Using benthic indicators to assess status at HUC levels: A robust hierarchical modeling approach

Decisions often need to be made about the ecological status of regions based on limited or variable quantities of data. The United States Forest Service, for example, has developed a subwatershed condition classification approach to classify status of National Forest subwatersheds (level 6 hydrological units) based on a variety of indicators including biological indicators. For forests in Western Virginia, benthic macroinvertebrate data are commonly collected and used to estimate biological status. Available data varies in sample size from 1 to 20 samples per HUC. One approach to improve estimation of the biological status of a region using the biological indicator is to use a model based estimate. A hierarchical model is proposed to aid in estimating the status of regions based on a HUC level 6 classification. The hierarchical model involves components associated with regressors at both the subwatershed and sampling level, a model for the error that allows for extreme values and for a spatial component. The model based estimates have lower mean square error than the estimates using the mean and the model may be used to estimate the status as well as the associated uncertainty or variability. The distribution of the multimetric is used to predict the risk category associated with each subwatershed.