3.1.

Dual-View Capability

In a field setting, it is very difficult to decouple the effect of the sky illumination from that of the solar illumination in the observed target reflectance across the full range of the solar irradiance spectrum. Many researchers have relied on atmospheric models to aid in the understanding the impact of the sky illumination. However, this approach can be prone to modeling errors if the model input parameters are not well characterized.^14,15 Typically, these parameters are determined by using a locally launched radiosonde. However, in many situations, this technique may also be inadequate if the sky conditions are rapidly changing. It is for these reasons that many researchers in the remote sensing community prefer to coincidently measure the downwelling illumination with the target reflectance and incorporate these additional measurements into their postprocessing routines to better compensate for temporal changes in illumination that occur throughout the course of a scan.^6,10,16

While the downwelling illumination measurements are useful for removing the temporal changes, these measurements do not compensate for the spatially dependent incoming illumination from nearby objects (i.e., adjacency effects). Specifically, when measuring an HCRF, it is necessary to understand the angular distribution of the incoming radiance. It is for this reason that the GRIT-T incorporates an upward-looking, dual-view capability, which allows for the measurement of the spatially dependent incoming illumination, similar to the dual FoV FIGOS goniometer.¹⁰

Both of the on-board spectrometers, one upward-looking and one downward-looking along the same optical axis, are Analytical Spectral Devices (ASD), Inc. FieldSpec 4 Hi-Res spectroradiometers.¹⁷ By using two spectrometers with the same make and model, we reduce the likelihood of interpolation errors when comparing one grid to another. These spectrometers are equipped with three detectors that cover the spectral range of 350 to 2500 nm: a silicon array for the visible and near-infrared (VNIR) spectral range of 350 to 1000 nm, a cooled InGaAs photodiode detector for the short-wave infrared (SWIR) range of 1001 to 1800 nm, and a cooled InGaAs photodiode detector for the SWIR spectral range of 1801 to 2500 nm. These systems are annually calibrated by the manufacturer to maintain both spectral and radiometric accuracy using National Institute of Standards and Technology (NIST)-traceable standards. In the spectral domain, the annual calibrations are certified to provide resolutions of 3 nm in the VNIR and 8 nm in the SWIR. In addition, these systems are able to achieve wavelength accuracies of 0.5 nm and wavelength reproducibility to within 0.1 nm.¹⁷ The annual radiometric calibration for the spectrometers results in an absolute radiance accuracy typically $\le 2.5\%$ between 655 and 2400 nm with uncertainty being greater at the ends of the wavelength range (typically $\le 3.6\%$ absolute accuracy at 350 nm). This calibration also indicates an uncertainty of 0.5% in reflectance from 350 to 2000 nm, with the uncertainty rising to 2% at 2400 nm. Each spectrometer can be fitted with multiple fore-optic attachments including: 1 deg, 3 deg, 5 deg, and 8 deg. The fiber optic with no fore-optic attachment has a 25-deg FoV.¹⁷ Normally, the 5-deg fore-optic attachment will be used on the downward-looking spectrometer, and an 8-deg fore-optic attachment will be used on the upward-looking spectrometer. This is done with the intent of monitoring a small, homogenous region on the ground while simultaneously capturing the sky conditions over a somewhat broader angular extent.

By designing the system so that the downward-looking spectrometer is aligned along the same optical axis as the upward-looking spectrometer, the GRIT-T measures the incoming radiance onto the target over the entire spherical grid while simultaneously characterizing the angular dependence of the scattered radiance from the surface. A third spectrometer is then used to ensure that any temporal changes in illumination are measured for compensation in the postprocessing routines; typically, this third spectrometer will monitor the reflected radiance from a ${\text{Spectralon}}^{\mathrm{TM}}$ calibration panel or monitor the downwelling radiance onto the surface by means of a cosine collector fore-optic attachment.¹⁶ This third spectrometer is located at a stationary location near the GRIT-T. Due to its lightweight frame design, the GRIT-T is able to add this dual-view capability without dramatically reducing the mobility of the system, and the software is able to integrate the third spectrometer into the overall control scheme since the control of all on-board devices operates on a separate internal network. Figure 3 shows the placement of the spectrometers along with the fiber optic cables and the design of the fore-optic pointing head. In addition to using a third spectrometer to monitor temporal changes in illumination conditions, we determined through laboratory studies that frequent measurements by the upward-looking and downward-looking spectrometers at the nadir orientation during the execution of a hemispherical scan pattern also can be used to monitor and compensate for variations in overall radiance levels that may occur over time. Using all of this information, we have developed postprocessing routines to ensure that all reflected radiance measurements have been corrected for variations in downwelling radiance levels over the course of the scan.

