A Data-driven Approach for Generating Leaf Tissue Nutrient Interpretation Ranges for Greenhouse Lettuce

Abstract

In the absence of controlled sufficiency studies, foliar interpretations for many horticultural crops are based on survey concentrations from small data sets. In addition, both survey and sufficiency ranges provide little interpretation regarding zones that are above or below the concentration range deemed “sufficient.” While providing a critical initial set of ranges, it was based on a limited set of data and therefore improvements in interpretation of data are needed. This study presents a novel method based on 1950 data points to create data-driven nutrient interpretation ranges by fitting models to provide more refined ranges of deficient (lowest 2.5%), low (2.5% to 25%), sufficient (25% to 75%), high (75% to 97.5%), and excessive (highest 2.5%). Data were analyzed by fitting Normal, Gamma, and Weibull distributions. Corresponding P values were calculated based on the Shapiro-Wilk test for normality for the Normal and Gamma distributions, and the Kolmogorov-Smirnov test was used for the Weibull distribution. The optimal distribution was selected based on the lowest Bayesian Information Criterion (BIC) value and visual fitness. The Weibull distribution best represented nitrogen, phosphorus, potassium, calcium, manganese, zinc, and copper, and the Gamma distribution best represented magnesium, sulfur, iron, and boron. Using the selected distributions, we propose a refined set of nutrient evaluation ranges for greenhouse-grown lettuce. These refined standards will aid growers and technical specialists in more accurately interpreting leaf tissue sample data.