A systematic review of the treatment of phosphorus in biogeochemical and ecological models

Aquatic biogeochemical and ecological models have become increasingly detailed, both in resolution and in the number of processes and components represented. It is timely to pause and take stock of these models. Do these models accurately reflect our current understanding of the biogeochemistry of aquatic systems? Is their predictive performance improving? Are they improving in the range of system properties and behaviours that can be predicted? Where there is variety in approaches or algorithms, can we demonstrate that one method is better than another and should be preferred? This review begins this process, focusing on how phosphorus cycles are represented in models of aquatic systems. A systematic review of 71 distinct published biogeochemical and ecological models of aquatic systems in 167 applications finds: Lake models and marine models are very similar, while river and catchment models differ. This appears largely to be due to different traditions within modelling communities rather than real differences in biogeochemical processes. River models tend to have simpler representations of phosphorus than lake or marine models. The performance of river models is usually quantitatively assessed using standard metrics (77% of publications include some performance metrics), while this is usually not the case for lake models (35% include some quantitative metric) or marine models (only 29%). Across all three domains, models are becoming more complex over time (Figure 1), but there is no clear evidence that this is improving the predictive performance of models. The appropriate degree of model complexity depends on a number of factors, including the resources and data available to support the model and the purpose for which the model is being developed. Simpler models can be more rigorously calibrated and evaluated and may have just as much predictive capacity, while physiologically-based models, which are usually more complex, may be more generalisable and better able to anticipate unexpected responses and emergent properties of the system. A range of phosphorus papers not included in current models is discussed, including the biogeochemistry of organic phosphorus, the biogeochemical implications of flow through sediments, and sediment drying and re-wetting. In some aquatic environments, these processes are likely to be important. It is concluded that a broader community toolkit of model algorithms is required