Open Access
Add to BibSonomy
Abstract
Sea fog represents a significant risk for safe navigation of sea vessels. Visibility LiDAR systems might offer a striking way to reduce the risks associated with sea fog, but they should be appropriately designed to provide a proper level of detection for reliable forewarning of sea fog. Here we analyze the performances of a visibility LiDAR system with the aim of achieving optimal detection operation. A series of echo signals are simulated under different visibility conditions addressing the influence of the various hardware parameters on the final system performances and defining an optimal visibility LiDAR configuration. Using the optimized parameters, a visibility LiDAR system was realized and tested in a field campaign on Hengsha Island (Shanghai). The experimental findings obtained by the visibility LiDAR are compared with results of a forward scattering visibility meter showing good consistency in homogeneous atmosphere, while even superior performances are observed for inhomogeneous atmospheric conditions. Our experimental results indicate that an optimized visibility LiDAR can provide an early warning for light fog located at a distance of 5 km, i.e. about 3.5 hours in advance to the spreading of the fog to the shore. These findings demonstrate the good performances of the visibility LiDAR developed in the present study in performing visibility measurements and its capability of providing sea-fog warning.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Sea fog is one of the most disastrous weather event with a significant impact on the safety of ships navigation, making a reliable and effective monitoring of sea fog of critical practical importance for navigation safety purposes [1–5]. With the development of new technology, a variety of atmospheric detection equipment has been continuously applied to sea-fog detection, such as buoy stations, seaside observation towers, satellite remote sensing, millimeter wave radar, LiDAR, and so forth. The occurrence of sea fog is random, and equipment commonly used for forward scattering visibility measurements can only obtain data over a limited scanning range, which makes them not the optimal choice to provide forewarning observations in highly changeable sea-fog conditions [6,7]. Satellite remote sensing can monitor sea-fog, but cloudy weather conditions significantly limit its effectiveness in sea fog detection under large cloud coverage [8,9]. Millimeter wave radars can effectively capture three-dimensional (3D) spatial structure of sea-fog and its variation, with the further advantage of long detection distances, but they suffer from large near-field blindness and low spatial resolution [10–12]. In terms of detecting sea fog, LiDAR can offer the advantage of a small blind area and a high spatial resolution. Atmospheric LiDAR technique is generally applied to characterize atmospheric aerosols and monitor their spatial and temporal variation. Hence, these LiDAR systems are namely designed to optimize the sensitivity to atmospheric aerosols that is enhanced at shorter laser wavelengths due to the larger efficiency of both the reference molecular and aerosol backscattering processes [13]. The operation at short wavelengths of a typical atmospheric aerosol LiDAR generally results in a rather limited horizontal detection range due to a much larger signal extinction, which makes it not well suited for sea fog detection. Nevertheless, this limitation is not intrinsic to the LiDAR technique that, in principle, might offer more advantages with respect to other approaches like visual inspection, transmission and forward scattering visibility meters, and so forth. In fact, LiDAR measurements can provide observations of offshore regions that cannot be easily probed by other methods. Moreover, scanning capabilities can allow gaining effective and useful information on the spatial distribution of the visibility over sea surfaces, thus providing accurate identification of sea fog and observation of its generation and dissipation processes [13]. To this end, the LiDAR system must be specifically designed for sea fog detection, but studies on sea fog detection performances of a visibility LiDAR in different weather conditions are still rather scarce [14–17].
The use of LiDAR for visibility detection has been steadily developing over the years and both instrumental developments and data analysis approaches have been progressively explored to improve its performances [14–23]. For example, in 1982, Gaumet et al. developed a LiDAR system to measure atmospheric visibility by using the slope method to invert the experimental data [14]. They compared LiDAR visibility results with those obtained by transmissometers along both horizontal and slant paths. Their results showed that the values of the visibility obtained by the LiDAR technique in foggy conditions were in good agreement with simultaneous transmissometer data. Next, in 2005, Xie et al. used the slope method to obtain visibility measurements from data collected by a mobile Raman–Mie scattering LiDAR [15]. Their results showed that LiDAR is characterized by high reliability and can achieve a measurement uncertainty of less than 20%. Recently, Zhu et al. explored the use of superconducting nanowire single photon detector (SNSPD) exploiting its characteristics of high detection efficiency and low dark counts to improve the detection distance of sea fog registering LiDAR echo signals from fog located at distances of ≈40-80 km and predicting a range of measurement up to about 180 km [24]. In 2017, Pantazis et al. developed a 3D scanning LiDAR to estimate horizontal, slant, and vertical visibilities for tower of aircraft controllers and meteorologists [16]. On the basis of these studies, some of the authors of the present paper developed a scanning LiDAR operating at a laser wavelength of 532 nm and proposed an algorithm to retrieve extinction coefficient and visibility distribution information applying it to the detection of offshore sea-fog in 2018 [17]. A direct comparison between the visibility data obtained by scanning LiDAR and forward scattering visibility instrument showed the superiority of the former in providing accurate real-time monitoring of sea-fog and of its temporal evolution. All these studies seem to suggest that in the field of offshore sea fog observations, dedicated LiDAR systems located on coasts or islands might offer a useful and effective way to set up early-warning systems of sea-fog with improved performances.
