Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Developing a Comprehensive Oil Spill Detection Model for Marine Environments
Next Article in Special Issue
Tree Species Classification from UAV Canopy Images with Deep Learning Models
Previous Article in Journal
Harbor Detection in Polarimetric SAR Images Based on Context Features and Reflection Symmetry
Previous Article in Special Issue
Identifying Old-Growth Forests in Complex Landscapes: A New LiDAR-Based Estimation Framework and Conservation Implications
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 16
Issue 16
10.3390/rs16163078
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Derivation and Evaluation of LAI from the ICESat-2 Data over the NEON Sites: The Impact of Segment Size and Beam Type

by
Yao Wang
1,2,* and
Hongliang Fang
2
1
Chongqing Key Laboratory of GIS Application, School of Geography and Tourism, Chongqing Normal University, Chongqing 401331, China
2
LREIS, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2024, 16(16), 3078; https://doi.org/10.3390/rs16163078
Submission received: 25 July 2024 / Revised: 11 August 2024 / Accepted: 19 August 2024 / Published: 21 August 2024
(This article belongs to the Special Issue LiDAR Remote Sensing for Forest Mapping)
Browse Figures
Figure 1
<p>The spatial distribution of (<b>a</b>) ICESat-2 data over the 12 National Ecological Observatory Network (NEON) sites and (<b>b</b>) an example of ICESat-2 data at the BART site. The background is a land cover map from the National Land Cover Database (NLCD).</p> "> Figure 2
<p>Statistics of point density along track distance for (<b>a</b>) ICESat-2 strong beam, (<b>b</b>) ICESat-2 weak beam, and (<b>c</b>) ALS data for different segment sizes (20 m, 100 m, and 200 m). DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.</p> "> Figure 3
<p>Comparison between the ICESat-2 LAI and the ALS LAI for different segment sizes and beam types. The upper (<b>a</b>–<b>c</b>), middle (<b>d</b>–<b>f</b>), and lower (<b>g</b>–<b>i</b>) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.</p> "> Figure 4
<p>Comparison between the LAI values derived from strong-beam ICESat-2 and ALS for DF, EF, MF, and WET. The upper (<b>a</b>–<b>d</b>), middle (<b>e</b>–<b>h</b>), and lower (<b>i</b>–<b>l</b>) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.</p> "> Figure 5
<p>The correlation between ICESat-2 LAI and ALS LAI of each NEON site for different segment sizes and beam types. See <a href="#remotesensing-16-03078-t001" class="html-table">Table 1</a> and <a href="#remotesensing-16-03078-f002" class="html-fig">Figure 2</a> for the site names and land cover types, respectively.</p> "> Figure 6
<p>Comparison between the ICESat-2 LAI from all beams, strong beams, and weak beams and the CGLS LAI. The upper (<b>a</b>–<b>c</b>), middle (<b>d</b>–<b>f</b>), and lower (<b>g</b>–<b>i</b>) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.</p> "> Figure 7
<p>Comparison between the LAI derived from all-beam ICESat-2 and CGLS for DF, EF, MF, and WET. The upper (<b>a</b>–<b>d</b>), middle (<b>e</b>–<b>h</b>), and lower (<b>i</b>–<b>l</b>) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.</p> "> Figure A1
<p>The variation in LAI bias at different <math display="inline"><semantics> <mrow> <msub> <mi>ρ</mi> <mi>v</mi> </msub> </mrow> </semantics></math>/<math display="inline"><semantics> <mrow> <msub> <mi>ρ</mi> <mi>g</mi> </msub> </mrow> </semantics></math> values. The black dashed line represents the 1/3 value used in this study.</p> "> Figure A2
<p>The example profile of ICESat-2 photons along track distance (ATD) for DF (<b>a</b>,<b>b</b>), EF (<b>c</b>,<b>d</b>), MF (<b>e</b>,<b>f</b>), and WET (<b>g</b>,<b>h</b>) types. The left and right panels correspond to strong and weak beams, respectively. The classified photons are from ATL08 data products. The top of the canopy, canopy photons, and ground photons are marked as light-green, forest-green dots, and orange dots, respectively. DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.</p> "> Figure A3
<p>The distribution and seasonal variation in field LAI for overall, overstory, and understory at typical NEON sites. The ratio is understory LAI divided by overall LAI.</p> "> Figure A4
<p>The ATL08 photon classification (left panel) and composed ATL08 and manual photon classification (right panel) at four example sites. (<b>a</b>,<b>e</b>) SERC site, (<b>b</b>,<b>f</b>) DELA site, (<b>c</b>,<b>g</b>) BART site, and (<b>d</b>,<b>h</b>) DSNY site.</p> ">
Versions Notes

Abstract

:
The leaf area index (LAI) is a critical variable for forest ecosystem processes. Passive optical and active LiDAR remote sensing have been used to retrieve LAI. LiDAR data have good penetration to provide vertical structure distribution and deliver the ability to estimate forest LAI, such as the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2). Segment size and beam type are important for ICESat-2 LAI estimation, as they affect the amount of signal photons returned. However, the current ICESat-2 LAI estimation only covered a limited number of sites, and the performance of LAI estimation with different segment sizes has not been clearly compared. Moreover, ICESat-2 LAIs derived from strong and weak beams lack a comparative analysis. This study derived and evaluated LAI from ICESat-2 data over the National Ecological Observatory Network (NEON) sites in North America. The LAI estimated from ICESat-2 for different segment sizes (20, 100, and 200 m) and beam types (strong beam and weak beam) were compared with those from the airborne laser scanning (ALS) and the Copernicus Global Land Service (CGLS). The results show that the LAI derived from strong beams performs better than that of weak beams because more photon signals are received. The LAI estimated from the strong beam at the 200 m segment size shows the highest consistency with those from the ALS data (R = 0.67). Weak beams also present the potential to estimate LAI and have moderate agreement with ALS (R = 0.52). The ICESat-2 LAI shows moderate consistency with ALS for most forest types, except for the evergreen forest. The ICESat-2 LAI shows satisfactory agreement with the CGLS 300 m LAI product (R = 0.67, RMSE = 1.94) and presents a higher upper boundary. Overall, the ICESat-2 can characterize canopy structural parameters and provides the ability to estimate LAI, which may promote the LAI product generated from the photon-counting LiDAR.

