Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Exploring Functional Gene XsPDAT1’s Involvement in Xanthoceras sorbifolium Oil Synthesis and Its Acclimation to Cold Stress
Previous Article in Journal
Parameter Estimations on Measurement Accuracy for Thermal Conductivity of Wood Using the Transient Plane Source Method
Previous Article in Special Issue
Response of Individual-Tree Aboveground Biomass to Spatial Effects in Pinus kesiya var. langbianensis Forests by Stand Origin and Tree Size
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
Forests
Volume 15
Issue 10
10.3390/f15101821
forests-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

Unveiling Anomalies in Terrain Elevation Products from Spaceborne Full-Waveform LiDAR over Forested Areas

by
Hailan Jiang
1,2,
Yi Li
3,
Guangjian Yan
3,*,
Weihua Li
3,
Linyuan Li
4,
Feng Yang
5,
Anxin Ding
1,
Donghui Xie
3,
Xihan Mu
3,
Jing Li
2,
Kaijian Xu
1,
Ping Zhao
1,
Jun Geng
6 and
Felix Morsdorf
7
1
School of Resources and Environmental Engineering, Hefei University of Technology, Hefei 230009, China
2
State Key Laboratory of Remote Sensing Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100101, China
3
State Key Laboratory of Remote Sensing Science, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China
4
State Key Laboratory to Efficient Production of Forest Resources, Beijing Forestry University, Beijing 100083, China
5
College of Soil and Water Conservation Science and Engineering, Northwest A&F University, Xi’an 712100, China
6
School of Civil Engineering, Hefei University of Technology, Hefei 230009, China
7
Remote Sensing Laboratories, Department of Geography, University of Zurich, 8057 Zurich, Switzerland
*
Author to whom correspondence should be addressed.
Forests 2024, 15(10), 1821; https://doi.org/10.3390/f15101821
Submission received: 7 September 2024 / Revised: 11 October 2024 / Accepted: 14 October 2024 / Published: 17 October 2024
(This article belongs to the Special Issue Methodology and Theory of Forest Parameters Estimation Using Multi-Source Remote Sensing)
Browse Figures
Figure 1
<p>Study area and the geolocation of GLAS and GEDI data.</p> "> Figure 2
<p>Flowchart of this study.</p> "> Figure 3
<p>Spatial distribution of terrain elevation anomalies in the GLAS and GEDI datasets.</p> "> Figure 4
<p>Scatter plot of terrain elevation estimates obtained from GLAS (<b>a</b>) and GEDI (<b>b</b>) vs. the terrain elevation derived from airborne laser scanning (ALS) as a reference. A0 denotes the default algorithm of the GEDI L2A product.</p> "> Figure 5
<p>Details of terrain elevation outliers from GLAS: scatter plot of terrain elevation from data acquired during nighttime (<b>a</b>) and daytime (<b>b</b>) before removing outliers, scatter plot (<b>c</b>), transmitted waveforms (<b>d</b>), the histogram of the data acquisition time (<b>e</b>), and the histogram of Signal-to-Noise Ratio (SNR) (<b>f</b>) of source laser shot of the outliers.</p> "> Figure 6
<p>Details of terrain elevation outliers from GEDI: scatter plot of terrain elevation from data acquired during nighttime (<b>a</b>) and daytime (<b>b</b>) before removing outliers, scatter plot (<b>c</b>), transmitted waveforms (<b>d</b>), the histogram (<b>e</b>) of the beam type (<b>e1</b>) and data acquisition time (<b>e2</b>), and the histogram of sensitivity (<b>f</b>) of source laser shot of the outliers.</p> "> Figure 7
<p>Examples with small (upper panel) and large (lower panel) terrain elevation error: the three-dimensional scene (<b>left</b>), the transmitted waveform (<b>middle</b>), and the return waveform (<b>right</b>) of GLAS and GEDI with the terrain elevation from product and airborne laser scanning (ALS) data illustrated. In the lower panel (<b>right</b>), the ALS terrain elevation is not indicated since the GLAS or GEDI terrain elevation exceeds ALS by more than 330 m (see outlier-27 and outlier-12 indicated in green circles in <a href="#forests-15-01821-f005" class="html-fig">Figure 5</a> and <a href="#forests-15-01821-f006" class="html-fig">Figure 6</a>c). A&lt;<span class="html-italic">n</span>&gt; (<span class="html-italic">n</span>: 1–6) denotes the terrain elevation from six different algorithm groups of GEDI.</p> "> Figure 8
<p>Scatter plot of canopy height estimates for laser shots of terrain elevation anomalies obtained from GEDI L2A product versus the canopy height derived from airborne laser scanning (ALS) as a reference (the legend of the point density applies to all the figures). A0 denotes the default algorithm (<b>a</b>), and A&lt;<span class="html-italic">n</span>&gt; (<span class="html-italic">n</span>: 1–6) denotes the other six algorithm groups (<b>b</b>–<b>g</b>).</p> "> Figure A1
<p>Probability density of “sensitivity” of “power” and “coverage” beams estimated by different algorithms (<b>a</b>–<b>f</b>) of GEDI using the data with “sensitivity &gt; 0.90” in all footprints. A0 denotes the default algorithm setting (<b>a</b>), and A&lt;n&gt; (n: 1–6) denotes the other six algorithm groups (<b>b</b>–<b>g</b>).</p> "> Figure A2
<p>Original GLAS (<b>upper panel</b>) and GEDI (<b>lower panel</b>) waveform examples of terrain elevation anomalies with terrain elevation provided by GLAS and GEDI product indicated. A0 denotes the default algorithm, and A&lt;<span class="html-italic">n</span>&gt; (<span class="html-italic">n</span>: 1–6) denotes the other six algorithm groups of GEDI.</p> ">
Versions Notes

Abstract

:
Anomalies displaying significant deviations between terrain elevation products acquired from spaceborne full-waveform LiDAR and reference elevations are frequently observed in assessment studies. While the predominant focus is on “normal” data, recognizing anomalies within datasets obtained from the Geoscience Laser Altimeter System (GLAS) and the Global Ecosystem Dynamics Investigation (GEDI) is essential for a comprehensive understanding of widely used spaceborne full-waveform data, which not only facilitates optimal data utilization but also enhances the exploration of potential applications. Nevertheless, our comprehension of anomalies remains limited as they have received scant specific attention. Diverging from prevalent practices of directly eliminating outliers, we conducted a targeted exploration of anomalies in forested areas using both transmitted and return waveforms from the GLAS and the GEDI in conjunction with airborne LiDAR point cloud data. We unveiled that elevation anomalies stem not from the transmitted pulses or product algorithms, but rather from scattering sources. We further observed similarities between the GLAS and the GEDI despite their considerable disparities in sensor parameters, with the waveforms characterized by a low signal-to-noise ratio and a near exponential decay in return energy; specifically, return signals of anomalies originated from clouds rather than the land surface. This discovery underscores the potential of deriving cloud-top height from spaceborne full-waveform LiDAR missions, particularly the GEDI, suggesting promising prospects for applying GEDI data in atmospheric science—an area that has received scant attention thus far. To mitigate the impact of abnormal return waveforms on diverse land surface studies, we strongly recommend incorporating spaceborne LiDAR-offered terrain elevation in data filtering by establishing an elevation-difference threshold against a reference elevation. This is especially vital for studies concerning forest parameters due to potential cloud interference, yet a consensus has not been reached within the community.

