Cranfield University1, Scottish Water2
Introduction
Water utilities are increasingly adopting catchment management as an approach to improve the quality of raw drinking water in response to rising treatment costs, deteriorating raw water quality and a drive for industry sustainability. The heterogeneity present in large drinking water supply catchments along with the range of substances that must be considered under the EU Drinking Water Directive makes it difficult for utilities to select and spatially target interventions to improve raw water quality. This, along with pressure from economic regulators to spend customers' money effectively, means that there is a need for modelling methodologies that assess key multiple pollutant risk processes so that water quality interventions can be appropriately selected and targeted.
This paper describes how the CaRPoW (Catchment Risk to Potable Water) framework which assesses multiple pollutant processes within a relatively simple generic modelling framework can be used to select and target interventions for the high risk areas of a Scottish Water supply catchment
Methods/Materials
Criteria for the modelling framework were devised within an early dialogue with water industry professionals involved in catchment management, to ensure that the framework would be fit for purpose. The criteria followed the structure of Graves et al. (2005) which included setting criteria on categories such as systems represented, model objectives, viewpoint of analysis (i.e. water utility), spatio-temporal scale, generation and use of data, model platform and model inputs and outputs.
In keeping with the set criteria CaRPoW has been developed on the conceptual understanding of the pollutant transfer continuum which links together Source-Mobilisation-Delivery (S-M-D) processes in a Source-Pathway-Receptor type template (Haygarth et al., 2005; Granger et al., 2010). Within the framework both Source and Mobilisation are represented at the 'field' scale, where 'field' is defined as a unique combination of soil, land use and drainage along with the inclusion of actual field boundaries. In contrast, Delivery represents the link between mobilised pollutants at the field edge and the water body in which raw drinking water is abstracted and is therefore based on principles of hydrological connectivity at the catchment scale. Total risk (representing the g of pollutant per unit area that reach the abstraction point) is calculated by combining the Source [amount of pollutant available], Mobilisation [proportion of pollutant mobilised to field edge] and Delivery [proportion delivered to the abstraction point] coefficients. This presents a modular modelling structure in which the three main components of risk are transparent to the end user. Interventions can be selected for high risk fields according which of the three risk components is the highest e.g. source based interventions are deemed most appropriate if source component is most dominant component of total risk.
Modelling methodologies have been developed on an annual time scale for each module and pollutant (pesticides, sediment and nutrients). Source and mobilisation methodologies are generally unique to each pollutant however the connectivity method is applied to all pollutants in the same way. Once risk outputs have been generated they are compared between pollutants to identify those fields with the highest risk between multiple pollutants. Fields with shared high risk are deemed the most suitable for intervention.
Initially the framework has been applied to six pesticides in the River Ugie catchment, North East Scotland. The River Ugie has a 330 km2 catchment area with a mixed arable/grassland land use and supplies drinking water to the town of Peterhead and its surrounding area. Model outputs are validated against spatially distributed catchment water quality data.
Results from nutrient and sediment modelling will also be outlined in the presentation but not in this abstract.
Results and Discussion
Initial results are mixed for the 6 pesticides the framework was applied to when compared against water quality data. Generally for the four pesticides that are broadcast applied (Chlorotoluron, CMPP, Metazachlor, Metaldehyde) aggregated risk values for each sub-catchment correlate well spatially with sub-catchment loadings from a 3 year water quality dataset (P < 0.05). However for the other two pesticides (MCPA, and 2, 4-D) results are not as well matched to loading data. Following discussion with local agronomists we hypothesise that these pesticides are often spot applied to particular weed problems therefore making it difficult to predict (in the absence of detailed usage data) where they are applied and their loading based on land use alone.
When risk outputs are compared against one another using spearman's rank strong relationships are shown between the herbicides (Chlorotoluron, CMPP, MCPA, 2, 4-D) which are generally applied to similar cereal crops. Metazachlor conversely does not share any strong relationships with any of the herbicides and only a weak correlation with Metaldehyde. We suggest that this is because it is only applied to Oilseed Rape in the catchment which the other herbicides are not applied to.
Although results are preliminary and more model refinement is required, the outputs and risk comparisons for the 6 pesticides gives the water utility greater spatial insight into catchment pesticide risk. This has allowed Scottish Water to use the risk outputs in conjunction with pesticide loading data to prioritise sub-catchments in the catchment for investment. It is envisaged that once the model has been further refined and is better validated against pesticide loading data the higher spatial resolution risk characterisation will be used to assess individual farms and fields.
Conclusions
Following consultation from catchment management practitioners on the requirements for an intervention selection and spatial targeting tool, the CaRPoW framework was developed. CaRPoW is a modular framework based on the principles of the Source-Mobilisation-Delivery continuum. Methodologies to represent each component have been developed for pesticides and applied to the River Ugie catchment in North East Scotland. Although model performance is limited by data limitations (principally the lack of usage data) and uncertainties in calculating pesticide loads from weekly samples and ungauged flows it provides practitioners with more information about the nature and components of pollutant spatial risk.
Haygarth, P. M., Condron, L. M., Heathwaite, a L., Turner, B. L., and Harris, G. P. (2005). The phosphorus transfer continuum: linking source to impact with an interdisciplinary and multi-scaled approach. The Science of the total environment, 344(1-3), 5–14.
Granger, S.J., Bol, R., Anthony, S., Owens, P.N., White, S.M. and Haygarth, P.M. (2010). Towards a Holistic Classification of Diffuse Agricultural Water Pollution from Intensively Managed Grasslands on Heavy Soils. Advances in Agronomy. 105, 83-115.
Graves, A.R., Burgess, P.J, Liagre, F., Terreaux, J.-P and Dupraz. C. (2005) Development and use of a framework for characterising computer models of silvoarable economics. Agroforestry Systems. 65. 53-65.