1. Introduction
Only satellite observations offer a continuous survey of the state of the atmosphere over the whole globe. Most current satellite instruments are radiometers, measuring reflected, scattered, and emitted radiation from the earth’s surface, atmosphere, and clouds. To convert the measured radiances into cloud properties, clouds have to be first distinguished from clear-sky situations and then their properties have to be determined using radiative transfer models.
Long time series (more than twenty years) of these measurements are available from imagers, using infrared (IR) and visible (VIS) atmospheric window channels as well as from the Television Infrared Observation Satellite-N (TIROS-N) Observational Vertical Sounders (TOVS), using CO2 and H2O sensitive channels. Radiances measured from near the center of a CO2 absorption band are only sensitive to the upper atmosphere while radiances from the wings of the band (away from the band center) see successively lower levels of the atmosphere. Microwave radiation measured around the O2 absorption band passes through aerosols and most clouds, since these wavelengths are much greater than aerosol and cloud particles.
The International Satellite Cloud Climatology Project (ISCCP; Rossow and Schiffer 1999) offers the cloud climatology with the best diurnal sampling (3 hourly) and spatial resolution (7 km sampled every 30 km), using IR atmospheric window and VIS (day only) radiance measurements from imagers on the suite of geostationary and polar orbiting weather satellites. Cloud-top temperature, Tcld, is first retrieved assuming that all clouds are black bodies. During daytime, when VIS radiances are available to retrieve cloud VIS optical thickness, τcld, Tcld of transmissive clouds (τcld < 5.5) is corrected for the radiation transmitted from below, decreasing Tcld (or increasing cloud height) as a function of τ. During night time, however, semitransparent cirrus (high ice clouds) may falsely be identified as midlevel clouds (by not being able to use VIS radiances to correct the cloud height).
In contrast, cirrus properties obtained from TOVS, with its relatively high spectral resolution in the IR, are especially reliable day and night (Wylie et al. 1994; Stubenrauch et al. 1999a; Wylie and Menzel 1999). For climate studies, it is important to understand how cloud properties are perceived by these different instruments and methods. Detailed comparisons have shown that these datasets agree quite well (Jin et al. 1996; Stubenrauch et al. 1999c), with a better sensitivity of TOVS to cirrus clouds. Discrepancies can be explained by differences in temperature profiles, horizontal heterogeneities (partial cloud cover) and vertical heterogeneities (multilayer clouds). For example, in the case of thin cirrus overlying low clouds, using TOVS provides the properties of the cirrus, whereas the use of a VIS channel by ISCCP leads to inference of a midlevel cloud.
The TOVS Path-B dataset (Scott et al. 1999) provides global atmospheric temperature and water vapor profiles as well as cloud and surface properties at a spatial resolution of 1° latitude × 1° longitude. At present, the dataset covers the time period from July 1987 to June 1995. The data and cloud property retrieval are described in section 2.
A study of the time evolution of cloud properties from the original TOVS Path-B dataset has revealed an artificial increase of low clouds in the Tropics after the Mt. Pinatubo eruption in June 1991 (Stubenrauch and Eddounia 2001). The revision of the cloud detection is discussed in section 3.
Cloud height of the revised TOVS Path-B dataset has been evaluated (Stubenrauch et al. 2005) by using vertical profiles of backscattered radiation from quasi-simultaneous lidar in space technology experiment (LITE). The cloud height determined by TOVS in general corresponds well to the height of the apparent middle of the cloud system at 1° spatial resolution. Pressure distributions of the highest cloud layer weighted by effective cloud amount (ECA; see section 2) confirms that the average height of high clouds (with cloud pressure smaller than 440 hPa) is largest in the Tropics, due to a higher tropopause, and that in these regions there are nearly no cloud systems with the highest cloud layer in the middle troposphere. The Southern Hemisphere (SH) midlatitudes are mostly covered by low-level clouds. Regional and seasonal distributions of bulk microphysical properties of large-scale semitransparent cirrus obtained from the revised TOVS Path-B data have been published by Rädel et al. (2003) and Stubenrauch et al. (2004).
This article concentrates on the physical properties of clouds in general and on their seasonal and diurnal variability. Section 4 presents average retrieved cloud properties compared to ISCCP. Seasonal and diurnal cycles of high, midlevel, and low clouds are investigated in sections 5 and 6. As an example of the utility of this dataset, the seasonal and diurnal cycles of high clouds associated with deep convection over the American and African continents and over Indonesia are compared in section 7. Section 8 summarizes the features of this dataset.
2. Data and cloud property retrieval
Since 1979, the TOVS instruments aboard the National Oceanic and Atmospheric Administration (NOAA) Polar Orbiting Environmental Satellites have measured radiation emitted and scattered from different levels of the atmosphere. The TOVS system consists, in particular, of two sounders: the High-Resolution Infrared Radiation Sounder (HIRS/2) with 19 IR spectral channels between 3.7 and 15 μm and one VIS channel (0.7 μm) and the Microwave Sounding Unit (MSU) with four microwave channels around 5 mm. The IR channels have been chosen around absorption bands of CO2, centered at 4.3 and 15 μm, a water vapor absorption band centered at 6.7 μm, and an ozone absorption band around 9.3 μm. Observations are made from morning and afternoon NOAA satellites, at about 0730 and 1930 local time (LT) as well as 1400 and 0200 LT, respectively. A problem in establishing a long time series of cloud properties from satellite observations can be the drift of the satellite observation time (Wylie et al. 2005). The afternoon satellites NOAA-11 and NOAA-14 have drifted by 4 and 3.5 h in 6 yr, respectively. Cloud properties with a strong diurnal cycle can therefore be affected. To avoid this effect, TOVS Path-B cloud properties shown in section 4 have been averaged only from NOAA-10 and NOAA-12 observations at 0730 and 1930 LT. On the other hand, for the analysis of the diurnal cycle of cloud properties in section 6 we have taken advantage of the NOAA-11 satellite drifting by using two years of data near the beginning (1989 and 1990) and near the end of the satellite period (1993 and 1994). This leads in total to six observation times per day: 0200, 0430, 730, 1400, 1630, and 1930 LT.
The Improved Initialization Inversion (3I) algorithm (Chédin et al. 1985; Scott et al. 1999) converts the measured radiances into physical properties of the atmosphere and surface. It is based on a fast line-by-line radiative transfer model (4A; Scott and Chédin 1981) and a dataset for the initial guess of the atmospheric temperature profile retrieval (Chédin et al. 1985; Chevallier et al. 1998). This Thermodynamic Initial Guess Retrieval (TIGR) dataset has been generated from a huge collection of radiosonde measurements of temperature, humidity and pressure. It consists of brightness temperatures of the HIRS and MSU channels which have been simulated by the 4A radiative transfer model from clear-sky radiosonde measurements of about 2000 atmospheric temperature and humidity profiles, classified into five air masses. Systematic biases between observed and simulated brightness temperatures (TB) due to the radiative transfer model, to instrument calibration (satellite-to-satellite consistency) or to unexpected events (such as the Mt. Pinatubo eruption) are removed by applying corrections to the measured HIRS brightness temperatures (Scott et al. 1999). This method also takes care of changes in the CO2 concentration over time (Chédin et al. 2003). These bias adjustment corrections were obtained from a collocated radiosonde-satellite dataset (DSD5 dataset; Uddstrom and McMillin 1993), provided by the National Satellite, Data, and Information Service (NESDIS) of NOAA for the period between 1987 and 1995. As an example, Fig. 1 shows the bias adjustment corrections of HIRS channel 8 (at 11 μm) as a function of time, separately for ocean and land, at night, for two airmass types (tropical and midlatitude). A sudden increase of these bias adjustment corrections of about 0.5 to 1 K after the eruption of Mt. Pinatubo in June 1991 is observed, which then decrease slowly until 1994. The difference between values after 1994 and values before 1991 is due to satellite-to-satellite inconsistency.