1. Introduction

Spaceborne full-waveform Light Detection and Ranging (LiDAR) represents an advanced active remote sensing technology capable of acquiring earth surface elevation with a high vertical resolving ability [1]. This technology, renowned for its exceptional ability to penetrate small gaps within tree crowns, facilitates the derivation of forest structure and the detection of ground beneath forest canopies. Presently, spaceborne full-waveform LiDAR sensors offering freely available products are the Geoscience Laser Altimeter System (GLAS) [2] and the Global Ecosystem Dynamics Investigation (GEDI) [3]. With a fifteen-year interval in sensor development, notable disparities exist in critical parameters between the GLAS and the GEDI, including the footprint diameter (~65 m vs. ~25 m), pulse duration (~4 ns vs. ~15 ns), and transmitted power (~75 mJ vs. ~10 mJ). These missions are designed to provide vertical information products about land surfaces [4,5,6], encompassing terrain elevation.
Terrain elevation anomalies denote samples with significant deviations between spaceborne LiDAR-derived products and reference elevations. In the domain of spaceborne full-waveform LiDAR product applications or assessments, data filtering is commonly employed to ensure the utilization of high-quality data for land surface studies [7,8,9,10,11,12,13]. However, anomalies persist even after data filtering in terrain elevation studies [14,15,16], indicating that knowledge about anomalies is necessary for optimal data utilization.
Enhancing our understanding of terrain elevation anomalies in spaceborne full-waveform LiDAR data is imperative. Despite over two decades having elapsed since the advent of the GLAS, and the current operational status of the widely used GEDI, our understanding of anomalies in GLAS and GEDI datasets remains limited. For instance, there exists a divergence in perspectives regarding the origins of these anomalies, with some researchers attributing them to rainfall [14], while others posit timing errors in the pipeline as the primary cause [15]; however, detailed explanations, especially from the waveform perspective, have not been provided. The limited understanding of anomalies thus far can be attributed to the fact that studies rarely simultaneously focus on terrain elevation and waveform analysis. Typically, product assessment studies address outliers by directly removing them after comparing terrain elevation products with values from airborne laser scanning (ALS) point cloud data or high-resolution Digital Elevation Models (DEMs) [17,18,19], yet these studies often overlook waveform explorations. Conversely, research focusing on deriving parameters like leaf area index [20,21] and canopy height [11,22] by processing waveforms might disregard waveforms with abnormal elevations since they neither depend on nor are concerned with the absolute elevation of each return signal. Although the emphasis is primarily on “normal” data, comprehending abnormal data is crucial for improving spaceborne LiDAR applications. Nevertheless, research specifically targeting anomalies remains scarce. Therefore, a dedicated study on terrain elevation anomalies is essential and urgently needed for a comprehensive understanding of spaceborne full-waveform data, which not only enables optimizing data utilization in land surface studies but also enhances exploring applications beyond their current scope.
Our objective is to investigate the waveform characteristics and discuss the origins of terrain elevation anomalies in GLAS and GEDI products to improve our understanding of spaceborne LiDAR waveforms. Given the prevalent application of the GLAS and the GEDI in estimating forest parameters, our focus is specifically on data from forested areas. Terrain elevation data were extracted from elevation parameters available in GLAS and the GEDI products, with anomalies identified through comparison with high-resolution ALS measurements. Anomalies were investigated in detail using both the ALS-generated three-dimensional point cloud and the DEM, as well as the transmitted waveform (i.e., the digitized outgoing pulse of the laser) and return waveform (i.e., the pulse received at the instrument after reflection from the Earth’s surface) of spaceborne LiDAR. This study potentially provides initial specialized insights into anomalies, crucial for optimizing the applications of spaceborne full-waveform LiDAR data.

2. Materials and Methods

2.1. Study Area

The study area is Aargau, the tenth-largest canton in northern Switzerland, covering an area of 1404.4 km2. This region was selected for its substantial forest cover, approximately 500 km2, which accounts for over one-third of the canton’s total area, as well as the availability of high-resolution airborne LiDAR data. Approximately 63% of the forests consist of pure or mixed deciduous species, with beech (32%) and spruce (26%) being the predominant tree species. The geolocations of GLAS and GEDI footprints are depicted in Figure 1. The forest boundary data were obtained from the forestry department of canton Aargau. The terrain elevation relative to the reference ellipsoid derived from airborne laser scanning (ALS) data ranges from 262 m to 878 m, with a mean value of 504 m and a standard deviation of 101 m. The terrain slope in forested regions varies from 1° to 46°, with a mean of 13° and a standard deviation of 8°. The mean canopy height within the area covered by the spaceborne laser footprints, calculated from the ALS point cloud data, is 30 m with a standard deviation of 7 m, and the mean fractional vegetation cover is 0.54 with a standard deviation of 0.17.

2.2. Airborne Laser Scanning Data

ALS data were acquired between 23 March 2019, and 21 April 2019, using the RIEGL LMS-VQ780i long-range laser scanner with a central wavelength of 1550 nm and a laser beam divergence of 0.25 mrad. Aerial surveys were conducted by Milan Geoservice GmbH (Kamenz, Germany). The average operating flight altitude was 1250 m above ground level, with a maximum scanning zenith angle of 30°. The classified three-dimensional point cloud data were supplied as LAS files, utilizing a Swiss Cartesian coordinate system of CH-1903+ (LV95) (EPSG code: 2056), and the positional accuracy is < 0.07 m in vertical and < 0.15 m in horizontal direction. Pulse density (i.e., number of laser pulses per square meter) ranged from 10 pulses to 50 pulses per square meter, with a mean value of 23 and a standard deviation of 3. We generated a DEM with a spatial resolution of 0.5 m using the extracted ALS point cloud within each GLAS and GEDI footprint. For terrain elevation comparison, we adopted the mean terrain elevation derived from our generated DEM as a reference.

2.3. Spaceborne Full-Waveform LiDAR Data Processing

We extracted terrain elevation data from elevation parameters available in two different datasets: the “GLAS/ICESat L2 Global Land Surface Altimetry Data” (GLA14 product, NSIDC, Boulder, CO, USA), release 34 [6], and the “GEDI Level 2A Footprint Elevation and Height Metrics” (L2A product, NASA, Greenbelt, MD, USA), version 2 [5]. The products are provided in HDF file format, and we used Python code to read the parameters from the HDF file and extract the necessary parameters into a shapefile. The transmitted and reflected waveforms, provided in the GLA01 and L1B products for the GLAS and the GEDI, respectively, were extracted to examine the transmitted and return signals of laser shots of terrain elevation anomalies. Records from different products were linked using the unique ID of each laser shot. Parameters used for data filtering to ensure data quality, geolocation information, transmitted and return waveform, and solar elevation angle extraction are provided in Table 1. To ensure consistency with ALS data, we conducted a coordinate system transformation using the transformer function in the “pyproj” package in Python.

2.3.1. GLAS Data Processing

We extracted terrain elevation and other key parameters from GLA14 products, downloaded from the National Snow and Ice Data Center (NSIDC) via https://nsidc.org/data/icesat/data.html (accessed on 1 March 2020), which cover data acquisition from 20 February 2003, to 11 October 2009. It is important to note that GLA14 does not directly provide terrain elevation at the location of the last Gaussian mode but offers relevant parameters for terrain elevation computation. “d_elev” represents the elevation at the centroid of the return waveform, and other parameters, including “d_GmC”, “d_gpCntRngOff”, “d_ldRngOff”, “d_satElevCorr”, and “d_deltaEllip” from GLA14, were utilized alongside “d_elev” to acquire the terrain elevation. Specifically, GLAS elevations are presented as ellipsoidal heights relative to the Topex ellipsoid; “d_deltaEllip” (with a value of 0.707 m), which represents the difference between surface elevation from the Topex ellipsoid and the WGS84 ellipsoid, was used to convert the ellipsoidal height from Topex to WGS84. Furthermore, the elevation was converted from the elevation at the location of the centroid to the elevation at the location of the last Gaussian mode using “d_GmC”, “d_gpCntRngOff”, and “d_ldRngOff”; d _ s a t E l e v C o r r serves as a correction to the elevation in cases of saturated waveforms. Following the method described by [24] for calculating the surface elevation, we computed the elevation at the location of the last Gaussian mode as the terrain elevation:
T e r r a i n   e l e v a t i o n = d _ e l e v − ( d _ g p C n t R n g O f f − d _ l d R n g O f f ) − d _ G m C + d _ s a t E l e v C o r r − d _ d e l t a E l l i p
Readers are encouraged to consult [24] and the GLA14 dictionary for more details about the meanings of these parameters. To ensure the inclusion of high-quality data, we filtered data samples after extracting the key parameters listed in Table 1 to a shapefile using the “elev_use_flg” with a value of 0, which indicates that elevations recorded should be utilized. Furthermore, “elv_cloud_flg” from the GLA05 product [25] was used to identify the presence of clouds, denoted by a value of 1.

