Radiative heating of the ISCCP upper level cloud regimes and its impact on the large-scale tropical circulation

Authors


Abstract

[1] Radiative heating profiles of the International Satellite Cloud Climatology Project (ISCCP) cloud regimes (or weather states) were estimated by matching ISCCP observations with radiative properties derived from cloud radar and lidar measurements from the Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) sites at Manus, Papua New Guinea, and Darwin, Australia. Focus was placed on the ISCCP cloud regimes containing the majority of upper level clouds in the tropics, i.e., mesoscale convective systems (MCSs), deep cumulonimbus with cirrus, mixed shallow and deep convection, and thin cirrus. At upper levels, these regimes have average maximum cloud occurrences ranging from 30% to 55% near 12 km with variations depending on the location and cloud regime. The resulting radiative heating profiles have maxima of approximately 1 K/day near 12 km, with equal heating contributions from the longwave and shortwave components. Upper level minima occur near 15 km, with the MCS regime showing the strongest cooling of 0.2 K/day and the thin cirrus showing no cooling. The gradient of upper level heating ranges from 0.2 to 0.4 K/(day∙km), with the most convectively active regimes (i.e., MCSs and deep cumulonimbus with cirrus) having the largest gradient. When the above heating profiles were applied to the 25-year ISCCP data set, the tropics-wide average profile has a radiative heating maximum of 0.45 K day-1 near 250 hPa. Column-integrated radiative heating of upper level cloud accounts for about 20% of the latent heating estimated by the Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR). The ISCCP radiative heating of tropical upper level cloud only slightly modifies the response of an idealized primitive equation model forced with the tropics-wide TRMM PR latent heating, which suggests that the impact of upper level cloud is more important to large-scale tropical circulation variations because of convective feedbacks rather than direct forcing by the cloud radiative heating profiles. However, the height of the radiative heating maxima and gradient of the heating profiles are important to determine the sign and patterns of the horizontal circulation anomaly driven by radiative heating at upper levels.

1 Introduction

[2] The tropics have been recognized as a source of climate variability at a number of temporal and spatial scales [Hoerling et al., 2001; Selten et al., 2004; Deser et al. 2004]. Diabatic heating associated with clouds and precipitation in tropical convective systems is a key component that drives the variation of the large-scale circulation. The vertical shape of the diabatic heating largely depends on the precipitation and cloud types (Houze, 1982, 1989), and the large-scale circulation is sensitive to variations in the vertical heating structure and its geographical distribution [e.g., Geisler, 1981; Hartmann et al., 1984; Sui and Lau, 1989; Wu et al., 2000; Schumacher et al., 2004]. As such, it is useful to delineate the heating associated with each cloud type to better understand their dynamical impact.

[3] Diabatic heating is composed of latent and radiative components [although vertical eddy transports of sensible heat are another relevant component when considering large grid sizes]. Schumacher et al. [2004] tested the large-scale response to the latent heating distribution associated with varying fractions of stratiform rain. Their work indicated that the geographically varying top-heavy latent heating profile associated with higher stratiform rain fractions resulted in a more realistic upper-level circulation response in an idealized climate model.