After multispectral cloud detection at HIRS spatial resolution (17 km at nadir) described in section 3 is performed, the HIRS radiances are averaged separately for clear and cloudy pixels within 100 km × 100 km regions. Cloud properties are determined from the averaged cloudy pixel radiances assuming that all cloudy pixels are covered by a single homogeneous cloud layer. The average cloud pressure pcld and the average effective cloud emissivity εcld over cloudy pixels are obtained from four radiances in the 15-μm CO2-absorption band (with peak responses from 400- to 900-hPa levels in the atmosphere) and one in the 11-μm IR atmospheric window by minimizing a weighted χ2 (Stubenrauch et al. 1999b). Empirical weights reflect the effect of the brightness temperature uncertainty within a TIGR airmass class on these radiances at the various cloud levels. Effective cloud amount (ECA in %) corresponds to εcld weighted by the fraction of cloudy HIRS pixels (CA in %) over a 1° latitude × 1° longitude grid.
We have taken the ISCCP definition to distinguish high clouds (pcld < 440 hPa), midlevel clouds (440 hPa < pcld < 680 hPa), and low clouds (pcld > 680 hPa). High clouds can be further separated into high opaque clouds (cloud type thick Ci, with εcld > 0.95, corresponding approximately to a VIS optical thickness τcld > 6) and cirrus (including both cloud types: Ci, with 0.95 > εcldcld > 0.5 or 6 > τcld > 1.4, and thin Ci, with εcld < 0.5 or τcld < 1.4). Midlevel and low clouds are only separated into two thickness classes, respectively, corresponding to Ast and St with ECA > 50%, and Acu and Cu with ECA < 50%. This cloud-type definition is slightly different from the ISCCP definition of nine cloud types according to pcld and τcld, with limits at τcld = 23 and τcld = 3.6 for all three pcld intervals. Because of the much better spatial resolution and the higher threshold ISCCP certainly selects better high convective clouds (Cb with τcld > 23) than TOVS.
3. TOVS Path-B multispectral cloud detection and its revision
Reliable cloud detection is very important, not only to determine cloud properties but also to identify clear-sky regions for the determination of atmospheric and surface properties. ISCCP associates clear-sky conditions with low IR and VIS spatial and temporal variability, and then clouds are detected through a variable IR–VIS threshold test, comparing the measured radiances to statistically inferred clear-sky composite radiances (Rossow and Garder 1993). Many other approaches have been published (e.g., Baum et al. 1997; Smith and Taylor 2004). Cloud detection thresholds vary with surface emissivity, atmospheric moisture, and viewing angle, and it is therefore difficult to determine globally applicable thresholds. The TOVS multispectral cloud detection is based on the following:
Interchannel regression tests: In the case of clear sky, the brightness temperature of one channel (most efficiently an MSU channel, because microwaves probe through nonprecipitating clouds) can be simulated by a weighted sum of brightness temperatures from HIRS channels that sound different depths of the atmosphere. Regression coefficients for different air masses have been obtained by using the TIGR dataset. The presence of clouds leads to a difference between the measured brightness temperature and the reconstructed brightness temperature.
Atmospheric window temperature differences: Clouds absorb less and reflect more radiation at shorter wavelengths. Surface temperature estimates at 3.7, 4, and 11 μm are each calculated by a weighted sum of brightness temperatures from four to five HIRS channels, to remove contributions of atmospheric water vapor and the effect of surface emissivity. Regression coefficients for different air masses have again been obtained by using the TIGR dataset. Using surface temperature estimates instead of brightness temperatures in the cloud detection has the advantage that thresholds do not change regionally or seasonally.
Spatial heterogeneity: In the case of clear sky, adjacent HIRS spots within 100 km × 100 km boxes should have similar surface temperature estimates, and the maximum surface temperature should not be too large compared to all other spots.
MSU surface emissivity: Sea ice and snow have a larger MSU surface emissivity than clouds.
A study of the effect of the Mt. Pinatubo eruption in June 1991 on the properties of high clouds (Luo et al. 2002) has revealed that over the whole Tropics TOVS Path-B effective high cloud amount was only little affected by this event. However, volcanic aerosols with an optical thickness exceeding 0.1 (McCormick et al. 1995) slightly increased VIS reflectances. Since the ISCCP cloud property retrieval uses the VIS channel to correct IR Tcld of transmissive clouds for radiation transmitted from below, it leads to an underestimation of ISCCP high cloud amount of 4.5% in the Tropics, but only of 2.5% over the globe.
ISCCP total cloud amount was not affected, because ISCCP uses dynamic cloud detection thresholds. Figures 2 show that the TOVS Path-B average clear sky fraction (100% − CA) decreased after the Mt Pinatubo eruption by about 15% in the Tropics (20°N–20°S) and by about 5% over the globe. On the other hand, TOVS low cloud amount increased (not shown). Since the ISCCP total cloud amount stayed stable after this event (Rossow and Schiffer 1999), one may assume that the stratospheric volcanic aerosols were falsely detected as low clouds by the TOVS instruments. Therefore, a systematic analysis and improvement of the original multispectral cloud detection appeared necessary.
Considering Figs. 2, the clear-sky fraction of the original TOVS Path-B data decreases by about 3% when the NOAA-12 satellite comes into operation (from September 1991 onward), although the equator crossing times of NOAA-10 and NOAA-12 are the same. This difference can be explained by a lack of satellite intercalibration of the HIRS VIS channel: By removing the VIS albedo test from the original TOVS Path-B cloud detection (Stubenrauch et al. 1999a) the systematic clear-sky fraction difference due to satellite change disappears. However, the clear-sky fraction increases by about 4% over the globe.
Two further tests of the original cloud detection have also been found to misidentify volcanic aerosol as cloud: a test on the difference between surface temperature estimates at 4 μm and at 11 μm and, over ocean, a test on the difference between a climatological monthly mean sea surface temperature and the surface temperature estimate at 11 μm (Table 1 of Stubenrauch et al. 1999a). Therefore these tests have been removed. To obtain a similar cloud amount as before, the remaining cloud test thresholds had to be slightly readjusted. This was done by comparing distributions of the test variables between clear-sky and cloudy scenes of the DSD5 dataset. Examples of these distributions are shown in Figs. 3. Further controlling of the thresholds was done by investigating frequencies of clouds in general and of clouds retrieved with εcld = 0 (probably clear sky falsely identified as cloud), which have been detected by a single cloud detection test.
