2000 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE
2000IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 9, SEPTEMBER 2005Cloud Statistics Measured With theInfrared Cloud Imager (ICI)Brentha Thurairajah, Member, IEEE, and Joseph A. Shaw, Senior Member, IEEEAbstract—The Infrared Cloud Imager (ICI) is a ground-basedthermal infrared imaging system that measures spatial cloudstatistics with a 320 240-pixel uncooled microbolometer detector array. Clouds are identified from the residual radiance thatremains after water vapor emission is removed from radiometrically calibrated sky images (the water vapor correction relieson measurements of precipitable water vapor and near-surfaceair temperature). Cloud amount, the percentage of an ICI imagecontaining clouds, is presented for data from Atmospheric Radiation Measurement (ARM) sites at Barrow, AK in February–April2002, Lamont, OK in February–April 2003, and Barrow, AK inMarch–April 2004. In Oklahoma, the percent cloud cover determined from full ICI images was slightly higher than that foundfrom a single-pixel time series, suggesting that cloudiness may beunder sampled by vertically viewing lidars or radars under highlyvariable conditions. Full-image and single-pixel statistics agreedmore closely for Arctic clouds, which tend to be uniform for longperiods of time. Good agreement is found in comparing cloudamount from ICI and active remote sensors during day and night,but much worse agreement is found between ICI and the ARMWhole Sky Imager during nighttime relative to daytime, indicatingthe importance of the diurnally consistent ICI measurements.Index Terms—Clouds, infrared imaging, infrared radiometry,terrestrial atmosphere.I. INTRODUCTIONCLOUDS fill an important role in maintaining the radiationbudget of the atmosphere by preventing incoming solar radiation from reaching the Earth and outgoing thermal radiationfrom escaping into space. This radiative impact of clouds, whichtend to warm or cool the Earth depending on their optical properties and spatial distribution, is a key component in the globalclimate [1]. Thus, understanding the spatial distribution and radiative properties of clouds can contribute to improved climateand weather predictions. Spatial cloud distribution also is a keyfactor in understanding and modeling cloud radiation feedbackmechanisms [2].While satellite sensors achieve global coverage, groundbased sensors provide improved radiometric contrast betweenManuscript received September 16, 2004; revised June 1, 2005. This workwas supported in part by the Communications Research Laboratory (Tokyo,Japan), in part by the National Oceanic and Atmospheric Administration ArcticResearch Office, and in part by the Office of Biological and Environmental Research of the U.S. Department of Energy as part of the Atmospheric RadiationMeasurement program.B. Thurairajah was with the Electrical and Computer Engineering Department, Montana State University, Bozeman, MT 59717 USA. She is now withthe Atmospheric Science program, University of Alaska, Fairbanks, AK 99775USA.J. A. Shaw is with the Electrical and Computer Engineering Department,Montana State University, Bozeman, MT 59717 USA (e-mail: jshaw@ece.montana.edu).Digital Object Identifier 10.1109/TGRS.2005.853716clouds and the background. This is especially true at high latitudes, where satellites have difficulty distinguishing betweenclouds and the underlying surface [3]. Other ground-basedcloud measuring instruments generally are either wide-angle,spatially resolving passive imagers or zenith-viewing activesensors. For example, the Total Sky Imager (TSI) [4] mea–nm from the entire skysures visible skylightdome during daytime, while the Whole Sky Imager (WSI)[5] provides hemispherical coverage with different detectiontechniques during day and night. The WSI has approximately70-nm wide imaging bands centered at 450, 650, and 800 nm,identifying clouds from red/blue ratios during the day andfrom star maps at night (with gaps in operation near sunriseand sunset). More complete diurnal coverage is provided byzenith-viewing active sensors, such as the Micro-Pulse Lidar(MPL) [6] or mm-wave cloud radar (MMCR) [7]. Unlessoperated in a scanning mode, the active sensors produce spatialcloud statistics by assuming that temporal statistics of clouds ata fixed location are equal to spatial statistics along the directionof the mean wind (Taylor’s Hypothesis [8]), which may nothold under conditions such as developing or dissipating clouds[9]. The combination of active sensors and passive imagingsensors may provide enhanced information about the verticaland horizontal cloud distribution, including cloud overlap thatcan dramatically alter the net radiative impact of the clouds(consider, for example, the difference between two overlappingcloud layers that completely obscure the sun and two nonoverlapping cloud layers that allow direct solar illumination of thesurface).The Infrared Cloud Imager (ICI) is a ground-based