A predictive model of mercury fish tissue concentrations for the southeastern United States

We developed a statistical model for predicting mercury concentrations in fish tissue in four southeastern states in the United States with an emphasis on identifying important predictor variables. A number of variables that could influence mercury fish tissue concentration, including proximity to sources of mercury, environmental factors affecting mercury movement and transformation, and factors affecting mercury accumulation, were considered. The model consists of three components using classification and regression three modeling, generalized additive modeling, and universal kriging, respectively. Each modeling component accounts for a different level of variation in fish tissue mercury concentration. Important factors for predicting mercury fish tissue concentrations are: (1) location, (2) species, (3) water body pH, and (4) fish weight. South central Arkansas and southeast Mississippi are the two "hot spots" with high fish tissue mercury concentrations. In addition, relatively high mercury levels were found near the Arkansas-Louisiana border and the mid-section of Mississippi.