Validation and metrics for emissions detection by satellite

Validation and metrics for emissions detection by satellite

Introduction

Detecting and quantifying greenhouse gas emissions from individual sites by satellite remote sensing has emerged as a powerful new method in recent years. As more and more players enter the field, based on a variety of technologies for both instrumentation and data processing, there is a need for standardized methods for evaluating the performance of these systems. This document is focused on the specific example of satellite-based methane detection – but the principles can easily be applied to detection of other gases, and to detection from different platforms (such as aircraft).

These measuring systems must be validated using the statistical metrics of sensitivity, specificity and quantification accuracy. The sensitivity tells us how effective the system is at detecting an emission that is truly present (the rate of true positives). The sensitivity increases for higher (true) emissions rates, and approaches zero for very low rates – this leads to the important concept of a detection limit. The specificity tells us how good we are at suppressing false positives – declared detections when no emission of the gas of interest is truly occurring. False positives are highly undesirable by the end-users, including industrial operators seeking to monitor and manage their emissions – so there is typically a strong requirement for high specificity. Finally, the quantification accuracy must be measured and specified to give credibility to emission rates and their uncertainties reported by the measurement provider.

An effective emissions detection system has the following requirements, all of which must be satisfied simultaneously.

(a) a well-characterized detection limit based on a Probability of Detection curve
(b) a well-characterized, high specificity (low rate of false positives)
(c) well-characterized quantification accuracy

In the sections that follow, we elaborate on all these concepts with concrete examples of how they should be implemented in practice.

Read the full paper here: Validation and metrics for emissions detection by satellite