How to appropriately manage mathematical model parameters for accuracy and reliability: A case of monitoring levels of particulate emissions in ecological systems

In the last few years there have been various initiatives aimed at reaching a global consensus on carbon emission limitations. While thisgoal may finally be achieved, the fast evolving ways of data collection, analysis and dissemination pose new challenges to the accuracyand reliability of the way the stipulated limitations are measured. This paper proposes a novel methodological approach to learning rulesfrom data using particulate emissions data from a UK industrial firm which is legally required to meet well-specified standards. The datawere collected by continuously monitoring levels of emission from a coal-fired boiler using a particulate monitoring system over regularintervals and the government-imposed requirements were assessed hourly over each 24 hour period. Parameters in a uniform andmixture of two normal models are used to highlight potential technical loopholes that may cause the firm to either "pass" or "fail" theparticulate emission limitation test. The main idea of the paper is attaining perfection in measurements and predictions by buildingsharable environmental conditions for scientific decision making. Despite being based on ecological data, the findings of the paper cutacross disciplines and hence it makes recommendations for a unified modeling process that would capture not only the envisionedenvironmental goals globally but also other phenomena which may be subjected to parameter-dependent learning algorithms