Assessment of forecast uncertainty with OTAMIN
OTAMIN is designed to automatically associate confidence intervals to hydrologic and hydraulic forecasts obtained from diverse deterministic models that are used for real-time applications. This tool was developed thanks to the PhD work of Bourgin (2014) at INRAE.
Forecasts uncertainty is estimated using the method QUOIQUE (also developed by Bourgin, 2014), which aims at describing relative errors behavior based on past observations and simulations. Errors are classified regarding the range of simulated values and the forecast lead-time, which constitutes an easy tool to anticipate future errors knowing forecasts and lead-time. QUOIQUE is based on the hypothesis that errors are temporally stable.
The architecture of OTAMIN
OTAMIN is divided into three tools:
- An analysis tool that helps the user to evaluate the method and the uncertainty temporal stability thanks to a calibration-validation process.
- A calibration tool that creates error abacus which are then used for real-time uncertainties assessment.
- A real-time tool that provides uncertainties applied to real-time forecasts using the abacus created during the calibration (see Figure 1).
Figure 1: Confidence intervals for a forecast