Here we report on the design of a visibility LiDAR system optimizing its configuration for fog detection, over a long distance, through simulations that address the influence of various experimental parameters. A series of echo signals are generated, under different visibility conditions, evidencing the effect of the various hardware factors (i.e., laser transmitter wavelength and pulse energy, optical aperture of the receiving telescope, etc.) on the system target performances and defining a final design with optimal parameters. Anticipating our findings, we observe that optimal performances predicted by our simulations identify a wavelength of 1064 nm as the more suitable, which is different from the one used in the previous investigation by some of the authors of the present paper [17]. Therefore, the results of a field campaign measurements carried out by using a visibility LiDAR system with the designed characteristics (wavelength 1064 nm, pulse energy 200 μJ at a repetition rate of 10 kHz) are presented and compared with data provided by a forward scattering visibility meter, eventually supporting the reliability of the designed system and the feasibility of the visibility LiDAR approach in sea fog detection.
2. Background
For a single laser pulse, the power of the atmospheric backscatter echo signal P(λ, X) received from a distance X (units: m) at the transmitter emission wavelength λ (nm) can be expressed as [13,23]:
For aerosol, the extinction to backscattering coefficients ratio SP(λ, X) (aerosol LiDAR ratio) is defined as:
The value of SP(λ, X) is related to aerosol size scale, refractive index and detection wavelength, and ranges generally between 10 and 100 sr [26]. For horizontal detection in homogeneous atmosphere, Sp(λ, X) can be considered independent of the distance X becoming only a function of the wavelength λ, i.e. Sp(λ, X)= Sp(λ).For molecules the extinction to backscattering coefficients ratio Sm(λ, X) (molecule LiDAR ratio) is constant and according to Rayleigh scattering is expressed as [27]:
Therefore, substitution of Eqs. (2)–(5) into Eq. (1) allow to write the power of the atmospheric backscattering echo signal in the following form:
3. Optimization of LiDAR parameters
Our aim is to design a visibility LiDAR that can work reliably in different weather conditions. Taking into account that in most cases the visibility is greater than 5 km, we selected as target figure of the LiDAR system performances a detection range larger than 5 km in daytime operation and 10 km at night, respectively. In order to achieve such target performances, we carried out a series of simulations addressing the influence of the various hardware parameters with the aim of obtaining the optimal selection for the visibility LiDAR system.
As for the transmitter wavelength, we select the values of 1550, 1064, 532 and 355 nm on the base of the various available laser sources. The sky spectral radiance for these four wavelengths, in daytime conditions, correspond to 0.025, 0.05, 0.12 and 0.08 W m-2sr-1 nm-1, in order of decreasing wavelength, as obtained from the solar spectrum [23,30]. Obviously, sky spectral radiance depends on the experimental conditions, but we have considered the ones corresponding to the sun zenith angle that represent the worst case in terms of its influence on the background noise. Moreover, suitable detector choices for the four wavelengths are InGaAs-APD (1550 nm), Si-APD (1064 nm) and PMTs (532 and 355 nm), whose typical values of the quantum efficiency are 10%, 3%, 40%, 30%, respectively [31].
Considering the atmospheric visibility V as given by the meteorological optical range [32], the following relationship with the total extinction coefficient holds:
Where, αp(550) and αm(550) are the aerosol and molecular extinction coefficients at 550 nm. The visibility for the other wavelengths can be calculated using identical formulas. According to the U.S. Standard atmospheric model, for a wavelength of 550 nm, the molecular extinction coefficient αm(550) at an altitude of 100 m is equal to 1.12888×10−5 m-1. Therefore, assuming known the horizontal visibility V, the corresponding aerosol extinction coefficient is given by:As we are dealing with a visibility LiDAR for sea fog detection, here we consider that the system should be located on the coast and the main contribution of atmospheric particulates can be ascribed to marine aerosols.
As for the value of the Ångström exponent [33], we consider the median value obtained from those reported by Catrall et al. [34] for marine aerosol that is equal to 0.8. According to the Rayleigh and Mie scattering [35], the molecular and aerosol extinction coefficients at wavelength λ can, then, be obtained as:
The value of the LiDAR ratio Sp can be very dependent on geographical and climatological characteristics of the observational site and different values, even at the same wavelength, can be found in the literature [34,36–40]. For example, for oceanic aerosols, Catrall et al. reported values of the LiDAR ratio at 550 nm varying in the range (30 ± 10) sr obtained by selected ground-based measurements with sun photometers in the frame of the AERONET global network [34]. For clean maritime conditions, Muller et al. reported values of Sp varying in the range 20-40 sr and 20-30 sr in Sagres (Portugal) and Hulule island (Indian Ocean), respectively, using a Raman LiDAR at a wavelength of 532 nm [36,37]. In another measurement campaign carried out in Hulhuhe island (Maldive), the same group observed a significant change of Sp at 532 nm due to air masses coming from India with values ranging from 50 to 60 and going up to ≈90 [38]. In addition, Ansmann et al. reported values of Sp for marine aerosols below 800 m ranging from 30 to 65 sr, with an average value of ≈40 sr [39].
