The original NIR film filter with Lmax > 720 nm was made by overlaying of 7 ... campaigns in riverine and estuarine ecosystems in the Sea of Azov region on 19.
10 Photogrammetry and remote sensing
MEASUREMENTS OF CHLOROPHYLL-A IN VIVO FLUORESCENCE AND WATER REFLECTANCE BY CONSUMER DIGITAL CAMERAS FOR REMOTE SENSING DATA VERIFICATION Dr. Vasiliy Povazhnyy1, Andrey Povazhnyy 2, Anna Zolotareva1 1 2
South Scientific Centre of RAS, Russia Multicharts LLC, Russia
ABSTRACT Two modifications of consumer digital cameras were proposed, allowing for quick assessments of chlorophyll-A concentration in case II waters during the ground truth campaigns. To measure chlorophyll-A in vivo fluorescence, the images in highsensitivity mode (ISO 6400) were taken through red (> 650 nm) glass filter with flashlight excitation through blue glass (450 nm) filter in dark chamber of 1l-volume. The digital brightness of selected part of JPEG-processed images was recorded in red channel and the sensitivity up to 1,0 ug/l was reached with inexpensive device. To measure water reflectance in red and NIR regions for the satellite NIR-algorithms verification, the images of the water surface were taken in normal mode (ISO 100) through polarizing filter, equipped with red (>620 nm) and NIR (>730 nm) film filters. The white plate for standard reflectance was pictured on every scene. The digital brightness for the red and NIR parts of JPEG images was recorded in red channel, the NIR values were subtracted from red ones to yield the reflectance in 620 – 730 band and NIR to red reflectance relation was compared with chlorophyll-A and TSS values in situ. The good agreement with TSS and Secchi depth data was obtained for NIR reflectance, while the chlorophyll assessments proved to be reliable for high (> 30 ug/l) summer bloom concentrations. The progress of consumer digital cameras allows for future development of reliable cloud-based algorithms for ground truth and water quality monitoring. Keywords: chlorophyll-A, fluorescence, reflectance, digital camera INTRODUCTION Satellite estimations of chlorophyll-A (chl-A) in turbid productive case II waters were challengeable for a long time because of their optical complexity. The «blue-togreen» algorithms have severely overestimated chl-A in these waters due to colored dissolved organic matter (CDOM) and total suspended solids (TSS) interference. As the narrow red and near infra-red (NIR) -channels were introduced to MODIS and other modern scanners, the new algorithms for local water bodies were successfully developed to overcome the problem. Up to date the universal NIR/red algorithm is hardly found, which urges for more ground truth data for case II waters [1]. Determination of chl-A in vivo fluorescence (F) allows for express assessment of chl-A concentration in water, omitting the filtration and extraction stages. The typical fluorometer for F determination uses broadband blue light (400-500 nm) excitation and
16th International Multidisciplinary Scientific GeoConference SGEM 2016
reads the F signal in red (c.a. 670 nm) spectral region. These instruments are routinely applied in oceanographic profiling systems and moorings, as well as for inland water bodies’ monitoring. The F registration method is known to have several draw-backs, including CDOM and TSS interference, self-shading in high phytoplankton concentration or large-volume samples, physiological state- and species-specific F output. The periodic calibrations with standard chl-A measurements are usually carried out to account for local environmental variations. The proper calibrated fluorometer for F registration is the unmatched tool for high-resolution ground truth campaigns [2]. Water reflectance (R) measurements provide the opportunity to get the ground truth data with the remote ship- or air-born instrument at daytime. Several models of spectrometers were introduced to assess chl-A, CDOM, TSS and other water environmental parameters by R measurements in different spectral regions. Two classes of compact spectrometers to measure R were proposed. The first class of instruments uses the adjusted diffraction grid to scan the measured object with 2 – 5 nm spectral resolution. The second class of devices obtains the data simultaneously in several fixed spectral regions via broadband (15 – 50 nm) filters. R measurements with certain assumptions are identical to satellite data except for atmospheric correction and could be directly used in optical models for local water environments and ground truth [3]. The instruments for express assessments of chl-A by F and R are still barely found in many scientific institutions of developing countries, with traditional laborious methods of chl-A, CDOM and TSS measurements prevailing. This limits the amount of data, available for ground truth in case II waters and motivates the