passivesensor that measures downwelling atmospheric radiance in the8–14 m wavelength band [10]–[14]. ICI data are used withprecipitable water vapor and air temperature measurements toidentify clouds and calculate cloud amount continuously withno difference in sensitivity during day and night. This systemwas developed as part of a Japan–U.S. joint venture to study theArctic atmosphere, led by the Japanese Communications Research Laboratory (CRL, which recently became the NationalInstitute of Information and Communications Technology,www.nict.go.jp). The ICI has been deployed at Poker FlatResearch Range (PFRR), Alaska in 2000–2001 [10] and at theDepartment of Energy’s Atmospheric Radiation Measurement(ARM) sites [16] in Barrow, AK in February–April 2002[10]–[14], Lamont, OK as part of the Cloudiness Intercomparison Campaign from mid February to late April 2003 [10], [15],and again in Barrow, AK for the Arctic Winter RadiometricExperiment in March and April 2004 [10], [17]. Results fromBarrow, AK in 2002 and 2004 and Lamont, OK in 2003 arepresented here.0196-2892/ 20.00 2005 IEEE
THURAIRAJAH AND SHAW: CLOUD STATISTICS MEASURED WITH THE ICI2001III. DATA ANALYSISA. Water Vapor CorrectionFig. 1. Block diagram of the ICI optics box. A microbolometer-array infraredcamera alternately views blackbody calibration sources and the sky.II. INSTRUMENT DESCRIPTIONThe Infrared Cloud Imager uses an uncooled microbolometerdetector array [18] to measure the downwelling atmospheric radiance. The prototype system has a relatively narrow field ofview, 18 13.5 , which has been adequate for demonstratingthe capability and developing calibration and data analysis techniques. Because the ICI requires no cryogens or other coolingmechanisms for the detector array, it can be deployed in anunattended mode at remote field sites. The present system hastwo blackbodies, whereas previous versions had only one blackbody, which reduces both calibration uncertainty and relianceon stability of the instrument housing temperature [10]. Futureversions of the ICI will have wide angle imaging capability toobtain cloud cover in a significant fraction of the sky dome.The primary ICI system components are the infrared camera,two blackbody calibration sources, a gold-plated beam-steeringmirror, and control electronics (see Fig. 1). The beam-steeringmirror is rotated by a stepper motor to alternately view the blackbodies and the sky, the latter through a sky port that opens whenheavy precipitation is not indicated by a precipitation sensor.The optics are housed in an optics box that sits outside, connected via Arctic cables to a computer inside a nearby buildingor shelter. The system is controlled remotely over a networkconnection.The detector is an uncooled microbometer array containing320 240 pixels, each measuring 50 50 m. One of the largearea blackbodies floats at the temperature of the optics-box inC) and the other is thermo-electricallyterior (typicallyC).controlled over a range of 0 C to 100 C (typicallyPrior to the 2004 deployment in Barrow the ICI contained onlyone blackbody, which was operated at 22.6 C to monitor theICI calibration offset; the gain was measured in the laboratoryand varied less than 5% as long as the enclosure temperature remained in the range of 15 C to 20 C [10]. With the presenttwo-blackbody system, radiometric calibration uncertainty isapproximately 2% [10].The system averages two subsequent images of the blackbodyand the sky (a larger number of images can be averaged to reduce noise). Each sky image is calibrated with a linear calibration equation that uses the measured gain and offset to produceradiometric images with units of band-averaged radiance (wattsper square meter steradian) [10]. Clouds are identified and classified from the radiance value in each pixel.The ICI measures downwelling atmospheric radiance thatincludes emission from clouds (if present) and atmosphericgases. The magnitude of the downwelling radiance increaseswith cloud optical thickness and cloud temperature, but alsowith water vapor content [10]. Other gases that contribute to–m and carbonthe ICI radiance include ozonedioxide (near the longwave edge of the 8–14- m ICI band).We do not compensate for ozone directly because its emissionis not highly variable in space or time and because ozone isprimarily in the stratosphere where it is masked by the lowerclouds. On the other hand, CO emission tends to increase thetemperature dependence of the clear-sky radiance, so we haveincorporated a temperature-dependent water vapor correctionthat includes the small amount of CO emission that is detectedon the longwave edge of the ICI band. This correction and thetemperature independent correction used in Barrow are bothdescribed in the following paragraphs.Within the ICI optical bandwidth, water vapor is the mosthighly variable emission source other than clouds, so we remove the water vapor emission from ICI images to arrive at aresidual radiance that is