After applying the revised cloud detection tests summarized in Table 1, the average global clear sky fraction is about 26% over ocean and 31% over land. The revised cloud detection reduces the artificial clear-sky fraction decrease just after the Mt. Pinatubo eruption from 20% to 2% over tropical ocean.
4. Average and regional cloud properties
Table 2 gives an overview of average cloud properties and cloud type amounts over the globe and over selected regions and seasons, as obtained from revised TOVS Path-B and ISCCP D2 data (Rossow et al. 1996). ISCCP D2 data provide monthly averages and cloud type statistics over 2.5° latitude × 2.5° longitude, from three-hourly observations. The TOVS Path-B averages shown in Table 2 are from NOAA-10 and NOAA-12 data, with observation times at 0730 and 1930 LT.
According to ISCCP and TOVS retrievals about 70% of the earth’s surface is covered by clouds, with slightly more cloudiness over ocean than over land. From surface observations cloudiness was estimated slightly lower: 65% over ocean (Warren et al. 1988) and 52% over land (Warren et al. 1986). The average cloud temperature is about 261 K, which is 28° colder than the average surface temperature. Clouds are lower and slightly denser over ocean than over land, according to mean εcld. It is interesting to note that the global averages from TOVS Path-B and ISCCP agree quite well, even though they have been obtained from different instruments and with different temporal and spatial resolution.
However, if one considers the amounts of cloud types in Table 2, the instruments which were used for the observations play an important role. (The ISCCP results do not considerably change when using only observation times close to those of the NOAA-10 and NOAA-12 satellites.) Whereas the globe is covered by 22% high clouds, 20% midlevel clouds, and 27% low clouds according to ISCCP, the corresponding cloud type amounts reported by TOVS Path-B are 30%, 12%, and 31%. As mentioned in the introduction, the higher spectral resolution of TOVS is better adapted to identify high semitransparent clouds. This is also confirmed by the high cloud frequency of 33% (averaged from NOAA morning and afternoon satellite HIRS measurements from 1979 to 2001) reported by the University of Wisconsin (UW) NOAA Pathfinder analysis (Wylie et al. 2005).
Surface observers see clouds from below: leading to 18% high clouds and 46% low clouds over ocean (Warren et al. 1988) and 27% high clouds and 22% low clouds over land (Warren et al. 1986). These values agree slightly better with those of TOVS Path-B over land than over ocean. According to Wang et al. (2000) using rawinsonde data, there are more multilayered cloud systems over ocean than over land. However, overlap is accounted for in the climatologies of Warren et al. (1986, 1988). Therefore the smaller amount of high clouds over ocean than over land inferred from surface reports may be due to a larger threshold optical depth for visibility of cirrus in the presence of marine boundary layer haze.
From Table 2 we also deduce that high opaque clouds (Cb for ISCCP or thick Ci for TOVS) cover only a small area of the globe of about 2.5%. A similar value of 3% is reported by Wylie et al. (2005). Cirrus clouds (including cirrostratus and cirrocumulus for ISCCP or Ci and thin Ci for TOVS) have a much larger horizontal extent of about 25% in the midlatitudes (30°–60°N and 30°–60°S) to 45% in the Tropics (15°N–15°S). ISCCP detects 4% and 24% less cirrus in these regions, respectively.
Seasonal differences are greater in the Northern Hemisphere (NH) than in the SH midlatitudes, probably linked to the distribution of land masses. These differences are more pronounced in the TOVS Path-B dataset than in the ISCCP dataset, with more cirrus in summer and more midlevel clouds in winter.
Figures 4 present the latitudinal dependence of ECA and pcld as annual means, as well as means for December–February (DJF) and June–August (JJA). The mean standard deviations of these properties are also shown in these figures. They have been calculated as the standard deviations of the monthly means over the latitudinal bands, averaged over the whole time period. Annual mean cloud amount (CA) and pcld from ISCCP are presented for comparison. One has to keep in mind that the temporal resolution of ISCCP is better and that ECA should always be smaller than CA by a factor corresponding approximately to εcld.
ECA is lowest in the subtropics (15°–30°N and 15°–30°S) with an average of about 0.3, and it increases toward higher latitudes to up to 0.6 and 0.7 in the NH and SH, respectively. A local maximum at about 2°N of about 0.5 indicates the ITCZ. The migration of the ITCZ into the summer hemisphere is apparent on ECA as a narrow maximum of 0.57 at about 7°N and a much broader maximum of about 0.48 from 2° to 10°S, respectively. Regional variability within each latitude band is estimated by the standard deviation of monthly mean ECA to about 0.1. Seasonal variability is in general smaller than regional variability, except in the subtropics with an ECA difference of 0.15 between summer and winter, linked to the migration of the ITCZ. The latitudinal behavior of ECA and CA of ISCCP are very similar, with CA on average 0.2 to 0.3 larger than ECA.
Considering effective high cloud amount (EHCA), separately over ocean and over land for JJA and DJF (Fig. 5), one observes that in JJA high clouds of the ITCZ have about the same EHCA over ocean and land, whereas in DJF the ITCZ is mostly apparent over land. ISCCP daytime high cloud amount (HCA) has nearly the same value as TOVS EHCA, especially at the ITCZ. Only in midlatitudes over land ISCCP HCA is about 0.1 higher than TOVS EHCA. However, average εcld of high clouds from TOVS Path-B lies between 0.2 (SH subtropics) and 0.7 (midlatitude winter). This is another sign that ISCCP is less sensitive to high cirrus clouds. Effective cloud amount of midlevel and of low clouds (not shown) have a minimum in the Tropics and increase toward higher latitudes, with a stronger increase for low clouds over ocean and a stronger increase for midlevel clouds over land. Surface observations (Warren et al. 1988) show over ocean about 15% cumulus in the Tropics and subtropics, and 20% stratus, the latter increasing toward higher latitudes (up to 60%).
Average pcld is lowest in the Tropics with about 450 hPa, and it increases toward higher latitudes to up to 700 hPa. The migration of the ITCZ is similarly apparent in pcld. Mean pcld of ISCCP varies from 520 hPa to 700 hPa. This smaller spread compared to TOVS is linked to the nondetection of cirrus by ISCCP in the case of multilayer clouds (Stubenrauch et al. 1999c). An increase of about 50 hPa in pcld in the subtropical subsidence regions in winter indicates more stratocumulus in these regions during the winter season. The standard deviation of monthly mean pcld within each latitudinal band is about 50 hPa in the midlatitudes and increases toward the ITCZ to about 100 hPa. The high longitudinal pcld variability in the subtropics and Tropics can be explained by observations by Mace and Benson-Troth (2002), showing that in the Tropics highest cloud layers consist of high cirrus and low boundary layer clouds, without transition to midlevel clouds.