2.3.2. GEDI Data Processing

We extracted terrain elevation from the GEDI L2A product [5], downloaded from NASA Earthdata (https://www.earthdata.nasa.gov/, accessed on 1 March 2022), which covers data acquisition from 30 April 2019, to 19 June 2021. The parameter “elev_lowestmode” represents the elevation of the lowest mode’s center relative to the WGS84 ellipsoid, estimated using the default algorithm of the GEDI (indicated as A0 in this study). elev_lowestmode_a<n> (n: 1–6) denotes terrain elevation estimates derived from six distinct algorithms (hereafter referred to as A1, A2, …, A6). These algorithms define thresholds for signal start and end, as well as smoothing width for noise and signal height metrics above ground level based on different standards [26]. Notably, default terrain elevation estimates differ from these six algorithms, depending on plant functional type [27].
We implemented specific filters to the GEDI data samples to ensure data quality following the conversion of the HDF file to a shapefile, from which the necessary parameters listed in Table 1 were extracted. The filters include ① degrade_flag = 0, indicating non-degraded GEDI samples; ② quality_flag = 1, signifying the most useful data; and ③ sensitivity > 0.90, an indicator of the SNR [10] and penetration capability [28], which is defined as the maximum canopy cover through which the GEDI can detect the ground with a 90% probability [28]. The GEDI provides sensitivity for each algorithm, denoted as sensitivity_a<n> (n: 1–6), representing sensitivity from six different algorithms. The selection of the threshold values and the three parameters aligns with most established practices in previous studies [10,14,17,18].

2.4. Terrain Elevation Anomalies Extraction and Analysis

The Z-score method was used to detect terrain elevation anomalies. After extracting terrain elevation from spaceborne LiDAR products ( T 1 ) and ALS data ( T 2 , the reference terrain elevation), we calculated the terrain elevation error (E) as E = T 1 − T 2 . We calculated the Z-score of E (i.e., Z ( E ) ) for each laser shot by
Z ( E ) = ( E − E ¯ ) / σ ( E )
where E ¯ and σ ( E ) denote the average value and standard deviation of E for all the laser shots studied, respectively. Terrain elevation anomalies were detected by identifying laser shots with an absolute value of Z ( E ) greater than three. To further investigate these anomalies, we extracted the transmitted and received waveforms for laser shots identified as anomalous and the three-dimensional ALS point cloud data located in the spaceborne footprints. We computed the Pearson correlation coefficient between the transmitted waveform of each anomalous laser shot and the mean transmitted waveform of other shots to investigate if the anomalies were due to abnormal outgoing laser pulses. Additionally, we assessed whether the anomalies were algorithm-dependent by comparing terrain elevation data derived from different processing algorithms of the GEDI dataset using the coefficient of determination ( R 2 ), root mean square error (RMSE), and the mean absolute error (MAE). We also considered data acquisition time and beam type, and analyzed the statistical characteristics of the SNR and “sensitivity” of the laser shots with anomalies, including the mean, maximum, minimum, and standard deviation, to investigate the characteristics of the received waveform of anomalies. Selected transmitted and received waveforms were illustrated for inspection, and the three-dimensional ALS point cloud was used to provide additional insights into objects within the spaceborne footprints. Refer to Figure 2 for the flowchart of this study.

3. Results

3.1. Comparison of Terrain Elevations Acquired from Spaceborne and Airborne LiDAR Data

In examining the terrain elevations from the GLAS and estimated by default algorithm settings (A0) of the GEDI, we found outliers—90 in the GLAS dataset and 138 in the GEDI dataset—with significantly higher elevation values compared to the reference data. Their spatial distribution in the study area is illustrated in Figure 3. These outliers exhibited elevation errors ranging from hundreds to thousands of meters (Figure 4). Additional details on terrain elevation derived from the other six GEDI algorithms are presented in Table 2. Terrain elevation anomalies appear to be similar across different GEDI algorithms, with minimal variation observed. Specifically, the ranges and standard deviations of R 2 , RMSE, and MAE for terrain elevation anomalies are 0.51 to 0.57 ( σ = 0.02 ), 246.2 m to 290.1 m ( σ = 17.2   m ), and 226.8 m to 272.6 m ( σ = 17.7   m ), respectively, among different algorithms. This suggests that the observed anomalies are not due to the method used for calculating terrain elevation. Specifically, the sample size (n) varies among different algorithms because the GEDI provides sensitivity values for each algorithm that can differ from one another (see Figure A1); however, a consistent sensitivity threshold of 0.90 was applied during data filtering.

3.2. Characteristics of Terrain Elevation Anomalies

All the terrain elevations of anomalies are higher than the reference values from ALS for both the GLAS and the GEDI (Figure 5 and Figure 6c). The mean, maximum, minimum, and standard deviation of the differences between terrain elevations from the GLAS and the ALS-generated DEM are 1428.9 m, 2721.6 m, 264.6 m, and 754.6 m, respectively. In contrast, the corresponding statistics for the differences between terrain elevations from the GEDI and the ALS-generated DEM are 272.8 m, 357.6 m, 55.9 m, and 70.2 m, respectively.
Results show that anomalies were found in data acquired during both the day and night (Figure 5 and Figure 6a,b) and were not confined to a certain season. For the GLAS, 46% of the anomalies were acquired during the day and 54% at night (Figure 5e). In contrast, for the GEDI, 97% of the anomalies were acquired during the day and 3% at night (Figure 6(e2)). Terrain elevation anomalies were detected in three product files (4.5%) in the GEDI datasets, with average values and standard deviations ( σ ) of their solar elevation angles, the number of shots (n), and the acquisition season and year as follows: −20.15° ( σ = 0.01 ° , n = 4, spring, 2019), 9.76° ( σ = 0.12°, n = 66, winter, 2019), and 21.39° ( σ = 0.04°, n = 68, summer, 2019). For the GLAS, terrain elevation anomalies were found in four product files (18%), with average values and standard deviations of solar elevation angles, the number of shots (n), and the acquisition season and year as follows: −0.71° ( σ = 0.02°, n = 25, autumn, 2003), −19.95° ( σ = 1.64°, n = 24, summer, 2006), 15.12° ( σ = 0.02°, n = 23, autumn, 2004), and 25.13° ( σ = 0.04°, n = 19, spring, 2005). We observed that the anomalies were characterized by low absolute solar elevation angles (less than 26°) and occurred during both the daytime and nighttime without being confined to any particular season. Furthermore, although the GEDI employs both full power and coverage beams with varying energy levels, anomalies were observed in both types of beams: 95% in full-power beams and 5% in coverage beams (Figure 6(e1)). Terrain elevation anomalies in the GEDI were not found to appear only in a specific beam type.
Furthermore, the presence of anomalies is unrelated to the outgoing signal, as evidenced by the high Pearson correlation coefficients between each transmitted waveform of abnormal laser shots and the mean transmitted waveform of other shots. For the GLAS dataset, the average Pearson correlation coefficient is 0.94, with a maximum of 0.99, a minimum of 0.84, and a standard deviation of 0.04; the average p-value is 3.98 × 10 − 15 , with a maximum of 9.8 × 10 − 14 and a minimum of 0. For the GEDI dataset, the average Pearson correlation coefficient is 0.93, with a maximum of 0.99, a minimum of 0.74, and a standard deviation of 0.06; the p-value is always 0. Both the p-values for the GLAS and the GEDI are significantly less than 0.05, indicating that the correlation is significant. The transmitted laser pulse of the GLAS or the GEDI is close to a Gaussian distribution. See Figure 5d and Figure 6d for the transmitted waveforms of laser shots with terrain elevation anomalies and the mean transmitted waveform of other laser shots. Consequently, both the Pearson correlation coefficients and the visual inspection of the transmitted waveforms do not reveal any abnormal patterns of the outgoing signals.
The return waveform data for laser shots of terrain elevation anomalies exhibit a low SNR for the GLAS, with an average SNR of 16.5 and a standard deviation of 7.0 (Figure 5f), in contrast to an average SNR of 76 and a standard deviation of 19 for other laser shots. For the GEDI, sensitivity—an indicator of the SNR [10]—had an average value of 0.97 with a standard deviation of 0.02 (see the histogram in Figure 6f) for anomalous shots in contrast to an average sensitivity of 0.95 and a standard deviation of 0.04 for other laser shots.

