The Landscape Ecology of Tropical Secondary Forest in Montane Costa Rica

Multinomial logistic models of land use/land cover in montane Costa Rica and landscape pattern analysis showed that relative to agriculture, secondary forest occurred closer to old-growth forest, further from roads, in forest reserves, and at higher elevations. Collinearity between explanatory variables yielded simple multivariate models; proportion of surrounding old growth predicted secondary forest most accurately. An old-growth matrix [mean patch size (MPS) 24.5 ha], located mainly within protected areas, dominated elevations greater than 2500 m. A matrix of agriculture (MPS 23.5 ha), with smaller patches (approximately 9 ha) of secondary forest and old growth, dominated elevations from 1500 to 2500 m. Combining secondary forest with old growth decreased forest patch number and increased MPS from 7.3 to 37.1 ha. I concluded that: (a) secondary forest pattern is nonrandom, so ancillary data will aid its mapping with satellite imagery. The variables elevation, agriculture distance, road distance, and population density distinguished secondary forest from old growth with 74% accuracy; (b) socioeconomic and biological forces probably interact to create these secondary forest patterns; and (c) the strong association between secondary forest and old growth supports the concept that tropical forest recovery depends on the landscape structure of remnant forest. Received 16 February 1999; accepted 20 August 1999.