Biomass Estimation of Xerophytic Forests Using Visible Aerial Imagery: Contrasting Single-Tree and Area-Based Approaches

Abstract

A large part of arid areas in tropical and sub-tropical regions are dominated by sparse xerophytic vegetation, which are essential for providing products and services for local populations. While a large number of researches already exist for the derivation of wall-to-wall estimations of above ground biomass (AGB) with remotely sensed data, only a few of them are based on the direct use of non-photogrammetric aerial photography. In this contribution we present an experiment carried out in a study area located in the Santiago Island in the Cape Verde archipelago where a National Forest Inventory (NFI) was recently carried out together with a new acquisition of a visible high-resolution aerial orthophotography. We contrasted two approaches: single-tree, based on the automatic delineation of tree canopies; and area-based, on the basis of an automatic image classification. Using 184 field plots collected for the NFI we created parametric models to predict AGB on the basis of the crown projection area (CPA) estimated from the two approaches. Both the methods produced similar root mean square errors (RMSE) at pixel level 45% for the single-tree and 42% for the area-based. However, the latest was able to better predict the AGB along all the variable range, limiting the saturation problem which is evident when the CPA tends to reach the full coverage of the field plots. These findings demonstrate that in regions dominated by sparse vegetation, a simple aerial orthophoto can be used to successfully create AGB wall-to-wall predictions. The level of these estimations' uncertainty permits the derivation of small area estimations useful for supporting a more correct implementation of sustainable management practices of wood resources.

Data collection methods for forest inventory: a comparison between an integrated conventional equipment and terrestrial laser scanning

“…The software is used to structure the database characteristics, to record the field measurements and for the geospatial processing. In order to get reference ground data this technology was used in remote sensing studies based both on passive (Bernasconi et al 2017, Brovkina et al 2018) and active sensors (Brovkina et al 2017, Tockner et al 2017.…”