A protocol to delimitate homogeneous zones from joint spatial variability at regional scale

GIANNINI KURINA, F.; BALZARINI, M.; HANG, S.

Marrakech

Congreso; 61st World Statistics Congress; 2017

Joint spatial variability of soil and climate variables offers the opportunity to delimitate edaphoclimatic areas. We develop a zoning protocol. Homogeneous zones were identified in Cordoba, Argentina, using data from 355 sites with soil and climate data at regional scale. A two-step algorithm that considers the spatial correlation of variables for clustering sites was proposed. In the first step, the MULTISPATI-PCA method was used to obtain linear combinations of original data. In the second, Spatial Principal Components were used as input variables of the Fuzzy k-mean clustering. The protocol was run with and without pre-processing data with kriging in order to increase spatial resolution. The number of clusters was established by several external and internal validation indices. Previous spatial modelling and the use of spatially constrained PCA improved zoning in the sense of obtaining more contiguous zoning. Five edaphoclimatic areas were delimited in Cordoba, which were evaluated by cross-validation and external variables. The proposed protocol can refine environmental interpretations and advance efforts to improve natural resources managements.