1. Introduction
Optical remote sensing has emerged as a powerful tool for monitoring water quality on a global scale [
1,
2,
3] and for achieving a better understanding of various aquatic ecosystems. It uses data obtained from space- or air-borne sensors to obtain near real-time information regarding the constituents of the water column across vast and potentially inaccessible areas.
Water colour remote sensing relies on the interaction between light and water constituents, typically in the visible (400–700 nm) and near-IR (700–1000 nm) parts of the spectrum. When sunlight penetrates a water body, its spectral properties are modified based on the absorption and scattering characteristics of water, depending on its dissolved and particulate constituents and their concentrations. Part of this altered light is scattered back to the surface and potentially captured by optical sensors and is called the water-leaving radiance (L
w, in W m
−2 sr
−1 nm
−1). For comparison with satellite imagery, researchers mostly use the remote sensing reflectance (R
rs, in sr
−1), defined as the ratio of L
w to the downwelling irradiance at the water surface (E
s, in W m
−2 nm
−1). R
rs contains information about the water body and its constituents, making it a critical parameter for assessing water quality and other aquatic characteristics. For instance, it is used to derive chlorophyll-a and CDOM concentrations or turbidity, but also for detecting submerged macrophytes (e.g., seagrass and kelp beds) or bottom depths in shallow waters [
4,
5,
6].
In situ radiometric measurements, if they are valuable as such for conducting environmental studies [
7], are also key to the development and validation of water colour algorithms by linking space-based observations to ground-based knowledge. High-quality in situ measurements are also required for the system’s vicarious calibration of spaceborne sensors. For instance, a community work [
8] recommends that in situ reference measurements in oligotrophic and mesotrophic areas should reach an uncertainty lower than 3–4% for the L
w in the blue–green spectral regions and 5% in the red.
Despite the importance of Rrs measurements, too little data have been collected in the field. Most of the difficulty arises from the costs of materials and campaigns. Also, the technicity required to manipulate radiometric devices does not guarantee the success of a campaign and the quality of the measured data.
Various methods exist for measuring R
rs in the field [
9,
10,
11], but the Skylight-Blocked Approach (SBA) [
12,
13,
14] is noteworthy for its ability to directly measure L
w instead of deriving it from the upwelling radiance. This protocol allows high-quality measurements, even in complex environments (e.g., in stratified or shallow waters, or in the presence of a scattered cloud cover). It consists of measuring the upwelling radiance close to the water surface using a nadir-view radiometer equipped with a shield screening skylight at its open end, thus preventing sky- and sun-glint from entering the sensor FOV. The signal recorded by the radiance sensor represents the actual radiance emerging underneath the water surface, L
w. It is expected not to be polluted by the signal coming from the light reflecting at the water surface, such as for a standard “above-water” protocol [
15]. Concomitant and collocated measurements of E
s allow for the derivation of the R
rs. Manually operated SBA methods (used in [
16,
17,
18]) require constant attention to maintain the radiance sensor immersed at the right depth and in the correct position with respect to the sun and deployment platform during the entire acquisition sequence. If standard protocols [
19] recommend long-lasting sequences of acquisition (typically 5–10 min), rough water conditions often make this exercise challenging and wearing and fully monopolises an operator dedicated to this measurement. Knowing that, a few authors [
14,
20,
21] have designed floating systems implementing the SBA method, herein called “radiometric buoys”. These radiometric buoys allow for easier deployment and enable measurements to be taken at a reasonable distance from the research vessel (typically 10–20 m), in a location where the aquatic light field will not be affected by the presence of the vessel. However, optimal geometric conditions, which are crucial for obtaining accurate SBA measurements, are not necessarily ensured. For instance, radiometric buoys can rotate freely under the effect of winds or waves, causing the radiance sensor to measure the L
w of a water volume in the direct sun shadow of the buoy, thus biasing L
w measurements [
22]. These floating devices are also subject to wave-induced tilt, which moves the sensors away from the nadir view. Changes in the sun-relative zenith angle of the irradiance sensor (i.e., the difference between the sun zenith angle and the zenith angle of the radiometer) can cause a large bias, particularly if the radiometer is tilted towards (or away from) the sun [
23]. The tilt issue is typically addressed by measuring the tilt of the instruments with onboard inclinometers and later discarding impaired radiometric measurements in the post-processing step. However, there is currently no measuring system that ensures the correct sun-relative azimuth of the radiance sensor throughout the entire acquisition sequence. It relies on intensive manipulation control during the whole acquisition (for handheld SBA), or on manual quality control during the post-processing step (for all SBA implementations), impeding the assurance of de facto accurate and consistent measurements in the field.