1. Introduction

The leaf area index (LAI) is defined as one-half of the total green leaf area per unit ground area [1,2] and is a critical variable for ecosystem processes and land surface models [3,4]. Both direct and indirect measurement methods can be used to obtain forest LAI [5,6], whereas large-scale in situ measurements of forest LAI are difficult and labor-intensive. Compared with field measurements, remote sensing gives the ability to estimate LAI over a large area [7]. Passive optical remote sensing has been widely used to retrieve LAI and generate LAI products through canopy reflectance and vegetation indices (VIs) [8]. However, passive optical remote sensing faces the saturation effect because canopy reflectance and VI gradually stop changing with the increase in LAI [8,9,10,11]. Given the advantage of directly detecting the vertical structure of forests, active light detection and ranging (LiDAR) remote sensing can alleviate the saturation effect and presents great promise for estimating forest LAI [12,13,14]. The LiDAR-derived LAIs present a higher upper boundary (>5) [10,15] and capture >70% of the variations in the field LAI [16,17].
LiDAR systems can be categorized into discrete return, full waveform, and photon counting based on the detection technology [12]. Meanwhile, LiDAR systems can be mounted on different platforms (terrestrial-based, airborne-based, and spaceborne-based) [14]. Airborne LiDAR data from discrete return and full waveform have been used for forest LAI estimation from the site and regional scales with the development of laser scanners and platforms [14,18,19]. The international public airborne LiDAR data projects, such as Goddard’s LiDAR, Hyperspectral and Thermal (G-LiHT) airborne imager [20], the Global Airborne Laser Scanning Data Providers Database (GlobALS) [21], and the National Ecological Observatory Network (NEON) [22], offer a great amount of data to the LAI estimation.
Spaceborne LiDAR is commonly used to estimate forest LAI at the regional or global level with the advantage of global data coverage [23,24]. As the first spaceborne full waveform LiDAR, the Geoscience Laser Altimeter System (GLAS) onboard the Ice, Cloud, and Land Elevation Satellite (ICESat) has been explored to map LAI over large areas and shows moderate agreement with reference data (R2 > 0.65) [24,25,26]. However, the 70 m footprint size and 170 m along-track sampling interval make GLAS sensitive to slope and sparsely distributed [27,28]. Moreover, the GLAS mission was stopped in 2009. The Global Ecosystem Dynamics Investigation (GEDI) installed on the International Space Station is a new spaceborne full waveform LiDAR with a 25 m footprint size and a 60 m along-track sampling interval [29]. The improvement in footprint size and sampling interval promotes the large-scale LAI estimation from GEDI [28,29,30]. GEDI LAI presents good agreement with field data (R2 > 0.75 and RMSE < 0.96) [28,31]. However, the large footprint size and along-track sampling interval of full-waveform LiDAR limited the near-continuously LAI estimation [32].
The Advanced Topographic Laser Altimeter System (ATLAS) onboard the Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2), as a follow-on mission to ICESat, is a new spaceborne photon-counting LiDAR [33,34]. Compared to ICESat/GLAS full-waveform LiDAR, ICESat-2/ATLAS has a small footprint size of 14 m and a high laser pulse repetition rate with a sampling interval of 0.7 m, which produces a relatively dense sampling interval and high point density along the track direction [35,36]. ICESat-2 can characterize three-dimensional vegetation information and deliver the ability to estimate canopy height, canopy cover, and forest aboveground biomass in different forest types [37,38,39,40,41]. Meanwhile, ICESat-2 data were also explored for LAI estimation. Zhang et al. [42] made the first attempt to estimate LAI from ICESat-2 at two forest sites, and their results showed moderate agreement (R > 0.63 and RMSE < 2.55) with MODIS and Sentinel-2 outputs. Recently, Guo et al. [43] also calculated LAI from ICESat-2 at one forest site, and the derived LAI was found to be consistent with the airborne laser scanning (ALS) estimation (R = 0.71). However, the current ICESat-2 LAI estimation was only conducted at a limited number of forest sites, and the time difference between ICESat-2 and reference ALS data may lead to a limited conclusion. Thus, there is a need to better understand the performance of ICESat-2 for LAI estimation in large areas.
The segment size and beam type are crucial for estimating forest structural parameters using the ICESat-2 data [44]. The segment size represents an along-track span of photons from a single ground track, and the beam type describes the transmitted energy and is divided into strong and weak beams with an energy ratio of 4:1 [45]. The setting and choice of ICESat-2 data segment size and beam type affect the number of signal photons returned, which are important for LAI estimation [46]. The current ICESat-2 LAI validation study neglects the scale difference between ICESat-2 and reference data [42], and the effect of segment size on LAI estimation is rarely explored [43,47]. Furthermore, existing studies have only used the strong beam to estimate LAI [42,43], whereas the weak beam has not been evaluated. Therefore, there is a need to evaluate the capability of LAI estimation with different segment sizes and comparative analysis of ICESat-2 LAI derived from strong and weak beams.
The NEON ALS data span a wide range of elevations, slopes, canopy covers, and land cover types. Therefore, the concurrent acquisition of ICESat-2 and ALS data at NEON sites enables the evaluation of the performance of LAI estimation with ICESat-2 over large areas. This study aims to derive and evaluate LAI from the ICESat-2 data over the NEON sites in North America. The specific objectives are to (1) evaluate the accuracy of ICESat-2 LAI by comparing these with concurrent ALS data, (2) explore the optimal segment size and the impact of strong and weak beams in LAI estimation, and (3) compare the ICESat-2 LAI with the Copernicus Global Land Service (CGLS) 300 m LAI product.

2. Study Area and Data

2.1. Study Area and Field Data

The study area is in the 12 NEON sites (https://www.neonscience.org/, accessed on 5 May 2023) (Figure 1; Table 1). The main land cover types of the 12 NEON sites include deciduous forest (DF), evergreen forest (EF), mixed forest (MF), and woody wetlands (WET). The elevation ranges from 20 to 2149 m, the mean annual precipitation from 655 to 2451 mm, and the mean annual temperature from 0.3 to 22.5 °C.
The field LAI data of typical sites were derived from the digital hemispherical photography (DHP) data provided by the Ground-Based Observations for Validation (GBOV) service [48,49].

2.2. ALS Data

NEON ALS data were acquired from surrounding the sites during peak greenness [22]. In this study, the ALS data from the 12 NEON sites were obtained from April to August 2019 [50] (Table 2). The LiDAR instrument was an Optech Gemini system, and the laser beam wavelength and beam divergence were 1064 nm and 0.25 mrad, respectively.

2.3. ICESat-2 Data

The ICESat-2/ATLAS has three pairs of laser beams with a wavelength of 532 nm, and there is an across-track spacing of 3.3 km in two neighboring pairs [51]. Each beam pair was split into strong and weak beams with an energy ratio of 4:1 [52]. The two laser beams of each pair were separated by 90 m and 2.5 km in the across-track and along-track distance, respectively. The laser footprint size was nominally 14 m, and the sampling interval was 0.7 m in the along-track direction.
Data from the ATL08 Version 5 were used in this study. The ATL08 data list the photon classification flags and provide both the terrain and canopy heights at the 100 m data segment size [35]. Each segment of ATL08 consisted of five consecutive 20 m segments from the ATL03 data. Using the collection time and extent, the ATL08 data at the 12 NEON sites were downloaded from the National Snow and Ice Data Center (Table 1).

2.4. CGLS LAI Collection

The CGLS provides global 300 m LAI data with a temporal resolution of 10 days. The 300 m LAI product was produced from Sentinel-3 OLCI instantaneous top-of-canopy reflectance or PROBA-V daily top-of-the-atmosphere reflectance under a neural network [53]. The production of the 300 m CGLS LAI product includes a two-step process [54,55]. Instantaneous or daily reflectance was first transformed into instantaneous LAI estimation. A dedicated compositing scheme (including smoothing and gap filling) was then used to generate the 10-day product values.

2.5. Ancillary Dataset

The National Land Cover Database (NLCD) 2019 with 30 m resolution focused on providing the NLCD land cover products [56,57]. The land cover types of the ICESat-2 segment are identified by the NLCD data. The ICESat-2 segments were categorized into four different types during the validation process: deciduous forest (DF), evergreen forest (EF), mixed forest (MF), and woody wetlands (WET).

3. Methodology

3.1. Derivation of LAI from ICESat-2 Data

3.1.1. ICESat-2 Data Processing

ICESat-2 data were selected based on the quality flag and number of photons [45]. Each segment contains at least 50 signal photons within 100 m to accurately represent the surface. The selected ICESat-2 photons were divided into different segment sizes and beam types. The segment sizes were 20, 100, and 200 m along track distance, respectively. The beam types were strong and weak beams according to the atlas_beam_type flag. The photon point density of each segment was obtained by dividing the total number of photons by segment size.

3.1.2. ICESat-2 LAI Estimation

The classification of each photon was extracted from ATL08. The photons are classified as the noise, ground, and canopy returns. Flags 0, 1, 2, and 3 refer to noise, ground, canopy, and top-of-canopy photons, respectively [35]. The ICESat-2 LAI was derived from the Beer–Lambert method [42,58]:
L A I = 1 G ∗ ln 1 + R v ρ v ρ g ∗ R g
R v / R g was calculated as the ratio of canopy photons to the number of ground photons. ρ v / ρ g is the reflectance ratio of canopy and ground and was set as 1/3 at 532 nm [42,59]. G is the leaf projection function, and G = 0.5.

3.2. Deriving LAI from ALS

3.2.1. ALS Data Processing and Clipping

The laspy Python library (https://pypi.org/project/laspy/, accessed on 6 March 2023) was used to read ALS data. The preprocessing of ALS data includes removing the noisy points, normalizing the return height using NEON digital terrain model (DTM) data, and correcting the LiDAR intensity [60,61]. After data preprocessing, the ALS data were clipped to match the ICESat-2 segment data. The center geolocation of the segment in ATL08, spacecraft orientation, laser footprint size, and segment size were used to build the polygon. Then, the ALS data were clipped using the polygon. Finally, the ALS segment data were obtained with dimensions of 20 m × 14 m, 100 m ×14 m, and 200 m ×14 m.

3.2.2. LAI Estimation from ALS

The combined light penetration index (LPIRI) was used for LAI estimation, following the Beer–Lambert law [58,62]. The LPIRI presents an advantage for LAI estimation from ALS [28]. The LAI of each ALS segment was calculated as
L A I = 1 G ∗ ln L P I R I
L P I R I = 1 2 1 + 1 L P I R + 1 ρ v ρ g ∗ 1 − L P I I n t L P I I n t
L P I R = N g r d N a l l
L P I I n t = I g r d I a l l
where G and ρ v / ρ g are the same as in Equation (1), and their values are 0.5 and 1.5, respectively. N and I are return number and intensity, respectively, and the subscripts “grd” and “all” refer to the ground and total returns, respectively.

