Landscape Pattern of a Semiarid Mediterranean Area Using Remote Sensing Data- Application to the Wilaya of Saida (West of Algeria).

Abstract

The identification, delimitation and spatio-temporal evolution of landscape units is one of the major preoccupations in the Mediterranean region countries. To obtain reliable statistics and to understand the transformation mechanisms of agro-sylvo-pastoral ecosystems, an area selected at the interface of the steppe region and the Tellian region in the north-west of Algeria was chosen. This work aims to implement a landscape mapping approach for a semi-arid Mediterranean area using remote sensing data available through a vicinity analysis and to define the explanatory variables of each segmented landscape units. The landscape metric total core area is used to transform the land cover map from a discrete qualitative image to a continuous quantitative grid for each land cover type. The clustering analysis allows us to automatically identify boundaries between landscape units. The spatial logistic regression method was used to identify the predictive variables of landscape pattern by measuring the degree of correlation between landscape entity (land units) and natural and anthropic landscape factors (explicative variables).These latter represent undergoing landscape processes. The result is a conceptual model of landscape structure which can be considered as a useful tool for assessing landscape structure and process changes due to the ecological and economic context of this region.