Lake chloride concentrations and model predictions for 49,432 lakes in the Midwest and Northeast United States.

Lakes in the Midwest and Northeast United States are at risk of anthropogenic chloride
contamination, but we have little knowledge of the prevalence and spatial distribution of
the problem. The majority of salt pollution in north temperate regions stems from road salt
application but other chloride sources include water softeners, synthetic fertilizers, and
livestock excretion. Although chloride contamination of lakes is well documented, it is
unknown how many lakes are at risk of long-term salinization. We used a quantile regression
forest to leverage information from 2,773 lakes to predict the chloride concentration of all
49,432 lakes greater than 4 ha in a 17-state area. The QRF used 22 predictor variables,
which included lake morphometry characteristics, watershed land use, and distance to the
nearest interstate and road. Model predictions had an r2 of 0.94 for all chloride
observations, and 0.87 for predictions of the mean chloride concentration observed at each
lake.