4. Discussion

4.1. Origins of Terrain Elevation Anomalies and the Potential of GEDI in Atmospheric Science

The occurrence of terrain elevation anomalies in the GLAS and the GEDI is unrelated to outgoing laser pulses and is not confined to a certain data acquisition time (i.e., daytime or nighttime) or beam type but is instead associated with the scatter source of return signals. Outliers displaying significant differences between estimated terrain elevation and the reference value are evident in both the GLAS and the GEDI datasets. Figure 7 illustrates the transmit and return waveforms for instances where the absolute terrain elevation error of the GLAS or the GEDI is less than 0.85 m (upper panel) or exceeds 330 m (lower panel). It is noteworthy that terrain elevations from ALS were not depicted in the return waveforms (Figure 7, lower panel) due to considerable disparities between ALS and GLAS or GEDI elevations. Upon examining all the waveforms of these outliers, most exhibited similarities to the scenarios depicted in the figures. The outgoing laser pulse travels through the transmission medium to the land surface and then returns to the altimeter instrument. No abnormal situations were detected in the transmitted pulses, as evidenced by Figure 7 (middle) and all source laser shots of the outliers (Figure 5 and Figure 6d). Considering the overestimation of terrain elevation, ranging from hundreds to thousands of meters (Figure 5 and Figure 6c), and the high reflectance of clouds at 1064 nm (e.g., approximately 0.8 as reported in [29]), the return signals stored in GLAS and GEDI waveforms might not originate from a land surface but instead from clouds. Assuming the cloud to be a turbid medium, the nearly exponential decay trend of the return signals (refer to Figure 7 (right, lower panel) and additional examples in Figure A2) further supports our inference.
Clouds with low transmittance result in diminished energy reaching the receiver from the canopy. The measurement setup of LiDAR is mono-static, meaning that the transmitter and receiver share the same optical path. Full-waveform sensors record the highly detailed reflected energy of the outgoing laser pulse as it interacts with the reflecting surfaces. Assuming complete cloud coverage over the laser spot, a portion of the transmitted energy may pass through the cloud layer and ultimately reach the land surface, albeit with reduced intensity due to cloud transmittance. Subsequently, photons reflected from the forest canopy undergo an additional passage through the cloud layer. Consequently, acquiring return signals from the land surface via spaceborne full-waveform LiDAR becomes challenging in the presence of clouds. However, this circumstance underscores the potential utility of the GLAS and the GEDI in cloud height detection, which is crucial for modeling surface and atmospheric radiative fluxes [30]. In contrast to the conventional practice of detecting the location of the last Gaussian component as the terrain elevation, identifying the position of the first Gaussian component through Gaussian decomposition would acquire the elevation of the cloud top.
In comparison to the GLAS, GEDI products have seldom been reported in atmospheric science. The GLAS has been successfully utilized in detecting cloud layer height (e.g., GLA09 product [30]) using data from the 532 nm channel, while data from the 1064 nm channel are considered supplementary for denser clouds [31,32]. Although the primary science objective of the GEDI is not focused on clouds, the 1064 nm channel has the capability to detect and penetrate optically thick clouds [31]. In addition to cloud height, the temporal variation of return signals from clouds provides an attenuated backscatter profile that contains information about optical depth. For instance, larger optical thickness causes the exponential decay trend observed in the cloud’s return energy profile to become more pronounced. Given the path length information (the elevation difference from location A to location B) provided in the waveform, if an exponential fit is applied to the reflected waveform profile, the fitted coefficients will characterize the optical thickness. Thus, deriving cloud-top height and optical depth from the GEDI might be another potentially valuable application, which, although significant, has rarely been explored and reported in previous studies.
In terms of sampling capabilities, especially footprint size and vertical resolution, the GEDI outperforms the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) on Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) and the Cloud Profiling Radar (CPR) on CloudSat. The footprint diameter and vertical resolution are approximately 25 m and 0.15 m for the GEDI [3], 70 m and 30 m for the CALIOP [33], and 1.5 km and 240 m for the CPR [34], respectively. Although the GEDI dataset reveals a limited incidence of terrain elevation anomalies attributed to cloud effects (i.e., less than 0.5%) and is characterized by discontinuous sampling of the Earth, the GEDI is capable of providing more granular data in both footprint size and vertical resolution for atmospheric research. Moreover, although anomalies appear infrequently in our study area, their presence is concentrated along specific ground tracks of the sensors (see Figure 3), likely caused by cloud cover. This suggests that combining the GEDI with the widely used CALIOP and CPR may provide complementary datasets for tracking temporal changes in cloud cover. With a projected operational lifespan extending to 2031, the GEDI may serve as a unique and valuable supplementary dataset for atmospheric science applications. It would be valuable for weather prediction and atmospheric dynamics study due to its potential in estimating cloud height. Additionally, its potential in cloud optical depth estimation makes the GEDI an important data source for modeling radiative fluxes between the land surface and the atmosphere, which are essential for understanding energy budgets and the role of clouds in climate systems [30,35,36].

4.2. Data Filtering Recommendations for Forest Parameters Extraction to Avoid the Influence of Abnormal Return Signals

The extracted forest structure parameters from abnormal return signals may not appear as outliers in the same way as terrain elevation; however, it is important to note that the signals recorded in the received waveform do not originate from the forest and, therefore, such data should not be used. For instance, consider canopy height [37,38,39], which represents the vertical distance between the top of the canopy and the ground as determined from the waveform data. Each return waveform contains 544 records for the GLAS and an average of 869 records (varies depending on the specific land surface) in this study for the GEDI. Adjacent records in the waveform are spaced at 0.15 m intervals (with a measurement rate of 1 ns); consequently, the relative distance covered by the return signal ranges up to 81.6 m (i.e., 544 × 0.15 m) for the GLAS and an average of 130.4 m (i.e., 869 × 0.15 m) for the GEDI. Although the return signals originate from clouds rather than the forest, the canopy height derived from the outlier waveform did not exhibit significant abnormality (refer to Figure 8 for the GEDI canopy height product from laser shots of terrain elevation anomalies), thus impeding identification. Return signals originating from clouds rather than from the land surface result in an average RMSE of 20.6 m, MAE of 18.0, and R2 of 0.003 for the canopy height products derived from various algorithms of the GEDI in our study area. Incorrect canopy height in practical applications can significantly affect the estimation of forest biomass and carbon stock in models where canopy height is an input parameter [40,41,42]. For other forest structure parameters, such as the gap fraction—the probability that incoming beams do not intersect with any foliage elements and reach the ground—which relies on the computation of accumulated return energy from the forest canopy and ground, respectively, after the Gaussian decomposition of the waveform [20,21], the calculated gap fraction may appear normal (i.e., between 0 and 1) because it is a ratio of return energy, even if the return waveform originates from clouds. Additionally, for leaf area index retrieval, which depends on the gap fraction [20,43], the computed leaf area index may also appear in a normal range, like the example of canopy height. However, the uncertainties in gap fraction and leaf area index can introduce uncertainties in modeling forest growth and productivity [44], potentially affecting forest management decisions. As illustrated in Figure 3, anomalies in the GLAS and GEDI datasets are concentrated along specific ground tracks of the sensors rather than being randomly distributed across the study area. This pattern indicates that forest management strategies relying on parameters derived directly from spaceborne full-waveform LiDAR or on products that combine spaceborne LiDAR data with passive optical imagery (e.g., canopy height) may lead to inappropriate management decisions in certain regions if return signals from clouds are not properly accounted for.
Given the cloud interference, it is crucial to investigate effective methods for filtering out abnormal data; this is important not only for terrain elevation assessment and applications but also for the extraction of forest parameters that rely on data products from these sensors.
We analyzed the characteristics of Signal-to-Noise Ratio (SNR) and cloud flag, commonly employed for quality filtering in the GLAS, and the “sensitivity”, typically utilized for quality filtering in the GEDI. Although the GLA05 product includes a cloud flag parameter (elv_cloud_flg) for identifying cloud cover, a significant 97.4% of the data is flagged as cloud, suggesting a notable inaccuracy and rendering it unsuitable for data filtering. Additionally, although the waveforms of outliers in the GLAS are characterized by a low SNR, proposing an SNR threshold for quality filtering is challenging due to potential dependence on the utilized samples. For most outliers in the GEDI, the “sensitivity” hovers around 0.97, evidently exceeding the 0.90 threshold we set. The sensitivity parameter serves to identify “best” data [27] and has been widely adopted for quality filtering of GEDI data. For example, previous studies used a sensitivity threshold of 0.95, among other parameters, to mitigate cloud-contaminated observations [45]. However, our study cases suggest that “sensitivity” is algorithm-dependent (Figure A1). Importantly, a high “sensitivity” does not necessarily ensure that the return signals originate from the land surface, which has significant implications for future GEDI-related research. In conclusion, it is challenging to filter out laser shots with abnormal return signals using only the commonly employed quality filtering parameters for both the GLAS and the GEDI.
Since elevation information is a direct measurement from LiDAR, we recommend establishing an elevation-difference threshold between the GLAS or the GEDI terrain elevation and a reference dataset to avoid the use of abnormal return signals in the extraction of forest parameters from spaceborne full-waveform LiDAR. The minimum elevation difference between the terrain elevation provided by spaceborne LiDAR and the ALS reference values is approximately 265 m for the GLAS dataset and 55 m for the GEDI dataset. To maintain consistency, an elevation-difference threshold of 55 m was selected for both datasets. Therefore, in addition to the commonly used data filtering strategies, we recommend filtering data using “elev_use_flg = 0” and “ ∆ h > 55   m ” for the GLAS, and filtering data using “degrade_flag = 0”, “quality_flag = 1”, “sensitivity > 0.90”, and “ ∆ h > 55   m ” for the GEDI. Here, ∆ h indicates the elevation difference between the terrain elevation provided by spaceborne LiDAR and the reference elevation, such as that derived from the free Digital Elevation Model (DEM) product or ALS. Some studies related to forest structure or biomass have considered terrain by comparing terrain elevation from the spaceborne LiDAR product and reference elevations in data filtering for the GLAS [46,47] and the GEDI [7,12], while many have overlooked this factor. To the best of our knowledge, consensus regarding the consideration of terrain in quality-control filtering of spaceborne full-waveform LiDAR data has not been reached within the community. Therefore, it is imperative to heed this recommendation for data filtering when deriving forest parameters, including canopy height, relative height metrics, leaf area index (profile), fractional vegetation cover (profile), and biomass, from these missions as the return waveforms of outliers may be attributable to cloud interference.

