UNIVERSITY OF CALIFORNIA, SAN DIEGO 3 1822 04429 0690 UC SAN DIEGO LIBRARY UNIVERSITY OF CALIFORNIA, SAN DIEGO Offsite (Annex-JoIOSPHERIC OPTICS GROUP rnals) HNICAL NOTE NO. 259 3 1822 04429 0690 Version 1.1 Jul 2003 974.5 T43 no. 259 Daylight Visible/NIR Whole Sky Imagers for Cloud and Radiance Monitoring in Support of UV Research Programs For publication in Proceedings of SPIE Vol. 5156 Ultraviolet Ground- and Space-based Measurements, Models, and Effects III, SPIE, Bellingham, WA UNIVERSITY OF CALIFORNIA SAN DIEGO J. E. Shields R. W. Johnson M. E. Karr A. R. Burden J. G. Baker ...... . . . .. ARSI . oo.... ......... CAUN FORN ONIA Songs ees... 1888 08cee be The material contained in this note is to be considered proprietary in nature and is not authorized for distribution without the prior consent of the Marine Physical Laboratory. SCRIPPS INSTITUTION OF OCEANOGRAPHY MARINE PHYSICAL LAB San Diego, CA 92152-6400 UC SAN DIEGO LIBRARY UNIVERSITY OF CALIFORNIA, SAN DIEGO 3 1822 04429 0690 Daylight Visible/NIR Whole Sky Imagers for Cloud and Radiance Monitoring in Support of UV Research Programs For publication in Proceedings of SPIE Vol. 5156 Ultraviolet Ground- and Space-based Measurements, Models, and Effects III, SPIE, Bellingham, WA J. E. Shields, R. W. Johnson, M. E. Karr, A. R. Burden and J. G. Baker Efi le 片 ​... ·全国 ​. r. . . … - - --. .. ... . .… …….. ver . . . r: # r* a" ,. …. +/- - - --- --- - -- .. . . . … . . ……. .. . . …. . . . , " " """ " .." .. . . .…... ''' '' ' ' Daylight Visible / NIR Whole Sky Imagers for Cloud and Radiance Monitoring in Support of UV Research Programs Janet E. Shields*, Richard W. Johnson, Monette E. Karr, Art R. Burden, and Justin G. Baker Marine Physical Lab, Scripps Institution of Oceanography, University of California San Diego, 9500 Gilman Dr., La Jolla CA 92093-0701 ABSTRACT Measurements of UV radiation at the earth's surface may be highly impacted by the presence of clouds. In order to provide support for UV research, a Daylight Visible/NIR Whole Sky Imager was developed to provide cloud fraction assessment over the whole sky, as well as measurements of the radiance distribution over the full sky in several spectral bands. Radiances are determined in approximately 700,000 directions simultaneously with a given optical filter. Data may be acquired in seven spectral bands that may be selected for the application. The current instrument uses filters near 450 and 650 nm, open-hole, filters in the blue-green broadband and NIR long-pass, and two polarizers. Opaque and thin cloud fraction is determined from images acquired in the blue and red wavelengths. A more sophisticated version of the algorithm to detect thinner clouds and enable aerosol assessment is in development, and will be based on use of the NIR data in conjunction with the blue and red data. This paper will provide an overview of the instrument design and calibration, and sample sky radiance results. The cloud algorithms for determination of cloud fraction will be discussed, and the cloud imager results will also be presented. Key words: Cloud, radiance, calibration, UV, cloud fraction, sky radiance, whole sky imager 1. INTRODUCTION Whole Sky Imagers (WSI) are ground-based sensors which are used in UV research, global climate research, test site support, and other applications. They are designed to provide high quality digital imagery of sky conditions and, when combined with appropriate algorithms, provide assessment of cloud amount and location within the scene, absolute radiance distribution, and related atmospheric parameters. The Marine Physical Lab (MPL) at Scripps Institution of Oceanography has been very active in the development of WSI systems for the last two decades. The first WSI systems the group developed using digital imaging technology were deployed in the early 1980's, and were followed by fully automated systems in the mid to late 1980s (Johnson et al 1989, and Shields et al., 1993). WSI systems capable of full 24-hour autonomous operation for acquisition of day and night sky parameters were fielded in the early 90's, and have continued to evolve in capability (Shields et al 1998). The Deutsche Wetterdienst (DWD) has been very active in research related to solar radiation, biologically effective UV radiation, and a variety of atmospheric radiation research (Feister et al, 1998). Due to the long scanning times of many spectroradiometers measuring solar irradiance, the effect of cloud variability enhances the uncertainty in the spectral irradiance (spectral distortions) and of time integrals of broad-band irradiance such as UV-B and erythemal irradiance (Feister et al., 2003). Cloud flags for UV spectra have been derived from 1-minute pyranometer data as well a sunshine duration and 1-hourly cloud observations to classify the UV spectra according to typical 'optical' cloud conditions (Feister et al., 1998). Much more information can be