1. Introduction
Clouds are a reflection of a process involving the necessary ingredients of moisture and uplift. The vertical distribution of clouds is of fundamental importance to many areas of atmospheric science. The height of the cloud top and base, along with layering characteristics of clouds, has a significant effect on the global radiation balance (e.g., Slingo and Slingo 1988). While sufficient moisture in the atmosphere is a necessary condition for clouds to form, there are many nuances involving the interplay of atmospheric dynamics and condensation nuclei with water vapor that determine the type and placement of clouds. In this study, the vertical occurrence of cloud in several geographic regions is compared to variations in total precipitable water (TPW).
McClain (1966) pioneered a look at the relationship between cloud vertical structure and atmospheric moisture. Four manually classified cloud types from Television and Infrared Observation Satellite (TIROS) imagery were compared to mean 1000–500-mb relative humidity and a related quantity called saturation deficit. Results showed that the clouds manually classified as low- or midlevel stratiform or high cirriform were mainly associated with low column relative humidity, while deeper clouds occurred with high column relative humidity. Garand (1993) used a similar approach but with an objective classification of cloud type in visible and infrared imagery to estimate dewpoint depression in several layers and then to predict TPW. Rossow et al. (2005) and Rossow and Zhang (2010) used optical thickness and cloud-top pressure from the International Satellite Cloud Climatology Project (ISCCP) to estimate cloud vertical distributions and structure, and compared the predictions to CloudSat/Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) mission cloud vertical occurrence profiles. No information on water vapor was used. Rossow and Zhang (2010) point out the important distinction between cloud vertical structure and cloud vertical distribution. Cloud vertical structure is not observable except for measurements that can see through obscuring layers of clouds, such as the CloudSat/CALIPSO combined lidar and radar measurements. Cloud vertical distribution is a climatological quantity developed by observations of bases or tops. Peixoto and Oort (1996) developed a climatology of global relative humidity and compared this to observations of low, middle, and high cloud vertical distribution in a zonal mean sense. They found similar latitudinal variations in relative humidity (from radiosondes) and cloud cover (from surface observations). Limitations of cloud climatologies at that time to derive cloud vertical distributions were noted.
Anomalies of moisture have been used in climatological studies aimed toward forecasting applications. Junker et al. (2008) classified extreme precipitation events along the U.S. west coast by normalized anomalies of TPW derived from the National Centers for Environmental Prediction (NCEP) model reanalysis. The use of normalized anomalies facilitates a ranking of extreme weather events. Several other normalized variables were used by Grumm and Hart (2001) and Hart and Grumm (2001) to rank extreme winter storms including wind, temperature, height, and 850-mb moisture anomaly. Zeng (1999) combined the National Aeronautics and Space Administration (NASA) Water Vapor Project (NVAP) TPW and satellite cloud-top temperatures to estimate monthly precipitation in the tropics.
In a study of how CloudSat and CALIPSO transects across warm and cold fronts compare to the classic Norwegian model, Naud et al. (2010) used a compositing technique to identify issues with model cloud parameterization, humidity fields, and vertical motion. Analysis of relative humidity fields indicated that water vapor is not lifted enough in modeled midlatitude cyclones and that this is related to weak vertical velocities in the model. The relationship between sea surface temperature and TPW derived by Stephens (1990) over the ocean provides an example of the type of application in model diagnosis to which this technique could be applied. Cloud vertical frequency sensed from a variety of satellite instruments were compared to CloudSat and CALIPSO data by Wu et al. (2009), but with an emphasis on cloud-top measurements.
TPW over the globe is increasingly well observed from a variety of sensors (e.g., Kidder and Jones 2007; Boukabara et al. 2010), while cloud vertical structure is still poorly observed. Most previous work concerned the estimation of cloud vertical structure and moisture fields from visible and infrared cloud types to derive moisture profiles (e.g., Garand 1993) for numerical weather prediction.
In this paper we test the hypothesis that the TPW anomaly is reflective of changes in cloud vertical distribution. From subjective experience at the Cooperative Institute for Research in the Atmosphere (CIRA), values >200% of normal are nearly always cloudy, while values <50% of normal are nearly always clear. This observation was the inspiration for the current research. We explore this idea over three meteorologically distinct regions—the tropical ocean, the midlatitude ocean, and a midlatitude land area.
