MIDA, a software system to facilitate Model-Independent Data Assimilation and ecological forecasting

Abstract:

Models have become an essential tool to predict future states of the Earth system. Accurate prediction of the future states depends not only model structure but also parameterization. The model parameters can be constrained by data assimilation. However, applications of data assimilation are limited because of highly technical requirements, e.g. model-dependent coding. To reduce the technical burden of data assimilation applications, we have developed a model-independent data assimilation (MIDA) module. MIDA is a 3-step workflow including preparation, data assimilation process, and visualization. The first step prepares prior ranges of parameter values, a defined number of iterations, directory paths to access files of observations, simulation outputs and model executable and a output configuration file to map simulation outputs to observations. The data assimilation process calibrates parameters values to best fit the observations and estimates the posterior distributions of parameters. The calibration performance and posterior distributions will be automatically visualized in the final step. The 3-step workflow is conducted via a simple and interactive way for model users without code modification of original models. Thus, MIDA is agnostic of a specific model. In this presentation, we applied MIDA to four types of ecological models: the data assimilation linked ecosystem carbon (DALEC) model, a surrogate-based energy exascale earth system model (E3SM) land model, nine phenological models and a stand-alone biome ecological strategy simulator (BiomeE). The four applications all indicate that MIDA can be effectively implemented for different ecological models. By alleviating the burden of model-dependent coding, MIDA has the potential to facilitate data assimilation to models with observations and reduce uncertainties in model simulation or forecasting.