used to identify clouds. The water vaporradiance was determined with the MODTRAN radiative transfercode [19] as a function of precipitable water vapor (PWV integrated atmospheric column water vapor in cm). Although watervapor is present both above and below clouds, and only thatamount below the cloud would contribute to the ICI signal, simulations and experiments both have confirmed that subtractingthe entire water vapor emission is a self-correcting problem forcloud detection since high clouds are above essentially all of thewater vapor and thick clouds are easy to detect because of theirlarge residual radiance.For Barrow in winter, we found that the ICI data could becorrected with a fairly simple best-fit line relating band-averaged radiance to PWV without a temperature correction (PWVvaried from 0.17 to 1.66 cm in 2002 and from 0.06 to 1.1 cm in2004, but nearly always remained below 0.5 cm, and air temperC to 0 C). This linear relation was deature varied fromrived from radiative transfer modeling using the standard MODTRAN atmospheric models [20] ranging from cold, dry, Arcticwinter to humid tropical conditions to obtain a wide range ofPWV values (Fig. 2).In Oklahoma, however, the atmospheric conditions varied sowidely and the humidity was sufficiently high that we foundit necessary to incorporate a temperature dependence into theCwater vapor correction (the air temperature varied fromC and PWV varied over approximately 0.3–3 cm). Fortothis analysis we used the MODTRAN code to calculate fourdifferent response curves, each of which expresses integratedradiance as a function of PWV for an atmospheric model witha different near-surface air temperature. One of these curveswas calculated for the Tropical atmospheric model, one for the1976 U.S. Standard atmospheric model, one for the MidlatitudeWinter model, and one for the Subarctic Winter atmosphericmodel [20]. Fig. 3 shows the temperature profiles used in thesemodels, each of which has a corresponding set of profiles of
2002IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 9, SEPTEMBER 2005Fig. 2. Band-averaged atmospheric radiance as a function of precipitablewater vapor (centimeters) used for removing water vapor emission from ArcticICI images.Fig. 3. Atmospheric temperature profiles [20] used in the Modtrancalculations for the temperature-dependent water vapor correction. Thesurface air temperatures for these profiles are approximately 15 C, 0 C,15 C, and 27 C.0water vapor, ozone, carbon dioxide, and all other radiatively significant atmospheric species (because of this the correction alsoremoves the CO emission). While the model profiles do notmatch the atmospheric state exactly, they provide comparableresults to MODTRAN calculations using radiosonde profiles oftemperature and water vapor (processed image differences aresmaller than the ICI calibration uncertainty of approximately1–2 W m sr [10]). These curves allow ICI images to beprocessed without relying on infrequent radiosonde profiles.Fig. 4 shows the temperature-dependent water vapor correction curves, each of which was generated by varying thePWV in the corresponding MODTRAN atmospheric model (forwhich the temperature profile is shown in Fig. 3). Each responsecurve in Fig. 4 is characterized according to the near-surface airC,temperature in the corresponding temperature profile (C, andC). The magnitude of the water vapor0 C,correction applied to each ICI image is determined by theFig. 4. Temperature-dependent water vapor correction curves used to processOklahoma ICI data. Each curve is labeled with the near-surface air temperatureof the standard Modtran atmospheric model used to calculate radiance versusPWV (see Fig. 3).value of PWV measured by a nearby microwave radiometer(MWR) and the value of the 2-m air temperature from theSurface Meteorological Observation System (SMOS; data forboth the MWR and SMOS were obtained from the ARM dataarchive, www.arm.gov/data/). So, for example, an ICI imageacquired with a near-surface air temperature of 27 C would becorrected by the amount found from the top curve in Fig. 4 forthe measured value of PWV. A linear interpolation between thetwo nearest response curves is used when the near-surface airtemperature is between two curves.It is important to recognize that the near-surface air temperature is used only to choose which atmospheric model is closestto the actual conditions for each image; it is not used in placeof the cloud temperature or as the temperature of the air between the ICI and the cloud. Full, detailed profiles of air temperature, water vapor, and other gases were used in each MODTRAN atmospheric model [20] to calculate the band-integrateddownwelling radiance as a function of PWV, resulting in thecurves in Fig. 4 (note also that a least squares fit through allthe data points in Fig. 4 reproduces the single curve in Fig. 2).Atmospheric emission also increases with zenith angle, so anangle-dependent water vapor correction will be required in