2. Data sources
a. TPW anomaly
This work uses the blended multisensor TPW product described in Kidder and Jones (2007), which is now operational at the National Oceanographic and Atmospheric Administration (NOAA)/National Environmental Satellite, Data, and Information Service (NESDIS) and distributed to National Weather Service offices via the Advanced Weather Interactive Processing System (AWIPS). The blended TPW product is created from a mixture of polar orbiter microwave TPW retrievals [Advanced Microwave Sounding Unit-A (AMSU-A) and Special Sensor Microwave Imager (SSM/I)], which are augmented by Geostationary Operational Environmental Satellite (GOES) sounder and surface-based global positioning system (GPS) TPW retrievals over the United States, Alaska, Hawaii, and adjacent waters. A histogram adjustment described in Kidder and Jones (2007) is applied to the TPW retrievals to reduce intersensor differences and “seams” in the polar-orbiting component of the product. The input TPW retrievals over ocean are produced by the NOAA Microwave Surface Precipitation Products System (MSPPS) package (Ferraro et al. 2005). The product grid has a resolution of 16 km and is produced hourly. An example of the blended TPW product near the United States is shown in Fig. 1a. Note the extensive coverage of TPW over the ocean except for a few gaps between swaths in the tropical east Pacific. There are also some missing data over ocean in precipitation along the Intertropical Convergence Zone in the same region. The precipitation screening is important for this study, as only nonprecipitating scenes, at the scale of hundreds of square kilometers viewed by AMSU and SSM/I, are used to compute TPW over ocean. The land over the conterminous United States (CONUS) is well covered by interpolation of highly accurate surface-based GPS-derived TPW from the network of 200–300 stations reporting in 2007–09 (Rama Varma Raja et al. 2008). The GPS network is most dense over the southern Great Plains—an area we chose for further detailed analysis. Note that the GPS TPW measurements are all weather, so that precipitating results are included over land. The circular artifacts over Canada and Mexico are due to extrapolation of the sparse GPS coverage. These data-sparse regions are not included in our study. A 6-hourly interval between blended TPW products is used in this study, as that is roughly the amount of time needed for the global oceans to be refreshed by the five or six polar orbiting spacecraft carrying AMSU-A or SSM/I during 2007 to 2009.
Example of (a) blended TPW product and (b) corresponding TPW anomaly product for 0600 UTC 26 Sep 2008. Units of TPW are mm, while anomaly ranges from 0% to 200%. Anomalies >200% are shaded yellow.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
Future revisions to the blended TPW product will use inputs from the NOAA Microwave Integrated Retrieval System (MIRS), which is described in Boukabara et al. (2010). This system includes the exciting addition of passive microwave water vapor retrievals over land.
To give forecasters an idea about how abnormally moist or dry the blended TPW product is, the TPW values from the blended product are divided by the weekly mean TPW values from the NVAP dataset for 1988–99 (Randel et al. 1996). The weekly mean, without a moving window, is the operational method in use at NOAA to calculate the TPW anomaly. It was chosen to provide an adequate number of samples (84) from the 12-yr record of NVAP. NVAP is a daily, 1° resolution, global TPW and four-layer precipitable water vapor dataset. This creates the operational nearly global (70°N–70°S) TPW anomaly “Percent of Normal” product. An example of the TPW anomaly product is shown in Fig. 1b. Abnormally moist regions are shaded in blue, with >200% shaded yellow, while dry regions are shaded in brown. NVAP was created with TPW derived from SSM/I (ocean), NOAA TIROS Operational Vertical Sounder (TOVS; ocean and land), and radiosonde (land). There are 52 weekly mean fields used as normal from NVAP. The Percent of Normal product was initially developed at CIRA for research purposes but has proven useful for tracking atmospheric rivers, return flow of moisture from the Gulf of Mexico, and abnormally dry conditions associated with fire danger. The purpose of the TPW anomaly product is to highlight and quantify important moisture features for forecasters. In Fig. 1a, the plume of tropical moisture with values above 50 mm over the mid-Atlantic region is reflected by anomaly values from 150% to over 200% over Pennsylvania in Fig. 1b.
The TPW anomaly color table ranges from dark brown (0%) through white (100%) to cyan (200%). The Percent of Normal product is especially robust in midlatitudes and shows a wide range, with the bulk of values between 50% and 200% (½ and double the climatological TPW). In the tropics, even hurricanes do not generate values >200% since the background state is already extremely moist. So statistics and forecast rules based on the TPW anomaly are regionally meaningful, but cannot be transferred across widely different climate regimes.
b. CloudSat/CALIPSO 2B-GEOPROF-lidar cloud vertical profile product
Before the advent of the CloudSat/CALIPSO mission, routine measurements of cloud base heights were at sparse surface-based active instrument sites or by ceilometers and human observers at airports over land only. Surface observations by ceilometer or human observer are limited by obscuration of upper-cloud layers by low ceilings so they do not sense the entire vertical profile of cloud.
