Main content area

Combining high resolution satellite imagery and lidar data to model woody species diversity of tropical dry forests

George-Chacon, Stephanie P., Dupuy, Juan Manuel, Peduzzi, Alicia, Hernandez-Stefanoni, J. Luis
Ecological indicators 2019 v.101 pp. 975-984
cost effectiveness, ecosystems, environmental indicators, forest management, habitats, lidar, models, prediction, regression analysis, remote sensing, secondary forests, social welfare, species richness, texture, tropical dry forests, variance, vegetation structure, woody plants
Tropical dry forests provide goods and ecosystem services that rely on their diversity and are vital for human wellbeing. However, they are among the most threatened ecosystems due mainly to conversion to agriculture. Accurate estimations of species diversity in tropical secondary forests are needed for effective conservation and forest management. We assessed the separate and combined performance of remotely-sensed surrogates of habitat heterogeneity and vegetation structure complexity for predicting and mapping woody plant species richness and diversity in tropical dry forests. Here, we used image texture measures to calculate spectral variability from RapidEye imagery as an indicator of habitat heterogeneity as well as height and cover metrics from LiDAR data as surrogates of the complexity of vegetation structure. Separately, image texture measures and LiDAR metrics were used to explain variation in species richness and exp Shannon diversity calculated in 48 plots using multiple regression analysis. We also evaluated the relative importance of two sets of indicators to estimate species diversity using variation partitioning analyses. Habitat heterogeneity (image texture metrics) contributed most to explain variation in species richness (R2: 0.72–0.87), whereas complexity of vegetation structure (LiDAR metrics) was more important for diversity (R2: 0.68–0.74). However, a large percentage of variance of richness and diversity (58%–67%) was jointly explained by both factors and using models that combine them provided similar or higher prediction accuracy (R2: 0.68–0.89). We conclude that using image texture of high resolution imagery as an indicator of habitat heterogeneity allows precise and cost-effective estimations of species richness, while LiDAR metrics as a surrogate of vegetation structure complexity allow better estimations of diversity and that combining image texture and LiDAR provides the best estimates of species richness and diversity.