The present study aims to offer a novel solution, called Lake SkyWater (LSW), to reliably and easily acquire Rrs in lakes. Its qualities, reasonable cost, and ease of construction would enable the water colour community to better and more easily document the optical properties of water worldwide. In this article, we describe the LSW design, which implements the SBA scheme and ensures that the measurements are performed under optimal geometric conditions, evaluate its performance in the field, and discuss its advantages.
2. Lake SkyWater Design
2.1. General Idea and Naming Conventions
Radiometric buoys need to be easy to handle and transport, sturdy enough to hold the sensors, and small enough to minimise the impact on the ambient light field. We chose to keep the classic design of SBA buoys [
19] consisting of two radiometers in the nadir view on opposite sides of a float: one (the irradiance sensor) looks upward and measures E
s, and the other (the radiance sensor) faces downward and measures L
w.
The key addition of LSW is its motorised solution, which ensures that the radiance sensor is placed in optimal geometrical conditions with respect to the sun and buoy during the entire measurement sequence. For this purpose, LSW is composed of four sections (
Figure 1) detailed hereafter: the structure (in white), the rotating system (in orange), the radiometric system (in pink), and the embedded system (inside the green container). We designed the custom parts using Onshape and printed them using the Netpune 4 entry-level 3D printer (Elegoo, 518109 Shenzhen, China). These 3D models as well as the code for controlling the device are freely available on GitHub at
https://github.com/inrae/Lake-SkyWater (v1.0; accessed on 24 February 2025).
To facilitate the understanding of the results, we propose a naming convention for all the variables that we use hereafter.
Let X, Y, and Z be a referential centred on the LSW centroid, whose three axes are fixed: X points eastwards, Y northwards, and Z is orthogonal to the geoid (i.e., the water surface) and points upwards.
Let X’, Y’, and Z’ be a referential centred on LSW centroid, whose three axes rotate with the two arms to which the sensors are attached (X’ points in the direction of the radiance sensor).
When LSW is at rest (i.e., without any tilt): , with .
In the presence of waves, LSW rotates from its reference position such that , where R is a general 3D rotation matrix. By converting Cartesian coordinates to spherical coordinates, we obtain the following:
, the radial distance of ẑ′ in the (X, Y, Z) referential;
, the azimuth of ẑ’ in the (X, Y, Z) referential, defined on the interval [0°, 360°). The sun-relative azimuth angle of the tilt θt (defined on (−180°, 180°]) is then derived: , where SAA is the solar azimuth angle (defined on [0°, 360°));
, the polar angle of ẑ in the (X, Y, Z) referential, i.e., the tilt “amplitude”, defined on [0°, 180°].
The final angle of importance is θLw, the sun-relative azimuth angle of the radiance sensor. When LSW rotates (either because of waves or its rotating system), , where R is a general 3D rotation matrix. In the same way as for the tilt, we obtain , and thus, .
These three angles, along with their names and value ranges, are listed in
Table 1.
2.2. The Overall Structure
The structure of LSW (see
Figure 1) is composed of a float, a main frame, a small (3.6 L) waterproof container, and 3D printed parts. The float is a front tractor inner tube (major diameter of 50 cm and minor diameter of 14 cm). The frame is made of black anodised aluminium profiles (Systéal, 77185 Lognes, France). Black anodising was applied to make it more resistant and less reflective in water owing to its matte finish. It is composed of two parts connected by a slewing ring (see
Section 2.2). The first axis is fixed and attached to the floating body with 3D printed clamps, while the second (consisting of two 74 cm arms to which the radiometers are attached) can rotate, allowing the correct orientation of the radiometer measuring L
w. The waterproof container is used to safely store all the electronics: the embedded system and power supply (see
Section 2.4). It is attached to the rotating part or the structure using 3D printed clips, in the centre of the float. All 3D printed parts (clamps, clips, wedges, etc.) were made of PETG, which is ideal for functional 3D printing because it is UV- and water-resistant, stronger than PLA, and easier to print than ABS. These custom elements allow LSW to be adaptable to different hardware (sensors, floats, etc.) and protocols (e.g., by modifying the distance between the radiometers and the water surface).