The CloudSat and CALIPSO satellites are flying in close formation in a sun-synchronous polar orbit. CloudSat carries a 94-GHz cloud radar that is sensitive to medium- to large-sized cloud particles, rain, and snow. CALIPSO carries a dual-frequency (532 and 1024 nm), dual-polarization lidar that is sensitive to small cloud particles and aerosols. Both instruments are pointed down toward the surface of the earth in a slightly off-nadir orientation to reduce specular reflection to the radar from the earth’s surface and to the lidar from horizontally oriented ice crystals. As the spacecraft orbit the earth, each instrument generates a two-dimensional vertical curtain of data through the atmosphere. The orbits of the two spacecraft were chosen so that as one follows the other the footprints of the two instruments overlap and therefore sample the same portion of the atmosphere. By combining the data from the two instruments, a complete picture of cloud vertical structure is created along the orbit path.
The datasets used for this study include CloudSat’s level-2B geometric profile (2B-GEOPROF) and 2B-GEOPROF-lidar products. 2B-GEOPROF (Marchand et al. 2008) contains the cloud mask created from the radar observations at a 1.1-km horizontal resolution and a 500-m vertical resolution oversampled to 240 m. The cloud-mask field is a confidence value used to indicate the presence of cloud in each radar bin. A minimum value of 20 was chosen to include as many clouds as possible including those identified by “weak echoes” as described by Marchand et al. (2008). The 2B-GEOPROF product version used is P_R04. An uncertainty to consider when using this product is the effect of surface clutter in the radar data. As described in Marchand et al. (2008), this effect decreases the signal-to-noise ratio within ~1.2 km of the earth’s surface. This is more noticeable over land where surface clutter effects are more pronounced. As a result, the ability of the cloud-masking scheme near the surface is limited to recognizing light precipitation and drizzle, which have a higher radar reflectivity signal than low-level clouds. Therefore stratus clouds with tops below 1.2 km might be missed by the cloud-mask scheme. The 500-m vertical resolution could misplace cloud tops and bases up to ±250 m, but this error is random and has little to no effect on the statistics presented in this paper.
The 2B-GEOPROF-lidar product (Mace et al. 2009) contains CALIPSO lidar data that have been collocated with the radar footprint. Clouds too thin to be observed by the radar and reported by the 2B-GEOPROF cloud mask are obtained from the 2B-GEOPROF-lidar cloud fraction data field. This field reports the fraction of lidar profiles that indicate the presence of cloud in each radar bin. A minimum value of 50% is used in this study to indicate a cloudy bin. The 2B-GEOPROF-lidar product version used is P2_R04, which incorporates the updated version 3.01 vertical feature mask product from CALIPSO. Versions prior to 3.01 overestimated the frequency of occurrence of low-level clouds (Vaughan et al. 2010). The vertical resolution of the lidar ranges from 30 to 75 m, so placement of cloud tops and bases are generally more accurate than the radar can achieve.
The lidar component of the 2B-GEOPROF-lidar product will detect clouds below 1.2 km not detected by the radar when the lidar is not attenuated by optically thick upper cloud. Clouds may go undetected below 1.2 km when thick upper clouds occur. As will be shown, this is most likely at TPW anomalies >100%. Frequency of cloud occurrence below 1.2 km at these higher TPW anomaly values should be considered as a lower bound in the results presented here, and used cautiously.
c. Time–space matching
Each 2B-GEOPROF-lidar profile is collocated with a single TPW anomaly gridpoint value that is closest in both space and time. First the TPW anomaly product, time is chosen. The TPW anomaly products for each day are composited at 0000, 0600, 1200, and 1800 UTC. For each CloudSat/CALIPSO orbit, the TPW anomaly product generated for the time immediately after the overpass time is matched with the overpass. For example, an overpass occurring between 0700 and 1200 UTC is matched with the 1200 UTC TPW anomaly product for that day. Overpasses occurring after 1900 UTC are matched with the 0000 UTC TPW anomaly product for 0000 UTC the next day. This delay is used since the TPW anomaly time represents the compositing time of various polar orbiter swaths from the previous 6 hours (i.e., the contents of the 1800 UTC TPW anomaly file represent retrievals from the period 1200–1800 UTC). Then the TPW anomaly product spatial location is chosen. For each CloudSat cloud profiling radar (CPR) profile, this is simply the TPW anomaly gridpoint value with a latitude–longitude coordinate closest to the profile’s latitude–longitude coordinate. The TPW anomaly is generated in a Mercator projection with 16-km resolution at the equator. The grid size is 2500 (x direction) × 1437 (y direction). Because 2B-GEOPROF lidar has a significantly smaller footprint than the TPW anomaly gridpoint size, each TPW gridpoint is matched with several contiguous CPR profiles. This is typically 10–12 CPR profiles per grid box, with a maximum of 13 profiles. Some cloud types are smaller than the TPW gridpoint size (e.g., individual cumulonimbus towers), so there is some uncertainty in the exact relationship between these small clouds and TPW anomaly.
