Laserfiche WebLink
JoEllen Turner <br />970 - 864 -7682 p.9 <br />eorch Environmental Pollution Consultina, LLC October 2, 2012 <br />The rationale for my selection of data was not easy since I wanted to compare data from the <br />same depth profile for each data set (not always possible) to ensure that I was comparing apples <br />with apples. Thus, I eliminated data when I was uncertain about the depth from which they were <br />sampled or if they were not sampled to the same depth. For instance I eliminated sample TP05 <br />for two reasons 1) data was only available for the top 72 cm (granted that inches was mistakenly <br />used as the unit here instead of cm which is the SI unit used for samples TP01 -04). I maintain my <br />previous opinion that the substitute soil (samples TP06 -0$) is different from the undisturbed soil <br />(Samples TPOI -04). My conclusions are in general agreement with the conclusions derived by <br />Aaron Dejoia (see details above and his report). I used the entire depth profile for 4 soil cores <br />when calculating the averages for samples B 1-4. The reason that I only took a look at 4 samples <br />and not 10 was that n=4 (number of soil core replicates) for the undisturbed soil (after <br />eliminating TP05 see comments above) and n =3 for the substitute soil - thus to compare "apples <br />with apples" I decided to just use the same or close to the same number of soil core replicates for <br />the 3 sample types (see further comments below). Based on Aaron Dejoia's report I would have <br />come to the same general conclusions if I had included all datasets. My use of the term "similar <br />to Si " was based on quick review of the data for the other soil cores - I did not try to imply that I <br />was quantitative or precise with that statement. Based on my quick review of the other data sets <br />for the B lift it appears to me that the conclusions were reasonable. For instance, I estimated a pH <br />for the B lift of 7.67 and Aaron Dejoia calculated an average pH of 7.6 for the entire dataset (I <br />was off by less than 1 %) and my estimate of the EC value was off by 7% - in conclusion I <br />slightly overestimated the pH and underestimated the EC. <br />My major concern is why only 3 samples were selected for the substitute soil but 10 to 90 <br />samples were collected for the other soils including both the B and A lift? By selecting only 3 <br />samples it might not account for the true variability within the substitute soil. In fact, I strongly <br />believe that the large confidence interval observed for the substitute soil (which also means that <br />the numbers presented for the soil substitute are less precise; see below) is due to the very <br />limited number of samples used for this calculation. <br />The impact of only using 3 samples for the calculation of the confidence interval of the pH and <br />EC is severe. Furthermore, Mr. Dejoia incorrectly uses the confidence interval determined for <br />each soil to compare the mean values and concludes that e.g., the pH of the substitute soil is not <br />statistically different from most of the other soils due to the large confidence interval for that <br />particular dataset. An ANOVA test is needed to make comparisons among mean values (see <br />details below). Another example of the problem related to not using the same number and type of <br />samples for analysis is exemplified in figure 4 of Aaron Dejoia's report in which the mixed <br />topsoil (A and B lift mixed) has a "claimed" significantly higher EC value than both the A and B <br />lift. Table 1 in Deioia's report shows the number of replicates used for calculation of all <br />chemical data e.g., 90 replicates for Lift A and 3 replicates for the substitute soil. In brief, more <br />replicates results in a better representation of the bulk soil and leads to more precise estimates of <br />the actual number for the data set. For me to trust that e.g., the given average numerical values <br />are good estimates of the bulk parameters for the substitute soil and that the substitute soil is not <br />statistically different from some of the other soils, the number of replicates have to be similar to <br />or higher than the soils to which it is being compared. If the data sets were of the same size and <br />a normal distribution can be assumed then an ANOVA analysis could be used for comparison of <br />the mean values. <br />81 Page <br />PLTF 002482 <br />