Background

Most lidar applications over the past several decades have focused on bathymetry, water surface height statistics, and biological activity using airborne scanning systems (Brock and Purkis, 2009; Guenther, 1985, 2007; Klemas, 2011). Bathymetric mapping generally includes coincident 532- and 1064-nm lidar, often in conjunction with hyperspectral imagery (Ackermann, 1999; Guenther, Tomas, and LaRocque, 1996; Krabill et al., 2002; Lillycrop, Pope, and Wozencraft, 2002) and high scan-rate systems such as the Experimental Advanced Airborne Research lidar (EAARL) (Bonisteel et al., 2009; McKean et al., 2009; Nayegandhi, Brock, and Wright, 2009; Wright et al., 2014) and the Scanning Hydrographic Operational Airborne Lidar Survey (SHOALS) (Irish and Lillycrop, 1999; Lillycrop, Irish, and Parson, 1997). Both high and low energy commercial systems are employed depending on environmental conditions. High energy systems that offer deep penetration but sparse pixel spacing include the Hawkeye II (Tulldahl and Wikstrom, 2012), the Laser Airborne Depth Sounder (LADS) MK3, and the Coastal Zone Mapping and Imaging Lidar (CZMIL) (Feygels et al., 2012; Fuchs and Mathur, 2010) systems. Low-altitude systems (< 3,000 m above ground) typically employ approximately 250-m swath widths 200–700-m above ground yielding vertical accuracies of 15 cm over 1 m spatial scale.

Low energy commercial systems suitable for shallow water and high spatial density observations include EAARL (Wright and Brock, 2002), Riegl VQ-880 series (Pfennigbauer et al., 2011), Optech's Aquarius (Pan et al., 2015), and the High-Resolution Quantum Lidar System (HRQLS) (Degnan et al., 2011). Example experimental low-energy photon counting systems include the low altitude Swath Imaging Multipolarization Photon counting Lidar (SIMPL) (Dabney et al., 2010; Harding et al., 2011) and the high-altitude (20-km above ground) Cloud Physics Lidar (CPL) (McGill et al., 2002). The photon counting systems, when combined with smaller telescopes and the elimination of automatic gain control, offer up to two orders of magnitude greater receiver performance than analog lidars (Krainak et al., 2010).

Space-based retrievals of water properties have evolved over the past two decades. The first generation Geoscience Laser Altimeter System (GLAS) aboard ICESat-1 (Abshire et al., 2005) consisted of a single beam, low repetition rate (O[10²]Hz), high pulse energy (O[10] mJ) lidar with an approximately 70-m footprint and along track spacing of about 170 m. Inland water observations were successfully explored with accuracies in the cm to decimeter range, and its height products were used in a number of both lake and river studies (Birkett et al., 2010; Calmant, Seyler, and Cretaux, 2008; Harding and Jasinski, 2004; Zhang et al., 2011a, 2011b). Future mission concepts, in addition to ICESat-2, that will employ photon counting detectors include the Lidar Surface Topography (LIST), the Active Sensing of CO₂ Emissions over Nights, Days, and Seasons (ASCENDS), and the Aerosols-Clouds-Ecosystem (ACE) missions.

In addition to range determination, the analysis of satellite-observed specular reflectance has allowed retrieval of additional water properties (Barrick, 1968; Bufton, Hoge, and Swift, 1983; Lancaster, Spinhirne, and Palm, 2005; Menzies, Tratt, and Hunt, 1998). Lancaster, Spinhirne, and Palm (2005) used the near-nadir ICESat GLAS reflectance to estimate ocean surface albedo. Menzies, Tratt, and Hunt (1998) were the first to examine sea surface directional reflectance and wind speed using the Lidar In-space Technology Experiment (LITE) instrument aboard the space shuttle. Hu et al. (2008) examined surface wind speed variability using NASA's Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) backscatter data employing the Cox and Munk slope variance–wind speed relations. Several satellite lidar studies over oceans have focused on wind speed retrieval that relies on knowledge of backscatter distribution from wave slope facets (Hu et al., 2008; Lancaster, Spinhirne, and Palm, 2005; Menzies, Tratt, and Hunt, 1998). CALIOP observations over Tampa Bay were also used to investigate subsurface scattering (Barton and Jasinski, 2011).