gained from automated cloud imaging with the WSI, including cloud distribution, and the associated absolute radiance distribution. As a result of this need to characterize cloud fields during the solar radiation measurements including the UV region, a new Daytime WSI was developed by MPL, for use at DWD in the UV research program. The system was developed to acquire high resolution images in several spectral bands in the visible and NIR wavelengths, and to acquire imagery * ishields@ucsd.edu; phone (858) 534-1769; fax (858) 822-0665 suitable for the automatic determination of clouds within the images. The instrument is shown in Figure 1, in its environmental housing. It includes an automated solar occultor, designed to minimize stray light within the optics and thus enable acquisition of accurate radiometric data. The optics includes a fisheye lens, with nearly 180 degree field of view. A sample image, acquired near sunset with cirrus clouds overhead, is shown in Figure 2. The instrument evolved from earlier versions of the WSI, as discussed in Section 2. Its development, calibration, cloud algorithms, and current research at MPL will be discussed in the later sections of this paper. Figure 1: Daytime Visible/NIR WSI fielded at DWD Figure 2: WSI image acquired near sunset at MPL 2. THE DEVELOPMENT OF WSI SYSTEMS AT MPL The original concept for the Whole Sky Imagers at MPL evolved out of the group's Atmospheric Optics program, a measurement and modeling program using multiple sensors for monitoring sky radiance, atmospheric scattering coefficient profiles, and other parameters related to vision through the atmosphere (Johnson et al 1980). In particular, the first automated WSI was conceived as combining the features of the all-sky camera with the scanning radiometer systems that provided quantitative measurements of sky radiance distribution. Early systems were based on digital cameras (sometimes CCD, sometimes Charge Injection Device or CID systems), with fisheye lenses, optical filter changers, relay optics to provide the proper image size and location, equatorial sun occultors to provide shading for the lens, and early versions of personal computers for automated control. Figure 3 shows some of this evolution. The film- based all-sky camera in use in a 1963 deployment is shown in Figure 3a, and the automated Day-only WSI developed in the mid-1980's based on CID technology is shown in Figure 3b. With the use of very low noise 16 bit CCD cameras and an occultor modified to handle both sun and moon, these systems were further developed into the Day/Night WSI shown in Figure 3c. Some typical images from the Day/Night WSI are shown in Figure 4. To place the development of the Daylight Visible/NIR WSI into the context of related developments at MPL, Figure 5 shows other systems that evolved at MPL for other applications. These include a Day/Night WSI designed for remote sites and real-time processing, shown in Fig. 6a. This system was designed for applications in military test support. Figure 6b shows two cameras, a visible and NIR system designed for airborne use with a UAV (Shields et al, 2003). This system is being used for climate research. Figure 6c shows imagery from a new mockup for a full scene imager with optical zoom for surveillance and tactical applications. This full scene imager uses a unique type of optical zoom developed at MPL, for which the patent is pending. The general concept is to image the full scene with one or more fisheye lenses or other wide angle lenses. The image from the fisheye is then split with a beam splitter, and one of the image planes is inspected with a microscope objective and second camera, providing a very high resolution image of selected regions of interest. The choice of the region for high resolution view can be made by a motion detector or by a user with a touch screen. The Daylight Visible/NIR WSI is intermediate in complexity between the Day/Night WSI system shown in Figure 3c and the airborne fisheye imagers shown in Figure 5b. It is intended to be a viable and superior replacement for the older Day WSI shown in Figure 3b, which used components that are now largely outdated. CAMERA'S VEW ALL-SKY CAMERA ASSEMBLY (a) (b) (c) Figure 3: Some of the WSI Systems developed at MPL that contributed to the development of the Daylight Visible/NIR WSI: a) the All-Sky Camera used in 1963; b) the Day-only WSI used in the 1980's; c) the Day/Night WSI used in 1990's and currently in use. Figure 4: Sample imagery from the Day/Night WSI for sunlight, moonlight, and starlight conditions. (a) (b) (c) Figure 5: Additional Imaging Systems in the WSI family at MPL: a) Day/Night WSI for real-time processing; b) Airborne visible and NIR systems; c) A full scene imager with patent zoom 3. DESIGN OF THE DAYLIGHT VISIBLE / NIR WSI It was possible to simplify the design of the new Daylight WSI system, in comparison with the Day/Night WSI, because