5. Seasonal cycle of cloud properties
High clouds are best observed by satellite instruments, because they view the atmosphere from above. Midlevel and low clouds can only be observed from passive remote sensing aboard satellites when there are no higher clouds above which obscure the view. The emphasis is therefore placed on the high cloud analysis here.
a. High clouds (pcld < 440 hPa)
Figure 6 (left) presents time series of the amount of the three high cloud types defined in section 2, for NH midlatitudes (30°–60°N), NH subtropics (30°–15°N), Tropics (15°N–15°S), SH subtropics (15°–30°S), and SH midlatitudes (30°–60°S).
Mesoscale high opaque clouds (thick Ci) cover about 1%–4% of the different latitude bands. The amount is constant (2.5%) in the Tropics and varies from 1% to 2.5% in the subtropics (with a maximum during summer). There are slightly more of these clouds in the NH midlatitudes than in the SH midlatitudes, slight maxima corresponding to winter storm tracks.
In the Tropics there are slightly more thin Ci (24%) than Ci (21%). The NH subtropics have more thin Ci than the SH subtropics. The strong seasonal variation of Ci and thin Ci in the subtropics reflects the migration of the ITCZ toward the summer hemisphere. It is interesting to note that the NH subtropical summer maxima of Ci are about 4% lower in 1987, 1992, and 1993 than during the other years (less Ci in the northern part of the monsoon region), which is at the end of warm El Niño–Southern Oscillation (ENSO) events, which started in 1986 and 1991. The year 1993 was at the very end of a multiyear ENSO event. The midlatitudes are covered by about 15% Ci, but the amount of thin Ci is higher in the NH (10%) compared to 7.5% in the SH. In the NH midlatitudes thin Ci have a pronounced seasonal cycle whereas Ci do not have a seasonal cycle. However, in the SH midlatitudes Ci have a pronounced seasonal cycle and thin Ci have no seasonal cycle. By investigating ocean and land separately (Figs. 9 and 10) one observes that over ocean Ci appear most often in winter, both in NH and SH, which is plausible because they are part of the winter storm tracks. Over NH land Ci and thin Ci appear most often in summer, whereas over SH land only thin Ci have this seasonal cycle. Since there is much more land in NH midlatitudes than in SH midlatitudes, the opposite seasonal cycles of Ci over ocean and over land cancel each other. During the period of eight years, these high cloud amounts remain quite stable.
To study the seasonal cycle in more detail, the monthly data have been averaged over the time period, and mean annual cycles of HCA, EHCA, and of the amount of these three high cloud types are presented in Figs. 7 to 10, distinguishing five latitude bands at six different observation times (0200, 0430, 0730, 1400, 1630, and 1930 LT), separately over ocean and over land. Error bars are not drawn in these figures, because statistical errors of all values are negligible. Standard deviations of EHCA (not shown) are smallest in the Tropics and subtropics (4%–8%). They vary over SH midlatitude ocean from 10% (in winter) to 18% (in summer), whereas over NH midlatitude ocean they vary from 9% (in summer) to 16% (in winter–spring). Over NH land standard deviations of EHCA have a strong diurnal variation, from 6% at 0200 LT to 14% at 1400 LT. From Figs. 7 to 10 one concludes the following.
In general, the seasonal cycle of EHCA is stronger over land than over ocean. It is strongest in the subtropics due to the migration of the ITCZ toward the summer hemisphere. The seasonal cycle of HCA is in general slightly stronger than of EHCA, with the latter the more relevant variable for radiative transfer. While in general the seasonal variation is larger than the diurnal variability, both are of the same order over tropical land.
1) Tropics (15°N–15°S)
This region has the largest HCA and EHCA with 45% and 25% over ocean (Fig. 7, left) and 60% and 35% over land (Fig. 8, left). A seasonal cycle, with a minimum in July and maxima in spring and autumn, is only recognizable over land, with an amplitude of EHCA varying from about 7% at 0730 LT to 13% at 1630 LT. The seasonal cycle of EHCA is similar for all three high cloud type amounts (Figs. 9 and 10). Whereas over ocean (Fig. 9) the amount of thin Ci (25%) is slightly higher than that of Ci (20%), there is more Ci (28%) than thin Ci (23%) over land (Fig. 10). There is also more thick Ci over land (3%) than over ocean (2%), with its strongest seasonal cycle at the time of its maximum (early evening), having an amplitude of about 2.5%.
2) Subtropics (15°–30°N, 15°–30°S)
These regions show the strongest HCA and EHCA seasonal cycles of about 20% and 10% over ocean (Fig. 7) and 30% and 20% over land (Fig. 8), which can be easily explained by the migration of the ITCZ toward the summer hemisphere, producing maxima in HCA and EHCA and in all high cloud-type amounts during this season. The seasonal cycle is slightly stronger over SH land than over NH land, which can probably be linked to the fact that the center of the ITCZ over NH land in July is mostly situated outside the northern limit of the subtropics whereas it is well within the latitude band for SH land during January.
3) Midlatitudes (30°–60°N, 30°–60°S)
In these regions, HCA and EHCA have a relatively small seasonal cycle of about 5% and 3% over ocean (Fig. 7). Over NH land the seasonal cycle is largest in late afternoon of about 10% and 5%, for HCA and EHCA, respectively (Fig. 8). Over ocean the maximum of EHCA of 20% occurs in winter and is related to the winter storm tracks, with a maximum of thick Ci amount of 4% in winter and a minimum below 1% in late summer. Note that HCA already starts to increase in autumn over the NH oceans. Over land the maximum of EHCA of 23% is in summer. However, over SH land there is also another maximum in early winter, probably due to stronger oceanic influence. The seasonal cycle of EHCA corresponds mostly to the one of Ci amount. Over ocean there is more Ci than thin Ci, with less thin Ci in the SH (7%) than in the NH (12%).
b. Midlevel clouds (440 hPa > pcld > 680 hPa) and low clouds (pcld > 680 hPa)
Figure 6 (right) presents time series of the amount of the two midlevel cloud types and low cloud types defined in section 2, for the five latitude bands. It is recalled that these statistics are only collected during periods when there are no higher clouds above. Only a small amount of midlevel clouds (with equal amount of Ast and Acu of less than 5%) is observed in the Tropics and subtropics. This agrees with combined radar and lidar retrievals in the Tropics by Mace and Benson-Troth (2002) and with an analysis of Stratospheric Aerosol and Gas Experiment (SAGE) II data by P.-H. Wang et al. (2001). There are more Ast than Acu in the midlatitudes, with a much stronger seasonal cycle in the NH, leading to 15% Ast in winter. There are about 10% of St and Cu in the Tropics, increasing toward 15% in the subtropics. In the midlatitudes there are again more thicker and extended clouds with a maximum of St of 30% in the SH in summer. In the NH maximum of St occurs in early summer. The seasonal cycle of St is stronger than that of Cu in midlatitudes, and it is greater in the SH than in the NH midlatitudes. The cloud amounts remain quite stable during the eight-year period.