3.3. Comparison of ICESat-2 LAI with ALS and CGLS Data

In this study, the LAI derived from the ICESat-2 data was compared with those from ALS at the same segment size. Additionally, the LAI from ICESat-2 was compared with the CGLS LAI product. To minimize the difference between the segment size of ICESat-2 and the pixel size of CGLS, only the ICESat-2 data segment of sufficient length (greater than 50 m) in each CGLS LAI pixel was selected. The ICESat-2 LAI within the CGLS pixel was then compared with the CGLS LAI
The R, RMSE, rRMSE, and Bias metrics were used to assess the ICESat-2 LAI estimation:
R = ∑ i = 1 n x i − x ¯ y i − y ¯ ∑ i = 1 n x i − x ¯ 2 ∑ i = 1 n y i − y ¯ 2
R M S E = ∑ i = 1 n y i − x i 2 n
r R M S E = R M S E x ¯ × 100 %
B i a s = 1 n ∑ i = 1 n y i − x i
where x i and y i are the reference LAI and estimated LAI, respectively. x ¯ and y ¯ are the mean values of reference LAI and estimated LAI, respectively.

4. Results

4.1. Comparison of Photons from Strong and Weak Beams

The point density of ICESat-2 and ALS at different segment sizes is shown in Figure 2. For the strong beam, the point density increased with the segment size, except for the ABBY and OSBS sites (Figure 2a). The point density of the weak beam does not increase with the segment size, and the lowest density is presented at 100 m for most sites. The point density of the weak beam is >0.75 pts/m (Figure 2b), and the strong-beam photon density is almost twice that of weak beams. Strong beams received more photon signals than weak beams and more clearly depicted the forest structure and terrain (Figure A1). In contrast, the point density of ALS data is relatively higher than ICESat-2 (Figure 2c).

4.2. Comparison of LAI from ICESat-2 and ALS

ICESat-2 LAI estimation from different segment sizes and beam types are compared with those of ALS LAI (Figure 3; Table 3). ICESat-2 LAI exhibits a moderate correlation with ALS LAI (R ≥ 0.48), except for the weak beam at the 20 m segment. The LAI derived from strong beams (R > 0.62, Table 3) performs better than those of weak beams. The correlation between the ICESat-2 LAI and ALS LAI increased with segment size, and the LAI estimated from the strong beam at the 200 m segment shows the highest consistency (R = 0.67, Figure 3h). In contrast, the ICESat-2 LAI estimated from the weak beam at the 20 m segment gives the lowest consistency (R = 0.37, Figure 3c).
Figure 4 and Table 4 compare the LAI from ICESat-2 and ALS for different land cover types. WET shows good performance among all types, particularly for strong beams at 200 m (R = 0.61 and RMSE = 1.00, Figure 4l). EF shows the lowest agreement with the ALS LAI for the 20 m weak beam (R = 0.10 and RMSE = 2.55, Table 4). DF and MF are superior to EF and show moderate agreement with the ALS LAI (R > 0.47, Table 4), whereas DF is overestimated in high-LAI regions (Figure 4a,e,i). The performance of ICESat-2 improved with segment size for MF and WET (Table 4). However, ICESat-2 shows a non-significant increase in EF with segment size.
The correlation between ICESat-2 LAI and ALS LAI of each NEON site is presented in Figure 5. In general, the ICESat-2 LAI at most sites is moderately correlated with the ALS LAI, whereas DELA and TEAK show a slight correlation with the ALS LAI (R < 0.24). For most sites, the strong-beam LAI performs better than the weak-beam LAI, especially for the DF sites. However, the weak beam is slightly better than the strong beam at the SCBI, HARV, and UNDE sites. The consistency between the ICESat-2 and ALS LAI also increases with segment size, particularly for weak beams (Figure 5). Conversely, LAI from the EF site is only slightly affected by the beam type and segment size, especially for the OSBS site (R ≥ 0.59, Figure 5).

4.3. Comparison of ICESat-2 LAI and CGLS LAI

Figure 6 compares the ICESat-2 and CGLS LAI for different beam types and segment sizes. The ICESat-2 LAI shows a moderate agreement with the CGLS LAI (R ≥ 0.50). The LAI derived from strong beams (R ≥ 0.66, Figure 6b,e,h) performs better than those of weak beams. The correlation between ICESat-2 LAI and ALS LAI is almost unchanged with segment size. Compared to the CGLS LAI, the ICESat-2 LAI shows a higher upper boundary. The strong beam ICESat-2 obtains the LAI in both high and low regions (Figure 6b,e,h), whereas the LAI from the weak beam focuses solely on the low regions (Figure 6c,f,i).
Figure 7 compares the LAI derived from the all-beam ICESat-2 and CGLS for different land cover types. DF and WET perform better than the other types (R ≥ 0.50), whereas DF is more scattered in the high LAI regions. EF shows the lowest correspondence (R = 0.40, RMSE = 2.59, Figure 7j) and is significantly overestimated in all LAI regions. MF sites are overestimated in the high LAI regions. In contrast, WET shows a slightly systematic underestimation in all regions (Figure 7d,h,l). The performance of ICESat-2 shows no significant change with segment size.

5. Discussion

5.1. Performance of ICESat-2 LAI

ICESat-2 LAI moderately correlates with the ALS LAI and CGLS LAI, and the highest correlation is produced from the strong beam (ALS: R = 0.67 and RMSE = 2.55, Figure 3; CGLS: R = 0.67 and RMSE = 1.94, Figure 6). The performance of ICESat-2 is comparable to other studies, which exhibit similar consistency between the LAI derived from ICESat-2 and the ALS LAI (R2 = 0.63, RMSE = 1.03) [47] and MODIS LAI (R = 0.667, RMSE = 2.433) [42]. The results demonstrate the effectiveness of ICESat-2 LAI and show that the ICESat-2-derived LAI is reliable.
As shown in Figure 3 and Figure 6, ICESat-2 LAI presents overestimation for the different segment sizes and beam types. The ρ v / ρ g is an important parameter in ICESat-2 LAI estimation (Equation (1)). One reason for the overestimation may be attributed to the ρ v / ρ g setting. Figure A1 shows the variation in LAI bias at different ρ v / ρ g values and indicates the appropriate ρ v / ρ g value can reduce bias. The reflectances of the canopy and ground vary with environmental conditions and forest sites [58]. In this study, a constant value (=1/3) was adopted for the LAI algorithm at different forest sites. However, this static ratio may not be suitable for different sites or biome types [28]. Ignoring the spatial variation in the canopy and ground reflectance may cause overestimation in LAI estimation. The linear regression of canopy photon against ground photon can be used to update the ρ v / ρ g in the future [63]. Another reason may be that due to the low number of photons returned per outgoing ATLAS pulse, the canopy cover or LAI derived from the ratio of the number of canopy photons to the total number of terrain and canopy photons overestimate the ALS measures [39,47].
The ICESat-2 shows a higher upper boundary than the CGLS LAI at different forest sites, whereas the latter saturates at around 6.0 (Figure 6 and Figure 7). The saturation of LAI in the passive optical remote sensing data was already reported in earlier studies [8,64]. In contrast, ICESat-2 emits pulses to penetrate the forest canopy and detect the structure of the canopy and understory canopy, effectively alleviating the saturation of passive optical remote sensing data [42]. However, ICESat-2 presents some differences with CGLS. One reason for the difference may be due to the different spatial resolutions. The CGLS LAI pixel size is 300 m × 300 m, whereas the ICESat-2 footprint size is 14 m. Although the ICESat-2 segment size is 14 m × 200 m ideally, there is still a large difference in coverage. Another reason may be due to the misclassification of land cover types [43]. The CGLS LAI algorithm depends on the land cover type to provide vegetation structural information. The CGLS pixels recognized as homogeneous vegetation may consist of different vegetation types in the NLCD and induce uncertainty in the LAI estimation.