several features were not required for this application. It was not necessary to acquire imagery at night, resulting in relaxed performance requirements (and lesser costs) for the imager. Also, the thermal environmental requirements were somewhat more modest in comparison with the Day/Night system which must operate in extreme environments such as the Arctic, Tropics, and American Deserts. In addition, the solar occultor could be considerably simplified, because it did not need to shade the large lens required on the Day/Night WSI, it did not need to shade the moon, and it did not need to handle changes in orientation for ship-board applications. These reduced requirements enabled considerable savings in design and fabrication cost. On the other hand, it was still necessary that the full 2n hemisphere be imaged on a digital camera. A camera with approximately 1k x 1k pixel resolution was needed, and several spectral filter bands were desired. Thus we could not simplify this system as much as we simplified the airborne systems shown in Fig. 5b. Since radiometric calibration was desired, it was necessary to retain a solar occultor to minimize stray light in the image. Thus the basic application led to the design of an instrument which is intermediate in complexity and capability in comparison with other systems developed at MPL. One of the primary design characteristics of this system is the use of relay optics to transfer the image plane from the fisheye lens to the CCD sensor. This approach enabled us to include a dual-wheel filter changer behind the lens, and also acquire the full hemisphere view on the CCD imager of choice. The optical relay was also designed to limit the angle of the rays as they penetrated the interference filters to 5 degrees or less, thus minimizing the slight spectral shifting normally associated with interference filters and off-axis rays. This design retained a good modulation transfer function, so that resolution was limited by the camera pixel resolution, and not by the image line pair density. Considerable design time was required to verify that required image quality could be maintained with primarily off-the- shelf lenses and practical (cost-effective) lens mount machining tolerances. Measurements of the achieved point spread function with the as-built device demonstrated that this had been achieved. Radiometric studies of the existing Day/Night WSI data enabled the selection of a CCD camera with sufficient dynamic range, absolute sensitivity, and Signal/Noise characteristics to acquire the full range of sky images under normal daylight conditions with reasonable flexibility in spectral filter selection. The Photometrics SenSys 1600 CCD camera with 12- bit digitization was chosen for this application. Additional important features included the ability to use 3 different fixed gain settings, and the ability to acquire dark images, which are both very important in enabling absolute radiometry. The internal hardware with the lens mounts, relay optics, filter changer, and CCD camera system are shown in Figure 6. The filter passbands selected by the sponsor are shown in Figure 7. (This plot taken from Feister et al, 2000.) These include a red and blue filter, a broadband visible filter, a broadband NIR filter, and open-hole (camera response). In addition, the system can select either of two crossed polarizers. Several of the filters are buffered with neutral density filters to provide similar throughput, and the overall system response is also modified with a 2 log neutral density filter, which can be removed if very narrow spectral band measurements are desired. To shade the fisheye lens and minimize stray light, the equatorial occultor design used on the original Day WSI was modified to enable use at the higher latitudes now required. A camera housing was designed which provides a sealed and purged environment for the optics, yet enables cooling of the air-cooled CCD camera. The equatorial occultor and portions of the camera housing may be seen in Figure 8. The environmental housing shown in Figure 1 includes protection for the cooler/heater and the camera power brick. The control package (not shown here) includes the control computer and an electronics accessory control panel to allow control of the system either manually or via computer. The control computer includes a program for automated data acquisition. This program controls the camera, filter changer, and occultor, and also includes logic for determining an appropriate choice of gain and exposure for the current lighting conditions. Data are acquired and archived between sunrise and sunset. Normally a second system is used for processing the data in archival mode. BLUE RED POL BG39 RG850 CAMERA7046 TRANSMISSION OR RESPONSIVITY ni 300 400 900 1000 1100 500 600 700 800 WAVELENGTH (nm) Figure 7: WSI spectral filters and polarizers Figure 6: WSI relay optics, filter