Question: How do properties of different vegetation components vary along ecotones of semi-deciduous forest islands, and can the depth of edge influence (DEI) of the components be detected using a novel combination of analyses?

Location: Comoé National Park (CNP), NE Ivory Coast.

Methods: Along eight transects at semi-deciduous forest islands tree individuals > 20 cm DBH were mapped. At one transect, tree and shrub individuals down to 1 cm DBH were measured and cover of species was estimated. Split moving window dissimilarity analysis (SMWDA) and moving window regression analysis (MWRA) were combined to detect statistical significance of borders in multivariate vegetation data along continuous transects, to determine the width of associated ecotones, and, thus, the DEI towards the forest interior.

Results: For trees > 20 cm DBH, a distinct boundary formation was detected, dominated by the semi-fire resistant tree species Anogeissus leiocarpus. The median of DEI towards the forest interior was 55 m. Ecotone detection with all species present revealed an interlocked sequence of ecotones for grasses, herbs, woody climbers, shrubs and trees, with each of these ecotones being narrower than the overall ecotone. DEI ranged from 10 m for grasses up to 120 m for trees and shrubs.

Conclusions: The coherent set of analyses applied proved to be an objective method for detecting borders and the width of associated ecotones. The patterns found may be explained by successional processes at the forest-savanna border. The DEI measured for the forest islands in the nearly undisturbed semi-natural system of the CNP is of relevance to concepts of core-area analysis and the protection of forest interior species in semi-deciduous forests in tropical West Africa.

Nomenclature: Lebrun & Storck (1991–1997).

Abbreviations: CNP = Comoé National Park; DEI = Depth of edge influence; MWRA = Moving window regression analysis; SED = Squared Euclidean distance; SMWDA = Split moving window dissimilarity analysis.