Using climatological values of aerosol size distributions, Ackermann estimated LiDAR ratio for different tropospheric aerosol types at the typical wavelengths emitted by a Nd:YAG lasers (355 nm, 532 nm and 1064 nm) [40]. For maritime aerosols, the analysis showed an increase of the LiDAR ratio with the wavelength and a striking dependence on the relative humidity. Moreover, also the uncertainties on Sp depend on the relative humidity and can be in the ranges ±(14-18)% at 355 nm, ±(8-13)% at 532 nm and ±(1-5)% at 1064 nm. In an attempt to take into account a wavelength variability of the LiDAR ratio, we considered values of Sp consistent with available data at wavelength of 532 nm, selecting Sp(532 nm)) = 40 sr, and rescaled the values at other wavelengths according to the spectral dependence addressed by Ackermann for relative humidity values of 40-80%, eventually selecting for the simulation values of the LiDAR ratio Sp(λ) at a wavelength λ of 1550, 1064, 532 and 355 nm of 80, 60, 40 and 20 sr, respectively. For any given value of the visibility V, the corresponding value of the atmospheric aerosol and molecular extinction coefficients can be estimated according to Eqs. (12)–(15), and the simulated LiDAR echo signal can be derived by means of Eq. (10). It is worth noticing here that our simulations namely aim at assessing the influence of the various system parameters on visibility measurements carried out at the sea front by LiDAR technique in order to select the optimal experimental conditions. This, in turn, can allow defining the parameters achieving the longest distance over which a sea fog can be detected, which is very important for an effective early warning system. Therefore, our simulations are not meant to gain direct information on the features of LiDAR echoes produced by the sea fog, for which its composition as well as effects of multiple scattering processes should be considered. The identification of the sea fog by LiDAR is associated to a significant drop of the measured visibility.
According to a real-life scenario, a set of parameter values were considered for the visibility LiDAR system, as summarized in Table 1. In particular, the values of some parameters like the laser pulse energy E0, the receiver telescope field of view θ and radius r, and the interference filter bandwidth Δλ were varied in a selected range, reported in Table 1, in order to address their influence on the LiDAR echo, whereas the other parameters listed were set at the given typical value. The optimal values of E0, θ, r and Δλ predicted by the simulations are also reported in Table 1. In the following, when the effect of a specific parameter is going to be illustrated, it is meant that that one is left free to vary while the other three variable parameters are fixed the their optimal values, meanwhile the rest of the system features are as listed in Table 1. The laser repetition rate is set at 10 kHz, whereas the laser beam axis is located at 75 mm from the axis of the receiver telescope.
Table 1. Values of the parameters of the visibility LiDAR used in the simulation for laser wavelengths of 1550, 1064, 532 and 355 nm. The optimal values of E0, θ, r and Δλ are also reported.
We consider first the influence of laser wavelength on the SNR. Figure 1 reports simulated curves of the SNR at the four wavelengths (i.e., 1550, 1064, 532 and 355 nm) for a value of V=10 km, both in nighttime (Fig. 1(a)) and daytime (Fig. 1(b)) conditions. The maximum detection distance Xmax is associated to a SNR value of 3, i.e. SNR(λ, Xmax) = 3. From Fig. 1, it can be clearly seen that the maximum detection distance Xmax, i.e. the value of the distance at which each curve intersects the horizontal x-axis, scales with the wavelength, both for daytime and nighttime conditions, becoming progressively shorter as the wavelength reduces. For example, in daytime conditions, Xmax decreases from ≈ 14 km at 1550 nm to ≈ 6 km at 355 nm. This is because, for the same weather conditions, larger wavelengths imply smaller extinction coefficients and correspondingly lower signal attenuation, thus leading to longer detection distances. On the other hand, in terms of the SNR, Fig. 1 suggests that shorter wavelengths are more advantageous for short detection distances. For instance, at detection distances lower than 2 km the wavelength of 355 nm is the best performing one. This is due to the larger backscattering efficiency at 355 nm with respect to the other larger wavelengths that results in higher values of its SNR. However, direct comparison of the SNR variation with distance for 355 nm and 1064 nm evidences that the former displays higher SNR at detection distances shorter than ≈(4.0-4.5) km, whereas the SNR performance of the latter starts prevailing as soon as detection distances longer than ≈(4.0-4.5) km are concerned (the lower and upper values of the detection distance occurring for daytime or night operation, respectively). In addition, Fig. 1(b) shows that for wide range operation, e.g. for distances larger than ≈3 km, the best performance in terms of SNR are achieved at 1550 nm. Such an analysis highlights how two important factors defining the visibility LiDAR performance as Xmax and SNR display a noteworthy opposite dependence on the wavelength. As the proposed visibility LiDAR of concern to the present study aims at detecting sea fog on a long detection distance in order to gain a wide range of sea visibility distributions the simulation results suggest a priority order for the selection of the laser transmitter wavelength going from infrared to ultraviolet range. However, it is worth to notice that laser sources and detectors operating at a wavelength of 1550 nm are still rather expensive. In turn, this indicates that the most appropriate choice for the wavelength of the transmitter for a sea fog visibility LiDAR can wisely fall on a laser system emitting at a wavelength of 1064 nm. In the following, further parameter optimization of the visibility LiDAR system is carried out fixing the laser light at this wavelength.
Fig. 1. SNR curves for LiDAR wavelengths of 355 (purple), 532 (green), 1064 (fuchsia) and 1550 nm (red), under the condition of 10 km visibility at (a) nighttime and (b) daytime.
We turn now to the analysis of the effect of the pulse energy E0, at the selected wavelength of 1064 nm. Simulating the SNR at 4 different values of E0, namely 50, 100, 150 and 200 μJ, the variation of the maximum detection distance Xmax as a function of the visibility V has been obtained, as for example reported in Fig. 2 for daytime conditions. As said above, in the simulation E0 is varied while the other system parameters remain fixed at the values listed in Table 1, with the other three varying parameters set at their optimal value. As expected, the data on Fig. 2 show at a fixed value of the visibility, the higher the laser pulse energy E0 the longer the detection distance Xmax. Interestingly, for low visibility conditions (e.g. less than 1 km), the dependence of Xmax on E0 is very weak since the laser beam is always highly attenuated on a rather short initial spatial range. Instead, for good or high visibility conditions one can appreciate a progressive rise of Xmax on E0. However, it is worth observing that the improvement in Xmax progressively reduces as E0 increases. For instance, at high visibility conditions a 4× increase of E0 from 50 μJ to 200 μJ leads to a gain of ≈50% only on the value of Xmax. This, in turn, suggests that any additional rise of pulse energy does not improve the maximum visibility range any further. As a target visibility range is reached at the two highest values of E0 considered in the simulation, we therefore set E0=200 μJ as the most appropriate value of the laser pulse energy benefitting from the fact that the slight reduction of Xmax occurring at 150 μJ, in good visibility conditions, is beneficial to maintain good performances of the visibility LiDAR over long operational periods despite of the unavoidable reduction of the laser transmitter energy with usage time.