The analysis of lidar returns from photon counting systems is, in many ways, similar to analysis using analog systems (Churnside, Naugolnykh, and Marchbanks, 2014; Guenther, 1985; Guenther, LaRocque, and Lillycrop, 1994). A principal difference is that, instead of analyzing a full waveform return from a single pixel illuminated by a high energy analog pulse, an equivalent, but not identical, histogram must first be generated from along track returns. The required track length depends on surface reflectance, atmospheric conditions, and solar background. In general, aggregations of at least 100 signal photons are sufficient for mean height analysis (Jasinski et al., 2015). For dark targets such as water, experience with MABEL indicates that about 0.5 to 1 signal photons per meter are returned (Jasinski et al., 2015). Range is measured from the time difference in between the laser pulse and the reflected light. Return intensity provides information on target characteristics. Factors affecting signal performance include Fresnel scattering from the water surface, water volume scattering and absorption, clouds, solar background, and bottom reflectance. While conceptually simple, execution requires precise measurements and timing. Optical water clarity is the most limiting factor for depth detection (Sinclair, 2008). In general, lidar technology can detect light down to about three times the Secchi depth (Estep, Lillycrop, and Parson, 1994; Sinclair, 1999) under ideal conditions. Recommended guidelines to achieve optimal performance include flying at night, low wind conditions, clear water, low altitude, and maximum sounding energy (Sinclair and Penley, 2007). Analysis of data from high-altitude aircraft platforms must also account for atmospheric scattering and delay and for aircraft pitch, roll, and yaw perturbations. Procedures to compare the various lidar waveform processing algorithms of different systems are available (Parrish et al., 2011; White et al., 2011).

ICESat-2 ATLAS Mission

The soon-to-launch Advanced Topographic Laser Altimeter System (ATLAS) is the only instrument on the polar-orbiting Ice, Cloud, and Land Elevation Satellite (ICESat-2) mission. ICESat-2 is a Tier 1 mission recommended by the National Research Council (NRC, 2007). Its principal objectives are to quantify polar ice sheet contributions to sea level change and the linkages to climate conditions, quantify regional signatures of ice sheet changes, estimate sea ice thickness from freeboard measurements, and quantify and map vegetation height over a two-year period (Abdalati et al., 2010). However, the ICESat2 mission will also develop inland water and ocean data products. The Inland Water data product, or ATL13, will consist of principally the mean and standard deviation of water surface height for ICESat-2 transects over global lakes, rivers, and near coastal regions.

ATLAS is configured as a six-beam laser altimeter utilizing a high repetition rate (10k Hz), short pulse width, 532-nm laser transmitter with photon counting detectors, as shown in Figure 1. The spacing is configured to observe local cross slope within a beam pair and wide spatial coverage between the three sets of pairs. Each beam pair consists of a comparatively low energy (40 μJ) and strong energy (121 μJ) beam to better observe the full dynamic range of dark (water, vegetation) and bright (snow, ice) targets, respectively (McGill et al., 2013; Zwally et al., 2011).

Figure 1.

Schematic of ICESat-2 ATLAS six beam configuration. Credit: ICESat-2 Project Office.

ICESat-2/ATLAS is thus significantly different from its predecessor, ICESat/GLAS, which fired at a much lower rate (40 Hz) but employed ~80 mJ lasers for full waveform detection (Abshire et al., 2005; Schutz et al., 2005). In addition to the higher repeat frequency, ATLAS will offer near-continuous 0.70-m ground spacing with approximately 14-m footprints compared to GLAS's 170-m spacing and 70-m footprints. Each returned photon will be time-tagged with a vertical precision of approximately 30 cm, depending on surface and atmospheric characteristics (personal communication, Thomas Neumann, ICESat-2 Project Office). ATLAS also utilizes a narrower instrument field of view (FOV) to limit the observation of solar photons. The ATLAS system will thus provide higher measurement sensitivity with lower resource requirements. A summary of ATLAS parameters is shown in Table 1.

