Open Access
Add to BibSonomy
Abstract
Self-calibration of UV cameras was demonstrated for the first time. This novel method has the capability of real-time continuous calibration by using the raw images at 310 nm and 330 nm without changing the viewing direction or adding any additional equipment. The methodology was verified through simulations and experiments and demonstrated to be of greatly improved effectiveness and accuracy. The errors of self-calibration mothed are estimated by comparison with the differential optical absorption spectroscopy (DOAS) approach, and it can be reduced to 1.8% after filter transmittance corrections. The results show that the self-calibration method appears to have great potential as a future technique for quantitative and visual real-time monitoring of SO2 emissions from ships and other point sources (such as oil refineries, power plants, or more broadly, any industrial stack) when the field of view (FOV) of the system is not completely covered by the SO2 plumes.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The ultraviolet (UV) cameras developed in the 2006 enabled monitoring of the SO2 emission rate more accurate due to its unique advantage in providing unprecedented insights into plume dynamics with high spatial and temporal resolution [1]. Focus of this novel gas imaging technology has been initially on volcano hazard monitoring because the SO2 emission rate is one of the most important indicators of volcanic activity [2–4]. With the improvement of UV cameras and the development of UV filter technology, the SO2 cameras was recently employed for measuring SO2 pollution escaped from anthropogenic sources such as power plants, garbage burning plants, oil refineries and other industrial stacks [5,6]. Some efforts were also made to employed this advanced technology to monitor the much lower concentrations of SO2 in the plumes from ships at sea or at anchor [7].
Despite the major advantages of the SO2 cameras over other remote sensing technologies (such as DOAS [8,9] and FTIR [10]), the limitation in its difficult calibration suggest that currently this novel system may only be more suitable for monitoring stationary pollution sources, and that the calibration mothed needs further development to make SO2 cameras serve as an effective method to monitor ship emissions for regulatory purposes [11].
For the SO2 cameras, which use a set of UV band-pass filters centered at 310 nm (to capture the absorption features produced by the SO2 molecules) and 330 nm (to correct for radiative transfer effects of aerosol) for SO2 quantitative imaging, the received signal is not measured as a function of wavelength λ like DOAS, but instead the integral intensity of the incident spectrum over the filter transmittance window [12,13]. Unlike the optical density τ(λ) retrieved with DOAS, the weighted average optical density $\hat{\tau }$ obtained from SO2 cameras is not linear proportional to the SO2 column density SSO2. Therefore, the calibration process is essential for the SO2 cameras to accurately determine the numerical correspondence relationship between $\hat{\tau }$ and SSO2 [11]. And the calibration may be better in real-time or near real-time because it heavily depends on the solar zenith angle (SZA) and the total O3 column of the local atmosphere.
There are two commonly used calibration methods for the SO2 camera system. The empirical calibration using several absorption cells with different SO2 pathlength concentration is currently the most straightforward method [14,15]. However, using this calibration method need to change the viewing direction and place calibration cells in front of the camera frequently, which makes it not suitable for applications in monitoring SO2 emissions from moving ships. Calibration of the camera using SO2 cell can be conducted in the laboratory (as employed in [7]), but it will be hard to estimate and predict the calibration uncertainty because the sky conditions (e.g., SZA, O3 column and presence of clouds) has a crippling effect on the calibration curves. The up-to-date intrinsic calibration achieved by co-locating a DOAS instrument with the SO2 camera makes the real-time continuous calibration feasible [11]. However, except the complexity and additional cost, it is quite difficult to exactly know the area to which the DOAS telescope is directed and obtain the best correlation between the corresponding AA data from the SO2 camera and the SO2 CD recorded by the DOAS at the same time.
Our motivation for this work is to develop a self-calibration method for real-time monitoring of SO2 ship emissions with UV cameras. This proposed method has the capability to calibrate the continuously measured data using the raw images at 310 nm and 330 nm continuously without changing the viewing direction or adding any additional equipment (such as cells or spectrometers). The organization of this paper is as follows: we briefly describe the principal characteristics of SO2 cameras and for the theoretical basis for pertinent aspects of spectral consideration. The detailed principle of the self-calibration methodology is described, and the issues that influence self-calibration are identified. The main results of the ship experiments conducted near Shanghai Port are presented. For purpose of verifying the accuracy and effectiveness of this novel method, we compare the it with the calibration approach with DOAS system. In order to further enhance the calibration accuracy, the effective filter transmittance corrections are made. This is followed by a detailed error comparison of the calibration time series between the two calibration methods. We conclude with comments on how this technology might be improved by account for changes in camera sensitivity due to the presence of aerosol in the plume, and the radiative transfer between the sun and the instrument, including both “radiative dilution” and multiple scattering. Here we report, to the best of our knowledge, the first self-calibration method for real-time monitoring of SO2 emissions with UV cameras.