Fig. 2. Variation of the visibility LiDAR maximum detection range Xmax as a function of the visibility V as obtained from simulated SNR curves at four different values of the single pulse energies E0: 50 μJ (light blue), 100 μJ (orange), 150 μJ (red) and 200 μJ (yellow), in daytime conditions.
As next parameter, we analyze the receiver telescope field-of-view θ, for a laser transmitter wavelength of 1064 nm, pulse energy of 200 μJ and beam divergence 0.2 mrad, the other system characteristics being fixed as in Table 1. Visibility LiDAR signals were simulated, in daytime conditions, for various values of θ in the range (0.05-0.6) mrad and the values of Xmax as a function of the visibility V were estimated. Typical results are displayed in Fig. 3(a) for 4 different values of the receiver field-of-view. Figure 3(b), instead, reports the detection distance Xmax as a function of θ as obtained for a visibility of 5 km in both diurnal and night operation. Figure 3(a) shows that for low visibility (V < 1 km), there is negligible influence of the fields-of-view on the value of Xmax, whereas at the maximum visibility analyzed of 15 km an improvement of ≈35% occurs by passing from θ=0.6 mrad to θ=0.2 mrad. On the other hand, Fig. 3(b) indicates an optimal value for θ matching the beam divergence, as indeed expected, because lower values of θ implies a reduction in the collected echo signal while higher values lead to an increase of the sky background contribution. However, the detection distance Xmax experiences a negligible variation by increasing θ to values larger than 0.2 mrad in night operation conditions. A good compromise can be achieved by considering the average behavior between the two conditions reported in Fig. 3(b) that shows a relatively flat curve for receiver field angles larger than the optimal value of 0.2 mrad with a difference between θ=0.2 mrad and θ=0.4 mrad of only 5%. Hence, the most appropriate choice for the receiver field of view angle is set 0.4 mrad, i.e. 2× the laser beam divergence that also limits the possible influences of parameters fluctuations on the system performances.
Fig. 3. (a) Visibility LiDAR maximum detection distance Xmax vs visibility as obtained from simulations for 4 different values of receiver fields-of-view: 0.1 (light blue), 0.2 (orange), 0.4 (red) and 0.6 mrad (yellow). (b) Variation of the maximum detection distance Xmax as a function of the receiver field-of-view θ for a visibility V of 5 km in daytime and night conditions; the average of the values of Xmax obtained in the two conditions is also shown.
For a laser transmitter wavelength of 1064 nm, pulse energy of 200 μJ and a receiving telescope field-of-view of 0.4 mrad, we carried out simulations addressing the influence of the receiver telescope radius for different visibility conditions on the visibility LiDAR system performances. The results for daytime operation are reported in Fig. 4. Figure 4(a) shows the dependence of the maximum detection distance Xmax as a function of the visibility for three different values of the receiver radius, r. One can observe that at each visibility value, the larger r the longer Xmax. At low visibility conditions (V < 1 km), there is a rather weak influence of the receiving telescope radius on Xmax, whereas for a visibility of 15 km Xmax increases by only 11% when passing from r=0.05 m to r=0.075 m. Figure 4(b) reports the variation of Xmax vs r as well as its corresponding relative change rate ρ=(1/Xmax)×(dXmax/dr) (m-1), for a visibility of 5 km, in daytime conditions. The data of Fig. 4(b) clearly indicate that the increase of Xmax with r, due to an increase of the echo signal collected, is accompanied by a simultaneous reduction of ρ on r consequent to the corresponding rise of the background noise contribution. A receiver telescope radius r = 0.05 m seems to represent a preferable option in order to get good performances at affordable costs; in fact, larger telescopes make the system more expensive without resulting into a very significant improvement of its performances. Hence, the selected value of the receiver radius is 0.05 m in this work.
Fig. 4. (a) Visibility LiDAR maximum detection distance Xmax as a function of the visibility for three different values of the receiver radius r: 0.025 (light blue), 0.05 (orange) and 0.075 m (red); (b) Variation of the maximum detection distance Xmax (left axis, solid curve) and of its relative rate of change, i.e. ρ=(1/Xmax)×(dXmax/dr), (right axis, dotted curve) as a function of the receiver radius r for a visibility of 5 km.
