A theoretical and empirical investigation into the economic relationship between forested watersheds and water treatment costs

Abstract

Forests around the world are believed to perform important chemical and nutrient retention functions. Chemical concentration levels have been found to be lower in surface water bodies located in areas with a higher forest cover. There is increasing interest from both academics and policymakers in understanding the economic value behind these nature-based services provided by forests. Including forest cover as green infrastructure in integrated source water protection and management strategies is believed to enhance their overall economic efficiency by improving water treatability. However, the empirical evidence base linking forest cover and forest management to water treatability and treatment costs is limited, and largely absent in Canada, one of the most resource-abundant regions in the world. In order to justify investments in forest cover as green infrastructure it is vital to understand the economic benefits involved, in particular in relation to drinking water treatment. The main objective of this PhD thesis is to further analyze the relationship between forest land and water treatment, both theoretically and empirically using Canada as a case study area.

The first chapter of this PhD thesis aims to provide a theoretical framework for better understanding the costs and benefits of investment decisions in the provision of safe drinking water. More specifically, a cost minimization function is specified to reach a given water quality standard, for example based on World Health Organization guidelines. The costs are based on two possible treatment approaches that can be adopted, denoted as grey and green infrastructure, where grey infrastructure represents the traditional water treatment technologies and green infrastructure consists of forest cover (e.g. forest protection or re-afforestation). Compared to grey infrastructure, green infrastructure has been found to be less costly, but riskier to implement than grey infrastructure to improve water treatability due to the lack of engineering control and environmental uncertainties surrounding causal dose-response relationships between forest cover and water quality. An optimal control model is developed to guide social planners in combining these two complementary types of infrastructure in the most cost-effective way given assumptions about the age structure of forests, risk levels, risk aversion, and the discount rate used to value future water service delivery from green infrastructure. Any optimal allocation between grey and green infrastructure is based on balancing the marginal net benefits of both types of infrastructure. Including wildfires as an additional risk, makes green infrastructure less attractive, among others because of the introduction of additional costs such as forest protection costs and reforestation costs. More forest means a higher risk of forest fires and hence damage costs and increases the uncertainty surrounding the delivery of the water service. Accounting for the co-benefits of forests as a carbon trap increases the likelihood of investing in green infrastructure, because it reduces the risk of forest fire in the long term and hence the forest protection costs, but is highly dependent on the applied discount rate to factor these long-term benefits into present-day decision-making.

The second chapter in this PhD makes use of available empirical data for the province of Ontario in Canada, and focuses on the potential role of forest cover in potentially reducing drinking water incidents, reflecting on concerns in the first chapter about the effectiveness of green infrastructure as a means of source water protection. The publicly available Ontario drinking water quality and enforcement data base contains all drinking water incidents over a particular fiscal year that failed existing water quality standards in Ontario. The database lists all incidents, so-called adverse events, related to municipal water sources. By linking this database (n=228) to geographical information retrieved from the Ontario Land Cover (GIS) database, a set of interconnected spatial regression models are estimated, aiming to assess the relationship between forest cover and drinking water rates and between drinking water rates and drinking water safety. In the latter case, the drinking water rates are used as a proxy for the drinking water treatment costs. To this end, a spatial instrumental variable model is estimated to improve our understanding about the aforementioned (reverse) causal relationships, i.e. how drinking water rates influence incidence rates and vice versa incidence rates in turn impact water rates. A key finding is that forest cover significantly reduces the number of adverse events and drinking water rates.

In the third and final chapter of this PhD thesis, use is made of another important database, the biennial Drinking Water Plants Survey conducted by Statistics Canada for the country as a whole. The survey aims to gain insight into the financial treatment costs, water treatment characteristics, and water plant customers. The survey data are confidential and can only be accessed on-site in Statistics Canada in Ottawa after requesting permission and going through an extensive (legal) screening procedure of both student and supervisor. The collected data provides detailed insight in different treatment cost categories that can help to assess how specific cost categories are influenced by surrounding land cover across Canada. Using the detailed water treatment costs in similar spatial econometric regression models (n=1,373), accounting for potential spillover effects between neighbouring water service units, a significant negative relationship is found for Canada as a whole between forest cover and total drinking water treatment costs and the material costs incurred in drinking water treatment, whilst accounting for a range of individual water treatment plant characteristics, such as treatment capacity, treatment technology, and population served. In conclusion, in this PhD thesis I demonstrate that surrounding forest cover has a significant negative effect on water rates and incidence rates in Ontario and I show that surrounding forest cover significantly reduces water treatment costs across Canada as a whole. However, the regression models estimated in this PhD thesis are based on various far-reaching assumptions which could not be verified. These include, most importantly, the assumption that there exists a direct relationship between water rates and water treatment costs in Ontario and the assumption that the spatial analysis conducted at the level of census sub-divisions in both Ontario and Canada as a whole is able to capture upstream-downstream relationships between land cover upstream and the quality of the water intake downstream in the watersheds providing water to the drinking water treatment plants. More research is needed to validate these key assumptions.