5.2. Impact of Segment Size, Beam Type, and Land Cover Type on ICESat-2 LAI

ICESat-2 LAI moderately correlates with the ALS LAI at different segment sizes (R > 0.47, Figure 3 and Table 3). It may be attributed to the sufficient number of photons in the LAI estimation. The results also show that the consistency between the ICESat-2 and ALS increased with segment size (Figure 3 and Table 3). The ICESat-2 LAI derived from the 200 m segment is strongly correlated with the ALS LAI (R = 0.67, Figure 3h), followed by 100 m and 20 m, respectively (Table 3). These results are consistent with the findings in the previous study [43]. This may be attributed to the point density increasing with segment size (Figure 2a). For vegetated surfaces, the number of returned photons ranges from zero to four in a single footprint [35]. As the segment size increased, more photons were binned within the segment. Therefore, more canopy information was captured, leading to an increase in LAI estimation accuracy.
The accuracy of strong-beam LAI was better than that of weak-beam LAI (Table 3 and Table 4). The energy of the strong beam is larger than the weak beam [33], resulting in a high point density in the strong beam that penetrates the canopy more easily (Figure 2a). Consequently, the LAI estimated from the strong beams is more accurate. However, the weak beams also provide useful data for LAI estimation (R = 0.52, Table 3).
The ICESat-2 LAI performance varies with the land cover types (Table 4). For EF, the accuracy of strong-beam LAI was significantly superior to the weak-beam LAI, particularly for the 100 m segment. The superior strong-beam performance is attributed to the fact that the number of canopy photons returned from the strong beam is larger than that of the weak beam for the dense EF canopies (Figure A2) [65]. The WET vegetation is scattered, and the received photon point density is higher than those of the other forest types (Figure A2); in this case, the weak beam performs slightly inferior to the strong beam. For DF and MF, the weak beam shows the potential to estimate LAI because of the lower canopy cover than EF and the enhanced photon returns from weak beams (Figure A2).

5.3. Impact of Canopy Cover on ICESat-2 LAI

The ICESat-2 LAI was moderately consistent with the ALS LAI at most sites, except for DELA and TEAK (Figure 5). The low correlation at DELA and TEAK may be attributed to high canopy cover and complicated terrain conditions, respectively. The DELA site had a high LAI value (>6) during the peak growing season (Figure A3), which indicates a high canopy cover that reduces the penetration ratio of photons. TEAK exhibited the highest slope among all sites (slope = 17.7°, Table 1). Previous studies indicated that steep slopes and dense vegetation can influence photon classification, where near-ground noise is recognized as ground photon and canopy signal photons are misclassified as noise [51,66]. On the other hand, a lower canopy cover may enhance photon penetration. For example, the ICESat-2 LAI was consistent with the ALS at the OSBS and DSNY sites (R > 0.45), which have relatively low LAI values (<3, Figure A3).
The strong-beam LAI performed better than the weak beam for most sites (Figure 5). For the deciduous forest sites (CLBJ, SERC, and UKFS), there was a clear performance difference between the strong and weak beams. This may be because of moderate canopy cover, and strong beams can obtain more canopy and ground photon signals. However, the strong beam shows a slight difference from the weak beam at the evergreen forest sites (Figure 5). This may be attributed to the high canopy cover. Although the ICESat-2 strong-beam energy was higher than the weak beam energy, the return photon signal from the strong beam may not be sufficient for LAI estimation in dense vegetation with high cover.
The weak beam was superior to the strong beam for the HARV and UNDE sites, especially for a large segment size (Figure 5). This was attributed to the decrease in the number of segments as the segment size increased. From 20 m to 100 m to 200 m, the segment numbers of the weak beams were 20 to 14 to 1 and 25 to 10 to 5 for HARV and UNDE, respectively. Uncertainty in LAI estimation may be induced by insufficient segment numbers.

5.4. Limitations and Prospects

The ICESat-2 signal contained ground, canopy, and noise photons. In this study, classification flags from the ATL08 product were used to estimate LAI. Photon classification is a key step in the LAI estimation [42]. The photon classification in ATL08 indicated some misclassification in dense vegetation and steep terrain slopes [46,66]. The misclassification may have led to the LAI errors in this study [47]. In this study, we further manually classified the photons based on the ATL08 data product (Figure A4). The ICESat-2 LAI from the reclassification photon shows a higher accuracy (Table A1). Several algorithms have been proposed to reclassify the misclassified canopies or noise photons in ICESat-2 [66,67]. In the future, utilizing these new algorithms to optimize ICESat-2 photon classification may improve LAI estimation.
The direct comparison between ICESat-2 LAI and field data can accurately evaluate the performance of ICESat-2. However, in this study, the field LAI from GBOV at typical NEON sites was used to describe canopy cover rather than validate ICESat-2 LAI. This is because only three plots with a size of 20 m × 20 m or 40 m × 40 m are measured using the DHP system for each NEON site [68]. The limited number of field data and the small area of field plot size cannot fully evaluate the performance of ICESat-2 LAI, especially for the different segment sizes. In the future, we will conduct field measurements and collect more field data to validate ICESat-2 LAI.

6. Conclusions

The evaluation of LAI estimation with different segment sizes and comparative analysis of ICESat-2 LAI derived from strong and weak beams are significant for the global LAI mapping from the photon counting LiDAR. In this study, we derived and evaluated LAI from ICESat-2 over the NEON sites in North America. The impact of different segment sizes and beam types was explored for the ICESat-2 LAI estimation. The LAI estimated from ICESat-2 for different segment sizes (20, 100, and 200 m) and beam types (strong beam and weak beam) were compared with those from the ALS and CGLS data. The research results showed the following:
  • The strong beam at a 200 m segment size shows the best LAI estimation compared with those from the ALS data (R = 0.67) and is recommended for future LAI estimation.
  • The strong-beam LAI performs better than the weak beam, and the weak beam also presents the potential to estimate LAI when the number of return photons is sufficient.
  • The ICESat-2 LAI partly alleviates the saturation effect and shows satisfactory agreement with the CGLS LAI product (R = 0.67, RMSE = 1.94).
  • The ρ v / ρ g setting and photon classification algorithm are important for ICESat-2 LAI estimation, and specific ρ v / ρ g value and new classification algorithm are needed to improve LAI estimation.
Overall, the ICESat-2 can characterize canopy structural parameters and provides a new dataset for estimating LAI both regionally and globally, which may promote the LAI product generated from the photon-counting LiDAR.

Author Contributions

Conceptualization, Y.W.; methodology, Y.W.; software, Y.W.; validation, Y.W.; formal analysis, Y.W.; investigation, Y.W.; writing—original draft preparation, Y.W.; writing—review and editing, Y.W. and H.F.; visualization, Y.W.; supervision, Y.W.; project administration, Y.W.; funding acquisition, Y.W. and H.F. All authors have read and agreed to the published version of the manuscript.

Funding

This research was supported by the Science and Technology Research Program of Chongqing Municipal Education Commission (KJQN202300546), the Science Foundation of Chongqing Normal University (22XLB029), and a grant from the State Key Laboratory of Resources and Environmental Information System.

Data Availability Statement

The ICESat-2 data are downloaded from https://nsidc.org/data/icesat-2, accessed on 5 April 2023; the CGLS LAI data are downloaded from https://land.copernicus.eu/en/products/vegetation/leaf-area-index-300m-v1.0, accessed on 20 November 2023; the ALS data are available from https://data.neonscience.org/; and the NLCD data are available from https://www.mrlc.gov/data/nlcd-2019-land-cover-conus.

