Development of Minimum Levels for Sustainable Nutrition Soil Guidelines Using Data Mining and Distribution Fitting Software.

With pressures to reduce nutrient inputs at golf courses on the rise, new guidelines that identify the minimum nutrient requirements for turf (in contrast to guidelines based upon the average or median survey values) are needed. However, to insure that the guidelines are robust and broadly applicable, they must be based on a large number of soil samples and must be analyzed with high-resolution analytical tools. The Minimum Levels for Sustainable Nutrition (MLSN) guidelines presented here utilized a dataset of soil samples collected over the past 20 years that were all analyzed by a single laboratory – Brookside Laboratories (New Knoxville, OH). To help deal with the fact that soil test results are usually not normally distributed, we utilized a data mining approach (where new patterns are discovered from analysis of large data sets) and employed EasyFit software from Mathwave (http://www.mathwave.com) to identify the three-parameter log-logistic distribution for each soil nutrient evaluated. This distribution allowed us to identify the 10^th percentile range (where 10% of samples are lower than the threshold), and to then to use this threshold value as the minimum guideline level for each nutrient. Kolmogorov Smirnov goodness of fit for the log-logistic distribution was significant (0.02< p < 0.07) for parameters evaluated. This type of data mining approach, which analyzes already-existing databases, will become a more important research tool as funding for turfgrass research and extension dwindles, and therefore generation of new data becomes more challenging.