For each CPR granule, all profiles occurring in one of the three study regions are selected for use. Each profile is assigned a TPW anomaly value if the anomaly value is present (e.g., nonprecipitating and no swath gap between polar orbiter passes), and the presence of clouds in the GEOPROF-lidar cloud mask is noted. This includes both the presence of clouds anywhere in the profile (either yes or no) and the vertical location of the clouds if the profile has clouds. Then the profiles are binned by TPW anomaly (bin width of 5% anomaly). Each TPW anomaly bin contains the number of profiles in the bin, the number of cloudy profiles, and the number of clouds occurring at each vertical height. This process is done for each granule, so that the data for all granules is simply a summation of the data for each individual granule. From here the frequency of cloudy profiles, the vertical distribution frequency of clouds, and cloud thickness statistics can be calculated.
3. Results
In this exploration, data from December to February (DJF) and June to August (JJA) over three meteorologically distinct regions are examined for 2007–10. The first date used is 1 June 2007, and the last date used is 28 February 2010. The three regions are indicated in Fig. 2 and are
The midlatitude North Pacific (NPAC), which sees frequent extratropical cyclones and is of particular interest for atmospheric rivers that affect the U.S. West Coast. This region is almost completely ocean, except for some very small islands. Widespread marine stratus is often observed in the eastern portion of NPAC. Geographic limits are 30°–40°N, 180°–130°W.
An expanded Niño 3.4 region (Niño). The Niño 3.4 region is strictly defined as from 5°N to 5°S and from 120° to 170°W. In our study, we expand this region to 15°N–15°S to capture the seasonal migration of the ITCZ. No land is in this region. Geographic limits are 15°S to 15°N, 170°–120°W.
A land region over the U.S. Mississippi Valley (MSVL). While most land regions have no blended TPW values, relatively dense surface-based GPS measurements are available in this region (Rama Varma Raja et al. 2008). This region was also chosen for its relatively uniform surface topography. Geographic limits are 30°–40°N, 100°–85°W.
Location map of the three study regions named NPAC, Niño, and MSVL. NPAC spans 30°–40°N, 180°–130°W. Niño spans 15°S to 15°N, 170°–120°W. MSVL spans 30°–40°N, 100°–85°W.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
The number of 2B-GEOPROF-lidar profiles in the three regions for DJF and JJA are shown in Figs. 3 and 4, respectively, binned by TPW anomaly. The solid lines are all observations, while the dotted lines are the number of profiles where cloud at any level was indicated in 2B-GEOPROF lidar. The dashed vertical lines at each extreme of the TPW anomaly are the limits of where <0.01% of the data occurs. The difference between the dotted (cloudy) and solid (all) lines diminishes as the TPW anomaly increases. Once the anomaly nears 150%, the lines are nearly indistinguishable in the three regions. This is evidence for the initial inspiration for this research that the TPW anomaly functions to some limited extent as a cloud mask.
Number of 2B-GEOPROF-lidar profiles per 5% TPW anomaly bin for (a) NPAC, (b) Niño, and (c) MSVL for DJF of 2007–09. Solid lines are all profiles, while dotted lines are profiles with cloud at any level. The two dashed vertical lines at each end of the TPW anomaly are the limits beyond which <0.01% of the data lie.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
As in Fig. 3 but for JJA of 2007–09.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
In Fig. 4, the NPAC and Niño plots have a roughly Gaussian shape in JJA. In DJF, NPAC becomes more skewed to low anomaly values, while Niño also shifts to drier values. It is hypothesized that the drier values in Niño are due to the northward movement of the ITCZ and increasing compensating subsidence between 15°N and 15°S. In the Niño region, the range of anomalies is generally limited to between 50% and 150%. This reflects the tropical nature of this region and the high background TPW values. The NPAC exhibits a larger range than Niño, especially in DJF (Fig. 3). MSVL behaves very differently in JJA, which is somewhat Gaussian, versus DJF when the distribution is broadened with several local maxima. In DJF over MSVL a few rare values up to 300% occur, which are not seen in the other areas. The weekly background TPW here can be on the order of 5–10 mm, so extreme TPW anomalies can be generated over land in winter months, which is not possible over the oceans. Over MSVL, the TPW is derived from GPS only, which retrieves TPW even in precipitating conditions unlike for the NPAC and Niño passive microwave retrievals. These factors contribute to the extremely high values over MSVL. The weather regime in MSVL has a large seasonal change. In summer, MSVL experiences a humid tropical regime, with weak intermittent frontal passages. In winter, polar air masses often penetrate equatorward of this region. The DJF NPAC and MSVL curves are similar—both are affected by the midlatitude storm track.
a. Cloud vertical distribution versus TPW anomaly
In this section we show results of vertical frequency of cloud occurrence versus TPW anomaly for NPAC, Niño, and MSVL during DJF and JJA. Cloud presence is determined by the top and base of the up-to-five cloud layers reported in 2B-GEOPROF lidar. To first illustrate the complementary nature of the lidar and radar inputs, we show radar- and lidar-only results for NPAC in DJF from 2007–09. These are shown in Fig. 5. The dashed lines are height above mean sea level at 5-km intervals from 0 to 15 km. Note the distinct physical difference between the lidar-only cloud frequency and the radar cloud frequency. The lidar detects more upper-level clouds at roughly between 8 and 12 km MSL and detects clouds at a higher altitude; however, the radar detects thicker cloud throughout the troposphere. The 2B-GEOPROF-lidar product leverages these differences and combines the results to detect cloud throughout the troposphere.