Fig. 3

(a) Image of GRIT-T with the two ASD fiber optic cables that are guided above the shield and along the pointing arm. Once guided, the two ASD fiber optic cables are inserted into the GRIT-T pointing head, either upward or downward, and two ASD fore-optics are attached. (b) The GRIT-T pointing head with the two ASD fore-optics and ASD fiber optic cables.

3.2.

Full Automation

One of the main limitations of the first-generation system GRIT¹ was that of operator fatigue resulting from the manual operation of the instrument in harsh environments. The device required a considerable amount of user input to operate with the significant potential for user error from lapses in operator attention. For this reason, the GRIT-T was designed to be fully autonomous during a scan. To perform a GRIT-T scan, the user sets the scan pattern in the graphical user interface (GUI) provided by the integrated GRIT-T software. Once the user starts the scan with the GUI, the system moves the sensor to the predefined positions on the spherical grid and saves the data from the spectrometers to an on-board computer. In addition to the data saved from the spectrometers, the system saves the corresponding time-stamped data from the other on-board devices. These on-board devices include the VectorNav VN-300 inertial measurement unit (IMU) and GPS, which provides the heading information; the NIST-Traceable Rieker Precision Inclinometer, which provides the leveling information; and the encoders, which provide the pointing information relative to the system frame of reference provided by the on-board VectorNav IMU/GPS system. As the scan progresses, the user can monitor the status of the scan, the health of subsystems, the real-time spectra from the spectrometers, and the positioning information from the on-board GPS-enabled IMU. The diagnostic information available on the field view tab of the GUI is shown in Fig. 4.

Fig. 4

Diagnostic information during measurement observed through the GUI of the goniometer system. The interface shows the spectra in digital counts for the upward- and downward-looking spectrometers. It also monitors the progress of the generated scan pattern, the power supply, and the GPS data provided by the IMU. Shown: field view tab, one of 8 command, control, and monitoring tabs available with the GUI.

The interface in Fig. 4 provides the user with a comprehensive status of the measurement/scan being performed. In this specific interface, the user has the ability to begin, pause, or end the measurement at anytime. It can monitor the progress of the generated scan pattern, the operation of the embedded computers, the roll and pitch of the frame, the power supply, and the health of the GPS data provided by the IMU. Furthermore, this interface allows the user to monitor the real-time spectra in digital counts from the upward- and downward-looking spectrometers, the time since last optimization and the battery voltage of the spectrometers. The “status” indicators change from green to red to warn the user if any errors are being sent from the spectrometers or the subsystems.

3.3.

Target-Plane Tracking

One of the key goals of the GRIT-T design was to ensure that the view angles requested by the user were achieved without effects due to parallax. Parallax error occurs when a goniometric system uses a measurement plane that is predefined and the measured scattering surface differs in height from the predefined measurement plane of the goniometer. To avoid this, the GRIT-T design implements “target-plane tracking,” which is defined to be a measurement that ensures that the center of the FoV on the target surface remains the same regardless of changes in view angle or sensor distance from the surface. In order to achieve this stringent goal, GRIT-T utilizes a pointing arm, a pointing head, and a laser range finder that is accurate to within $\pm 2\mathrm{mm}$ and repeatable to within $\pm 0.5\mathrm{mm}$. Before performing a scan, the user inputs the parameters of the scan pattern in the GUI. Then, the system is commanded to measure the distance to the target at nadir using the laser range finder. Using this information in combination with the known geometry of the instrument frame and components, the system calculates motor movements required for the pointing arm and the pointing head, which both rotate to achieve the minimum parallax error for the commanded view angle. As previously mentioned, the pointing arm was designed to move the sensor from $-70\mathrm{deg}$ to $+70\mathrm{deg}$ in zenith. The rotating sensor head was designed to achieve the ability to aim the sensor at the same spot in the center of the azimuthal ring for every possible sensor viewing orientation. Thus, the pointing arm performs the large zenith movements, and the pointing head performs the small counter-rotated zenith offsets that ensure accurate tracking of the same spot on the target. This technique ensures that the GRIT-T can achieve the requested view angles with respect to the target without the undesired effect of wandering spot errors during measurements of the surface radiance. In addition, based on feedback from the on-board Rieker Precision Inclinometer, GRIT-T levels itself using an embedded control algorithm, which automatically adjusts the three system actuators to ensure that a standard frame of reference, is achieved during scans. No matter what height the GRIT-T ring is above the target surface, the measurement plane is always adjusted to the target surface using the recorded height from the laser range finder when at nadir. GRIT-T then automatically calculates the rotations of the pointing arm and point head needed to ensure tracking of the same point on the ground, as described above.