Table 1.

Summary comparison of the principal ATLAS and MABEL instrument parameters.

An additional unique feature of ICESat-2 is its two orbit modes. Above approximately +/− 65 deg latitude, ATLAS will operate in a repeat track mode over designated reference tracks similar to ICESat in order to obtain continuous time series of ice sheet change along those tracks. Below +/− 65 deg, however, ICESat-2 will systematically point left or right off the reference tracks in subsequent orbits in order to conduct a two-year global mapping of vegetation. Additional scheduled off-pointing is also planned to observe targets of opportunity and calibration/validation sites.

MABEL Prototype Instrument

The Multiple Altimeter Beam Experimental Lidar (MABEL) was built as a high-altitude prototype of the ATLAS instrument (McGill et al., 2013), but possessing additional beams and flexibility to test variations in the ICESat-2 concept. In this capacity it serves several purposes, including validation of ICESat models of instrument performance, evaluation of the photon counting system in the 532-nm band, providing experiment data over actual ICESat-2 targets, and development of retrieval algorithms of ICESat-2 data products. From 2012 through 2015, major flight experiments were conducted in Greenland, the east coast of the United States, the western United States, and Alaska. In all these experiments, MABEL was flown aboard either the ER-2 or Proteus Aircraft at 20 km or above 95% of the Earth's atmosphere. The high-altitude platform more realistically replicates the impact of clouds that ICESat-2 will encounter, and that will need to be addressed in the retrieval algorithms.

A summary comparison of the relevant ATLAS and MABEL instrument parameters is provided in Table 1. A unique feature of MABEL is that it possesses much flexibility in the configuration of several main lidar parameters. For example, it possesses up to 16 active channels at 532 nm and 8 at 1064 nm with changeable viewing angles, as shown in Figure 2. Laser repetition rate can be varied from 5 to 25 Hz. At 5 kHz and at an aircraft ground speed of 200 m/s, a pulse is thus emitted every 4 cm. Laser mean pulse width is 2 ns.

Figure 2.

Typical MABEL beam configuration uses up to 16 active channels at 532 nm and 8 at 1064 nm.

Aim of This Study

The purpose of the present study is to analyze MABEL along track profiles of water surface height over inland and near-shore waters, and to evaluate what features can be derived from the ICESat-2/ATLAS instrument. The analysis is pertinent in the development of planned retrieval algorithms for the ICESat-2 Inland Water Body Height data product (ATL13). The primary ATL13 products are surface water height statistics including mean, standard deviation, and slope. However, they cannot be derived without considering additional processes that affect the retrieval, including the subsurface backscatter from the water column, the impact of a possible bottom signal in shallow areas, and meteorology. Analyses of the five cases reported herein serve to evaluate both the feasibility of the ATLAS photon counting lidar system for water surface profiling and to define the quality limits of the ATL13 data product.

Site Selection

Three sites from within the above experiments were selected to evaluate MABEL data over a range of inland and near-shore water bodies and operating conditions: i) Upper Chesapeake Bay near Gooses Reef on both September 22, 2012, at 00:56 UTC (evening local time) and September 25, 2013, at 16:51 UTC (noon local time), ii) Atlantic Ocean near Virginia Beach on September 20, 2013, at 22:23 UTC (early evening local time), and iii) Lake Mead on February 24, 2012, at 6:15 UTC (night local time).

The 2012 flights were flown aboard the ER-2 at an altitude of 20 km and an air speed of about 750 km/hr out of Dryden Air Force Base and the Wallops Flight Facility for the lake Mead and Chesapeake Bay cases, respectfully. The 2013 flights were flown aboard the Proteus out of Langley Air Force Base at about 12 km altitude and air speed of approximately 500 km/hr. Aircraft used are shown in Figure 3.

Figure 3.