4.3. Future Research Perspectives

This study identified several phenomena that are interesting but challenging to explain. For example, the frequency of abnormal cases differs significantly between the GLAS and the GEDI, with rates of 18.2% and 4.5%, respectively. Anomalies are present in a limited number of product files, specifically four HDF files for GLAS, accounting for 18.2% (i.e., four out of twenty-two), and three HDF files for GEDI, representing 4.5% (i.e., three out of sixty-six). When considering the percentage of abnormal cases in the total number of laser shots, GLAS shows a rate of 16.7% (i.e., 90 out of 541), while GEDI has a rate of less than 0.5% (i.e., 138 out of 29,662). Moreover, a notable difference exists in the range of terrain elevation anomalies between spaceborne LiDAR products and reference elevations from airborne LiDAR data. Specifically, the mean differences in terrain elevations of anomalies from the GLAS and the GEDI compared to the ALS-generated DEM are 1428.9 m and 272.8 m, respectively. These anomalies were detected both during the daytime and the nighttime for the GLAS, whereas they predominantly occurred during the daytime for the GEDI (97%). These phenomena can be attributed to occasional cloud cover; however, a more comprehensive study by the atmospheric science community and the sensor design community would contribute to a deeper understanding.
This study focused on forested regions in a canton of northern Switzerland, characterized by complex terrain and a temperate forest ecosystem. The largest difference in the terrain includes an elevation variation of 616 m and a slope variation of 45°. The appearance of elevation anomalies is primarily driven by return signals from clouds rather than the underlying forest. Consequently, the findings, particularly regarding the sources of anomalies identified in this study, should be regarded as generalizable. However, the occurrence rate of anomalies may vary across different terrains and forest ecosystems, a complexity further influenced by the variability of cloud formations in different regions of the Earth. Future research in tropical rainforests and boreal coniferous forests, for instance, could not only further validate and expand upon the findings of this study but could also evaluate the 55 m elevation-difference threshold we recommended for product utilization.

5. Conclusions

In this study, we conducted an exploration of anomalies in both GLAS and GEDI terrain elevation products. Unlike conventional assessment studies that seldom concentrate on waveforms and three-dimensional scenes, and waveform application studies that rarely focus on the true terrain elevation, we utilized both transmitted and return waveforms from spaceborne sensors, alongside a three-dimensional point cloud and its generated DEM from airborne LiDAR for this investigation. Despite substantial differences in footprint size, pulse duration, and transmitted power between sensors, we found considerable similarities in terrain elevation anomalies across the GLAS and the GEDI. We observed a low signal-to-noise ratio and a nearly exponential decay trend in the return waveform of laser shots of anomalies and found an overestimation of terrain elevation in anomalies, ranging from hundreds to thousands of meters. Importantly, our findings suggest that the scattering source is the primary cause of elevation anomalies, with return signals of outliers possibly originating from clouds rather than the land surface. This highlights the potential of deriving cloud-top height from the GEDI, suggesting promising prospects for applying GEDI data in atmospheric science, despite it not being the primary science objective of the GEDI and receiving scant attention in prior studies. Given the interference of clouds, incorporating spaceborne LiDAR-offered terrain elevation into data filtering is particularly crucial for studies aiming to derive forest parameters, which are the predominant applications of spaceborne full-waveform LiDAR, from the GLAS and the GEDI. However, a consensus on this has not yet been reached within the community. Therefore, we recommend implementing a threshold of 55 m for the elevation difference (i.e., the difference between the terrain elevation provided by spaceborne full-waveform LiDAR product and the reference terrain elevation from the free DEM product, airborne laser scanning data, or other high-precision data sources), in addition to the commonly used data filtering strategies for each dataset. This threshold was derived from data analysis conducted in a temperate forest ecosystem with approximately 500 km2 of forest coverage in northern Switzerland. Overall, this study provides initial insights into anomalies, crucial for optimizing the applications of spaceborne full-waveform LiDAR data.

Author Contributions

Conceptualization, H.J., Y.L. and G.Y.; methodology, H.J. and Y.L.; formal analysis, W.L. and L.L.; writing—original draft preparation, H.J., D.X., X.M., F.Y. and A.D.; writing—review and editing, G.Y., J.L., P.Z., K.X., J.G. and F.M.; visualization, Y.L.; supervision, G.Y. and F.M. All authors have read and agreed to the published version of the manuscript.

Funding

This work was funded by the key program of the National Natural Science Foundation of China (NSFC) (Grant No. 42301364), the Open Fund of State Key Laboratory of Remote Sensing Science (Grant No. E06102010A), the Fundamental Research Funds for the Central Universities (Grant No. JZ2023HGQA0143), the National Natural Science Foundation of China (NSFC) (Grant Nos. 42090013, 42301369, 42301363, 42401413), and the Fundamental Research Funds for the Central Universities (Grant No. JZ2022HGTB0253).

Data Availability Statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

Acknowledgments

The authors would like to thank the anonymous reviewers for their helpful and valuable comments that have greatly improved the paper.

Conflicts of Interest

The authors declare no conflicts of interest.

