Michael (Iggy) Litaor, Assaf Israeli and Ofer M. Shir – Tel-Hai College and MIGAL Research Institute Upper Galilee, Israel
The objective of precision agriculture (PA) is to increase crop production system efficiency, productivity, and profitability while reducing negative environmental impacts by employing inputs at variable rates. The first step to achieve this end, the soil spatial variations must be quantified.
We hypothesized that cost-effective soilscape sampling at a model farm (37 ha located at Jezreel Valley, northern Israel) can be achieved using large ancillary datasets gleaned from proximal soil survey such as apparent electrical conductivity – ECa in vertical and horizontal modes and multispectral imagery.
There are challenging theoretical questions concerning effective soil-sampling using a minimal set of points while aiming for maximal information. Such questions require rigorous mathematical modelling and involve the perspective of tradeoffs and multi-objective (Pareto) optimization, since minimizing the number of points and simultaneously maximizing the amount of information lie in direct competition.
This study demonstrated the usefulness of using ancillary data to quantify soil variability by delineate the boundaries of site specific management units (SSMUs). The use of Pareto front algorithm employing conditional functions such as conditional Latin Hypercube sampling and Max-Min dispersion provides an optimal spatial coverage of soil sampling and analysis. Fuzzy c-mean clustering of the ancillary data identified 4 SSMUs while the Pareto front algorithm recommended 22 soil locations. We independently tested the SSMUs classification by running binary tree-model with 24 soil attributes analyzed from all soil locations and found that 90.9% of the top soils (0-30 cm) were correctly classified.
We intend to use the advanced computational platform developed in this study to further reduce soil sampling and analysis by using thermal and hyperspectral analyses.