development of robust and affordable alternative for modern fluorometers and spectrometers. The modern digital camera could be considered as a measuring tool, processing the analog signal of complementary metal-oxide-semiconductor (CMOS) sensor into the digital product. The final digital image is most frequently obtained by splitting the total light field into three (red, green and blue) broadband color channels by Bayer filter, mounted over the CMOS sensor, and interpolating the raw data by JPEG algorithm. The final product presents the set of pixels, with digital brightness units (varying from 0 to 255) in three channels, calculated for each pixel. In spite of deep processing of initial color and brightness data of the scene, the RGB JPEG-compressed image to some extent is capable to represent the initial data on different narrow parts of visible and NIR spectral brightness with additional filters added to the device and proper calibration made. Most of the consumer class cameras’ CMOS sensors are to some extent sensitive to NIR radiation, which allows for NIR and red bands selective registration on the image. The recent advances in camera sensors’ sensitivity and signal-to-noise ratio allow for registration of extremely low brightness signal even with inexpensive devices. The aim of the study was to develop and preliminary calibrate the methods for chl-A in vivo fluorescence and water reflectance measurements with consumer digital cameras, used as fluorometer and spectrometer respectively for ground truth support of satellite data. MATERIALS AND METHODS The simple modification of consumer digital camera was proposed to allow for automatic registration of F signal with significant sensitivity. The waterproof model
10 Photogrammetry and remote sensing
Coolpix AW120 (Nikon Corp., Japan) was selected due to high light sensitivity and GPS module presence. To excite F in the sample, the flashlight of the camera was covered with blue glass filter (maximum transmittance L max= 450 nm, bandwidth at 50 % transmittance L1/2 = 80 nm), while the lens of the camera was covered with deep red glass filter (Lmax > 670 nm) to reduce the noise signal from the flashlight. The sensitivity of the camera was set to maximum level (ISO 6400), the resolution of the image was set to maximum and the white balance settings were always kept at “cloud sky” preset. The camera was mounted in a dark box, optically connected to black measuring flow cell of c.a. 1l-volume. Special attention was payed to avoid flashlight blinking on optical surfaces and chamber walls. The automatic data acquisition and processing procedures were applied in the installation. The camera was connected to PC via USB 2.0. The open code program digiCamControl [4], written in C#, was used for remote control of camera’s performance. The program was in turn controlled by original script, written in Python. In automatic operation mode the original script launched the digiCamControl in fixed periods of time to obtain the image. The image was then automatically found in autosave folder and processed with the help of open-source Python Image Library subscripts [5]. The original script was cropping the image with sub-scripts by the pre-set mask, and the average digital brightness in red channel was calculated for the selected part of the image. The measured values, together with selected data from EXIF file were automatically transferred via GSM connection to Internet server. The original image was saved on the hard drive. Instead of automatic image acquisition and processing, the F values can also be obtained by the described camera modification and measured manually by any available image-processing program. Similar modifications were made to utilize the digital camera as a spectrometer with broadband fixed spectral intervals. The consumer class camera Finepix S6500 (Fuji Photo Film Co., Japan) was used due to compatibility with 58 mm screw glass filters. The original NIR film filter with L max > 720 nm was made by overlaying of 7 layers of automotive tint film with 10 % transparency, known to be transparent for NIR. The red film filter for 3D movie glasses was used as a broadband red filter (L max > 630 nm). To increase the NIR to red sensitivity of the device, two layers of 10 % tint film were placed over the red filter. The filters were cut into half-circles and placed onto glass UV-C filter (Hoya, Japan). The whole set of filters was screwed onto camera lens together with polarizing glass filter (Fujimi, Japan) to reduce glare influence to measured water reflectance. The camera was operated manually in “P” (Program AE) mode with normal sensitivity and “cloud sky” white balance settings. To obtain the information on water reflectance in the selected bands, images were taken from board of the ship with camera facing downwards at 45° angle. The broad white plate was pictured in the lower part of every scene to be present on both “red” and “NIR” parts of the image. That was made to account for actual spectral variations of the incident sunlight. The digital brightness of the selected “white plate” and “water” parts of images was measured manually in Photoshop CS3 package (Adobe). The R were calculated for the NIR channel as the relation of “water” to “white plate” digital brightness values. The R values for red channel were calculated similarly with the