Acknowledgments

We thank NASA for providing the ICESat-2 products, the CGLS for providing the 300 m LAI data, and NEON and GBOV for supplying valuable validation data.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A. The Variation in LAI Bias at Different ρ v / ρ g Values

Figure A1 shows the variation in LAI bias at different ρ v / ρ g values, and the bias decreases with ρ v / ρ g value.
Remotesensing 16 03078 g0a1
Figure A1. The variation in LAI bias at different ρ v / ρ g values. The black dashed line represents the 1/3 value used in this study.
Figure A1. The variation in LAI bias at different ρ v / ρ g values. The black dashed line represents the 1/3 value used in this study.
Remotesensing 16 03078 g0a1

Appendix B. An Example Profile of Icesat-2 Photons along Track Distance for Different Land Cover Types

Figure A2 shows an example profile of ICESat-2 photons along track distance for different land cover types. Strong beams receive more photon signals than weak beams and depict the forest structure and terrain more clearly. The strong beam is moderately different from the weak beam for WET and DF when obtaining the canopy profiles and terrain, while a large difference is presented at EF and MF.
Remotesensing 16 03078 g0a2
Figure A2. The example profile of ICESat-2 photons along track distance (ATD) for DF (a,b), EF (c,d), MF (e,f), and WET (g,h) types. The left and right panels correspond to strong and weak beams, respectively. The classified photons are from ATL08 data products. The top of the canopy, canopy photons, and ground photons are marked as light-green, forest-green dots, and orange dots, respectively. DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.
Figure A2. The example profile of ICESat-2 photons along track distance (ATD) for DF (a,b), EF (c,d), MF (e,f), and WET (g,h) types. The left and right panels correspond to strong and weak beams, respectively. The classified photons are from ATL08 data products. The top of the canopy, canopy photons, and ground photons are marked as light-green, forest-green dots, and orange dots, respectively. DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.
Remotesensing 16 03078 g0a2

Appendix C. The Field LAI from GBOV at Typical NEON Sites

The DELA site shows a high LAI value (>6) during the peak growing season, while the OSBS and DSNY sites have relatively low LAI values (Figure A3).
Remotesensing 16 03078 g0a3
Figure A3. The distribution and seasonal variation in field LAI for overall, overstory, and understory at typical NEON sites. The ratio is understory LAI divided by overall LAI.
Figure A3. The distribution and seasonal variation in field LAI for overall, overstory, and understory at typical NEON sites. The ratio is understory LAI divided by overall LAI.
Remotesensing 16 03078 g0a3

Appendix D. The Comparison of ICESat-2 LAI from Different Photon Classifications

Figure A4 shows the reclassified photons at four example sites, and Table A1 shows the improvement in reclassified photons on ICESat-2 LAI estimation.
Remotesensing 16 03078 g0a4
Figure A4. The ATL08 photon classification (left panel) and composed ATL08 and manual photon classification (right panel) at four example sites. (a,e) SERC site, (b,f) DELA site, (c,g) BART site, and (d,h) DSNY site.
Figure A4. The ATL08 photon classification (left panel) and composed ATL08 and manual photon classification (right panel) at four example sites. (a,e) SERC site, (b,f) DELA site, (c,g) BART site, and (d,h) DSNY site.
Remotesensing 16 03078 g0a4
Table A1. The comparison of ICESat-2 LAI from different photon classifications.
Table A1. The comparison of ICESat-2 LAI from different photon classifications.
SiteATL08 ClassificationATL08 and Manual Classification
RRMSEBiasRRMSEBias
SERC0.751.451.390.791.351.14
DELA0.163.111.720.202.921.37
BART0.551.482.400.611.462.05
DSNY0.611.000.680.690.860.51

