Exceedance Probability and Predictor Data for Uranium and Radon concentrations in New Hampshire Groundwater

Statewide maps of the probability of exceeding a given concentration of either uranium (U) or radon (Rn) in New Hampshire groundwater (represented as statewide rasters) are the product of statistical analyses and are available here. The dependent variables in these statistical models were either 1) the natural log of Rn concentrations or 2) a dichotomous variable indicating the exceedance of 1 microgram per liter (μg/L) U in groundwater samples. In the case of U, a dichotomous variable was used because of the large number of measured samples below the detection limit. A boosted Regression Tree (BRT) model was selected for the U analysis with an exceedance probability of 1 ug/L because that was near the center of observed measured concentrations. Ordinary Least Squares was selected for the Rn analysis, which permitted predictions for multiple exceedance values (300, 2000, and 4000 ug/L selected) from a single model. Raster datasets included in this data release are: 1) rasters of input predictor variables for the state of New Hampshire at 30-m resolution. – Each 30-meter cell has a unique cell value (GRID_CODE) and the associated attributes used as predictors. One set of 3 rasters (North, Central, and South) is used for Rn exceedance probability calculations and the other set of 3 rasters were used for U exceedance probability calculations; 2) output rasters of the probabilities of exceeding 200, 2000, and 4,000 pCi/L for radon and 1 μg/L for uranium, at 30-m resolution. – These exceedance probability rasters are presented in integer format to store the data more efficiently. For this reason, the value of each cell must be divided by 100,000 to obtain the actual exceedance probability (a number that can range from 0 to 1).