The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring
The Global Ecosystem Dynamics Investigation (GEDI) LiDAR provides new spaceborne vegetation canopy structural information including relative canopy height products defined with respect to 25 m diameter footprints. The GEDI geolocation requirement is that each 25 m footprint center is horizontally ge...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2021-12-01
|
| Series: | Science of Remote Sensing |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2666017221000110 |
| _version_ | 1818461503766921216 |
|---|---|
| author | David P. Roy Herve B. Kashongwe John Armston |
| author_facet | David P. Roy Herve B. Kashongwe John Armston |
| author_sort | David P. Roy |
| collection | DOAJ |
| description | The Global Ecosystem Dynamics Investigation (GEDI) LiDAR provides new spaceborne vegetation canopy structural information including relative canopy height products defined with respect to 25 m diameter footprints. The GEDI geolocation requirement is that each 25 m footprint center is horizontally georeferenced to within 10 m (1σ), assuming normally distributed geolocation errors with a 0 m mean and a 10 m standard deviation. The impact of this geolocation uncertainty on the reliability of forest canopy height estimation is examined considering Airborne Laser scanner (ALS) and GEDI data acquired in 2014 and 2019 respectively. A total of 445 GEDI footprints acquired over 2000 ha of unforested and tropical secondary forest in the western Democratic Republic of the Congo with vegetation heights ranging from 1 m to 42 m are considered. Airborne true color 10 cm imagery and an ALS derived canopy height model are examined to contextualize the results. GEDI waveforms are simulated from the ALS data at the reported locations of the GEDI footprints and used to derivehˆ95,hˆ85,hˆ75relative heights that define the canopy height relative to the ground below which 95%, 85% and 75% of the simulated cumulative waveform energy is returned. A Monte Carlo simulation is undertaken, moving the centers of each GEDI footprint with 300 randomly generated position errors modelled using the GEDI geolocation uncertainty (0 m mean, 10 m standard deviation), and each time simulating the GEDI waveform from the ALS data. Relative heights are extracted from the 300 simulated GEDI waveforms and their variation, defined by the 25th and 75th percentiles, and the interquartile range (IQR) (75th - 25th percentiles), are quantified to provide insights into the impact of the GEDI geolocation uncertainty on forest canopy height retrieval. The IQR accounts for 50% of the variation in the forest canopy height due to GEDI geolocation uncertainty. High IQR values, greater than or comparable to the relative height derived from the ALS data at the GEDI reported footprint location are shown to occur where the footprint covered or was adjacent to spatially heterogeneous canopies, including canopies with small forest stands, holes in the vegetation canopy, and forest edges. This is a concern for the use of GEDI data acquired over these conditions which are prevalent in many forest systems. The impact of GEDI geolocation uncertainty on tropical forest change monitoring is demonstrated by comparing the GEDI h95 product footprint values (sensed in 2019) with simulated hˆ95 values derived from the ALS data (sensed in 2014) at the GEDI product reported footprint location and at the 300 shifted footprint locations. GEDI footprints where five-year canopy height changes, and not changes due to artefacts associated with the GEDI geolocation uncertainty or the GEDI simulator, are attributed conservatively. Differences among the six algorithm setting group GEDI h95 product relative height values are evident and influential on the change attribution. PlanetScope 3 m imagery sensed in 2019 are examined to provide qualitative evidence that support the efficacy of the approach for forest height reduction monitoring. The simulation approach described in this study provides a route to determine if forest canopy height change found by comparing multi-temporal data (for example, GEDI with previously collected ALS data or GEDI data) is significant relative to errors imposed by the GEDI geolocation. The treatment of change is simple, and recommendations for improvements to detect more subtle change are made. The study was undertaken using the Release 1.0 GEDI data and suggests, pending planned geolocation improvement, the need to accommodate for GEDI geolocation uncertainty, particularly over canopies that are spatially fragmented or that have heterogeneous three dimensional structure at scales comparable to the 25 m GEDI footprint dimension. |
| first_indexed | 2024-12-14T23:47:11Z |
| format | Article |
| id | doaj.art-96ecf1add6bf4cbd8a64d458ef6c7955 |
| institution | Directory Open Access Journal |
| issn | 2666-0172 |
| language | English |
| last_indexed | 2024-12-14T23:47:11Z |
| publishDate | 2021-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Science of Remote Sensing |