Vertical frequency of cloud occurrence binned by TPW anomaly in 5% bins for NPAC region for DJF 2007–09. (a) Clouds only detected by lidar, and (b) clouds only detected by radar. Dashed lines indicate height above mean sea level at 5-km spacing.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
Cloud occurrence probability versus TPW anomaly plots for the three study regions are shown in Figs. 6 and 7 for DJF and JJA, respectively. These plots should be interpreted along with the number of observations shown in Figs. 3 and 4, as very low amounts of samples occur at each end of the anomaly scale. These plots all encouragingly show the tendency first noted by McClain (1966, his Tables 1 and 3)—that of increasing cloud coverage and depth versus some measure of atmospheric moisture.
2B-GEOPROF-lidar vertical cloud frequency binned by TPW anomaly for DJF 2007–09. The three study regions are (a) NPAC, (b) Niño, and (c) MSVL. Horizontal dashed lines are height above mean sea level at 5-km spacing.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
As in Fig. 6 but for JJA 2007–09.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
The DJF results in Fig. 6 have similarities and differences for the three regions. We discuss these plots in more detail.
1) DJF results: NPAC
The boundary layer cloud probability is nearly constant with cloud frequencies of 30%–50% at TPW anomalies from 50% to 150%. At a 160% TPW anomaly, the boundary layer cloud becomes indistinguishable from deep cloudiness. The low cloud regime to about 130% represents conditions within the moist marine boundary layer. Few clouds appear aloft where subsidence prevails. A paucity of cloud in midlevels between roughly 2 and 10 km is observed. Clouds begin to thicken downward below 10 km at >100% anomalies. The DJF results in NPAC are influenced by strong baroclinic waves, which are not as intense in JJA. The strong upward motion with these systems injects more water vapor into the upper troposphere and thus forms or leads to increased persistence of high clouds.
2) DJF results: Niño
This tropical region is meteorologically distinct from NPAC. Low cloud increases as the anomaly increases. A low-cloud signal is detectable to an anomaly of about 150%. The range of the anomaly is generally limited to between 50% and 150% (Fig. 3b). The upper-cloud probability is less than 20% until the anomaly exceeds 100%. There is a greater than 25% chance of clouds at any level from 0 to 10 km for anomalies >120%. At the highest TPW analysis anomaly values cloud frequency does not reach 100%, explained by the inclusion of occasional gaps in cloudiness due to compensating downward motion induced by convection. It should be noted that 2007–09 was dominated La Niña conditions, and results in this region should be interpreted in light of this strong forcing mechanism.
Increasing column moisture has a distinctly different effect on clouds rooted in the boundary layer in the Niño versus the NPAC region. In the NPAC, boundary layer clouds do not become deeper as the anomaly increases (Fig. 6a) until it reaches beyond 140%. In contrast, the clouds in the Niño region begin to grow vertically as the anomaly increases from 90% to 150%.
3) DJF results: MSVL
The most striking feature of the MSVL region in this time period is the very broad range of anomaly values, as is also seen in Fig. 3c. MSVL TPW results are formed by interpolation of surface GPS measurements and the anomaly field is computed by dividing the TPW value by the NVAP weekly mean. Over MSVL, the weekly mean is derived from the integration of radiosondes and weighted blending with the NOAA operational TOVS TPW retrievals (Randel et al. 1996). It is possible that there exist biases between the two different approaches, which are emphasized in the winter. The denominator term (weekly mean TPW) is smaller in DJF, leading to a more extreme range of anomaly values. Notice how the distribution of samples in Fig. 3c over MSVL has multiple minima and maxima. The cloud frequency occurrence in Fig. 6c shows a general trend toward more cloud at higher anomaly values, but is not as monotonic as for NPAC and Niño. The results over MSVL in JJA (Fig. 7) are radically different from the DJF results. The MSVL DJF results are much more difficult to draw interpretations from than the NPAC or Niño results, but they do illustrate the challenges of comparing TPW anomaly and cloud vertical structure over wintertime land.
Results of the same form as in Fig. 6 but for JJA are shown in Fig. 7. NPAC and MSVL are quite different from DJF, while Niño is less cloudy at high anomalies but is otherwise similar to Niño in DJF.
