Carbon efficiency analysis in the provision of drinking water: Estimation of optimal greenhouse gas emissions

Assessing carbon efficiency (CE) in the provision of drinking water services is essential to achieve a net-zero greenhouse gas (GHG) urban water cycle. Previous studies evaluating the CE of water companies are very scarce and employed parametric and non-parametric. Both methodological approaches present limitations such as overfitting issues or require assumptions about the production technology which could lead to less reliable efficiency scores. To overcome these limitations, in this study, and for the first time, we estimated CE of English and Welsh water companies using the Efficiency Analysis Trees (EAT) approach. This technique brings together machine learning and non-linear programming techniques to estimate production frontier and efficiency scores. It also allowed us to quantify the optimal level of GHG emissions in the provision of water services and estimate potential GHG savings. Bootstrap truncated regression methods were employed to quantify the impact of operating characteristics on CE of water companies. The optimal level of GHG emissions was estimated to be between 0.062 and 133.03 tons of CO2 equivalent (CO2eq) per year and per connected property. The average CE was at the level of 0.632. This means that GHG emissions could reduce by 36.8% to maintain the same level of water services. Equivalently, this corresponds to a reduction of 488,321 tons of CO2eq per year. Water only companies exhibited a better performance than water and sewerage companies with an average CE of 0.785 and 0.540, respectively. The performance of the English and Welsh water companies decreased over time. In 2011 the average CE was 0.772 whereas it went down to 0.534 in 2020. It was also estimated that on average water companies could reduce 0.034 tons of CO2eq per cubic meter of drinking water supplied and 16.16 tons of CO2eq/ connected property per year. The regression results showed that topography and water treatment complexity had a significant impact on CE. The conclusions of this study are relevant for policy makers to define policies toward a low-carbon urban water cycle.