Stochastic modelling of phosphorus transfer from agricultural land to aquatic ecosystems.

Abstract

Phosphorus is a limiting nutrient in many freshwater ecosystems and increases in its concentration can lead to eutrophication. Effective management of P in freshwaters requires quantitative estimates of P supply from all significant sources. Simple export coefficient models aim to predict annual diffuse-source nutrient transfers from catchments on the basis of constituent land use. They are attractive water quality management tools, largely due to their low input data requirements. However, the export coefficient for each land use (designed to account for all the controls on P loss including soil type, topography and prevailing meteorological conditions) must be selected from a wide range of published values. This selection is uncertain as although some sort of calibration (to match predicted with observed fluxes by altering export coefficients) may be performed, this will be poorly constrained as different combinations of export coefficients may produce similar predicted fluxes. In addition, this simple approach does not account for inter-annual variations in P losses due to climatic variations and does not explicitly account for topographic controls, distance of fields to the receiving water body or soil type. The overall aim of this thesis was to investigate modifications to the basic export coefficient model which would improve its applicability to ungauged catchments whilst retaining low, readily obtainable data requirements, without the need for extensive calibration. The modified model, “Stochastic Estimation of Phosphorus Transfer In Catchments” (SEPTIC), has been developed using GIS to exploit spatially referenced information on ■ Slope and specific cumulative area drained, derived from digital elevation data. For any given land use, fields on steep slopes adjacent to the stream network are likely to contribute more phosphorus than those on shallow slopes far from streams. ■ Soil type, using the UK Hydrology of Soil Types (HOST) classification to estimate “standard percentage runoff’ which, in turn, is used to estimate the propensity of a given soil type for exporting phosphorus. This information was used to constrain the export coefficients, which are randomly sampled from probability distributions, constructed from the range of published values, in a large number of iterations (Monte Carlo simulation). Meteorological data (hydrologically effective rainfall) is also used in the model to predict for inter-annual variations driven by changes in hydrology. The model produces frequency distributions of outputs which can be compared with the sample statistics (mean and confidence intervals) of observed fluxes. A field experiment was carried out to explore P distribution in soil and sediment deposited at field boundaries and to determine whether the model would require refinement to include these. The model has been applied to two catchments in Scotland (Greens Bum and Leet Water) for which a limited amount of data on observed P losses are available. For the Greens Bum, the model performs well in the years of application, with the predicted load always within the measured load mean ± 1SEM. For the Leet Water, the model performs reasonably well, with an overlap between the standard deviation of the predicted load and the standard error of the mean of the estimated measured load for many years. Although there is room for further development and improvement, SEPTIC represents a step forward in export coefficient modelling and is a useful screening tool for environmental managers.