2. Spectral consideration
SO2 is a toxic gas, and exhibits significant spectral features in the ultraviolet region between 240–340 nm, due to vibrational bands of the electronic transition $\tilde{B}{}^1{B_1} \leftarrow \tilde{X}{}^1{A_1}$ [16]. The strongest feature at wavelength region shorter than 300 nm is not suitable for ground-based remote sensing of SO2 because O3 absorption dominates in this region and the sun shone could not break through the ozone layer. The incident solar spectral intensity and the scattered solar spectral intensity reaching the ground, which are calculated using MODTRAN [17] at a resolution of 0.1 nm, are shown on the Fig. 1(a). The deep ultraviolet region is opaque and thus unsuitable for ground-based use. Fortunately, in the narrow band region centered at 310 nm, the atmospheric absorption is relatively weak, and the absorption feature of SO2 is rather strong. Therefore, almost all of the SO2 camera systems locate the maximum filter transmission of the signal channel at 310 nm with a standard deviation of 3 nm [1,4,7,12]. Another band-pass filter centered at 330 nm is usually used in the reference channel to quantify the amount of light attenuation not originating from SO2 absorption [13]. The spectral transmittance of the pair of filters used in our system is shown in Fig. 1(c).
Fig. 1. The limb spectral radiance of the O2 infrared atmospheric band as a function of tangent height.
Because the background concentration of SO2 in the atmosphere is very low (typically less than 10 parts per billion [4]), the atmospheric extinction due to SO2 can be ignored [shown blue line in Fig. 1(b)]. While the O3 absorption dominates in this region [shown as red line in Fig. 1(b)]. Therefore, the background sky images recorded by both the signal and reference channels are influenced only by the variations in ozone optical density. For high solar zenith angles (SZA), the average optical path length through the stratospheric ozone layer is considerably longer than for lower SZAs. Figure 2(a) shows the scattered solar spectral intensity varying with wavelength for four different SZA (0°, 20°, 40° and 60°). As the absorption cross-section of O3 increases significantly towards deep ultraviolet wavelengths, the signal channel (310 nm) is particularly influenced by variations in ozone optical path length, while wavelengths transmitted though the reference are relatively less affected by ozone absorption. This difference is the fundamental source of self-calibration of the SO2 camera systems.
Fig. 2. Top traces: the scattered solar spectral intensity varying with wavelength for four different SZA (0°, 20°, 40° and 60°). Middle traces the intensity ratio of the two channels varying with SZAs at different atmospheric conditions. Bottom traces: the calculated calibration coefficients varying with the negative logarithm of the intensity ratio of the two channels.
The intensity ratio of the two channels can be given by
Figure 2(b) shows the intensity ratio of the two channels varying with SZAs at different atmospheric conditions (atmospheric visibility or presence of clouds). As illustrated, intensity ratio is approximately the monotone function of SZA, i.e. ozone optical density, and suffers litter from atmospheric conditions.
The calibration coefficient of the UV camera system can be obtained from the linear fitting of the numerical relationship between ${\tau _{S{O_2}}}$ and ${S_{S{O_\textrm{2}}}}$, which is the typical way to find the desired calibration. The two commonly used calibration methods, inserting SO2 cells into the light path and measuring the column density (CD) with an additional DOAS system, are all based on this principle. In reality, the calibration curve of the SO2 camera can also be calculated directly, if the scattered solar radiation, the transmission curves of the two filters and the quantum efficiency of the detector are given.
The physical relationship between SO2 optical density τSO2 and the SO2 column density SSO2 depends on SZA, the total O3 column, the spectral transmittance the band-pass filter and the quantum efficiency of the camera. It can be calculated following the theoretical description provided in Sect. 2.1 of [6].
Figure 2(c) shows the calculated calibration coefficients $\kappa$ varying with the negative logarithm of the intensity ratio of the two channels at different atmospheric conditions. Two features can be seen in Fig. 2: first, a clear decrease in the intensity ratio (A to B) with increasing SZA and a significant increase of calibration coefficient with increasing value of the negative logarithm of the intensity ratio can be observed; second, the calibration coefficient is approximately the monotone function of the negative logarithm of the intensity ratio, and the relationship between them is little affected by atmospheric visibility or presence of clouds. Therefore, the calibration curve of the SO2 camera can be clearly and easily determined if the ratio of the two background images of the two channels is known.