An example illustrating the simultaneously rotating sensor pointing head and arm appears in Fig. 5. Here, the GRIT-T is operating in a field setting and measuring the radiance from a sediment surface at a 70-deg zenith angle with respect to the target. Tests performed in laboratory settings show that the final system achieves the goal of tracking the center point of the azimuthal ring to within 10 mm, even at the most extreme operating distances. During alignment of the GRIT-T pointing, we used nominal target distances of 632 and 732 mm obtained at nadir from the laser range finder, as these distances bound the range of most of the typical operating distances used in field and laboratory settings. To perform alignment, the downward-looking spectrometer and associated fore-optic were replaced with an optical alignment laser. The laser was optically aligned with a custom-designed holder, and the center of the ground instantaneous FoV was marked on leveled grid paper. The system then performed a hemispherical measurement sequence, in which the laser was positioned azimuthally in 45-deg intervals over a range from 0 deg to 360 deg. At each azimuthal orientation, the laser was oriented at zenith angles of 0 deg and 65 deg. At each of these orientations of the laser, the center of the optically aligned laser spot on the leveled grid paper was recorded for later analysis. Figure 6 shows examples of the results of performing these alignment tests in our laboratory.

Fig. 5

The GRIT-T as it measures a “true” 70-deg zenith angle with respect to the target. Here, the GRIT-T pointing head is counter-rotated to ensure that the same spot on the target is measured at each view angle position. The magnitude of this counter-rotation is determined for every zenith view angle.

Fig. 6

The results of performing alignment calibration operations on the GRIT-T system. These images show a coordinate system that is relative to the center of the GRIT-T C-shaped ring. Tests were done for scenarios where the sensor to target distance was (a) 632 mm and (b) 732 mm. The legend keys show the azimuth and zenith orientation of each recorded sensor orientation of the hemispherical scan pattern.

This figure shows that the center of the target of interest deviates by no more than a radius of $\sim 5\mathrm{mm}$ from the center of the target of interest at a height of 632 mm and no more than 10 mm at a height of 732 mm. These radii, the FoV of the spectrometers, which is determined by the choice of fore-optic, the height above the ground, and the view geometry on the hemisphere, determine the total area sampled by the spectrometer. The amount of offset in the alignment of the optical axis of the FoV with respect to the center of the target of interest can have a drastic impact on the total extent of the surface that will be included in the directional reflectance measurements. For example, if a 5–deg fore-optic attachment is used for the downward-looking spectrometer of the GRIT-T system and the distance from the target surface to the sensor at nadir is 732 mm, then it can be shown through basic trigonometry that the sampled area when the fore-optic is oriented at the nadir position is $3208{\mathrm{mm}}^2$, and the radius of this sampled area is 32 mm. At extreme zenith angles, the extent of the sampled region increases significantly due to the elongation of the major axis of the sampled, elliptical area by a cosine factor. For a sensor orientation at a zenith angle of 65 deg, the observed area is $7588.27{\mathrm{mm}}^2$, and the major axis extends 75 mm. During the course of a scan, this elliptically sampled area will be rotated azimuthally by 360 deg. This means that an annular region of the target surface will be represented in the directional reflectance measurements at extreme zenith angles that are not included in measurements taken close to the nadir orientation. Ideally, the target surface of interest should be considered homogenous so that this does not become an issue. However, geophysical parameters, such as roughness, grain size, and moisture content, can potentially vary even on local scales, meaning that the total extent of the sampled annular region covered at oblique angles, but not observed at nadir, should be kept as close as possible to the region observed at nadir. In the scenario mentioned, having an optical alignment that deviates by 50 mm, for example, from the desired center of interest, increases the extent of the sampled area at oblique angles by $\sim 180\%$ when compared to a perfectly aligned system. These types of deviation would be more likely for any system having a fixed measurement plane such as that of the earlier GRIT¹ goniometer system. The GRIT-T system, on the other hand, achieves only a 27% increase in sampled area at oblique angles at a deviation corresponding to the maximum uncertainty of the alignment when compared to a perfectly aligned system.

