Towards Methane Detection Onboard Satellites

TL;DR

The paper addresses the challenge of detecting methane plumes from hyperspectral satellite data onboard spacecraft without the heavy preprocessing of orthorectification. It introduces UnorthoDOS, an unorthorectified dataset, and evaluates a UNet-based per-tile classification and segmentation pipeline against orthorectified data and a baseline methane detector. Results show that models trained on unortho data perform comparably to those trained on orthorectified data, with UNet significantly outperforming the mag1c baseline, especially in segmentation for strong plumes. This work enables real-time methane detection with reduced computational and data-downlink demands, and it provides publicly available datasets and code to support onboard deployment across upcoming missions.

Abstract

Methane is a potent greenhouse gas and a major driver of climate change, making its timely detection critical for effective mitigation. Machine learning (ML) deployed onboard satellites can enable rapid detection while reducing downlink costs, supporting faster response systems. Conventional methane detection methods often rely on image processing techniques, such as orthorectification to correct geometric distortions and matched filters to enhance plume signals. We introduce a novel approach that bypasses these preprocessing steps by using \textit{unorthorectified} data (UnorthoDOS). We find that ML models trained on this dataset achieve performance comparable to those trained on orthorectified data. Moreover, we also train models on an orthorectified dataset, showing that they can outperform the matched filter baseline (mag1c). We release model checkpoints and two ML-ready datasets comprising orthorectified and unorthorectified hyperspectral images from the Earth Surface Mineral Dust Source Investigation (EMIT) sensor at https://huggingface.co/datasets/SpaceML/UnorthoDOS , along with code at https://github.com/spaceml-org/plume-hunter.

Figures (3)

• Figure 1: Schematic representation of the unorthorectified data generation process.
• Figure 2: Overview of the detection pipeline. EMIT images are preprocessed into orthorectified and unorthorectified datasets using spatial jittering as data augmentation. The ML models provide an onboard early detection system by performing classification for the TIP satellite and segmentation for the CUE satellite.
• Figure 3: Visualisation of semantic segmentation results on 2 example tiles from the orthorectified \ref{['fig:ortho']} and the unorthorectified \ref{['fig:unortho']} datasets each. From Left to Right in each sub-figure: L1B tiles (only RGB bands shown for visualisation) overlaid with ground truth methane plume annotations; predicted semantic segmentation plume masks from UNet; segmentation predictions from mag1c foote2020fast