4) JJA results: NPAC
NPAC shows clouds at 0–2 km not varying as a function of the TPW anomaly. There are fewer high clouds in DJF, perhaps because of weaker vertical motions in summer. A paucity of cloud between 2 and 8 km is noticed until an anomaly of 130%, when this height range becomes cloudier.
5) JJA results: Niño
The Niño region is less cloudy in JJA than in DJF. An interesting feature, also present in DJF, is that the boundary layer tops increase in height from about 1 to 3 km as the anomaly increases to 140%. Upper-level cloud shows a weak increase with TPW anomaly. This is the type of behavior that could be examined in a forecast model, or it could be directly compared with TPW. The deep-layer cloud frequency, at anomaly values greater than about 140%, is greater in DJF than JJA.
6) JJA results: MSVL
The MSVL region shows the most radical change from winter to summer. The large wintertime range of TPW anomaly has been replaced by a compact regime with nearly all values between 50% and 160% (see Fig. 4c). There is a surprising lack of clouds below 2 km at anomalies <120%. This might be due to the previously described inability of the radar to sense boundary layer clouds and lidar attenuation by upper-tropospheric clouds. In Fig. 7c, note that upper-tropospheric clouds are becoming deeper and more common as the TPW anomaly increases. Upper-level clouds between 10 and 15 km do show an upward trend with anomaly. The pattern is very similar to Niño, but with a lack of low clouds.
b. Cloud cover probability versus TPW anomaly
Probability of cloud for the three study regions is shown in Figs. 8 and 9 for DJF and JJA, respectively. The solid line is cloud at any level, while the dotted line is cloud frequency for any clouds >2 km MSL. The 2-km threshold was chosen to eliminate any possibility of surface clutter in the CloudSat radar returns, thus providing a more reliable, though less robust, measure of cloud probability. Figure 8, and Fig. 9 for JJA, reveals that the TPW anomaly has little correlation with low cloud in the NPAC region because of the frequent low cloud cover. The Niño region displays a consistent upward trend of cloud probability with TPW anomaly, and at an anomaly of 150% there is a nearly 100% chance of cloud cover. MSVL shows correlation intermediate between NPAC and Niño.
Probability of cloud cover for DJF 2007–09 binned by TPW anomaly. Solid line is for cloud at any level, dotted line is cloud above 2 km MSL. Results are shown for (a) NPAC, (b) Niño, and (c) MSVL. The two vertical dashed lines are as in Fig. 3 and indicate the bounds beyond which <0.01% of the data lie.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
As in Fig. 8 but for JJA 2007–09. The two vertical dashed lines are as in Fig. 4 and indicate the bounds beyond which <0.01% of the data lie.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
c. Integrated cloud thickness versus TPW anomaly
We have observed deeper clouds with more vertical development as the atmosphere moves closer to saturation. We define integrated cloud thickness as the total thickness of 240-m vertical resolution bins, which were classified as cloudy in 2B-GEOPROF lidar. Mean total thickness for the three regions for DJF and JJA is shown in Figs. 10 and 11, respectively. One standard deviation above and below the mean is indicated by the dotted lines. All regions in all seasons do show increasing total cloud thickness as TPW anomaly increases. The relationship is weakest over MSVL in DJF, which is a problematic season and region as described previously. In JJA, MSVL shows a more monotonic relationship.
Mean cloud thickness (km) binned by TPW anomaly for DJF 2007–09. Results are shown for (a) NPAC, (b) Niño, and (c) MSVL. One standard deviation above and below the mean is indicated by dotted lines. The two vertical dashed lines are as in Fig. 3 and indicate the bounds beyond which <0.01% of the data lie.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
As in Fig. 10 but for JJA 2007–09. The two vertical dashed lines are as in Fig. 4 and indicate the bounds beyond which <0.01% of the data lie.
Citation: Journal of Hydrometeorology 13, 2; 10.1175/JHM-D-11-049.1
To determine if the change in cloud thickness with increasing TPW anomaly seen in Figs. 10 and 11 is significant, we use a two-sample independent Student’s t test with unequal variances and sample sizes. For each data region for JJA and DJF, the test samples used are chosen as the samples in 5% TPW anomaly bins possessing the smallest (nonzero) and largest mean thickness values. Only data points within the central 99.99% portion of the data are considered (i.e., the region between the vertical dotted lines) in order to prevent contamination from outlier data. Table 1 displays the minimum and maximum thickness values associated variances and sample sizes, t-test value, degrees of freedom (DOF), and critical t-test value for a one-tailed t test at 99% significance for each season and data region. The one-tailed test is chosen because of the hypothesis that the mean cloud thickness increases with increasing TPW anomaly. Because CloudSat samples clouds every 1.1 km along its orbit, and because many cloud types do not vary significantly in properties over 1.1-km horizontal distance, it is unlikely that every vertical profile in each sample is completely independent of its immediate neighbors. Therefore, we chose to calculate the effective sample sizes used in the t test as one-tenth the actual sample size. This roughly corresponds to an independent vertical profile every 11 km. This is a reasonable compromise between cumuliform clouds, which vary significantly over horizontal distances much less than 11 km, and stratiform and certain cirriform clouds, which can vary significantly only for distances much greater than 11 km. The t-test value for all regions for JJA and DJF is much greater than the critical t-test value for 99% significance, so it is clear that cloud thickness is a function of TPW anomaly.