3.3.1.

Accuracy, mobility, and rigidity features

In order to create a system that was accurate, mobile, and field-portable, virtually all of the GRIT-T components had to be custom-designed. In some ways, since the system was intended to be used in austere coastal and desert environments, ensuring accurate movements became one of the most challenging tasks. The ability to track the same spot on the target within $\pm 10\mathrm{mm}$ at any reasonable height while measuring, for example, along the sloped surface of a sand dune, required absolute knowledge of every mechanical motion in the GRIT-T design. To obtain this knowledge, the GRIT-T utilizes an independent absolute encoder for every rotational motion on the system. These encoders ensure that the user knows where the system is positioned and pointed, at all times, including the error due to the mechanical play internal to the gearboxes. Furthermore, the GRIT-T utilizes sealed, ingress-protected components to increase reliability and aid in ensuring that even in the sandiest of environments, the accuracy of the mechanical movements would not be compromised.

While the absolute encoders work well to reference a particular movement to the frame, they do not provide overall knowledge of the system’s geolocation. To obtain geolocation, the GRIT-T utilizes a dual-antenna GPS-enabled IMU (VectorNav VN-300), and an industrial-grade inclinometer (NIST-Traceable Rieker Inclinometer). These two devices, when integrated, provide outstanding geolocation information, timing, roll, pitch, and reliable heading. While many earlier field-portable goniometers have leveraged GPS for geolocation and timing, the IMU utilized by the GRIT-T has dual GPS antennas that eliminate the need to obtain heading via a standard magnetometer. Instead, the IMU uses two precisely placed, high-gain GPS antennas at a well-known baseline: the GPS antennas sit atop the two actuators on each end of the GRIT-T azimuth ring. This ensures that the heading of GRIT-T is known to within 0.3 deg despite the presence of electronics and stepper motors nearby that would make magnetometer measurements unreliable. Thus, due to the on-board GPS/IMU, the GRIT-T provides the user with excellent knowledge of system geolocation as well as solar position. The GPS antennas (and their ground planes) can be seen in Fig. 1 on top of the actuators that are positioned at the two ends of the ring.

Also shown in Fig. 1 are the actuators that allow the GRIT-T to accommodate a wide variety of terrain. These three sealed devices are used to ensure that the system is level before a scan begins. Controlled by one of the embedded computers, these actuators are used to automatically level the system with the aid of the industrial-grade inclinometer. This allows the system to be leveled before the start of a scan and to report the roll and pitch to within $\pm 0.1\text{degree}$ throughout the course of the scan; this information is included in the output accompanying each spectral measurement from the two on-board spectrometers, along with the current state of all other on-board devices. Since the roll and pitch are monitored throughout the course of the scan, the GRIT-T postprocessing routines can incorporate information regarding the leveling of the system. With each movement of the system, the GRIT-T provides the user with the “true” view angle based on the actual position of the sensor reported by the absolute encoders, in addition to the commanded sensor orientations, both of which are reported to the user in the GUI. This automated system leveling, combined with the target-plane tracking, allows the system to measure the same spot of the target on a sloped surface, without compromising angular accuracy, as shown in Fig. 7.

Fig. 7

The GRIT-T as it measures the radiance reflected from a sloped sand surface. Note that the goniometric system is leveled and still able to track the same spot on the target. Also note that, on a sloped surface, the maximum zenith angle that can be reached is reduced, since the distance from the sensor to the target is enlarged.