The last parameter of interest deserving analysis is the interference filter bandwidth Δλ (nm). Therefore, we carried out simulations at various values of Δλ for different visibility conditions, keeping the values of the other system parameters as listed in Table 1. A Gaussian spectral linewidth of 0.2 nm of the 1064 nm laser source was considered in the simulations. The results are summarized in Fig. 5(a) shows the variation of Xmax with visibility at four different values of Δλ and in Fig. 5(b) the dependence of Xmax on Δλ for a visibility of 5 km in daytime operation. As a general indication, a narrower bandwidth should yield a longer detection distance. Figure 5(a) indicates that while for low visibility levels (V< 1 km) similar detection distances are achieved almost independently of the filter bandwidth, at a high visibility of 15 km the value of Xmax rises by ≈27% for a 10× increase of Δλ, passing from 0.3 nm to 3 nm bandwidth. On the other hand, Fig. 5(b) shows a reduction of the detection distance for bandwidths smaller than 0.4 nm, which although limit the amount of sky background noise also decreases the signal power eventually reaching the detector after passing through the filter. For values of the bandwidth larger than 0.4 nm, the signal power at the detector is not affected anymore but the contribution of background noise increases, thus worsening the SNR and limiting the performance. This seems to suggest that a filter bandwidth of 0.4 nm might provide an optimal choice to maximize the detection distance. However, it is worth considering that fluctuation of the operating temperature can influence both the response of the interference filter as well as the emission linewidth of the laser source, therefore a larger filter bandwidth of 1 nm is considered a more appropriate choice to keep good-level performances and assure reliable working operation of the system in different ambient conditions.
Fig. 5. (a) Visibility LiDAR maximum detection distance Xmax as a function of the visibility for three different values of the interferential filter bandwidth Δλ: 0.3 nm (light blue), 0.5 nm (orange), 1 nm (red) and 3 nm (yellow); (b) Variation of the LiDAR detection distance Xmax as a function of the filter bandwidth Δλ for 5 km visibility in daytime conditions.
Finally, Fig. 6 reports the detection distance of the visibility LiDAR system according to simulations of a system with the parameters listed in Table 1, both for daytime and night, as a function of the visibility V. The results of Fig. 6 predict values of the maximum detection distances larger than 6 km and 10 km for daytime and night operations, respectively at a visibility level of 5 km, which fits with our target requirements for the visibility LiDAR system for sea fog early warning.
Fig. 6. Visibility LiDAR maximum detection distance Xmax as a function of the visibility for daytime (orange) and night (light blue) conditions.
4. Results and discussion
A visibility LiDAR system with design characteristics resulting from the performance assessment study provided by the simulations discussed in the previous section and whose working parameters are listed in Table 1 was realized and tested. The laser transmitter provided 1064 nm pulses of 200 μJ at a repetition rate of 10 kHz. The laser beam axis was located at a distance of 75 mm from the axis of the receiver telescope, whose radius was 50 mm. This new system at 1064 nm differs with respect to the one used in a previous study carried out by some of the authors of the present paper at a wavelength of 532 nm [17] that was designed on the base of previous literature work. In this respect, it is worth noticing that the longer wavelength of 1064 nm used here is characterized by a narrower dynamic range that leads to the possibility of probing longer distances without suffering from detector saturation effects due to the large signal echo recorded from short ranges. In fact, sounding the same maximum distances with shorter wavelength of 532 nm can also be possible by increasing the laser pulse energy, but this would be accompanied by a corresponding increase of the signal at short range with associated possible saturation effects, which would limit its performances.
In 2019, the visibility LiDAR was installed on Hengsha Island (31.297424 °N, 121.849609 °E), Chongming District, Shanghai, China (see photos in Fig. 7). The system was also provided with a scanning capability as shown in the close-up image of Fig. 7(a) and in the schematic diagram of Fig. 7(b). Figure 7(c) reports a picture showing the sea surface as seen from the installation site towards the south. The visibility LiDAR was placed on a tower at a height of 25 m above the ground, as shown in Fig. 7(d). The scanning azimuthal range of the visibility LiDAR can range from 0° to 360°, with 0° corresponding to north, 90° to east, 180° to south, and 270° to west. Scanning angle resolution and integration time were set to 2° and 10 s, respectively, and the average scanning cycle time was 30 min. LiDAR echo signals were registered during the angular scan, obtaining a visibility distribution map with the LiDAR system at its center. The data were processed by using a horizontal visibility inversion algorithm [17], which overcame initial value sensitivity and did not require assuming the LiDAR ratio.
Fig. 7. Photos of the system for a testing campaign carried out of Hengsha Island (Shangai, China) in 2019: (a) close-up image of the scanning visibility LiDAR; (b) on site installation diagram; (c) image of the site towards south showing the sea surface in the distance; (d) image of the tower on which the LiDAR system was installed.
In June 2019, a forward scattering visibility meter [6] was installed and located 23 m below the visibility LiDAR, at a horizontal distance of 10 m. Simultaneous measurements were carried out by the two instruments. As the forward scattering visibility meter is to the south of LiDAR, the visibility obtained by the LiDAR scanning in the 180° direction was compared with that of the forward scattering visibility meter.
