A pedometric technique to delimitate soil-specific zones at field scale

作者：Castro Franco Mauricio; Cordoba Mariano A; Balzarini Monica G; Luis Costa Jose

来源：Geoderma, 2018, 322: 101-111.

DOI：10.1016/j.geoderma.2018.02.034

• Argentina
• Digital soil mapping
• Random forest
• Spatial principal components
• Precision agriculture

Delimitation of soil types within a farm field is key for site-specific crop management. An alternative to this, is to develop pedometric techniques that allow an efficient combination of soil survey information and high-resolution terrain attribute data. The aim of this study was to present and evaluate a pedometric technique to delimit soil-specific zones at field scale by coupled Random forest, fuzzy k-means clustering and spatial principal components algorithms (RF-KM-sPCA) and by using information from soil surveys and terrain attributes derived from a digital elevation model. The protocol involves three-steps: 1) automatic classification of small (20x20m) spatial units (SU) using the knowledge of the soil map units present in the farm landscape, 2) aggregation of SUM at farm scale and 3) validation of soil-specific zones. For the first step, we used the random forest algorithm with 10 terrain attributes. For the second step, KM-sPCA algorithms were used to cluster within field SU accounting for autocorrelation. For the third step, apparent soil electrical conductivity and yield maps was used to validate the delimitation of soil-specific zones. This technique produced more contiguous zones than other cluster methods which do not use spatiality. Six farm fields with highly differences in soils were partitioned by the proposed pedometric strategy. Apparent soil electrical conductivity and yield maps present significant differences among zones in all experimental fields. This analytic strategy, based in easy-to-obtain data, could be used to improve precision agricultural managements.

• 出版日期2018-7-15