Comparing the impact of weather forecasts, observation and parameterization uncertainties when predicting water stress in vineyards

Année de publication

2025

We analyze the impact of different uncertainty sources when using of a crop water stress model with weather forecasts in order to make crop water stress predictions. Three uncertainty sources are jointly studied: (i) the observation uncertainty, coming from the use of past weather data not directly located on the field under study, (ii) the input uncertainty, coming from imperfect values of critical crop model parameters, (iii) the forecast uncertainty coming from the difficulty to predict future weather data. These three uncertainty sources are modeled and their impact characterized using a dedicated double-stage sensitivity analysis that allows to tackle two issues: dependencies among input factors and multivariate outputs. The analysis is performed on a large database of 12000 scenarios covering 5 years of weather data observations and forecasts from Integrated Forecast System Ensemble Prediction System (IFS-EPS) associated to 10 different locations in France and various simulated soil conditions. The results show that all uncertainty sources matter, but not in the same ranking depending on lead time. In particular, forecast uncertainty becomes in average the dominant source from the 5th day of forecast, accounting for 64% of the variability at a 7-day lead and 86% at a 15-day lead. Our analysis also highlights the existence of several interpretable dynamic patterns of uncertainty dominance, where each of the two other uncertainty sources (past weather data and cropping system parameterization) can dominate the overall uncertainty for some specific part of the forecasting period, stressing that no uncertainty source can be a priori neglected.</div>

  Accès à la notice sur le site du portail documentaire de Météo-France

  Liste complète des notices publiques