Assessing the sensitivity of land-atmosphere coupling strength to boundary and surface layer parameters in the WRF model over Amazon

• Published in: Atmospheric research Vol. 234; p. 104738
• Main Authors: Wang, Chen, Qian, Yun, Duan, Qingyun, Huang, Maoyi, Berg, Larry K., Shin, Hyeyum H., Feng, Zhe, Yang, Ben, Quan, Jiping, Hong, Songyou, Yan, Junhua
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.04.2020
• Subjects: Amazon region; Amazonia; L-A coupling strength; Model uncertainties; Parameter sensitivity analysis; troposphere; uncertainty; variance; WRF model; Amazonia; WRF model; Amazon region; Parameter sensitivity analysis; L-A coupling strength; Model uncertainties
• ISSN: 0169-8095; 1873-2895
• DOI: 10.1016/j.atmosres.2019.104738

Modeling tools can be used to diagnose regional land-atmosphere (L-A) coupling strength in the absence of sufficient observations, but subject to uncertainties associated with parameters in model physical parameterizations. Different sensitivity analysis (SA) approaches may lead to different conclusions about the underlying sensitivities. In this study, we quantify simulation uncertainties related to parameter perturbations, and use different approaches to conduct parameter SA on the WRF model pertaining to L-A coupling strength for simulations over the Amazon region. A total of twenty parameters from the Yonsei University (YSU) planetary boundary layer (PBL) and the revised MM5 surface layer (SL) schemes were selected in this analysis. Three different SA methods, the Morris One-at-A-Time (MOAT) method, the Multivariate Adaptive Regression Splines (MARS) method, and the Sobol’ method, were employed to analyze seven WRF-simulated variables and five L-A coupling metrics. Results show that 1) parameter perturbations cause large simulation uncertainties which are comparable to those in the observations; 2) three different SA methods give consistent L-A coupling strength outcomes; 3) six out of the twenty parameters contribute 80%–95% of the total variance in the metrics analyzed, and first-order effects dominate over interaction effects; 4) the twelve variables/metrics of interest show similar sensitivity patterns to the selected parameters, which is consistent across all the methods used. Physical mechanisms for how the sensitive parameters act in determining the L-A coupling strength and associated variables also are illustrated. Our results will help quantifying L-A coupling strength and establishing a basis for parameter calibration over the Amazon region. •Six out of 20 parameters contribute 80%–95% of the total variance in L-A coupling strength metrics over Amazon.•The contributions from parameter interactions are relatively small compared to first-order effects.•The twelve variables/metrics of interest show similar patterns of sensitivity to the selected parameters.