Mapping forest age and characterizing vegetation structure and species composition in tropical dry forests

Land use changes generate a mosaic of forest patches with different ages of abandonment (i.e. succession) intermingled with other land uses. Mapping the successional age of vegetation is crucial to understand carbon accumulation patterns and the recovery of vegetation structure, diversity, and composition of forests over time. The overall objective of this research was to produce maps portraying secondary vegetation age classes and to assess how successional age classes can be related to vegetation structure, diversity and composition in two types of tropical dry forests (TDF) in the Yucatan Peninsula. We used a two-stage image classification process. First, SPOT-5 imagery were segmented and then classified using a Random Forests method. Second, the classified images were post-processed to rectify any classification errors. Additionally, we evaluated the association be-tween the different forest age classes and vegetation structure, species richness and composition using a separate Random Forests classification of field plot data. Post-processing improved the accuracy of the Random Forests classifications by 14.19% and 16.28% for the tropical semi-deciduous and semi-evergreen forests, to attain final accuracy values of 91% and 88.37%, respectively. Vegetation structure, richness and composition were all strongly associated with successional age, accounting for 77.7% and 84.7% of the total variation among forest age classes for the tropical semi-deciduous and semi-evergreen forests respectively. Therefore, the forest age maps obtained can be related to attributes of vegetation structure, diversity and composition that are useful for biodiversity conservation, forest management and climate change mitigation.

Automated summary

The study aimed to produce maps of secondary vegetation age classes and evaluate the correlation between successional age classes and vegetation structure, diversity, and composition. Through a two-stage image classification process and separate Random Forests classification of field data, results showed a strong association between successional age classes and vegetation attributes.