Consideration of Scale in Remote Sensing of Biodiversity

Abstract

A coherent and effective remote sensing (RS) contribution to biodiversity monitoring requires careful consideration of scale in all its dimensions, including spatial, temporal, spectral, and angular, along with biodiversity at different levels of biological organization. Recent studies of the relationship between optical diversity (spectral diversity) and biodiversity reveal a scale dependence that can be influenced by the RS methods used, vegetation type, and degree and nature of disturbance. To better understand these issues, we call for multi-scale field campaigns that test the effect of sampling scale, vegetation type, and degree of disturbance on the ability to detect different kinds of biodiversity, along with the development of improved models that incorporate both physical and biological principles as well as ecological and evolutionary theory. One goal of these studies would be to more closely match instrumentation and sampling scales to biological definitions of biodiversity and so improve optical diversity (spectral diversity) as a proxy for biodiversity. The ultimate goal would be to design and implement a truly effective, “scale-aware” global biodiversity monitoring system employing RS methods. Such a system could improve our understanding of the distribution and functional importance of biodiversity and enhance our ability to manage ecosystems for resilience and sustainability in a changing world.

Plant spectral diversity as a surrogate for species, functional and phylogenetic diversity across a hyper‐diverse biogeographic region

“…One approach to measuring biodiversity with remote sensing is spectral diversity (also termed optical diversity, see Gamon et al., 2020). This approach rests upon the assumption that electromagnetic radiation reflected from plants (referred to here as reflectance spectra or spectral signatures) is a signal of their underlying chemical, structural and physiological traits, as well as their taxonomic identity (Jacquemoud & Ustin, 2019; Ustin & Jacquemoud, 2020).…”

Predicting species distributions and community composition using satellite remote sensing predictors

“…This is not a small issue as no species occur within their complete geographic range, i.e., there are discontinuities in the species occurrence patterns that can be detected only at local or ecological scales. In addition, although RS-products are useful covariates for the spatial modelling of biodiversity, that allow reducing the uncertainty associated with environmental covariates in model predictions 30 , the coarse grain of current RS-products hampers accurate predictions of biodiversity at ecological scales 30,35,48 . In other words, the broad spatial resolution of current RS-products (usually with a pixel size of 30 arc-seconds or ~ 1 km) limits our ability to capture environmental features at ecological scales (e.g., microtopography, soil moisture, landscape structure) that ultimately determine the coexistence of species locally.…”

Multi‐temporal assessment of grassland α‐ and β‐diversity using hyperspectral imaging

“…This contribution, integrating spectral, spatial, and temporal scales, is valuable in light of the prospective opportunities afforded by forthcoming space‐borne missions such as NASA’s Surface Biology and Geology (Schneider et al 2019), the German Environmental Mapping and Analysis Program (EnMAP; Stuffler et al 2007), and the Italian Hyperspectral Precursor of the Application Mission (PRISMA; Pignatti et al 2013). The pixel size of these planned space‐borne imagers is expected to be approximately 30 m. Based on previous findings (Gamon et al 2020), such spatial resolution would be coarse to match the scale of field plots or map α‐diversity in grasslands, but might be suitable for assessing β‐diversity or for providing contextual information for airborne and field sampling campaigns. In general, spatial resolution of the data should be comparable to the phenomenon being observed (Levin 1992), a rule that is rarely observed in remote sensing due to the constraints of sensor and platform design.…”

“…Although many government satellite data sources are now accessible and free of charge, the caveat of such platforms is their coarse spectral and spatial resolution. Previous studies have shown that the ability of remote sensing to detect biodiversity in grasslands decreases significantly at coarse spatial and spectral resolutions (Gamon et al 2020). We thus posit that using airborne hyperspectral data with fine spatial resolution (1 m pixels or better) is imperative to meaningful multi‐temporal analyses of plant diversity in grasslands.…”