Identification of site-specific management zones from combination of soil variables
Site-specific management demands the identification of homogeneous subfield regions within the field or management zones (ZM). However, due to the spatial variability of soil variables, determination of ZM from several variables, is often complex. Although the zonification or delimitation of MZ may...
| Autores principales: | , , , |
|---|---|
| Formato: | article |
| Lenguaje: | Español |
| Publicado: |
Corporación colombiana de investigación agropecuaria - AGROSAVIA
2019
|
| Acceso en línea: | http://revistacta.agrosavia.co/index.php/revista/article/view/239 http://hdl.handle.net/20.500.12324/35122 |
| Sumario: | Site-specific management demands the identification of homogeneous subfield regions within the field or management zones (ZM). However, due to the spatial variability of soil variables, determination of ZM from several variables, is often complex. Although the zonification or delimitation of MZ may be univariate, it is more appropriate to consider all variables simultaneously. Fuzzy k-means clustering (KM) and principal component analysis (PCA) are multivariate analyses that have been used for zonification. Nevertheless, PCA and KM have not been explicitly developed for georeferenced data. Novel versions of PCA, known as Multispati-PCA (PCAe), incorporate spatial autocorrelation among data of neighbor sites of regionalized variables. The objective of this study was to propose a new analytical tool to identify homogeneous zones from the combination of KM and PCAe on multiple soil variable data. The performance of proposed method was assessed through comparison of the average yields obtained in each zone delimited by combination of KM with PCA, as well as KM on the original variables and the new proposed method KM-PCAe. The results showed that KM-PCAe was the only method able to identify zones statistically different in terms of production potential. PCAe and its combination with KM are useful tools to map spatial variability and to identify ZM within fields from georeferenced data |
|---|