changer, and sensor Figure 8: WSI solar equatorial occultor and camera housing 4. CALIBRATION OF THE WSI FOR ABSOLUTE RADIANCE DISTRIBUTION In order to calibrate the imager for absolute radiance, one must first ensure that the data are appropriate for calibration. That is, the design of the sensor, and choice of purchased components, must be made so that raw data of sufficiently high quality are acquired. Probably the most important factor is the use of the solar occultor that shades the full lens and optical dome from direct sunlight. While internal blocking could be used to prevent blooming of the CCD, this would not be sufficient for radiometric purposes. One way to look at this is that we want essentially all of the scattered light that we sense to have been scattered by the atmosphere, not by the optics. The CCD imager's performance specifications also impact the radiometrics, and one nice characteristic of this camera is that the CCD is thermally stabilized, so that absolute sensitivities and dark levels are quite stable and repeatable. Sensor relative response, as characterized by the system linearity, turned out to be more of a challenge for this camera system than anticipated. We normally acquire linearity data in two ways, first with a fixed exposure setting and variable light levels from the optical bench (using FEL lamps traceable to NIST), and secondly with a fixed input flux level but variable exposures. With the Day/Night WSI sensor, these two approaches yield consistent results, with linearities of better than one percent over most of the range, and 2 – 3% on the extremes of the range. Although similar results were obtained with the SenSys camera, the calibration was somewhat more complex. Both systems have a mechanical shutter, and with the SenSys camera the relationship between the initiation of the opening of the mechanical shutter, the start of the electronic integration, the end of electronic integration, and the release of the mechanical shutter are determined by user-control variables. Even when these control variables are optimized, the resulting system sensitivity behaves in a slightly complex manner as a function of exposure time and flux level. As an example, Figure 9 shows the dark- corrected data from a linearity measurement acquired varying gain and exposure settings, with at fixed lamp position (for each gain). Plotted in this format, the data appear reasonably linear. However, if the percentage deviation with respect to a linear response is plotted, as shown in Figure 10, it becomes apparent that significant errors in calibrated radiance would result if the linearity were not properly measured and corrected. Extensive measurements were taken, in order to de-couple the impact of the exposure setting and the impact of changes in the flux level impinging on the CCD. The best correction was found to be an additive correction to exposure time, and use of a spline fit to account for the remaining non-linearity of the system. With these corrections, remaining uncertainties which could be attributed to non-linearity dropped to 0.2% or less. To calibrate the sensors for absolute sensitivity, several measurement sets were required. First the relative transmittance of each filter, and the relative sensitivity of the CCD, were measured. These values are used with the absolute spectral irradiance of each lamp to compute the effective calibration lamp irradiances in the WSI passbands using Equation 1. 4000 FT - - - - - - 3000 € - - - - - - Signal 2000 TTTTTTTTTTTTTT - - - - - - - 1000 - - Gain 1 Gain 2 - - - Gain 3 ..................................................................... 5.-.-..- 2000 4000 6000 8000 Exposure Signal vs Exposure Figure 9: System Linearity for 3 gain settings; each gain setting acquired with a fixed lamp position and variable exposure 1.101 . . . ..... . . . . . wroti Gain 1 Gain 2 Gain 3 - - % Non-linearity I 1.00 - - do 0.95 0 . 90 یییییییییللییییییییلییییللللیییییلللا 4000 0.90 0 1000 2000 3000 Signal % Non-linearity with Eo=0, So=0 Figure 10: Resulting Percent Non-linearity for the data shown in Figure 9 - SE S Tz 1... Tanda ssa Tal... Tanda Eq. 1 In Equation 1, we define E = Effective lamp irradiance Ea = Lamp spectral irradiance Sa = Sensor spectral sensitivity or responsivity Ta = Spectral transmittance for filters 1 through n Next, measurements were acquired with a 3-meter calibration bar, using FEL lamps traceable to NIST, as well as FEL lamps calibrated by our sponsors. The lamps are directed at a lambertian plaque in order to provide a known radiance to the sensor. The lamp position was varied such that seven measurements over a 1-log range could be acquired. That is, approximately 7 measurements were taken with flux levels varying by a factor of 10. These signals should result in redundant calibration constant determinations. One measure of the uncertainty in the calibrations is the percentage STD