[4] Previous modeling work also highlighted the importance of elevated heating associated with the radiative impact of mid- to upper tropospheric cloud on the large-scale tropical circulation. Ramaswamy and Ramanathan [1989] focused on the impact of the absorption of solar radiation by cirrus clouds on the thermal structure of the tropical atmosphere (i.e., upper level warming), while Slingo and Slingo [1988] and Randall et al. [1989] examined the impact of longwave absorption of upper level clouds on the general circulation (which leads to increased precipitation in the tropics, among other impacts). Sherwood et al. [1994] studied the combined shortwave and longwave radiative impact of clouds above 600 hPa and found large-scale circulation responses similar to previous studies, including the weakening of the Hadley and Walker circulations when upper level clouds were excluded from their model simulations. Lohmann and Roeckner [1995] showed that the radiative warming by upper level clouds is similar to the impacts of increased sea surface temperatures and that cirrus clouds increase the climate sensitivity of their model, especially through radiative-convective-dynamical coupling. Zender and Kiehl [1997] further highlighted the sensitivity of climate to anvil by comparing the diagnostic and prognostic anvil formulations and found a stronger tropical circulation with the prognostic anvil paramerterizations. Each of these studies used global climate models (GCMs) capable of radiative-convective feedbacks to modify cloud radiative properties and examine the resulting large-scale response. However, their heating structures were fully model dependent, and the vertical profiles of heating varied among studies (e.g., the longwave atmospheric cloud radiative forcing profiles in the studies by Slingo and Slingo [1988] and Randall et al. [1989] exhibited different signs above 12 km, which will lead to different circulations aloft). These variations stress the need for tropics-wide observations of the vertical structure of radiative heating associated with upper level clouds, in part to assess the validity of GCM studies.

[5] Hybrid modeling-observation studies have also been performed to examine the radiative impact of clouds on the large-scale circulation. Wang and Rossow [1998] interactively changed the cloud vertical structure in a GCM to study the impact of changes in the vertical gradient in radiative heating associated with quasi-realistic cloud distributions. They found that clouds with the same cloud-top pressure but different cloud-layer thickness produce very different Hadley circulations. Thus, clouds with similar top-of-atmosphere forcing (or vertical integral of heating) may have large variability in cloud radiative atmospheric forcing in the vertical. Bergman and Hendon [2000] performed a set of GCM simulations forced by relatively coarse-resolution, reanalysis- and satellite-derived cloud radiative heating profiles and indicated that tropical cloud radiative forcing strengthens the latent heating driven large-scale circulation, although they found that the impact was larger at low levels. To date, similar studies with a climatology of high-resolution radiative heating inputs (and with a focus on upper level clouds) have not been done due to a lack of observationally based data.

[6] The International Satellite Cloud Climatology Project (ISCCP) was established in 1982 and has since provided long-term satellite datasets of cloud properties [Rossow and Schiffer, 1991; Rossow and Schiffer, 1999]. Of particular use in delineating the radiative impact of certain cloud types is the ISCCP cloud regimes, which are objectively identified based on daytime observations of cloud top pressure and optical thickness [Jakob and Tselioudis, 2003; Rossow et al., 2005]. By combining ISCCP cloud regimes and cloud and precipitation data from other sources, several studies have investigated the cloud radiative properties and latent heating profiles for major tropical and midlatitude cloud modes [Jakob et al., 2005; Jakob and Schumacher, 2008; Oreopoulos and Rossow, 2011; Haynes et al., 2011]. For example, Jakob et al. [2005] used the radiative retrievals from the Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Nauru site in the tropical western Pacific (TWP) to characterize the radiative flux for four major ISCCP cloud regimes (suppressed shallow cloud, suppressed thin cirrus, convectively active deep cloud, and convectively active cirrus). Jakob and Schumacher [2008] further expanded the application of the cloud regime data to Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR) observations to estimate the latent heating profiles associated with ISCCP cloud regimes in the TWP.

[7] Department of Energy ARM cloud radars and lidars allow high vertical and temporal resolution calculations of cloud radiative properties [Comstock et al., 2002; Comstock and Jakob, 2004; McFarlane and Evans, 2004; Mather et al., 2007; McFarlane et al., 2007]. By matching ISCCP cloud regimes with cloud radiative information from the DOE ARM TWP sites, this study builds a look-up table of radiative heating profiles associated with ISCCP upper level cloud regimes that is then applied across the tropics. An idealized climate model is further used to examine the sensitivity of the large-scale circulation to the three-dimensional variation of the radiative and latent heating of tropical convective cloud systems. This study only looks at the direct impact of the diabatic heating on the model response. An interactive modeling framework that includes convective interactions is left for future work.

