Applying Quality Control Measures to Validate Airborne Lidar Bathymetry in Diverse Environments

Abstract

Airborne lidar (light detection and ranging) technology represents an evolution of traditional photogrammetry due to its capability to produce dense and accurate three-dimensional representation of Earth's surfaces. Installed on a variety of mobile platforms, airborne lidar systems (ALS) can generate vast amounts of positional information with high planimetric accuracy and detail, using artificially produced laser pulses, commonly in the near-infrared spectrum (1064 nm). The technology is effective in mapping topographic details, and researchers have applied the information to various applications, such as urban mapping, forestry, environmental assessment, and hazard management. Technological advancements in sensors, auxiliary hardware, and computing software have enabled the manufacturing of more efficient and versatile systems, and the addition of a visible (green) wavelength (515-532 nm) has revolutionized the shallow water mapping industry. Dual wavelength "bathymetric" airborne lidar systems (ALB) can support a wide range of applications in hydrographic surveying, coastal zone monitoring, and riverbed and lake mapping, and has lately also been used in cryospheric studies. With access to a variety of remotely sensed information, researchers can utilize datasets sourced from different sensors and platforms to conduct interdisciplinary studies, contributing to the discovery of new patterns and processes in geomorphology. To use airborne lidar data effectively, researchers need to validate data quality in terms of accuracy, completeness, and consistency. Particularly with airborne lidar bathymetry, measurements require verification by conducting rigorous in-situ campaigns and a comparison to datasets sourced from other applicable methods.

In the research described in this thesis, I utilized two airborne lidar systems with bathymetric capabilities (Leica Chiroptera I and 4X) and conducted a variety of supplemental surveys to investigate the quality control aspects related to measuring topographic and bathymetric surfaces and water depths.

In the Frio River in southwestern Texas, springs provide recharge along the outcrop portion of the Edwards Aquifer, and the river empties into the Nueces River. Because of the recent drought causing environmental concerns related to pathogens and bacteria build-up in non-flowing pool conditions, and the related concerns about recreational business loss, the Texas Water Development Board inquired about and funded an extensive airborne lidar survey. Consequently, in 2017, UT Austin researchers acquired airborne lidar data, supplemented with sonar soundings and GNSS (Global Navigation Satellite System) surveys in the river watershed. The in-situ campaign included sampling of turbidity and water transparency at multiple basins, and the mean turbidity was less than 1 nephelometric unit. Therefore, the transparency was high, and the river bottom was visible to the observer, creating an ideal environment for mapping with bathymetric lidar. As an unconventional validation method, I conducted waveform analysis to confirm water surface heights and depths, at locations where the river riparian zone was covered with vegetation and blocking the GNSS signals. I further validated the findings with sonar and GNSS measurements, where applicable. The results differed depending on the basin characteristics; however, the mean lidar bathymetry and sonar depths agreed within 16% or better depending on the basin characteristics (0.14 m to 0.24 m). Comparison to sonar soundings produced a mean absolute depth difference of 0.13 m, and the agreement between the measurements was high (R²=0.72). I concluded the research by verifying the importance of conducting in-situ measurements even in almost ideal environmental conditions and demonstrated the feasibility of conducting waveform analysis at locations where in-situ observations are not feasible.