References

  1. Chen, J.M.; Black, T.A. Defining leaf area index for non-flat leaves. Plant Cell Environ. 1992, 15, 421–429. [Google Scholar] [CrossRef]
  2. GCOS. The 2022 GCOS ECVs Requirements (GCOS 245); WMO: Geneva, Switzerland, 2022. [Google Scholar]
  3. Treuhaft, R.N.; Asner, G.P.; Law, B.E.; Van Tuyl, S. Forest leaf area density profiles from the quantitative fusion of radar and hyperspectral data. J. Geophys. Res. Atmos. 2002, 107, ACL 7-1–ACL 7-13. [Google Scholar] [CrossRef]
  4. Bonan, G.B.; Patton, E.G.; Finnigan, J.J.; Baldocchi, D.D.; Harman, I.N. Moving beyond the incorrect but useful paradigm: Reevaluating big-leaf and multilayer plant canopies to model biosphere-atmosphere fluxes–a review. Agric. For. Meteorol. 2021, 306, 108435. [Google Scholar] [CrossRef]
  5. Yan, G.; Hu, R.; Luo, J.; Weiss, M.; Jiang, H.; Mu, X.; Xie, D.; Zhang, W. Review of indirect optical measurements of leaf area index: Recent advances, challenges, and perspectives. Agric. For. Meteorol. 2019, 265, 390–411. [Google Scholar] [CrossRef]
  6. Weiss, M.; Baret, F.; Smith, G.J.; Jonckheere, I.; Coppin, P. Review of methods for in situ leaf area index (LAI) determination Part II. Estimation of LAI, errors and sampling. Agric. For. Meteorol. 2004, 121, 37–53. [Google Scholar] [CrossRef]
  7. Fang, H.L.; Zhang, Y.H.; Wei, S.S.; Li, W.J.; Ye, Y.C.; Sun, T.; Liu, W.W. Validation of global moderate resolution leaf area index (LAI) products over croplands in northeastern China. Remote Sens. Environ. 2019, 233, 19. [Google Scholar] [CrossRef]
  8. Fang, H.L.; Baret, F.; Plummer, S.; Schaepman-Strub, G. An Overview of Global Leaf Area Index (LAI): Methods, Products, Validation, and Applications. Rev. Geophys. 2019, 57, 739–799. [Google Scholar] [CrossRef]
  9. Ganguly, S.; Schull, M.A.; Samanta, A.; Shabanov, N.V.; Milesi, C.; Nemani, R.R.; Knyazikhin, Y.; Myneni, R.B. Generating vegetation leaf area index earth system data record from multiple sensors. Part 1: Theory. Remote Sens. Environ. 2008, 112, 4333–4343. [Google Scholar] [CrossRef]
  10. Shao, G.; Stark, S.C.; de Almeida, D.R.A.; Smith, M.N. Towards high throughput assessment of canopy dynamics: The estimation of leaf area structure in Amazonian forests with multitemporal multi-sensor airborne lidar. Remote Sens. Environ. 2019, 221, 1–13. [Google Scholar] [CrossRef]
  11. Canisius, F.; Fernandes, R. Evaluation of the information content of Medium Resolution Imaging Spectrometer (MERIS) data for regional leaf area index assessment. Remote Sens. Environ. 2012, 119, 301–314. [Google Scholar] [CrossRef]
  12. Wang, Y.; Fang, H. Estimation of LAI with the LiDAR Technology: A Review. Remote Sens. 2020, 12, 3457. [Google Scholar] [CrossRef]
  13. Gao, G.; Qi, J.; Lin, S.; Hu, R.; Huang, H. Estimating plant area density of individual trees from discrete airborne laser scanning data using intensity information and path length distribution. Int. J. Appl. Earth Obs. Geoinf. 2023, 118, 103281. [Google Scholar] [CrossRef]
  14. Guo, Q.H.; Su, Y.J.; Hu, T.Y.; Guan, H.C.; Jin, S.C.; Zhang, J.; Zhao, X.X.; Xu, K.X.; Wei, D.J.; Kelly, M.G.; et al. Lidar Boosts 3D Ecological Observations and Modelings: A Review and Perspective. IEEE Geosci. Remote Sens. Mag. 2021, 9, 232–257. [Google Scholar] [CrossRef]
  15. Richardson, J.J.; Moskal, L.M.; Kim, S.H. Modeling approaches to estimate effective leaf area index from aerial discrete-return LIDAR. Agric. For. Meteorol. 2009, 149, 1152–1160. [Google Scholar] [CrossRef]
  16. Zheng, G.; Ma, L.X.; Eitel, J.U.H.; He, W.; Magney, T.S.; Moskal, L.M.; Li, M.S. Retrieving Directional Gap Fraction, Extinction Coefficient, and Effective Leaf Area Index by Incorporating Scan Angle Information From Discrete Aerial Lidar Data. IEEE Trans. Geosci. Remote Sens. 2017, 55, 577–590. [Google Scholar] [CrossRef]
  17. Zhao, K.G.; Popescu, S. Lidar-based mapping of leaf area index and its use for validating GLOBCARBON satellite LAI product in a temperate forest of the southern USA. Remote Sens. Environ. 2009, 113, 1628–1645. [Google Scholar] [CrossRef]
  18. Tian, L.; Qu, Y.; Qi, J. Estimation of Forest LAI Using Discrete Airborne LiDAR: A Review. Remote Sens. 2021, 13, 2408. [Google Scholar] [CrossRef]
  19. Jiang, H.L.; Cheng, S.Y.; Yan, G.J.; Kuusk, A.; Hu, R.H.; Tong, Y.Y.; Mu, X.H.; Xie, D.H.; Zhang, W.M.; Zhou, G.Q.; et al. Clumping Effects in Leaf Area Index Retrieval From Large-Footprint Full-Waveform LiDAR. IEEE Trans. Geosci. Remote Sens. 2021, 60, 1–20. [Google Scholar] [CrossRef]
  20. Cook, B.; Corp, L.; Nelson, R.; Middleton, E.; Morton, D.; McCorkel, J.; Masek, J.; Ranson, K.; Ly, V.; Montesano, P. NASA Goddard’s LiDAR, Hyperspectral and Thermal (G-LiHT) Airborne Imager. Remote Sens. 2013, 5, 4045–4066. [Google Scholar] [CrossRef]
  21. Stereńczak, K.; Laurin, G.V.; Chirici, G.; Coomes, D.A.; Dalponte, M.; Latifi, H.; Puletti, N. Global Airborne Laser Scanning Data Providers Database (GlobALS)—A New Tool for Monitoring Ecosystems and Biodiversity. Remote Sens. 2020, 12, 1877. [Google Scholar] [CrossRef]
  22. Kampe, T.U.; Johnson, B.R.; Kuester, M.; Keller, M. NEON: The first continental-scale ecological observatory with airborne remote sensing of vegetation canopy biochemistry and structure. J. Appl. Remote Sens. 2010, 4, 043510. [Google Scholar] [CrossRef]
  23. Tang, H.; Dubayah, R.; Brolly, M.; Ganguly, S.; Zhang, G. Large-scale retrieval of leaf area index and vertical foliage profile from the spaceborne waveform lidar (GLAS/ICESat). Remote Sens. Environ. 2014, 154, 8–18. [Google Scholar] [CrossRef]
  24. Tang, H.; Ganguly, S.; Zhang, G.; Hofton, M.A.; Nelson, R.F.; Dubayah, R. Characterizing leaf area index (LAI) and vertical foliage profile (VFP) over the United States. Biogeosciences 2016, 13, 239–252. [Google Scholar] [CrossRef]
  25. Yang, X.; Wang, C.; Pan, F.; Nie, S.; Xi, X.; Luo, S. Retrieving leaf area index in discontinuous forest using ICESat/GLAS full-waveform data based on gap fraction model. ISPRS-J. Photogramm. Remote Sens. 2019, 148, 54–62. [Google Scholar] [CrossRef]
  26. Tang, H.; Brolly, M.; Zhao, F.; Strahler, A.H.; Schaaf, C.L.; Ganguly, S.; Zhang, G.; Dubayah, R. Deriving and validating Leaf Area Index (LAI) at multiple spatial scales through lidar remote sensing: A case study in Sierra National Forest, CA. Remote Sens. Environ. 2014, 143, 131–141. [Google Scholar] [CrossRef]
  27. Wang, Y.; Ni, W.; Sun, G.; Chi, H.; Zhang, Z.; Guo, Z. Slope-adaptive waveform metrics of large footprint lidar for estimation of forest aboveground biomass. Remote Sens. Environ. 2019, 224, 386–400. [Google Scholar] [CrossRef]
  28. Wang, Y.; Fang, H.; Zhang, Y.; Li, S.; Pang, Y.; Ma, T.; Li, Y. Retrieval and validation of vertical LAI profile derived from airborne and spaceborne LiDAR data at a deciduous needleleaf forest site. GISci. Remote Sens. 2023, 60, 2214987. [Google Scholar] [CrossRef]
  29. Dubayah, R.; Blair, J.B.; Goetz, S.; Fatoyinbo, L.; Hansen, M.; Healey, S.; Hofton, M.; Hurtt, G.; Kellner, J.; Luthcke, S.; et al. The Global Ecosystem Dynamics Investigation: High-resolution laser ranging of the Earth’s forests and topography. Sci. Remote Sens. 2020, 1, 100002. [Google Scholar] [CrossRef]
  30. Wang, C.; Jia, D.; Lei, S.; Numata, I.; Tian, L. Accuracy Assessment and Impact Factor Analysis of GEDI Leaf Area Index Product in Temperate Forest. Remote Sens. 2023, 15, 1535. [Google Scholar] [CrossRef]
  31. Brown, L.A.; Morris, H.; Meier, C.; Knohl, A.; Lanconelli, C.; Gobron, N.; Dash, J.; Danson, F.M. Stage 1 Validation of Plant Area Index From the Global Ecosystem Dynamics Investigation. IEEE Geosci. Remote Sens. Lett. 2023, 20, 1–5. [Google Scholar] [CrossRef]
  32. Hancock, S.; McGrath, C.; Lowe, C.; Davenport, I.; Woodhouse, I. Requirements for a global lidar system: Spaceborne lidar with wall-to-wall coverage. R. Soc. Open Sci. 2021, 8, 211166. [Google Scholar] [CrossRef]
  33. Markus, T.; Neumann, T.; Martino, A.; Abdalati, W.; Brunt, K.; Csatho, B.; Farrell, S.; Fricker, H.; Gardner, A.; Harding, D.; et al. The Ice, Cloud, and land Elevation Satellite-2 (ICESat-2): Science requirements, concept, and implementation. Remote Sens. Environ. 2017, 190, 260–273. [Google Scholar] [CrossRef]
  34. Popescu, S.C.; Zhou, T.; Nelson, R.; Neuenschwande, A.; Sheridan, R.; Narine, L.; Walsh, K.M. Photon counting LiDAR: An adaptive ground and canopy height retrieval algorithm for ICESat-2 data. Remote Sens. Environ. 2018, 208, 154–170. [Google Scholar] [CrossRef]