obtained between the seven computed calibration constants from the above 7 measurements in each filter. For the red and visible filters, we achieved self- consistencies of 0.1% and 0.2% in those measurements taken in Gain 2. The blue filter was measured in Gain 3, near the low end of the scale, and for this case the STDs were not as good, with uncertainties of about 2% due to the effects of the linearity in Gain 3. Results were also slightly degraded for those passbands such as NIR and open whole which included significant amounts of NIR energy. We appeared to have some stray “light” in the NIR within the room, resulting in STD's of about 2-3% in those filters affected by the NIR. (We typically attain self-consistencies of 0.1% to 0.2% with the Day/Night WSI, however it has NIR light beyond 900 nm blocked, and only uses NIR filters near 800 nm.) Absolute calibration measurements were taken with three lamps, calibrated at three different locations. The comparison of the results from the 3 lamps was very reasonable. All measurement results were within the 2-3% accuracy expected with FEL lamps. In comparison with the lamp that was chosen by the sponsors as the primary source, one of their lamps yielded results about 1.5% to 2% higher, and the MPL lamp traceable to NIST lamp yielded results about 1% lower. Additional calibration measurements were acquired to characterize noise, stability, and other parameters potentially affecting performance. Three of these calibrations that should be mentioned briefly are the dark calibration, flat field calibration, and rolloff calibration. The dark image is the image acquired when the mechanical shutter is closed, but the thermally-generated electrons are collected for the normal collection period associated with a given exposure. The average dark current increased slowly as a function of exposure time, as anticipated. However, at long exposures, a few pixels had much higher dark current than the average (we assume due to crystal defects). While this type of response is expected for CCDs, we do not normally see it in the Day/Night WSI's. The Day WSI system has its CCD temperature stabilized at +10C, as opposed to the Day/Night CCD's, which are stabilized at 40C. At -40C, these spurious points are well suppressed. We found the high points in the dark images to be very repeatable, and we found that the dark correction (subtracting the dark image from the field image) was quite effective in removing them from the field imagery. Regarding the flat field calibrations, these calibrations are intended to measure variations in effective sensitivity of pixels. We found the flat field images for this system to be much more uniform than on the Day/Night CCD, because the Day WSI system uses relay optics, rather than a fiber optic taper, to convert the image size. While a fiber optic taper is has certain advantages in the design of a system, it does slightly degrade the uniformity of the system due to its physical structure; the relay system avoids this non-uniformity source. Regarding the rolloff calibration, this is the calibration that characterizes the change in effective sensitivity as a function of angle of incidence of the light. It is normally due to a combination of Fresnel losses, vignetting of the optics near the edge of the field of view, and most importantly, changes in the effective field of view. As will be discussed below, the field of view per pixel for the lens used in this system changed in such a way that the effective rolloff was quite nominal for this system. The calibration does include a correction for the rolloff, but it was fairly small except near the edge of the field of view of the system. It is perhaps worth noting that in the interim since this system was calibrated and delivered to the sponsor, we have developed an improved method of measuring flat field and rolloff using an integrating hemisphere, which results in improved ability to assess the radiance calibration near the horizon. The calibration also includes an angular calibration of the system field of view, in order to provide a mapping between object space (as characterized by zenith and azimuth angle) and image space (characterized by pixel position x,y). This system uses a Sigma 8 mm f4 fisheye lens, which provides sufficient throughput for a daytime system at a reduced cost in comparison with the lens used on the Day/Night WSI. The lens is approximately equi-distant, that is the relationship between object-space zenith angle and the fractional radius on the spherical image is close to linear. Figure 11 shows a comparison between a linear relationship, the Nikon lens used in the Day/Night WSI, and the Sigma lens used in the Daylight Visible/NIR WSI. The resolution per pixel depends on this curve; the result for the Sigma lens is shown in Figure 12. It can be shown that in a strictly linear lens, the solid angle per pixel drops off according to sin 0 / 0. In many ways, having a lens which is somewhat non-linear in zenith angle is useful, because it mitigates this drop in solid angle. The resulting change in solid angle per pixel is close to 0 in the Nikon lens, and the Sigma lens has a slight increase in solid angle per pixel as a result of this effect. As a result, the Sigma lens has very little measured lens rolloff as noted above. The angular calibration measurements are used to generate maps or equations of solid angle per pixel for the imagery. 