Quantifying Uncertainties for Cloud Retrievals using Variance-Based Decomposition and Global Sensitivity Index

 

Authors

Shaocheng Xie — Lawrence Livermore National Laboratory
Xiao Chen — Lawrence Livermore National Laboratory
Qi Tang — Lawrence Livermore National Laboratory

Category

QUICR: Quantification of Uncertainty in Cloud Retrievals

Description

There are large uncertainties in the retrieval products of cloud microphysical properties using US Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) ground-based cloud remote sensing observations. Quantitative knowledge about the retrieval uncertainties is critical to accurately evaluate global climate models. As a recent progress made by the Atmospheric System Research (ASR) Quantification of Uncertainty in Cloud Retrievals focus group (QUICR), we develop a novel uncertainty quantification methodology for cloud retrievals using variance-based decomposition and global sensitivity index. In this approach, empirical orthogonal function (EOF) expansion is applied to ARM ground-based observations. The principal component (PC) in the EOF expansion is considered as random variable, and hence the uncertainty of input data profiles is automatically imbedded in the EOF expansion. This method enables probabilitization of a retrieval process by adding normally distributed perturbations into sample-means of input data profiles within an observation time window. Therefore, this approach effectively facilitates objective validation of climate models against cloud retrievals under a probabilistic framework for rigorous statistical inferences. This method has been demonstrated to characterize the uncertainties of the cloud ice water content retrieved from the ARM program’s baseline cloud retrieval algorithm (MICROBASE) for an ice cloud case observed at the Southern Great Plains site on March 9, 2000. Based on this approach we greatly reduces dimension of random variables (up to a factor of 50), and hence the large ensemble of random sampling becomes computationally feasible. Also, the cross correlations in the input data profiles are revealed, and hence making the estimated uncertainties more accurate. Moreover, the variance-based global sensitivity index analysis, part of this method, attributes the output uncertainties to each individual source, thus providing directions for improving retrieval algorithms and observation strategies. We are extending this approach to ensemble MICROBASE cloud retrievals with comprehensive UQ information for model evaluation for the entire March 2000 IOP. This work was performed under the auspices of the U. S. DOE by Lawrence Livermore National Laboratory under Contract No. DE-AC52-07NA27344. The release number is LLNL-ABS-666460.