Figures 7 and 8 also present the seasonal cycle of effective mid-cloud amount (EMCA) and effective low cloud amount (ELCA) over five latitude bands at six different observation times, separately over ocean and over land. Since the emissivities for these clouds are close to unity, MCA and LCA are very similar and are therefore not shown.
Over tropical and subtropical ocean, EMCA is very small (5%), whereas it is between 10% and 15% over ocean in the midlatitudes. In all cases seasonal and diurnal variabilities are small. Over tropical and subtropical land EMCA is slightly higher. Only over NH land is there a pronounced seasonal cycle (of about 12%) with a maximum of 22% in winter.
ELCA over ocean is about 25%–30% in the midlatitudes, decreasing toward the Tropics down to 15%. ELCA is in general larger in the SH than in the NH over ocean. The seasonal cycle of low clouds (about 10%) is stronger than that of high clouds in the midlatitudes, and it is different in NH and SH; whereas in the NH the minimum of ELCA is located during summer, it is located during winter in the SH. According to surface observations (Warren et al. 1988) there is more stratus in summer than in winter over midlatitude ocean in both hemispheres. Wang et al. (2000) have observed 10% more multiple layer cloud systems in summer than in winter over NH midlatitude ocean, whereas there is no seasonal change over SH midlatitude ocean. Therefore the increase of stratus from winter to summer according to surface observations could be hidden in the satellite observations by the simultaneous appearance of higher clouds. The seasonal cycle over NH midlatitude ocean is also slightly different in the evening and night than during the morning and early afternoon: The minimum during summer is less pronounced in the morning and early afternoon. In the subtropics one observes again different seasonal cycles in the NH and SH. Whereas the NH has no remarkable seasonal cycle in ELCA, the SH has its maximum in August, certainly linked to the occurrence of stratocumulus in the subsidence regions off the west coasts of South America, Africa, and Australia. Over land the diurnal variation is so strong that no coherent seasonal cycle can be detected.
6. Diurnal cycle of cloud properties
The diurnal cycle of clouds and of precipitation has already been studied in different ways and using numerous datasets, for example Dai (2001) has analyzed precipitation data spanning the period from 1975 to 1997 from about 15 000 stations. He found that drizzle and nonshowery precipitation occur most frequently during night (0000 to 0400 LT) over ocean and in the early morning (0600 LT) over land. Showery precipitation is most frequent during night (0000 to 0400 LT) in the tropical and southern oceans and in the late afternoon over land in summer. The latter is related to solar heating of the ground whereas oceanic precipitation may be linked to low-level convergence induced by pressure tides and high relative humidity at night. A comparison of the vertical structure of precipitation between the Amazon and Africa by Geerts and Dejene (2005), using Tropical Rainfall Measuring Mission (TRMM) satellite data, has shown that in tropical Africa precipitation systems tend to be deeper and more intense than in the Amazon which could be explained by differences in convective available potential energy (CAPE). Storms in Africa are in generally most common in the late afternoon, whereas weaker shallow systems occur more often around noon. ISCCP C2 data (Rossow and Schiffer 1991) have been used by Cairns (1995) and by Bergman and Salby (1996) to describe and explain the diurnal cycle of clouds over the globe. Their findings were that high cloud cover is maximum at about 1700 LT over ocean and persists during night over land, with a minimum at 1100 LT. Low cloud cover is maximum at 0400 LT over ocean and at 1300 LT over land. Furthermore the relative diurnal amplitudes could be linked to diurnal solar amplitude, high cloud fraction, cloud height (low clouds over ocean), surface temperature and its diurnal amplitude (over land), and atmospheric moisture content (low clouds over land).
A regional study by Wylie and Woolf (2002) using Geostationary Operational Environmental Satellite Visible and Infrared Spin Scan Radiometer (VISSR) Atmospheric Sounder (GOES-VAS) satellite data has shown that cirrus clouds identified by this IR atmospheric sounder and by ISCCP follow a similar diurnal cycle over land. The diurnal amplitude in the GOES-VAS data, however, was about 1.5 times larger than in the ISCCP data, due to the higher sensitivity of GOES-VAS to thin cirrus. In the following the diurnal behavior of the TOVS Path-B effective cloud amount and cloud type amount is presented. In section 7 we compare diurnal cycles of high clouds in the tropical convergence zones of the Americas, Africa, and Asia. Even if there are two gaps of about 6 h (from 0730 to 1400 LT and from 1930 to 0200 LT), this dataset still provides complementary information to ISCCP data, because cirrus identification is also available during night from TOVS.
From Fig. 7 we have already deduced that the diurnal variation of high and midlevel clouds over ocean is negligible. Low maritime clouds have no diurnal cycle in winter midlatitudes. The diurnal variability of low clouds is in general larger than that of the other cloud types. Figure 8 shows that the diurnal variability of clouds over tropical land is as large as their seasonal variation. Over land at higher latitudes the diurnal variation of high clouds is maximal during summer. In the following the diurnal cycles of clouds are highlighted over tropical land (Figs. 11, 12), as well as for high clouds at higher latitudes during summer (Figs. 11, left). Figure 12 also compares the diurnal cycles of high, midlevel, and low clouds over land at higher latitudes during winter. The diurnal cycle of high clouds over land is further detailed by distinguishing the amounts of different high cloud types in Fig. 13. Finally the diurnal cycle of low clouds over ocean during summer is presented in Fig. 11 (right). ISCCP daytime cloud amounts are also shown on these figures for comparison. There are maximal five ISCCP values available at observation times between 0600 and 1800 LT, in winter at midlatitudes only three (at 0900, 1200 and 1500 LT).
a. Clouds over land in Tropics and in summer
According to Fig. 11 (left), TOVS Path-B HCA and EHCA have a minimum in the morning and increase during the day by about 10%. EHCA is about 15% (midlatitudes) to 22% (subtropics and Tropics) smaller than HCA, and its diurnal cycle is also slightly smaller. HCA values from ISCCP are between EHCA and HCA values of TOVS Path-B. The SH midlatitudes are an exception with values larger than the HCA values of TOVS Path-B. These are mostly identified as thin cirrus by ISCCP and by TOVS. However, the land cover in SH midlatitudes is very small with the most southern parts of South America, Africa, and Australia. In the NH the diurnal cycles agree well.
In the Tropics the diurnal cycles are similar in January (Fig. 11) and in July (Fig. 12), with ISCCP determining a stronger diurnal cycle, probably due to the better spatial resolution. MCA, shown in Fig. 12 (middle row), is greater late at night and early in the morning than in the afternoon and evening. The decrease of MCA during the day is also confirmed by ISCCP. EMCA however does not have a pronounced diurnal cycle. Note that MCA of ISCCP is about 5% greater in the Tropics than MCA of TOVS, identified by the latter as cirrus. An earlier analysis of ISCCP data (Cairns 1995) has shown that the maximum occurrence of high clouds tends to be in the evening and night and the maximum occurrence of midlevel clouds late at night and in the early morning. This is consistent with our results both for MCA and also with the observation of Ci late at night, since ISCCP falsely identifies thin Ci as midlevel clouds during night. LCA over tropical land has a maximum in the early afternoon, observed both by TOVS and by ISCCP. The amplitude of the diurnal cycle is about 10% (TOVS) to 15% (ISCCP). However ELCA has no pronounced diurnal cycle, indicating that the low clouds are stratus in the morning and cumulus in the afternoon. The diurnal cycle of midlevel clouds is negligible in the other regions (not shown) and the one of low clouds (with no high clouds above) is similar to the one over ocean (see section 6c).
