Evaluating, interpreting, and communicating performance of hydrologic/water quality models considering intended use: A review and recommendations
Additional Document Info
Previous publications have outlined recommended practices for hydrologic and water quality (H/WQ) modeling, but limited guidance has been published on how to consider the project's purpose or model's intended use, especially for the final stage of modeling applications - namely evaluation, interpretation, and communication of model results. Such guidance is needed to more effectively evaluate and interpret model performance and more accurately communicate that performance to decision-makers and other modeling stakeholders. Thus, we formulated a methodology for evaluation, interpretation, and communication of H/WQ model results. The recommended methodology focuses on interpretation and communication of results, not on model development or initial calibration and validation, and as such it applies to the modeling process following initial calibration. The methodology recommends the following steps: 1) evaluate initial model performance; 2) evaluate outliers and extremes in observed values and bias in predicted values; 3) estimate uncertainty in observed data and predicted values; 4) re-evaluate model performance considering accuracy, precision, and hypothesis testing; 5) interpret model results considering intended use; and 6) communicate model performance. A flowchart and tables were developed to guide model interpretation, refinement, and proper application considering intended model uses (i.e., Exploratory, Planning, and Regulatory/. Legal). The methodology was designed to enhance application of H/WQ models through conscientious evaluation, interpretation, and communication of model performance to decision-makers and other stakeholders; it is not meant to be a definitive standard or a required protocol, but together with recent recommendations and published best practices serve as guidelines for enhanced model application emphasizing the importance of the model's intended use. 2014.