future wide-angle systems. However, for the limited angular fieldof the current system, neglecting this leads to an error of only0.25 W m sr at the image edge.The effectiveness of the water vapor correction is bestillustrated with an example. Fig. 5 shows ICI images fromC and PWVcmOklahoma with air temperaturebefore (top) and after (bottom) application of the temperature-dependent water vapor correction. Both images are colorcoded in units of band-average radiance. In these images thesky is mostly clear on the left-hand side and cloudy on theright-hand side, with brightness temperatures of approximately40 C and 0 C, respectively (Arctic winter clear-sky brightness temperature often is nearC). The clear (dark blue)regions in the water-vapor-corrected image [Fig. 5(b)] haveresidual radiance less than 2 W m sr , which is sufficiently
THURAIRAJAH AND SHAW: CLOUD STATISTICS MEASURED WITH THE ICI2003which comparisons with other sensors were used to refine thethreshold. For example, we compared ICI data from varyingsky conditions with the Actively Remotely Sensed Cloud Locations (ARSCL) value-added product, which is a data productgenerated from a combination of lidar, radar, and radiometerdata at the ARM measurement sites in a manner designed toproduce optimal cloud detection that avoids limitations of eachindividual sensor [21]. Our objective was not to force the ICIto reproduce the ARSCL data exactly, but rather to identifyperiods when the ICI measured either significantly less orsignificantly more clouds. We used data from a variety of ARMsensors for these periods to determine if the ICI was seeingreal clouds not detected by other sensors or if the ICI wasmissing real clouds that were detected by other sensors. In thisway, the best cloud-identification threshold was determined tobe 1.5 W m sr for Barrow data and 2.65 W m sr forOklahoma data.Sometimes with a constant threshold the ICI identifies thickhaze, fog, or aerosols as clouds, and at other times it missesthin cirrus clouds. Therefore, we have begun using an adaptive threshold (explained later) that allows a higher value togreatly reduce incidences of identifying haze as clouds and alower value to increase sensitivity to thin cirrus. In Section IV,we show results obtained with a constant threshold for Barrowdata and an adaptive threshold for the Oklahoma data.IV. ICI CLOUD MEASUREMENTSFig. 5. ICI images (top) before and (bottom) after the temperature-dependentwater vapor correction (note the different scales). The images are fromOklahoma on April 20, 2003, with air temperature 17 C andPWV 1:3 cm. The clear-sky region in the bottom image has residualradiance less than 2 W (m 1 sr), below the default cloud threshold of 2.65W (m 1 sr), indicating successful removal of noncloud atmospheric emission.less than the default Oklahoma cloud threshold of 2.65 (described in Section III-B) to be classified as clear. The residualW msr for the bright cloud on theradiance isright-hand side, 15 W m sr for the small patch at lowerleft, and 4.5 W m sr for the wispy patch between the twobrighter clouds. Therefore, Fig. 5 demonstrates that the watervapor correction is effective at separating cloudy and clearpixels, even in a region between clouds (which may not be asclear as a truly clear sky). Many clear images have residualradiance smaller than 1 W m sr , especially in recent ICIdata with the more stable calibration [10]. A wider variety ofimages and movies from the ICI can be seen elsewhere [10].B. Cloud ThresholdsFollowing water vapor correction, the residual radiance isused to identify and classify clouds. Clouds are identified inpixels where the residual radiance is greater than a thresholdvalue. The first threshold was determined, by considering theradiometric calibration uncertainty, ICI system stability, andwater vapor retrieval uncertainty, to be 1.5 W m sr , afterThe first-order statistical cloud property of interest in radiative transfer and climate models is the percent cloud cover,which is referred to here as cloud amount to distinguish itfrom cloud fraction, which usually is expressed as a functionof altitude. Images from the prototype ICI system are of sufficiently narrow angular coverage that they accurately representthe horizontal cloud cover in that portion of the sky; however,to avoid biasing the cloud amount with cloud sides, futurewide-angle versions of the ICI will be designed with a field), which has been shown byof view near 100 (zenithKassianov et al. [22] to provide optimal correlation betweencloud amount measured from a hemispherical imager and thequantity needed by modelers.Using the data analysis methods described in the previoussection, we computed monthly and weekly cloud amount fordata from the Barrow 2002, Oklahoma 2003, and Barrow 2004deployments. Fig. 6 shows histograms of monthly ICI cloudamount, determined as the fractional number of cloudy images(we recorded one image per 10 min in Barrow 2002 and oneimage per minute in Oklahoma 2003 and Barrow 2004). Inother words, the figure shows the number of pixels that have aresidual radiance greater than 1.5 W m sr for Barrow and2.65 W m sr for Oklahoma, divided by the total number ofpixels in an image, plotted in 10% bins.The monthly statistics for Barrow show that the rate of occurrence for 90% to 100% cloudiness was 39% in March 2002,65% in April 2002, 38% in March 2004, and 42% in the firstnine days of April 2004. In a study of Arctic cloud characteristics during the 1997–1998 SHEBA project Intrieri et al. [23] analyzed cloud zenith-viewing cloud radar and lidar data to determine that the monthly-average cloud occurrence increased from
