Cloud Coverage Based on All-Sky Imaging and Its Impact on Surface Solar Irradiance

G. Pfister National Institute of Water and Atmospheric Research, Lauder, Central Otago, New Zealand

Search for other papers by G. Pfister in
Current site
Google Scholar
PubMed
Close
,
R. L. McKenzie National Institute of Water and Atmospheric Research, Lauder, Central Otago, New Zealand

Search for other papers by R. L. McKenzie in
Current site
Google Scholar
PubMed
Close
,
J. B. Liley National Institute of Water and Atmospheric Research, Lauder, Central Otago, New Zealand

Search for other papers by J. B. Liley in
Current site
Google Scholar
PubMed
Close
,
A. Thomas National Institute of Water and Atmospheric Research, Lauder, Central Otago, New Zealand

Search for other papers by A. Thomas in
Current site
Google Scholar
PubMed
Close
,
B. W. Forgan Bureau of Meteorology, Melbourne, Victoria, Australia

Search for other papers by B. W. Forgan in
Current site
Google Scholar
PubMed
Close
, and
C. N. Long Pacific Northwest National Laboratory, Richland, Washington

Search for other papers by C. N. Long in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

In Lauder, Central Otago, New Zealand, two all-sky imaging systems have been in operation for more than 1 yr, measuring the total, opaque, and thin cloud fraction, as well as indicating whether the sun is obscured by clouds. The data provide a basis for investigating the impact of clouds on the surface radiation field. The all-sky cloud parameters were combined with measurements of global, direct, and diffuse surface solar irradiance over the spectral interval from 0.3 to 3 μm. Here, the results of ongoing analysis of this dataset are described. As a reference for the magnitude of the cloud influence, clear-sky irradiance values are estimated as a simple function of solar zenith angle and the earth–sun distance. The function is derived from a least squares fit to measurements taken when available cloud images show clear-sky situations. Averaged over a longer time period, such as 1 month, cloud fraction and surface irradiance are clearly negatively correlated. Monthly means in the ratio of the measured surface irradiance to the clear-sky value had a correlation coefficient of about −0.9 with means of cloud fraction for the months from July 2000 to June 2001. In the present work reductions in the surface irradiance and situations in which clouds cause radiation values to exceed the expected clear-sky amount are analyzed. Over 1 yr of observations, 1-min-averaged radiation measurements exceeding the expected clear-sky value by more than 10% were observed with a frequency of 5%. In contrast, a reduction of more than 10% below estimated clear-sky values occurred in 66% of the cases, while clear-sky irradiances (measured irradiance within ±10% of estimated clear-sky value) were observed 29% of the time. Low cloud fractions frequently lead to moderate enhancement, because the sun is often unobscured and the clouds are brighter than the sky that they hide. As cloud fraction increases the sun is likely to be obscured, causing irradiance values to fall well below clear-sky values. However, in the case of unobscured sun, there is a tendency for strongest enhancements when cloud fractions are highest. Enhancements, especially at high solar zenith angle, are also often observed in association with thin clouds.

Current affiliation: National Center for Atmospheric Research, Boulder, Colorado

Corresponding author address: G. Pfister, Atmospheric Chemistry Division, National Center for Atmospheric Research, P. O. Box 3000, Boulder, CO 80307-3000. pfister@ucar.edu

Abstract

In Lauder, Central Otago, New Zealand, two all-sky imaging systems have been in operation for more than 1 yr, measuring the total, opaque, and thin cloud fraction, as well as indicating whether the sun is obscured by clouds. The data provide a basis for investigating the impact of clouds on the surface radiation field. The all-sky cloud parameters were combined with measurements of global, direct, and diffuse surface solar irradiance over the spectral interval from 0.3 to 3 μm. Here, the results of ongoing analysis of this dataset are described. As a reference for the magnitude of the cloud influence, clear-sky irradiance values are estimated as a simple function of solar zenith angle and the earth–sun distance. The function is derived from a least squares fit to measurements taken when available cloud images show clear-sky situations. Averaged over a longer time period, such as 1 month, cloud fraction and surface irradiance are clearly negatively correlated. Monthly means in the ratio of the measured surface irradiance to the clear-sky value had a correlation coefficient of about −0.9 with means of cloud fraction for the months from July 2000 to June 2001. In the present work reductions in the surface irradiance and situations in which clouds cause radiation values to exceed the expected clear-sky amount are analyzed. Over 1 yr of observations, 1-min-averaged radiation measurements exceeding the expected clear-sky value by more than 10% were observed with a frequency of 5%. In contrast, a reduction of more than 10% below estimated clear-sky values occurred in 66% of the cases, while clear-sky irradiances (measured irradiance within ±10% of estimated clear-sky value) were observed 29% of the time. Low cloud fractions frequently lead to moderate enhancement, because the sun is often unobscured and the clouds are brighter than the sky that they hide. As cloud fraction increases the sun is likely to be obscured, causing irradiance values to fall well below clear-sky values. However, in the case of unobscured sun, there is a tendency for strongest enhancements when cloud fractions are highest. Enhancements, especially at high solar zenith angle, are also often observed in association with thin clouds.

Current affiliation: National Center for Atmospheric Research, Boulder, Colorado

Corresponding author address: G. Pfister, Atmospheric Chemistry Division, National Center for Atmospheric Research, P. O. Box 3000, Boulder, CO 80307-3000. pfister@ucar.edu

Introduction

Clouds are still one of the less understood components in the earth–atmosphere system, and the measurement of their radiative properties, from ground as well as from satellites, proves to be a challenge. Understanding the effects of clouds on the shortwave irradiance is of importance for radiative energy budget studies because clouds play a critical role in regulating the amount of solar radiation penetrating the earth's atmosphere. While long-term temporal or spatial averages define climate, cloud modulation of irradiance at shorter timescales affects available solar energy for plants and human use, and the risk of damage by solar radiation.

Clouds of the same type attenuate radiation by different amounts because of variations in their macrophysical characteristics, such as cloud coverage and geometry, and their microphysical characteristics, like optical thickness, liquid water content, or particle size distribution. These dependencies make it difficult to develop a quantitative relationship between cloud properties and the actual radiation field. Large spatial and temporal variability in cloudiness further complicates an understanding of their radiative effects.

Historically, ground-based cloud measurements have relied on human observations. These provide information about different cloud types when visible, but they are expensive and somewhat subjective. In a previous study at Lauder, Central Otago, New Zealand, a series of noontime all-sky photographs were used to assess the accuracy of satellite estimates of cloud cover and to identify the effects of clouds on radiation (McKenzie et al. 1998), but the study found a need for higher time resolution, necessitating an automated system. In the present study we make use of a hemispherical sky imaging technique (Long and DeLuisi 1998), which provides information about the fractional cloud coverage with a high temporal resolution, although it does not provide information about the cloud type, except for separating between thick (opaque) and thin clouds. We use over 1 yr of cloud-fraction measurements collected in Lauder in combination with measurements of the global, direct, and diffuse solar irradiance to investigate the influence of the cloud amount on the surface radiation field. The approach we use to quantify the cloud impact is based on the ratio of measured surface irradiance to expected clear-sky values. A nonlinear least squares fit was applied to measured clear-sky data for estimating radiation levels under cloudless conditions as a function of the solar zenith angle.