NASA ER-2 (a) and Proteus (b) aircraft used in the MABEL Experiments. Credits: NASA photo

Analyses

Once collected, data were first processed for individual photon geolocation by the ICESat Project Office. All MABEL data were grouped into granules of one-minute flight time. Accounting for the speed of the aircraft, each one-minute granule consists of about 99,000 photons covering a distance of about 12 km for ER-2 flights and 8 km for Proteus flights. Instantaneous photon height data are reported with respect to the WGS84 ellipsoid for all ice and land products. However, instantaneous heights were further converted in the case of inland water to orthometric height using the EGM96 Geoid.

Analysis of the data consisted of first plotting the along track heights of the individual MABEL photons to identify water surface, subsurface backscatter, and potential bottom reflectance. Several statistical parameters were then estimated including mean background rate, the rate of observed water surface photons (per meter of transect length), the observed mean geodetic and orthometric heights, the standard deviation of the water surface height and the MABEL subsurface attenuation coefficient. These are reported in Table 2 along with in situ observations when available.

Table 2.

Summary parameters of the MABEL experiments.

Also computed was an expression of the vertical profile of MABEL's observable subsurface backscattered signal photons. This was formulated as the ratio of the depth-dependent signal photon density to mean background density, SBR(d), written (after Schroeder, 1999) as

Because both the total observed return and the mean background can be computed directly from the observed vertical profile, and because the background can range over several orders of magnitude, Equation 1 is more conveniently rewritten as

Site 1: Middle Chesapeake Bay

The two Chesapeake Bay transects are shown in Figure 4. They represent contrasting day and night open water cases with moderate wind and turbidity with mostly clear sky conditions. Both transects consist of a one-minute acquisition along nearly identical 8-km reaches in the middle of the bay near NOAA's Gooses Reef buoy. The September 22, 2012, flight occurred during late evening local time and the September 25, 2013, flight during midday local time. There were no land crossings and water depth was greater than 10 m.

Figure 4.

Location map of high-altitude MABEL flights over Site 1 in both 2012 (green line) and 2013 (red line) on Chesapeake Bay near Gooses Reef buoy. Base map from Google Earth.

Plots of the georeferenced MABEL photon cloud returns from the atmosphere through the water column with respect to the WGS84 Geodetic height are shown in Figures 5a and b. The plots consist primarily of i) background photons throughout the atmosphere and water column, ii) a concentrated band of photons of about a meter wide representing the water surface, and iii) an additional band of subsurface backscattered photons extending a few meters below the water surface and diminishing with depth. The above profiles are typical of most MABEL water transects.

Figure 5.

Along track profile of MABEL-observed photons for Site 1 during the 2012 (a) and 2013 (b) flights on Chesapeake Bay near Gooses Reef.

The plots indicate notable differences in background rates, surface signal photon rates, and SBR penetration between the two dates. The 2012 late evening flight exhibits an almost negligible background rate of 0.00002 m⁻² for this nighttime flight. The 2013 midday flight, however, exhibits variable background along the flight line, shown in Figure 5b and in an expanded view in Figure 6, with a moderate background rate of 0.011 m⁻² between a distance of 2,000 and 4,300 m, followed by a low background rate of 0.0053 m⁻² over the distance 4,300–6,300 m. The different backgrounds for the same 2013 transect represent differences in cloud cover within the instrument field of view. Clouds increase the solar scattering while reducing the lidar surface signal.

Figure 6.

Expanded view of MABEL 2013 data at site 1, Chesapeake Bay near Gooses Reef buoy. Results also indicate SBR₁₀ depths of −1.3 m and −3.7 m for moderate (0.011m⁻²) and low (0.0053m⁻²) background, respectively, for the same turbidity of 2.9 NTU.

Analysis also indicates that for the 2013 Chesapeake cases, that occurred during daytime within a minute of each other, nearly four times as many water surface photons were detected, or 2.20 m⁻¹ versus 0.56 m⁻¹, in the low background segment compared to the moderate segment, respectively. As indicated in Figure 5b, however, the moderate background segment still easily possesses sufficient photons to clearly define the water surface.