90 - Sigma Lens ..... Nikon Lens - Linear Curve WSI VIS/NIR 7 0.30 - - - 0.1421 +0.0005319*x + 8.125E-006*x2 - - - - Angle (degrees) - - RESOLUTION (DEGREES PER PIXEL) - 0.05 0.00 0 10 20 30 40 50 60 70 ZENITH ANGLE (DEGREES) 80 90 0.0 RESOLUTION 0.2 0.4 0.6 0.8 1.0 Fractional Distance To Edge Figure 11: Zenith Angle Dependency of Sigma and Nikon lenses Figure 12: Resulting resolution per pixel for the Sigma lens Once all calibrations are completed, several calibration constant files are generated. These include the linearity corrections, rolloff corrections, absolute calibration constants, image size parameters, and related information. In archival post-processing, calibrations can be applied to the field images acquired in all spectral filters. In addition, the average radiance and the (cosine-weighted) irradiance, integrated over selected regions of interest and the whole sky, may be computed. These data are being processed and collected for use in modeling studies, particularly related to determination of clear air and thin cloud optical depth from WSI data. Sample results for a series of sky conditions are shown in Table 1. Radiances are given in units of watt / 12 m² um. Table 1 Sky Radiances at the Zenith extracted for several cases Case Solar Solar Sky Zenith Blue Vis Red NIR Open # Zen Azi Cover Cover 36 154 Clr Clr 69.4 | 40.1 23.0 6.73 | 14.2 34. | 178 Clr 69. 3 39.4 | 22.3 6.63 | 14.5 33 | 178 178 | Scat | Scat | 144 | 97.0 74.7 18.3 45.7 | 191 Scat Bkn | 206 155 154 | 41.9 111 34 34 152 Ocst Ocst | 25.1 | 14.9 20.3 8.80 14.3 81 1 212 Scat Thin | 26.7 15.1 10.2 3.80 6.92 84 218 Ocst Thin 20.9 12.5 8.48 3.81 16.07 In Table 1, looking at the clear cases (1 and 2), the radiances decrease with increasing wavelength, as anticipated for Mie scattering. The open hole radiance is an intermediate value, as anticipated. The radiances are much higher in cases 3 and 4, which include clouds within the region of interest. Case 5 shows the drop expected with overcast. Cases 6 and 7 are lower in spite of thin clouds due to the low winter solar zenith angle. Thus the relative changes in the radiances appear quite reasonable. Tables 2 and 3 show comparisons with previous data acquired by our group. Table 2 shows clear cases, and Table 3 shows overcast cases (where blue radiances did not happen to be available). While this is too limited a data sample to be a rigorous check, it does provide a "sanity” check. The first data set in each table is extracted from an extensive set of data acquired from a C-130 aircraft using a scanning radiometer using a photomultiplier system (Johnson et al 1980). The second data set in each table is extracted from Day/Night WSI data. The sun angle varied in the cases in Table 2, so data were selected for regions of interest with a 60 degree scattering angles with respect to the sun. The Day WSI data appear slightly high in the clear case, and slightly low in the overcast case. Although we have not made any kind of extensive data comparison, these data were not screened in any way, and were selected on the basis of similar sky conditions in the conveniently available data sets. They appear to show a reasonable comparison. Table 2 Comparison with C-130 and D/N WSI Data under Clear Skies Blue Red Radiance Radiance 35 9.0 Instrument Solar Zen C-130 Flt 379 T30 Photomultiplier Day/Night WSI 9 60 Jan 97 SGP Day WSI 34 134110000 30-32 | 7.1-7.5 - 47 14 Table 3 Comparison with C-130 and D/N WSI Data under Overcast Skies Instrument ROI Solar Zen 43 Red Radiance 60-70 Zenith C-130 Flt 422 Photomultiplier Day/Night WSI 13 Nov 97 SGP Day WSI 00143100000 55 Zenith 40-50 34 Zenith 21 It should be noted that a full calibrated image constitutes approximately 700,000 simultaneous measurements of radiance in all directions at approximately 1/5 degree spatial resolution. These data can be acquired routinely in all 7 spectral bands. Automated processing programs were written for handling this data. The large data amount is both a strength and a weakness of the system, since it is difficult to handle the sheer bulk of numbers that can be generated. Additionally, the processing program was written utilizing the image processing program provided by the camera vendor, and this vendor product has been somewhat problematic in terms of memory handling and associated stalls in processing. However, the vendor has worked with us in alleviating the worst of these problems, and the data processing is routinely continuing by the sponsor. 