In the Devils River study, the motivation was to address the potential data gap in stream bathymetry, which was critical to parameterize the physical stream-flow models for mapping the aquatic habitat locations of the Devils River minnow, an endangered species. It was challenging to measure the complete riverbed with an airborne lidar system and confirm depths where submerged vegetation (variable-leaf watermilfoil and emergent American water willow) blocked laser pulses. Therefore, I applied a novel approach and utilized a ground penetrating radar (GPR) from an inflatable boat with a shielded monostatic 200-MHz antenna that was mounted immediately above the water surface. Preliminary GPR measurements revealed a bias of 0.22 m (the mean depth was 1.77 m) to lidar bathymetry using the standard fresh-water permittivity rate (20° Celsius, radio waves propagation speed of 3.3 cm/ns, attenuation constant of 0.1 dB/m). To enhance the results, I adjusted the water-column permittivity (ε, 80 versus 78.4) based on the observed water temperature (24.5° Celsius), resulting in radio waves propagating quicker (3.7 cm/ns), and attenuating more slowly (0.08 dB/m), enhancing the height recordings especially in deeper parts of the river. As a result, the mean depths measured were equal (1.99 m) and agreement between the airborne lidar and GPR measurements was strong (R²=0.92). To validate the findings, I compared lidar bathymetry to sonar recordings, where applicable. In the upper basin, the comparison algorithm matched lidar bathymetry to sonar recordings at 53% in the user-defined thresholds (1 m radius, 0.5 m vertical spacing) and the mean absolute height difference was 0.11 m where depths were shallower than 2.75 m. In the lower basin, I observed a reduced concentration of submerged vegetation; however, depths were greater and influenced the agreement between the measurements adversely (R² = 0.72, versus 0.78 for the upper basin). Overall, the absolute depth difference was less than 0.1 m among comparison of all datasets. Nevertheless, I was able to build a bottom map, and successfully demonstrated the possibility of seamlessly combining datasets sourced from different sensors and confirmed the feasibility of using GPR for mapping riverbeds with submerged vegetation.

Lower Laguna Madre is a shallow, hyper-saline lagoon in southern Texas. It is large (800 km²) and has a unique seagrass ecosystem that serves as an essential nursery for species found in the Gulf of Mexico. The depth of the lagoon varies due to tides, and transparency is influenced by northerly winds that carry sediment from the surrounding sand dunes. In 2017, an airborne lidar survey was conducted and researchers completed an extensive in-situ campaign to build an updated geomorphic map of the lagoon and the surrounding areas. The in-situ data collection areas were clustered in the southern section of the lagoon; each area was unique in terms of turbidity, depth, and other environmental conditions such as floating and submerged vegetation, and varying bottom properties. My research motivation was to investigate the feasibility of measuring the lagoon bathymetry with airborne lidar at these diverse locations and investigate the accuracy of lidar measurements using satellite imagery and other survey methods. For this purpose, I conducted sonar surveys from a boat, also sampling turbidity and transparency. I adjusted the sonar sound wave propagation speed (1533.9, 1530.7, and 1537.5 m/s) by measuring the mean water temperature and salinity at each in-situ area and computed the sonar recordings accordingly. For the satellite-derived bathymetry, I observed tides that influenced the depths, and adjusted the surface heights to match the satellite sensing times prior to comparing them to the lidar bathymetry. To align all measurements with each other, I converted lidar measurements using the applicable geoid model and adjusted all measurements to NAVD 88 orthometric height. Further, I analyzed the pixel reflectance of the satellite imagery, classified the spectral variability, and excluded excessively turbid locations from further analysis. The lidar bathymetry indicated that 51% of the lagoon had depths greater than 0.4 m, with only 1.1% of the measurements exceeding 2 m. The deepest point recorded was at 3.35 m. Sonar recordings agreed in good confidence (R²=0.68) with lidar bathymetry and exceeded the Special-Order uncertainty standards (Order 1b) set by the International Hydrographic Organization (IHO), whereas feature detection is not specified, and the total vertical uncertainty (TVU) is less than 0.5 m at depths shallower than 10 m. In the less turbid locations (Area 1), the mean absolute depth difference between lidar bathymetry and sonar recordings was 5 cm, and RMSE was 14 cm. In Area 2, higher levels of turbidity decreased the comparison algorithm matching (40%) and the coefficient of determination (R²=0.38), resulting in an increased depth difference of 14 cm. Further comparison of lidar bathymetry to satellite imagery-derived depths revealed sufficient correlation (44-66 %) between the measurements due to the coarse pixel density of satellite imagery and the limitations of optical imaging technology in shallow and turbid waters. The mean absolute depth difference between lidar bathymetry and satellite-derived bathymetry was less than 13 cm, and RMSE was 15 cm. As a result, I demonstrated that higher levels of turbidity reduced the correlation rate between sonar and lidar measurements, impacting the depth accuracy adversely. However, the bias (the mean absolute depth difference) between lidar bathymetry, sonar and satellite-derived bathymetry was better than the IHO Order 1b standard. Furthermore, the study results demonstrated that optical imaging technology had depth resolving issues at varying turbid locations with depths exceeding 1 m, where lidar bathymetry was successful measuring deeper bottoms with more detail. Therefore, the study revealed the weaknesses and strengths of both technologies and demonstrated the practical and theoretical limitations of airborne lidar bathymetry and depths derived from satellite imaging in shallow waters with varying environmental conditions, filling a gap in the scientific literature, where both technologies have proven effective under ideal conditions.

