A statistical gap filling model for methane fluxes over an urban area in the Alps

There is consensus that climate change is mostly driven by anthropogenic greenhouse gas emissions. In addition to CO₂ emissions, which have been the subject of public debate for a long time, increased awareness of methane (CH₄) emissions has developed in recent years. CH₄ is considered the 2^nd most important contributor to radiative forcing, making it the most important non-CO₂ greenhouse gas.

Due to the relatively short lifetime of the gas in the  atmosphere compared to CO₂, the reduction of methane emissions can lead to a climate benefit on relatively short time horizons. In order to effectively reduce emissions, the polluters must be better understood and recognized. Here, we combine methane eddy covariance observations in combination with a variety of other trace gases and meteorological parameters that have been recorded since August 2020 in an Alpine city (Innsbruck, Austria) to investigate urban methane emissions. For an accurate comparison with bottom-up emission inventories we test different gap-filling methods with the help of meteorological parameters as well as other tracer fluxes, such as NO, NO₂, or CO₂. In order to quantify methane emissions in urban areas as annual totals, a complete, gap-free flux dataset is desired. We have developed different statistical gap filling models which are able to predict CH₄ fluxes at the study location. The method is based on a boosted regression tree model with a variety of meteorological and astronomical parameters, as well as other trace gas fluxes serving as input. Different combinations of these input parameters are tested for accuracy of their prediction. Contrasting other gap filling methods, used over uninhabited areas, adding gases like CO₂ or NO can serve as important additional predictors, because sectors related to combustion processes are considered as important contributors to CH₄ emissions.  In this presentation we discuss CH₄ flux measurements performed during the last 2,5 years over an urban area, and highlight first results on the performance of the developed gap filling models.