Integrating Peru forest inventory with remote-sensed data to describe tropical vegetation structure
Ivan Gonzalez, Northern Arizona University, ig299@nau.edu (Presenter)
Patrick Jantz, Northern Arizona University, patrick.jantz@nau.edu (Presenter)
Scott J. Goetz, Northern Arizona University, scott.goetz@nau.edu
Arana Alexs, Peruvian Forest service - SERFOR, earana@serfor.gob.pe
Patricia Duran, Peruvian Forest service - SERFOR, pduran@serfor.gob.pe
Ricardo de la Cruz, Peruvian Forest service - SERFOR, rdelacruz@serfor.gob.pe
The Amazon Forest is one of the largest intact regions in the world, and a key role in species diversity and carbon storage dynamics. Vegetation structure analysis is an opportunity to integrate currently available datasets from different sources that provide insights into the condition of the Amazon and other tropical forests.
In this work, we are integrating information from remote-sensed sources like GEDI and Sentinel, with the national forest inventory in Peru (<400 plots). These available datasets contain reflectance, vertical profiles, species lists, and socioeconomic details, respectively, among other kinds of information relevant to vegetation structure analysis. The purpose of this integration is to compare and validate the different data sources and also derive maps of different vegetation structure metrics, such as height and structural diversity. Also, we want to characterize different traits of Peruvian forests in terms of their vertical profiles. For this, we are extracting the relative height of GEDI L2A products, and L2B information like Plant Area Index profile (PAI) and Plant Area Volume Density profile (PAVD). Also, socioeconomic information retrieved from inventory plots is being considered to assess management strategies. In terms of the modeling approach, we are extracting vegetation traits from inventory plots and GEDI shots, then using imagery values (original bands, and vegetation indexes) for training random forest algorithms in Google Earth Engine. The prediction is being made on sentinel imagery in order to get estimated values of vegetation traits in all the pixels in the study area imagery.
Finally, we are working on the generation of an R-Shinny platform where users can access original atomic data from different sources, and the spatial aggregation of vegetation traits. The results we are developing are descriptive vertical profile curves for different metrics, wall-to-wall raster predictions, and a web-based platform capable of hosting, visualizing, and interacting with these products.
Associated Project(s):
Poster Location ID: 2-22
Presentation Type: Poster
Session: Poster Session 2
Session Date: Wed (May 10) 5:15-7:15 PM
CCE Program: BDEC