IRSN

Background

Accidents at nuclear power plants can lead to the contamination of vast territories by radionuclides with long half-life and the substantial loss of agricultural productions, depending on various factors, such as land-cover around the plant or meteorological conditions at the time of the accident. IRSN asked PGeostat to conduct a preliminary probabilistic assessment of the agricultural consequences (e.g. tons of cereals, number of head of livestock) of an accident occurring at a French nuclear power plant. Data were available at different scales: 1) harvested areas per municipality and department, 2) agricultural production (e.g. tons of cereals) at the department level, and 3) meteorological data to predict the dispersion of radionuclides at coarse scale (10x10 km squares). A key issue is the magnitude of prediction errors resulting from the manipulation of data at different scales.

Objectives

The main objectives of this project were to:

1. Explore the main sources of errors that can impact the upscaling of farm statistics to the scale of the output of meteorological models
2. Use this information to write a call for proposals (RFT) that would explore the propagation of errors over the entire French territory and for different crops (cereals, fruits, vegetables, livestock, milk production).

Results

• A geostatistical simulation-based methodology was developed to propagate uncertainty about agricultural productions through the methods implemented by IRSN to conduct change of spatial supports (e.g. upscaling, downscaling, and side-scaling).
• The methodology was used to simulate crop production at the municipality level for one French Department under different scenarios for the spatial pattern of yield (random versus spatially structured) and the magnitude of the temporal bias.
• Criteria were developed to quantify the accuracy of the predictions.