In the SH subtropics HCA of ISCCP in the early morning is by about 10% lower than one would expect from the diurnal cycle from TOVS. Here one has to keep in mind that when the sun is low the determination of cloud properties using VIS information is more difficult.
In all cases TOVS extends information during night, indicating that high cloudiness, increasing during the afternoon, persists longer during night in the Tropics and subtropics than in midlatitudes. Distinguishing the three high cloud type amounts in Fig. 13, one observes especially in the Tropics a maximum of thick Ci in the evening; Ci also increases during the afternoon, but persists longer during night and decreases in the morning. The maximum of thin Ci is in the early afternoon, and then its amount decreases slowly during the afternoon when thick Ci and Ci increase.
b. Clouds over land in winter
Whereas over tropical land the diurnal cycle of high clouds is similar for HCA and EHCA and similar in July and in January, there is a difference between summer and winter in the other regions: In winter (Fig. 12, left) HCA is minimal in the morning and increases toward early afternoon and then decreases again. The amplitude is greater in subtropics than in midlatitudes and greater in the NH than in the SH. Since in winter, daytime is shorter, ISCCP provides only three points in the midlatitudes, but it indicates this behavior in the subtropics, though with a much smaller amplitude. EHCA has no diurnal cycle, and considering Fig. 13 the diurnal cycle of HCA only originates from an increase of thin Ci in the afternoon. In NH midlatitudes in summer the maximum of convection in the afternoon, as observed by ISCCP, is not revealed by TOVS because of the coarse spatial resolution. The increasing amount of Ci shows that the convective core is accompanied by Ci. In the SH the effect is weaker, probably because of the small land portions (see above).
Figure 12 (middle row) show that MCA has a peak in the early morning in midlatitudes, which is less pronounced in EMCA. The maximum in midlevel cloud amount coincides with the peak of nonshowery precipitation in the early morning over land observed by Dai (2001).
LCA and ELCA have maxima in the early afternoon, in agreement with an earlier analysis of ISCCP data by Cairns (1995). The daytime ISCCP data in Fig. 12 (right) show however a much less pronounced peak in midlatitudes.
c. Low clouds over ocean in summer
The amplitudes of the diurnal cycle of LCA and ELCA in Fig. 11 (right) over ocean are less than 10%. Over tropical and subtropical ocean the maxima of LCA and ELCA are just after sunrise, then decreasing during day. This is in agreement with an earlier analysis of ISCCP data (Cairns 1995). In Fig. 11 ISCCP shows a slightly stronger decrease than TOVS in NH subtropics, probably again due to the better spatial resolution. In midlatitudes LCA and ELCA have broader maxima with a decrease only at sunset, a behavior that cannot be observed by ISCCP because of missing data.
7. Comparison of tropical convergence regions
High clouds within the ITCZ are intercompared between the Americas, Africa, and Asia over seasonal and diurnal time scales. Figure 14 present geographical maps of average EHCA for DJF and JJA. The most apparent features are the EHCA maxima in the ITCZ of the tropical summer hemispheres. For DJF EHCA maxima are located in South America and in Africa over land and in Asia over the islands of Indonesia and their surrounding ocean. For JJA EHCA maxima are in the Americas mostly over ocean and in Africa and in Asia over land. From these maps regions with maximum EHCA are identified over the three continents of the Americas, Africa, and Asia covering 0°–20°S and 45°W–70°W, 0°–20°S and 10°–30°E, 5°N–15°S and 90°–150°E during DJF, and covering 5°–20°N and 75°–120°W, 5°–15°N and 0°–30°E, 10°–25°N and 75°–115°E during JJA. For these regions the seasonal cycles of thick Ci, Ci, and thin Ci amounts were analyzed (not shown). The seasonal cycles are strongest and similar over the land regions, about 6%, 40%, and 15% for the three cloud types, respectively. The Americas during JJA exhibits slightly smaller seasonal cycles, whereas the Indonesian region has nearly no seasonal cycles with average amounts of about 3% thick Ci, 40% Ci, and 35% thin Ci.
The diurnal cycles are largest when high cloud amount is maximum. Figure 15 presents the diurnal cycles of thick Ci, Ci, and thin Ci amount over the southern tropical regions during February. The behavior and amount of high clouds are similar over South America and Africa with slightly more thick high clouds over the South American region. As over tropical and subtropical land in Fig. 13, we observe an increase of thick Ci toward the evening and persisting Ci during night. During day Ci decreases and thin Ci increases toward early afternoon. Considering the diurnal cycles of EHCA and Tcld of these high clouds over these three regions in Fig. 16, high cloud temperature decreases over South America and Africa by about 10 K during day, which implies that cloud tops are higher in the afternoon and evening. Over Indonesia, high clouds are in general colder than over South America and Africa, with a less strong diurnal cycle, except for the minimum in the afternoon in combination with an increase of EHCA. The difference in vertical structure of precipitation systems found by Geerts and Dejene (2005) between South America and Africa, the latter producing deeper systems, seems not to be directly reflected in the horizontal coverage and the mesoscale effective emissivity of high clouds. A study by Rickenbach (1999) of the evolution of cold cloud anvils associated with two tropical oceanic mesoscale convective systems during The Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) has shown that the cold cloud spreads when in the western Pacific upper tropospheric easterly winds overrun westerly winds. The coverage of these anvils is largest as the convection is already weakening. For more detailed comparisons of these three convergence regions one would have to combine the TOVS Path-B dataset with dynamical parameters derived for example from the European Centre for Medium-Range Weather Forecasts (ECMWF) 40-yr reanalyses (ERA-40; Simmons and Gibson 2000) with precipitation data.
8. Conclusions
Satellite instruments provide a continuous survey of physical properties of upper cloud layers covering the earth. The presented values have been obtained from eight years (1987–95) of revised TOVS Path-B data. Over this time period, global cloud amounts stayed stable within 2%. In general, ISCCP and TOVS Path-B give similar results: about 70% cloud cover with slightly more clouds over ocean than over land. Only about 3% of the total cloud cover comes from high opaque clouds, but about 30% from single-layer low clouds, with about 10% more in the SH midlatitudes than in the NH midlatitudes. However, TOVS Path-B identifies more cirrus than ISCCP, about 4% in the midlatitudes and up to 20% in the Tropics, where these clouds are the most abundant with an amount of about 45%. Infrared sounders have a good spectral resolution, and these results agree with the UW NOAA Pathfinder analysis using HIRS data.