A key factor in achieving mobility for this type of system is ensuring that it is lightweight enough to be easily carried by our team members. The GRIT-T was designed to be carried by just two people under normal circumstances on level terrain and by just three people under unusual circumstances on uneven terrain. To ensure a design that was lightweight, yet rigid enough to provide high accuracy, many of the GRIT-T components were refined using finite-element analysis and then custom-machined from aluminum. Due to the C-shaped design, the ring was one of the most challenging components to design and manufacture and was ultimately created from three separate aluminum slabs. This allowed for a structural layer, a geared layer, and a guide rail layer that were aligned with guide pins and secured using M12 bolts. Even though the ring required a large, solid cross section (roughly $60\mathrm{mm}\times 50\mathrm{mm}$) to reduce torsional rotation when the system was elevated, the aluminum construction ensured that the ring remained reasonably lightweight (19 kg). In order to allow for use in coastal environments with salt spray, the GRIT-T ring was black anodized and then painted with Krylon 1602 ultraflat black paint to reduce secondary scatter from the instrument frame. The components that contributed to the alignment of the carriage, i.e., the guide rail and the associated bearings, were either black anodized for wear protection or covered to minimize optical impact. All of the other components on the GRIT-T, with the exception of the spectrometers, are rated to at least the IP 54 standard for ingress of dust or water. The IP 54 standard indicates that equipment can operate in environments where water may be splashed on the component and ingress of dust is limited to such a degree that it does not limit operation of the instrument over multiple hours.¹⁸

3.4.

Minimized Obscuration

An important feature incorporated into the GRIT-T design was the minimization of obscuration due to self-shading. At small phase angles, i.e., at angles where the sun or other light source is behind the sensor, most systems cast a large, unwanted shadow onto the target plane due to the sensor carriage, and a lot of valuable information from the HCRF or BCRF is lost. In order to measure as much of the HCRF or BCRF as possible, the cross section of the sensor head needs to be minimized. Therefore, the design of the GRIT-T sensor carriage utilized a narrow pointing arm and a narrow pointing head that only cast shadows onto the target plane when the system is measuring close to the opposition direction. The sizes of these components were minimized while still including motors, gearboxes, and a laser range finder required to enable other features. An image of the pointing head can be seen in Fig. 3. To reduce shadows from overhead structures, the GRIT-T was designed with a C-shaped (i.e., 220 deg) ring that required no other structure to ensure rigidity. Furthermore, the ring was positioned low to the ground to ensure that it did not obscure the downwelling skylight. The remaining structure was then designed to be as short as possible while still meeting other requirements. The only other notable obscurations are from the spectrometers, the two leveling actuators at the ends of the ring, and the pointing arm control hardware. Using these design features, the open side of the ring is intended to be aimed toward the Sun, or, in a laboratory setting, the illumination source. This ensures that the number of points on the spherical grid that are impacted by self-shading from the direct illumination depend only on the width of the pointing head. With this in mind, the width of the pointing head was minimized to 30 mm. Typically, this allows for clean measurements in the solar principal plane for all points on the spherical grid except those with phase angles $\le 5\mathrm{deg}$.

3.5.

Digital Elevation Model

As a result of the GRIT-T’s ability to simultaneously rotate its pointing head and pointing arm, the system is also capable of maintaining a nadir view of the measurement plane throughout the entire hemisphere of its range of motion. By combining this capability with the use of the laser range finder on-board the system, the GRIT-T has a sampling mode to produce a DEM of the region surrounding the target. The effort to develop a DEM was motivated by the fact that surface topography on the scale of millimeters plays a significant role in the interaction of light with natural surfaces.¹⁹ Figure 8 shows an example of a DEM that was measured using the GRIT-T. Here, the sediment was collected during a field experiment in the Nevada desert and dried prior to being used in laboratory measurements. These data are then postprocessed and output as a point-cloud, and metrics of roughness, such as mean slope angle, random roughness, and sill variance, are calculated for use in photometric modeling.

Fig. 8

(a) Sediment collected in the Nevada desert and (b) a DEM of the sediment using the GRIT-T. The DEM has units of centimeters along the $x$-, $y$-, and $z$-axes. Note that the rocky appearance of the sediment surface is captured, allowing metrics of roughness to be calculated for analysis.