An Observational View of Relationships Between Moisture Aggregation, Cloud, and Radiative Heating Profiles
- 643 Downloads
Data from several coincident satellite sensors are analyzed to determine the dependence of cloud and precipitation characteristics of tropical regions on the variance in the water vapor field. Increased vapor variance is associated with decreased high cloud fraction and an enhancement of low-level radiative cooling in dry regions of the domain. The result is found across a range of sea surface temperatures and rain rates. This suggests the possibility of an enhanced low-level circulation feeding the moist convecting areas when vapor variance is large. These findings are consistent with idealized models of self-aggregation, in which the aggregation of convection is maintained by a combination of low-level radiative cooling in dry regions and mid-to-upper-level radiative warming in cloudy regions.
KeywordsConvective aggregation Radiation Water vapor Satellite Observations
Radiative-convective equilibrium, in which heating of the atmosphere by moist convection and precipitation balances radiative cooling, is an idealization of the Earth’s tropical atmosphere that neglects advective energy transport. Over a uniform surface, and in domains large enough to contain many convective elements, the null hypothesis would be that convection in radiative-convective equilibrium would be distributed roughly uniformly throughout the domain. It has been known for more than 20 years, however, that convection in numerical model simulations of radiative-convective equilibrium frequently gathers itself together, increasing in spatial scale until, under many circumstances, the entire domain contains only a single region of convection (Held et al. 1993; Bretherton et al. 2005; Stephens et al. 2008; Wing and Emanuel 2014; Tompkins and Craig 1998). The phenomenon of convective ‘self-aggregation’ was originally noted in cloud-resolving models in which deep convection is explicit (Held et al. 1993; Bretherton et al. 2005; Stephens et al. 2008; Wing and Emanuel 2014) but it also appears in global models (Bony et al. 2016; Shi and Bretherton 2014; Reed et al. 2015; Coppin and Bony 2015) in which convection is parameterized.
Self-aggregation has primarily been found to occur in simulations with warm Sea Surface Temperatures (SST) with a possible SST threshold below which aggregations does not occur (e.g., Wing and Emanuel 2014). A precise threshold remains elusive: Self-aggregation has been simulated at SSTs less than 300 K (Wing et al. 2016; Holloway and Woolnough 2016) and there are even simulations with an upper SST bound above which self-aggregation does not occur (Wing and Emanuel 2014). While studies have consistently found a relationship between temperature and self-aggregation, there is not a consensus on the specific nature of this relationship. The remainder of this paper will address aggregation in warm SST environments with the caveat that the phenomena may have broader applicability.
Models show that in a fully aggregated state, the atmosphere consists of a few very moist regions containing strong convection and a much larger dry, subsiding region. The contrast in humidity between dry and moist regions is strong so that the variance of the moisture field increases with the degree of aggregation (Bretherton et al. 2005; Wing et al. 2016). High clouds are less frequent in aggregated states, allowing increased longwave (LW) radiative cooling to space. The precise details of convectively aggregated states are still being explored (Muller and Bony 2015) but the frequent finding that aggregation increases with sea surface temperature suggests a stabilizing feedback on climate reminiscent of the ‘Iris hypothesis’ (Lindzen et al. 2001), albeit through different mechanisms than originally proposed (Mauritsen and Stevens 2015; Bony et al. 2016).
The degree to which the self-aggregation of convection is relevant to the Earth’s atmosphere is not entirely clear. As reviewed carefully by Holloway et al. (2017; this issue), the organization of convection has been thought of for many years as being intimately linked to mesoscale systems organized by gravity and other convectively coupled waves (e.g., Mapes 1993) or large-scale circulations. The Earth’s atmosphere does exhibit some characteristics of convective self-aggregation, including the tendency of more organized atmospheres to have lower humidity in clear areas, reduced domain-mean high cloudiness, and increased low cloudiness in non-convective areas, with corresponding impacts on surface and radiative fluxes (Tobin et al. 2012; Stein et al. 2017; Tobin et al. 2013).
