Canopy structure from space using GEDI lidar

Abstract

Large-scale, globally consistent characterizations of the Earth's terrestrial ecosystems are a critical component of international conservation and ecological monitoring initiatives. To help coordinate and operationalize these efforts, the Convention on Biological Diversity (CBD) and the Group on Earth Observations Biodiversity Observation Network (GEO BON) have devised a list of essential biodiversity variables (EBVs) that can be remotely sensed (Pereira et al. 2013). Among them, lidar (light detection and ranging)-derived EBVs stand out for their ability to accurately and consistently characterize three-dimensional (3D) properties of terrestrial ecosystems, such as biomass, vertical stratification, and topography (Coops et al. 2021). Lidar data have been instrumental across a range of ecological applications, including models of tree species richness (Fagua et al. 2021), community composition (Hakkenberg et al. 2018), wildlife habitat (Burns et al. 2020), wildfire behavior (Botequim et al. 2019), and microclimate (Davis et al. 2019). Recent advances in airborne and terrestrial lidar systems offer ecologists the ability to quantify ecosystem structure with unprecedented precision and accuracy – albeit at a single point in time, over limited spatial extents, with considerable logistical hurdles, and at relatively high costs. Spaceborne lidar, on the other hand, from missions like the National Aeronautics and Space Administration's (NASA's) Ice, Cloud, and land Elevation Satellites (ICESat-1 and ICESat-2) and the Global Ecosystem Dynamics Investigation (GEDI) largely resolve or mitigate these issues – providing open, consistent, multi-temporal data on forest structure at near-global extents (Markus et al. 2017; Dubayah et al. 2020). In this letter, we focus on the GEDI lidar mission, the first spaceborne lidar investigation specifically designed to measure canopy structure for ecosystem science (Dubayah et al. 2020). Perched on the International Space Station (ISS) and since April 2019, GEDI has been operationally collecting data, specifically: vertical waveform profiles of forest structure in the equatorial to mid-latitudes (± 52°). Raw waveforms are processed to derive higher level products like plant area index, plant area volume density, foliage height diversity, and aboveground biomass (Dubayah et al. 2022; Duncanson et al. 2022). Although GEDI's near-global coverage and temporal repeat frequency have the potential to transform global forest monitoring capabilities, some important considerations should be noted for ecologists hoping to employ these novel datasets. GEDI canopy height estimates possess a vertical accuracy of approximately 2–3 m (Beck et al. 2021). This degree of vertical accuracy is due primarily to uncertainty in the horizontal geolocation of GEDI footprints and sensor pointing precision, where small differences may be accentuated over the long distances that each laser pulse travels to/from the ISS. As such, slight lateral shifts in the geolocation of an individual GEDI footprint can result in discrepancies in estimated canopy heights, especially in heterogeneous canopies with abundant gaps. Despite these challenges, version 2 GEDI products possess a geolocational accuracy (mean ± standard deviation) of approximately 10 ± 2 m – already well within the range of many current satellite-derived data products, and with accuracy expected to improve in upcoming data releases (Beck et al. 2021). Additional factors affecting vertical accuracy include signal attenuation in dense canopies and difficulties with ground finding (a critical component of height estimation) in steep terrain. Where accuracy is of paramount concern, studies have demonstrated the ability to further improve geolocational accuracies to within a few meters using return pulse correlation matching (or “bullseye”) methods that address systematic geolocation error by co-registering GEDI with airborne lidar (Hancock et al. 2019). At present, GEDI does not by itself provide spatially continuous maps of canopy structure at moderately fine spatial resolutions (less than ~100 m). Unlike gridded (or “wall-to-wall”) imagery from optical satellites like Landsat, GEDI is a sampling instrument that records height measurements in spatially discrete ~25 m diameter footprints, where interstitial areas remain unsampled between footprints (60 m along-track) and between transects (600 m across-track) (Figure 1a). Thus, although GEDI's sampling design ensures near-global coverage, it comes at the expense of unsampled gaps and globally undersampled regions owing to limitations of the ISS's orbital geometry (eg relatively sparse overpass in equatorial regions) and the inability of near-infrared lidar to penetrate clouds. To mitigate against GEDI's discontinuous sampling scheme and provide analysis-ready data for the research community, the GEDI science team has released a series of gridded data products of structural features like canopy height and aboveground biomass. These spatially continuous maps are produced by statistically aggregating GEDI footprints to coarser spatial resolution (eg 0.1–1 km) raster grids (Tang et al. 2019; Dubayah et al. 2022) or by interpolating structural features between footprints using continuous ancillary remote-sensing datasets like optical imagery and radar (Figure 1c) (Healey et al. 2020; Potapov et al. 2021). Despite the benefit of continuous spatial coverage, multiple factors may affect the accuracy and consistency of these derived maps, including the quality and quantity of lidar observations and ancillary data, as well as natural variability in forest structural conditions across biogeographic gradients (Lang et al. 2022; Wang et al. 2022). Data users should therefore carefully consider the trade-offs between statistically derived or modeled continuous maps versus the higher precision and accuracy of GEDI footprint samples. By the end of its extended mission, currently set for March 2023, GEDI is expected to have systematically collected well over 20 billion quality observations of 3D canopy structure. Moreover, GEDI may be extended for several more years, which would greatly increase sampling density and allow for more comprehensive change detection. With near-global coverage, consistent measurements, and high-frequency revisit intervals, spaceborne lidar instruments like GEDI, as well as those currently in planning (eg NASA's Surface Topography and Vegetation mission), have the potential to revolutionize the characterization of large-scale ecosystem structure and dynamics, with far-reaching implications for ecological research and management. Financial support was provided by NSF DEB award 1924942. Data used in the preparation of this letter are publicly available.