Global cloud property models for real-time triage on board visible–shortwave infrared spectrometers

Steve Chien,Winston Olson-Duvall,Amruta Yelamanchili,Raffaele Vitulli,Brian H Kahn,Macey W Sandford,Robert O Green,David R Thompson

doi:10.5194/amt-13-7047-2020

Abstract

Abstract. New methods for optimizing data storage and transmission are required as orbital imaging spectrometers collect ever-larger data volumes due to increases in optical efficiency and resolution. In Earth surface investigations, storage and downlink volumes are the most important bottleneck in the mission's total data yield. Excising cloud-contaminated data on board, during acquisition, can increase the value of downlinked data and significantly improve the overall science performance of the mission. Threshold-based screening algorithms can operate at the acquisition rate of the instrument but require accurate and comprehensive predictions of cloud and surface brightness. To date, the community lacks a comprehensive analysis of global data to provide appropriate thresholds for screening clouds or to predict performance. Moreover, prior cloud-screening studies have used universal screening criteria that do not account for the unique surface and cloud properties at different locations. To address this gap, we analyzed the Hyperion imaging spectrometer's historical archive of global Earth reflectance data. We selected a diverse subset spanning space (with tropical, midlatitude, Arctic, and Antarctic latitudes), time (2005–2017), and wavelength (400–2500 nm) to assure that the distributions of cloud data are representative of all cases. We fit models of cloud reflectance properties gathered from the subset to predict locally and globally applicable thresholds. The distributions relate cloud reflectance properties to various surface types (land, water, and snow) and latitudinal zones. We find that taking location into account can significantly improve the efficiency of onboard cloud-screening methods. Models based on this dataset will be used to screen clouds on board orbital imaging spectrometers, effectively doubling the volume of usable science data per downlink. Models based on this dataset will be used to screen clouds on board NASA's forthcoming mission, the Earth Mineral Dust Source Investigation (EMIT).

Highlights

Imaging spectrometers, known as hyperspectral imagers, collect images in the form of three-dimensional data cubes: a two-dimensional image of the target surface in the field of view and swath of the instrument with a continuous spectrum in the third dimension
The globally representative sample set of imaging spectroscopy data provides a comprehensive sample of TOA reflectance for various surface types in space, time, and wavelength as well as a prediction model for screening cloud-contaminated data on board orbital imaging spectrometers
The method described for screening cloud-contaminated data on board orbital imaging spectrometers will at least double the volume of useful data for a fixed downlink size

Summary

Introduction

Known as hyperspectral imagers, collect images in the form of three-dimensional data cubes: a two-dimensional image of the target surface in the field of view and swath of the instrument with a continuous spectrum in the third dimension. At the time of this writing, NASA is studying dramatically enhanced imaging spectrometer architectures to provide measurements with global coverage (National Academies of Sciences, Engineering, and Medicine, 2018). The swath width and length of these instruments permit growing coverage areas. Due to the high-dimensional nature of these datasets, the duty cycle of these instruments is expected to be limited by data volume.

Methods

Results

Conclusion