A method to classify ecoclimatic arid and semiarid zones in circum-Saharan Africa using monthly dynamics of multiple indicators

Within the context of monitoring desertification using a network of observatories, one faces the problem of the representativeness of biophysical, ecological, and socioeconomic variations. Ideally, one would choose at least one observatory for each homogeneous zone using indicators that capture these variations. Here, we focused only on ecoclimatic aspects defined by the global physical and climatic conditions characterizing arid and semiarid zones. The challenging aim of this paper is to spatially identify the different patterns of ecoclimatic variations. For most of the indicators of arid or semiarid zones, temporal characteristics and interactions through a typical year are observed. In order to take into account these dynamics in the clustering approach, one must use a methodology that captures interactions between spatial location, time of measurement, and indicator measured. For this purpose, a multiway analysis, generalizing principal component analysis, has been used on internationally recognized ecoclimatic indicators that characterize arid and semiarid zones. Most of these indicators were used nondynamically or qualitatively to approve and certify observatories of the Reseau d'Observatoires pour le Suivi Ecologique a Long Termes network. A direct application of this method would be to assess homogeneity and validity of ecoclimatic characteristics for each observatory comparative to its regional location within its country or within the whole circum-Saharan region. Index Terms--Arid zone, classification, desertification, ecoclimatic, ecoregion, multiway analysis, pattern recognition, principal tensor analysis k-modes (PTAk), semiarid zone, tensor decomposition.