As the mechanisms responsible for self-aggregation in radiative-convective equilibrium become more robustly understood, observational tests focusing on those mechanisms become possible (see Holloway et al., this issue; Bony et al., this issue). Here, we expand on the relatively small literature (Tobin et al. 2012, 2013; Stein et al. 2017) examining how the structure of the atmosphere, and the clouds embedded in it, depends on the degree of organization. We exploit a range of colocated observations from the A-Train satellite constellation (Stephens et al. 2002; L’Ecuyer and Jiang 2010) to identify aggregated states in the tropical atmosphere and examine the cloud, precipitation, and radiative structure of these states. We use a new measure of organization based on the spatial variability of the water vapor field. The richness of the observations allows us to identify the circumstances under which aggregation is most frequent and to disentangle the effects of aggregation and mean environmental conditions on the cloud and humidity structure of the atmosphere. We emphasize the important distinction between the observed aggregation that is influenced by external forcing and the idealized concept of self-aggregation, which occurs in the absence of large-scale forcing. The results of this paper must be interpreted with the understanding that these are distinct phenomena.
2 Characterizing Aggregation in Clouds and Their Environment
2.1 Observations from the A-Train
We use data products from the A-Train constellation (Stephens et al. 2002; L’Ecuyer and Jiang 2010). These sensors include the Advanced Microwave Scanning Radiometer for EOS (AMSR-E), the Moderate resolution Imaging Spectroradiometer (MODIS), CloudSat, and CALIPSO. AMSR-E data is available from June 2002 to October 2011 when the instrument spun down. MODIS cloud data is used during this same period. CloudSat/CALIPSO data was used from the period June 2006 to February 2011. All the observations are nearly instantaneous snapshots as opposed to daily average quantities. Results that contain CloudSat/CALIPSO data use the 2006–2011 epoch, whereas results that do not use CloudSat/CALIPSO data use the full A-Train period.
Here, we use Column Water Vapor (CWV) from the AMSR-E sensor (Kawanishi et al. 2003) derived from the version 7 Remote Sensing System algorithms (Wentz and Meissner 2000, 2007). The CWV has an expected precision of 1 kgm−2. The data product is available over ocean surfaces on a 0.25° daily grid with the ascending orbital nodes separated from the descending orbital nodes.
Surface Rain Rate (RR) data are also derived from the AMSR-E sensor using the version 2 Goddard Profiling (GPROF) algorithm (Kummerow et al. 2011). Like the CWV data, the RR is available on a 0.25° daily grid with the ascending orbital nodes separated from the descending orbital nodes.
Cloud data are taken from the collection 5.1, Level 3 Aqua MODIS products. We use the Cloud top pressure (CTP) histograms in the Level 3 products, which are separately stored for ascending and descending nodes (Cloud_Top_Pressure_Day_Histogram_Counts and Cloud_Top_Pressure_Night_Histogram_Counts). The CTP histograms bin the observed cloud counts as a function of 11 bins in 100 hPa increments from the surface to the top of the atmosphere. In addition to the histogram, we use the CTP counts (Cloud_Top_Pressure_Day_Pixel_Counts and Cloud_Top_Pressure_Night_Pixel_Counts) variable in order to calculate cloud fractions from the histograms.
Cloud occurrence profiles are derived from the release-04 2B-Geoprof-Lidar product (Mace et al. 2009) which combines the CloudSat radar cloud mask (Marchand et al. 2008) with the CALIPSO lidar cloud mask (Vaughan et al. 2009). These data are stored on granules that correspond to single orbits; nadir-only sampling means that there is no overlap between the ascending and descending observations.
Precipitation incidence is used from the release-04 CloudSat 2C-Precip-Column product (Haynes et al. 2009; Smalley et al. 2013). Surface rain incidence is defined using the Precip_flag variable = 3, which corresponds to certain precipitation and a radar reflectivity exceeding 0 dBZ at an altitude of approximately 720 m.
Radiative heating profiles are taken from the CloudSat/CALIPSO 2B-Flxhr-Lidar product (Henderson et al. 2013), which combines meteorological analysis with the cloud and aerosol profile information from CloudSat and CALIPSO and a dynamic land surface as input to a radiative transfer model to compute the profile of radiative fluxes at 240 m vertical resolution. Pixel-level RMS differences between this product and the derived top of the atmosphere (TOA) fluxes from the Clouds and Earth Radiant Energy System (CERES) are 5.7 and 16.5 W m−2 for the longwave and shortwave, respectively. However, the biases in pixel-scale retrievals are less than 5 W m−2 (Henderson et al. 2013). We use a modified version of the 2B-Flxhr-Lidar product designed to estimate the diurnal mean fluxes. This modified product computes the shortwave fluxes using 12 different solar zenith angles to account for the diurnal precession of the incoming flux. This product does not, however, account for diurnal changes in the cloud or thermodynamic variables, since they are not directly observed by the A-Train constellation. The Flxhr-lidar product includes an estimate of the Cloud Radiative Effect (CRE). The CRE is calculated explicitly by performing the radiative transfer calculation twice: once all sky and once clear sky. The CRE is then calculated as the difference of the clear-sky calculation from the all-sky calculation. All fluxes that follow are defined positive downward.
