David Radcliffe(1), Oscar Shoumans(2), James Freer(3), and Faycal Bouraoui(4) (1) Department of Crop and Soil Science, University of Georgia, Athens, GA, USA, (2) Alterra, Wageningen, The Netherlands, (3) Environmental Science Department, Lancaster University, Lancaster, UK, (4) Eurpoean Commission-DG, Institute for Environment and Sustainability, Ispra, Italy
Dradclif@uga.edu Introduction
Much progress has been made in the last three decades in modeling pollutant transport. Today, there are many well-established computer models that are being used at different spatial and temporal scales to describe water, sediment, and phosphorus (P) transport. When these models are used properly, they can provide new information, but validation of models is still a problem. In this review, we describe how dynamic watershed-scale P models are being used in the US and Europe, the challenge presented by different temporal and spatial scales, and the uncertainty in model predictions.
How models are being used
Models of diffuse P transport are being used in a number of important ways. In the US, the federal Clean Water Act requires that water quality standards be set by states and water bodies monitored to determine if they are meeting these standards.
For water bodies that do not meet these standards, a total maximum daily load (TMDL) of the pollutant of concern must be set and a plan implemented to reduce the current loads to meet the TMDL. Models are used to determine the current maximum daily load and annual average load by interpolating among sparse data sets that usually provide, at best, weekly-to-monthly samples of P concentration, much of which is taken under base flow conditions. Models are also used to estimate the percent contribution from non-point sources and how that contribution is divided up among different non-point sources. However, there is a great deal of uncertainty in quantifying this distribution among non-point sources. The final step is to use models to develop different scenarios for reducing the current daily or annual load to the TMDL target, taking into account costs and the time that will be required to achieve the load reduction. Throughout the process, models with modern graphical displays, are being used as a tool for communicating with stakeholders in meetings and via websites. Outside the TMDL arena, models are playing a role in regional efforts to reduce P loading to critical water bodies such as the Chesapeake Bay and the New York City water supply system. Models have even been used as evidence in court cases such as the conflict between Arkansas and Oklahoma over P loading to drinking water reservoirs.
In Europe, the EC-Water Framework Directive (WFD; Directive 2000/60/EC) is one of the most important international driving forces to improve water quality, since 20% of all surface waters are seriously threatened with pollution and 60% of European cities overexploit their groundwater resources. Models will play a key role in the
implementation of the various directives affecting P applications and losses. Indeed the WFD requires Member States to perform an impact analysis, often done through modeling of various degrees of sophistication. Models will also play a key role in the evaluation of the river basin management plans that Member States have to develop in order to reach good chemical and ecological status by 2015. In the mean time large efforts and resources are being put into developing monitoring networks to assess the actual status of the water bodies and evaluate the efficiency of the measures taken by the different Member States. Models are being used to optimize the design of these networks and interpolate between data. The WFD requires Member States to manage at the watershed level and models are being used to understand processes and pathways operating at such large scales and also to provide an integrated response both in terms of resources (soil and water) and disciplines (ecology, hydrology, economics, etc.). Many types of models are being used and these were reviewed in the EU-EUROHARP-project (Schoumans and Silgram, 2003).
Model scale issues
In general, models can be differentiated in terms of the spatial and temporal scales for which they are best suited. In recent years, there has been a trend to scale up models from small catchments to large contributing watersheds of critical lakes, reservoirs, and estuaries. Most of these models are based on an implicit assumption that small-scale processes can be scaled up by using effective parameter values obtained through calibration (Kirchner, 2006). This has led to over-parameterized models that cannot be used to test hypotheses regarding non-point sources of P or transport processes using the monitoring data that is typically available. There is a need to develop models designed for the large watershed scale with fewer parameters and to design monitoring programs to test these models. However, selecting the most important processes and transport routes for the simplified models will require new thinking. Furthermore, lumping or eliminating processes and
pathways may limit the ability of the model to describe other scenarios. As models move to large-scale watersheds, in-stream processes become more important and the current P models differ substantially in how they describe these processes.
Temporal scales are also important and one role models may play is in showing the long response time that may be required to see improvements in water quality where P is the pollutant.
Model uncertainty
One of the reasons for a degree of skepticism among peer scientists and the public in regard to the use of models is the lack of any measure of the certainty (confidence limits) on model predictions. Despite the progress in developing new methods for quantifying model uncertainty, these methods are seldom used by modelers. Tools for measuring model uncertainty, by whatever method as long as the method is clearly identified, must become an integral part of models and be readily available for model users (Papenberger and Beven, 2006). Progress is being made along these lines in that both SWAT and HSPF now have autocalibration tools available through interfaces.
The future
There are a number of ways in which dynamic P models can be improved, beyond those already mentioned. Progress is being made in identifying critical source areas and this must be incorporated into models. A manure pool, separate from other soil P pools, is required for P when manure is applied to pasture. Better modeling of leaching of all P forms is needed (Schoumans and Chardon, 2003). Long-term nested watershed monitoring experiments for the calibration and validation of models is essential. More process-based modeling and monitoring of the effect of BMPs and combinations of BMPs at the local and watershed-scale is essential. In the near future, questions will arise regarding the impact of climate change on water quality.
From that point of view, many model processes need to be reconsidered, for example the snow melting process in the mountains and oxidation/reduction processes during long rewetting conditions in low lands. Looking longer range, we need to ask a number of questions:
o What should a ‘learning framework’ for P modeling and prediction look like?
Should we go about building models in a completely different way?
o How can we improve the way we evaluate models with limited data? Can we use ‘soft information’ or other ways of constraining models?
o Will any emergent technologies help us improve our ability to model and if so at what scales?
References
Kirchner, J.W., 2006. Getting the right answers for the right reasons: Linking measurements, analyses, and models to advance the science of hydrology. Water Resources Research.
42, W03S04.
Pappenberger, F. & Beven, K.J., 2006. Ignorance is bliss: Or seven reasons not to use uncertainty analysis. Water Resources Research. 42. W05302.
Schoumans, O.F. & Chardon, W.J., 2003. Risk assessment methodologies for predicting phosphorus losses. Journal of plant nutrition and soil science 166(4; aug), 403-408.
Schoumans, O.F. & Silgram, M., 2003. “Review and Literature Evaluation of Quantification tools for the assessment of Nutriënt Losses at catchment scale” Oslo (Norway): NIVA, 2003 (EUROHARP Report 1-2003) - ISBN 82-557-4411-5 - p. 120.