Pyplis — A Python software toolbox for the analysis of SO2 camera images for emission rate retrievals from point sources

Abstract

Ultraviolet (UV) SO2 cameras have become a common tool to measure and monitor SO2 emission rates, mostly from volcanoes but also from anthropogenic sources (e.g., power plants or ships). Over the past decade, the analysis of UV SO2 camera data has seen many improvements. As a result, for many of the required analysis steps, several alternatives exist today (e.g., cell vs. DOAS based camera calibration; optical flow vs. cross-correlation based gas-velocity retrieval). This inspired the development of Pyplis (Python plume imaging software), an open-source software toolbox written in Python 2.7, which unifies the most prevalent methods from literature within a single, cross-platform analysis framework. Pyplis comprises a vast collection of algorithms relevant for the analysis of UV SO2 camera data. These include several routines to retrieve plume background radiances as well as routines for cell and DOAS based camera calibration. The latter includes two independent methods to identify the DOAS field-of-view (FOV) within the camera images (based on (1) Pearson correlation and (2) IFR inversion method). Plume velocities can be retrieved using an optical flow algorithm as well as signal cross-correlation. Furthermore, Pyplis includes a routine to perform a first order correction of the signal dilution effect (also referred to as light dilution). All required geometrical calculations are performed within a 3D model environment allowing for distance retrievals to plume and local terrain features on a pixel basis. SO2 emission rates can be retrieved simultaneously for an arbitrary number of plume intersections. Hence, Pyplis provides a state-of-the-art framework for more efficient and flexible analyses of UV SO2 camera data and, therefore, marks an important step forward towards more transparency, reliability and inter-comparability of the results. Pyplis has been extensively and successfully tested using data from several field campaigns. Here, the main features are introduced using a dataset obtained at Mt. Etna, Italy on 16 September 2015.