Eutrophication from agricultural sources : a comparison of SWAT, HSPF and SHETRAN/GOPC phosphorus models for three Irish catchments - Final Report

Phosphorous has been implicated as the primary cause of the deterioration of surface water quality in Ireland. Extensive water quality surveys revealed that diffuse sources (runoff and subsurface flows) from agricultural land are the major contributors of phosphorus to surface waters. The mechanism of phosphorus movement from land to water can be described by a number of mathematical models that vary in modelling approaches and scales (plot, field and catchment). In this work three efficient mathematical models of diffuse source pollution in general and of phosphorus in particular have been applied, for the first time, to three Irish catchments (Clarianna (Co. Tipperary), Dripsey (Co. Cork) and Oona (Co. Tyrone)) in order to explore the suitability of these models in Irish conditions for future use in implementing the European Water Framework Directive (WFD). The models are: (i) Soil Water Assessment Tool (SWAT), (ii) Hydrological Simulation Programme FORTRAN (HSPF) and (iii) Système Hydrologique Européen TRAnsport (SHETRAN). The first two can model phosphorus production directly while the third can calculate fields of various hydrological variables relevant to phosphorous modelling. For the latter, a generic phosphorus modelling component called Grid Oriented Phosphorus Component (GOPC) has been developed here to model the phosphorus detachment and transport by taking as inputs the hydrological fields produced by any physically-based distributed catchment model such as SHETRAN.
The three models have been successfully transferred to Irish conditions and this required the build up of a database consisting of topographic, land use and soil maps, water quality and weather data. The models application involved two stages. In the first stage, hydrological variables (evaporation, runoff, etc.) within the catchment domains were simulated by each of the three models. The second stage uses the outputs of the first in order to estimate the amount of phosphorus loss from the catchments.
The SWAT, HSPF, and SHETRAN/GOPC models have been calibrated and then compared and assessed on the basis of their ability to fit and reproduce the flow discharges and phosphorus loads and concentrations for the three test catchments. The findings from the flow and phosphorus calibrations of SWAT and HSPF models were generally consistence with what have been found from previous studies outside Ireland. However, the simple structure of the first order kinetics method used for phosphorus modelling in HSPF has generally impeded the freedom of the phosphorus calibration which was noticeably difficult. Application of the SHETRAN model to the study catchments has illustrated the importance of carefully assigning the parameters related to the soil water modelling, particularly the parameters of the van Genuchten soil-hydraulic function, in order to obtain the best results. The GOPC performance has been found to depend significantly on the SHETRAN model which provided the required hydrological variables.
The flow comparison has showed that in the three catchments, the HSPF model was the best in simulating the mean daily discharges. Moreover discharge simulation from an independent validation run of the three models in the Oona catchment have also demonstrated the superiority of the hydrological component of HSPF. However, the best calibration results for daily total phosphorus loads in the study catchments has been achieved by the SWAT model. However from the validation in the Oona catchment the HSPF has been found better than the other two models, SWAT and GOPC, in simulating the total phosphorus loads. Generally the results of total phosphorus loads from the GOPC in the three catchments were quite good and this model has reproduced some observed values better than the best model, SWAT. Simulation of the daily dissolved reactive phosphorus loads by the three models in the study catchments have showed big differences between the simulated and the observed data in most of the cases. Results for mean daily total and dissolved reactive phosphorus concentrations from the three models were not as good as the results for the loads in the three catchments.