Appendix A

Forests 15 01821 g0a1
Figure A1. Probability density of “sensitivity” of “power” and “coverage” beams estimated by different algorithms (a–f) of GEDI using the data with “sensitivity > 0.90” in all footprints. A0 denotes the default algorithm setting (a), and A<n> (n: 1–6) denotes the other six algorithm groups (b–g).
Figure A1. Probability density of “sensitivity” of “power” and “coverage” beams estimated by different algorithms (a–f) of GEDI using the data with “sensitivity > 0.90” in all footprints. A0 denotes the default algorithm setting (a), and A<n> (n: 1–6) denotes the other six algorithm groups (b–g).
Forests 15 01821 g0a1
Forests 15 01821 g0a2
Figure A2. Original GLAS (upper panel) and GEDI (lower panel) waveform examples of terrain elevation anomalies with terrain elevation provided by GLAS and GEDI product indicated. A0 denotes the default algorithm, and A<n> (n: 1–6) denotes the other six algorithm groups of GEDI.
Figure A2. Original GLAS (upper panel) and GEDI (lower panel) waveform examples of terrain elevation anomalies with terrain elevation provided by GLAS and GEDI product indicated. A0 denotes the default algorithm, and A<n> (n: 1–6) denotes the other six algorithm groups of GEDI.
Forests 15 01821 g0a2