2.2 Characterizing Aggregation in the Water Vapor Field
Observational studies of convective self-aggregation to date (Tobin et al. 2012; Stein et al. 2017; Tobin et al. 2013) have quantified the degree of aggregation based on the degree to which cold clouds observed with a domain are spatially coherent, and these studies have also noted that clear areas tend to be less humid when convection is more aggregated. We invert this logic using a measure of aggregation defined by the inhomogeneity of the integrated water vapor field. Our motivations are partly practical: Definitions of aggregation based on clouds require processing high volumes of pixel-scale cloud observations, and cloud observations can be sensitive to the details of the observing system including sensor resolution, inherent sensitivity, and algorithmic choices (Pincus et al. 2012) while being subject to much larger high-frequency variability than is vapor. More importantly, it is useful to understand the degree to which variability in water vapor can exist independently of the systematic organization of convection.
All 5° regions are filtered for SSTs between 300 and 304 K, which include most of the warmest tropical SSTs. Higher SSTs are infrequent enough that sampling is problematic. The lower SST bound is motivated by modeling studies that suggest that SSTs over 300 K are the relevant regime for self-aggregation. We also note that results are robust down to SSTs of 296 K.
Daily data can occasionally be missing in areas of intense precipitation where high winds affect surface emissivity and large ice water contents cause scattering of the emission signal. These missing data points are disregarded in the calculation of the aggregation. The missing data are most likely the high tail of the CWV distribution and therefore may introduce some systematic bias in the calculation of α; however, this influence is somewhat mitigated by the normalization in Eq. 1 as both the standard deviation and mean will be biased low when data are missing.
3 Relationships Among Clouds, Humidity, and Aggregation
3.1 Cloudiness Depends on Sea Surface Temperature and Aggregation State
3.2 Cloudiness, Radiative Heating, and Convective Intensity
Figures 1, 2, 3, 4, and 5 mix many different convective states across the aggregation index α. Might these results be the result of systematic variation in convective activity with the vapor variance, rather than an indication of the aggregation of convection? To address this concern, we further stratify our results by the observed rain rate averaged over each 5° × 5° region following the approach of Stein et al. (2017). This admittedly rough metric for convective intensity is the best available from the A-Train observations. A better measure of convective area fraction might be gleaned from Global Precipitation Measurement (GPM) mission or Tropical Rainfall Measurement Mission radar observations, which can identify convective precipitation using the spatial variance of the radar reflectivity; however, these observations are rarely coincident with the A-Train data used in this study. The GPM mission includes the GPM Microwave Radiometer (GMI), which has similar characteristics to the AMSR-E radiometer, so it would be possible to examine GPM radar observations in terms of the aggregation index defined in this paper.
3.3 Does Vapor Aggregation Imply Convective Aggregation?
The results already shown, using a measure of aggregation defined by the water vapor field, show variations of cloudiness consistent with observations stratified by the connectedness of clouds themselves, suggesting that large-scale variance of water vapor and the mesoscale distribution of clouds are tightly linked. In this section, we explore these relationships more carefully.
Within each rain rate/SST regime, the moistest areas have the highest cloud fraction at all levels, while the driest show very little cloud in the middle atmosphere and enhanced low-level cooling. This picture is consistent with modeling results showing preferential convection in the moist regions that can be maintained (or caused) by an enhanced LW radiative cooling in the dry region (c.f. Figure 2 in Muller and Bony (2015)).
4 Summary and Discussion
We have explored the relationships among clouds, precipitation, radiation, and a measure of the convective aggregation given by the variance in the column water vapor field in large domains over the tropical oceans. This study was motivated by a number of modeling studies hypothesizing an increase in convective aggregation with warming SSTs. Cloud modeling studies of this convective self-aggregation robustly find that the water vapor variance increases with aggregation of the convection. Over ocean surfaces, column water vapor is well-measured and relatively continuous. We expect it provides a measure complementary to the infrared cloud observations that have been employed by previous observational studies of convective aggregation (Tobin et al. 2012; Holloway, this issue).
