A transformative approach to enhance the parameter information from microwave and infrared remote sensing measurements. Issue 3 (2nd July 2020)

• Record Type: Journal Article
• Authors: Koner, Prabhat K.
• Is Part Of: Big earth data. Volume 4:Issue 3(2020)
• Page Start: 322
• Page End: 347
• Publication Date: 2020-07-02
• Subjects: Infrared -- microwave -- remote sensing -- radiative transfer -- inverse problem -- parameter information; Earth sciences -- Periodicals; Earth sciences -- Research -- Periodicals; Geographic information systems Periodicals; 550
• DOI: 10.1080/20964471.2020.1776435
• Languages: English
• ISSNs: 2096-4471

Abstract

ABSTRACT: In observational science, data is the foundation of a scientific model; satellite-derived parameters serve as data for earth sciences models. The building of science is imprecise if data is ambiguous. Remote sensing 'big data' provides a wealth of information for unlocking the mysteries of earth sciences. The parameter estimation from remote sensing measurements is extremely ill-posed and the inverse method plays a significant role in extracting parameter information. In this paper, predominant stochastic inverse methods in satellite retrieval applications are critically investigated from different schools of thought and several basic flaws are revealed, e.g. error being treated as definite information. The major drawbacks of these methods include a high reliance on a priori information and binding the satellite retrievals to in situ measurements. A fundamentally different and transformative approach is explored as an alternative. A rational, reliable, and repeatable determination of geophysical parameter values from remote sensing measurements is possible using the total least squares based deterministic inverse method. It is a physical model-based data-driven optimization, where the error quantity is extracted from the problem itself for regularization on a case-by-case basis using singular vector decomposition of the augmented function of the Jacobian and the residual. By moving from the prevalent to the proposed inverse method, a paradigm shift in results from "information loss" to 'information gain' is achieved.