In the summer of 2022, UT Austin and National Aeronautics and Space Administration (NASA) researchers participated in the ICESat-2 validation/calibration (Cal/Val 2022) campaign, based in Thule Air Force Base (Thule AB), northwest Greenland. The campaign provided sea ice measurements acquired using airborne lidar systems, a Leica Chiroptera 4X and NASA's Land, Vegetation, and Ice Sensor (LVIS). Both sensors acquired data simultaneously at low altitudes (< 2200 ft.), enabling cross-checking of heights. Despite the Chiroptera's low altitude, slow cruising speed constraints, and other logistical and environmental challenges, measurements almost coincident with those of the ICESat-2 observatory were successfully collected. The principal objective of this research was to align Chiroptera measurements to those of ICESat-2 measurements and compare sea ice heights and bathymetric depths. Because sea ice is dynamic and achieving a 1:1 correspondence of satellite and airborne altimeter measurements (time and location) is not possible, an algorithm was developed to adjust the location of each measured point in the Chiroptera lidar dataset by aligning features with ICESat-2 datasets, using linear regression. The algorithm iteratively adjusted Chiroptera measurements for ice drift speed and bearing combinations within user-defined range and increments, to a theoretical location where ICESat-2 and Chiroptera were coincident. For quality-control purposes, I verified the calibration of the Chiroptera system with GNSS checkpoint heights that were established on the Thule AB runway, and further validated the measurements with ATL03 photon heights. The mean absolute height difference was less than a centimeter on both confirmations. Additionally, I compared LVIS and Chiroptera sea ice measurements in ellipsoidal heights and the results matched with less than 4 cm (mean absolute difference) on a 31 km long data swath, indicating a strong level of agreement (R²=0.98, RMSE=0.047 m). I segmented the long Chiroptera data swath lines (~ 3 km) and applied geophysical height corrections, identical to NASA's tide free computations, for direct comparison to ATL07 segment heights. The tide free system includes the orthometric heights and excludes permanent tide deformations. I used the least-squares method for absolute height comparison where approximately 300 Chiroptera returns were registered within a 12-m diameter footprint of a single ATL07 beam. Before applying the drift adjustments, the height comparison generated a high bias (RMSE= 0.25 m), and the agreement between the measurements was poor (R²= 0.24). Results improved with drift adjustment, particularly for the stronger ICESat-2 r-series sensor product (mean absolute height difference = 0.015 m, R² = 0.73). For bathymetric measurements of melt ponds and occasional leads (fractures in the sea ice), I merged bathymetric lidar data classes (Classes 7, 8 and 10), computed the sea surface and mean ice surface heights, and compared depths in the height domain to the ATL03 photons that registered within a 2 m radius of each lidar return. Overall, depths compared were shallow (< 3 m) and ICESat-2 registered slightly deeper returns (mean absolute depth difference = 1 cm). The RMSE was 16 cm and depths agreed with R² = 0.84. As a result, the study quantified the measurement bias between ICESat-2 and Chiroptera airborne lidar measurements and introduced a novel algorithm for adjusting sea ice for in terms of speed and bearing drift, enabling direct comparisons.

This PhD research provides a scientific approach to understanding practical and theoretical limitations of bathymetric airborne lidar surveys and demonstrates the significance of quantifying height and depth differences (biases) using supplementary survey methods, particularly at locations with varying environmental conditions. For this purpose, each case study demonstrates an application of a range of survey methods with the purpose of generating validated datasets to build topographic and bathymetric maps in confidence.