3. Self-calibration methodology
As with any new techniques or methods, it is necessary to start with a theoretical analysis or simulation, carrying out experimental study and then improve this technique or method based on the experimental results.
In order to verify the accuracy and effectiveness of this novel self-calibration method, we tested our approach by observing SO2 emissions from ships at sea near Shanghai Port on 11 April 2019. The SO2 imaging camera used in this experiment consists of a pair of UV sensitive back-illuminated CMOS sensor (Photometrics Prime95B) with 1200 × 1200 pixels, C-mount quartz lenses (Universe Kogaku UV1054B) with F number of 4.0, and two UV bandpass filters (Asahi Bunko Co.) with peak transmittance wavelengths at 310 nm and 330 nm for each channel. The main advantage of this camera is its fast sampling frequency (better than 40 Hz) and good signal-to-noise ratio (the quantum efficiency of the detector is about 50–55% at 310 nm and 330 nm), which could meet the characteristics of ship monitoring (fast moving and low SO2 concentration). For purpose of comparison, a DOAS is also deployed in this remote sensing system and installed between the two cameras. The DOAS system consists of an Ocean Optics USB2000+ spectrometer with a resolution of 0.035 nm, a 400 µm quartz optical fiber, and a telescope with a quartz lens same with the UV camera. The DOAS is mechanically attached to the camera optics and measures the SO2 CD at the approximate center of the camera image. The Field set-up of this remote sensing system is shown in Fig. 3. Both the SO2 camera and the DOAS are attached above a pan tilt, which is controlled through software to track the ship.
Figures 4(a) and 4(b) show raw images of the ship plume captured by the signal channel and the reference one of the SO2 camera system. The background sky images of the two channels can be obtained from the raw images using the 2-IM approach developed by [6]. Before artificial background generation, the raw images were processed following the protocols outlined in Kantzas et al. [12], particularly involving the subtraction of the dark current image, and optimizing image correlation though translation and rotation operations for best match or coincidence. The artificial background sky images generated though the 2-IM approach are shown in Figs. 4(c) and 4(d). For convenience of calculations, the sea surface part of the raw image is cropped and removed, and the sky one in the rectangular area marked by red dotted line is preserved. The average digital numbers (ADU) of the two background images are 13157 and 58885, which are used as the input parameter of the self-calibration mothed. According to the numerical relationship of the calibration coefficient and the negative logarithm of the intensity ratio of the two images [see Fig. 2(c)], the camera calibration curve can be clearly and easily determined.
Fig. 4. Example of raw images at λsig = 310 nm (a) and at λref = 330 nm (b) of the ship plume, as well as the artificial background sky generated by the 2-IM procedure: (c) 310 nm, (d) 330 nm.
To overcome the divergence of drift error over time, the calibration of the SO2 camera system should be conducted frequently, especially for the ship plume monitoring. The SO2 camera measures the optical density (OD) of SO2. Calibration could convert the OD arrays to SO2 CD images by using the parameterization of the $\kappa \textrm{ - }R$ relationship determined by Eq. (1) and (2). Figure 5(a) shows the camera calibration curve (the red solid line) obtained with the self-calibration method using the the the numerical relationship of the calibration coefficient and the negative logarithm of the intensity ratio of the two images [see Fig. 2(c)]. For purpose of comparison, the calibration curve calculated from the hyperspectral data of the background sky recorded with the UV spectrometer of the DOAS system is overlapped in the same figure (shown red solid line). The camera calibration curve is simulated using the method proposed by Lubcke et al. [11]. Note that in this experiment, the two cameras simultaneously capture the plume with the same FOV, while the UV spectrometer measures the spectral information of the plume-free area close to the FOV of the camera. After converting the SO2 OD to SO2 CD by multiplying with the calibration factor of 2976 ppm·m obtained from the self-calibration method, the color map of the SO2 image of the ship plume is achieve [see Fig. 5(b)].
Fig. 5. (a) Camera calibration curves obtained with the self-calibration mothed and simulated from a sky spectrum recorded with the UV spectrometer respective. (b) Colormap of the SO2 image of the ship plume retrieved from the UV cameras.
4. Effective filter transmittance corrections
As illustrated from Fig. 5, the calibration curves from the self-calibration method and the DOAS approach both show a linear relationship between SO2 OD and SO2 CD. However, the slopes derived by the two calibration methods differ by more than 10%. This may be the result of non-perpendicular illumination of the bandpass interference filters – effects that are not accounted for when calibrating with both the DOAS approach and the self-calibration method. The imaging instrument integrates the incident radiance, rather than the OD, over a finite wavelength bandwidth, and the shift of the filter's central wavelength and the decrease of the maximum transmittance due to non-perpendicular illumination may lead to a slight change of instrument's sensitivity to SO2. It is this issue that we wish to address in this section.