16th International Multidisciplinary Scientific GeoConference SGEM 2016
additional subtraction of NIR values from red ones, as the broadband red film filter proved to be transparent for NIR as well. The calibration of the described instruments was made during two spring field campaigns in riverine and estuarine ecosystems in the Sea of Azov region on 19 stations. The water temperature ranged from 12 to 16 ° C, the salinity ranged from 0,5 to 14 PSU and the Secchi depth values (as a proxy for TSS content) varied from 0,4 to 3 m. The spring phytoplankton community was dominated by diatom taxa, with Sceletonema costatum (Grev.) Cleve, 1873 prevailing in most of the estuarine samples. The fluorometer was installed into the ship pumping system, with the intake located at 0,7 m depth, while the water samples for calibration were collected by bucket from the surface layer. The records of the instrument, made at the exact time of water sampling, were used for calibration. To obtain “zero” values for the fluorometer, it was removed from the pumping system and filled with filtered (GF/F) sea water. The chl-A samples were filtered immediately onto GF/F filters, grinded and extracted in 90% acetone for 24 h in dark cool (+4 °C) environment. The extracts were measured by preliminary calibrated laboratory LED fluorometer after centrifugation with no acidification made to obtain chl-A + pheophytine-A values. Several additional measurements were made for spectrometer to get the idea of instrument’s range. Measurements were made in shallow productive turbid pond with extremely high chl-A concentrations (100 – 850 ug/l). The chl-A values were obtained by the same fluorometric method with necessary dilution made. RESULTS AND DISCUSSION The chl-A, measured in the Don river and the Sea of Azov during the field campaigns ranged from 1,2 to 34 ug/l. Good agreement was obtained for F and chl-A values in a wide range of environmental conditions (fig. 1).
Fig. 1. Calibration results for the designed fluorometer, obtained in case II waters.
10 Photogrammetry and remote sensing
The rather big flow cell, used in the installation, have influenced the character of the calibration curve obtained. The non-linearity of the curve at high chl-A concentrations appeared to be the result of self-shading and TSS influence on the F signal. At low chl-A conditions the F to chl-A ratio was c.a. 16 (brightness units/ug chlA). The typical SD values, obtained for the fluorometer, ranged from 1,5 to 2,5 (n=10). Thus the minimum detection limit of the instrument, calculated as triple SD, could be as low as 0,33 ug/l in low chl-A and low TSS waters. The typical range of chl-A concentrations in the Sea of Azov is 10 – 15 ug/l. The F to chl-A ratio in this range is c.a. 10, so the minimum detection limit of the instrument rises to c.a. 1 ug/l, still sufficient to detect the main peculiarities of chl-A distribution. As the F output for different taxonomic groups of algae was shown to vary significantly, the obtained calibration will hardly represent chl-A in summer cyanobacterial bloom conditions, and another calibration will be needed. Yet, with the automatic data acquisition and processing the in vivo fluorometer proved to be reliable, easy-to-get and sensitive instrument for the case II waters. The principle way of F measurements could easily be scaled both to professional cameras to yield high-sensitivity devices, and to mass market of smartphone cameras, creating the potential of broader communities’ involvement to water quality monitoring. With the online services provided for data processing, the big datasets would be created to help for better ground truth and expand the spatial and temporal resolution of existing chl-A data in case II and inland waters. The R values, obtained during two sampling campaigns, were analyzed to get proxies for Secchi depth (S) and chl-A. The process of remote data acquisition is influenced by more variable factors, then the contact measurements (e.g. F). They are water glare, ship position at the moment of data acquisition, angle of the incident light (time of day), clouds’ influence on the incident light spectrum etc. That makes the assessments, based on R, of less predictive power, then the results of contact measurements. The relation of NIR upcoming radiation (R NIR) to the S is shown on fig. 2.