2 Data and Methodology

2.1 ISCCP Cloud Regimes

[8] This study utilizes ISCCP tropical cloud regimes (or weather states), which are derived from the joint frequency distributions of cloud top pressure and optical thickness using a clustering algorithm (Rossow et al., 2005, Figure 1). The tropical cloud regime data covers the area of 15°S-15°N with 2.5° × 2.5° resolution. Only local daytime data are available due to the use of the visible channel in the retrieval of optical thickness [Jakob et al., 2005]. The data used in this study cover the period during July 1983 to December 2008 with a 3-hour sampling interval.

Figure 1.

Cloud top pressure cloud optical thickness histograms of the six cloud regimes for ISCCP tropical data during 1983–2008. RFO refers to the frequency of occurrence, and CCF is the average percent cloud cover inside the 280 km domain for each cloud regime.

[9] The ISCCP tropical cloud regime product contains six regimes with regimes 1–3 representing convectively active, strongly precipitating clouds and regimes 4–6 representing convectively suppressed conditions [Rossow et al., 2005; Jakob and Schumacher, 2008; Chen and Del Genio, 2009]. The first four regimes include a large fraction of upper-level clouds (Figure 1). In particular, regime 1 is indicative of deep convective systems with significant stratiform rain and anvil cloud (CD), regime 2 represents deep cumulonimbus clouds with significant cirrus present (CC), regime 3 is a mix of convective clouds with varying heights and less upper level cloud (MIX), and regime 4 is associated with thin cirrus (CI). Regimes 5 and 6, representing trade cumulus (CU) and stratus/stratocumulus (ST/SC) clouds, respectively, have very little cloud at upper levels (e.g., 440 hPa and above) and are hereafter excluded from this analysis.

[10] Figure 2 shows the geographical distribution of the first four ISCCP cloud regimes. Their occurrence frequencies have somewhat similar geographic patterns with maxima near regions of high rain accumulations. However, variations exist between regimes, with CD most closely linked to regions of tropical rainfall maxima, CC having relatively high occurrence over the Maritime Continent and west Pacific warm pool, MIX showing the largest occurrence over land, and CI occurring on the edges of more convectively active areas.

Figure 2.

Occurrence frequency of the first four cloud regimes from 3-hourly tropical ISCCP data for 1983–2008. Manus is indicated by M, and Darwin is indicated by D.

2.2 MMCR–MPL Radiative Heating

[11] Major goals of DOE ARM are to improve the understanding of cloud-radiation feedbacks in the atmosphere and their representation in climate models. Toward these goals, ARM set up several long-term measurement sites to represent a broad range of environmental conditions. The TWP sites at Manus (2.06°S, 147.42°E), Nauru (0.52°S, 166.92°E), and Darwin (12.42°S, 130.89°E) provide high vertical and temporal resolution cloud properties from a suite of active and passive remote sensing instruments, including the Millimeter Wavelength Cloud Radar [MMCR; McFarlane et al., 2007; Mather et al., 2007], the micropulse lidar [MPL; J. M. Comstock et al., Assessment of Uncertainty in Cloud Radiative Forcing through Retrieval Algorithm Differences: Analysis using 3-years of ARM data at Darwin, Australia, submitted to Journal of Geophysical Research, 2012, hereinafter referred to as Comstock et al., submitted manuscript, 2012]. These cloud properties and associated atmospheric state parameters can be used to calculate radiative heating profiles [Mather et al., 2007; Comstock et al., submitted manuscript, 2012]. In addition, cloud types observed over the TWP sites are typical of the wider tropical western Pacific region [Jakob and Tselioudis, 2003]. This study further assumes the cloud radiative heating profiles at the TWP sites are representative of convectively active regions in the broader tropics. Because the lidar signal can be blocked by lower-level clouds and the radar is not sensitive to optically thin clouds with small particles, a recent study indicates that the ground-based remote sensing observations miss much of the high thin cirrus (> 13 km) observed by satellite lidar measurements at Darwin (T. Thorsen et al., Cloud radiative effects over Darwin using ARM and A-train radar-lidar observations, submitted to Journal of Geophysical Research, 2012). These high thin cirrus clouds are most likely to be relevant to the CI regime.