| spelling | doaj.art-96ecf1add6bf4cbd8a64d458ef6c79552022-12-21T22:43:21ZengElsevierScience of Remote Sensing2666-01722021-12-014100024The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoringDavid P. Roy0Herve B. Kashongwe1John Armston2Center for Global Change and Earth Observations, Michigan State University, East Lansing, MI, USA; Department of Geography, Environment, and Spatial Sciences, Michigan State University, East Lansing, MI, USA; Corresponding author. Michigan State University, East Lansing, MI, USADepartment of Geography, Environment, and Spatial Sciences, Michigan State University, East Lansing, MI, USADepartment of Geographical Sciences, University of Maryland, College Park, MD, USAThe Global Ecosystem Dynamics Investigation (GEDI) LiDAR provides new spaceborne vegetation canopy structural information including relative canopy height products defined with respect to 25 m diameter footprints. The GEDI geolocation requirement is that each 25 m footprint center is horizontally georeferenced to within 10 m (1σ), assuming normally distributed geolocation errors with a 0 m mean and a 10 m standard deviation. The impact of this geolocation uncertainty on the reliability of forest canopy height estimation is examined considering Airborne Laser scanner (ALS) and GEDI data acquired in 2014 and 2019 respectively. A total of 445 GEDI footprints acquired over 2000 ha of unforested and tropical secondary forest in the western Democratic Republic of the Congo with vegetation heights ranging from 1 m to 42 m are considered. Airborne true color 10 cm imagery and an ALS derived canopy height model are examined to contextualize the results. GEDI waveforms are simulated from the ALS data at the reported locations of the GEDI footprints and used to derivehˆ95,hˆ85,hˆ75relative heights that define the canopy height relative to the ground below which 95%, 85% and 75% of the simulated cumulative waveform energy is returned. A Monte Carlo simulation is undertaken, moving the centers of each GEDI footprint with 300 randomly generated position errors modelled using the GEDI geolocation uncertainty (0 m mean, 10 m standard deviation), and each time simulating the GEDI waveform from the ALS data. Relative heights are extracted from the 300 simulated GEDI waveforms and their variation, defined by the 25th and 75th percentiles, and the interquartile range (IQR) (75th - 25th percentiles), are quantified to provide insights into the impact of the GEDI geolocation uncertainty on forest canopy height retrieval. The IQR accounts for 50% of the variation in the forest canopy height due to GEDI geolocation uncertainty. High IQR values, greater than or comparable to the relative height derived from the ALS data at the GEDI reported footprint location are shown to occur where the footprint covered or was adjacent to spatially heterogeneous canopies, including canopies with small forest stands, holes in the vegetation canopy, and forest edges. This is a concern for the use of GEDI data acquired over these conditions which are prevalent in many forest systems. The impact of GEDI geolocation uncertainty on tropical forest change monitoring is demonstrated by comparing the GEDI h95 product footprint values (sensed in 2019) with simulated hˆ95 values derived from the ALS data (sensed in 2014) at the GEDI product reported footprint location and at the 300 shifted footprint locations. GEDI footprints where five-year canopy height changes, and not changes due to artefacts associated with the GEDI geolocation uncertainty or the GEDI simulator, are attributed conservatively. Differences among the six algorithm setting group GEDI h95 product relative height values are evident and influential on the change attribution. PlanetScope 3 m imagery sensed in 2019 are examined to provide qualitative evidence that support the efficacy of the approach for forest height reduction monitoring. The simulation approach described in this study provides a route to determine if forest canopy height change found by comparing multi-temporal data (for example, GEDI with previously collected ALS data or GEDI data) is significant relative to errors imposed by the GEDI geolocation. The treatment of change is simple, and recommendations for improvements to detect more subtle change are made. The study was undertaken using the Release 1.0 GEDI data and suggests, pending planned geolocation improvement, the need to accommodate for GEDI geolocation uncertainty, particularly over canopies that are spatially fragmented or that have heterogeneous three dimensional structure at scales comparable to the 25 m GEDI footprint dimension.http://www.sciencedirect.com/science/article/pii/S2666017221000110GEDIForestChangeGeolocationAirborneLiDAR |
| spellingShingle | David P. Roy Herve B. Kashongwe John Armston The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring Science of Remote Sensing GEDI Forest Change Geolocation Airborne LiDAR |
| title | The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring |
| title_full | The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring |
| title_fullStr | The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring |
| title_full_unstemmed | The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring |
| title_short | The impact of geolocation uncertainty on GEDI tropical forest canopy height estimation and change monitoring |
| title_sort | impact of geolocation uncertainty on gedi tropical forest canopy height estimation and change monitoring |
| topic | GEDI Forest Change Geolocation Airborne LiDAR |
| url | http://www.sciencedirect.com/science/article/pii/S2666017221000110 |
| work_keys_str_mv | AT davidproy theimpactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring AT hervebkashongwe theimpactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring AT johnarmston theimpactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring AT davidproy impactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring AT hervebkashongwe impactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring AT johnarmston impactofgeolocationuncertaintyongeditropicalforestcanopyheightestimationandchangemonitoring |