H141-0025
Spatio-Temporal Interpolation of Cloud Data

Monday, 14 December 2020
Poster
Shane Grigsby1,2, Facundo Sapienza3, Tasha Snow1,4, Alice Cima3, Lindsey Justine Heagy3, Matthew Siegfried2, Fernando Perez3 and Jonathan Taylor5, (1)Cooperative Institute for Research in Environmental Sciences, Boulder, CO, United States, (2)Colorado School of Mines, Geophysics, Golden, CO, United States, (3)University of California, Berkeley, Statistics, Berkeley, CA, United States, (4)University of Colorado Boulder, Boulder, CO, United States, (5)Stanford University, Stanford, United States
Abstract:
Identification of cloud presence is a common remote sensing task, both for removal of cloud contaminated data, and for tracking of precipitation events. Reanalysis climate models and low-spatial high-temporal resolution NOAA observations provide a dense time series of cloud location, while moderate temporal and spatial resolution observations from MODIS and VIIRS can provide enhanced spatial detail for determining cloud presence. We develop a Bayesian approach to provide error-bounded predictions of cloud locations at arbitrary temporal and spatial scales, for simple incorporation with other coincident data sets of interest. Predictions can be constrained by simple proximity metrics (distance to observed or predicted clouds), be maximum likelihood at point in time, a stochastically simulated realization, or an error envelope. Predicted locations are observationally based, but physically informed using reanalysis wind vectors. Specific application of cloud masking is demonstrated for the ICESat, ICESat-2, and Landsat missions.

This work is part of the Jupyter meets the Earth project, supported by the NSF EarthCube program (awards 1928406 & 1928374).

<< Previous Abstract | Next Abstract >>