Figure 8 reports a comparison of the values of the visibility obtained during a set of measurements continuously carried out by the two systems over a period of twelve days in June 2019. Figure 8 demonstrates an overall, high-level of consistency between the results obtained by the two techniques providing a reliable and successful test of the accuracy and stability of the visibility LiDAR system after parameter optimization. However, discrepancy is also observed in very few cases. To address when a sizable difference might occur, we have selected two exemplificative cases, marked with A and B, in the set of measurements reported in Fig. 8. In order to clarify the reason of discrepancy, the range corrected signals (RCS) registered by the LiDAR in the two cases are displayed in Fig. 9(a); the RCS profiles without and with the correction by the LiDAR overlap function of the LiDAR system are reported, with (A, B) referring to the former and (A’, B’) to the latter case, respectively. At measurement time B, the RCS profile is well described by a linearly decreasing trend with an almost constant slope over a very large region going down to the LiDAR station, which indicates relatively homogeneous atmospheric conditions. In such a situation, a very good agreement between the two techniques is achieved. Instead, at measurement time A, the values of the visibility obtained by the two systems show a remarkable difference. In this situation, the RCS profile A’ of Fig. 9(a) evidences the presence of various aerosol structures in the range of 0–1.3 km, with a larger slope in the very first distance of about 200 m. Then, it is followed by a distance interval going from 1.3 to 5.5 km with a relatively smooth slope suggestive of a region characterized by a small extinction coefficient and a high visibility. In the following spatial range (5.5-10) km, the curve slope slightly rises but the visibility remains still high. Therefore, the LiDAR signal suggests a rather good visibility except for the region very close to the LiDAR station presenting the aerosol structures. The forward scattering visibility meter essentially measures the visibility at its location that in this case corresponds to the very first range of the LiDAR signal shown in Fig. 9 (profile A’), therefore leading to a value of the visibility lower with respect to the that estimated by the LiDAR. This, in turn points out that discrepancies observed in Fig. 8 between the two techniques namely occur in the case of inhomogeneous atmospheric conditions, in which the much shorter sampling path used by forward scattering visibility meter may provide an estimate of the visibility based on a more localized sampling of the atmospheric conditions than that used by the LiDAR. Our findings clearly evidence in this case that the measure provided by the forward scattering visibility meter is somewhat smaller with respect to the LiDAR estimate, but also the opposite situation can occur, as observed in few cases in Fig. 8(a). Notwithstanding the occurrence of some discrepancies, the overall consistency over the rather long observation time interval evidences a striking agreement between the two techniques for homogeneous atmospheric conditions, thus demonstrating a good accuracy of the visibility LiDAR measurements. Moreover, the actual measured value of the visibility at time B is ≈5 km (Fig. 8(b)), while the RCS profile B in Fig. 9 suggests that the effective distance of the signal is about 6.5 km, which is consistent with the results of our previous parameter optimization.
Fig. 8. Comparison of the visibility obtained by the optimized visibility LiDAR (light blue) and the forward scattering visibility meter (red) from (a) June 6 to 12 and (b) June 12 to 18, 2019.
Fig. 9. (a) RCS profiles and (b) visibility distribution maps (5 km radius) corresponding to the times A and B marked in Fig. 8(b).
Figure 9(b) shows a visibility distribution maps obtained by the LiDAR at times A and B, respectively. In Fig. 9(b), the radius of the colored region is 5 km. From this figure, it is still more evident that the atmospheric conditions at time A were rather inhomogeneous, whereas the situation at time B was much more homogeneous. All these experimental findings strongly support the capabilities and advantages of the visibility LiDAR system for long range visibility measurements.
Hereafter, we discuss further measurements carried out on February 6, 2019, when the weather in Shanghai was occasional light rain and with a southeast wind of grade 2. Figure 10 reports a set of visibility distribution maps obtained by the scanning visibility LiDAR at different times (UTC+8) on February 6, 2019. The maps allow analyzing the sea surface visibility distribution and sea fog evolution process. The visibility map of Fig. 10(a), registered at 03:30, shows a visibility larger than ≈10 km in most areas and evidence the presence of a light sea fog at a distance of approximately 5 km towards south with a local visibility in the range of 5–7 km. Figure 10(b), registered at 04:01, shows a sharp decrease of the visibility to values of 2–4 km in the southern area, whereas it remains still larger than 10 km in the north. Figure 11 shows the wind speed and direction registered on the observational site on February 6, 2019. Figure 11(a) indicates that at 04:00 the wind was directed prevalently toward south. The visibility map of Fig. 10(c) at 04:31 highlights that under the influence of this south-directed wind the fog gradually moves towards the north. Then, Fig. 11(b) evidences a progressive change into a southeast wind occurring from 04:00 to 07:00, while Figs. 10(d)–10(g) shows that the fog moves towards northwest gradually spreading to the position of the visibility LiDAR station. Then, Figs. 10(e)–10(j) highlight how the fog covers the whole area during its evolution, beginning to dissipate at later times as shown by the visibility improvement evidenced by the map reported in Fig. 10(k). Finally, Fig. 10(l) shows that after the fog has fully dissipated, the visibility increases again to values larger than 10 km.
Fig. 10. LiDAR visibility distribution map (5 km radius) at different times (UTC + 8) registered on February 6, 2019: (a) 03:30, (b) 04:01, (c) 04:31, (d) 05:01, (E) 05:31, (f) 06:02, (g) 07:02, (h) 07:32, (i) 09:03, (j) 09:33, (k) 11:04 and (l) 16:36.
Fig. 11. Wind speed and direction on February 6, 2019 at two different times (UTC+8): (a) 04:00 and (b) 07:00.
From the visibility distribution in the 180° direction, the values of the visibility at distances of 1 and 5 km at different time were extracted and reported in Fig. 12. The data enclosed in the red box show that the visibility at 1 km is larger than at 5 km at 03:30. After diffusion of the sea fog, the visibility at 1 km decreases gradually, finally reaching a minimum at 9:33. The present analysis clearly shows that at 03:30 the visibility LiDAR detected a light sea fog at a distance of 5 km which by 07:02 had spread to the shore. Therefore, the LiDAR system was capable of providing an early warning of the sea fog event with temporal advance of ≈3.5 hours.
Fig. 12. Time varying curves of the visibility at distances of 1 km (orange) and 5 km (purple) along the 180° direction on February 6, 2019.