5. CLOUD ALGORITHMS FOR CLOUD FRACTION Although a full discussion of the cloud algorithm is beyond the scope of this paper, this section provides an overview of the algorithm. The cloud algorithm is designed to identify the presence of opaque and thin clouds on a pixel by pixel basis. The cloud algorithm includes several steps, as discussed below. se promenade of persone che non sono present per a. The red and blue images are dark-corrected, and linearity and exposure corrections are applied. Then a red/blue ratio image is derived. b. Portions of the image which are expected to be blocked by the occultor or be outside the optical image are given a value of 0, or “no data”. This calculation is based on the occultor arm position readout. With the Day WSI system, a series of 7 occultor arms of varying length are used as the solar declination changes during the year. Although there are designated days to change the arms, the occultor was built with enough tolerance that the occultor arms could be changed about 2 days before or after the nominal change date. For this reason, the processing program also requires input files which identify when the occultor arm was actually changed, as well as any other changes in the geometry such as mechanical offsets in the arm position. b. The opaque cloud decision is based on a fixed threshold in the red/blue image ratio. We have found that for opaque clouds, the red/blue ratio is significantly enhanced in comparison with the clear sky to provide a good detection mechanism. Additionally, this approach does not normally lead to biasing as a function of solar zenith angle or look angle, and appears to be relatively robust over a variety of cloud types. c. For thin clouds, we found that the fixed red/blue ratio threshold was not effective. The thin clouds, rather than having a fixed spectral signature, appear to act more as a perturbation with respect to the clear sky spectral signature. The clear sky red/blue ratio tends to increase near the horizon, and in the solar aureole. We found that the thin clouds were much better characterized on the basis of the ratio between the red/blue ratio and a clear sky red/blue ratio for the same sun angle and look angle. d. Although the magnitude of the clear sky red/blue ratio distribution tends to increase with increasing haze amount, and increase at low sun altitudes, the shape of the distribution is relatively invariant as a function of season and haze amount. To create a clear sky library, we normalize clear-day red/blue ratio images with respect to the ratio at the 45-45 point. This is the point at 45 degree scattering angle from the sun, and 45 degree zenith angle. There are actually two such points in the sky, and the average of them is used to normalize the clear sky ratio image. Data are extracted for a series of clear days throughout a year. The result is a library of anticipated normalized ratios, which are a function of both solar zenith angle and look angle. e. To assess a given field image for thin clouds, after the no-data regions are identified and the opaque clouds are identified, we extract the appropriate data from the library, and correct for the normalization. The normalization factor correction depends on the solar zenith angle. Ideally, it should also depend on haze amount, but this is not included yet, as will be discussed below. The resulting clear sky ratio image is compared with the field ratio image; in regions where the field ratio image exceeds the clear sky ratio image by more than 20%, the pixel is identified as thin cloud. That is, the pixel is identified as thin cloud if the spectral signature is perturbed from the nominal clear-sky spectral signature by more than a given amount. In the resulting image, the clouds that are identified as optically opaque are colored white, and those that are identified as optically thin are colored yellow. Areas not identified as cloud are colored blue, and may correspond to either clear sky, or sky impacted by small-droplet aerosol (haze). Regions previously identified as no data are colored black. The coloring is also varied according to the magnitude of the red/blue ratio in the clear and opaque regions, and varied according to the magnitude of the red/blue ratio divided by the clear sky red/blue ratio in the thin cloud regions. By including this color variation, this enables the human viewing the image to make a visual assessment of whether the algorithm has done a reasonable job. Sample results for two skies are shown in Figure 13. Although the algorithm appears to do well for the test cases we have processed, we know that it has some difficulty distinguishing thin clouds from haze or aerosols. In order to address this concern, NIR filters were installed in the WSI systems. We first installed the NIR filters, in order to achieve a higher