The physical model for simulating the effective filter transmittance has been explained in detail by Kern et al. [18], and here just a brief description is given.
For this UV camera system, the bandpass filter is placed between the object lens and the detector, therefore, its effective transmittance can be given by
${\lambda _C}$ is the effective bandpass center depending on the filer’s refractive index ${n_F}$ and its central wavelength for perpendicular illumination ${\lambda _F}$
Figure 6 shows the filter transmittance curves measured by the UV spectrometer using the collimated halogen lamp as light source, and the calculated effective filter transmittance curves for this SO2 camera system. Note that the effective maximum transmittance is reduced by about 10.5% and more importantly, the effective center wavelength shifts by approximately 0.25 nm towards shorter wavelengths, which will inevitably lead to a change of camera system sensitivity to SO2.
Fig. 6. The filter transmittance curves (shown solid lines) measured by the UV spectrometer using the collimated halogen lamp as light source, and the calculated effective filter transmittance curves (shown dotted lines) for this SO2 camera system.
Figure 7 shows the camera calibration curves and SO2 image of the ship plume after effective filter transmittance corrections. This time the difference between slopes of the two calibration curves derived by the self-calibration method and the DOAS approach is less than 2%, well within the measurement uncertainty. Therefore, the accuracy and effectiveness of the self-calibration method are improved significantly by the filter transmittance corrections.
The time series of the relative error of the calibration factors obtained by the self-calibration method and the DOAS approach are shown in Fig. 8. Good agreement is achieved especially for first twenty seconds and the thirty seconds between 50 s and 80 s, with a relative difference of about 1.8%. The sudden changes of the difference taking place from 25 s to 30 s and during the ninetieth seconds may be caused by the rapid rotation of the pan tilt when tracking the ship.
Fig. 8. Time series of difference between calibration coefficients obtained from DOAS approach and self-calibration mothed.
Since the detector quantum efficiency at 310 nm and 330 nm is obtained from the product manual, small errors due to deviation from the actual value are possible. However, the self-calibration method represents a very promising technique for UV camera system, especially for ship SO2 plume monitoring, which has special need in real time performance and on-line calibration.
Certainly, the accuracy of the self-calibration method depends on the artificial generation of the background image, but it is difficult to quantify a priori its order of magnitude or constrain the error it may introduce. It is also important to mention that for the calculation of the average digital number of the background image, the instrumental vignetting may bring additional errors [12]. However, this perturbing influence could be avoided on the certain degree in the intensity ratio calculation.
5. Conclusions
In this research, we put forward a methodology of self-calibration for real-time monitoring of SO2 ship emissions with UV cameras, as we believe, for the first time. A theoretical basis for the pertinent aspects of self-calibration is given in detail, including the ultraviolet atmospheric radiative transfer, the physical relationship between SO2 optical density, the SO2 column density, and the image intensities of the background sky. The fact that the parameterization of this relationship is nearly not impacted atmospheric conditions is proved theoretically, which establishes the theory foundation for self-calibration of the SO2 camera systems. Based on the one-to-one correspondence relationship between the ratio of the SO2 column density to SO2 optical density and the intensity ratio of the background sky images at 310 nm and 330 nm, the calibration coefficient can be obtained without changing the viewing direction or adding any additional equipment (such as cells or spectrometers). For purpose of verify the effectiveness and accuracy of the proposed method, we conducted field experiments near Shanghai Port, where SO2 emissions from ships at sea were made. We compare the respective results of the self-calibration mothed and the DOAS approach. Calibration errors between them are effectively reduced to 1.8% after filter transmittance corrections. Compared with the traditional calibration approach, this new method is more convenient and practicable, especially for ship plume monitoring. Predictably, this promising technique may bring great improvement on the performance for SO2 camera system and become the standard calibration mothed for UV imaging measurements of SO2 emission in the near future.
Funding
National Key Research and Development Program of China (2017YFC0211900); National Natural Science Foundation of China (41975039, 61705253).
Acknowledgments
Prof. Shunsheng Gong is acknowledged for his suggestion and encouragement.
Disclosures
The authors declare no conflicts of interest.
