Combining spaceborne lidar from the Global Ecosystem Dynamics Investigation with local knowledge for monitoring fragmented tropical landscapes: A case study in the forest-agriculture interface of Ucayali, Peru.

Abstract

Improving our ability to monitor fragmented tropical ecosystems is a critical step in supporting the stewardship of these complex landscapes. We investigated the structural characteristics of vegetation classes in Ucayali, Peru, employing a co-production approach. The vegetation classes included three agricultural classes (mature oil palm, monocrop cacao, and agroforestry cacao plantations) and three forest regeneration classes (mature lowland forest, secondary lowland forest, and young lowland vegetation regrowth). We combined local knowledge with spaceborne lidar from NASA's Global Ecosystem Dynamics Investigation mission to classify vegetation and characterize the horizontal and vertical structure of each vegetation class. Mature lowland forest had consistently higher mean canopy height and lower canopy height variance than secondary lowland forest (μ = 29.40 m, sd = 6.89 m vs. μ = 20.82 m, sd = 9.15 m, respectively). The lower variance of mature forest could be attributed to the range of forest development ages in the secondary forest patches. However, secondary forests exhibited a similar vertical profile to mature forests, with each cumulative energy percentile increasing at similar rates. We also observed similar mean and standard deviations in relative height ratios (RH50/RH95) for mature forest, secondary forest, and oil palm even when removing the negative values from the relative height ratios and interpolating from above-ground returns only (mean RH50/RH95 of 0.58, 0.54, and 0.53 for mature forest, secondary forest, and oil palm, respectively) (p < .0001). This pattern differed from our original expectations based on local knowledge and existing tropical forest succession studies, pointing to opportunities for future work. Our findings suggest that lidar-based relative height metrics can complement local information and other remote sensing approaches that rely on optical imagery, which are limited by extensive cloud cover in the tropics. We show that characterizing ecosystem structure with a co-production approach can support addressing both the technical and social challenges of monitoring and managing fragmented tropical landscapes.