Modeling Diffuse Pollution Loads with Geo-spatial Data

Jocelyn Dela-Cruz (1), Gillian Glegg (1), Christopher Macleod (2), Phil Haygarth (2)

(1) Institute of Marine Studies, University of Plymouth, Devon (UK)
(2) Soil Science and Environmental Quality Team, Institute of Grassland and Environmental Research, North Wyke Research Station, Okehampton, Devon (UK)

Diffuse pollution loads to coastal waters from land use activities may be estimated by the empirically based export coefficient model (ECM), which relies on the use of coefficients that describe the amount of pollutant loss per unit of activity (area, animals, people etc) per unit time. This modeling approach is widely and preferentially used in place of the more detailed process-based modeling because of the low data requirements and simplicity of the model. For example, the amount of total phosphorus produced from growing crops may be estimated by simply multiplying a 'crop phosphorus export coefficient' by the area of the cropland. The area of the cropland may be derived from remote sensing (RS) and geographic information systems (GIS) data sets, whereas the coefficients may be obtained from the literature. Many studies have used existing coefficients to produce reliable estimates of diffuse pollution loads. However, most of these studies have been limited to the US, UK and other parts of Europe, where the existing coefficients were originally derived from long term water quality monitoring studies and/or from process-based laboratory studies. These existing coefficients are known to reflect specific land uses, topography, geology, soil types and climate within a catchment suggesting that they are unlikely to be transferable to areas where the catchment characteristics differ significantly from those in the US, UK and Europe.

Here we present a diffuse pollution loading model based on export coefficients derived from various RS and GIS data sets. The obvious advantage of this approach is the ability to generate coefficients for areas where long term monitoring data do not exist. We present coefficients generated from RS and GIS data sets currently available for the UK. We considered the UK to be an appropriate test region for our approach as we had access to long term (20 y) water quality data for over 40 catchments, but we also had access to a range of existing coefficients that are known to produce reliable estimates of diffuse pollution loads. Access to the water quality and existing coefficient data sets allowed us to statistically test whether our generated coefficients also produced reliable estimates of diffuse pollution loads. The coefficients were generated from a multiple linear regression analysis of the long term water quality data (dependent variable), rainfall, demographic data, and data extracted from a digital elevation model, land cover, soil type and geology type grids (independent variables). We discuss the feasibility of applying the multiple linear regression equation produced from the UK data set to other countries.