In this poster, we will present the most recent results of applying multisite parametric uncertainty quantification (UQ) workflow for dimensionality reduction of ACME Land Model followed by targeted, site-specific, low-dimensional accurate surrogate model construction. Surrogate modeling is the key ingredient of the presented work, as it presents a reasonable approximation of input-output maps, as well as provides efficient means for uncertainty propagation and global sensitivity analysis (GSA), otherwise called variance-based decomposition. Specifically, we develop polynomial chaos (PC) surrogates using Bayesian inference. However, the PC surrogate construction still requires a large ensemble of simulations, especially when the number of parameters is large. Here we apply a new procedure, the weighted Iterative Bayesian Compressive Sensing (WIBCS) algorithm, which allows a sparse, high-dimensional PC surrogate with very few model evaluations, also quantifying uncertainties due to lack of enough model simulations.

We first applied the WIBCS algorithm for the purpose of a GSA to determine the sensitivity of ALM to 65 parameters.  Several model output variables are analyzed including gross primary productivity (GPP), leaf area index (LAI), vegetation carbon and soil organic matter carbon at nearly 100 FLUXNET sites covering a broad range of multiple plant functional types (PFTs) and climates.  We find for all PFTs, generally 15 or fewer parameters drive most of the variance in the outputs.  Within a PFT for a given output, generally the same parameters appear as sensitive at each site while differences in parameters are evident among PFTs and different outputs. This sensitivity analysis then serves as the basis for more focused, lower-dimensional surrogate construction that will help for parameter calibration and improved land-surface model predictions.