5. Conclusions
A visibility LiDAR system was proposed and the LiDAR main design parameters were assessed on the base of simulation of the system performances. The simulations were carried out by generating a series of LiDAR echo signals as a function of the visibility and the influence of the different hardware parameters was critically evaluated. From the simulation predictions a set of optimal parameters for designing a visibility LiDAR with a maximum detection distance larger than 6 km in daytime and 10 km at night for weather conditions with a typical visibility of 5 km were defined. The developed visibility LiDAR system with the defined parameters was tested in a field experiment carried out at Hengsha Island in Shanghai. Continuous observations were carried out simultaneously with a forward scattering visibility meter, comparing the measurement results. The comparative experiment showed that both instruments provide consistent results under homogeneous atmosphere conditions. However, in inhomogeneous atmospheres, the visibility measurements obtained with the visibility LiDAR seem to suggest a higher degree of reliability. In addition, in February 2019, the visibility LiDAR was tested during a sea fog event showing the capability of providing an early warning for a light fog at a distance of 5 km of about 3.5 hours in advance. These results demonstrate that the visibility LiDAR has achieved good performances for long range visibility measurements and with the ability of providing reliable sea fog warnings.
Funding
China Scholarship Council.
Acknowledgments
J.X. thanks the China Scholarship Council (CSC) funding of a grant for the stay at Dipartimento di Fisica “Ettore Pancini”, Università di Napoli “Federico II”. This work was supported by Darsunlaser Technology Co., Ltd.
Disclosures
The authors declare that there are no conflicts of interest related to this article.
References
1. D. Koracin, C. E. Dorman, J. M. Lewis, J. G. Hudson, E. M. Wilcox, and A. Torregrosa, “Marine fog: A review,” Atmos. Res. 143, 142–175 (2014). [CrossRef]
2. S. Zhang, “Recent observations and modeling study about sea fog over the Yellow Sea and East China Sea,” J. Ocean Univ. China 11(4), 465–472 (2012). [CrossRef]
3. J. Lewis, D. Koracin, R. Rabin, and J. Businger, “Sea fog off the California coast: Viewed in the context of transient weather systems,” J. Geophys. Res.: Atmos. , 108(D15), 4457 (2003). [CrossRef]
4. J. M. Lewis, D. Koracin, and K. T. Redmond, “Sea fog research in the United Kingdom and United States - A historical essay including outlook,” Bull. Am. Meteorol. Soc. 85(3), 395–408 (2004). [CrossRef]
5. J. Chen, B. Han, Q. Yang, L. Wei, Y. Zeng, R. Wu, L. Zhang, and Z. Ding, “Analysis of a Sea Fog Episode at King George Island, Antarctica,” Atmosphere 10(10), 585 (2019). [CrossRef]
6. J. V. Winstanley and M. J. Adams, “Point visibility meter - forward scatter instrument for measurement of aerosol extinction coefficient,” Appl. Opt. 14(9), 2151–2157 (1975). [CrossRef]
7. R. Barrales-Guadarrama, A. Mocholí-Salcedo, M. E. Rodríguez-Rodríguez, V. R. Barrales-Guadarrama, and E. R. Vázquez-Cerón, “A New Forward-Scatter Visibility Sensor Based on a Universal Frequency-To-Digital Converter,” Instrum. Sci. Technol. 41(5), 445–462 (2013). [CrossRef]
8. J.-Y. Jeon, S.-H. Kim, and C.-S. Yang, “Fundamental Research on Spring Season Daytime Sea Fog Detection Using MODIS in the Yellow Sea,” Korean Journal of Remote Sensing 32(4), 339–351 (2016). [CrossRef]
9. Y.-F. Xiao, J. Zhang, and P. Qin, “An Algorithm for Daytime Sea Fog Detection over the Greenland Sea Based on MODIS and CALIOP Data,” J. Coastal Res. 90(sp1), 95–103 (2019). [CrossRef]
10. J. B. Mead, R. E. McIntosh, D. Vandemark, and C. T. Swift, “Remote-sensing of clouds and fog with a 1.4-mm radar,” J. Atmos. Oceanic Technol. 6(6), 1090–1097 (1989). [CrossRef]
11. M. Temimi, R. M. Fonseca, N. R. Nelli, V. K. Valappil, M. J. Weston, M. S. Thota, Y. Wehbe, and L. Yousef, “On the analysis of ground-based microwave radiometer data during fog conditions,” Atmos. Res. 231, 104652 (2020). [CrossRef]
12. D. G. Macfarlane, G. Wadge, D. A. Robertson, M. R. James, and H. Pinkerton, “Use of a portable topographic mapping millimetre wave radar at an active lava flow,” Geophys. Res. Lett. 33(3), L03301 (2006). [CrossRef]
13. Weitkamp, Claus, LIDAR : range-resolved optical remote sensing of the atmosphere (Springer, 2005).
14. J. L. Gaumet and A. Petitpa, “Lidar-transmissometer visibility comparisons over slant and horizontal paths,” J. Appl. Meteorol. 21(5), 683–694 (1982). [CrossRef]
15. C.-B. Xie, Y. Han, C. Li, G.-M. Yue, F.-D. Qi, A.-Y. Fan, J. Yin, S. Yuan, and J. Hou, “Mobile lidar for visibility measurement,” High Power Laser and Particle Beams 17, 971–975 (2005).