2.3. The Rotating System
A rotating system has been added compared to the classic design of SBA radiometric buoys [
19], enabling the radiometer measuring L
w to always be placed under optimal geometrical conditions, i.e., facing the sun so that it does not measure water in the direct shadow of the structure. The rotation is made possible by the use of a slewing ring driven by a stepper motor. We chose the Iglide
® PRT-04 slewing ring (igus GmbH, 51147 Köln, Germany), an on-the-shelf product with high performance, robust design, and moderate price. The motor (STEPPERONLINE, 211100 Nanjing, China) is a NEMA 17 stepper equipped with a high-precision planetary gearbox. This low-cost option has good accuracy (maximal backlash of 50 arc minutes) and provides a significant torque (maximal permissible torque of 10 N m), which is required for easily rotating the moving chassis and the attached radiometers.
2.4. The Radiometric System
Two synchronised radiometers are required to implement the SBA scheme. One for measuring L
w and the other to measure the surface irradiance E
s. A cone- or cylinder-shaped apparatus attached to the radiance sensor (hereinafter referred to as “screen”) is also needed to block sky- and sun-light reflection at the water surface, allowing the direct measurement of L
w. In this study, we used two RAMSES G2 VIS hyperspectral radiometers (TriOS Mess- und Datentechnik GmbH, 26180 Rastede, Germany), which are widely used in the water colour community [
24]. These two sensors operate in the 320–950 nm wavelength range, with a spectral sampling of ~3.3 nm and an adjustable integration time of 4 ms to 8 s. The radiance sensor has an FOV of 7°. They are lightweight (<1 kg) and have a low power consumption (~1 W), making them suitable for portable and autonomous use. Similar to the other 3D printed parts, the conical screen was made of matte black PETG. It measures ~9.5 cm in height and 5.8 cm in width and can be screwed to the radiometer using a custom-made adapter (collar). In this study, the immersion depth of the screen was ~3.7 cm. We calculated (considering the dimensions of the device) that this depth allowed the open end of the screen to remain below the water surface as long as the buoy was not tilted by more than 5°. The smaller the screen and its immersion depth, the better it is for self-shading, i.e., the contamination caused by the shadow of the radiance sensor and its attached screen in water (see [
25] for its evaluation).
2.5. The Electronic System
All electronics, i.e., the embedded system and power supply, are safely housed in a waterproof container placed in the centre of the buoy.
The embedded system is made of open-source components, and is composed of a Raspberry Pi 3 Model B+ (RPi; Raspberry Pi Ltd, CB4 0AB Cambridge, UK) and Tinkerforge modules (Tinkerforge GmbH, 33758 Schloß Holte-Stukenbrock, Germany) for I/O and add-on sensors. The RPi serves as a local controller and data logger. It controls the radiometers and the stepper motor and allows for wireless communication with the end user’s device (e.g., a laptop, a smartphone, etc.) via a self-hosted Wi-Fi network. All measured data are stored on the same SD card as the operating system (64-bit Raspberry Pi OS). Tinkerforge is a flexible and affordable system of building blocks that is typically used for prototypes and small batch production. These blocks can be divided into two categories: Bricks and Bricklets. Bricks are single-task stackable 44 cm PCBs. Bricklets can be connected to Bricks and allow for the addition of sensors (here, an inertial measurement unit (IMU) and a GNSS receiver) and IO interfaces (here, RS485/Modbus RTU for communicating with the two radiometers and the Silent Stepper Brick for controlling LSW stepper motor) to the system. The IMU + GNSS receiver combination is required for positioning the radiance sensor (see
Section 2.5), as well as for monitoring the location of the buoy. The geolocation of LSW is provided by the Tinkerforge GPS Bricklet 2.0. It is used to record a time series of latitude, longitude, and altitude at each sampling station, eventually monitoring the drift of the buoy during a measurement sequence, and to set the time of the system, avoiding a battery-backed real-time clock. This Bricklet is equipped with a FireFly X1 (GlobalTop Technology Inc., 741 Tainan, Taiwan), which is a compact low-powered multi GNSS module that can provide a positioning accuracy of up to 1.8 m CEP (circular error probability), and is connected by a U.FL connector to an external GPS antenna attached on top of the waterproof container. The absolute orientation of LSW is provided at any time by the Tinkerforge IMU Bricklet 3.0. It is used in combination with the location of the buoy and the current time to compute the angles illustrated in
Figure 1: θ
Lw, θ
t, and φ
t.
