Uncertainty Quantification for ECOSTRESS

ECOSTRESS is NASA's ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station. It is designed to retrieve evapotranspiration heat flux using spectroscopy in the thermal infrared. Uncertainty quantification for ECOSTRESS evapotranspiration extends the paradigm of simulation-based uncertainty quantification by simulating whole spatial fields at one time. In the original, non-spatial version of simulation-based uncertainty quantification (used for OCO-2 and AIRS), we created ensembles by drawing synthetic true state vectors from a distribution of potential states, one-at-a-time and independently. Here, we generate ensemble of spatial fields by simulating them from a fitted spatial statistical model. The model is fit to a parent exemplar field (such as the output of deterministic model). Then, we apply the ECOSTRESS retrieval to all pixels in all ensemble members. Each pixel has a stack of synthetic true ET values and corresponding stack of synthetic retrieved ET values, which are used to produce a stack of ET retrieval errors. These distributions can be used to characterize ET retrieval performance in each pixel individually, but this strategy allows us to do more: we can investigate the covariance properties of the errors. This is valuable for quantifying uncertainties in estimated spatial gradients using ECOSTRESS data products.

Median ET error, by pixel, for a scene in Nebraska, USA.	Distributions (over all pixels and all ensemble members) of retrieved ET error (ET minus synthetic true ET) for six scenes (A-G) in CONUS.

Selected Publications

Cawse-Nicholson, K., Braverman, A., Kang, Emily L., Li, M., Johnson, M., Halverson, G., Anderson, M., Haind, C., Gunson, M., and and Hook, S. (2020). “Sensitivity and uncertainty quantification for the ECOSTRESS evapotranspiration algorithm – DisALEXI,” International Journal on Applied Earth Observations and Geoinformation. Document (link to DOI).

Ma, P. and Kang, E.L. (2020). “A Fused Gaussian Process Model for Very Large Spatial Data,” Journal of Computational and Graphical Statistics, pp. 1-11. Document (link to DOI).

Ma, P., Kang, E.L., Braverman, A.J., and and Nguyen, H.M. (2019). “Spatial Statistical Downscaling for Constructing High-resolution Nature Runs in Global Observing System Simulation Experiments,” Technometrics, 61(3), pp. 322-340. Document.