The LSBR(d) profiles of the Chesapeake cases, shown in Figure 7, indicate the observable limits of MABEL's subsurface volume scattering. Results indicate that the LSBR(d) profile for 2013 decays faster for the moderate background segment compared to the low background 2013 case. The observable penetration of both 2013 cases is less than the 2012 Chesapeake case, indicating greater observability at night when there is no solar background.

Figure 7.

Signal to background ratio profiles versus depth, LSBR(d), for five cases presented in this study, expressed in Log₁₀ base. Also indicated is the LSBR₀ threshold level.

For quantitative comparison of the observable MABEL penetration, it is useful to choose a threshold level, say LSBR₀, representing the depth at which the signal-to-noise ratio equals one or Log₁₀(SBR) equals 0. Results shown in Table 2 and Figures 6 and 7 indicate that LSBR₀ equals 1.3 m and 3.7 m for the 2013 moderate and low backgrounds, respectively, despite having the same turbidity of 2.9 NTU. As defined, MABEL's observable LSBR₀ depth is not only a function of the intrinsic properties of the water but also the relative intensity of the incident signal photons compared to the background. Lower background makes it easier to discern a given signal strength. For the 2012 late evening case, this observable depth or LSBR₀ equals 6.8 m, a much deeper depth, resulting largely from the very low background.

Once LSBR₀ is defined, the attenuation of the MABEL subsurface backscattered signal can be explored and modeled as an exponential decay with depth. The water penetration of a 532-nm laser beam has been shown to decrease exponentially proportional to the diffuse attenuation coefficient (Guenther, 1985; Feygels et al., 2003). MABEL analyses yielded attenuation coefficients of α₅₃₂ = 0.91 m⁻¹ (R² = 0.53) and 0.56 m⁻¹ (R² = 0.84) for the moderate and low background cases, respectively. Lower R² generally occurs with the smaller LSBR₀ depth as there are fewer data to fit the subsurface decay. In situ measurements of diffuse solar light at 532.2 nm were also made along the 2013 transect using a free falling HyperPro II by Satlantic. Data were averaged over three casts for each location. The mean of the upwelling and downwelling diffuse attenuation coefficients were K₅₃₂ = 0.45 m⁻¹ (R² = 0.99) and 0.52 m⁻¹ (R² = 0.99) for the moderate and low background cases, respectively. The estimated MABEL-based attenuation is thus slightly higher with a lower R² than the in situ results. Error sources include difference in instrumentation, spatial variability in water turbidity over the length of the transect, and some differences in the precise time of acquisition.

Site 2: Atlantic Ocean near Virginia Beach

The second site analyzed was an east-west transect extending from the Atlantic Coast at Virginia Beach, just south of the mouth of the Chesapeake, eastward into the Atlantic on September 19, 2013, at 22:30 UTC (late afternoon local time). Figure 8 shows the transect location map, which is situated just south of the mouth of the Chesapeake Bay. A 20-second segment of about 2,000 MABEL photons is plotted in Figure 9. For this date, sky conditions were mostly clear, and the wind was from the east at 4.2 m/s. One additional feature not seen in the Chesapeake Bay cases is evidence of some wave structure throughout the transect. This is attributed to the MABEL flight being aligned nearly parallel to the wind direction.

Figure 8.

Location map of high-altitude MABEL flights over Site 2, Atlantic Ocean near Virginia Beach. Base map from Google Earth.

Figure 9.

Along track profile of MABEL-observed photons for Site 2, Atlantic Coast at Virginia Beach. LSBR₀ depth indicated at 9.3 m below surface. Estimated envelope of near shore subsurface signal photons suggests approximate location of bottom.

Also plotted on Figure 9 is the LSBR₀ depth estimated to be about 9.3 m. This comparatively high penetration is attributed to a combination of the lower turbidity of 2.2 NTU compared to the Chesapeake Bay cases, and a low background rate of 0.0003 m⁻². The attenuation coefficient is estimated to be α₅₃₂ = 0.55 m⁻¹ (R² = 0.95).

