Quantification of sources of atmospheric pollutants is crucial for regulatory purposes as well as for atmospheric science in general. Due to many physical limitations in observation and modeling, the existing methodologies have many simplifying assumptions, e.g. linear observation model or uncorrelated emission values, which cause inevitable bias in pollutant estimates. We propose to analyze and relax these assumptions to obtain more reliable pollutant estimates. We have already shown that the use of the Bayesian approach can reveal the temporal profile of pollutants as well as quantify associated uncertainty. Our main objective is to extend the Bayesian methodology to spatial-temporal pollutant emissions estimation implying new challenges such as ambiguity of pollutant origin or correlations between species or locations. Key application areas will be emissions of microplastics and microfibers (thanks to our long-term cooperation with the Norwegian institute for air research), volcanic emissions, ammonia emissions, and recent point-source atmospheric radionuclide releases.