Loading...
AI-assisted rapid classification of infrared emitters detected by VIIRS nightfire in oil and gas facilities
Brockman, Donoven L. Steiger, Thomas
Brockman, Donoven L.
Steiger, Thomas
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2026-04
Date Submitted
Keywords
Collections
Files
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
The categorization of oil and gas flares is important for environmental monitoring, regulatory compliance, and reducing the waste of natural gas, particularly in energy-deficient regions. Flares can be categorized into upstream, midstream, and downstream oil and gas operations, as well as non-oil-and-gas industrial sources, with primary interest in identifying upstream activities. The VIIRS Nightfire algorithm, based on data from the Visible Infrared Imaging Radiometer Suite (VIIRS), detects and characterizes heat sources such as gas flares using a subset of shortwave and midwave infrared bands. Gas flare detections are typically persistent over time and have temperatures exceeding 1300 K. As new emitters are continuously detected through near-real time updates to the Nightfire algorithm, there is a growing need for automated and scalable methods to classify them by facility, fuel and technology. In this study, three types of input: 1) high-resolution daytime imagery from Google Earth, 2) VIIRS Nightfire time series (including radiance, temperature, and radiant heat), and 3) geolocation data from OpenStreetMap for reverse geocoding — were incorporated into prompt-based workflows using large language models (LLMs), including Gemini, Grok and ChatGPT. The model-generated classifications were compared against independent interpretations by four human analysts to evaluate accuracy. The objective of this work is to develop and refine an automated AI-assisted, prompt-based workflow for flare classification, enabling efficient identification and large-scale inventory development of infrared emitters globally.
Associated Publications
Rights
Copyright of the original work is retained by the author.