- Calibration. Models are often used to predict potential impairment. Model calibration is the initial testing of a model and tuning it to reflect a set of field data. Field data are necessary to guide the modeler in choosing the empirical coefficients used to predict the effect of management techniques. An example of this is the data set of measured suspended sediment and bedload sediment by stream type and stability to establish dimensionless ratio sediment rating curves used in the WARSSS methodology. These data were not collected in all areas where the model would potentially be applied, thus another type of monitoring validation is helpful to determine if the model is appropriate for extrapolation to a particular region.
- Validation. Model validation involves testing of a model with a data set representing "local" field data. This data set represents an independent source different from the data used to develop the relation. Often this data is used to extend the range of conditions for which the model was developed. Due to the uncertainty of prediction, this step is very important prior to widespread application of model output. Models can be extremely helpful in comparative analysis, even if observed values depart from measured. It is important, however to be aware of the variability in the prediction. Often this monitoring outcome develops "tighter" relations or sub-sets of the initial relation, improving understanding of the processes being predicted. An example of this type of monitoring would be similar to the effectiveness monitoring of streambank erosion rates presented previously. However, beyond measuring bank erosion rate, the observer is additionally required to measure the same parameters used to predict streambank erosion: Bank Erosion Hazard Index, or BEHI, and Near-Bank Stress or NBS. The analysis involves plotting the observed values with the predicted values for the same prediction variables. In many cases, with sufficient numbers of observations, this monitoring can lead to improved local or regional models, adapted for unique soil types and vegetation. Validation modeling provides documentation not only how well the mitigation performed, but also on the performance of the model.
Validation modeling is designed to answer very specific questions at very specific sites/reaches. Design must be matched with a strong understanding of the prediction model. Validation modeling for the dimensionless ratio sediment rating curves would involve sampling sediment over the full range of streamflows in order to compare predicted to observed values. The measurements would need to be stratified by the stream type and stability rating used for the prediction. Monitoring personnel should consult the sediment measurement protocols developed by the U.S. Geological Survey titled "Field Methods for Measurement of Fluvial Sediment" (1999), Book 3 Chapter C2.
and "A Field Calibration of the Sediment-Trapping Characteristics of the Helly-Smith Bedload Sampler" (Emmett, 1980), Emmett, W.W. 1980. A field calibration of the sediment-trapping characteristics of the Helley-Smith bedload sampler. USGS Professional Paper 1139. 44 pp.
This paper may be purchased on the USGS Publications Warehouse Web site by following the "Purchase Now" link in the right column.
Continue Monitoring Section
|
|
|
