Using the Lens of Uncertainty in Cloud Data Records

Wednesday, 17 December 2014
Brent C Maddux, Cooperative Institute for Meteorological Satellite Studies, Madison, WI, United States and Steven A Ackerman, University of Wisconsin Madison, Madison, WI, United States
Uncertainty is the lens through which we must view data to determine if it is 'fit-for-purpose'. Yet it is the single most neglected aspect of satellite derived cloud data records. Without a measure of uncertainty it is difficult to assess the quality of cloud data for use in any range of studies, from global climate model assessments to regional cloud process case studies. Cloud data records from the Moderate Resolution Imaging Spectroradiometer(MODIS) are considered to be reliable, inter-consistent, and well calibrated, yet no measures of uncertainty are included in any level of data. As studies utilizing MODIS cloud data become more nuanced and complex and as scientists try to squeeze more information of smaller magnitudes out of the records, it is imperative to estimate uncertainties.

We will have two purposes for study:

1) Compare natural variability to measures of uncertainty at multiple spatiotemporal scales for cloud cover and top pressure
2) Show how estimates of uncertainty might be used to determine the fit-for-purpose use of cloud data

Measures of uncertainty are provided for the MODIS cloud cover and cloud top pressure at various spatiotemporal scales, ranging in time from daily observations to decadal averages for spatial scales of 1km to global for each source of uncertainty and combined for all sources. We will focus on the combined uncertainty from all sources and for illustrative purposes will present uncertainty introduced by cloud heterogeneity. Cloud heterogeneity is the single largest source of uncertainty in satellite derived cloud data records. Discretizing the continuum of cloud for the purposes of producing cloud data introduces uncertainty at all spatial scales. We divide cloud heterogeneity into three groups: heterogeneity across a cloud top or the ‘bumpiness’ of a cloud, e.g. cloud shadow; heterogeneity of cloud cover; heterogeneity changes across data record characteristics, e.g. pixel size over disparate view geometries. MODIS’s spatial resolution of 250m allows for the assessment of cloud heterogeneity on spatial scales of importance utilizing cloud data.

Effects of these three disparate cloud heterogeneities on cloud property data records ranging from 10-year cloud cover maps at 1km spatial resolution to regional and global timeseries analyses will be presented.