The power source of LSW—a 24V 2Ah NiMh battery—is connected to (i) the Tinkerforge Step-Down Power Supply, powering the stepper motor through the Silent Stepper Brick, (ii) the Tinkerforge HAT Brick, powering the RPi and all the Bricklets, and (iii) the two radiometers. It allows <20 h of measurements or ~1 h of usage of the stepper motor “at full speed” (see
Table S1 for the list of all the electronics and the power consumption of each item). Given that we need ~5 min per sampling station, i.e., ~75 R
rs spectra, this means we can measure up to 10 stations per battery.
2.6. The Acquisition Protocol and Output Data
The standard scenario used for field measurements includes the following stages: (i) switching on LSW, (ii) connecting to it through SSH, and (iii) starting the acquisition sequence using the provided command line interface (CLI). At the end of the measurements, the raw spectra are calibrated (i.e., corrected for the dark signal and non-linear response, using the relevant calibration coefficients), and summary graphs are generated. These can be used to check the data quickly. The measured data (
Table 2) can then be downloaded for further processing and visualisation.
When the acquisition sequence starts, two scripts are launched in parallel; the first controls the rotating system (see
Section 2.2) and the other controls the radiometric system (see
Section 2.3).
The pseudo-code of the orientation algorithm is provided below. Please, confer to
Section 2.1 and
Figure 1 for all the above-mentioned angles and their definition.
The script controlling the two radiometers allows for the acquisition of a given number of Lw and Es spectra. Both radiometers are triggered synchronously.
The logged data are summarised in
Table 2. The sun-relative azimuth of the radiance sensor θ
Lw and LSW tilt (both its amplitude φ
t and direction θ
t) are not saved as they can easily be derived from the absolute orientation of the buoy r, its location (latitude and longitude), and the time
1–2.
5. Discussion
5.1. LSW Radiometric Performances
The results of our experiment in Bimont reservoir demonstrated a high degree of concordance in Rrs measurements between LSW and the handheld SBA method, with a coefficient of determination > 0.99, a general accuracy (MdSA) of ~3.21%, and a bias (SSPB) of ~3.04%. These findings support the conclusion that LSW represents an effective implementation of the SBA protocol and constitutes a suitable alternative to handheld SBA measurements, offering enhanced autonomy and user-friendliness (as discussed in the following sections).
The R
rs measurement accuracy remained below the 5% threshold for the visible part of the spectrum (
Figure 8), as intended to meet the highest level of standard requirements for water colour validation practices [
34,
35,
36]. The lower accuracy at both ends of the spectrum was attributable to the very low L
w values of these oligotrophic waters.
Similarly, the dispersion was comparable for the two methods: below 4% in the visible portion of the spectrum (400–700 nm)—and even <3% between 400 and 625 nm—and up to 10% for wavelengths in the range of 700–800 nm. This is consistent with or even slightly better than previously published SBA measurements. For instance, Lee et al. [
14] obtained in Lake Michigan and Green Bay a dispersion between 3% and 5% in the 350–600 nm wavelength range, 8% below 700 nm, and >15% for wavelengths greater than 780 nm. In Honghu Lake, Tian et al. [
20] reported CVs < 4.5% in the visible spectrum and up to 13% at 750 nm. Finally, measurements carried out by Zibordi and Talone in the Western Black Sea [
37] had a mean CV < 7% between 400 and 700 nm—and even <4% for wavelengths shorter than 600 nm.
Moreover, the minor discrepancies observed between the R
rs spectra acquired by the two systems can be explained by several factors. First, the radiometric calibration of the two radiance sensors was not conducted simultaneously, and despite their close correspondence, it is not perfect (
Figure S1). The potential drift in the calibration coefficients of one or both sensors could contribute to the observed difference in the measurements, which is of a similar magnitude. Second, measurements were not taken at the exact same location, with handheld measurements being conducted closer to our boat, potentially introducing perturbations to the underwater light field.
