

Open the presentation in a new tab →