Clouds generally reduce the incoming solar radiation at the surface, but in some cases they may also cause radiation levels exceeding the expected clear-sky value. Herein, we refer to increased radiation levels as “enhancements” or “enhanced” radiation. Indications of cloud enhancement of diffuse irradiance in clear-sun conditions (conditions when the sun disk is not obscured by clouds) are well known (e.g., Robinson 1966). They have attracted particular interest in the ultraviolet (UV), where high intensity even for a short time can constitute a hazard. Nack and Green (1974) identified significant enhancement of global irradiances, and, more recently, Mims and Frederick (1994) described, for UV-B solar irradiance measured with high temporal resolution, increases of up to nearly 30% due to scattering from the side of cumulus clouds near the position of the sun. In a 1996 study in Lauder, Bodeker and McKenzie (1996) observed a significant enhancement in the measured surface radiation due to clouds. Estupinan et al. (1996) found that under special circumstances broken cumulus-type clouds might produce localized increases in UV-B irradiance of up to 27% over timescales of less than 1 h.

Segal and Davis (1992) presented evidence that surface measurements can be persistently affected by orographically related deep convection. Enhancements are more than canceled by reductions in irradiance when the sun is obscured, so that averages over time or area are almost always less under cloudy skies.

The magnitude of enhanced radiation has been also investigated by radiative transfer simulations. For example, Segal and Davis (1992) used a Monte Carlo radiative transfer model to simulate various deep convective cloud cases. Except for one low solar elevation case (solar zenith angle of 85°) where they calculated the surface irradiance to be enhanced by as much as 100%, their cases showed an enhancement on the order of about 20%–30%, which is in agreement with observations.

Under certain cloud conditions, regions that experience little or no attenuation in the direct downwelling solar radiation component might exhibit enhanced surface irradiance values over short time intervals (typically in the order of minutes). From a climatological point of view, or when large regions are considered, these enhanced radiation values are of importance only for the extent to which they offset reductions due to cloud. They have greater significance when the short time exposure to radiation is of interest, such as in biology (e.g., plant photosynthesis, harmful effects of UV-B) or economics (e.g., solar power, solar heating, light). Nonlinearities between the amount of radiation and its effects on organisms and materials, including threshold effects, can mean that an increase in the irradiance even over a short time period may be important.

Automated detection of cloud cover is a relatively new field, without extensive literature on instruments, algorithms, and their performance. In section 2 we describe the all-sky cameras used in this study, and the processing of measurements into cloud data. In section 3 we compare results from more than 1 yr of observations with the two systems to quantify the reproducibility of the technique. Section 4 describes the radiation data used in the analysis of section 5, which compares cloud and radiation data by statistical analysis. Section 6 uses the data to characterize the frequency of situations where clouds either cause enhancements or reductions in the surface radiation, and the net effect on surface irradiance. Section 7 summarizes these results and comments on future prospects.

Cloud information from all-sky imaging

At Lauder, New Zealand (45.038°S, 169.684°E; altitude, 370 m), two automated all-sky imaging systems provide information about the cloud coverage and distinguish when the sun is or is not obscured by clouds. The first instrument, hereinafter called Allsky1, was developed at the National Institute of Water and Atmospheric Research (NIWA) in Lauder and was installed in April 1998. The second all-sky imager (hereinafter Allsky2), a total sky imager (TSI) 880 built by Yankee Environmental Systems, Inc., began operating in September 1999. The two cameras cover about the same hemispherical field of view (FOV), but were placed about 300 m from each other to allow the possibility of using the images for stereographic analysis of cloud height.

The cameras are programmed to acquire one image per minute when the solar elevation is greater than about 3°. Each image is approximately 40 kB in size, and about 1 month of data can be stored on a CD-ROM. A few months after the installation of Allsky2, observations with Allsky1 were downgraded to backup mode and the sampling interval was increased to 10 min.

The basic design for all-sky imaging systems includes a camera and a wide-angle lens or hemispheric mirror. Both systems use a small weather-resistant digital camera looking down at a convex hemispheric mirror. To prevent flare in adjacent pixels, both systems use a shadow band to block the sun's reflection. Analysis of the digital image is based on the red-to-blue ratio (the ratio of red to blue light in a pixel) in each image pixel (nominal pixel resolution ∼0.32° and ∼0.28° per pixel for Allsky1 and Allsky2, respectively). White clouds have a much higher red-to-blue ratio than a clear blue sky, though this distinction can be confused by a pallid sky in hazy conditions, especially near the sun or the horizon, and by dark clouds illuminated by skylight.

When a cloud image is processed, a red-to-blue threshold, which is empirically derived, is used to distinguish between clear and cloudy pixels. This limit is not the same across the entire image but is determined as a function of the relative position of the image pixel and the sun. This is necessary because of the angular dependence of the scattering processes, and is also due to the effects of the increased pathlength with increasing solar zenith angle (Long and DeLuisi 1998; Long et al. 2001).

The estimated fractional cloud cover for an all-sky imager is calculated as the number of cloudy pixels divided by the total number of pixels within the FOV of the imager (FOV = 160° for both imagers) and is given in steps of 1% cloud fraction. Parameters derived from the all-sky cameras are a total, opaque (thick), and thin cloud fraction. Separation between thick and thin clouds is based on a second empirically derived threshold in the red-to-blue ratio, which varies with distance of the pixel from the sun and the horizon. The total cloud fraction is the sum of opaque and thin fractions. A further parameter that can be estimated from the all-sky cameras is the “sunshine parameter,” which is an indicator of whether the sun is obscured by clouds. The image-processing algorithm for Allsky1 makes use of the brightness of the flat, white area surrounding the mirror to calculate this parameter. The algorithm for Allsky2 looks along the sun-blocking strip on the mirror for an increase in brightness. If the brightness increases above an empirically derived threshold, then the sun disk is deemed unobscured by clouds, otherwise it is assumed to be obscured.

Figure 1 shows, as an example, cloud images (Figs. 1a and 1c), as well as the corresponding analysis images (Figs. 1b and 1d), taken with Allsky1 and Allsky2, respectively. The cloud images are digital pictures reflecting the actual cloud situation (e.g., Fig. 1 shows a situation of broken cloud coverage), while the analysis images are a graphical depiction of the results of the sky cover retrieval processing. The overall circular area seen in the cloud-processing images defines the area that is processed to estimate the cloud fraction. It is limited to the region within 80° of the zenith for both imagers. Cloud-free areas appear in dark gray, while the shadow band and camera are masked as black and, thus, ignored in the processing algorithm. White and light gray areas represent opaque and thin clouds, respectively.

The cross in the Allsky1 analysis image (Fig. 1b) and the dot in the Allsky2 image (Fig. 1d) denote the sun's position. The circle around the sun's location (hereinafter referred to as “sun circle”), as well as the marked area around the solar azimuth shown in the Allsky1 image in Fig. 1b, depict areas that are processed separately from the rest of the image. These regions are problematic for the cloud detection algorithm due to possible overloading effects in the charged couple device (CCD) camera.

For the circle region around the zenith (shown as a circle in the middle of the picture in Fig. 1b), a zenith circle cloud fraction is calculated for Allsky1. It is a helpful parameter for comparison with cloud fractions from, for example, lidar measurements. Cloud fractions for sun circle, zenith circle, and horizontal area are currently only available for Allsky1 and are not yet included in the automatic processing software for Allsky2.

Pixels in the sun circle, an approximate 20° FOV centered on the sun position, often appear to be cloudy even when they are not. This can occur because of forward scattering from thin cirrus clouds or boundary layer haze, which is subvisible elsewhere in the image, or because of CCD element energy overloading and bleeding. Because of these problems in the sun circle region, we apply a quality control procedure that uses the mean and standard deviation of the cloud fraction over an 11-min period centered on the image of interest. This quality control is applied to the 1-min all-sky images only. If both the whole-sky cloud fraction and its variance over the sun circle for an 11-min period are small, then it is very likely that the sun circle was free of clouds. Conversely, a large variability in the sun circle cloud fraction and significant cloudiness in the rest of the sky over the 11-min period suggests that at least some of the sun circle actually had some clouds in it at the time of interest. A quality flag is included in the data stream to provide the data user with information about this quality check. A test of the quality flag is the corresponding sunshine parameter. For those cases where the quality flag indicates that the sun circle area has no “real” clouds in it, Allsky1 identifies unobscured sun in 89%. The percentage is even higher, 98%, when comparing it with the Allsky2 sunshine parameter.

The cloud retrieval algorithm for Allsky1 estimates also an “edge-to-area ratio” as the number of pixels on cloudy/clear-sky boundaries divided by the number of pixels within clouds. This parameter is an indication of “average” cloud size and brokenness of cloud coverage. A high edge-to-area ratio results from broken clouds of a small diameter, while small edge-to-area ratios refer to extended clouds. Clear and completely covered skies have an edge-to-area ratio of zero.

Allsky1 and Allsky2 are both based on the same measurement principle but show differences with regard to their technical configuration. The main distinctions are different cameras and filter systems in Allsky1 and Allsky2. Allsky1 is an inexpensive “off the shelf” Web camera called “Quickcam.” Additional optics include a daylight filter to correct the color balance of the camera, which is set for indoor use, and a neutral density filter to reduce the light level and prevent saturation. Allsky2 has a camera module requiring only the neutral density filter.

The Allsky2 camera is supported by a “hockey stick” support (a pipe bent in an inverted L shape, firmly attached at the base with the bent horizontal section supporting the camera), the effect of which is masked in processing. The Allsky1 camera is supported by a tripod, which might cause uncertainties in cloud fraction of about 1% (because the legs of the tripod block a small part of the mirror). An additional disadvantage of this construction is that the tripod can cast a shadow on the flat white area around the mirror, which can interfere with calculations of the sunshine parameter. The domes of the two systems are similar, and in both cases the dome and shadow band rotate during the day to prevent direct sun strike to the camera.

With their different technical specifications, Allsky1 and Allsky2 also use slightly different cloud retrieval algorithms. These can lead to discrepancies, especially under complex cloud fields, as can the 300-m separation of the instruments, and some inconsistency in temporal assignment of the data by the acquisition system. In the following section we compare results of Allsky1 and Allsky2 and discuss the agreement in cloud fraction and the sunshine parameter.

Comparison of Allsky1 and Allsky2 cloud fractions

An intercomparison of the two all-sky systems was performed for all times for which data for both cameras are available (∼67 000 cloud images over the period from September 1999 to June 2001). In about 80% of the images the absolute difference in total and opaque cloud fractions is less than 10%. The percentage is about the same for the thin cloud fractions, but because the fraction of thin clouds in the images is, in general, rather small, it implies that the uncertainty in the thin cloud fraction is larger in comparison with opaque and total cloud fraction.

The mean value for the Allsky1 total cloud fraction over the whole dataset is calculated as 65.4%, with a standard deviation of 40.5%. Results are very similar for Allsky2, with a mean value of 65.9% and a standard deviation of 41%. Slightly larger differences are observed for the opaque cloud fraction with a mean value and standard deviation of 60.9% ± 41.8% and 57.9% ± 42.4% for Allsky1 and Allsky2, respectively. Thin cloud fractions are typically on the order of less than about 20% in both cameras. For Allsky1 a mean thin cloud fraction of 4.5% and a standard deviation of 7.5% are calculated; a mean of 8% and a standard deviation of 13.5% are derived for Allsky2. Mean values for the absolute difference between Allsky1 and Allsky2 total, opaque, and thin cloud fractions are 6.2%, 6.7%, and 6.8%, respectively. The standard deviation in all three cases is in the range of 12%. An absolute difference of less than or equal to 10% in cloud fraction is observed for 83% of the images in case of total cloud fraction, 80% in case of opaque cloud fraction, and 79% in case of thin cloud fraction.

The frequency distribution for the difference in total, opaque, and thin cloud fractions between the two imagers is shown in Fig. 2. From all images that were compared, 31% yield a difference in the thin cloud fraction between Allsky1 and Allsky2 of zero, 51% of the Allsky1 images detect lower thin cloud fractions when compared with Allsky2, and 18% of the images yield a higher thin cloud fraction for Allsky1. The opaque cloud coverage shows the opposite behavior, that is, more cases with Allsky1 cloud fractions greater than those of Allsky2 and fewer cases where Allsky1 shows lower cloud fractions. Of the images, 32% have identical opaque cloud fractions, 45% of the Allsky1 cloud fractions are greater than Allsky2, and 23% of the Allsky2 cloud fractions are greater than Allsky1 cloud fractions. The result for the total cloud fraction is similar to that for the thin cloud fraction but the bias is less pronounced: 38% of the images indicate zero difference, 24% of the images indicate that Allsky1 has higher cloud fractions, and 38% of the images indicate that Allsky2 data have larger cloud fractions.

On average, Allsky1 and Allsky2 differ by −3.5% in thin cloud fractions, by +3% in opaque, and by −0.5% in total cloud fractions (for which the differences in thin and opaque fractions partially cancel). The median of the difference reflects the findings mentioned above. It is zero for total and opaque cloud fractions but −1% for the thin cloud fraction, showing that in most cases Allsky2 detected slightly higher thin cloud fractions than Allsky1.

The correlation coefficient indicates a good agreement in total and opaque cloud fractions and larger deviations in the thin cloud fraction. The correlation coefficients between Allsky1 and Allsky2 for both the total and opaque cloud fractions are about 0.95, while the correlation is clearly worse for the thin cloud fraction, at only 0.43.

The sunshine parameter for both systems agrees in 89% of the cases that were investigated. The technique to determine the sunshine parameter is the same in both cameras, that is, by setting an empirically derived threshold for the brightness, although the measurement of the brightness and, consequently, the threshold differ for the two all-sky cameras. A mismatch in the sunshine parameter will predominantly occur in situations that cannot be unambiguously classified as either obscured or unobscured, for example, situations when the sun disk is partly covered by clouds or when the sun disk is covered by optically very thin clouds.

The intercomparison for the two cameras in opaque and thin cloud fractions is slightly improved if the analysis is restricted to solar zenith angles less than 70°. Mean values of the difference in thin and thick cloud fractions are highest at large solar zenith angle; the strongest effect is seen in the thin cloud fraction. For solar zenith angles below 70° the agreement shows no dependence on solar zenith angle. This is very likely due to the empirical threshold in the red-to-blue ratio that is used to separate opaque and thin clouds in the cloud retrieval algorithm. At large solar zenith angles clear sky is more white because of the larger pathlength, which complicates the separation of cloudy from clear pixels.

Figure 3 indicates that the differences between the two systems strongly depend on whether the sun is obscured by clouds, giving better agreement in cloud fraction for sun-obscured situations. For cases in which both imagers specify obscured sun (33 896 images), about 95% agree in total cloud fraction within 10% (Fig. 3b), while the percentage is only about 65% for images where the sun disk is not obscured (25 907 images, Fig. 3a). Based on these findings, cloud fractions from the all-sky imaging systems are, in general, consistent to about 10%. We, therefore, use this as an approximate bound on the inherent errors in the cloud coverage data in the analysis that follows.

Radiation measurements

Solar irradiance measurements at Lauder are part of the international Baseline Surface Radiation Network (BSRN). The goal of BSRN is to provide long-term, frequently sampled, state-of-the-art measurements of surface radiation fluxes by adhering to the highest achievable standards of measurement procedures, calibration, and accuracy.

As per BSRN specifications, the irradiance statistics collected for each minute are the mean, standard deviation, and maximum and minimum irradiance, based on 1-Hz sampling. Each 1-Hz sample is an integration measurement of at least one power cycle (50 Hz, or 20 ms). Together with the e-folding response time of the radiometers of about 5 s, it implies that high-frequency changes in the true irradiance with equivalent periods of 10 s or less are recorded with significantly higher uncertainty than averages over periods of 1 min or more. This is particularly relevant when considering situations in which the surface irradiance is enhanced, often to a peak of short duration.

From the complete set of BSRN data we used the global, direct, and diffuse irradiance measurements in our analysis. Solar irradiance is defined as the integrated radiative energy transported from all directions across a horizontal surface. Unless otherwise specified, horizontal is defined as a plane perpendicular to the direction to the zenith. Direct irradiance refers to that part of solar radiation that reaches the earth's surface as a collimated beam after selective attenuation in the atmosphere, while diffuse radiation includes downward-scattered and reflected radiation, excluding the circumsolar radiation. Global irradiance is defined as the sum of diffuse and direct irradiance. Within the BSRN, the uncertainty for the 1-min average values of the global and diffuse irradiance is about 4% or 15 W m−2, and for the direct component it is 3% or 10 W m−2. Further information about the BSRN equipment, calibration techniques, or the data processing is given in Ohmura et al. (1998) and the references therein.

Radiation measurements taken within the Lauder BSRN (beginning in August 1999) were combined with the Allsky1 and Allsky2 cloud datasets. The combined BSRN and all-sky datasets form the basis for the following investigation studying the relationship between cloud fraction and solar surface irradiance. The distance between the Allsky2 sky-imager systems and the radiation sensors that are located on the roof of one of the laboratory buildings in Lauder is approximately 5 m.

Analysis of radiation and cloud data

One method to quantify the cloud impact is to use estimates for the clear-sky surface irradiance as a reference and look at the ratio of measured to estimated clear-sky values. Estimates of clear-sky values can be derived from radiative transfer simulations or from a parameterization of clear-sky irradiance. We employ the latter technique—using the solar zenith angle as the primary factor to determine the downwelling solar irradiance under clear-sky conditions. As was shown, for example, by Cess et al. (1995), an empirical fit applied to measurements and taken under cloudless conditions can be used to calculate expected clear-sky values as function of the solar zenith angle.

According to Long and Ackerman (2000) a simple equation of the form Iclear = a cos(SZA)b serves to estimate the clear-sky irradiance Iclear, with SZA as the solar zenith angle, and a and b calculated from a least squares fit. The annual variation in the sun–earth distance is taken into account by accordingly adjusting the coefficient a. Clear-sky measurements were selected from the dataset by searching for times at which cloud fractions derived from both imagers are 0% (1580 data points). These clear-sky data were used to calculate the coefficients a and b (a = 1125.4 and b = 1.21). For clear-sky estimates, solar zenith angles are restricted to 75° or less because the quality of the fit diminishes if low elevation sun angles are included.

Because the site in Lauder is characterized by low aerosol loading and, in general, only small changes in the surface albedo throughout the year, this simple fit seems reasonable. This is confirmed by differences between observed and estimated clear-sky values of less than ±10%. For 89.8% of the clear-sky cases, the measured and estimated values are within ±5%, 99.7% agree within ±8%, and 100% within ±10%. Differences are partly due to uncertainties in the empirical fit and partly due to measurement errors but also have to be expected because a single fit cannot account for the temporal variability in other factors that influence the transfer of solar radiation through the atmosphere (day-to-day variations in the atmospheric water vapor content, the seasonal variation in surface albedo with changes in the vegetation, etc.). Lauder is generally snow free, and the few days with snow coverage were taken out of the dataset.

The Long and Ackerman (2000) technique was also used to obtain estimates of the direct radiation, because the measured-to-direct estimated clear-sky irradiance, too, provides information regarding the cloud impact on the surface radiation. Within the BSRN data stream the direct radiation incident on a surface normal to the zenith and the direct radiation incident normal to the direction to the sun are available. The empirical fit is accurate to within ±10% and is independent of whichever of these two parameters is used. However, we made use of the latter, and the fitting coefficients for this parameter were calculated as a = 1060 and b = 0.20.

An estimate for the diffuse irradiance under the clear sky was also performed. For this parameter we fit a fourth-order polynomial as a function of the SZA to the clear-sky measurements. As expected, the agreement between the measured and estimated clear-sky data is worse than that for global and direct irradiance, because the amount of clear-sky diffuse radiation is small and strongly influenced by the variability in the atmospheric aerosol content. The quality of the fit for this parameter is within the range of about ±20%.

Relationship between cloud fraction and solar radiation

As previously mentioned, in addition to the cloud fraction, cloud macro- and microphysical characteristics also strongly impact the surface radiation. Because of these various factors, a simple relationship between cloud fraction and downwelling surface radiation does not exist. Attempts to relate cloud cover and the surface radiation field have been made by Bais et al. (1993) and Frederick and Steele (1995). The latter developed a technique to estimate the attenuation of solar radiation from standard meteorological information on cloudiness and visibility. Taking cloud fraction, cloud ceiling altitude, and visibility into account, the parameterization used by Frederick and Steele (1995) led to differences between the observed and estimated radiation at the surface of greater than ±30% in about one-fourth of the cases.

Nevertheless, on longer timescales (e.g., 1 month) we observed a clear anticorrelation between cloud amount and the ratio of measured to estimated clear-sky irradiance, representing the overall reduction of downwelling radiation due to clouds. Monthly means of this ratio and the monthly means of the cloud fraction from Allsky2 were compared. Only data for Allsky2 have been used for calculating the monthly mean cloud fraction, because of gaps in the Allsky1 data series. For those months where both cameras were in operation simultaneously, the monthly mean total cloud fraction agrees to within a few percent, with a maximum difference of 8% and a minimum of 0.4%.

Results for monthly means of the irradiance ratio and the Allsky2 total cloud fraction are shown in Figs. 4a and 4b. In October 2000 no data were taken with the Allsky2 system because of technical difficulties. The monthly mean cloud fraction ranges from about 40% up to 67%, with lowest cloud fraction in March and April 2001, and the highest in July 2000. The corresponding median values are around 12% for March, 13% for April, and 98% for July. Averaged reductions in solar irradiance at the surface range from about 18% to 33% in comparison with expected clear-sky values; smallest reductions are observed for months with the lowest mean cloud fraction. As is somewhat evident in Fig. 4a, the correlation between the mean reduction in surface irradiance and the mean of the total/opaque cloud amount is high, with correlation coefficients of about −0.9.

The number of data points varies seasonally because of the differences in the length of the day and the removal of large zenith angle data (the number of data points varies from ∼10 000 to ∼20 000 in winter and summer, respectively). Although restricting the analysis to solar zenith angles of less than or equal to 75° removes early morning and evening observations, analysis of the all-sky cloud fractions does not show a preference for diminished (or enhanced) cloudiness for any given time of the day. Hence, eliminating low solar elevation data does not appear to significantly affect the statistics.

Cloud-related reduction and enhancement in solar radiation caused by clouds

Under certain conditions, clouds may actually enhance the downwelling solar radiation. Various examples for cloud enhancement situations are described in the literature, for example, Nack and Green (1974), Mims and Frederick (1994), Schafer et al. (1996), Estupinan et al. (1996), Thiel et al. (1997), or Segal and Davis (1992). Previous studies mainly concentrate on describing a limited number of selected situations. Instead, we attempt to classify situations that either reduce or enhance surface irradiance based on statistical analysis of the BSRN dataset, together with the cloud information from the all-sky imagers spanning more than 1 yr of observations.

Enhancements generally occur on short timescales; they tend not to be observed in data streams for which the mean irradiance is calculated using a temporal window of more than a few minutes. Herein, we use 1-min-averaged values to investigate the cloud impact on the downwelling surface solar radiation, but we also compare the frequency of occurrence of enhanced radiation levels for shorter (1-min peak values) and longer (5, 30, and 60 min) averaging periods.

One-minute-averaged values of global irradiance in the dataset show maximum values of up to about 1450 W m−2, which is about 6% higher than the solar constant at the mean Earth–Sun distance, and radiation values enhanced by up to about 70% in comparison with clear sky have been observed. Figure 5a shows the global irradiance and total cloud fraction for 16 December 2000 for a case in which the surface downwelling solar radiation is clearly enhanced. In Fig. 5b, the direct and diffuse components are shown for the enhanced time period (around 1500 NZST). In this example, 1-min-averaged values of the global irradiance are enhanced by as much as 20% over that of clear-sky estimates.

It can be seen that the enhancement in the global irradiance is associated with little or no reduction in the direct component and an increase in the diffuse component; that is, the cloud fields responsible for enhanced surface irradiance do not significantly reduce the direct beam but increase scattering and, thus, the diffuse component.

For the case illustrated in Fig. 5, manual inspection of the Allsky1 and Allsky2 images suggests altocumulus clouds around the position of the sun during the time of enhanced surface radiation. The accompanying all-sky data show an increase in the cloud fraction together with an unobscured sun disk for times when enhanced radiation values are observed (sunshine parameter not shown), whereas for the preceding period, in which the surface irradiance shows the typical daily variation of a clear-sky day, the measured cloud fraction is approximately zero.

Of special interest in the context of enhanced radiation values is their frequency of occurrence and the characterization of cloud situations associated with enhanced surface irradiance. For this reason, a statistical analysis was performed for time periods for which Allsky1, Allsky2, and BSRN measurements are available. Because the estimated clear-sky global irradiance is used to gauge the cloud impact, the analysis is restricted to solar zenith angles less than or equal to 75°. Further, only the data for which differences in the total cloud fraction between the two sky imagers are less than 10% are considered (∼45 000 data points; the results look very similar if these restrictions are not applied).

Figure 6 shows the ratio between the measured global irradiance and the clear-sky estimate as function of the total cloud fraction. In addition to the requirement that the cloud fraction for the two imagers agrees within ±10%, we limit the comparison to those cases in which their sunshine parameter matches. This helps to reduce the impact of errors due to differences in the sunshine parameter (this restriction reduces the number of data points by 8%). Figure 6a includes all of the selected data points, Fig. 6b includes only data where the corresponding sunshine parameter for both imagers identified unobscured sun, and, similarly, Fig. 6c includes only cases with obscured sun.

Despite the large scatter in the data, sun-free and sun-obscured cases are noticeably distinct. For those cases in which the sun was not shaded by clouds, most of the data points are concentrated around an irradiance ratio of 1.0 or higher. As is evident from Fig. 6b, clear-sky values are not necessarily associated with a zero cloud fraction, indicating that clear-sky values might be also observed under cloudy conditions. This happens when the reduction of the direct radiation component caused by clouds is compensated by an increase in the diffuse component due to enhanced scattering. Similarly, the same fractional cloud cover might either cause reduced or enhanced radiation values; in the next section we will identify the circumstances that lead to either enhanced or reduced irradiance.

From Fig. 6b (unobscured sun) it can be seen that there are some data points that show enhanced irradiance values but, at the same time, a cloud fraction of 100%. Partly this can be explained by uncertainties in the cloud fractions (e.g., if the cloud imagers do not detect a small gap in the cloud field near the sun's position; as we mentioned before the sun circle is a problematic region in cloud retrieval), but enhancements actually might occur at completely covered sky in a case in which the sun disk is covered by optically thin clouds that do not reduce the direct sunlight by a significant amount. It is also evident from Fig. 6b that largest values for the ratio of measured to estimated clear-sky surface irradiance occur predominantly in situations in which the cloud fraction has high values. Looking at the magnitude of enhancements in terms of percentage increase of the measured irradiance over the expected clear-sky value (not shown here), we observe this increase with cloud fraction at all solar zenith angles. Though, looking at the absolute value of the enhanced radiation in terms of irradiance units, this is only true for solar zenith angles of up to about 60°, while for larger solar zenith angles the dependency flattens. Absolute enhancements reach values of up to about 400 W m−2 at high sun and up to about 200 W m−2 at low sun.

Most of the data points in the sun-obscured case (Fig. 6c) are associated with high cloud fractions and show an irradiance ratio below 1. Values for which the irradiance ratio exceeds 1.0 for the sun-obscured case (∼1% of the data, Fig. 6c) are most likely an effect that is caused by uncertainties in the sunshine parameter or due to the strict separation into a sun-obscured and an unobscured sun class. Clouds that partially cover the sun disk, or optically thin clouds covering the sun disk, might cause the brightness to drop below the threshold used to determine whether the sun is obscured but might actually increase the global irradiance value. Irradiance ratios slightly smaller than 1.0 might result from errors in estimating the clear-sky irradiance.

To identify cloud situations associated with reduced or enhanced radiation values, the ratio of measured to estimated clear-sky global irradiance was binned using the following irradiance ratios: 0.0–0.5, 0.5–0.7, 0.7–0.9, 0.9–1.1, 1.1–1.3, 1.3–1.5, and 1.5–2.0. The bins ranging from 0.0 to 0.9 are referred to as “reduced,” while the bins ranging from 0.9 to 1.1 are referred to as “clear sky,” and those greater than 1.1 are “enhanced.” The frequency distribution for the different bins is shown in Fig. 7a. Of all cases that were investigated, 66% fall into the reduced class, while for 44% the reduction amounts to more than 50%. Clear-sky values (measured value within ±10% of expected clear-sky value) are observed in about 29% of the cases, while enhanced cases occur with a frequency of about 5%. Measured irradiance values exceed the expected clear-sky level by more than 30% in 1.3% of the cases.

Observations of enhanced radiation mainly occur over short time periods and strongly depend on the averaging time of the radiation measurements. If the 1-min maximum instead of the 1-min-averaged values are considered in this study, the frequency of enhancement situation increases from 5% to about 8%. As mentioned earlier, the 1-min maximum irradiance is subject to higher uncertainties than the 1-min-averaged values. Nevertheless, for 87% of enhanced situations the standard deviation during the 1-min period is <10% of the 1-min-averaged global irradiance, and this confirms that an enhanced 1-min maximum value is very likely due to actual enhanced irradiance and not caused by the higher measurement uncertainty. On the other hand, increasing the averaging period to 5 min reduces the frequency of enhanced surface irradiance to about 3.5%, and averaging the surface irradiance over 30 min reduces the frequency of enhanced cases to ∼1%. For 1-h mean values we did not observe irradiance ratios higher than 1.1 in the dataset.

For each of the specified ranges of the measured-to-estimated clear-sky irradiance ratio the mean, median, quartiles, outlier values, and far outlier values for various Allsky1 and radiation parameters were calculated. Figures 7b–h show Tukey plots (Tukey 1977) for various cloud and radiation parameters as a function of the measured-to-estimated clear-sky global irradiance ratio. The thick horizontal line in the plots represents the median. The boxes are bound by the 0.25 and 0.75 quartiles, which means that 50% of the data points are located within the range spanned by this box. Error bars denote the minimum and maximum values, not including outliers and far outlier values. Outliers are defined as points that are more than 1.5 times the interquartile range (range spanned by the 0.25 and 0.75 quartiles) away from the median, and far outlier values have a distance to the median of 3 times the interquartile range. Tukey plots are shown for the total (Fig. 7b) and thin (Fig. 7c) cloud fraction, the edge-to-area ratio (Fig. 7f), the ratio of measured to estimated direct irradiance (Fig. 7g), and the measured-to-estimated clear-sky diffuse irradiance (Fig. 7h). The sunshine parameter (Fig. 7d) and the sun circle cloud fraction (Fig. 7e) are also illustrated. Results for the opaque cloud fraction are not shown but look very similar to the total cloud fraction. The percentage contributions of outlier and far outlier values are not included in the plots but, for the sake of completeness, are given in Table 1.

Recall that the Allsky1 sun circle cloud-fraction data stream is limited to those periods for which the temporal sampling frequency is at least 1 min (data from September 1999 to June 2000). However, this does not significantly affect the results, because the frequency distribution of reduced, clear-sky, and enhanced scenes looks very similar for both datasets.

The results show clearly that reductions in the global irradiance are in most cases associated with a high total and opaque cloud fraction (Fig. 7b, opaque cloud fraction not shown). The sunshine parameter is below a value of 150 (i.e., the threshold in Allsky1 to separate obscured from unobscured sun) for most reduced-irradiance scenes (Fig. 7d), indicating that the sun disk was obscured by clouds. Consistent with the small sunshine parameter, the direct radiation is also reduced. The ratio of the measured to estimated clear-sky direct irradiances (Fig. 7g), which is close to zero for significant reductions in the global irradiance ratio (indicating that the direct beam is nearly completely obscured by optically thick clouds), increases for larger irradiance ratios, that is, less strongly reduced cases, and is consistent with a decrease in the total cloud fraction. Figure 7h compares the measured diffuse irradiance with a clear-sky value for the same SZA, clearly indicating the diffuse flux increases for both reductions and enhancements in the surface irradiance.

For the 29% of irradiance measurements within the interval around clear-sky values, (Fig. 7a), the all-sky systems observed smaller total cloud coverage as compared with the reduced global irradiance cases. It can be seen that the observed clear-sky irradiance does not require cloudless conditions because it is also observed in case of large cloud fractions (Fig. 7b). To achieve clear-sky values it is important that the direct radiation beam is nearly unaffected, as is shown by the direct irradiance ratio of ∼1 and the high sunshine parameter. The direct irradiance ratio is around 1.0 for most clear-sky and enhanced bins (Fig. 7g), although, as our calculations showed, in 17% (7%) of the clear-sky (enhanced) cases the measured direct component is more than 10% (30%) lower than the expected clear-sky level. This implies that clear-sky (enhanced) irradiance levels can also be attained even if the direct irradiance is reduced by clouds, as long as this decrease is offset by an increase in the diffuse component.

The enhanced cases (i.e., where the irradiance ratio is greater than 1.1) are associated with an increase in the total cloud fraction, and we observe a tendency of stronger enhancements in cases of larger total (and thin) cloud fraction. However, the range of the cloud-fraction values is still smaller when compared with the reduced cases. This happens because as cloud fraction increases the sun is more likely to be obscured, which causes a reduction in the surface irradiance. The outcome of this is that most frequently enhanced cases will be observed at moderate values of cloud fraction (when clouds are present, but they are not blocking the sun), while the majority of the reduced cases are observed when the sky is covered by clouds to a large extent, and the chance that clouds obscure the sun is high. For images in which the cloud fraction is in the range 10%–20% (∼10 500 images), reduced cases occurred in 3%, clear-sky cases in 96%, and enhanced cases in 0.5%. Cloud fractions within 40%–60% (∼1400 images) caused 42% reduced, 26% clear-sky, and 31% enhanced cases, while for cloud fractions in the range 80%–100% (∼30 000 images), the percentage for reduced, clear-sky, and enhanced cases is 91%, 5%, and 4%, respectively.

The sun circle cloud fraction shows the impact of near-sun clouds for the case of enhanced surface radiation. This parameter is also high for reduced cases and shows low values only for clear-sky cases. The high sun circle cloud fraction in combination with the high values for the sunshine parameter and the direct ratio of ∼1 confirm that clouds in the vicinity of the sun but not shading the sun play an important role in enhanced cases.

The large median value for the thin cloud fraction suggests that it contributes to enhanced surface irradiance, too (Fig. 7c). In contrast to broken cloud fields, for which reflections at clouds edges can actually increase the amount of diffuse irradiance (by “focusing” the available energy), which in turn can lead to an enhancement in surface global irradiance, enhanced surface irradiance values in cases of homogeneous thin clouds (cases in which the sun disk is partially or completely covered by thin clouds) are very likely a result of the angular redistribution of the diffuse radiation and the fact that irradiance is defined as the cosine-weighted solar flux. Optically thin clouds produce multiple scattering so that the angular distribution of the diffuse radiation field becomes nearly isotropic. The cosine weighting for the direct beam remains unchanged but the small amount of direct radiation that is turned into diffuse radiation experiences a higher cosine weighting at large solar zenith angles. This means the actual total flux will not be enhanced through a surface normal to the sun's direction, but only through a horizontal surface normal to the zenith.

Statistics of the dataset confirm that enhanced cases in association with thin clouds mainly occur at large solar zenith angles. Enhancements for scenes in which the thin cloud fraction as seen by the all-sky cameras is less than a few percent and the opaque cloud fraction dominates occur at all solar zenith angles, while for those cases in which a high, thin cloud and low, opaque cloud fraction are observed, enhanced radiation values are seen only for solar zenith angles greater than ∼50°.

A direct irradiance ratio around 1 for the enhanced bins confirms that enhanced radiation values occur only when there is a significant contribution from the direct beam. The high edge-to-area ratio indicates that broken clouds of small diameter cover the sky. The probability that solar radiation is reflected by cloud edges, but that clouds do not shade the direct beam, is higher than in the case of a small value for the edge-to-area ratio (as observed for the reduced cases).

Conclusions and prospects

Results of a study about the cloud impact on surface irradiance are presented. The work is based on cloud-coverage information from all-sky imaging systems and measurements of the downwelling global, direct, and diffuse solar irradiance in the wavelength range 0.3–3 μm. We used irradiance values expected for clear-sky situations as a reference to investigate the cloud impact on the surface radiation field. The reference values were generated by applying a least squares fit with the solar zenith angle as a variable to all global and direct irradiance data for which the corresponding all-sky images defined the sky as being free of clouds. The agreement between the estimated and the measured clear-sky values is clearly better than ±10% for solar zenith angles less than 75°.

The clear-sky reference values were also used to look at the changes in the surface radiation that are caused by clouds on a monthly basis. A comparison of the monthly mean irradiance fraction (ratio of the measured to expected clear-sky global irradiance) and the monthly median cloud fraction for the period from July 2000 to July 2001 provides a clear negative correlation with correlation coefficients of about −0.9.

The combined cloud and radiation dataset has also been analyzed with regard to the frequency of occurrence of situations in which clouds cause radiation levels at the surface that exceed the expected clear-sky values. The presence of clouds tends to increase the downwelling diffuse solar irradiance over that of corresponding clear-sky amounts; for example, optically thin clouds appear bright white, an indication that they are enhancing the diffuse radiation significantly. Thus, unless the clouds are optically thick enough to attenuate the diffuse solar radiation below clear-sky amounts, clouds will act to increase the diffuse radiation field. Consequently, whether the surface irradiance is enhanced over that which is expected under a clear sky depends on the direct beam. The same fractional cloud cover will usually cause a reduction below clear-sky values in the global irradiance if the direct beam is attenuated, that is, if optically thick clouds obscure the sun and the cloud-induced increase in the diffuse component cannot compensate for this loss (or is reduced). However, it might also cause an enhancement above clear-sky levels if the direct beam is only slightly obscured, that is, the sun disk is not obscured by clouds or is obscured by optically thin clouds.

Clouds associated with enhanced surface irradiance have been identified based on statistical analysis. For over 1 yr of 1-min-averaged global irradiance data, a reduction of more than 10% in the incoming solar irradiance in comparison with clear-sky values occurred about 66% of the time, while nearly clear sky values were observed in 29% of all cases. Radiation enhancements (measured values exceed the estimated clear-sky values by more than 10%) were detected with a frequency of 5% for the measurement site, but there might be regions where the probability for cloud situations favoring enhanced radiation levels is higher, for example, regions dominated by convective clouds (Segal and Davis 1992).

Our results show that enhancements, that is, situations in which the surface irradiance exceeds expected clear-sky values, occur mainly under broken clouds when the direct component of solar radiation is nearly unaffected, while the diffuse radiation is increased in comparison with clear-sky values. Enhanced surface irradiance is associated with partly cloudy conditions for which clouds do not obscure the sun, and with optically thin clouds that may obscure the sun but do not reduce the direct radiation by as much as they enhance the diffuse irradiance. The former mechanism is important for high sun angles, and can produce large absolute enhancement (in terms of watts per square meter), while the latter applies to low sun angles and can produce large relative enhancements (in terms of percent) in cases of large cloud fraction.

Although the cloud fraction is a dominant factor influencing the amount of solar radiation reaching the surface, it is also obvious that information about the cloud coverage alone, even together with the identification of whether the sun is obscured, is not sufficient to explain the actual radiation field, especially if short time variations are considered. The spatial distribution of clouds and their optical properties also strongly influence the cloud impact. Nevertheless it is shown in this analysis that parameters from all-sky imaging provide a helpful source of information for analyzing the influence of clouds in the radiation field and that all-sky cameras are suitable instruments to perform long-term measurements of local cloud coverage with high temporal resolution.

Acknowledgments

This work was carried out as part of NIWA Contract CO1X0033 for the New Zealand Foundation for Research Science and Technology (FRST). We gratefully acknowledge Jill Scott and Mike Kotkamp for help with the operational requirements and data archival of the instruments. The authors thank the anonymous reviewers for helpful comments.

REFERENCES

  • Bais, A. F., C. S. Zerefos, C. Meleti, I. C. Ziomas, and K. Tourpali. 1993. Spectral measurements of solar UVB radiation and its relation to total ozone, SO2, and clouds. J. Geophys. Res. 98 D3:51995204.

    • Search Google Scholar
    • Export Citation
  • Bodeker, G. E. and R. L. McKenzie. 1996. An algorithm for inferring surface UV radiation including cloud effects. J. Appl. Meteor. 35:18601877.

    • Search Google Scholar
    • Export Citation
  • Cess, R. D. Coauthors,. 1995. Absorption of solar radiation by clouds: Observations versus models. Science 267:496499.

  • Estupinan, J. G., S. Raman, G. H. Crescenti, J. J. Streicher, and W. F. Barnard. 1996. Effect of clouds and haze on UV-B radiation. J. Geophys. Res. 101 D11:1680716816.

    • Search Google Scholar
    • Export Citation
  • Frederick, J. E. and H. D. Steele. 1995. The transmission of sunlight through cloudy skies: An analysis based on standard meteorological information. J. Appl. Meteor. 34:27552761.

    • Search Google Scholar
    • Export Citation
  • Long, C. N. and J. J. DeLuisi. 1998. Development of an automated hemispheric sky imager for cloud fraction retrievals. Preprints, 10th Symp. on Meteorological Observations and Instrumentation, Phoenix, AZ, Amer. Meteor. Soc., 171–174.

    • Search Google Scholar
    • Export Citation
  • Long, C. N. and T. P. Ackerman. 2000. Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res. 105:15;th60915;th626.

    • Search Google Scholar
    • Export Citation
  • Long, C. N., D. W. Slater, and T. Tooman. 2001. Total Sky Imager (TSI) model 880 status and testing results. Atmospheric Radiation Measurement Program Tech. Rep. ARM TR-006, 36 pp. [Available online at http://www.arm.gov/docs/documents/tech_reports/index.html.].

    • Search Google Scholar
    • Export Citation
  • McKenzie, R. L., K. J. Paulin, G. E. Bodeker, J. B. Liley, and A. P. Sturman. 1998. Cloud cover measured by satellite and from the ground: Relationship to UV radiation at the surface. Int. J. Remote Sens. 19:29692985.

    • Search Google Scholar
    • Export Citation
  • Mims III, F. M. and J. E. Frederick. 1994. Cumulus clouds and UV-B. Nature 371:291.

  • Nack, M. L. and A. E. S. Green. 1974. Influence of clouds, haze, and smog on the middle ultraviolet reaching the ground. Appl. Opt. 13:24052415.

    • Search Google Scholar
    • Export Citation
  • Ohmura, A. Coauthors,. 1998. Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research. Bull. Amer. Meteor. Soc. 79:21152136.

    • Search Google Scholar
    • Export Citation
  • Robinson, G. D. 1966. Solar Radiation. Elsevier, 347 pp.

  • Schafer, J. S., V. K. Saxena, and J. J. De Luisi. 1996. Observed influence of clouds on ultraviolet-B radiation. Geophys. Res. Lett. 23:26252628.

    • Search Google Scholar
    • Export Citation
  • Segal, M. and J. Davis. 1992. The impact of deep cumulus reflection on the ground-level global irradiance. J. Appl. Meteor. 31:217222.

    • Search Google Scholar
    • Export Citation
  • Thiel, S., K. Steiner, and H. K. Seidlitz. 1997. Modification of global erythemally effective irradiance by clouds. Photochem. Photobiol. 65:969973.

    • Search Google Scholar
    • Export Citation
  • Tukey, J. W. 1977. Explanatory Data Analysis. Addison-Wesley, 506 pp.

Fig. 1.
Fig. 1.

(a), (c) Cloud image and (b), (d) analysis image taken with (a), (b) Allsky1 and (c), (d) Allsky2 at 1322 NZST at 21 Oct 2001. The cross in (b) and the dot in (d) indicate the position of the sun. Lines in (b) denote the horizontal area (trapezoid area around the horizon), sun circle (circle around the sun's position), and zenith circle (circle around zenith)

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 2.
Fig. 2.

Frequency distribution for the difference [in percent cloud fraction (CF)] between Allsky1 and Allsky2 total, opaque, and thin cloud fraction

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 3.
Fig. 3.

Absolute difference in total, opaque, and thin CF between Allsky1 and Allsky2 if only images are considered for which the sunshine parameter from both imagers is (a), (c), (e) true, which means unobscured sun, or (b), (d), (f) false, which means sun obscured by clouds

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 4.
Fig. 4.

(a) Monthly mean of total and opaque CF from Allsky2 against monthly mean irradiance fraction (ratio of measured to estimated clear-sky global solar irradiance). (b) Bar plot showing the mean of the total cloud fraction and mean irradiance fraction for the individual months. Values for Jan–Jun are for 2001, Jul–Dec for 2000

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 5.
Fig. 5.

(a) Daily course of global irradiance and total CF for Allsky2 for 16 Dec 2000. (b) Global, direct, and diffuse irradiance for the period in the afternoon during which enhanced radiation values occurred. The estimated clear-sky global irradiance is shown, too

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 6.
Fig. 6.

Ratio of measured-to-estimated clear-sky global irradiance as function of the total CF, for (a) entire dataset as mentioned in the text and (b) only data for which the Allsky2 sunshine parameter identified unobscured sun. (c) Same as (b), but for situations in which the sun was obscured by clouds

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Fig. 7.
Fig. 7.

(a) Frequency of occurrence for reduction (irradiance ratio below 0.9), clear-sky (irradiance ratio between 0.9 and 1.1), and enhancement (irradiance ratio above 1.1) scenes. (b)–(h) Tukey plots showing median, quartiles, minimum, and maximum for different Allsky1 and radiation parameters as function of the ratio of measured to estimated clear-sky global irradiance

Citation: Journal of Applied Meteorology 42, 10; 10.1175/1520-0450(2003)042<1421:CCBOAI>2.0.CO;2

Table 1.

Percentage contribution of outside and far outside values for selected parameters shown in Fig. 7

Table 1.
Save
  • Bais, A. F., C. S. Zerefos, C. Meleti, I. C. Ziomas, and K. Tourpali. 1993. Spectral measurements of solar UVB radiation and its relation to total ozone, SO2, and clouds. J. Geophys. Res. 98 D3:51995204.

    • Search Google Scholar
    • Export Citation
  • Bodeker, G. E. and R. L. McKenzie. 1996. An algorithm for inferring surface UV radiation including cloud effects. J. Appl. Meteor. 35:18601877.

    • Search Google Scholar
    • Export Citation
  • Cess, R. D. Coauthors,. 1995. Absorption of solar radiation by clouds: Observations versus models. Science 267:496499.

  • Estupinan, J. G., S. Raman, G. H. Crescenti, J. J. Streicher, and W. F. Barnard. 1996. Effect of clouds and haze on UV-B radiation. J. Geophys. Res. 101 D11:1680716816.

    • Search Google Scholar
    • Export Citation
  • Frederick, J. E. and H. D. Steele. 1995. The transmission of sunlight through cloudy skies: An analysis based on standard meteorological information. J. Appl. Meteor. 34:27552761.

    • Search Google Scholar
    • Export Citation
  • Long, C. N. and J. J. DeLuisi. 1998. Development of an automated hemispheric sky imager for cloud fraction retrievals. Preprints, 10th Symp. on Meteorological Observations and Instrumentation, Phoenix, AZ, Amer. Meteor. Soc., 171–174.

    • Search Google Scholar
    • Export Citation
  • Long, C. N. and T. P. Ackerman. 2000. Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res. 105:15;th60915;th626.

    • Search Google Scholar
    • Export Citation
  • Long, C. N., D. W. Slater, and T. Tooman. 2001. Total Sky Imager (TSI) model 880 status and testing results. Atmospheric Radiation Measurement Program Tech. Rep. ARM TR-006, 36 pp. [Available online at http://www.arm.gov/docs/documents/tech_reports/index.html.].

    • Search Google Scholar
    • Export Citation
  • McKenzie, R. L., K. J. Paulin, G. E. Bodeker, J. B. Liley, and A. P. Sturman. 1998. Cloud cover measured by satellite and from the ground: Relationship to UV radiation at the surface. Int. J. Remote Sens. 19:29692985.

    • Search Google Scholar
    • Export Citation
  • Mims III, F. M. and J. E. Frederick. 1994. Cumulus clouds and UV-B. Nature 371:291.

  • Nack, M. L. and A. E. S. Green. 1974. Influence of clouds, haze, and smog on the middle ultraviolet reaching the ground. Appl. Opt. 13:24052415.

    • Search Google Scholar
    • Export Citation
  • Ohmura, A. Coauthors,. 1998. Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate research. Bull. Amer. Meteor. Soc. 79:21152136.

    • Search Google Scholar
    • Export Citation
  • Robinson, G. D. 1966. Solar Radiation. Elsevier, 347 pp.

  • Schafer, J. S., V. K. Saxena, and J. J. De Luisi. 1996. Observed influence of clouds on ultraviolet-B radiation. Geophys. Res. Lett. 23:26252628.

    • Search Google Scholar
    • Export Citation
  • Segal, M. and J. Davis. 1992. The impact of deep cumulus reflection on the ground-level global irradiance. J. Appl. Meteor. 31:217222.

    • Search Google Scholar
    • Export Citation
  • Thiel, S., K. Steiner, and H. K. Seidlitz. 1997. Modification of global erythemally effective irradiance by clouds. Photochem. Photobiol. 65:969973.

    • Search Google Scholar
    • Export Citation
  • Tukey, J. W. 1977. Explanatory Data Analysis. Addison-Wesley, 506 pp.

  • Fig. 1.

    (a), (c) Cloud image and (b), (d) analysis image taken with (a), (b) Allsky1 and (c), (d) Allsky2 at 1322 NZST at 21 Oct 2001. The cross in (b) and the dot in (d) indicate the position of the sun. Lines in (b) denote the horizontal area (trapezoid area around the horizon), sun circle (circle around the sun's position), and zenith circle (circle around zenith)

  • Fig. 2.

    Frequency distribution for the difference [in percent cloud fraction (CF)] between Allsky1 and Allsky2 total, opaque, and thin cloud fraction

  • Fig. 3.

    Absolute difference in total, opaque, and thin CF between Allsky1 and Allsky2 if only images are considered for which the sunshine parameter from both imagers is (a), (c), (e) true, which means unobscured sun, or (b), (d), (f) false, which means sun obscured by clouds

  • Fig. 4.

    (a) Monthly mean of total and opaque CF from Allsky2 against monthly mean irradiance fraction (ratio of measured to estimated clear-sky global solar irradiance). (b) Bar plot showing the mean of the total cloud fraction and mean irradiance fraction for the individual months. Values for Jan–Jun are for 2001, Jul–Dec for 2000

  • Fig. 5.

    (a) Daily course of global irradiance and total CF for Allsky2 for 16 Dec 2000. (b) Global, direct, and diffuse irradiance for the period in the afternoon during which enhanced radiation values occurred. The estimated clear-sky global irradiance is shown, too

  • Fig. 6.

    Ratio of measured-to-estimated clear-sky global irradiance as function of the total CF, for (a) entire dataset as mentioned in the text and (b) only data for which the Allsky2 sunshine parameter identified unobscured sun. (c) Same as (b), but for situations in which the sun was obscured by clouds

  • Fig. 7.

    (a) Frequency of occurrence for reduction (irradiance ratio below 0.9), clear-sky (irradiance ratio between 0.9 and 1.1), and enhancement (irradiance ratio above 1.1) scenes. (b)–(h) Tukey plots showing median, quartiles, minimum, and maximum for different Allsky1 and radiation parameters as function of the ratio of measured to estimated clear-sky global irradiance

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 2375 693 65
PDF Downloads 1726 449 39