Results of the t test for cloud thickness.
4. Conclusions
In this work, the winter and summer TPW anomaly for 2007–09 was compared to cloud vertical structure measurements from CloudSat/CALIPSO over three regions. The principal findings are
The cloud vertical structure varies greatly with the NOAA operational TPW anomaly but with important regional differences. Over NPAC, low clouds are not strongly affected by the TPW anomaly, while over Niño low clouds become deeper with increasing anomalies. At a value greater than 150% of normal TPW, cloud occurrence probability below 10 km nearly always exceeds 20%.
Seasonal differences are large over MSVL and NPAC, and less over Niño. MSVL in winter provides the most inconclusive results with the largest anomaly range. In summer, MSVL resembles Niño to some extent, but with few low clouds. This may be due to land surface clutter effects in the radar component of 2B-GEOPROF lidar.
Integrated cloud thickness increases with TPW anomaly, with the strongest relationship over the Niño region.
a. Diagnosis of forecast and climate model hydrometeorological cycle
The TPW anomaly and cloud vertical profiles derived from satellite in this study could be independently derived from model fields and compared to the results shown here. These results and types of plots could be used to understand behavior and errors in the model clouds in a manner similar to Naud et al. (2010).
b. Estimation of cloud base
This has traditionally been a very difficult analysis in atmospheric science. The method of using ISCCP-derived cloud-top pressure and optical depth to estimate cloud vertical structure (Rossow et al. 2005; Rossow and Zhang 2010) might benefit from an additional input of TPW anomaly information to utilize the relationships found here. The TPW anomaly over ocean could be created back to 1987, when the first SSM/I data became available.
The findings of this study could be useful to an analyst in the following way: consider an infrared window channel satellite image with an overcast of cold high clouds. Without other information, an analyst might not be able to distinguish between a high cirrus layer versus a multilayer cloud with a cloud within a few kilometers of the surface. Knowledge that clouds are in an environment with a 150% of normal TPW anomaly would support an interpretation of a multilayer cloud, or alternatively a single deep convective cloud. These results cannot specify which is true at any given time, but they do indicate in a statistical sense whether a single high cloud or high cloud with more cloud below is likely. In the NPAC region, the likelihood of boundary layer cloud is independent of TPW anomaly, so this idea is less useful there.
c. Optical attenuation
The TPW anomaly product, through its relationship to cloud occurrence shown here, might serve as input to a decision aid to assess confidence in the forecast. Earth imaging applications such as high-resolution (submeter) commercial imagery from satellite seek to detect features on the surface with a minimal amount of atmospheric attenuation. The anomaly product could have value in choosing whether to image two equally cloud-free areas, but one with less relative column moisture than the other and a more pristine view.
The relationship between TPW anomaly and cloud occurrence is not expected to be monotonic everywhere. For instance, in a subtropical high-pressure region under a strong inversion, increasing the TPW implies an increase in the boundary layer relative humidity or a deepening of the boundary layer. In the Niño region (Figs. 6b and 7b), the top of the boundary layer cloud increases as the TPW anomaly varies from 80% to 120%. At the same time, the base slightly lowers. Since most TPW occurs in the lowest few kilometers of the atmosphere [a 2-km scale height is a common assumption (e.g., Stephens 1990)], processes such as detrained moisture from convection can moisten upper levels with little impact on the total column value.
An interesting question not addressed in this study is the relative influence of vertical motion and high TPW in producing clouds. Are there more clouds in a high-TPW atmosphere because of the abundance of moisture or because vertical uplift is high?
In future work, this study could be expanded upon to examine layered rather than total precipitable water vapor anomalies, for instance using the microwave profile retrievals from the MIRS system (Boukabara et al. 2010). MIRS will also provide retrievals over land with consistent precipitation screening, and thus allow exploration of varied geographic regions beyond the limited MSVL region presented here.
Acknowledgments
This research was supported by the DoD Center for Geosciences/Atmospheric Research at Colorado State University under Cooperative Agreement W911NF-06-2-0015 with the Army Research Laboratory. Matt Sapiano and Brian McNoldy of the CSU Department of Atmospheric Science provided valuable input.