[12] The data used in this study are the MMCR and MPL derived cloud water content (CWC, including both ice water and liquid water content) and broadband heating profiles from Manus during January 2002 to January 2008 (heating data from 2007 is not available) and Darwin during January 2006 to December 2008 with ~30 m vertical resolution and 2-min temporal resolution (Manus and Darwin data are hereafter called the MMCR-MPL dataset). Note that although the periods for the Manus and Darwin data are not the same, results from the full dataset and subsets of the full dataset are similar such that we choose to use the full datasets for better sampling and statistics. The cloud frequency profiles over Manus and Darwin for the first four ISCCP cloud regimes are shown in Figure 3. The profiles were calculated by counting valid MMCR–MPL CWC values (i.e., CWC greater than zero) at each level and then dividing by the number of cloud occurrence of the associated ISCCP cloud regime. Because the MMCR and MPL are vertically pointing, we averaged the MMCR and MPL cloud profiles in a 1-hour range centered at the ISCCP data time to better match the 2.5°× 2.5° grid box of the ISCCP data.

Figure 3.

MMCR–MPL cloud frequency profiles for ISCCP cloud regimes CD, CC, MIX, and CI over Manus and Darwin.

[13] For Manus, the CD regime (Figure 3, top left) shows a cloud occurrence near 50% throughout most of the troposphere with a maximum frequency of 54% at 11 km, consistent with the ISCCP cloud regime joint distribution of high, thick clouds associated with deep convective and stratiform rain (Figure 1). The CD regime at Darwin shows a similar profile but has ~15% less cloud frequency on average. The CC regime (Figure 3, top right) is more weighted toward upper level cloud with a maximum occurrence of 42% at 11 km for Manus and 12 km for Darwin. These profiles are also consistent with ISCCP observations of deep convection, but with less of the optically thick upper level cloud associated with stratiform rain regions. The MIX and CI regimes (Figure 3, bottom panels) have less upper level cloud (i.e., maximum occurrences ranging from 28-36%) compared to the first two regimes, with Darwin maxima 1.5 km higher than Manus (i.e., 13 km vs 11.5 km). While there are some differences in percent coverage and height maxima between the Manus and Darwin occurrence frequency profiles, the differences are consistent between ISCCP regimes. It is also worth noting that the Manus profiles in Figure 3 are generally consistent with the cloud fraction profiles shown by Jakob et al. [2005] and Chen and Del Genio [2009] using the Active Remote Sensing of Clouds (ARSCL) data product, providing confidence in our retrieval methods.

[14] The radiative heating profile for each ISCCP cloud regime was obtained by averaging the MMCR–MPL infrared and solar radiative forcing profiles (i.e., heating rates with the clear-sky component removed) when the ISCCP cloud regime was observed over Manus and Darwin for the 1-hour period centered at the ISCCP time. Since the ISCCP cloud regime data is limited to daytime observations, we normalized the ISCCP-matched MMCR–MPL solar heating to all-day (i.e., 24-hour mean) heating. To obtain the normalized heating, the temporal mean diurnal variation of the column-integrated shortwave (SW) heating was calculated based on solar forcing profiles from MMCR–MPL data (Figure 4). A weighting parameter can be obtained from this distribution for SW heating at each time, inline image, where inline image is the whole data period mean SW heating rate. This weighting was applied to the SW heating profile for the ISCCP cloud regimes, that is, +swhr (i) is divided by w(i) to obtain the approximate all day mean SW heating rate. The longwave (LW) heating rate does not show significant diurnal variability, so no weighting is needed. In addition, seasonal and geographical variations (i.e., day of year and solar zenith angle variations with latitude) in solar radiative heating were ignored since these variations are relatively small in the tropics.

Figure 4.

Diurnal variation of column integral of solar heating rate in MMCR–MPL data.

