SELECTION OF DYNAMICAL MODEL USING ANALOG DATA ASSIMILATION

RUIZ, JUAN; VALERIE MONBET; PIERRE AILLLIOT; PIERRE TANDEO; TRANG, THI TUYET

Oxford

Congreso; Climate Informatics 2020; 2020

Data assimilation is a relevant framework to merge adynamical model with noisy observations. When various models are incompetition, the question is to find the model that best matches theobservations. This matching can be measured by using the modelevidence, defined by the likelihood of the observations given themodel. This study explores the performance of model attributionbased on model evidence computed using data-driven dataassimilation, where dynamical models are emulated using machinelearning methods.