  35. Neuenschwander, A.; Pitts, K. The ATL08 land and vegetation product for the ICESat-2 Mission. Remote Sens. Environ. 2019, 221, 247–259. [Google Scholar] [CrossRef]
  36. Neumann, T.A.; Martino, A.J.; Markus, T.; Bae, S.; Bock, M.R.; Brenner, A.C.; Brunt, K.M.; Cavanaugh, J.; Fernandes, S.T.; Hancock, D.W.; et al. The Ice, Cloud, and Land Elevation Satellite–2 mission: A global geolocated photon product derived from the Advanced Topographic Laser Altimeter System. Remote Sens. Environ. 2019, 233, 111325. [Google Scholar] [CrossRef]
  37. Guerra-Hernández, J.; Narine, L.L.; Pascual, A.; Gonzalez-Ferreiro, E.; Botequim, B.; Malambo, L.; Neuenschwander, A.; Popescu, S.C.; Godinho, S. Aboveground biomass mapping by integrating ICESat-2, SENTINEL-1, SENTINEL-2, ALOS2/PALSAR2, and topographic information in Mediterranean forests. GISci. Remote Sens. 2022, 59, 1509–1533. [Google Scholar] [CrossRef]
  38. Narine, L.L.; Popescu, S.C.; Malambo, L. Synergy of ICESat-2 and Landsat for Mapping Forest Aboveground Biomass with Deep Learning. Remote Sens. 2019, 11, 1503. [Google Scholar] [CrossRef]
  39. Queinnec, M.; White, J.C.; Coops, N.C. Comparing airborne and spaceborne photon-counting LiDAR canopy structural estimates across different boreal forest types. Remote Sens. Environ. 2021, 262, 112510. [Google Scholar] [CrossRef]
  40. Narine, L.; Malambo, L.; Popescu, S. Characterizing canopy cover with ICESat-2: A case study of southern forests in Texas and Alabama, USA. Remote Sens. Environ. 2022, 281, 113242. [Google Scholar] [CrossRef]
  41. Feng, T.; Duncanson, L.; Montesano, P.; Hancock, S.; Minor, D.; Guenther, E.; Neuenschwander, A. A systematic evaluation of multi-resolution ICESat-2 ATL08 terrain and canopy heights in boreal forests. Remote Sens. Environ. 2023, 291, 113570. [Google Scholar] [CrossRef]
  42. Zhang, J.; Tian, J.; Li, X.; Wang, L.; Chen, B.; Gong, H.; Ni, R.; Zhou, B.; Yang, C. Leaf area index retrieval with ICESat-2 photon counting LiDAR. Int. J. Appl. Earth Obs. Geoinf. 2021, 103, 102488. [Google Scholar] [CrossRef]
  43. Guo, D.; Hu, R.; Song, X.; Li, X.; Lin, H.; Zhang, Y.; Gao, L.; Zhu, X. Exploring Photon-Counting Laser Altimeter ICESat-2 in Retrieving LAI and Correcting Clumping Effect. IEEE Trans. Geosci. Remote Sens. 2023, 61, 1–9. [Google Scholar] [CrossRef]
  44. Neuenschwander, A.L.; Magruder, L.A. Canopy and Terrain Height Retrievals with ICESat-2: A First Look. Remote Sens. 2019, 11, 1721. [Google Scholar] [CrossRef]
  45. Rehman, K.; Fareed, N.; Chu, H.-J. NASA ICESat-2: Space-Borne LiDAR for Geological Education and Field Mapping of Aeolian Sand Dune Environments. Remote Sens. 2023, 15, 2882. [Google Scholar] [CrossRef]
  46. Moudrý, V.; Gdulová, K.; Gábor, L.; Šárovcová, E.; Barták, V.; Leroy, F.; Špatenková, O.; Rocchini, D.; Prošek, J. Effects of environmental conditions on ICESat-2 terrain and canopy heights retrievals in Central European mountains. Remote Sens. Environ. 2022, 279, 113112. [Google Scholar] [CrossRef]
  47. Wang, Y.; Wang, C.; Yang, X.; Nie, S. Verification of Leaf Area Index Retrieved by ICESAT-2 Photon-Counting Lidar with Airborne Lidar. In Proceedings of the IGARSS 2022—2022 IEEE International Geoscience and Remote Sensing Symposium, Kuala Lumpur, Malaysia, 17–22 July 2022; pp. 7305–7308. [Google Scholar]
  48. Brown, L.A.; Meier, C.; Morris, H.; Pastor-Guzman, J.; Bai, G.; Lerebourg, C.; Gobron, N.; Lanconelli, C.; Clerici, M.; Dash, J. Evaluation of global leaf area index and fraction of absorbed photosynthetically active radiation products over North America using Copernicus Ground Based Observations for Validation data. Remote Sens. Environ. 2020, 247, 111935. [Google Scholar] [CrossRef]
  49. Li, S.; Fang, H.; Zhang, Y.; Wang, Y. Comprehensive evaluation of global CI, FVC, and LAI products and their relationships using high-resolution reference data. Sci. Remote Sens. 2022, 6, 100066. [Google Scholar] [CrossRef]
  50. National Ecological Observatory Network (NEON). Discrete Return LiDAR Point Cloud (DP1.30003.001); National Ecological Observatory Network (NEON): Boulder, CO, USA, 2023. [Google Scholar] [CrossRef]
  51. Neuenschwander, A.; Guenther, E.; White, J.C.; Duncanson, L.; Montesano, P. Validation of ICESat-2 terrain and canopy heights in boreal forests. Remote Sens. Environ. 2020, 251, 112110. [Google Scholar] [CrossRef]
  52. Narine, L.L.; Popescu, S.; Neuenschwander, A.; Zhou, T.; Srinivasan, S.; Harbeck, K. Estimating aboveground biomass and forest canopy cover with simulated ICESat-2 data. Remote Sens. Environ. 2019, 224, 1–11. [Google Scholar] [CrossRef]
  53. Fuster, B.; Sánchez-Zapero, J.; Camacho, F.; García-Santos, V.; Verger, A.; Lacaze, R.; Weiss, M.; Baret, F.; Smets, B. Quality Assessment of PROBA-V LAI, fAPAR and fCOVER Collection 300 m Products of Copernicus Global Land Service. Remote Sens. 2020, 12, 1017. [Google Scholar] [CrossRef]
  54. Baret, F.; Weiss, M.; Verger, A.; Smets, B. Atbd for Lai, Fapar and Fcover from Proba-V Products at 300 Mresolution (Geov3). Imagines_rp2.1_atbd-lai 300 m. Issue 1.73. Available online: https://land.copernicus.eu/global/sites/cgls.vito.be/files/products/ImagineS_RP2.1_ATBD-LAI300m_I1.73.pdf (accessed on 22 October 2023).
  55. Verger, A.; Descals, A. Algorithm theorethical basis document, Leaf Area Index (LAI), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Fraction of Green Vegetation Cover (FCover) Collection 300m Version 1.1. Available online: https://land.copernicus.eu/global/sites/cgls.vito.be/files/products/CGLOPS1_ATBD_LAI300m-V1.1_I1.10.pdf (accessed on 22 October 2023).
  56. Wickham, J.; Stehman, S.V.; Sorenson, D.G.; Gass, L.; Dewitz, J.A. Thematic accuracy assessment of the NLCD 2016 land cover for the conterminous United States. Remote Sens. Environ. 2021, 257, 112357. [Google Scholar] [CrossRef]
  57. Homer, C.; Dewitz, J.; Jin, S.; Xian, G.; Costello, C.; Danielson, P.; Gass, L.; Funk, M.; Wickham, J.; Stehman, S.; et al. Conterminous United States land cover change patterns 2001–2016 from the 2016 National Land Cover Database. ISPRS-J. Photogramm. Remote Sens. 2020, 162, 184–199. [Google Scholar] [CrossRef]
  58. Tang, H.; Dubayah, R.; Swatantran, A.; Hofton, M.; Sheldon, S.; Clark, D.B.; Blair, B. Retrieval of vertical LAI profiles over tropical rain forests using waveform lidar at La Selva, Costa Rica. Remote Sens. Environ. 2012, 124, 242–250. [Google Scholar] [CrossRef]
  59. Neuenschwander, A.L.; Magruder, L.A. The Potential Impact of Vertical Sampling Uncertainty on ICESat-2/ATLAS Terrain and Canopy Height Retrievals for Multiple Ecosystems. Remote Sens. 2016, 8, 1039. [Google Scholar] [CrossRef]
  60. García, M.; Riaño, D.; Chuvieco, E.; Danson, F.M. Estimating biomass carbon stocks for a Mediterranean forest in central Spain using LiDAR height and intensity data. Remote Sens. Environ. 2010, 114, 816–830. [Google Scholar] [CrossRef]
  61. Luo, S.; Wang, C.; Xi, X.; Nie, S.; Fan, X.; Chen, H.; Ma, D.; Liu, J.; Zou, J.; Lin, Y.; et al. Estimating forest aboveground biomass using small-footprint full-waveform airborne LiDAR data. Int. J. Appl. Earth Obs. Geoinf. 2019, 83, 101922. [Google Scholar] [CrossRef]
  62. Ni-Meister, W.; Jupp, D.L.B.; Dubayah, R. Modeling lidar waveforms in heterogeneous and discrete canopies. IEEE Trans. Geosci. Remote Sens. 2001, 39, 1943–1958. [Google Scholar] [CrossRef]
  63. Purslow, M.; Hancock, S.; Neuenschwander, A.; Armston, J.; Duncanson, L. Can ICESat-2 estimate stand-level plant structural traits? Validation of an ICESat-2 simulator. Sci. Remote Sens. 2023, 7, 100086. [Google Scholar] [CrossRef]
  64. Bacour, C.; Baret, F.; Béal, D.; Weiss, M.; Pavageau, K. Neural network estimation of LAI, fAPAR, fCover and LAI×Cab, from top of canopy MERIS reflectance data: Principles and validation. Remote Sens. Environ. 2006, 105, 313–325. [Google Scholar] [CrossRef]
  65. Urbazaev, M.; Hess, L.L.; Hancock, S.; Sato, L.Y.; Ometto, J.P.; Thiel, C.; Dubois, C.; Heckel, K.; Urban, M.; Adam, M.; et al. Assessment of terrain elevation estimates from ICESat-2 and GEDI spaceborne LiDAR missions across different land cover and forest types. Sci. Remote Sens. 2022, 6, 100067. [Google Scholar] [CrossRef]
  66. Zhao, R.; Ni, W.; Zhang, Z.; Dai, H.; Yang, C.; Li, Z.; Liang, Y.; Liu, Q.; Pang, Y.; Li, Z.; et al. Optimizing ground photons for canopy height extraction from ICESat-2 data in mountainous dense forests. Remote Sens. Environ. 2023, 299, 113851. [Google Scholar] [CrossRef]