[15] Figure 5 shows the SW and LW profiles for the four ISCCP cloud regimes for Manus and Darwin, while Figure 6 shows the profiles of net radiative heating for Manus and Darwin and the average profile for the two sites. The CD and CC regimes have mean radiative heating peaks close to 1.0 K/day at 11.5 km (Figure 6, top panels), with a larger and higher heating maximum at Darwin for the CD regime. The difference at Darwin is likely due to the apparently higher, thinner cloud suggested by the cloud fraction profiles in Figure 3. This leads to stronger SW heating at Darwin as seen in Figure 5. The CD regime also has a mean maximum cooling of -0.3 K/day near 15 km, while the CC regime shows a weaker cooling maximum of 0.05 K/day at 15 km. The CD regime is composed of more optically thick cloud than the CC regime (Figure 1), which contributes to relatively stronger LW cooling aloft, especially over Manus (Figure 5). It is also interesting to note the similarity of the CD mean profile (Figure 6) to the anvil heating in Houze's [1982] idealized MCS, albeit the CD profile is elevated by a few km.

Figure 5.

Radiative LW and SW heating profiles for ISCCP cloud regimes CD, CC, MIX, and CI at Manus and Darwin based on MMCR–MPL retrievals.

Figure 6.

Radiative heating (LW + SW) profiles for ISCCP cloud regimes CD, CC, MIX, and CI at Manus and Darwin and the average of the two sites based on MMCR–MPL retrievals.

[16] The MIX and CI regimes have mean radiative heating maxima at 12 km (Figure 6, bottom panels), which is slightly higher than the CD and CC regimes. However, the MIX maximum of 0.7 K/day is slightly weaker than the CD and CC regimes, likely because of the lower cloud fractions (Figure 3), and the CI maximum of 1.2 K/day is slightly stronger, likely because the thinness of the cirrus promoting SW heating over LW cooling (Figure 5). In addition, the MIX and CI regimes show little to no cooling near cloud top.

[17] Overall, Darwin shows stronger heating and cooling than Manus and thus stronger heating gradients. The strong maxima and vertical gradients in the radiative heating profiles, which range from 0.2 to 0.4 K/(day∙km) from heating maximum to cooling minimum, suggests that the dynamical impact of the radiative heating profile of tropical upper level clouds could be significant. In addition, both the SW and LW components contribute significantly to the net radiative heating vertical structure.

[18] To extrapolate the MMCR-MPL radiative heating profiles to the rest of the tropics, we multiplied the heating profile from Manus, from Darwin, or averaged over the two sites by the mean occurrence frequency of the corresponding ISCCP cloud regime at every 2.5˚× 2.5˚ grid between 15˚S and 15˚N. While this methodology is less accurate than explicit grid-by-grid radiative transfer calculations, it is much more efficient for estimating radiative heating profiles over large space and time scales, which is made possible by the long-term, three hourly ISCCP observations across the tropics. The profiles estimated by this method also maintain the primary features of the high-quality MMCR-MPL radiative heating retrievals for the representative cloud types. The large cloud regime occurrence frequencies for 1-hour sampling at Manus (28%, 20%, 24%, and 17% for regimes CD, CC, MIX, and CI, respectively) and Darwin (19%, 11%, 18%, and 8% for regimes CD, CC, MIX, and CI, respectively) imply that it is reasonable to use this “look-up table” method to estimate tropics-wide radiative heating in convectively active regions. It should be noted that the CloudSat Cloud Profiling Radar (CPR) also provides tropics-wide cloud observations and radiative heating profile retrievals [L'Ecuyer et al., 2008], but its more limited time and space sampling compared to ISCCP hinders climatological applications.