References

  1. Chen, Q. Assessment of Terrain Elevation Derived from Satellite Laser Altimetry over Mountainous Forest Areas Using Airborne Lidar Data. ISPRS J. Photogramm. Remote Sens. 2010, 65, 111–122. [Google Scholar] [CrossRef]
  2. Afzal, R.S.; Yu, A.W.; Dallas, J.L.; Melak, A.; Lukemire, A.T.; Ramos-Izqueirdo, L.; Mamakos, W. The Geoscience Laser Altimeter System (GLAS) Laser Transmitter. IEEE J. Sel. Top. Quantum Electron. 2007, 13, 511–535. [Google Scholar] [CrossRef]
  3. 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]
  4. Dubayah, R.; Tang, H.; Armston, J.; Luthcke, S.; Hofton, M.; Blair, J.B. GEDI L2B Canopy Cover and Vertical Profile Metrics Data Global Footprint Level V002 [Data Set]; U.S. Government, NASA EOSDIS Land Processes Distributed Active Archive Center: Sioux Falls, SD, USA, 2021. [Google Scholar]
  5. Dubayah, R.; Hofton, M.; Blair, J.B.; Armston, J.; Tang, H.; Luthcke, S. GEDI L2A Elevation and Height Metrics Data Global Footprint Level V002 [Data Set]; U.S. Government, NASA EOSDIS Land Processes Distributed Active Archive Center: Sioux Falls, SD, USA, 2021. [Google Scholar]
  6. Zwally, H.J.; Schutz, R.; Bentley, C.; Bufton, J.; Herring, T.; Minster, J.; Spinhirne, J.; Thomas, R. GLAS/ICESat L2 Global Land Surface Altimetry Data, Version 34 [Data Set]; NASA National Snow and Ice Data Center Distributed Active Archive Center (NSIDC DAAC): Boulder, CO, USA, 2014. [Google Scholar]
  7. Dubayah, R.; Armston, J.; Healey, S.P.; Bruening, J.M.; Patterson, P.L.; Kellner, J.R.; Duncanson, L.; Saarela, S.; Ståhl, G.; Yang, Z.; et al. GEDI Launches a New Era of Biomass Inference from Space. Environ. Res. Lett. 2022, 17, 095001. [Google Scholar] [CrossRef]
  8. Guerra-Hernández, J.; Pascual, A. Using GEDI Lidar Data and Airborne Laser Scanning to Assess Height Growth Dynamics in Fast-Growing Species: A Showcase in Spain. For. Ecosyst. 2021, 8, 1–17. [Google Scholar] [CrossRef]
  9. Hoffrén, R.; Lamelas, M.T.; de la Riva, J.; Domingo, D.; Montealegre, A.L.; García-Martín, A.; Revilla, S. Assessing GEDI-NASA System for Forest Fuels Classification Using Machine Learning Techniques. Int. J. Appl. Earth Obs. Geoinf. 2023, 116, 103175. [Google Scholar] [CrossRef]
  10. Wang, C.; Elmore, A.J.; Numata, I.; Cochrane, M.A.; Shaogang, L.; Huang, J.; Zhao, Y.; Li, Y. Factors Affecting Relative Height and Ground Elevation Estimations of GEDI among Forest Types across the Conterminous USA. GIScience Remote Sens. 2022, 59, 975–999. [Google Scholar] [CrossRef]
  11. Hayashi, M.; Saigusa, N.; Oguma, H.; Yamagata, Y. Forest Canopy Height Estimation Using ICESat/GLAS Data and Error Factor Analysis in Hokkaido, Japan. ISPRS J. Photogramm. Remote Sens. 2013, 81, 12–18. [Google Scholar] [CrossRef]
  12. Zhu, X.; Nie, S.; Wang, C.; Xi, X.; Lao, J.; Li, D. Consistency Analysis of Forest Height Retrievals between GEDI and ICESat-2. Remote Sens. Environ. 2022, 281, 113244. [Google Scholar] [CrossRef]
  13. Cobb, A.R.; Dommain, R.; Sukri, R.S.; Metali, F.; Bookhagen, B.; Harvey, C.F.; Tang, H. Improved Terrain Estimation from Spaceborne Lidar in Tropical Peatlands Using Spatial Filtering. Sci. Remote Sens. 2023, 7, 100074. [Google Scholar] [CrossRef]
  14. Adam, M.; Urbazaev, M.; Dubois, C.; Schmullius, C. Accuracy Assessment of Gedi Terrain Elevation and Canopy Height Estimates in European Temperate Forests: Influence of Environmental and Acquisition Parameters. Remote Sens. 2020, 12, 3948. [Google Scholar] [CrossRef]
  15. Pronk, M.; Ledoux, H.; Eleveld, M. Assessing Vertical Accuracy and Spatial Coverage of ICESat-2 and GEDI Spaceborne Lidar for Creating Global Terrain Models. Remote Sens. 2024, 16, 2259. [Google Scholar] [CrossRef]
  16. Narin, O.G.; Lindenbergh, R.; Abdikan, S. Multi-Criteria Strategy for Estimating GEDI Terrain Height. In Proceedings of the 10th International Conference on Recent Advances in Air and Space Technologies, RAST 2023, Istanbul, Turkey, 7–9 June 2023. [Google Scholar]
  17. 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]
  18. Liu, A.; Cheng, X.; Chen, Z. Performance Evaluation of GEDI and ICESat-2 Laser Altimeter Data for Terrain and Canopy Height Retrievals. Remote Sens. Environ. 2021, 264, 112571. [Google Scholar] [CrossRef]
  19. Enßle, F.; Heinzel, J.; Koch, B. Accuracy of Vegetation Height and Terrain Elevation Derivedfrom ICESat/GLAS in Forested Areas. Int. J. Appl. Earth Obs. Geoinf. 2014, 31, 37–44. [Google Scholar] [CrossRef]
  20. Jiang, H.; Yan, G.; Tong, Y.; Cheng, S.; Yang, X.; Hu, R.; Li, L.; Mu, X.; Xie, D.; Zhang, W.; et al. Correcting Crown-Level Clumping Effect for Improving Leaf Area Index Retrieval from Large-Footprint LiDAR: A Study Based on the Simulated Waveform and GLAS Data. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2021, 14, 12386–12402. [Google Scholar] [CrossRef]
  21. 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]
  22. Duncanson, L.I.; Niemann, K.O.; Wulder, M.A. Estimating Forest Canopy Height and Terrain Relief from GLAS Waveform Metrics. Remote Sens. Environ. 2010, 114, 138–154. [Google Scholar] [CrossRef]
  23. Cui, L.; Jiao, Z.; Zhao, K.; Sun, M.; Dong, Y.; Yin, S.; Li, Y.; Chang, Y.; Guo, J.; Xie, R.; et al. Retrieval of Vertical Foliage Profile and Leaf Area Index Using Transmitted Energy Information Derived from ICESat GLAS Data. Remote Sens. 2020, 12, 2457. [Google Scholar] [CrossRef]
  24. Shu, S.; Liu, H.; Frappart, F.; Huang, Y.; Wang, S.; Hinkel, K.M.; Beck, R.A.; Yu, B.; Jones, B.M.; Arp, C.D.; et al. Estimation of Snow Accumulation over Frozen Arctic Lakes Using Repeat ICESat Laser Altimetry Observations—A Case Study in Northern Alaska. Remote Sens. Environ. 2018, 216, 529–543. [Google Scholar] [CrossRef]
  25. Zwally, H.J.; Schutz, R.; Bentley, C.; Bufton, J.; Herring, T.; Minster, J.; Spinhirne, J.; Thomas, R. GLAS/ICESat L1B Global Waveform-Based Range Corrections Data, Version 34 [Data Set]; NASA National Snow and Ice Data Center Distributed Active Archive Center (NSIDC DAAC): Boulder, CO, USA, 2014. [Google Scholar]
  26. Hofton, M.; Blair, J.B. Algorithm Theoretical Basis Document (ATBD) for GEDI Transmit and Receive Waveform Processing for L1 and L2 Products. 2019. Available online: https://lpdaac.usgs.gov/documents/581/GEDI_WF_ATBD_v1.0.pdf (accessed on 1 January 2024).
  27. Beck, J.; Armston, J.; Hofton, M.; Luthcke, S.; Tang, H. GLOBAL Ecosystem Dynamics Investigation (GEDI) Level 2 User Guide Version 2 (University of Maryland, NASA Goddard Space Flight Center, KBR Greenbelt). 2021; pp. 1–25. Available online: https://lpdaac.usgs.gov/documents/986/GEDI02_UserGuide_V2.pdf (accessed on 1 January 2024).
  28. Hancock, S.; Armston, J.; Hofton, M.; Sun, X.; Tang, H.; Duncanson, L.I.; Kellner, J.R.; Dubayah, R. The GEDI Simulator: A Large-Footprint Waveform Lidar Simulator for Calibration and Validation of Spaceborne Missions. Earth Sp. Sci. 2019, 6, 294–310. [Google Scholar] [CrossRef] [PubMed]
  29. Ding, N.; Shao, J.; Yan, C.; Zhang, J.; Qiao, Y.; Pan, Y.; Yuan, J.; Dong, Y.; Yu, B. Near-Ultraviolet to near-Infrared Band Thresholds Cloud Detection Algorithm for Tansat-Capi. Remote Sens. 2021, 13, 1906. [Google Scholar] [CrossRef]
  30. Hart, W.D.; Spinhirne, J.D.; Palm, S.P.; Hlavka, D.L. Height Distribution between Cloud and Aerosol Layers from the GLAS Spaceborne Lidar in the Indian Ocean Region. Geophys. Res. Lett. 2005, 32, L22S06. [Google Scholar] [CrossRef]
  31. Spinhirne, J.D.; Palm, S.P.; Hart, W.D.; Hlavka, D.L.; Welton, E.J. Cloud and Aerosol Measurements from GLAS: Overview an Initial Results. Geophys. Res. Lett. 2005, 32, 1–5. [Google Scholar] [CrossRef]
  32. Wang, X.; Cheng, X.; Gong, P.; Huang, H.; Li, Z.; Li, X. Earth Science Applications of ICESat/GLAS. Int. J. Remote Sens. 2011, 32, 8837–8864. [Google Scholar] [CrossRef]
  33. Winker, D.M.; Vaughan, M.A.; Omar, A.; Hu, Y.; Powell, K.A.; Liu, Z.; Hunt, W.H.; Young, S.A. Overview of the CALIPSO Mission and CALIOP Data Processing Algorithms. J. Atmos. Ocean. Technol. 2009, 26, 2310–2323. [Google Scholar] [CrossRef]
  34. Protat, A.; Bouniol, D.; O’Connor, E.J.; Baltink, H.K.; Verlinde, J.; Widener, K. CloudSat as a Global Radar Calibrator. J. Atmos. Ocean. Technol. 2011, 28, 445–452. [Google Scholar] [CrossRef]
  35. Lambert, F.H.; Webb, M.J.; Yoshimori, M.; Yokohata, T. The Cloud Radiative Effect on the Atmospheric Energy Budget and Global Mean Precipitation. Clim. Dyn. 2015, 44, 2301–2325. [Google Scholar] [CrossRef]
  36. Zhou, C.; Zelinka, M.D.; Klein, S.A. Impact of Decadal Cloud Variations on the Earth’s Energy Budget. Nat. Geosci. 2016, 9, 871–874. [Google Scholar] [CrossRef]
  37. Li, Y.; Gao, S.; Fu, H.; Zhu, J.; Hu, Q.; Zeng, D.; Wei, Y. Error Analysis and Accuracy Improvement in Forest Canopy Height Estimation Based on GEDI L2A Product: A Case Study in the United States. Forests 2024, 15, 1536. [Google Scholar] [CrossRef]
  38. Doyog, N.D.; Lin, C. Generating Wall-to-Wall Canopy Height Information from Discrete Data Provided by Spaceborne LiDAR System. Forests 2024, 15, 482. [Google Scholar] [CrossRef]
  39. Ni, W.; Li, Z.; Wang, Q.; Zhang, Z.; Liu, Q.; Pang, Y.; He, Y.; Li, Z.; Sun, G. Forest Heights Extraction Using GF-7 Very High-Resolution Stereoscopic Imagery and Google Earth Multi-Temporal Historical Imagery. J. Remote Sens. 2024, 4, 0158. [Google Scholar] [CrossRef]
  40. Zhang, G.; Ganguly, S.; Nemani, R.R.; White, M.A.; Milesi, C.; Hashimoto, H.; Wang, W.; Saatchi, S.; Yu, Y.; Myneni, R.B. Estimation of Forest Aboveground Biomass in California Using Canopy Height and Leaf Area Index Estimated from Satellite Data. Remote Sens. Environ. 2014, 151, 44–56. [Google Scholar] [CrossRef]
  41. 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]
  42. Duncanson, L.; Kellner, J.R.; Armston, J.; Dubayah, R.; Minor, D.M.; Hancock, S.; Healey, S.P.; Patterson, P.L.; Saarela, S.; Marselis, S.; et al. Aboveground Biomass Density Models for NASA’s Global Ecosystem Dynamics Investigation (GEDI) Lidar Mission. Remote Sens. Environ. 2022, 270, 112845. [Google Scholar] [CrossRef]
  43. Jiang, H.; Cheng, S.; Yan, G.; Kuusk, A.; Hu, R.; Tong, Y.; Mu, X.; Xie, D.; Zhang, W.; Zhou, G.; et al. Clumping Effects in Leaf Area Index Retrieval from Large-Footprint Full-Waveform LiDAR. IEEE Trans. Geosci. Remote Sens. 2022, 60, 1–20. [Google Scholar] [CrossRef]
  44. Clark, D.B.; Olivas, P.C.; Oberbauer, S.F.; Clark, D.A.; Ryan, M.G. First Direct Landscape-Scale Measurement of Tropical Rain Forest Leaf Area Index, a Key Driver of Global Primary Productivity. Ecol. Lett. 2008, 11, 163–172. [Google Scholar] [CrossRef]
  45. Rishmawi, K.; Huang, C.; Zhan, X. Monitoring Key Forest Structure Attributes across the Conterminous United States by Integrating Gedi Lidar Measurements and Viirs Data. Remote Sens. 2021, 13, 442. [Google Scholar] [CrossRef]
  46. Wang, Y.; Li, G.; Ding, J.; Guo, Z.; Tang, S.; Liu, R.; Chen, J. A Combined GLAS and MODIS Estimation of the Global Distribution of Mean Forest Canopy Height. Remote Sens. Environ. 2016, 174, 24–43. [Google Scholar] [CrossRef]
  47. Gwenzi, D.; Lefsky, M.A. Modeling Canopy Height in a Savanna Ecosystem Using Spaceborne Lidar Waveforms. Remote Sens. Environ. 2014, 154, 338–344. [Google Scholar] [CrossRef]
Forests 15 01821 g001
Figure 1. Study area and the geolocation of GLAS and GEDI data.
Figure 1. Study area and the geolocation of GLAS and GEDI data.
Forests 15 01821 g001
Forests 15 01821 g002
Figure 2. Flowchart of this study.
Figure 2. Flowchart of this study.
Forests 15 01821 g002
Forests 15 01821 g003
Figure 3. Spatial distribution of terrain elevation anomalies in the GLAS and GEDI datasets.
Figure 3. Spatial distribution of terrain elevation anomalies in the GLAS and GEDI datasets.
Forests 15 01821 g003
Forests 15 01821 g004
Figure 4. Scatter plot of terrain elevation estimates obtained from GLAS (a) and GEDI (b) vs. the terrain elevation derived from airborne laser scanning (ALS) as a reference. A0 denotes the default algorithm of the GEDI L2A product.
Figure 4. Scatter plot of terrain elevation estimates obtained from GLAS (a) and GEDI (b) vs. the terrain elevation derived from airborne laser scanning (ALS) as a reference. A0 denotes the default algorithm of the GEDI L2A product.
Forests 15 01821 g004
Forests 15 01821 g005
Figure 5. Details of terrain elevation outliers from GLAS: scatter plot of terrain elevation from data acquired during nighttime (a) and daytime (b) before removing outliers, scatter plot (c), transmitted waveforms (d), the histogram of the data acquisition time (e), and the histogram of Signal-to-Noise Ratio (SNR) (f) of source laser shot of the outliers.
Figure 5. Details of terrain elevation outliers from GLAS: scatter plot of terrain elevation from data acquired during nighttime (a) and daytime (b) before removing outliers, scatter plot (c), transmitted waveforms (d), the histogram of the data acquisition time (e), and the histogram of Signal-to-Noise Ratio (SNR) (f) of source laser shot of the outliers.
Forests 15 01821 g005
Forests 15 01821 g006
Figure 6. Details of terrain elevation outliers from GEDI: scatter plot of terrain elevation from data acquired during nighttime (a) and daytime (b) before removing outliers, scatter plot (c), transmitted waveforms (d), the histogram (e) of the beam type (e1) and data acquisition time (e2), and the histogram of sensitivity (f) of source laser shot of the outliers.
Figure 6. Details of terrain elevation outliers from GEDI: scatter plot of terrain elevation from data acquired during nighttime (a) and daytime (b) before removing outliers, scatter plot (c), transmitted waveforms (d), the histogram (e) of the beam type (e1) and data acquisition time (e2), and the histogram of sensitivity (f) of source laser shot of the outliers.
Forests 15 01821 g006
Forests 15 01821 g007
Figure 7. Examples with small (upper panel) and large (lower panel) terrain elevation error: the three-dimensional scene (left), the transmitted waveform (middle), and the return waveform (right) of GLAS and GEDI with the terrain elevation from product and airborne laser scanning (ALS) data illustrated. In the lower panel (right), the ALS terrain elevation is not indicated since the GLAS or GEDI terrain elevation exceeds ALS by more than 330 m (see outlier-27 and outlier-12 indicated in green circles in Figure 5 and Figure 6c). A<n> (n: 1–6) denotes the terrain elevation from six different algorithm groups of GEDI.
Figure 7. Examples with small (upper panel) and large (lower panel) terrain elevation error: the three-dimensional scene (left), the transmitted waveform (middle), and the return waveform (right) of GLAS and GEDI with the terrain elevation from product and airborne laser scanning (ALS) data illustrated. In the lower panel (right), the ALS terrain elevation is not indicated since the GLAS or GEDI terrain elevation exceeds ALS by more than 330 m (see outlier-27 and outlier-12 indicated in green circles in Figure 5 and Figure 6c). A<n> (n: 1–6) denotes the terrain elevation from six different algorithm groups of GEDI.
Forests 15 01821 g007
Forests 15 01821 g008
Figure 8. Scatter plot of canopy height estimates for laser shots of terrain elevation anomalies obtained from GEDI L2A product versus the canopy height derived from airborne laser scanning (ALS) as a reference (the legend of the point density applies to all the figures). A0 denotes the default algorithm (a), and A<n> (n: 1–6) denotes the other six algorithm groups (b–g).
Figure 8. Scatter plot of canopy height estimates for laser shots of terrain elevation anomalies obtained from GEDI L2A product versus the canopy height derived from airborne laser scanning (ALS) as a reference (the legend of the point density applies to all the figures). A0 denotes the default algorithm (a), and A<n> (n: 1–6) denotes the other six algorithm groups (b–g).
Forests 15 01821 g008
Table 1. A summary of the GLAS and GEDI products used.
Table 1. A summary of the GLAS and GEDI products used.
MissionGLASGEDI
Products usedGLA14: used to extract terrain elevationL2A: used to extract terrain elevation
GLA01: mainly used to illustrate the waveform L1B: mainly used to illustrate the waveform
GLA05: used to extract the cloud flag
VersionRelease 34Version 2
Terrain parameterd_elev (elevation relative to the Topex ellipsoid)elev_lowestmode (elevation relative to the WGS84 ellipsoid)
Other key parameters usedd_satElevCorr: used to correct the elevation; d_deltaEllip: used to convert elevation from the Topex to WGS84 ellipsoidquality_flag, degrade_flag, sensitivity: used to filter the samples
d_gpCntRngOff, d_ldRngOff, d_GmC: used to compute the elevation at the location of the last modebeam: used to identify the “power” and “coverage” beams
d_SolAng: used to judge the data acquisition time (daytime, d_SolAng > 0)solar_elevation: used to evaluate the data acquisition time (daytime, solar_elevation > 0)
d_maxRecAmp, d_sDevNsOb1: used to compute the Signal-to-Noise Ratio (SNR) [23], usually used to quantify the level of signal relative to the level of background noise. ( S N R = d _ m a x R e c A m p / d _ s D e v N s O b 1 )
d_lon, d_lat: geolocation of the footprint lon_lowestmode, lat_lowestmode: geolocation of the footprint
r_tx_wf, r_rng_wf: transmitted and return waveformtxwaveform, rxwaveform: transmitted and return waveform
Table 2. Statistical items for terrain elevation anomalies and normal data from different GEDI algorithms (A0—default algorithm, A1—algorithm 1, A2—algorithm 2, …, A6—algorithm 6).
Table 2. Statistical items for terrain elevation anomalies and normal data from different GEDI algorithms (A0—default algorithm, A1—algorithm 1, A2—algorithm 2, …, A6—algorithm 6).
AlgorithmTerrain Elevation AnomaliesNormal Data
nR2RMSE (m)MAE (m)nR2RMSE (m)MAE (m)
A01380.54267.2250.329,5240.998.55.0
A11450.52290.1272.627,6000.9910.76.5
A21590.56257.0237.930,6760.998.14.8
A31480.52288.3270.428,8800.999.75.9
A41460.51288.5269.227,6000.9910.76.5
A51510.57246.2226.830,9680.999.97.0
A61520.53269.6249.630,1950.998.45.1
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

