Toward a forest biomass reference measurement system for remote sensing applications

Abstract

Forests contribute to climate change mitigation through carbon storage and uptake, but the extent to which this carbon pool varies in space and time is still poorly known. Several Earth Observation missions have been specifically designed to address this issue, for example, NASA's GEDI, NASA-ISRO's NISAR and ESA's BIOMASS. Yet, all these missions' products require independent and consistent validation. A permanent, global, in situ, site-based forest biomass reference measurement system relying on ground data of the highest possible quality is therefore needed. Here, we have assembled a list of almost 200 high-quality sites through an in-depth review of the literature and expert knowledge. In this study, we explore how representative these sites are in terms of their coverage of environmental conditions, geographical space and biomass-related forest structure, compared to those experienced by forests worldwide. This work also aims at identifying which sites are the most representative, and where to invest to improve the representativeness of the proposed system. We show that the environmental coverage of the system does not seem to improve after at least the 175 most representative sites are included, but geographical and structural coverages continue to improve as more sites are added. We highlight the areas of poor environmental, geographical, or structural coverage, including, but not limited to, Canada, the western half of the USA, Mexico, Patagonia, Angola, Zambia, eastern Russia, and tropical and subtropical highlands (e.g. in Colombia, the Himalayas, Borneo, Papua). For the proposed system to succeed, we stress that (1) data must be collected and processed applying the same standards across all countries and continents; (2) system establishment and management must be inclusive and equitable, with careful consideration of working conditions; and (3) training and site partner involvement in downstream activities should be mandatory.

Keywords: Earth Observation; aboveground biomass; carbon; forest vegetation; permanent plots; representativeness; validation.

FIGURE 1

Relative environmental (top), geographical (centre) and structural (bottom) dissimilarities (%) over global forested areas with respect to conditions covered by potential forest biomass reference measurement sites (n = 195, top and centre; n = 118, bottom). Blank continental areas and hollow points (bottom), respectively, correspond to forested areas and sites not sampled (yet, for those within ±51.6° latitude) by GEDI. Relative dissimilarity was categorized for display purposes. Non‐forested areas are in grey. The map projection is EASE‐Grid 2.0 (epsg:6933), a global, equal‐area protection and spatial resolution is 5 km. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

FIGURE 2

Relative dissimilarities for different types of distances and subsets of potential forest biomass reference measurement sites. There are 1,728,368 contributing cells (5 km spatial resolution) for the environmental (left) and geographical (centre) density plots, and 829,256 for the structural (right) density plot because of GEDI discrete sampling and ISS‐orbit limited spatial coverage (±51.6° latitude). The X‐axis was cropped to 30% of relative dissimilarity for display purposes, excluding ca. 0.045% of the overall data.

FIGURE 3

Difference in relative environmental (top), geographical (centre) and structural (bottom) dissimilarities between a set of 100 randomly selected cells (median of 200 runs used) and the 100 most representative potential forest biomass reference measurement (FBRM) sites. A network made up of randomly selected cells is less representative of local conditions than one made up of the 100 most representative potential FBRM sites, wherever the difference in relative dissimilarity is positive. Difference in relative dissimilarity was categorized for display purposes. Non‐forested areas are in grey. Blank continental areas within ±51.6° latitude (bottom) correspond to areas not yet sampled by GEDI, and hollow points to sites not among the 100 most representative potential FBRM sites. The map projection is EASE‐Grid 2.0 (epsg:6933), a global, equal‐area protection and spatial resolution is 5 km. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

FIGURE 4

Relative dissimilarities versus number of locations for different types of distances and selection strategies. Only numbers of locations, n, which are multiples of 5 are used here. Lines and shaded areas correspond to the median and interquartile range of relative dissimilarity values over global forested areas, respectively.