In the observations presented here, we see a reduction in the area of high cloud cover and an associated increase in the longwave cooling of the atmospheric column to space, as the degree of aggregation increases. We further observe an increase in the low cloud cover with increased aggregation. The enhanced cooling occurs in the dry regions of the domain, reinforcing the moist-static energy gradient between moist and dry regions. Modeling studies suggest that this radiative effect acts as a positive feedback contributing with other processes to maintain the organization of convection in moist areas (Muller and Bony 2015). It is important here to draw a distinction between the initiation and maintenance of the aggregated state. The Muller and Bony study finds that cooling rates localized at cloud top on the order of ~ 13 K/day are required for the initiation, whereas broad lower tropospheric cooling on the order of ~ 2 K/day is helpful but not necessary for maintaining aggregation. The observations shown here are on the order of the ~ 2 K/day helpful for the maintenance of the aggregated state.
This picture of convective organization is consistent across a range of SSTs and rain rates, which we take as a loose proxy for convective intensity. Sorting the results by rain rate provides some relevance to the cloud-climate-feedback problem in the context of radiative-convective equilibrium. In particular, the results show that regions with very different cloud morphology and associated radiative effects can produce the same mean rain rate. It follows that if the aggregated state becomes more prevalent as SST warms, the Earth system may be able to produce the required rainfall to balance the radiative cooling of the atmosphere while having a significantly reduced amount of high cloud.
An important point not addressed by this study is the issue of the spatial scale over which convection might be expected to aggregate as the climate warms. Will the aggregation tend to occur on the mesoscale, the global scale, or at some scale in between? The cyclical boundary conditions and constraint of mass continuity may mean that modeling studies of self-aggregation are more relevant to global-scale circulations than to the mesoscale. This study supports the view that aggregated convection on the synoptic scale produces an environment with a bimodal moisture distribution including dry regions that produce a positive radiative cooling feedback on the convective circulations. These reinforcing radiative feedbacks on convection have also been noted in interannual variability as manifested in the El Nino Southern Oscillation (Rädel et al. 2016) and are implicit in the global-scale narrowing of the inter-tropical convergence zone (Wodzicki and Rapp 2016).
We emphasize that this study cannot confirm observationally that convection does indeed self-aggregate; testing this hypothesis mechanistically will require targeted observations and analysis (Holloway et al., this issue; Bony et al., this issue). Our survey does demonstrate that certain features of the cloud morphology present in model simulations of self-aggregation are also present in observations of the routine aggregation found in the tropical atmosphere. This suggests that aggregation of convection in Earth’s atmosphere, whatever the mechanism, provides a useful conceptual model through which to view important aspects of cloud feedbacks on climate (Mapes 2016).
This paper originates from discussions during the International Space Science Institute (ISSI) Workshop on ‘Shallow clouds and water vapor, circulation and climate sensitivity.’ This research was primarily carried out by ML at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration under RTOP/WBS (105357/967701.02.01.02.08). RP was supported by the National Science Foundation under award ATM-1138394. AMSR data are produced by Remote Sensing Systems and were sponsored by the NASA AMSR-E Science Team and the NASA Earth Science MEaSUREs Program. Data are available at www.remss.com. CloudSat data were obtained from the CloudSat Data Processing Center (http://www.cloudsat.cira.colostate.edu/). MODIS data were obtained from the Goddard DAAC.
- Holloway CE, Wing AA, Bony S, Muller C, Masunaga H, L’Ecuyer TS, Turner DD, Zuidema P (2017) Observing convective aggregation. Surv Geophys. doi: 10.1007/s10712-017-9419-1. http://link.springer.com/10.1007/s10712-017-9419-1. Accessed 30 Aug 2017
- Stein THM, Holloway CE, Tobin I, Bony S (2017) Observed relationships between cloud vertical structure and convective aggregation over tropical ocean. J Clim 30:2187–2207. https://doi.org/10.1175/JCLI-D-16-0125.1
- Wentz FJ, Meissner T (2000) AMSR Ocean Algorithm, Version 2, report number 121599A-1, Remote Sensing Systems, Santa Rosa, CA, 66 ppGoogle Scholar
- Wentz FJ, Meissner T (2007) AMSR_E Ocean Algorithms, report number 051707, Remote Sensing Systems, Santa Rosa, CA, 6 ppGoogle Scholar
- Wentz FJ, Meissner T, Gentemann C, Brewer M (2014) Remote sensing systems AQUA AMSR-E daily environmental suite on 0.25 deg grid, version 7.0. Remote Sensing Systems, Santa Rosa, CA. at www.remss.com/missions/amsre
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.