16th International Multidisciplinary Scientific GeoConference SGEM 2016
Fig. 2. Relation of upcoming radiation in NIR (R in NIR), measured with designed spectrometer, to Secchi depth (S). Note the log scale for S. It is obvious that the variation of RNIR was too great to make the solid predictions of S, based on the mentioned measurements only. Data acquisition at different Solar angles could be one of main reasons for R’s robust performance. Definitely more data is needed to better understand the possibility of R NIR use for S assessments. The relation of RNIR to water reflectance in red band (Rred), normalized for S (measured by RNIR as well) to actual chl-A is shown on fig 3.
Fig. 3. Calibration results of water reflectance in red and NIR bands, measured by designed spectrometer. Note the double log scale. The model of simple spectrometer, described in this paper definitely has some serious limitations for the application in low chl-A waters. Yet, the extremely high
10 Photogrammetry and remote sensing
concentrations of chl-A give the distinctive signal in RNIR to Rred relation, with the possible minimum to measure as low as 30 ug chl-A/l. As the instrument and optical model will be developing, the minimum detection limit of the design will become clear. CONCLUSION The automated monitoring of water quality and ground truth of environmental parameters is not possible without development of affordable and rather simple instruments. Two modifications of consumer digital cameras were proposed, allowing for quick assessments of chlorophyll-A concentration in case II waters during the ground truth campaigns. To measure chlorophyll-A in vivo fluorescence by digital camera, the images in high-sensitivity mode were taken through red glass filter with flashlight excitation through blue glass filter in dark chamber of 1l-volume. The digital brightness of selected part of JPEG-processed images was recorded in red channel and the sensitivity up to 1,0 ug/l was reached with inexpensive device. To measure water reflectance in red and NIR regions for the satellite NIR-algorithms verification, the images of the water surface were taken through polarizing filter, equipped with red and NIR film filters. The white plate for standard reflectance was pictured on every scene. The digital brightness for the red and NIR parts of JPEG images was recorded in red channel, the NIR values were subtracted from red ones to yield the reflectance in 620 – 730 band and NIR to red reflectance relation was compared with chlorophyll-A and TSS values in situ. The good agreement with TSS and Secchi depth data was obtained for NIR reflectance, while the chlorophyll assessments proved to be promising for high (> 30 ug/l) summer bloom concentrations. The modifications of consumer digital cameras to measure chl-A in fluorometer and spectrometer modes, described in the paper, could be considered as possible alternatives for standard instruments for case II waters studies. The potential exists for the cloud-based algorithms for mass measurements of chl-A to be developed as applications, utilizing smartphone cameras and creating big data on this valuable environmental parameter. ACKNOWLEDGEMENTS The research was supported by Russian applied research grant RFMEFI60714X0059 contract # 14.607.21.0059 REFERENCES [1] Moses, W. J., Gitelson, A. A., Berdnikov, S., & Povazhnyy, V. Estimation of chlorophyll-a concentration in case II waters using MODIS and MERIS data— successes and challenges. Environmental Research Letters, 4(4), 045005. 2009. [2] Karabashev G. S. Fluorescence in the Ocean: Diss. – Monterey Institute of International Studies, 1990. [3] McClure, W. F., David Moody, D. L. Stanfield, and Osamu Kinoshita. "Hand-held NIR spectrometry. Part II: An economical no-moving parts spectrometer for measuring chlorophyll and moisture." Applied spectroscopy 56, no. 6: 720-724. 2002 [4] http://www.digicamcontrol.com [5] http://www.pythonware.com