5.2. Optimisation of the Viewing Geometry: The Motorised Rotating System
As demonstrated above, LSW has significant advantages for the accurate measurement of R
rs. Using LSW, the duration of individual measurements can be significantly shortened. Indeed, as seen in
Section 4.1 and discussed here, the rotating system of LSW allows the downward-looking radiometer measuring L
w to be successfully maintained in the direction of the sun throughout the acquisition sequence (the sun-relative azimuth distribution is shown in
Figure S2 and
Table S2). Consequently, we obtain more valid spectra (i.e., spectra acquired under optimal geometric conditions) in less time than using a standard non-motorised radiometric buoy as in [
19]. The importance of not measuring a water volume in the shadow of the deployment structure is well illustrated in the literature [
22], but also in this experiment. We can use the incident at the 1st station to illustrate this: the mean R
rs at 540 nm in shadow is approximately ~15% lower than when it is measured towards the sun (
Figure 3). Therefore, it is important to determine the orientation of the radiance sensor for each measurement. This reinforces the pertinence of using an IMU onboard LSW to record essential metadata throughout the acquisition sequence: the absolute orientation of the radiometers with respect to the sun and its location. The latter facilitates the monitoring of potential buoy drift during measurements, which is also useful for selecting the appropriate pixel in satellite matchups if an image is simultaneously taken from space. The sun-relative azimuth of the radiance sensor θ
Lw can be derived from the absolute orientation of the buoy and is used to control the stepper motor (see
Section 2.5). LSW absolute orientation is also used to derive LSW tilt: both its amplitude φ
t and its direction θ
t. These three geometric values are crucial for the post-processing steps (or quality control) to eliminate or correct data that have not been acquired in optimal geometric conditions (e.g., due to waves or during the initialisation phase before its stabilisation in the optimal configuration; see
Figure 6). The L
w spectra measured in the direct shadow of the buoy can be readily filtered using θ
Lw. They can also be filtered using a threshold on φ
t, as recommended in [
19] (
) and applied in [
37] (
). Finally, θ
t is important because a given φ
t will not have the same impact on the measured E
s if the buoy is tilted towards the sun or in the opposite direction. However, LSW does not yet offer a means of actively correcting its tilt as it does for the radiance sensor azimuth.
Given the importance of the IMU in acquiring this geometric information, the selection of a robust chip is highly recommended. The brief incident that occurred during the 1st acquisition sequence (as mentioned briefly in
Section 4.1 and illustrated in
Figure 3) highlighted that. The IMU onboard LSW (specifically its magnetometer) experienced an anomaly, manifesting as a ramp in its north reference. This resulted in the rotating system compensating for a non-existent rotation, consequently orienting the radiance sensor in a direction nearly opposite to that of the sun. Subsequent investigations revealed that this malfunction was due to the self-calibration feature of this particular IMU (BNO055 sensor; Bosh Sensortec GmbH, 72770 Reutlingen, Germany), which cannot be deactivated. While this does not invalidate the design of our prototype, it is advisable for future iterations of LSW to utilise alternative chips (e.g., BNO085/ICM-20948/LSM9DS1 or ISM330DHCX/LSM6DSOX/LSM6DS3TR-C + LIS3MDL).
As observed in
Section 4.1 and
Figure S2, the orientation control of LSW functions correctly with only basic instructions (see
Section 2.5). The sun-relative azimuth angles of the radiance sensor at the time of acquisitions are centred at −1.34°, with Q1 and Q3 at −21.86° and 16.44°, respectively. The numbers would further improve with the exclusion of the 2nd station, where we were slowly moving to compensate for the drift we had during the 1st acquisition (see
Figure 2), resulting in LSW being towed behind the boat. The performance of the orientation control system could be improved by fine-tuning the stepper motor parameters (the aimed velocity and its acceleration/deceleration). Its stability could be improved by refining the orientation control script, for instance by implementing a Kalman filter to predict the future orientation of the system. Such improvements should reduce the amplitude of the oscillation of θ
Lw around 0 (see
Figure 3,
Figure 4,
Figure 5 and
Figure 6a). Note that part of the θ
Lw dispersion also stems from the occasional rotation of the base (i.e., the float and the bottom half of the structure that is attached to it) instead of the payload. This phenomenon is attributable to multiple factors: (i) the circular shape of the float, (ii) the minimal fluid friction on the float, and (iii) the relatively low weight of the base relative to the payload. We attempted to address this issue by attaching a small (~29 × 21 cm
2) PVC clippable fin to the device. While this modification mitigated the problem, the dimensions of the centreboard and its immersion depth of approximately 10 cm were insufficient.
