Durham University

Department of Geography

Departmental Research Projects

Risk-Based Modelling of Diffuse Agricultural Pollution

A research project of the Department of Geography.


A joint project between Geography (Durham) and the Centre for Sustainable Water Management (Professor Louise Heathwaite, University of Lancaster).

There is growing realisation that whilst that the localised restoration of river reaches can be effective in relation to certain ecosystem attributes, there are certain attributes that need a larger scale of analysis. This is enshrined in the new EU Water Framework Directive, which advocates holistic analysis but it applies to any land management activities that are spread across the landscape (e.g. agricultural activities) or diffuse but which integrate together to create particular point problems (e.g. flood risk, nutrient loading, fine sediment accumulation in river gravels). It has proved exceptionally difficult to demonstrate the extent to which these diffuse activities are responsible for some of these point problems. However, it is probable that more sensitive land management will lead to at least some amelioration of such point problems, especially with respect to fine sediment delivery, its linkage to degradation of fish habitat and phosphorous delivery, and the retention of overland flow on the landscape during flood events. This application seeks to develop and to test a new approach to implementing better land management in order to reduce the point impacts of diffuse activities. It recognises that the restoration of entire catchments will prove prohibitively expensive unless management activities:

  1. focus upon those parts of the catchment where restoration is likely to have most beneficial impact at the catchment scale; and/or
  2. use basic process understanding to identify the best form of restoration method

For instance, if it can be shown that a part of a catchment has low spawning potential due to fine sediment accumulation, and that this can be linked to a fine sediment delivery problem, then establishing how the fine sediment connects to the channel is a first step to improved environmental management. This connectivity may be effectively managed through river side buffer zones that are targeted upon the areas of greatest connectivity. This requires a landscape scale analysis, which is increasingly feasible for two reasons: (a) the progressive development of remote sensing and GIS technologies which allows the acquisition of highly detailed information on the spatial variation in habitat (e.g. in bed material characteristics, planform river morphology) and the controls upon that habitat (e.g. sources of fine sediment that may reduce habitat quality); and (b) new analytical tools that allow the determination of additional habitat relevant information (e.g. on the connectivity between sources of fine sediment and the drainage network). In this project we are developing and testing a tool for this purpose: SCIMAP.


The aim of the research is to develop a framework for the analysis of the relative risk of different locations within the catchment (in relation to their land use, land management etc.) in relation to different environmental requirements within receiving water bodies (e.g. fish habitat). The basis of the analysis is the joint consideration of the probability of a unit of land producing a particular environmental risk and then of that produced risk actually reaching the drainage network. Hydrologically well-connected and risky land uses should be the prime focus of management activities, and hence the result is a method for determining where efforts should be concentrated in order to achieve environmental protection.


The aim of the approach developed here is to quantify the risk associated with diffuse sources of land management activity where they impact upon surface runoff generation. By quantifying the risk, and notably the source of the riskiness, it aims to provide a tool for focusing land management activities where they are needed. Thus, it does not provide forecasts, but rather is a decision-making tool, that seeks to prioritise where concern should be directed. In this application, we seek to develop the approach to deal with: (a) fine sediment problems (e.g. in relation to fisheries decline); (b) pollutants associated with fine sediment production (e.g. sediment bound phosphorous); (c) microbiological risk; and (d) the delivery of ecosystem functions, such as water retention on the landscape, that may have have biodiversity and wildlife benefits. In developing the approach, we assume the following:

  1. The focus of the approach is diffuse pollution. Thus, sources of environmental risk are predominantly associated with distributed land use activities. Different land use activities, at the field scale, will produce different environmental risks
  2. As surface water moves across a catchment, its riskiness is controlled by the rate at which risk is acquired, which is a function of the sum of the risks of the contributing land units
  3. The level of risk from the impacts of surface runoff will be greater where high levels of risk are more readily hydrologically-connected to the stream network by overland flow than areas that are more poorly connected. This is a process that, according to current research, occurs at the sub field scale (Lane et al., 2004)
  4. A high level of risk at a point in a stream (i.e. a risky upstream contributing area) may be countered by a large upslope contributing area, as this is indicative of a high level of potential dilution
  5. Establishing risk in absolute terms will be a challenge because of: (a) uncertainties associated with the exact impacts of land management (e.g. fertiliser application rates); (b) the difficulty of establishing the exact nature of surface connectivity and the associated process rates; and (c) the fact that even resolution of (a) and (b) leaves problems as delivery processes are dynamic and hence time dependent (i.e. the absolute risk changes as a function of time).

However, for many management activities, it may only be necessary to determine the level of relative risk within a catchment. For instance, given a water quality problem that is perceived as being linked to diffuse sources, it may be sufficient for a land manager to know where, within a catchment, management activities need to be prioritised, rather than, in absolute terms, the magnitude of those problems. This opens up the probability for exploring surface-based diffuse pollution within a probabilistic framework.

Given the above, the framework proposed is designed to yield relativistic information upon the sources of catchment risk in order to optimise and to prioritise land management decisions. The approach is based upon:

  1. determination of land use risk or locational risk for each location within the landscape (this can include both diffuse and point sources of risk)
  2. evaluating the ease with which each location on the landscape surface is connected to the river network, called the connectivity risk
  3. integrating through the product of locational risk and connectivity risk to locations of interest within the drainage network (this may be 30 m reaches, say, within the main river network), and taking into account the possible mitigating effect that comes from dilution processes
  4. simulating how the effects of land management activities (e.g. land use change, set aside, buffer zones etc.) impact upon the prioritisation of risk (i.e. as an activity is modified, the system is recalculated, to prioritise new locations for management)

The locational risk estimation is based upon conventional application of export coefficient type approaches (ALH). The connectivity risk is based upon a new approach to estimation of the probability that a location in the landscape will be hydrologically-connected to the river network (Lane et al., 2004), and which is based upon the analysis of high resolution digital topographic data in relation to distributed rainfall data. The approach has two prime advantages. First, as the framework is relativist rather than absolute, the level of relativism will depend upon the spatial scale over which the risk is evaluated. This is a significant appeal of the proposed approach as it means that the analysis is relevant at multiple spatial scales. The scale at which information is yielded depends upon the chosen risk integration points. Integration at the sub-catchment scale allows determination of the relative risk associated with each sub-catchment and hence which sub-catchments need to be explored in further detail. Integration for each river reach within a sub-catchment will allow prioritisation of where within the sub catchment, field-scale land management activities might be required.

Second, ideally, as land management measures are adopted to reduce risk, it is possible to recalculate the relative risk. As risk is accumulated, this will change the relative risk of all downstream integration points and so allows the representation of the downstream benefits of a given upstream land mitigation measure. It will then change the identification of the relative risks of individual river reaches and/or sub-catchments, so allowing refocus upon different reaches. This should lead to more optimal environmental protection for a given management effort as the beneficial impacts of given activity are propagated through the system at each stage of the management activity.


From the Department of Geography

From other departments