References
1. T. Mori and M. Burton, “The SO2 camera: A simple, fast and cheap method for ground-based imaging of SO2 in volcanic plumes,” Geophys. Res. Lett. 33(24), L24804 (2006). [CrossRef]
2. T. Mori and M. Burton, “Quantification of the gas mass emitted during single explosions on Stromboli with the SO2 imaging camera,” J. Volcanol. Geotherm. Res. 188(4), 395–400 (2009). [CrossRef]
3. G. Tamburello, E. P. Kantzas, A. J. S. McGonigle, A. Aiuppa, and G. Giudice, “UV camera measurements of fumarole field degassing (La Fossa crater, Vulcano Island),” J. Volcanol. Geotherm. Res. 199(1-2), 47–52 (2011). [CrossRef]
4. G. J. S. Bluth, J. M. Shannon, I. M. Watson, A. J. Prata, and V. J. Realmuto, “Development of an ultra-violet digital camera for volcanic SO2 imaging,” J. Volcanol. Geotherm. Res. 161(1-2), 47–56 (2007). [CrossRef]
5. J. F. Smekens, M. R. Burton, and A. B. Clarke, “Validation of the SO2 camera for high temporal and spatial resolution monitoring of SO2 emissions,” J. Volcanol. Geotherm. Res. 300, 37–47 (2015). [CrossRef]
6. M. Osorio, N. Casaballe, G. Belsterli, M. Barreto, A. Gomez, J. A. Ferrari, and E. Frins, “Plume Segmentation from UV Camera Images for SO2 Emission Rate Quantification on Cloud Days,” Remote Sens. 9(6), 517 (2017). [CrossRef]
7. A. J. Prata, “Measuring SO2 ship emissions with an ultraviolet imaging camera, Atmos,” Meas. Tech. 7(5), 1213–1229 (2014). [CrossRef]
8. M. Boichu, C. Oppenheimer, V. Tsanev, and P. R. Kyle, “High temporal resolution SO2 flux measurements at Erebus volcano, Antarctica,” J. Volcanol. Geotherm. Res. 190(3-4), 325–336 (2010). [CrossRef]
9. G. G. Salerno, M. R. Burton, C. Oppenheimer, T. Caltabiano, V. I. Tsanev, and N. Bruno, “Novel retrieval of volcanic SO2 abundance from ultraviolet spectra,” J. Volcanol. Geotherm. Res. 181(1-2), 141–153 (2009). [CrossRef]
10. A. Wiacek, L. Li, K. Tobin, and M. Mitchell, “Characterization of trace gas emissions at an intermediate port,” Atmos. Chem. Phys. 18(19), 13787–13812 (2018). [CrossRef]
11. P. Luebcke, N. Bobrowski, S. Illing, C. Kern, J. M. Alvarez Nieves, L. Vogel, J. Zielcke, H. Delgado Granados, and U. Platt, “On the absolute calibration of SO2 cameras,” Atmos. Meas. Tech. 6(3), 677–696 (2013). [CrossRef]
12. E. P. Kantzas, A. J. S. McGonigle, G. Tamburello, A. Aiuppa, and R. G. Bryant, “Protocols for UV camera volcanic SO2 measurements,” J. Volcanol. Geotherm. Res. 194(1-3), 55–60 (2010). [CrossRef]
13. C. Kern, F. Kick, P. Luebeck, L. Vogel, M. Woehrbach, and U. Platt, “Theoretical description of functionality, applications, and limitations of SO2 cameras for the remote sensing of volcanic plumes,” Atmos. Meas. Tech. 3(3), 733–749 (2010). [CrossRef]
14. M. P. Dalton, I. M. Watson, P. A. Nadeau, C. Werner, W. Morrow, and J. M. Shannon, “Assessment of the UV camera sulfur dioxide retrieval for point source plumes,” J. Volcanol. Geotherm. Res. 188(4), 358–366 (2009). [CrossRef]
15. C. Kern, P. Luebcke, N. Bobrowski, R. Campion, T. Mori, J.-F. Smekens, K. Stebel, G. Tamburello, M. Burton, U. Platt, and F. Prata, “Intercomparison of SO2 camera systems for imaging volcanic gas plumes,” J. Volcanol. Geotherm. Res. 300, 22–36 (2015). [CrossRef]
16. R. Kullmer and W. Demtroder, “Vibronic coupling in SO2, and its influence on the rotational structure of the bands in the 300-330 nm region,” Chem. Phys. 92(2-3), 423–433 (1985). [CrossRef]
17. P. Wang, K. Y. Liu, T. Cwik, and R. Green, “MODTRAN on supercomputers and parallel computers,” Parallel Comput. 28(1), 53–64 (2002). [CrossRef]
18. C. Kern, C. Werner, T. Elias, A. J. Sutton, and P. Luebcke, “Applying UV cameras for SO2 detection to distant or optically thick volcanic plumes,” J. Volcanol. Geotherm. Res. 262, 80–89 (2013). [CrossRef]