Jiang, H.; Li, Y.; Yan, G.; Li, W.; Li, L.; Yang, F.; Ding, A.; Xie, D.; Mu, X.; Li, J.; et al. Unveiling Anomalies in Terrain Elevation Products from Spaceborne Full-Waveform LiDAR over Forested Areas. Forests 2024, 15, 1821. https://doi.org/10.3390/f15101821

AMA Style

Jiang H, Li Y, Yan G, Li W, Li L, Yang F, Ding A, Xie D, Mu X, Li J, et al. Unveiling Anomalies in Terrain Elevation Products from Spaceborne Full-Waveform LiDAR over Forested Areas. Forests. 2024; 15(10):1821. https://doi.org/10.3390/f15101821

Chicago/Turabian Style

Jiang, Hailan, Yi Li, Guangjian Yan, Weihua Li, Linyuan Li, Feng Yang, Anxin Ding, Donghui Xie, Xihan Mu, Jing Li, and et al. 2024. "Unveiling Anomalies in Terrain Elevation Products from Spaceborne Full-Waveform LiDAR over Forested Areas" Forests 15, no. 10: 1821. https://doi.org/10.3390/f15101821

APA Style

Jiang, H., Li, Y., Yan, G., Li, W., Li, L., Yang, F., Ding, A., Xie, D., Mu, X., Li, J., Xu, K., Zhao, P., Geng, J., & Morsdorf, F. (2024). Unveiling Anomalies in Terrain Elevation Products from Spaceborne Full-Waveform LiDAR over Forested Areas. Forests, 15(10), 1821. https://doi.org/10.3390/f15101821

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