Expanding high-latitude satellite methane data using a genetic algorithm optimisation technique.

Atmospheric methane climate data sets are crucial for monitoring and mitigating global warming and associated climate change impacts. The Greenhouse Observing Satellite (GOSAT) has been monitoring atmospheric methane since 2009 with near-surface sensitivity and forms a robust long-term climate data record that is essential for tracking global emission trends and understanding methane budget. University of Leicester has developed an operational global methane dataset from level 1 GOSAT data using a proxy retrieval method that is advantageous in mitigating the effects of aerosol scattering and instrumental errors. Year-round monitoring of atmospheric methane at high-latitudes is important in the context of Arctic amplification due to excessive warming that can trigger several climate feedbacks loops such as, thawing permafrost-carbon release feedback and snow - albedo feedback. However, data acquisition over high-latitudes is limited by challenging due to frequent cloud cover and low solar illumination resulting in a significant data deficit during winter season. As an attempt to resolve this limitation, this study has used a Non-dominated Sorting Genetic Algorithm II (NSGA II) approach to optimise the post-retrieval quality filters of University of Leicester GOSAT Proxy methane datasets to increase the data volume with least impact on the data quality compared to the ground-observations. If we loosen the QA filters to capture data under challenging conditions, data quality will naturally deteriorate due to introduction of noisy measurements. NSGA method is specifically suited for addressing problems with inherently conflicting objectives like in this case. In our problem fitness of each solution is evaluated based on a combination of the number of valid GOSAT observations obtained at high latitudes and the Root Mean Square Error (RMSE) between collocated GOSAT retrievals and ground-based data from the Total Carbon Column Observing Network (TCCON) network. Results suggest GA-optimized QA filters lead to an approximately 20% increase in valid satellite soundings over high latitudes with less than 1 ppb increase in RMSE between GOSAT and TCCON soundings. The optimised data set has significant data gain over the high-latitudes with more than double gain during winter. This work demonstrates the potential of GAs for improving greenhouse gas measurement coverage and volume in challenging high-latitude regions while maintaining while maintaining the accuracy of the data.