  67. Nie, S.; Wang, C.; Xi, X.; Luo, S.; Li, G.; Tian, J.; Wang, H. Estimating the vegetation canopy height using micro-pulse photon-counting LiDAR data. Opt. Express 2018, 26, A520–A540. [Google Scholar] [CrossRef] [PubMed]
  68. Meier, C.; Everhart, J.; Jones, K. TOS Protocol and Procedure: Measurement of Leaf Area Index. 2018. Available online: https://data.neonscience.org/api/v0/documents/NEON.DOC.014039vK (accessed on 7 June 2023).
Remotesensing 16 03078 g001
Figure 1. The spatial distribution of (a) ICESat-2 data over the 12 National Ecological Observatory Network (NEON) sites and (b) an example of ICESat-2 data at the BART site. The background is a land cover map from the National Land Cover Database (NLCD).
Figure 1. The spatial distribution of (a) ICESat-2 data over the 12 National Ecological Observatory Network (NEON) sites and (b) an example of ICESat-2 data at the BART site. The background is a land cover map from the National Land Cover Database (NLCD).
Remotesensing 16 03078 g001
Remotesensing 16 03078 g002
Figure 2. Statistics of point density along track distance for (a) ICESat-2 strong beam, (b) ICESat-2 weak beam, and (c) ALS data for different segment sizes (20 m, 100 m, and 200 m). DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.
Figure 2. Statistics of point density along track distance for (a) ICESat-2 strong beam, (b) ICESat-2 weak beam, and (c) ALS data for different segment sizes (20 m, 100 m, and 200 m). DF, EF, MF, and WET refer to deciduous forest, evergreen forest, mixed forest, and woody wetlands, respectively.
Remotesensing 16 03078 g002
Remotesensing 16 03078 g003
Figure 3. Comparison between the ICESat-2 LAI and the ALS LAI for different segment sizes and beam types. The upper (a–c), middle (d–f), and lower (g–i) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Figure 3. Comparison between the ICESat-2 LAI and the ALS LAI for different segment sizes and beam types. The upper (a–c), middle (d–f), and lower (g–i) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Remotesensing 16 03078 g003
Remotesensing 16 03078 g004
Figure 4. Comparison between the LAI values derived from strong-beam ICESat-2 and ALS for DF, EF, MF, and WET. The upper (a–d), middle (e–h), and lower (i–l) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Figure 4. Comparison between the LAI values derived from strong-beam ICESat-2 and ALS for DF, EF, MF, and WET. The upper (a–d), middle (e–h), and lower (i–l) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Remotesensing 16 03078 g004
Remotesensing 16 03078 g005
Figure 5. The correlation between ICESat-2 LAI and ALS LAI of each NEON site for different segment sizes and beam types. See Table 1 and Figure 2 for the site names and land cover types, respectively.
Figure 5. The correlation between ICESat-2 LAI and ALS LAI of each NEON site for different segment sizes and beam types. See Table 1 and Figure 2 for the site names and land cover types, respectively.
Remotesensing 16 03078 g005
Remotesensing 16 03078 g006
Figure 6. Comparison between the ICESat-2 LAI from all beams, strong beams, and weak beams and the CGLS LAI. The upper (a–c), middle (d–f), and lower (g–i) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Figure 6. Comparison between the ICESat-2 LAI from all beams, strong beams, and weak beams and the CGLS LAI. The upper (a–c), middle (d–f), and lower (g–i) panels correspond to the all, strong, and weak beams at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Remotesensing 16 03078 g006
Remotesensing 16 03078 g007
Figure 7. Comparison between the LAI derived from all-beam ICESat-2 and CGLS for DF, EF, MF, and WET. The upper (a–d), middle (e–h), and lower (i–l) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Figure 7. Comparison between the LAI derived from all-beam ICESat-2 and CGLS for DF, EF, MF, and WET. The upper (a–d), middle (e–h), and lower (i–l) panels correspond to the different land cover types at segment sizes of 20 m, 100 m, and 200 m, respectively. The solid line and dashed line indicate the fitting line and 1:1 line, respectively.
Remotesensing 16 03078 g007
Table 1. Information on ALS data, ICESat-2 data, and CGLS data over 12 NEON sites. The land cover class comes from 2019 NLCD data.
Table 1. Information on ALS data, ICESat-2 data, and CGLS data over 12 NEON sites. The land cover class comes from 2019 NLCD data.
Site NameCodeDominant NLCDElevation (m)Slope (°)ALS Data
(2019)		ICESat-2 Data
(2019)		CGLS Data
(2019)
Abby RoadABBYEvergreen forest36515.014 July29 July31 July
Bartlett Experimental ForestBARTMixed forest27415.625 August3 September10 September
Lyndon B. Johnson National GrasslandCLBJDeciduous forest2725.220 April16 May20 May
Dead LakeDELAEvergreen forest254.329 April7 May10 May
Disney Wilderness PreserveDSNYWoody wetlands204.315 April7 March10 March
Harvard ForestHARVMixed forest3487.511 August9 August10 August
Ordway-Swisher Biological StationOSBSEvergreen needleleaf464.515 April5 May10 May
Smithsonian Conservation Biology InstituteSCBIDeciduous broadleaf35213.224 May8 May10 May
Smithsonian Environmental Research CenterSERCDeciduous broadleaf336.515 May2 May10 May
TeakettleTEAKEvergreen forest214917.714 June7 June10 June
University of Kansas Field StationUKFSDeciduous forest3225.226 May10 May10 May
UNDERCUNDEMixed forest5215.98 June2 June10 June
Table 2. Information on the NEON airborne LiDAR system and product.
Table 2. Information on the NEON airborne LiDAR system and product.
Acquisition DateApril to August 2019
SensorOptech Incorporated Airborne Laser Terrain Mapper (ALTM) Gemini
Beam wavelength1064 nm
Footprint diameter0.25 m (at 1000 m flying height), 0.8 m in wide beam divergence mode
Sampling density1–4 points per square meter
Horizontal accuracy<5–15 cm; 1 σ
Elevation accuracy<5–35 cm; 1 σ
Derived productsDTM and CHM
Product resolutionUniform grid (1 m × 1 m)
Vertical datumGEOID12A
Table 3. Comparison between the ICESat-2 LAI and the ALS LAI for different segment sizes and beam types.
Table 3. Comparison between the ICESat-2 LAI and the ALS LAI for different segment sizes and beam types.
Segment Sizes (m)Beam TypeRRMSErRMSEBias
20all0.582.3554.09%2.04
strong0.632.3351.08%1.99
weak0.372.3761.74%2.15
100all0.612.6658.44%2.47
strong0.662.6555.42%2.41
weak0.482.6864.47%2.58
200all0.642.5059.92%2.30
strong0.672.5558.00%2.25
weak0.522.3963.79%2.39
Table 4. Statistics from validation of the ICESat-2 LAI using ALS LAI for different land cover types.
Table 4. Statistics from validation of the ICESat-2 LAI using ALS LAI for different land cover types.
Land Cover TypesStatistics20 m100 m200 m
All BeamsStrong BeamsWeak BeamsAll BeamsStrong BeamsWeak BeamsAll BeamsStrong BeamsWeak Beams
Deciduous forest (DF)R0.600.660.480.590.600.590.650.630.73
RMSE2.182.162.142.612.852.232.402.771.65
rRMSE (%)53.2051.0155.4762.3260.7664.6463.8964.2056.40
Bias2.282.372.132.642.682.592.392.512.21
Evergreen forest (EF)R0.270.370.100.300.420.130.270.340.22
RMSE2.572.542.552.962.862.972.812.712.80
rRMSE (%)55.1452.9055.5456.3752.4357.6956.8950.6459.74
Bias2.982.813.303.603.174.203.432.894.23
Mixed forest (MF)R0.520.510.600.620.590.740.660.660.68
RMSE1.851.831.691.951.931.961.771.722.31
rRMSE (%)26.1725.4729.8825.4224.7428.9123.7723.0132.73
Bias1.601.641.112.572.642.132.62.612.51
Woody wetlands (WET)R0.500.530.460.540.600.470.560.610.50
RMSE1.721.691.791.411.291.571.121.001.30
rRMSE (%)73.7673.3474.8977.0670.2886.2268.6162.1777.75
Bias1.031.011.090.800.820.770.670.680.66
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wang, Y.; Fang, H. Derivation and Evaluation of LAI from the ICESat-2 Data over the NEON Sites: The Impact of Segment Size and Beam Type. Remote Sens. 2024, 16, 3078. https://doi.org/10.3390/rs16163078

AMA Style

Wang Y, Fang H. Derivation and Evaluation of LAI from the ICESat-2 Data over the NEON Sites: The Impact of Segment Size and Beam Type. Remote Sensing. 2024; 16(16):3078. https://doi.org/10.3390/rs16163078

Chicago/Turabian Style

Wang, Yao, and Hongliang Fang. 2024. "Derivation and Evaluation of LAI from the ICESat-2 Data over the NEON Sites: The Impact of Segment Size and Beam Type" Remote Sensing 16, no. 16: 3078. https://doi.org/10.3390/rs16163078

APA Style

Wang, Y., & Fang, H. (2024). Derivation and Evaluation of LAI from the ICESat-2 Data over the NEON Sites: The Impact of Segment Size and Beam Type. Remote Sensing, 16(16), 3078. https://doi.org/10.3390/rs16163078

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop