• Baum, B. A., , and J. D. Spinhirne, 2000: Remote sensing of cloud properties using MODIS airborne simulator imagery during SUCCESS. III: Cloud overlap. J. Geophys. Res., 105 , 1179311804.

    • Search Google Scholar
    • Export Citation
  • Baum, B. A., and Coauthors, 1995: Satellite remote sensing of multiple cloud layers. J. Atmos. Sci., 52 , 42104230.

  • Chen, T., , Y. Zhang, , and W. B. Rossow, 2000: Sensitivity of atmospheric radiative heating rate profiles to variations of cloud layer overlap. J. Climate, 13 , 29412959.

    • Search Google Scholar
    • Export Citation
  • Gupta, S. K., , W. L. Darnell, , and A. C. Wilber, 1992: A parameterization for longwave surface radiation from satellite data: Recent improvements. J. Appl. Meteor., 31 , 13611367.

    • Search Google Scholar
    • Export Citation
  • Hahn, C. J., , and S. G. Warren, 1999: Extended edited cloud reports from ships and land stations over the globe, 1952–1996. Numerical Data Package NDP-026C, Carbon Dioxide Information Analysis Center, Oak Ridge, TN, 79 pp.

  • Hahn, C. J., , and S. G. Warren, 2002: Cloud climatology for land stations worldwide, 1971–1996. Numerical Data Package NDP-026D, Carbon Dioxide Information Analysis Center, Oak Ridge, TN, 35 pp.

  • Heidinger, A. K., , R. Frey, , and M. J. Pavolonis, 2004: Relative merits of the 1.6 and 3.75 micron channels of the AVHRR/3 for cloud detection. Can. J. Remote Sens., 30 , 182194.

    • Search Google Scholar
    • Export Citation
  • Ho, S., , B. Lin, , P. Minnis, , and T. Fan, 2003: Estimates of cloud vertical structure and water amount over tropical oceans using VIRS and TMI data. J. Geophys. Res., 108 .4419, doi:10.1029/2002JD003298.

    • Search Google Scholar
    • Export Citation
  • Ignatov, A., , I. Laszlo, , E. Harrod, , K. Kidwell, , and G. Goodrum, 2004: Equator crossing times for NOAA, ERS and EOS sun-synchronous satellites. Int. J. Remote Sens., 25 , 52555266.

    • Search Google Scholar
    • Export Citation
  • Inoue, T., 1985: On the temperature and effective emissivity determination of semi-transparent cirrus clouds by bi-spectral measurements in the 10 μm window region. J. Meteor. Soc. Japan, 64 , 8899.

    • Search Google Scholar
    • Export Citation
  • Jin, Y., , and W. B. Rossow, 1997: Detection of cirrus overlapping low-level clouds. J. Geophys. Res., 102 , 17271737.

  • Lin, B., , P. Minnis, , B. Wielicki, , D. R. Doelling, , R. Palikonda, , D. F. Young, , and T. Uttal, 1998: Estimation of water cloud properties from satellite microwave, infrared and visible measurements in oceanic environment: 2. Results. J. Geophys. Res., 103 , 38873905.

    • Search Google Scholar
    • Export Citation
  • Morcrette, J. J., , and C. Jakob, 2000: The response of the ECMWF model to changes in the cloud overlap assumption. Mon. Wea. Rev., 128 , 17011732.

    • Search Google Scholar
    • Export Citation
  • Nasiri, S. L., , and B. A. Baum, 2004: Daytime multilayered cloud detection using multispectral imager data. J. Atmos. Oceanic Technol., 21 , 11451155.

    • Search Google Scholar
    • Export Citation
  • Pavolonis, M. J., , and J. R. Key, 2003: Antarctic cloud radiative forcing at the surface estimated from the AVHRR polar pathfinder and ISCCP D1 datasets, 1985–93. J. Appl. Meteor., 42 , 827840.

    • Search Google Scholar
    • Export Citation
  • Pavolonis, M. J., , and A. K. Heidinger, 2004: Daytime cloud overlap detection from AVHRR and VIIRS. J. Appl. Meteor., 43 , 762778.

  • Pavolonis, M. J., , A. K. Heidinger, , and T. Uttal, 2005: Daytime global cloud typing from AVHRR and VIIRS: Algorithm description, validation, and comparisons. J. Appl. Meteor., 44 , 804826.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2002: The CloudSat mission and the A-Train. Bull. Amer. Meteor. Soc., 83 , 17711790.

  • Wang, J., , and W. B. Rossow, 1998: Effects of cloud vertical structure on atmospheric circulation in the GISS GCM. J. Climate, 11 , 30103029.

    • Search Google Scholar
    • Export Citation
  • Wang, J., , W. B. Rossow, , and Y. Zhang, 2000: Cloud vertical structure and its variations from a 20-yr global rawinsonde dataset. J. Climate, 13 , 30413056.

    • Search Google Scholar
    • Export Citation
  • Warren, S. G., , C. J. Hahn, , and J. London, 1985: Simultaneous occurrence of different cloud types. J. Climate Appl. Meteor., 24 , 658667.

    • Search Google Scholar
    • Export Citation
  • Wielicki, B. A., , R. D. Cess, , M. D. King, , D. A. Randall, , and E. F. Harrison, 1995: Mission to Planet Earth: Role of clouds and radiation in climate. Bull. Amer. Meteor. Soc., 76 , 21252153.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Broadband infrared heating rate profiles computed for a standard tropical atmosphere. The solid line shows the profile for an atmosphere with a cirrus cloud overlying a low-level water cloud. The dotted line shows the profile computed for an atmosphere with a single cirrus cloud. The total column optical depths and radiative cloud-top temperatures are the same in both atmospheres.

  • View in gallery

    Pixel-level detection of cirrus overlap using the AVHRR technique applied to MODIS data at 1915 UTC 4 Apr 2003. The data are from the tropical eastern Pacific. (top) RGB image using bands 1 (0.65 μm), 6 (1.6 μm), and 31 (11 μm). (bottom) Pixels that passed the AVHRR cirrus cloud overlap test as a gray mask.

  • View in gallery

    Variation of the equator crossing time (local) for the NOAA polar orbiting satellites. The triangles represent the July and January months chosen to have approximately the same equator crossing time.

  • View in gallery

    Mean percentage of AVHRR pixels being classified as cirrus overlap for the 4 yr (1982, 1986, 1991, and 1998) of July data. Values south of 60°S are not shown due to the lack of daytime results. The spatial resolution is 0.5°.

  • View in gallery

    Mean percentage of AVHRR pixels being classified as cirrus overlap for the 4 yr (1983, 1987, 1992, and 1999) of January data. Values north of 60°N are not shown due to the lack of daytime results. The spatial resolution is 0.5°.

  • View in gallery

    Zonal mean distribution of the percentage of AVHRR pixels typed as overlapped cirrus. Results represent the mean based on the four years of data and have a spatial resolution of 0.5°. Values for regions where daytime results where not available are missing.

  • View in gallery

    July zonal distribution of the mean value of the percentage of AVHRR pixels over land classified as cirrus overlap. CLAVR-x results represent mean values for the four years of data processed. Hahn and Warren results represent mean values from 1971 to 1996. Spatial resolution of the CLAVR-x results in 0.5°. Spatial resolution of the Hahn and Warren results is 5°. First number in parentheses is the mean value from 60°S to 60°N. Second number in parentheses is the mean value from 20°S to 20°N.

  • View in gallery

    January zonal distribution of the mean value of the percentage of AVHRR pixels over land classified as cirrus overlap. CLAVR-x results represent mean values for the 4 yr of data processed. Hahn and Warren results represent mean values from 1971 to 1996. Spatial resolution of the CLAVR-x results in 0.5°. Spatial resolution of the Hahn and Warren results is 5°. First number in parentheses is the mean value from 60°S to 60°N. Second number in parentheses is the mean value from 20°S to 20°N.

  • View in gallery

    July and January zonal distributions of the mean value of the percentage of AVHRR pixels classified as ice clouds that are also typed as cirrus overlap. CLAVR-x results represent mean values for the four years of data processed. Spatial resolution of results in 0.5°. Values for regions where daytime results where not available are missing.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 4 4 0
PDF Downloads 9 9 3

Global Daytime Distribution of Overlapping Cirrus Cloud from NOAA’s Advanced Very High Resolution Radiometer

View More View Less
  • 1 NOAA/NESDIS/Office of Research and Applications, Madison, Wisconsin
  • 2 Cooperative Institute for Meteorological Studies, University of Wisconsin—Madison, Madison, Wisconsin
© Get Permissions
Full access

Abstract

Data from the National Oceanic and Atmospheric Administration’s (NOAA’s) Advanced Very High Resolution Radiometer (AVHRR) instrument are used to provide the mean July and January global daytime distributions of multilayer cloud, where multilayer cloud is defined as cirrus overlapping one or more lower layers. The AVHRR data were taken from multiple years that were chosen to provide data with a constant local equator crossing time of 1430–1500 local time. The cloud overlap detection algorithm is used in NOAA’s Extended Clouds from AVHRR (CLAVR-x) processing system. The results between 60°N and 60°S indicated that roughly 20% of all clouds and roughly 40% of all ice clouds were classified as cirrus overlapping lower cloud (cirrus overlap). The results show a strong July–January pattern that is consistent with the seasonal cycle in convection. In some regions, cirrus overlap is found to be the dominant type of cloud observed. The distributions of overlapping cirrus cloud presented here are compared with results from other studies based on rawinsondes and manual surface observations. Comparisons are also made with another satellite-derived study that used coincident infrared and microwave observations over the tropical oceans during a 6-month period

Corresponding author address: Andrew K. Heidinger, 1225 West Dayton, Madison, WI 53706. Email: Andrew.Heidinger@noaa.gov

Abstract

Data from the National Oceanic and Atmospheric Administration’s (NOAA’s) Advanced Very High Resolution Radiometer (AVHRR) instrument are used to provide the mean July and January global daytime distributions of multilayer cloud, where multilayer cloud is defined as cirrus overlapping one or more lower layers. The AVHRR data were taken from multiple years that were chosen to provide data with a constant local equator crossing time of 1430–1500 local time. The cloud overlap detection algorithm is used in NOAA’s Extended Clouds from AVHRR (CLAVR-x) processing system. The results between 60°N and 60°S indicated that roughly 20% of all clouds and roughly 40% of all ice clouds were classified as cirrus overlapping lower cloud (cirrus overlap). The results show a strong July–January pattern that is consistent with the seasonal cycle in convection. In some regions, cirrus overlap is found to be the dominant type of cloud observed. The distributions of overlapping cirrus cloud presented here are compared with results from other studies based on rawinsondes and manual surface observations. Comparisons are also made with another satellite-derived study that used coincident infrared and microwave observations over the tropical oceans during a 6-month period

Corresponding author address: Andrew K. Heidinger, 1225 West Dayton, Madison, WI 53706. Email: Andrew.Heidinger@noaa.gov

1. Introduction

Knowledge of the vertical profile of cloudiness in the atmosphere is a fundamental piece of information often missing in standard satellite-derived cloud data. In recognition of this, the National Aeronautics and Space Administration (NASA) is funding new satellite missions such as Cloud Satellite (CLOUDSAT) and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) with active sensors to provide information on the three-dimensional distribution of cloudiness (Stephens et al. 2002). Methods to discriminate multilayer from single-layer cloudiness using passive satellite observations have been published previously (Baum et al. 1995; Jin and Rossow 1997; Lin et al. 1998; Baum and Spinhirne 2002; Nasiri and Baum 2004). In this study, the results of a recently published algorithm described by Pavolonis and Heidinger (2004, hereafter PH04) are used. What is unique about this algorithm is that it is applicable over land and ocean, unlike the previous techniques that use microwave observations. While the methodology described in papers by Baum et al. (1995) and Nasiri and Baum (2004) is applicable over land and water, results of its global application over a long period have not been published yet. One other benefit of the approach outlined in PH04 is that it can be applied to the roughly 25 years of existing Advanced Very High Resolution Radiometer (AVHRR) data; other methods require additional spectral channels not available on most AVHRRs or collocated sounder data. Therefore, results of this study appear to be one of the first multiyear global (land + ocean) surveys of daytime multilayer cloud in the remote sensing literature derived solely from satellite observations.

As described by PH04, the multilayer cloud situation that is most often detected corresponds to the situation of semitransparent cirrus cloud overlapping a lower-level cloud of moderate optical thickness. The term cirrus overlap will be used throughout this paper to refer to the multilayer cloud conditions detected by this algorithm. The observations used here are taken from the AVHRR flown on the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites. Global Area Coverage (GAC) AVHRR data, which have a nominal spatial resolution of about 4 km, are used in this study.

Examples of the importance of the vertical profile of cloudiness on derived radiative fluxes and the impact of cloud on climate can be found in the literature. For example, Wang and Rossow (1998) used global circulation model (GCM) simulations to demonstrate that large-scale features, such as the strength of Hadley circulation and the distribution of precipitation, are sensitive to the prescription of the cloud vertical structure. Further, Morcrette and Jakob (2000) found that the surface and top of the atmosphere radiative fluxes varied significantly in the European Centre for Medium-Range Weather Forecasts (ECMWF) general circulation model when the cloud overlap scheme was varied. Gupta et al. (1992) and Wielicki et al. (1995) showed that the earth radiation budget will be largely influenced by the vertical location and coverage of clouds.

As stated above, most standard cloud datasets derived from satellite data ignore cloud overlap. In addition, many of these satellite-derived cloud products are used to infer radiative fluxes and heating rates. For example, Pavolonis and Key (2003) used cloud data from the International Satellite Cloud Climatology Project (ISCCP) to compute radiative fluxes in the polar regions. In general, ISCCP and other passive satellite products assume single-layer clouds. In ISCCP, the properties of the single-layer clouds are defined by using a visible reflectance to derive a cloud optical depth and a window channel brightness temperature to define the cloud-top temperature.

One of the most obvious and direct errors caused by treating a multilayer cloud as a single-layer cloud occurs in the profile of longwave radiative heating. Figure 1 illustrates the potential errors in the broadband heating rate and it shows two heating rate profiles computed using a standard tropical atmosphere. For the multilayer cloud simulation, the cirrus cloud was positioned between 200 and 240 hPa with a visible optical depth of 2 and the lower-level cloud was positioned between 900 and 920 hPa with an optical depth of 8. As described by PH04, this scenario is one that should be easily detected as overlapped cirrus. The single-layer cirrus cloud was constructed to produce the same top of the atmosphere visible reflectance and window channel radiance as the multilayer simulation. Therefore, the single-layer cloud simulation represents the heating profile derived from the information available from most current passive satellite systems that ignore cloud overlap. The primary effect is that strong cloud-top cooling and cloud base warming of the lower cloud and the cloud-base warming of the higher cloud are replaced by a warming at the base of the single-layer cloud (which falls outside the clouds in the multilayer case). In addition, the cloud-top warming of the single-layer cloud is displaced downward due to erroneous cloud top height assignment. In general, the heating rate profiles of the single-layer cloud show less gradient than those in the multilayer simulation. As demonstrated later, the presence of multilayer clouds is quite common in some regions. It is clear from simulations, like the one presented here, that treating multilayer clouds as single-layer clouds poses severe limitations for deducing the radiative impact of clouds on the atmosphere from cloud products that ignore multilayer clouds. A more rigorous discussion of these effects is presented by Chen et al. (2000).

2. Cirrus overlap detection algorithm

The algorithm used to detect overlapping clouds in this study was described by PH04. The cloud overlap detection algorithm was written to be part of a larger cloud-typing algorithm. Specifically, the AVHRR cloud overlap algorithm is run within the cloud-type algorithm used by NOAA in the Extended Clouds from AVHRR (CLAVR-x) system. All results shown in this paper are produced using the system operated and run at NOAA.

The AVHRR provides measurements with central wavelengths of 0.63, 0.86, 3.75, 10.8, and 12.0 μm. Newer versions of the AVHRR allow for measurements at 1.6 μm to be substituted for the 3.75-μm measurements. Because this change comes late in the AVHRR data record, no results from the AVHRR with the 1.6-μm channel are shown. As noted by PH04, the presence of more spectral channels on the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Visible/Infrared Imager/Radiometer Suite (VIIRS) allows for additional techniques for detecting cloud overlap.

The physical basis of this approach is that for single-layer clouds, the variation of the 0.63-μm reflectance (R0.63) and the 10.8- minus 12.0-μm split-window brightness temperature difference (T11T12) should follow a well-defined curve. For single-layer clouds of moderate optical thickness, plane-parallel radiative transfer predicts that as R0.63 increases, T11T12 should decrease (Inoue 1985). The detection of cloud overlap is then just the detection of clouds that do not behave radiatively as single-layer clouds. For example, if a single-layer cloud is physically split into a semitransparent high cloud and a moderately thick low cloud (as done in the heating rate simulation in the last section), the temperature separation between the two layers will cause an elevated T11T12 value that is inconsistent with the value of R0.63. In practice, the thresholds for the detection of cloud overlap using R0.63 and T11T12 are functions of solar and viewing angles. Additional constraints are also applied to help prevent all single-layer cloud edges and single-layer thin cirrus located in regions of strong sunglint from passing the cloud overlap test. More specifically, T11 must be <270 K and R0.63 > 30%. These constraints may cause some thin cirrus (optical depth <0.5) over lower cloud or cirrus of optical depth 0.5–1.0 over broken or thin lower cloud to be missed.

As described in PH04, the optimal performance of this method occurs when a cirrus cloud with an optical depth between 0.5 and 3 occurs over a lower and warmer cloud with an optical depth greater than 5. In addition, our algorithm also works best when the clouds completely fill the 4-km field of view. Validation of the algorithm using cloud vertical profiles derived from surface-based radar indicates that many cirrus overlap scenarios that fall outside of the optimal conditions are detectable. As pointed out by Warren et al. (1985), cirrus often occur above altostratus (As), stratus (St), and cumulus (Cu). Our validation studies show that most occurrences of cirrus over all of these cloud types are detectable. The algorithm is not designed, however, to detect the occurrence of midlevel cloud (i.e., As) over low-level cloud (St, Cu) though the detection of some midlevel cloud over low-level cloud is possible.

One limitation of this technique is that surfaces that are highly reflective at 0.63 μm can cause single-layer cirrus to be falsely detected as cirrus overlap. To limit this false detection rate, the detection of cirrus overlap is prohibited for surfaces that are classified as either barren desert or perennially covered by snow or ice. This algorithm feature explains the total lack of cloud overlap occurring over desert regions, Greenland, and Antarctica in the following results. The cloud atlases of Warren et al. (1985), however, indicate that the occurrence of multilayer cloud over deserts and polar regions is relatively minor. The cloud overlap algorithm is used over all other land surfaces even though snow cover may be present as certain times of the year. This may result in an increase of false detection if single-layer cirrus are common over snow-covered land. To reduce the false detection of cirrus overlap over snow-covered land, the algorithm was modified to include a test of the 3.75-μm reflectance. The physical basis for this test is that water clouds have a high 3.75-μm reflectance while ice clouds (cirrus) and snow/ice surfaces have low values of the 3.75-μm reflectance. If the cirrus is optically thin, cirrus over water should have a higher 3.75-μm reflectance than cirrus over snow. Therefore, pixels with small values of the 3.75-μm reflectance are prohibited from being classified as multilayer. The description of this modification is presented in Pavolonis et al. (2005). With this constraint, the risk for false multilayer detection over snow-covered surfaces is reduced. As pointed out in PH04, the additional spectral information of the MODIS, VIIRS, and the Advanced Baseline Imager (ABI) provide for more accurate detection of cirrus overlap over snow-covered and desert surfaces.

To illustrate the performance of this algorithm at the pixel level, Fig. 2 shows an example from a region east of Hawaii illustrating the performance of the cloud overlap detection at the pixel level for a scene with significant amounts of cirrus overlap. The data shown in Fig. 2 are actually taken from MODIS at 1915 UTC 4 April 2003. Using MODIS data allows for validation of the AVHRR algorithm using channels missing from the AVHRR. The spatial resolution of the MODIS data is 1 km, compared to the roughly 4-km resolution of the AVHRR GAC data used in this study. The spatial and spectral differences between AVHRR and the comparable MODIS channels do not appear to significantly affect the performance of the algorithm. The top image in Fig. 2 shows a RGB image using the 0.65-, 16-, and 11-μm channels from MODIS. In this RGB, the ice clouds appear purple, warm water clouds appear yellow, and midlevel clouds appear white/off-white. The bottom image shows the results of the AVHRR cirrus overlap detection (shown in gray) as described in PH04. By visual inspection, once can clearly discern that many of the ice clouds are overlying lower water clouds and that the cirrus overlap detection algorithm appears to be working well. PH04 provides a rigorous validation of this technique based on surface-based measurements and includes a sensitivity analysis.

3. Data

The AVHRRs on the afternoon NOAA polar-orbiting satellites have provided a spectrally and spatially consistent global dataset from 1981 to the present. Only the potential inclusion (exclusion) of the 1.6 (3.75)-μm channel on the NOAA-KLM series of spacecraft starting in 1998 presents a discontinuity in the radiometric record (Heidinger et al. 2004). Therefore the AVHRR provides a unique set of consistent imager observations potentially useful for decadal-scale climate studies. In recognition of this, the NOAA/National Environmental Satellite, Data, and Information Service (NESDIS) Office of Research and Applications (ORA) is developing the capability to reprocess and analyze the entire AVHRR global record from both the morning and afternoon satellites. This study is an initial project within that larger effort.

In this study, only data from July and January are processed. These months were chosen to serve as a surrogate for the actual summer and winter seasons. Due to current limitations in the reprocessing infrastructure, it was not possible to process the entire AVHRR record for this study. While the goal of this study is an initial look at the global distribution of cirrus overlap from satellite data, future studies will include the goals of annual and monthly variability.

One of problems with using the AVHRR data for climate studies is the drift in the equator crossing time of the orbits (Ignatov et al. 2004). Figure 3 shows the variation in the local time that the afternoon AVHRRs crossed the equator in the ascending node. In general, the afternoon AVHRRs were launched in orbits that crossed the equator in the early afternoon and over time the orbits drifted such that equator time became later and later in the afternoon. Because the goal of this study was to make one of first global surveys of cirrus overlap from satellite data, it was decided to ignore the complexity of interpreting this continual change in equator crossing time by selecting periods when the equator crossing times were the same. The triangles in Fig. 3 mark the July and January months used in this effort. The July data used in this study are taken from 1982, 1986, 1991, and 1998 while the January data used in this study are taken from 1983, 1987, 1992, and 1999. These years include data from NOAA-7, -9, -11, and -14, respectively. Data from NOAA-16 were not used because of the potential discontinuity due to the 1.6-μm channel replacing the 3.75-μm channel during daytime operation. The equator crossing times were roughly 1430 and 1500 (LT) for the July and January months, respectively. Because NOAA-7 and NOAA-9 were launched into later orbits than NOAA-11 and NOAA-14, the data from NOAA-7 and NOAA-9 are taken earlier in the life of those satellites than the data from NOAA-11 and NOAA-14.

In summary, the AVHRR data processed here were chosen to be the minimum amount of data that would provide a meaningful look at the global distribution of cloud overlap and its variation between winter and summer. Future work will include more data (including the data from the morning satellites) and will focus on assessing annual, monthly, and diurnal variations.

4. Results

While CLAVR-x produces pixel-level cloud properties, the spatial resolution of all cirrus overlap data used here is 0.5°, which is the nominal resolution of the gridded cloud properties produced by NOAA. The fraction of all pixels that contain cirrus overlap is computed once per day using all of the daytime pixels that fall into a grid cell. The daily values are then averaged over each month and the monthly averages, for each year, are then averaged together to make the final multiyear monthly averages. The fraction of all (single-layer and cirrus overlap) ice clouds that overlap lower clouds and the fraction of all cloudy (the sum of totally cloudy plus probably cloudy) pixels that are cirrus overlap are also calculated in an analogous manner.

a. Global occurrence of cloud overlap

The first result shown here is the global distribution of the fraction of all pixels within each grid cell that contained cirrus overlap. Figure 4 shows the global distribution for all of the July months (1982, 1986, 1991, and 1999) used in this study. The South Polar region is not shown on this map since this region is without sunlight in July. Figure 4 shows that the global distribution of cirrus overlap is concentrated in a few regions, notably those with active convection. The most prominent region of cirrus overlap is Southeast Asia (0°–30°N, 60°–90°E), which corresponds with the Asian monsoon. While cirrus overlap is detected in significant numbers throughout the intertropical convergence zone (0°–10°N) in July, very frequent occurrences of cirrus overlap are detected in the ITCZ just west of Central America (120°–90°W). In these two regions, more than 35% of all pixels are determined to be cirrus overlap for the four July months studied. Another area of significant cirrus overlap in the Tropics occurs over tropical Africa north of the equator (0°–10°N) and is associated with intense summertime convection. Significant areas of cirrus overlap detected in the Tropics in July also include some regions in the Southern Hemisphere such as the Indian Ocean and the oceanic regions surrounding Indonesia. Outside of the Tropics, the Northern and Southern Hemisphere storm tracks are regions where significant cirrus overlap is detected. As seen in Fig. 4, the maximum amount of overlap detected in the storm tracks is generally less than the maximum amount observed in the Tropics.

The global distribution of cirrus overlap for all pixels for the four January months (1983, 1987, 1992, and 1999) are given in Fig. 5. The high-latitude regions in the Northern Hemisphere are not shown owing to the absence of sunlight in January. As was the case in the July results, the most striking features in the global cirrus overlap distribution in January are in the Tropics. Compared to the July distribution, the maximum over Southeast Asia has disappeared and has been replaced by a maximum over Indonesia (0°–10°S, 90°–150°E). The cirrus overlap seen north of the equator in tropical Africa in July has moved south of the equator in January (0°–20°S, 0°–30°E). In addition, Madagascar shows a large amount of cirrus overlap in January compared to none in July. The significant amount of cirrus overlap off the coast of Central America in July has disappeared in January but a significant amount of cirrus overlap is seen over the tropical South American continent (0°–20°S, 30°–90°W). These patterns of cirrus overlap in the Tropics are clearly driven by the annual cycle in the ITCZ and the solar heating–driven convection over land. It is notable that only tropical Australia lacks significant cirrus overlap compared to the other tropical landmass regions in the Southern Hemisphere. However, more cirrus overlap was detected in Australia as a whole in January than in July. As was the case in July, significant amounts of cirrus overlap are detected in the midlatitude storm tracks in January. Higher values of cirrus overlap are detected in the midlatitude storms tracks in the Southern Hemisphere in July than in January. In the Northern Hemisphere, more cirrus overlap is detected in the North Pacific than in the North Atlantic in July with the opposite pattern occurring in January. It should also be noted that some of the cirrus overlap shown over land in the Northern Hemisphere midlatitudes may be slightly overestimated due to the presence of snow cover.

The zonal means of the July and January distributions of the percentage of all pixels classified as cirrus overlap are shown in Fig. 6. Once again zonal means for the sunless high-latitude regions in winter are not shown. Evident in both curves are the three maxima associated with each hemisphere’s storm tracks and the ITCZ. The maximum in the Arctic in July is most likely false and is due to the overexaggerated classification of cirrus over sea ice as cirrus overlap. Tables 1 and 2 provide a quantitative evaluation of the qualitative results apparent in Figs. 4 –6. The values not in parentheses in Tables 1 and 2 are the fraction of all pixels determined to be cirrus overlap from 60°S to 60°N and separately for each 30° zonal band between 60°S and 60°N. The values inside the parentheses are the standard deviations expressed as a percentage of the mean value. The standard deviations are computed from the four individual monthly means used to compute the final mean value and are, therefore, an indication of the annual variation in the monthly means. Areas within 30° of each pole are excluded either due to a lack of sunlight or to avoid the potential uncertainties in the presence of snow/ice surfaces. Table 1 shows the July results, while Table 2 gives the January results. The values are given for all pixels and separately for pixels over land and water. The relative distribution of cirrus overlap presented in these tables is consistent with that evident in the figures. The movement in the distribution of cirrus overlap in the Tropics from July to January is clearly evident. Tables 1 and 2 demonstrate that the largest difference in the July and January cirrus overlap distribution occurs in the Tropics with the summer tropical zone having significantly more cirrus overlap than the winter tropical zone. For the midlatitude zones, the shift in cirrus overlap is opposite of that seen in the tropical zones with the winter midlatitude zone having higher values of cirrus overlap than the summer midlatitude zone. The magnitude of the July–January shift in cirrus overlap in the midlatitudes is much less than that in the tropical zones. The highest fraction of cirrus overlap over land occurs over tropical landmasses in the summer hemisphere with values of 15% in July and 21% in January. The relative difference in the amount of cirrus overlap in the tropical zones between land and ocean is less in July than in January. In the southern midlatitude zone, more cirrus overlap is observed over ocean than land in both months. No differences between the cirrus overlap over land and ocean is observed in the northern midlatitudes zones.

Analysis of the standard deviations in Tables 1 and 2 shows that they are generally smallest (as a percentage of the mean) in the summer tropical zones and in midlatitude zones in July. For the entire extrapolar globe, the standard deviation is higher in January than in July. Only one zone (July land from 30°S to 0°) has a standard deviation that exceeds 20% of the mean value. These small standard deviations indicate that the results shown here are consistent over the 4 yr studied and, therefore, that similar results would be obtained from other years. Analysis of the effects of the El Niño–Southern Oscillation (ENSO) cycles and other annual and seasonal variations will be studied later.

1) Comparison to estimates of multilayer cloud from manual surface observations

Over land, surface observations of cloud cover have been compiled by manual observations at many sites. Hahn and Warren (1999, 2002) have analyzed records from surface observations of cloud cover into an atlas of the monthly occurrence of several cloud types. Two of the cloud types compiled by this study were the occurrence of high cloud and the occurrence of high cloud when no other cloud layers were present. The difference between these cloud amounts should be the amount of high cloud that is multilayer. The multilayer high cloud amount derived in this way should be equivalent to the percentage of all pixels typed as overlapped cirrus shown in Figs. 4 –6. Note that the results of Hahn and Warren (2002) also provide information on the occurrence of multilayer midlevel cloud. Our analysis of the AVHRR algorithm indicates that overlap between midlevel and low level is often not detectable and we, therefore, did not include the midlevel cloud overlap in these comparisons.

Figure 7 shows a zonal comparison of the overlapped high cloud from Hahn and Warren (2002) and this study for the July data. The Hahn and Warren results are based on July data from 1971 to 1996 while the data from this study are from the 4 yr listed previously. The results for January are given in Fig. 8. The zonal profiles from the AVHRR and surface observations show a similar pattern and magnitude for both months. To help compare these results, some statistics were written in the legend of each figure. The first number written in the legend entry for each curve is the mean value from 60°S to 60°N and the second number is the mean value from 30°S to 30°N. Figure 7 shows that for most of the Tropics in July the CLAVR-x zonal mean of cirrus overlap is slightly less than that from Hahn and Warren but the CLAVR-x zonal mean value of cirrus overlap in the midlatitudes exceeds that from Hahn in Warren. In terms of the 30°S–30°N mean value, CLAVR-x is less than that of Hahn and Warren (11% to 13%) but for the 60°S–60°N mean value, CLAVR-x exceeds Hahn and Warren (9.7% to 8.6%). While the manual observer’s ability to detect multilayer cloud is most likely superior to that from an automated algorithm applied to satellite data, if the lower-level cloud is overcast, the detection of the upper-layer cloud by manual observation may be difficult. For this reason, values of cirrus overlap from CLAVR-x that exceed those from manual observation are possible and may explain some of the differences seen in Figs. 7 and 8. It is also important to note that the validation of this technique against surface-based radar (PH04) indicated that the overestimation of the technique was small. Therefore, even though the results below indicate regions where the CLAVR-x estimates of cirrus overlap exceed those from Hahn and Warren, we do not feel that this necessarily implies that the CLAVR-x technique is overestimating cirrus overlap.

The comparison of the zonal distribution of cirrus overlap over land from CLAVR-x and Hahn and Warren shown in Fig. 8 shows a similar level of agreement as seen in Fig. 7. Unlike the July comparison, the CLAVR-x estimates of cirrus overlap exceed those from Hahn and Warren for most of the zones except between 0° and 25°N. The mean values from CLAVR-x and Hahn and Warren from 30°S to 30°N are identical to two significant digits (15%). As was the case in the July comparison, the CLAVR-x estimate of mean value from 60°S to 60°N slightly exceeds that from Hahn and Warren (11% to 9.4%). As pointed out earlier, there is concern that the AVHRR algorithm will overestimate the amounts of cirrus overlap in the presence of snow/ice surfaces. If the agreement between the CLAVR-x and the Hahn and Warren zonal profiles in cirrus overlap is to be believed in the Northern Hemisphere in Fig. 8, the overdetection of cirrus overlap in the presence of snow/ice surfaces is not a major issue with the algorithm for these extrapolar regions. In summary, these comparisons indicate that the AVHRR algorithm for cirrus overlap is able to produce zonal profiles of cirrus overlap over land that are in rough agreement with those reported by manual observers. This agreement extends to the shape of zonal profiles, their magnitude, and their variation from July to January.

b. Percentage of all clouds determined to be cirrus overlap.

While the previous results showed the percentage of all pixels determined to be cirrus overlap, Tables 3 and 4 show the percentage of all cloudy pixels determined to be cirrus overlap. In these results, the clear and probably clear pixels from the CLAVR-x cloud mask were excluded from the analysis; therefore the percentages are generally higher in Tables 3 and 4 than in Tables 1 and 2.

Tables 3 and 4 show many similarities with Tables 1 and 2. For example, Tables 3 and 4 demonstrate that the lowest percentage of cloudy pixels being classified as cirrus overlap occurs in the winter tropical zones. As was the case with Tables 1 and 2, the differences between July and January are much less in the midlatitude zones than the Tropics with each winter midlatitude zone having a slightly higher percentage of clouds being cirrus overlap than the summer midlatitude zone. In general, roughly 20% of all cloudy pixels are classified as cirrus overlap. Except for the winter tropical zone in January, the percentage of cloudy pixels with cirrus overlap is higher over land than over water; however, the land–ocean difference is not significant in most zones. The zones with the largest percentage of clouds detected to be cirrus overlap are the winter midlatitude land zones with values of 27% in July and 29% in January. The zones with the lowest percentages of clouds being detected as cirrus overlap occur in winter tropical zones with values as low as 10% being reported for the 30°S–0° zone in July. As was the case with Tables 1 and 2, the standard deviations expressed as a percentage of the mean values in Tables 3 and 4 indicate the results from these 4 months are probably not sensitive to the months chosen.

1) Comparison to estimates of multilayer cloud from raobs

The results in Tables 3 and 4 can be compared to results from an extensive analysis of rawinsonde (raobs) data described by Wang et al. (2000). In this study, a quality controlled global set of raobs was analyzed to derive the presence of cloud layers based on the separation of the temperature and de wpoint temperature profiles. One product of this analysis was the global distribution of the occurrence of multiple layers of cloud. Table 5 summarizes the global occurrence of multilayer cloud from Wang et al. (2000), based on the full 20 yr of data. Their values are reported in the raobs column of Table 5 and are given as the percentage of all cloudy profiles that were determined to be multilayer. The raob results are summarized here to show the percentage of cloudy profiles that had two layers of clouds and the percentage that had at least two cloud layers. Because of the known limitation of the cirrus overlap detection, we assume that most cirrus overlap cases detected here are high semitransparent cirrus over lower clouds with significant vertical separation.

As pointed out in Wang et al. (2000), there are limitations to the raob analysis. For example, their analysis indicates that 20%–30% of the highest level clouds are missed by the raobs due to loss of sensitivity to humidity at cold temperatures. However, Wang and Rossow (1998) showed that the missed cloud tended to be very thin layers, which are also likely missed by the technique used here. Based on the global results for two-layer clouds, the mean pressure of the lower layer was 850 hPa and the mean pressure of the upper layer was 490 hPa. For the standard tropical atmosphere, this pressure difference corresponds to a roughly 30-K difference in cloud-top temperatures between the two layers. From the sensitivity study of PH04, a temperature difference of roughly 35 K is needed for detection of cirrus overlap. However, Wang et al. (2000) also indicate that the distribution of heights of the topmost cloud layer is broad and ranges up to 12 km with a small maximum near 7.5 km. Using the same standard tropical atmosphere and assuming the same mean cloud-top temperature for the lower cloud, top-layer clouds higher than 7.5 km should present a temperature difference greater than 40 K, which should allow for detection of cirrus overlap from the method described in PH04. However, it is clear that many of the multilayer cloud systems detected by Wang et al. (2000) are comprise water clouds over water clouds with insufficient vertical separation to allow for detection using the PH04 algorithm. Given all of the above limitations, we do not feel a quantitave comparison between these results has any meaning. Our goal here is to show the qualitative agreement in the rough magnitude and geographical distribution of multilayer cloud with the unique study of Wang et al. (2000).

As Table 5 shows, the raob two-layer results show little land–ocean difference, which is consistent with the values from this study. The relevant AVHRR results reported in Tables 3 and 4 are repeated in Table 5. In terms of magnitude, the AVHRR values are roughly two-thirds of the raob two-layer values. One possible explanation for these differences is that raob analysis includes cloud scenarios where the AVHRR algorithm for cirrus overlap detection fails, such as for multilayer conditions with very optically thin cirrus, optically thick cirrus, or optically thin low cloud. While the agreement between the raobs and CLAVR-x is less than that seen with the analysis of manual cloud observations shown earlier in Figs. 7 and 8, these results do indicate that the AVHRR approach is able to detect at least the majority of the two-layer multilayer cloud situations. If the raob multilayer results are correct, the AVHRR is not able to detect the majority of all cloud overlap conditions. As described above, this finding is consistent with our belief that the algorithm is able to detect only a subset of the multilayer cloud systems that could be identified from the raob analysis, possibly due to insufficient vertical separation between cloud layers.

c. Percentage of ice clouds determined to be cloud overlap

As stated before, the cirrus overlap detection functions within the larger CLAVR-x cloud-type algorithm. The CLAVR-x cloud-type algorithm classifies each cloudy pixel as being either fog, liquid, supercooled liquid, opaque ice, nonoverlapped cirrus, or cirrus overlap. In addition to computing the cirrus overlap results, the distributions of all the CLAVR-x cloud types were computed. Tables 6 and 7 show the percentage of all ice cloud pixels that were determined to be cirrus overlap for July and January. The ice cloud pixels include those classified as opaque ice, nonoverlapped cirrus, and cirrus overlap.

Tables 6 and 7 show the percentage of ice cloud pixels determined to be cirrus overlap for the July and January data. In comparison to the previous tables, one obvious feature of Tables 6 and 7 is the reduction in variation with the range of values being roughly 30%–60%. Also, roughly 40% of all ice clouds from 60°S to 60°N are determined to be cirrus overlap. In general, the midlatitude zones show a higher percentage of ice cloud to be cirrus overlap than the tropical zones. All midlatitude zones have values exceeding 40% while only one tropic zone (January 30°S–0° land) has a value exceeding 40%. There is little land–ocean difference in the percentage of ice cloud typed as cirrus overlap except for the 30°S–0° zone where the values over land are several percentage units higher than those over ocean.

1) Comparison to estimates of multilayer cloud from infrared and microwave observations

As mentioned in the introduction, other techniques have been developed to detect the presence of multilayer cloud from satellites. A recent study (Ho et al. 2003) applied a multilayer cloud detection technique to microwave and thermal emission observations from the Tropical Rainfall Measuring Mission (TRMM). The technique is described in Lin et al. (1998) and the physical principal applied is that the microwave observations are sensitive to the water cloud and the infrared observations are more sensitive to the ice cloud. When the infrared-derived and microwave-derived cloud temperature estimates differ by a predefined value, a multilayer cloud situation is assumed to be present. Ho et al. (2003, hereafter HO-03) applied this technique to TRMM data from January to August 1998. Due to the orbit of TRMM, HO-03 analyzed the distribution of multilayer cloud from roughly 35°S to 35°N.

In HO-03, spatial patterns of multilayer cloud similar to Figs. 4 and 5 of this study are shown (see Figs. 9 and 10 of HO-03). The general distribution and seasonal shift of the multilayer clouds are comparable. However, the AVHRR results show more structure and variation. Whereas the AVHRR technique does not distinguish between high or midlevel multilayer clouds, the results of HO-03 do. However, only the multilayer high cloud results from HO-03 will be discussed here. Figure 9 shows the zonal profile of the percentage of all ice clouds that were also typed as cirrus overlap. Figure 9 of this study can be compared to Fig. 11 of HO-03. While the results of HO-03 exclude land regions, the results in Fig. 9 include all pixels. It was found that the inclusion of the land pixels did not substantially alter the zonal profiles. In this analysis, it is assumed that the high cloud statistics from HO-03 are physically consistent with the ice cloud statistics presented here. The results of HO-03 indicate that roughly 25%–45% of the ice clouds are actually cirrus overlap between 35°S and 35°N, in rough agreement with Fig. 9. To more quantitatively compare these results, the values in Fig. 9 were recomputed by excluding the land pixels and averaging over the 10° bins used to derive the results of HO-03. Based on a comparison of these numbers and the figures in HO-03 (not shown), it was found that the results in Fig. 9 are slightly greater on average than those in HO-03 and that the mean difference between the results of these two studies is roughly 5%. The maximum difference between the two was approximately 15% and the minimum difference was less than 2%. Figure 9 of this study shares some qualitative features with HO-03. For example, both show a relative maximum of roughly 45% at 5°N in the January data. However, the HO-03 plots shows no peak in the Tropics in July while Fig. 9 shows a July peak similar to that in January. In summary, both studies are in rough agreement on the zonal distribution and the percentage of high clouds that are overlapped by cirrus.

5. Conclusions

This study applied a recently published and validated cirrus overlap detection algorithm to AVHRR data to produce one of the first global surveys of the cirrus overlap distribution using satellite data. The data used in this study were taken from July and January for 4 yr, which were chosen from the entire record to provide a dataset with the same local time (1430–1500). The cirrus overlap algorithm was run within the CLAVR-x software used by NOAA to produce the global distributions of the occurrence of cirrus overlap.

The results indicated that roughly 12% of all pixels between 60°N and 60°S can be classified as cirrus overlap. For the same region, roughly 18%–20% of all cloudy pixels are determined to be cirrus overlap. When considering only ice cloud pixels, the probability of cirrus overlap occurring was increased to roughly 40%. While the probability of occurrence of cirrus overlap within 60°N and 60°S varied little from January to July, the zonal and regional distributions of cirrus overlap varied significantly. The shifts in the patterns of cirrus overlap appear to be well correlated with the shift in the patterns of convection. In general, more cirrus overlap is detected over land than ocean. The zonal profiles of cirrus overlap over land were found to be in rough agreement with those based on many years of manual surface observations (Hahn and Warren 2002).

The cirrus overlap results over land from this study also agreed well with non-satellite-derived results from Wang et al. (2000), where rawinsondes were used to derive the vertical distribution of cloudiness. Over ocean, the rawinsonde-derived estimates of two-layer clouds are roughly 10% units larger than the numbers produced in this study. Owing to the sparse distribution of rawinsondes over the ocean, it is unclear if this discrepancy is due to sampling error or a limitation in the AVHRR algorithm. In any event, if the results of Wang et al. (2000) are taken as a true estimate of the occurrence of two-layer cloud, the AVHRR algorithm appears to detect the majority and spatial distribution of the two-layer situations. Finally, the percentage of ice clouds determined to be cirrus overlap was found to be in rough agreement with the study of Ho et al. (2003), which also found that approximately 40% of all ice clouds in the Tropics are actually cirrus overlap.

This analysis will continue and expand as part of the AVHRR cloud climatology being generated within NESDIS/ORA. Eventually all AVHRR data will be reprocessed allowing for a full analysis of the cirrus overlap climatology. In addition, the prevalence of cirrus overlap in many regions warrants the inclusion of cirrus overlap knowledge in the estimation of other cloud properties. Incorporation of the effects of cirrus overlap is a future goal for the real-time AVHRR cloud processing done at NOAA.

Acknowledgments

This work was funded by the NOAA/NESDIS Polar Program and the NOAA/NESDIS Office of Research and Applications Data Stewardship Initiative. The views, opinions, and findings contained in this report are those of the authors and should not be construed as an official National Oceanic and Atmospheric Administration or U.S. government position, policy, or decision.

REFERENCES

  • Baum, B. A., , and J. D. Spinhirne, 2000: Remote sensing of cloud properties using MODIS airborne simulator imagery during SUCCESS. III: Cloud overlap. J. Geophys. Res., 105 , 1179311804.

    • Search Google Scholar
    • Export Citation
  • Baum, B. A., and Coauthors, 1995: Satellite remote sensing of multiple cloud layers. J. Atmos. Sci., 52 , 42104230.

  • Chen, T., , Y. Zhang, , and W. B. Rossow, 2000: Sensitivity of atmospheric radiative heating rate profiles to variations of cloud layer overlap. J. Climate, 13 , 29412959.

    • Search Google Scholar
    • Export Citation
  • Gupta, S. K., , W. L. Darnell, , and A. C. Wilber, 1992: A parameterization for longwave surface radiation from satellite data: Recent improvements. J. Appl. Meteor., 31 , 13611367.

    • Search Google Scholar
    • Export Citation
  • Hahn, C. J., , and S. G. Warren, 1999: Extended edited cloud reports from ships and land stations over the globe, 1952–1996. Numerical Data Package NDP-026C, Carbon Dioxide Information Analysis Center, Oak Ridge, TN, 79 pp.

  • Hahn, C. J., , and S. G. Warren, 2002: Cloud climatology for land stations worldwide, 1971–1996. Numerical Data Package NDP-026D, Carbon Dioxide Information Analysis Center, Oak Ridge, TN, 35 pp.

  • Heidinger, A. K., , R. Frey, , and M. J. Pavolonis, 2004: Relative merits of the 1.6 and 3.75 micron channels of the AVHRR/3 for cloud detection. Can. J. Remote Sens., 30 , 182194.

    • Search Google Scholar
    • Export Citation
  • Ho, S., , B. Lin, , P. Minnis, , and T. Fan, 2003: Estimates of cloud vertical structure and water amount over tropical oceans using VIRS and TMI data. J. Geophys. Res., 108 .4419, doi:10.1029/2002JD003298.

    • Search Google Scholar
    • Export Citation
  • Ignatov, A., , I. Laszlo, , E. Harrod, , K. Kidwell, , and G. Goodrum, 2004: Equator crossing times for NOAA, ERS and EOS sun-synchronous satellites. Int. J. Remote Sens., 25 , 52555266.

    • Search Google Scholar
    • Export Citation
  • Inoue, T., 1985: On the temperature and effective emissivity determination of semi-transparent cirrus clouds by bi-spectral measurements in the 10 μm window region. J. Meteor. Soc. Japan, 64 , 8899.

    • Search Google Scholar
    • Export Citation
  • Jin, Y., , and W. B. Rossow, 1997: Detection of cirrus overlapping low-level clouds. J. Geophys. Res., 102 , 17271737.

  • Lin, B., , P. Minnis, , B. Wielicki, , D. R. Doelling, , R. Palikonda, , D. F. Young, , and T. Uttal, 1998: Estimation of water cloud properties from satellite microwave, infrared and visible measurements in oceanic environment: 2. Results. J. Geophys. Res., 103 , 38873905.

    • Search Google Scholar
    • Export Citation
  • Morcrette, J. J., , and C. Jakob, 2000: The response of the ECMWF model to changes in the cloud overlap assumption. Mon. Wea. Rev., 128 , 17011732.

    • Search Google Scholar
    • Export Citation
  • Nasiri, S. L., , and B. A. Baum, 2004: Daytime multilayered cloud detection using multispectral imager data. J. Atmos. Oceanic Technol., 21 , 11451155.

    • Search Google Scholar
    • Export Citation
  • Pavolonis, M. J., , and J. R. Key, 2003: Antarctic cloud radiative forcing at the surface estimated from the AVHRR polar pathfinder and ISCCP D1 datasets, 1985–93. J. Appl. Meteor., 42 , 827840.

    • Search Google Scholar
    • Export Citation
  • Pavolonis, M. J., , and A. K. Heidinger, 2004: Daytime cloud overlap detection from AVHRR and VIIRS. J. Appl. Meteor., 43 , 762778.

  • Pavolonis, M. J., , A. K. Heidinger, , and T. Uttal, 2005: Daytime global cloud typing from AVHRR and VIIRS: Algorithm description, validation, and comparisons. J. Appl. Meteor., 44 , 804826.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2002: The CloudSat mission and the A-Train. Bull. Amer. Meteor. Soc., 83 , 17711790.

  • Wang, J., , and W. B. Rossow, 1998: Effects of cloud vertical structure on atmospheric circulation in the GISS GCM. J. Climate, 11 , 30103029.

    • Search Google Scholar
    • Export Citation
  • Wang, J., , W. B. Rossow, , and Y. Zhang, 2000: Cloud vertical structure and its variations from a 20-yr global rawinsonde dataset. J. Climate, 13 , 30413056.

    • Search Google Scholar
    • Export Citation
  • Warren, S. G., , C. J. Hahn, , and J. London, 1985: Simultaneous occurrence of different cloud types. J. Climate Appl. Meteor., 24 , 658667.

    • Search Google Scholar
    • Export Citation
  • Wielicki, B. A., , R. D. Cess, , M. D. King, , D. A. Randall, , and E. F. Harrison, 1995: Mission to Planet Earth: Role of clouds and radiation in climate. Bull. Amer. Meteor. Soc., 76 , 21252153.

    • Search Google Scholar
    • Export Citation
Fig. 1.
Fig. 1.

Broadband infrared heating rate profiles computed for a standard tropical atmosphere. The solid line shows the profile for an atmosphere with a cirrus cloud overlying a low-level water cloud. The dotted line shows the profile computed for an atmosphere with a single cirrus cloud. The total column optical depths and radiative cloud-top temperatures are the same in both atmospheres.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 2.
Fig. 2.

Pixel-level detection of cirrus overlap using the AVHRR technique applied to MODIS data at 1915 UTC 4 Apr 2003. The data are from the tropical eastern Pacific. (top) RGB image using bands 1 (0.65 μm), 6 (1.6 μm), and 31 (11 μm). (bottom) Pixels that passed the AVHRR cirrus cloud overlap test as a gray mask.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 3.
Fig. 3.

Variation of the equator crossing time (local) for the NOAA polar orbiting satellites. The triangles represent the July and January months chosen to have approximately the same equator crossing time.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 4.
Fig. 4.

Mean percentage of AVHRR pixels being classified as cirrus overlap for the 4 yr (1982, 1986, 1991, and 1998) of July data. Values south of 60°S are not shown due to the lack of daytime results. The spatial resolution is 0.5°.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 5.
Fig. 5.

Mean percentage of AVHRR pixels being classified as cirrus overlap for the 4 yr (1983, 1987, 1992, and 1999) of January data. Values north of 60°N are not shown due to the lack of daytime results. The spatial resolution is 0.5°.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 6.
Fig. 6.

Zonal mean distribution of the percentage of AVHRR pixels typed as overlapped cirrus. Results represent the mean based on the four years of data and have a spatial resolution of 0.5°. Values for regions where daytime results where not available are missing.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 7.
Fig. 7.

July zonal distribution of the mean value of the percentage of AVHRR pixels over land classified as cirrus overlap. CLAVR-x results represent mean values for the four years of data processed. Hahn and Warren results represent mean values from 1971 to 1996. Spatial resolution of the CLAVR-x results in 0.5°. Spatial resolution of the Hahn and Warren results is 5°. First number in parentheses is the mean value from 60°S to 60°N. Second number in parentheses is the mean value from 20°S to 20°N.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 8.
Fig. 8.

January zonal distribution of the mean value of the percentage of AVHRR pixels over land classified as cirrus overlap. CLAVR-x results represent mean values for the 4 yr of data processed. Hahn and Warren results represent mean values from 1971 to 1996. Spatial resolution of the CLAVR-x results in 0.5°. Spatial resolution of the Hahn and Warren results is 5°. First number in parentheses is the mean value from 60°S to 60°N. Second number in parentheses is the mean value from 20°S to 20°N.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Fig. 9.
Fig. 9.

July and January zonal distributions of the mean value of the percentage of AVHRR pixels classified as ice clouds that are also typed as cirrus overlap. CLAVR-x results represent mean values for the four years of data processed. Spatial resolution of results in 0.5°. Values for regions where daytime results where not available are missing.

Citation: Journal of Climate 18, 22; 10.1175/JCLI3535.1

Table 1.

Percentage of all pixels with cirrus overlap from the July data. Values in parentheses are the standard deviations (expressed as a percentage of the mean) based on the individual monthly mean values for each year (1982, 1986, 1991, and 1998).

Table 1.
Table 2.

As in Table 1 except for January data (1983, 1987, 1992, and 1999).

Table 2.
Table 3.

Percentage of all cloudy pixels with cirrus overlap from the Jul data. The values in parentheses are the standard deviations (expressed as a percentage of the mean) based on the individual monthly mean values for each year (1982, 1986, 1991, and 1998).

Table 3.
Table 4.

As in Table 3, except for the January data (1983, 1987, 1992, and 1999).

Table 4.
Table 5.

Comparison of the percentage of all clouds that are cirrus overlap to the multilayer cloud percentages from the raobs based on the analysis of Wang et al. (2000).

Table 5.
Table 6.

Percentage of all ice cloud pixels with cirrus overlap from the July data. The values in parentheses are the standard deviations (expressed as a percentage of the mean) based on the individual monthly mean values for each year (1982, 1986, 1991, and 1998).

Table 6.
Table 7.

As in Table 6, except for the January data (1983, 1987, 1992, and 1999).

Table 7.
Save