| Road salt (NaCl) has been used extensively in North America for highway deicing activity which has become essential for satisfying the mobility and traffic safety demand of our society in the winter time. However, highway deicing activity has also caused many environmental concerns. This study applies nonparametric statistical methodologies to perform analyses on the laboratory analytical salinity data obtained from a variety of samples taken at carefully selected surface water, groundwater and vadose sites along a highway section in Saskatchewan. Background samples were taken at the end of 1992 before the newly constructed highway section opened to traffic. Succeeding samples were collected from 1993 to 1997 and from 2000 to 2005 when the highway section received normal deicing activity. The analytical salinity data obtained from the laboratory tests of the samples includes the concentrations of sodium, chloride, calcium, magnesium, potassium and sulfate, as well as TDS, EC, SAR and pH. Statistical comparisons of different data sets involve both paired and independent groups arranged in the form of background vs. post-background sampling, before-after the sampling gap (1998--1999), inter-seasons and inter-sites. Some robust statistical methods dealing with the data below the detection limit are applied to enhance the accuracy of data analysis. Proper robust methods are used to detect significant monotonic trends for each salinity parameter. The slopes, the magnitude of trends, are also estimated in terms of unit per year. Multiple linear regression models are established for the sodium and chloride concentrations.;Based on the collected data, vadose sites V1 and V6 are the sites most affected by road salt. Top 100mm soils at the shoulder areas of the two sites show signs of excessive sodium levels. Significant long-term increase trends have been found for road salt ions and other salinity parameters at all three types of sampling sites. Estimated trend slopes vary from less than 0.5 mg/L per year to 245 mg/L per year, depending on the media and the background data. |