16. A. Pantazis, A. Papayannis, and G. Georgousis, “Lidar algorithms for atmospheric slant range visibility, meteorological conditions detection, and atmospheric layering measurements,” Appl. Opt. 56(23), 6440–6449 (2017). [CrossRef]
17. J. Xian, Y. Han, S. Huang, D. Sun, J. Zheng, F. Han, A. Zhou, S. Yang, W. Xu, Q. Song, L. Wei, Q. Tan, and X. Li, “Novel Lidar algorithm for horizontal visibility measurement and sea fog monitoring,” Opt. Express 26(26), 34853–34863 (2018). [CrossRef]
18. F. Castagnoli, M. Morandi, I. Pippi, and B. Radicati, “Lidar system for visibility monitoring,” Opt. Quantum Electron. 15(3), 261–268 (1983). [CrossRef]
19. A. Cohen, “Horizontal visibility and measurement of atmospheric optical depth of lidar,” Appl. Opt. 14(12), 2878–2882 (1975). [CrossRef]
20. M. Chen, D.-S. Sun, H.-J. Li, F.-H. Shen, and H.-Y. Xia, “Detective method and analysis of atmospheric slant visibility for lidar,” Infrared Laser Engineering 35, 156–160 (2006).
21. L. Tang, Z. Shu, J. Dong, B. Yue, F. Shen, J. Dong, and D. Sun, “Measurement of slant visibility and its iteration method with diode-laser lidar,” Qiangjiguang Yu Lizishu 22(5), 963–967 (2010). [CrossRef]
22. B. Yue, J.-J. Dong, D.-S. Sun, F.-H. Shen, J.-H. Dong, and Z.-F. Shu, “Measuring method of atmospheric slant visibility with semiconductor lidar,” Infrared Laser Engineering 38, 135–139 (2009).
23. R. M. Measures, Laser Remote Sensing: Fundamentals and Applications (Krieger, 1992).
24. J. Zhu, Y. Chen, L. Zhang, X. Jia, Z. Feng, G. Wu, X. Yan, J. Zhai, Y. Wu, Q. Chen, X. Zhou, Z. Wang, C. Zhang, L. Kang, J. Chen, and P. Wu, “Demonstration of measuring sea fog with an SNSPD-based Lidar system,” Sci. Rep. 7(1), 15113 (2017). [CrossRef]
25. F. G. Fernald, B. M. Herman, and J. A. Reagan, “Determination of aerosol height distributions by lidar,” J. Appl. Meteorol. 11(3), 482–489 (1972). [CrossRef]
26. S. J. Doherty, T. L. Anderson, and R. J. Charlson, “Measurement of the lidar ratio for atmospheric aerosols with a 180degree backscatter nephelometer,” Appl. Opt. 38(9), 1823–1832 (1999). [CrossRef]
27. J. Strutt, “On the transmission of light through an atmosphere containing small particles in suspension, and on the origin of the blue of the sky,” Philos. Mag. 47, 375–394 (1899).
28. T. Y. Nakajima, T. Imai, O. Uchino, and T. Nagai, “Influence of daylight and noise current on cloud and aerosol observations by spaceborne elastic scattering lidar,” Appl. Opt. 38(24), 5218–5228 (1999). [CrossRef]
29. J. D. Spinhirne, “Micro pulse lidar,” IEEE Trans. Geosci. Electron. 31(1), 48–55 (1993). [CrossRef]
30. B. M. Herman, S. R. Browning, and R. J. Curran, “The Effect of Atmospheric Aerosols on Scattered Sunlight,” J. Atmos. Sci. 28(3), 419–428 (1971). [CrossRef]
31. G. Ribordy, J. Gautier, H. Zbinden, and N. Gisin, “Performance of InGaAs/InP Avalanche Photodiodes as Gated-Mode Photon Counters,” Appl. Opt. 37(12), 2272–2277 (1998). [CrossRef]
32. H. Koschmieder, “Atmospheric haze and visibility,” Naturwissenschaften 26(32), 521–528 (1938). [CrossRef]
33. A. Ångström, “The parameters of atmospheric turbidity,” Tellus 16(1), 64–75 (1964). [CrossRef]
34. C. Cattrall, J. Reagan, K. Thome, and O. Dubovik, “Variability of aerosol and spectral lidar and backscatter and extinction ratios of key aerosol types derived from selected Aerosol Robotic Network locations,” J. Geophys. Res. 110(D10), D10S11 (2005). [CrossRef]
35. G. Mie, “Articles on the optical characteristics of turbid tubes, especially colloidal metal solutions,” Ann. Phys. 330(3), 377–445 (1908). [CrossRef]
36. D. Müller, K. Franke, F. Wagner, D. Althausen, A. Ansmann, J. Heintzenberg, and G. Verver, “Vertical profiling of optical and physical particle properties over the tropical Indian Ocean with six-wavelength lidar: 2. Case studies,” J. Geophys. Res.: Atmos. 106(D22), 28577–28595 (2001). [CrossRef]
37. D. Müller, A. Ansmann, F. Wagner, K. Franke, and D. Althausen, “European pollution outbreaks during ACE 2: Microphysical particle properties and single-scattering albedo inferred from multiwavelength lidar observations,” J. Geophys. Res.: Atmos. 107(B9), ESE 3-1–ESE 3-13 (2002). [CrossRef]
38. D. Müller, K. Franke, A. Ansmann, D. Althausen, and F. Wagner, “Indo-Asian pollution during INDOEX: Microphysical particle properties and single-scattering albedo inferred from multiwavelength lidar observations,” J. Geophys. Res. 108(D19), 4600 (2003). [CrossRef]
39. A. Ansmann, F. Wagner, D. Althausen, D. Müller, A. Herber, and U. Wandinger, “European pollution outbreaks during ACE 2: Lofted aerosol plumes observed with Raman lidar at the Portuguese coast,” J. Geophys. Res. 106(D18), 20725–20733 (2001). [CrossRef]
40. J. Ackermann, “The Extinction-to-Backscatter Ratio of Tropospheric Aerosol: A Numerical Study,” J. Atmos. Oceanic Technol. 15(4), 1043–1050 (1998). [CrossRef]