contrast between thin cloud and the background sky. The clouds, with scattering particles of a micron or more, tend to have nearly equal scattering coefficients as a function of wavelength through the visible region. In contrast, the background sky scatters preferentially in the blue, with the wavelength variance depending on drop size distribution and other parameters. In general, clear skies have greater wavelength dependency than hazier skies. Thus there should be a greater contrast between the clouds and the background sky in the NIR than is observed in the red image. Similarly, there should be a greater contrast between the clouds and the background sky in the NIR/blue ratio than there is in the red/blue ratio. WSI POTSDAM APRIL 7, 2002 11:30 UTC OPAQUE: 0.21 THIN: 0.15 Figure 13: Sample Cloud Decision Algorithm results. Figure 14 shows two ratio images for the same case, normalized to the same value at the zenith. The first ratio is the red/blue ratio, and the second is the NIR/blue ratio. Figure 15 shows another example of the NIR/blue ratio. Figure 14: A red/blue ratio, and a NIR/blue ratio for a thin cloud case Figure 15: Another NIR/blue ratio image We are currently in the process of upgrading the algorithm to use the NIR/blue ratio. This should not only allow the algorithm to detect thin clouds more accurately, but should also avoid the problem of falsely identifying enhanced aerosol haze as thin cloud. We have also noted that the NIR/red ratio tends to increase in regions with thin cloud, but decrease in regions of enhanced aerosol. We expect to evaluate this behavior and potentially make assessments of aerosol amount. There is also interest in developing algorithms to determine thin-cloud optical depth with this system. The Daylight Visible/NIR WSI system has performed reasonably well, and provides the ability to assess the general sky condition from visual images, as well as quantitative measurements of cloud cover and sky radiance distribution. We are pleased to have a viable replacement for the original Day WSI that served us well for so many years. This system has much improved image quality over the original Day system. As a result, we have been able to develop the capability of determining absolute radiance distribution as well as cloud fraction. With the NIR filter capability, cloud algorithm improvements are also in development. 7. ACKNOWLEDGEMENTS We would like to express our appreciation to the Deutscher Wetterdienst for their support of this work and their expert handling of the WSI. The WSI was first installed at Meteorologisches Observatorium Potsdam, and later at Meteorologisches Observatorium Lindenberg. We would especially like to thank Dr. Uwe Feister, for his help and support in scientific analysis, hardware analysis, and project coordination. 8. REFERENCES 1. R. W. Johnson, W. S. Hering, and J. E. Shields, Automated Visibility and Cloud Cover Measurements with a Solid State Imaging System, University of California, San Diego, Scripps Institution of Oceanography, Marine Physical Laboratory, SIO 89-7, GL-TR-89-0061, NTIS No. ADA216906, 1989. 2. J. E. Shields, R. W. Johnson, and T. L. Koehler, Automated Whole Sky Imaging Systems for Cloud Field Assessment, Fourth Symposium on Global Change Studies, American Meteorological Society, 1993. 3. J. E. Shields, R. W. Johnson, M. E. Karr, and J. L. Wertz, Automated Day/Night Whole Sky Imagers for Field Assessment of Cloud Cover Distributions and Radiance Distributions, Tenth Symposium on Meteorological Observations and Instrumentation, American Meteorological Society, 1998. 4. U. Feister and K. Gericke, Cloud Flagging of UV Spectral Irradiance Measurements, Atmosph. Res. 49, 115 – 139, 1998. 5. U. Feister, A. Kaifel, R. Grewe, J. Kaptur, O. Reutter, M. Wolhfart and K. Gericke, First Performance Results of Two Novel Spectroradiometers developed for Fast Scanning of Solar Spectral UV Irradiance, International Symposium on Optical Science and Technology, SPIE the International Society for Optical Engineering, 2003 6. R. W. Johnson, W. S. Hering, J. I. Gordon, B. W. Fitch, and J. E. Shields, Preliminary Analysis and Modeling Based Upon Project OPAQUE Profile and Surface Data, University of California, San Diego, Scripps Institution of Oceanography, Visibility Laboratory, SIO Ref. 80-5, AFGL- TR-0285, NTIS No. ADB-052-1721, 1980. 7. J. E. Shields, R. W. Johnson, M. E. Karr, A. R. Burden, and J. G. Baker, Calibrated Fisheye Imaging Systems for Determination of Cloud Top Radiances from a UAV, International Symposium on Optical Science and Technology, SPIE the International Society for Optical Engineering, 2003. 8. U. Feister, J. Shields, M. Karr, R. Johnson, and M. Woldt, Ground-based Cloud Images and Sky Radiances in the Visible and Near Infrared Region from Whole Sky Imager Measurements, EUMP31, EUMETSAT Satellite Application Facility Workshop, Dresden Germany, 20-22 November 2000, 79-88.