Furthermore, although a distinct bottom is not identified even near the shore, evidence of an approximate implied bottom may be possible when there is a sharp drop in LSBR. Drawn on Figure 9 is an estimated envelope of MABEL's subsurface signal photons in the vicinity of the shore. The envelope suggests that the water depth extends down to about 4 m at a distance of about 200 m from shore. Although precise measurements of bathymetry were not recorded at the time of the MABEL flight, the depth of the envelope curve is consistent with current bathymetric soundings available from the National Ocean Service Hydrographic Data Base, NOAA National Centers for Environmental Information ( https://www.ngdc.noaa.gov/mgg/bathymetry/hydro.html).

A vertical histogram of the water surface height computed from the aggregated along track MABEL photon elevations is shown in Figure 10. Photon heights are plotted with respect to the WGS84 Ellipsoid. Orthometric heights using the EGM96 Geoid are also provided in Table 1. This histogram does not represent the true statistical distribution of the surface photons as the effect of the instrument impulse response is convolved with the returned signal. The ICESat-2 ATL13 Inland Water Height Data Product algorithm deconvolves the MABEL signal, providing an estimate of the true representation of the distribution of the surface variability also shown in Figure 10. The estimated true water surface height distribution for the Site 2 case yields a standard deviation of 0.21 m and an offset of about 0.30 m in the mode from the MABEL observations.

Figure 10.

Histograms of the water surface photons for Site 2, Atlantic Ocean at Virginia Beach, for (a) the MABEL observations and (b) the estimated true surface distribution after deconvolution.

Site 3: Lake Mead

This case represents a night flight over a relatively clear water body with turbidity equal to 1.6 NTU. The MABEL overpass of February 24, 2012, transected the western portion of Lake Mead in a southwest to northeast direction as shown in Figure 11. The transect represents two granules of data, or about 2 minutes of acquisition, covering about 24 km.

Figure 11.

Location map of high-altitude MABEL flights over Site 3, Lake Mead. Base map from Google Earth.

The corresponding plot of the MABEL photons is shown in Figure 12, with the southwestern corner of the lake on the left. During the flight approximately 91,000 photons were recorded. Because of the nighttime and clear sky conditions, there was an extremely low background count of 0.00008 m⁻². Several features are clearly identified. First, starting at the edge of the lake and traversing across, several islands are noted. To the far right of the figure, after passing over a large island nearly 60 m high, the aircraft reaches the edge of the lake. Subsurface backscatter results in an estimated LSBR₀ depth of 9.2 m. The attenuation coefficient is estimated to be α₅₃₂ = 0.37 m⁻¹ (R² = 0.73).

Figure 12.

Along track profile of water body and near-shore MABEL-observed photons for Site 3, Lake Mead, on February 24, 2012, 6:15-17 UTC.

Unlike the previous cases, the bathymetry of Lake Mead is very apparent in the vicinity of many of the shorelines of the lake edges and islands. To see this more clearly, an expanded view of the photons is plotted in Figure 13 for the southwest shore. Prior to plotting, data were first processed to remove an instrument after pulse at about 1.4 m depth. The near-shore bottom of the lake is observed as an extension of the shoreline to a depth of nearly 9 m. The corresponding histograms of the surface and subsurface photons are shown for an open water stretch of 2 km in Figure 14a and a near-shore stretch of 100 m in Figure 14b. In the open water segment, the water depth is much greater than the LSBR₀ depth of 9.2 m and no bottom signal is detected. For the near-shore profile, a bottom bump in the histogram is clearly identified at about a depth of 2 m. Although a detailed map is not available, these results are consistent with the NOAA Nautical Chart 18687 of the National Ocean Service Coast Survey ( http://www.oceangrafix.com/chart/zoom?chart=18687).

Figure 13.

Expanded view of MABEL observations at land water crossing of Lake Mead on the southwest shore. Results show penetration of the 532-nm channel into the water column and the presence of lake bottom along the western edge, up to a depth of about 10 m.

Figure 14.

Histograms of MABEL data for (a) a deep, open water 2-km stretch near the center of the Lake Mead, and (b) a near-shore 100-m stretch near southwestern Lake Mead. Typical bump in near-shore histogram indicates lake depth of about 2 m.