REFERENCES
Boukabara, S.-A., Garrett K. , and Chen W. , 2010: Global coverage of total precipitable water using a microwave variational algorithm. IEEE Trans. Geosci. Remote Sens., 48, 3608–3621.
Ferraro, R. R., and Coauthors, 2005: NOAA operational hydrological products derived from the Advanced Microwave Sounding Unit. IEEE Trans. Geosci. Remote Sens., 43, 1036–1049.
Garand, L., 1993: A pattern recognition technique for retrieving humidity profiles from Meteosat or GOES imagery. J. Appl. Meteor., 32, 1592–1607.
Grumm, R. H., and Hart R. E. , 2001: Standardized anomalies applied to significant cold season weather events: Preliminary findings. Wea. Forecasting, 16, 736–754.
Hart, R. E., and Grumm R. H. , 2001: Using normalized climatological anomalies to rank synoptic-scale events objectively. Mon. Wea. Rev., 129, 2426–2442.
Junker, N. W., Grumm R. H. , Hart R. , Bosart L. F. , Bell K. M. , and Pereira F. J. , 2008: Use of normalized anomaly fields to anticipate extreme rainfall in the mountains of Northern California. Wea. Forecasting, 23, 336–356.
Kidder, S. Q., and Jones A. S. , 2007: A blended satellite total precipitable water product for operational forecasting. J. Atmos. Oceanic Technol., 24, 74–81.
Mace, G. G., Zhang Q. , Vaughan M. , Marchand R. , Stephens G. , Trepte C. , and Winker D. , 2009: A description of hydrometeor layer occurrence statistics derived from the first year of merged Cloudsat and CALIPSO data. J. Geophys. Res., 114, D00A26, doi:10.1029/2007JD009755.
Marchand, R., Mace G. G. , Ackerman T. , and Stephens G. , 2008: Hydrometeor detection using Cloudsat—An earth-orbiting 94-GHz cloud radar. J. Atmos. Oceanic Technol., 25, 519–533.
McClain, E. P., 1966: On the relation of satellite viewed cloud conditions to vertically integrated moisture fields. Mon. Wea. Rev., 94, 509–514.
Naud, C. M., Del Genio A. D. , Bauer M. , and Kovari W. , 2010: Cloud vertical distribution across warm and cold fronts in CloudSat–CALIPSO data and a general circulation model. J. Climate, 23, 3397–3415.
Peixoto, J. P., and Oort A. H. , 1996: The climatology of relative humidity in the atmosphere. J. Climate, 9, 3443–3463.
Rama Varma Raja, M. K., Gutman S. I. , Yoe J. G. , McMillin L. M. , and Zhao J. , 2008: The validation of AIRS retrievals of integrated precipitable water vapor using measurements from a network of ground-based GPS receivers over the contiguous United States. J. Atmos. Oceanic Technol., 25, 416–428.
Randel, D. L., Vonder Haar T. H. , Ringerud M. A. , Stephens G. L. , Greenwald T. J. , and Combs C. L. , 1996: A new global water vapor dataset. Bull. Amer. Meteor. Soc., 77, 1233–1246.
Rossow, W. B., and Zhang Y. , 2010: Evaluation of a statistical model of cloud vertical structure using combined CloudSat and CALIPSO layer profiles. J. Climate, 23, 6641–6653.
Rossow, W. B., Zhang Y. , and Wang J. , 2005: A statistical model of cloud vertical structure based on reconciling cloud layer amounts inferred from satellites and radiosonde humidity profiles. J. Climate, 18, 3587–3605.
Slingo, A., and Slingo J. M. , 1988: The response of a general circulation model to cloud longwave forcing, Part 1: Introduction and initial experiments. Quart. J. Roy. Meteor. Soc., 114, 1027–1062.
Stephens, G. L., 1990: On the relationship between water vapor over the oceans and sea surface temperature. J. Climate, 3, 634–645.
Vaughan, M., and Coauthors, 2010: Strategies for improved CALIPSO aerosol optical depth estimates. Preprints, 25th Int. Laser Radar Conf., St. Petersburg, Russia, IRC, International Association of Meteorology and Atmospheric Physics, 4 pp. [Available online at http://www-calipso.larc.nasa.gov/resources/pdfs/ILRC_LaRC_2010/Vaughan_ILRC25_2010.pdf.]
Wu, D. L., and Coauthors, 2009: Vertical distributions and relationships of cloud occurrence frequency as observed by MISR, AIRS, MODIS, OMI, CALIPSO, and CloudSat. Geophys. Res. Lett., 36, L09821, doi:10.1029/2009GL037464.
Zeng, X., 1999: The relationship among precipitation, cloud-top temperature, and precipitable water over the tropics. J. Climate, 12, 2503–2514.
