| |
|
| Submitter's Name |
Rajendra Paudel |
| Session Name |
Poster Session: Hydrologic, Biogeochemical and Ecological Processes 2 |
| Category |
Hydrologic, biogeochemical and ecological processes |
| Poster Number |
232 |
| |
| Author(s) |
Rajendra Paudel, Soil and Water Science Department, University of Florida |
| |
Joong-Hyuk Min,
Soil and Water Science Department, University of Florida |
| |
James Jawitz, Soil and Water Science Department, University of Florida |
| |
|
| |
Phosphorus biogeochemical model complexity and prediction performance in a large stormwater treatment wetland of south Florida |
| |
|
| |
Reconciling the trade-offs in the level of model process complexity is a continuing challenge for mechanistic model application. Although complex models tend to be more detailed, they are not able to explain the complexity of the real systems they seek to describe. In many cases, the lack of input data is a limiting factor for an appropriate application of complex models. In most effective models, the cost of added complexity has to be in balance with the benefits of increased accuracy. The hypothesis evaluated here is that biogeochemical model prediction accuracy in treatment wetlands is increased as the modeled process complexity increases, but the benefits of increased accuracy compared to the costs of added complexity is small. We first developed four different phosphorus (P) biogeochemical models, coupled with a two-dimensional hydrodynamic model, in a 147-ha cell of a 1544-ha stormwater treatment wetland designed to help protect the greater Everglades, FL, USA. The level of complexity in the models was increased sequentially by adding internal P cycling processes. Model prediction accuracy was assessed based on outlet total P concentration data. A complexity index was developed for each model, and the overall model effectiveness was determined based on both accuracy and complexity. Results revealed that the P biogeochemical model with more complexity did not guarantee significantly better simulation accuracy compared to the simpler one. Also, results showed that the most complex model was not necessarily the most “effective” in simulating transient total P behavior. Results from this study could provide valuable insights for effective model selection based on the available data and modeling goal, and powerful diagnostic information to support future model development. |
| |