Determinants of efficiency in anaerobic bio-waste co-digestion facilities: A data envelopment analysis and gradient boosting approach

Highlights: • Data envelopment analysis used to compute efficiency in anaerobic co-digestion. • Dataset comprised of industrial co-digestion facilities in the US and Germany. • Stochastic gradient boosting used to modeled determinants of efficiency. • Model hyperparameters were tuned and model was cross validated for high accuracy. • Pre-treatment, digester technology, and co-digestion were important features. -- Abstract: The objective of this study is to explore the determinants of efficiency in industrial scale co-digestion facilities. The methodology applied to finding these determinants involved a two-stage process. First, data envelopment analysis was applied to input-output data from 386 operational anaerobic digestion facilities across Germany and the USA to compute project efficiency scores. These efficiency scores were converted into a binary target variable based on the 75th percentile threshold of efficiency scores. Secondly, a tuned and cross-validated stochastic gradient boosted classification model was built to determine the effect of technical parameters on a project meeting this threshold or not. The results were as follows. Firstly, 75th percentile technical efficiencies for German and US projects were 0.82 and 0.46 respectively. Secondly, the stochastic gradient boosting model had classification accuracies of 0.77 and 0.75 for German and US projects respectively, indicating high suitability for isolating determinants of efficiency. Thirdly, the most influential features included pre-treatment stages and pre-treatment types, digester technology type, and the presence/absence of co-digestion, indicating that decision makers should focus on these technologies. This research has important implications for industrial-scale biogas facilities seeking to enhance operational efficiency and policymakers engaged in technology promotion related to anaerobic digestion.