Finally, because of the geometrical information collected by LSW during radiometric acquisitions, we observed that only retaining measurements performed when the sun-relative azimuth of the radiance sensor is <±20° (as carried out in [
37]) may be excessively restrictive. Indeed, the spectra measured at a greater sun-relative azimuth during our experiment in Bimont reservoir did not appear to be affected by the shadow of the LSW structure. This is consistent with the modelling results presented in [
22], which indicate that the azimuth effect was negligible when the heading of the SBA system was within ±120° of the ideal direction determined by the sun azimuth.
5.3. LSW Structure Advantages
In addition to recording ancillary geometric data that are useful for quality control, LSW addresses the main sources of uncertainty (other than self-shading) in field measurements via SBA. First, as mentioned in the previous section, its orientation system prevents the acquisition of L
w spectra in the direct shadow of the structure, which can occur with more conventional designs (as in [
14,
37]). This also allows LSW to be less susceptible to the effects of high solar zenith angles than alternative designs such as the radiometric buoy FOBY [
20], as the radiance sensor is positioned between the float and the sun, rather than in the centre of the float (and thus potentially in its direct shadow). Second, the utilisation of an airtight screen significantly reduces the likelihood of fore-optics contamination, which occurs when the lens of the radiance sensor is submerged owing to wave action. Finally, the risk of the open end of the screen rising above the water surface is mitigated (i) by the carefully dimensioned attached screen and (ii) by the easy adjustment of the immersion depth of the screen and/or of the height of the radiance sensor, facilitated by the fixation clamp that secures it. In this study, the open end of the black screen was inserted below the water surface by 3.7 cm. Refs. [
14,
20] utilised depths of 5 cm and 2 cm below the surface, respectively. We did not characterise how the immersion depth of the screen affected SBA measurements, but since we employed the same radiometers (size and FOV) and screen for the two SBA protocols, it should impact them in a similar manner. Further experimentation would be necessary to quantitatively investigate the effect of the immersion depth of the screen on R
rs from SBA measurements under varying wind and wave conditions and water types.
Another important consideration was to ensure the user-friendliness of our device, which resulted in LSW being an easy-to-use solution for obtaining high-quality Rrs spectra. First, it allows for a significantly easier deployment compared with handheld SBA. Second, it is fully controllable via a smartphone over Wi-Fi using a straightforward yet powerful Python CLI, eliminating the need for a rugged laptop and obviating the use of long and often tangled cables. Furthermore, it does not require constant attention, and data are automatically acquired, enabling users to focus entirely on other tasks or measurements (e.g., water samplings or measurements of aerosol optical depth), which can represent a substantial time-saving and quality of life improvement. Additionally, in the case of the handheld SBA, maintaining the radiometer at the chosen immersion depth during the entire acquisition sequence can be challenging. Wave action or the movement of on-board operators can induce vessel tilt. Operator fatigue can also impact performance, particularly when utilising a sufficiently long boom to minimise interference from the boat. Furthermore, the tilt of the vertical pole (and, therefore, of the irradiance sensor) was not recorded and probably differed slightly from that on the hanging radiance sensor. While this can be an issue when collecting “real-world data”, it should not impact this comparison because we used the same Es spectra/source for the two methods.
Due to all these ameliorations, high levels of accuracy were achieved with considerably less effort by simply deploying LSW onto the water surface, whereas we had to manoeuvre the boat throughout the measurement sequences to ensure that the radiance sensor used for the handheld SBA remained on the right side of the vessel.
In anticipation of its use for various field campaigns (including in remote locations), LSW has been made easy to transport. It is suitable for commercial air travel as it can be entirely disassembled, comprises small components, and does not contain lithium-based batteries. It also features a rapid separating mechanism that facilitates transport in the field, without requiring complete disassembly, and can be swiftly separated into two compact sections (1: the float and the bottom half of the structure and 2: the payload) using only four screws.
