NEON-SD: A 30-m Structural Diversity Product Derived from the NEON Discrete-Return LiDAR Point Cloud

Abstract

Structural diversity (SD) characterizes the volume and physical arrangement of biotic components in an ecosystem which control critical ecosystem functions and processes. LiDAR data provides detailed 3-D spatial position information of components and has been widely used to calculate SD. However, the intensive computation of SD metrics from extensive LiDAR datasets is time-consuming and challenging for researchers who lack access to high-performance computing resources. Moreover, a lack of understanding of LiDAR data and algorithms could lead to inconsistent SD metrics. Here, we developed a SD product using the Discrete-Return LiDAR Point Cloud from the NEON Aerial Observation Platform. This product provides SD metrics detailing height, density, openness, and complexity at a spatial resolution of 30 m, aligned to the Landsat grids, for 211 site-years for 45 Terrestrial NEON sites from 2013 to 2022. To accommodate various ecosystems with different understory heights, it includes three different cut-off heights (0.5 m, 2 m, and 5 m). This structural diversity product can enable various applications such as ecosystem productivity estimation and disturbance monitoring.

Fig. 1

NEON field survey sites across the United States ((a): Alaska, (b): Hawaii, (c): Puerto Rico, (d): Contiguous US). Colors indicate land cover classes from National Land Cover Database.

Fig. 2

Examples of data excluded using three different cut-off heights in sample pixels of deciduous (a, SERC), mixed (b, ABBY) and evergreen forests (c, ABBY) for individual 30 x 30 m pixels.

Fig. 3

Maps of structural diversity in 2021_TALL_6. Vegetation Area Index (VAI), Deep Gap Fraction (DGF), and Foliage Height Diversity (FHD) with a cut-off of 0.5 m (a–c), 2 m (d–f), and 5 m (g–i); Mean Canopy Height (MCH) (j); Top Rugosity (k), Point Density (l), and NLCD land cover (m).

Fig. 4

Height comparisons between the ground surveys and our products at NEON plots. The solid black line represents the 1:1 line. The red line and surrounding grey shadow represent the regression line and 95% confidence intervals, respectively.

Fig. 5

Comparisons of quantile 99^th and 50^th heights between NEON-SD and GEDI by land cover (a,b) and point density (c,d). The colors of points and lines represent forest type in (a,b), and surrounding grey shadow represent the regression line and 95% confidence intervals. Low to high categories of point density (c,d) were classified by the mean ± 1 standard deviation of the point density. The solid black line represents the 1:1 line.

Fig. 6

Comparisons between NEON-SD’s VAI and Spectra-LAI from NEON. R² (a) and RMSD (b) with grain size for site-years 2021_MLBS_4 (24 points/m²), 2O21_SOAP_5 (20 points/m²), and 2021_TALL_6 (18 points/m²). R² (c) and RMSD (d) with grain size for site-years 2017_MLBS_2 (5 points/m²), 2O19_SOAP_4 (5 points/m²), and 2019_TALL_5 (6 points/m²).

Fig. 7

Maps of NEON-SD’s VAI (cut-off height = 0.5 m, grain size = 1 m), Spectra-LAI from NEON and NLCD land cover. VAI (a), LAI (b), and land cover (c) for site-year 2021_MLBS_4; VAI (d), LAI (e), and land cover (f) for site-year 2O21_SOAP_5; and VAI (g), LAI (h), and land cover (i) for site-year 2021_TALL_6.