Accuracies in Mapping Secondary Tropical Forest Age from Sequential Satellite Imagery

Abstract Multitemporal satellite images have been used to differentiate, measure the areal extent of, and determine transitional probabilities among different secondary forest age classes in the tropics. In these studies, temporal gaps in the satellite archive are common due to difficulties associated with availability of cloud-free data, costs, temporal discontinuities between satellite sensors, etc. In this probability study, examples of multitemporal image sequences with data gaps are considered in order to illustrate error rates associated with estimating the age of secondary tropical forest. Results indicate that on the order of 8–32% of scene pixels may be misclassified with respect to age if land cover transitions which occur within the data gap(s) are ignored. Using secondary forest regeneration rates from the literature, calculations showed that these gap errors caused the above-ground biomass to be overestimated by 8–23%. The error rates are scene-dependent. The results are presented to flag a potentially significant source of error.