File(s) stored somewhere else
Please note: Linked content is NOT stored on University of Manchester and we can't guarantee its availability, quality, security or accept any liability.
Slides – Tomasz Janus – Open Research Conference 2024
Slides used by Tomasz Janus for the University of Manchester Open Research Conference 2024
Title: Development of transparent low-emission hydropower expansion strategies for Myanmar using bespoke open-source software and explainable AI
Abstract: We present the outcomes of our 3-year-long project on the development of open-source software and a transparent methodology for automated estimation of reservoir greenhouse gas (GHG) emissions using global and publicly accessible geospatial data. We developed two pieces of software that are currently in public domain. (1) GeoCARET (https://github.com/tomjanus/geocaret) is a Geospatial CAtchment and REservoir analysis Tool. It automates the process of delineating reservoirs and catchments and deriving reservoir and catchment specific properties from global datasets such as elevation maps, land cover maps, and various remote-sensing data, (2) RE-Emission (https://github.com/tomjanus/reemission) is our software for estimating GHG emissions from reservoirs. It relies on substantial amounts of input data that need to be sourced manually, or alternatively, can be calculated with GEOCaret. We applied our software to estimate emissions of 200+ reservoirs in Myanmar. We also developed a methodology for interpreting the emission results using explainable AI (xAI). The methodology is shared as a collection of Python and R scripts on GitHub. We present the key components of our research and highlight the opportunities, challenges and potential threats arising from research carried out in a fully transparent and open manner. We give our perspective on sharing code and significant volumes of data stemming from large projects such as ours, that require different sharing platforms and approaches. Since our software relies on many input datasets that come with their own licenses, we would like to share our experience in addressing licensing issues during software release. Finally, we would like to address the problem that is common in research software, i.e. the often complex installation and execution procedures that form a barrier for adoption by people with no or little IT background. In particular, we will talk about packaging and running software requiring multiple dependencies using Docker containers.