2004IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 9, SEPTEMBER 2005Fig. 6. Monthly cloud cover statistics derived from ICI data for (top) Barrow,AK 2002, (middle) Lamont, OK 2003, and (bottom) Barrow, AK 2004.80% in March to almost 90% in April. The results from ICIshow a similar trend in increasing cloudiness, but with lowervalues that may be a result of the different locations, the ICIbeing deployed in Barrow (coastal) and the SHEBA measurements being taken further north on the Arctic Ocean pack ice.Although the ICI data in March 2002 are limited to only abouthalf a month because of the unavailability of MWR water vapordata, we have a continuous record from March 2004 with oneimage each minute. The bimodal nature of cloudiness at Barrow(clear or overcast) is typical of what has been observed by Intrieri et al. [23] and other previous investigators, and suggeststhat perhaps wide-angle imaging may not be so critically neededin the Arctic.The overall monthly statistics from Oklahoma depict a similar trend toward mostly clear or mostly cloudy skies, with approximately 40% cloudiness in both March and April 2003. Themonthly frequency of sky cover derived from the broadbandshortwave radiometers used at the SGP ARM site show that overa period of six years from 1996 to 2001, the monthly cloudinessvaried from 25% to 45% for both March and April and the trendof seeing mostly clear and/or cloudy skies is similar to what isseen by the ICI [24].Beyond just measuring cloudiness, one of the questionswe set out to explore with the ICI measurements is whetherthere are significant differences in cloud amount measured byzenith-viewing and imaging sensors. Statistics determined fromzenith sensors are assumed equal to spatial statistics throughTaylor’s hypothesis, which states that the covariance in time isrelated to the covariance in space by the speed of the mean flowwhen the turbulence is small compared to the mean flow [8];however, Sun and Thorne [9] indicated that this assumption mayFig. 7. Scatter plots of single-pixel and full-image ICI cloud statistics from(top) Alaska 2004 and (bottom) Oklahoma 2003. The single pixel predictsslightly less cloudiness than the full image for the Oklahoma data, but thiseffect is nearly absent in the Alaska data.break down with dissipating or developing clouds, multilayeredclouds, or when the wind speed and direction are changing.Although the ICI determines spatial statistics directly from themeasured radiance without invoking Taylor’s hypothesis, it isdifficult to say much about the difference between temporaland spatial statistics by comparing data from different sensorsbecause of the differing instrument sensitivities, thresholds,and retrieval techniques. Therefore, to address this questionwithout relying on a comparison of data from different sensors,we compared statistics calculated from a single ICI pixel tostatistics calculated from full ICI images. Cloud statistics werecalculated from the single-pixel radiance time series as thepercentage of samples for which that pixel’s residual radianceexceeded the cloud threshold. The same analysis was appliedalso to each full image, but the average cloud amount (CA) inpercent was determined for images asCAcloudy pixels per imagetotal of pixels per image(1)Fig. 7 shows scatter plots of the single-pixel and full-imageICI cloud amount data from Barrow 2004 and Oklahoma 2003.These plots show that single-pixel measurements slightly underestimate cloudiness relative to the full ICI images in Oklahoma. However, in the Barrow 2004 data the single-pixel andfull-image cloud amounts are very nearly equal except on justa few days. The Barrow 2004 experiment was characterized by