The seasonal and diurnal cycles of CA and ECA of high clouds, uppermost midlevel and low clouds are stronger over land than over ocean. The seasonal cycle of CA is in general stronger than the one of ECA, the latter the relevant variable for radiative transfer. The seasonal cycle of EHCA is strongest in the subtropics (10% over ocean and 25% over land) due to the migration of the ITCZ toward the summer hemisphere. In midlatitudes over ocean the maximum of Ci is in winter (linked to frontal storm tracks) and of thin Ci in summer. In midlatitudes over land in winter there is a maximum of midlevel clouds (in early morning), probably linked to the troughs of the storm tracks created over ocean, and a maximum of cirrus in summer. Maximum ELCA over ocean occurs in winter in SH subtropics in the early morning hours (occurrence of stratocumulus in the subsidence regions west off the coasts of South America, Africa, and Australia) and in NH midlatitudes without diurnal cycle. Over land in winter the maximum is in the early afternoon, accompanied in the midlatitudes by thin Ci.
Over ocean the diurnal variation of high and midlevel clouds is negligible. Maritime single-layer low clouds in midlatitudes have no diurnal cycle in winter. During summer their amount increases at sunrise and decreases at sunrise. Over tropical land and in the other regions in summer, the maximum of mesoscale high opaque clouds occurs in the evening. Cirrus also increases during the afternoon and persists during night and early morning. The maximum of thin cirrus is in the early afternoon, then decreasing slowly while cirrus and high opaque clouds increase. A small increase of midlevel clouds during night is also observed. TOVS extends information of ISCCP during night, indicating that high cloudiness, increasing during the afternoon, persists longer during night in the Tropics and subtropics than in midlatitudes.
By comparing seasonal cycles of high cloud amount between the Americas, Africa, and Indonesia, one observes the strongest seasonal cycle over the two land regions the Americas and Africa, whereas there is a nearly constant amount of cold high clouds over the Indonesian islands surrounded by ocean. The diurnal cycles are largest when HCA is maximum. High cloud amount and cloud temperature as well as their diurnal behavior are similar over South America and Africa during DJF, with even more Ci during the afternoon over South America than over Africa. Therefore, deeper precipitation systems over Africa than over South America do not seem to be directly reflected in the horizontal coverage and large-scale effective emissivity of high clouds. The next step is to pursue a more detailed study with combined datasets.
Monthly averages and instantaneous retrievals of different cloud physical properties of this revised TOVS Path-B dataset are available at http://ara.Lmd.polytechnique.fr. An extension of the TOVS Path-B dataset is in progress.
Acknowledgments
The TOVS Path-B dataset has been processed at IDRIS, the computer center of CNRS. This work has been supported by CNRS and by the Commission of the European Community, during the FP5 project CIRAMOSA (EVK2-2000-00597). The authors also want to thank R. Armante and F. Eddounia for their support during the production and evaluation of this dataset, as well as G. Abdelaziz for helping improving some figures. Special thanks to W. B. Rossow for many stimulating discussions and to A. Tomkins, S. Warren, A. DelGenio, and an anonymous reviewer for their fruitful comments that improved the text.
REFERENCES
Baum, B. A., V. Tovinkere, J. Titlow, and R. M. Welch, 1997: Automated cloud classification of global AVHRR data using a fuzzy logic approach. J. Appl. Meteor., 36 , 1519–1540.
Bergman, J. W., and M. L. Salby, 1996: Diurnal variations of cloud cover and their relationship to climatological conditions. J. Climate, 9 , 2802–2820.
Cairns, B., 1995: Diurnal variations of cloud from ISCCP data. Atmos. Res., 37 , 133–146.
Chédin, A., N. A. Scott, C. Wahiche, and P. Moulinier, 1985: The improved initialization inversion method: A high resolution physical method for temperature retrievals from satellites of the TIROS-N series. J. Climate Appl. Meteor., 24 , 128–143.
Chédin, A., S. Serrar, N. A. Scott, C. Crevoisier, and R. Armante, 2003: First global measurement of midtropospheric CO2 from NOAA polar satellites: Tropical zone. J. Geophys. Res., 108 .4581, doi:10.1029/2003JD003439.
Chevallier, F., F. Cheruy, N. A. Scott, and A. Chédin, 1998: A neural network approach for a fast and accurate computation of longwave radiative budget. J. Appl. Meteor., 37 , 1385–1397.
Dai, A., 2001: Global precipitation and thunderstorm frequencies. Part II: Diurnal variations. J. Climate, 14 , 1112–1128.
Geerts, B., and T. Dejene, 2005: Regional and diurnal variability of the vertical structure of precipitation systems in Africa based on spaceborne radar data. J. Climate, 18 , 893–916.
Jin, Y., W. B. Rossow, and D. P. Wylie, 1996: Comparison of the climatologies of high-level clouds from HIRS and the ISCCP. J. Climate, 9 , 2850–2879.
Luo, Z., W. B. Rossow, T. Inoue, and C. J. Stubenrauch, 2002: Did the eruption of the Mt. Pinatubo volcano affect cirrus properties? J. Climate, 15 , 2806–2820.
Mace, G. G., and S. Benson-Troth, 2002: Cloud-layer overlap characteristics derived from long-term cloud radar data. J. Climate, 15 , 2505–2515.
McCormick, M. P., L. W. Thomason, and C. R. Trepte, 1995: Atmospheric effects of the Mt. Pinatubo eruption. Nature, 373 , 399–404.
Rädel, G., C. J. Stubenrauch, R. Holz, and D. L. Mitchell, 2003: Retrieval of effective ice crystal size in the infrared: Sensitivity study and global measurements from the TIROS-N operational vertical sounder. J. Geophys. Res., 108 .4281, doi:10.1029/2002JD002801.
Rickenbach, T. M., 1999: Cloud-top evolution of tropical oceanic squall lines from radar reflectivity and infrared satellite data. Mon. Wea. Rev., 127 , 2951–2976.
Rossow, W. B., and R. A. Schiffer, 1991: ISCCP cloud data products. Bull. Amer. Meteor. Soc., 72 , 2–20.
Rossow, W. B., and L. C. Garder, 1993: Cloud detection using satellite measurements of infrared and visible radiances for ISCCP. J. Climate, 6 , 2341–2369.
Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80 , 2261–2287.
Rossow, W. B., A. W. Walker, D. Beuschel, and M. Roiter, 1996: International Satellite Cloud Climatology Project (ISCCP): Description of new cloud datasets. WMO Tech. Doc. 737, World Climate Research Programme (ICSU and WMO), 115 pp.
Scott, N. A., and A. Chédin, 1981: A fast line-by-line method for atmospheric absorption computations: The 4A automized atmospheric absorption atlas. J. Appl. Meteor., 20 , 801–812.
Scott, N. A., and Coauthors, 1999: Characteristics of the TOVS Pathfinder Path-B data set. Bull. Amer. Meteor. Soc., 80 , 2679–2701.
Simmons, A. J., and J. K. Gibson, 2000: The ERA-40 Project Plan. ERA-40 Project Rep. Series 1, ECMWF, 63 pp.
