Gedi Data Within Google Earth Engine: Potentials And Analysis For Inland Surface Water Monitoring

Inland surface water is the source of about 60% and a key component of the hydrological cycle. The monitoring of inland surface water is fundamental to understanding the effects of climate change on this key resource and preventing water stresses. Water levels traditionally measured by ground instruments like gauge stations are expensive and have high maintenance costs. Conversely, Earth Observation technologies can nowadays collect frequent and regular data with continuous monitoring of water reservoirs, reducing monitoring costs.

 

With the availability of new data, the need for a capable computation tool is crucial. Google Earth Engine (GEE), a cloud-based computation platform capable of integrating a high variety of datasets with powerful analysis tools [1], has recently added the Global Ecosystem Dynamics Investigation (GEDI) [4] to its wide archive. 

 

The GEDI [2] instrument, hosted onboard the International Space Station,  is a geodetic-class, light detection and ranging (LiDAR) system, having a 25 m spot (footprint) on the surface over which 3D structure is measured. The footprints are separated by 60 m along-track, with an across-track distance of about 600 m. The measurements are made over the Earth's surface nominally between the latitudes of 51.6° and -51.6°. GEDI was originally developed to enable radically improved quantification and understanding of the Earth’s carbon cycle and biodiversity. 

 

The available literature highlights that the quality of GEDI data is variable and impacted by several factors (e.g., latitude, orbit). Our preliminary analysis is focused on the accuracy assessment of the GEDI data, at first addressing the problem of outliers detection and removal, and secondly comparing the water levels measured by GEDI with reference ground truth; thus, we considered four lakes in Northern Italy for which level measurements from gauge stations are available.

The proposed outlier detection consists of two steps for each GEDI passage over water surfaces.

The first step is based on two flags implanted within GEDI bands. Specifically, the “quality_flag” indicates if the considered footprint has valid waveforms (1=valid, 0=invalid), due to anomalies in the energy, sensitivity, and amplitude of signals; the “degrade_flag” indicates the degraded state of pointing (saturation intensity of returned photons might reduce the accuracy of measurements) and/or positioning information (GPS data gap, GPS receiver clock drift).

The second step relies on the robust version of the standard 3σ test, implemented considering the NMAD (Normalized Median Absolute Deviation): every GEDI measurement not within -/+3*NMAD from the median is removed as outlier.

To assess the outlier detection procedure and to preliminarily evaluate the accuracy of the GEDI data, we compared the water levels inferred from the median of GEDI measurements after outlier removal with the contemporary water levels from hydrometric stations at four major lakes (Como, Garda, Iseo, Maggiore) in Northern Italy [3]. The comparison is ongoing over the period from GEDI activation until June 2022, for 3 years.

References

[1] Cardille, et al., 2022. Cloud-Based Remote Sensing with Google Earth Engine.

[2] Dubayah, et al., 2021. GEDI L3 gridded land surface metrics, version 1

[3] Enti Regolatori dei Grandi Laghi, 2022. Home Page - Laghi. www.laghi.net.

[4] University of Maryland, 2022. GEDI ecosystem lidar