THURAIRAJAH AND SHAW: CLOUD STATISTICS MEASURED WITH THE ICIFig. 8. Scatter plots of MPL and ICI daily cloud amount (top) with a constantICI threshold and (bottom) with an adaptive ICI threshold to increase cirrussensitivity and decrease haze sensitivity. The symbols indicate days for whichthe adaptive threshold algorithm identified significant haze ( ), cirrus ( ), orneither ( ). 2particularly persistent periods of clear or overcast skies (multiple hours to days); however, the clouds were more variableon the few days when the cloud amount was noticeably lessfor the single pixel than for the full images. Several of thesedays included highly variable cirrus, and the others had brokenlow clouds with intermittent clear periods. The Alaska 2004data (top) have correlation coefficient of 0.999 (between the ICIfull-image and single-pixel cloud amounts) and rms differenceof 1.56, while the Oklahoma 2003 data have correlation coefficient of 0.988 and rms difference of 7.69.In accordance with the expectation that Taylor’s hypothesis applies best to clouds that are not rapidly changing, thesingle-pixel and full-image data agree most closely duringthe dominant periods of steady clear or overcast conditions inthe Alaska 2004 experiment, but disagree more often (and bya larger amount) for the Oklahoma 2003 experiment, whichexperienced dramatically higher cloud variability. Given themodest field of view of the prototype ICI system, this analysis does not exactly represent the difference between zenithand full-sky imager data, but does suggest that there may bemeasurable differences in cloud cover statistics derived fromzenith and imaging sensors with variable clouds. Establishingthis more fully will require analysis of a longer dataset.The need for the adaptive cloud-identification thresholdmentioned earlier is illustrated in Fig. 8. The top scatter plot2005shows MicroPulse Lidar (MPL) and ICI daily cloud amountfrom March and April 2003 in Oklahoma, with a constant2.65 W m sr ICI threshold. Although there is reasonableagreement between the two sensors (correlation 0.752, rms24.5), sometimes the ICI constant threshold detects haze ascloud and fails to detect thin cirrus clouds. For the data shownin Fig. 8 we relied primarily on cloud lidar data to trigger theadaptive threshold because of the lidar’s high sensitivity toboth cirrus and haze, although another useful cloud indicator isthe variance of a broadband solar irradiance time series. Thecurrent adaptive threshold algorithm checks for the presenceof cirrus and low-level haze in an MPL time series during avariable-length running time window (typically several hoursto a day). If cirrus is present and low-level haze is not, the ICIthreshold is reduced to a lower value, found to be optimal at1.5 W m sr . Low-level haze is detected in the MPL databy examining the strength of the lidar backscatter in the bottom500 m. If the logarithm of this low-level MPL backscatterpower exceeds 4.5, the algorithm increases the ICI thresholdto at least 3.5 W m sr , with higher thresholds for muchstronger low-level lidar signals (not exceeding 6 W m sr .This algorithm will miss thin cirrus during periods of thicklow-level haze, but will not miss low-level clouds because evenwith a higher threshold such clouds are nearly impossible tomiss because of their large residual radiance.The bottom scatter plot in Fig. 8 shows the same data asthe top plot, but with the adaptive ICI threshold. The symbols indicate days on which the constant threshold missed cirrusclouds that were seen consistently in MPL data, while thesymbols indicate days on which the constant threshold classified low-level haze as clouds (haze is indicated by high relative humidity and near-surface MPL backscatter). The adaptivethreshold results in correlation 0.937 and rms 12.45 (although these adaptive threshold results are a result of postexperiment data processing, the procedure can be automated as longas the ancillary data streams are available). Research is continuing into improved adaptive algorithms, including using aerosolsize distributions and scattering coefficients to help identify periods of low-level haze.We also investigated the day-night consistency of ICI cloudstatistics (Fig. 9). Since the ICI measures thermal emission,its cloud sensitivity and retrieval accuracy are not expectedto vary significantly over a diurnal cycle. A key tradeoff inground-based measurement of cloud statistics has been thatzenith-viewing active sensors (e.g., radars and lidars) providevertically resolved measurements with high consistency overa diurnal cycle, but do not provide information about cloudsaway from the zenith when not scanning. Conversely, imagingsensors provide horizontal spatial information, but have difficulty doing so consistently during both day and night. Someimaging sensors (e.g., TSI) provide daytime data only, whileothers (e.g., WSI) measure daytime and nighttime data withdifferent spatial resolution, bandwidth, and retrieval technique(red-blue ratios in day and star maps at night).In Fig. 9 we show scatter plots of cloud data from the ARSCLcloud product [21] and ICI for daytime (top) and nighttime(bottom). The lack of a significant day-night difference provides evidence of the ICI’s consistency over a diurnal cycle (theARSCL active sensors have essentially constant daytime andnighttime sensitivity). The daytime scatter plot has a correlation
2006IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 9, SEPTEMBER 2005
THURAIRAJAH AND SHAW: CLOUD STATISTICS MEASURED WITH THE ICIREFERENCES[1] V. Ramanathan, B. Subasilar, G. J. Zhang, W.