Smith, J. A., and J. P. Taylor, 2004: Initial cloud detection using EOF components of high-spectral-resolution infrared sounder data. J. Appl. Meteor., 43 , 196–210.
Stubenrauch, C. J., and F. Eddounia, 2001: Cloud property survey from satellite observations using vertical sounders (TOVS Path-B) and imagers (ISCCP). Proc. Workshop on Ion–Aerosol–Cloud Interactions, Geneva, Switzerland, CERN, CERN Yellow Rep. CERN-2001-007, 63–74.
Stubenrauch, C. J., A. Chédin, R. Armante, and N. A. Scott, 1999a: Clouds as seen by infrared sounders (3i) and imagers (ISCCP). Part I: Evaluation of cloud parameters. J. Climate, 12 , 2189–2213.
Stubenrauch, C. J., A. Chédin, R. Armante, and N. A. Scott, 1999b: Clouds as seen by infrared sounders (3i) and imagers (ISCCP). Part II: A new approach for cloud parameter determination in the 3I algorithms. J. Climate, 12 , 2214–2223.
Stubenrauch, C. J., W. B. Rossow, N. A. Scott, and A. Chédin, 1999c: Clouds as seen by infrared sounders (3i) and imagers (ISCCP). Part III: Spatial heterogeneity and radiative effects. J. Climate, 12 , 3419–3442.
Stubenrauch, C. J., F. Eddounia, and G. Rädel, 2004: Correlations between microphysical properties of large-scale semi-transparent cirrus and the state of the atmosphere. Atmos. Res., 72 , 403–423.
Stubenrauch, C. J., F. Eddounia, and L. Sauvage, 2005: Cloud heights from TOVS Path-B: Evaluation using LITE observations and distributions of highest cloud layers. J. Geophys. Res., 110 .D19203, doi:10.1029/2004JD005447.
Uddstrom, M. J., and L. McMillin, 1993: A collocation archive of radiosonde and satellite data: 1979 to 1993. NOAA Tech. Memo., 54 pp.
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 , 3041–3056.
Wang, P-H., R. E. Veiga, L. B. Vann, P. Minnis, and G. S. Kent, 2001: A further study of the method for estimation of SAGE II opaque cloud occurrence. J. Geophys. Res., 106 , 12603–12613.
Warren, S. G., C. J. Hahn, R. M. Chervin, and R. L. Jenne, 1986: Global distribution of total cloud cover and cloud type amounts over land. NCAR Tech. Note NCAR/TN-273+STR, 29 pp. and 200 maps.
Warren, S. G., C. J. Hahn, J. London, R. M. Chervin, and R. L. Jenne, 1988: Global distribution of total cloud cover and cloud type amounts over ocean. NCAR Tech. Note NCAR/TN-317+STR, 42 pp. and 170 maps.
Wylie, D. P., and W. P. Menzel, 1999: Eight years of cloud statistics using HIRS. J. Climate, 12 , 170–184.
Wylie, D. P., and H. M. Woolf, 2002: The diurnal cycle of upper-tropospheric clouds measured by GOES-VAS and the ISCCP. Mon. Wea. Rev., 130 , 171–179.
Wylie, D. P., W. P. Menzel, and K. I. Strabala, 1994: Four years of global cirrus cloud statistics using HIRS. J. Climate, 7 , 1972–1986.
Wylie, D. P., D. L. Jackson, W. P. Menzel, and J. J. Bates, 2005: Trends in global cloud cover in two decades of HIRS observations. J. Climate, 18 , 3021–3031.
Time series of monthly mean bias adjustment corrections (simulated TB − observed TB) of HIRS channel 8 (at 11 μm), separately for land and ocean and for tropical (trp) and midlatitude (midl) air masses. Observations are from the DSD5 dataset, using HIRS on board NOAA-10 from June 1987 until August 1991 and HIRS on board NOAA-12 from September 1991 until May 1995. The observation time of both satellites is approximately 0730 and 1930 LT.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Time series of clear sky (1 − cloud amount) × 100%, (top) over the globe and (bottom) over the Tropics, separately (left) over ocean and (right) over land, using different TOVS cloud detection algorithms. Data from June 1987 until August 1991 have been obtained from the NOAA-10 satellite, from September 1991 until May 1995 from the NOAA-12 satellite. The observation time of both satellites is approximately 0730 and 1930 LT.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Frequency distributions of (top) [Σ biTi(∼15 μm) − TB(MSU2)] and (bottom) [Ts(3.7 μm) − Ts(4.0 μm)] for clear-sky (solid line) and cloudy scenes (broken line), night measurements only, in the Tropics (left) over ocean and (right) over land.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
TOVS Path-B effective (top) cloud amount and (bottom) cloud pressure together with standard deviations (below) within 2° latitude bands as function of latitude, for all seasons and for JJA and DJF. The latitudinal behavior of ISCCP cloud amount and cloud pressure (averaged over all seasons) is also shown. Averages are from 1987 to 1995.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
TOVS Path-B EHCA and ISCCP HCA as a function of latitude (top) over ocean and (bottom) over land, separately for JJA and DJF. Averages are from 1987 to 1995.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
TOVS Path-B time series of amounts of (left) cloud types thick Ci, Ci, and thin Ci and of (right) cloud types Ast, Acu, St, and Cu, for five latitude bands. Data are from NOAA-10 and NOAA-12 satellites.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Seasonal variation of (left) HCA (symbols) and EHCA (symbols and lines), (middle) EMCA, and (right) ELCA over five latitude bands for six different observation times, separately over ocean.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Seasonal variation of thick Ci, Ci, and thin Ci amounts over ocean, separately for five latitude bands and for six different observation times (same symbols as in Fig. 7).
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Diurnal variation (left) of HCA, EHCA over land in summer, and (right) of LCA and ELCA over ocean in summer, for five latitude bands. Data are averaged over July in the NH and over January in the Tropics and in the SH. HCA and LCA from ISCCP are also presented during daytime.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Same as Fig. 11 except over land in winter.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Diurnal variation of thick Ci, Ci, and thin Ci amounts over land for five latitude bands, separately in January and in July. In the first row are also given the Cb amounts obtained from ISCCP.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Geographical maps of EHCA (%) (top) for DJF, and (bottom) for JJA. Data have been averaged from 1987 to 1995.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Diurnal variation of high cloud type amounts (thick Ci, Ci, and thin Ci from top to bottom) over southern tropical convergence regions in the Americas, Africa, and Indonesia, averaged over February.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Diurnal variation (top) of high cloud temperature and (bottom) of effective high cloud amount over southern tropical convergence regions in the Americas, Africa, and Indonesia, averaged over February.
Citation: Journal of Climate 19, 21; 10.1175/JCLI3929.1
Revised TOVS Path-B multispectral cloud detection.
Average cloud properties and cloud type amounts from 8 yr (1987–95) of TOVS Path-B and of ISCCP (in parentheses) analyses. Cloud amount from surface observations is 65% over ocean (Warren et al. 1988) and 52% over land (Warren et al. 1986).
