Characterization of the Observational Covariance Matrix of Hyper-Spectral Infrared Satellite Sensors Directly from Measured Earth Views

Carmine Serio,David C Tobin,Guido Masiello,Pietro Mastro

doi:10.3390/s20051492

Abstract

The observational covariance matrix, whose diagonal square root is currently named radiometric noise, is one of the most important elements to characterize a given instrument. It determines the precision of measurements and their possible spectral inter-correlation. The characterization of this matrix is currently performed with blackbody targets of known temperature and is, therefore, an output of the calibration unit of the instrument system. We developed a methodology that can estimate the observational covariance matrix directly from calibrated Earth-scene observations. The technique can complement the usual analysis based on onboard blackbody calibration and is, therefore, a useful back up to check the overall quality of the calibration unit. The methodology was exemplified by application to three satellite Fourier transform spectrometers: IASI (Infrared Atmospheric Sounder Interferometer), CrIS (Cross-Track Infrared Sounder), and HIRAS (Hyperspectral Infrared Atmospheric Sounder). It was shown that these three instruments are working as expected based on the pre-flight and in-flight characterization of the radiometric noise. However, for all instruments, the analysis of the covariance matrix reveals extra correlation among channels, especially in the short wave spectral regions.

Highlights

The characterization of the radiometric noise, or more in general the observational covariance matrix, of remote sensing instrumentation is an important step to fully exploit the related observations for the estimation of geophysical parameters
The Principal Component Analysis (PCA) methodology developed in [3,4] was used to assess the radiometric characteristics of three major high-spectral-resolution infrared sensors, which have been put in orbit to pursue the objective of improving Numerical Weather Forecasts
We showed that IASI-C behaves according to what was expected based on the analysis performed during the pre-flight and the commissioning phase, which ended May 2019

Summary

Introduction

The characterization of the radiometric noise, or more in general the observational covariance matrix, of remote sensing instrumentation is an important step to fully exploit the related observations for the estimation of geophysical parameters. The observational covariance matrix, Sε , is an ingredient of the so-called Level 1 (L1) data. For infrared sensors, such as those we deal with in this paper, L1 data correspond to the calibrated spectral radiance. The diagonal of Sε corresponds to the variances of the observations and their square root is what is normally referred to as the radiometric noise. This expresses the precision of measurements because of instrument noise. Off-diagonal terms are negligible; non-zero covariances can arise because of post-elaboration of L1 data, such as mathematical apodization (e.g., see [1]), but they can reflect unexpected noise sources, such as mechanical micro-vibrations, thermal effects and so on [2,3,4]

Objectives

Methods

Results

Conclusion