2000 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 9, SEPTEMBER 2005 Cloud Statistics Measured With the Infrared Cloud Imager (ICI) Brentha Thurairajah, Member, IEEE, and Joseph A. Shaw, Senior Member, IEEE Abstract—The Infrared Cloud Imager (ICI) is a ground-base
IEEE 3 Park Avenue New York, NY 10016-5997 USA 28 December 2012 IEEE Power and Energy Society IEEE Std 81 -2012 (Revision of IEEE Std 81-1983) Authorized licensed use limited to: Australian National University. Downloaded on July 27,2018 at 14:57:43 UTC from IEEE Xplore. Restrictions apply.File Size: 2MBPage Count: 86Explore furtherIEEE 81-2012 - IEEE Guide for Measuring Earth Resistivity .standards.ieee.org81-2012 - IEEE Guide for Measuring Earth Resistivity .ieeexplore.ieee.orgAn Overview Of The IEEE Standard 81 Fall-Of-Potential .www.agiusa.com(PDF) IEEE Std 80-2000 IEEE Guide for Safety in AC .www.academia.eduTesting and Evaluation of Grounding . - IEEE Web Hostingwww.ewh.ieee.orgRecommended to you b
Signal Processing, IEEE Transactions on IEEE Trans. Signal Process. IEEE Trans. Acoust., Speech, Signal Process.*(1975-1990) IEEE Trans. Audio Electroacoust.* (until 1974) Smart Grid, IEEE Transactions on IEEE Trans. Smart Grid Software Engineering, IEEE Transactions on IEEE Trans. Softw. Eng.
898 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 43, NO. 4, APRIL 2005 Vertex Component Analysis: A Fast Algorithm to Unmix Hyperspectral Data José M. P. Nascimento, Student Member, IEEE, and José M. Bioucas Dias, Member, IEEE Abs
IEEE TRANSACTIONS ON IMAGE PROCESSING, TO APPEAR 1 Quality-Aware Images Zhou Wang, Member, IEEE, Guixing Wu, Student Member, IEEE, Hamid R. Sheikh, Member, IEEE, Eero P. Simoncelli, Senior Member, IEEE, En-Hui Yang, Senior Member, IEEE, and Alan C. Bovik, Fellow, IEEE Abstract— We propose the concept of quality-aware image, in which certain extracted features of the original (high-
IEEE Robotics and Automation Society IEEE Signal Processing Society IEEE Society on Social Implications of Technology IEEE Solid-State Circuits Society IEEE Systems, Man, and Cybernetics Society . IEEE Communications Standards Magazine IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology IEEE Transactions on Emerging .
Standards IEEE 802.1D-2004 for Spanning Tree Protocol IEEE 802.1p for Class of Service IEEE 802.1Q for VLAN Tagging IEEE 802.1s for Multiple Spanning Tree Protocol IEEE 802.1w for Rapid Spanning Tree Protocol IEEE 802.1X for authentication IEEE 802.3 for 10BaseT IEEE 802.3ab for 1000BaseT(X) IEEE 802.3ad for Port Trunk with LACP IEEE 802.3u for .
IEEE 1547-2003 IEEE P1032 IEEE 1378-1997 Controls IEEE 2030-2011 IEEE 1676-2010 IEEE C37.1 Communications IEC 61850-6 IEC TR 61850-90-1 & IEEE 1815.1-2015 IEC TR 61850-90-2 Cyber & Physical Security IEEE 1686-2013 IEEE 1402-2000
If you are starting on Advanced Level, we advise that you enrol onto the EIMF 2 Day Bookkeeping course, as there is an assumption at this level that you have a good knowledge of Double Entry Bookkeeping. Syllabus You will learn complex accounting techniques including maintaining cost accounting records and the preparation of reports and returns .
