The effects of the Penry Wellfield on well-water quality

Ohio Journal of Science, The, Dec, 2007 by Keith O. Mann

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

In order to explore further the effects of pumping upon water-quality parameters, a number of additional regression analyses were conducted using a slightly modified data set. Instead of using the measured water-quality parameters, the change in value (determined by calculating the difference between sampling round 7 and 4) was used. The independent variables also differed for this portion of the analysis: the effects of water-level elevation, change in water-level, length of open-bore exposure, and distance to Well TW-5, were investigated. Subtracting the water level measured during sampling round 7 from the water level measured just prior to the Pumping Phase produced the change-in-water-level variable and the length of open-bore exposure was calculated by subtracting the elevation of the water level in the well during sampling round 7 from the elevation of the casing bottom (only eight wells had water levels that fell below the top of the casing). Table 5 contains the results of linear regression analyses performed on these new independent variables and the changes in parameter values. The results indicate that only one of the 28 models possessed an adjusted [R.sup.2] value above 0.20, while 25 of the remaining 27 models produced adjusted R2 values below or equal to 0.15.

Recovery Phase

Table 6 contains the results of the regression analyses performed on data from sampling round 9, the last sampling round. The effects of both geographic variables (west-to-east and distance-from-well TW-5) were investigated during this portion of the analyses. Again, the results are very similar to those of sampling round 7, the last sampling round of the Pumping Phase, and show that the models usually produced the highest adjusted [R.sup.2] values for alkalinity. Additionally, hardness and specific conductance often display higher adjusted [R.sup.2] values than were documented during the Background or Pumping Phases; however, all adjusted [R.sup.2] values remain low (below 0.31), with 17 of 21 values below or equal to 0.20. Multivariate analyses showed better correlation between water quality measurements and the independent variables, as one would expect. Again alkalinity, specific conductance, and dissolved solids displayed the highest correlations, with only four of 28 values larger than 0.30, yet less than 0.42. Table 6 also displays the results of both linear and multivariate regression models run on the pooled data set (sampling rounds 8 and 9) of the Recovery Phase. These results essentially repeat the results obtained for sampling round 9.

The statistical results on hardness concentration, for both sampling round 9 and the pooled data of sampling rounds 8 and 9, are peculiar: models that contain either geography or distance-from-well TW-5, show similar correlations with some of the highest adjusted [R.sup.2] values recorded during the study. This is surprising because these independent variables order the spatial distribution of the data quite differently (and in some instances oppositely): the geography variable orders the data from west to east while the distance variable arranges it radially outward from well TW-5. This curious result clearly demonstrates that regression models with [R.sup.2] values of 0.45 are essentially meaningless in a practical setting.