Data- and parameter-induced uncertainty estimation for Land Surface Models. A SiB3 case study in the LBA domain

Uncertainty in model predictions is associated with data-, parameter-, and model structure-induced uncertainty and the determination/estimation of the different contributions is currently given a wide range of attention within the hydrologic community. Using a variety of single- and multiple criterion methods for sensitivity analysis and inverse modeling we analyze and compare the behavior of a state of the art land surface model, the Simple Biosphere Model 3 (SiB3) in terms of model sensitivity and how it is affected by data error. We also estimate the uncertainty in model predictions associated with parameter and data uncertainty. In particular, we focus on the behavior of model predictions for turbulent heat and carbon fluxes, soil moisture, and soil temperature. We use data collected from three hydrometeorological towers (Santarem Km 67 / Km 83) within the LBA domain. The methods used are based on Markov Chain Monte Carlo simulations with several Metropolis type algorithms for the parameter uncertainty estimation, perturbation methods for the data induced uncertainty, while generalized sensitivity and variance methods are used for the sensitivity analysis. The influence that the specific location exerts upon the model simulation is also analyzed. Overall we have performed in excess of 500 thousand simulations for our analysis.