Furthermore, LSW is cost-effective, adaptable, and readily implementable. First, excluding the two radiometers, our device with its motorised orientation system costs just a few hundred euros. It is also significantly more economical than incorporating a second radiance sensor into a standard radiometric buoy, as proposed in [
9], to avoid the measurement of L
w in the shadow of the device. Second, all the electronics, the stepper motor, slewing ring, waterproof container, etc., are open-source and/or are commercially available components. Additionally, the other parts are inexpensive as they are 3D-printed using affordable material and allow for adaptability to already owned sensors (provided they use Modbus RTU, a de facto standard transmission protocol) and any available components at the time of construction (for instance, any tube with sufficient buoyancy will suffice as a float). The height of the payload and of the radiometers is also freely adjustable using infinitely stackable wedges and/or the clamps that secure the sensors. It enables the use of different radiometers and/or protocols (e.g., above-water/contactless SBA; [
38]), as well as the adjustment of the immersion depth of the screen. Finally, the structure comprises aluminium profiles—rendering it lightweight and robust—which are widely available and easily cut to the required dimensions. Assembly is achieved using widespread aluminium M4/M5 screws and nuts.
5.4. LSW Application in Ocean and Coastal Environments?
Despite its name, the use of LSW in coastal waters can be considered. Radiometric buoys implementing SBA have already been successfully used in coastal waters and open oceans [
37,
39]. However, operating such a device in a marine environment might be more complicated than in freshwater systems. First, the presence of larger waves would likely necessitate substantial data filtering during post-processing. It would also require the selection of a stepper motor with more torque (which implies larger batteries). Second, operating LSW in seawater would necessitate the use of a stepper suitable for saltwater submersion, which would slightly increase the cost of our device. We also expect an increase in both pitting and galvanic corrosion. Although the aluminium structure of LSW is anodised, and the two stainless-steel radiometers are already covered with rubber, minor galvanic corrosion may potentially occur between the aluminium screws and the stainless-steel clamps that maintain the radiometers. However, a protective coating can be applied, or, alternatively, the clamps can be made of plastic.
6. Conclusions and Perspectives
Our work is part of a broader effort by the international community to provide tools to enrich and catalogue the collection of accurate in situ measurements of water reflectance worldwide, thereby improving our knowledge of coastal and inland waters. This effort is reflected in the creation and maintenance of measurement networks such as AERONET-OC [
34,
40] and WATERHYPERNET [
41], initiatives like SeaBASS [
42], OC-CCI [
43], and GLORIA [
44], as well as the development of portable sensing solutions, which are necessary for studying remote or inaccessible water bodies. In line with the latter, LSW is a promising tool for obtaining high-quality water-leaving radiance measurements in lakes or any stagnant water bodies, offering a balance of accuracy, usability, and affordability that could benefit both research and monitoring applications in aquatic remote sensing. Our device addresses key challenges in “on-water” field radiometry through its motorised rotating system, which maintains the radiance sensor under optimal geometrical conditions (i.e., facing the sun). A comparison with handheld SBA measurements demonstrated excellent agreement, with a coefficient of determination >0.99 and a general accuracy (MdSA) of ~2.41% for remote sensing reflectance (R
rs) in the visible spectrum (400–700 nm). LSW offers several advantages over existing SBA implementations:
A shorter acquisition sequence due to automated sensor positioning (or a longer acquisition time with excellent stability);
A reduced risk of fore-optics contamination;
An improved ease of use and deployment;
Cost-effectiveness and an adaptability to various sensors.
The system’s design allows for easy transport, rapid (dis)assembly, and straightforward operation via smartphone control. Its modular construction, using widely available components and custom 3D-printed parts, facilitates customisation and self-made repairs. While the prototype performed well overall, areas for potential improvement were identified, including refinement of the orientation control algorithm and addressing the occasional rotation of the float. Future studies should focus on optimising these aspects and conducting more extensive field trials across diverse water bodies and environmental conditions. Attaching both a full-sky and an above-water imaging camera is also under consideration. It would help to better characterise cloud distribution and hypothetical suboptimal measurement conditions, as well as observing what is being measured.