3 Tropics-Wide Diabatic Heating

3.1 ISCCP Cloud Regime Radiative Heating

[19] Figure 7 shows vertical cross sections averaged from 15˚S to 15˚N of the ISCCP cloud regime radiative heating using the MMCR-MPL lookup table based on heating profiles averaged over the two sites. In general, longitudinal variations in the radiative heating resemble the longitudinal variation of precipitation, with maxima over Africa (15–30° E), the Indian Ocean and West Pacific (90–180°E), and South America (75–50°W). This implies a close relationship between the first four ISCCP cloud regimes and deep convective cloud systems. However, the MIX regime radiative impacts extend further into the oceanic intertropical convergence zone (ITCZ) regions and the CI regime radiative heating maxima occur on the periphery of the heating maxima in the CD and CC regimes (cf. to the ISCCP cloud regime occurrence maps in Figure 2). The vertical cross sections in Figure 7 also highlight the relative contribution of each cloud regime's radiative heating, with the largest values of heating for upper-level clouds occurring in the MIX and CI regimes (~0.24–0.28 K/day) and largest values of cooling occurring in the CD regime (~0.06 K/day). In addition, there is relatively more radiative heating at upper levels than mid levels compared to Bergman and Hendon [1998], who also used ISCCP observations to calculate cloud radiative forcing profiles.

Figure 7.

Longitude-height cross sections of radiative heating for ISCCP cloud regimes CD, CC, MIX, and CI averaged over 15°S–15°N from 1983 to 2008 using the look-up table based on radiative heating profiles averaged over the Manus and Darwin sites.

3.2 Comparisons to TRMM PR Latent Heating

[20] While the radiative heating associated with upper level clouds in the tropics may be important to tropical dynamics in the absence of other factors, the large-scale tropical circulation responds to the total diabatic heating (i.e., the radiative plus latent heating components) associated with tropical convective cloud systems. This section evaluates the relative importance of radiative and latent heating to the total diabatic heating of the tropics. Climatological mean latent heating profiles were retrieved based on TRMM PR convective and stratiform rainrates during 1998–2007 [Schumacher et al., 2004]. Both the radiative and latent heating profiles were interpolated from their original resolution to the pressure levels of the CAM3 model (26 levels in total).

[21] Figure 8 shows the geographical distribution of TRMM PR latent heating compared to the ISCCP radiative heating for regimes 1–4 (i.e., CD + CC + MIX + CI) using the look-up table based on heating profiles averaged over the two sites at 250 hPa. Note that because the climatology of occurrence frequency for each ISCCP cloud regime during 1983–2008 looks very similar to that during 1998–2007 (not shown), we will continue to use the 1983–2007 ISCCP radiative heating to compare to the TRMM PR latent heating. As in Schumacher et al. [2004], the heating fields were tapered at the boundaries (20°S and 20°N for latent heating and 15°S and 15°N for radiative heating) to avoid sharp discontinuities in the subtropics in the model results, which will be discussed in the next section. The upper level latent heating pattern mimics the tropical precipitation field since surface rainfall is one of its main constraints. The radiative heating of tropical upper level clouds is also centered on regions of large tropical rainfall but shows a weaker signal over the central Pacific ITCZ and a much weaker overall magnitude compared to the upper level latent heating. The radiative heating map in Figure 8 is consistent with the calculations in Sohn [1999] and with the 5–12 km column-integrated net radiative heating in Figure 9 of L'Ecuyer and McGarragh [2010].

Figure 8.

TRMM PR latent heating from 1998 to 2007 and ISCCP radiative heating for cloud regimes 1–4 (i.e., CD + CC + MIX + CI) from 1983 to 2008 at 250 hPa. Latitudinal boundaries are tapered to avoid discontinuities in subsequent GCM runs. The ISCCP radiative heating uses the look-up table based on the average profile from Manus and Darwin.

Figure 9.

(Left) The tropics-wide (15°S–15°N) average radiative heating (RH) using look-up tables based on the profiles from Manus and Darwin and the average of the two sites for ISCCP cloud regimes 1–4 (i.e., CD + CC + MIX + CI). (Right) Tropics-wide average latent heating (LH) derived from the TRMM PR (from 1998 to 2007) and tropics-wide LH + RH based on RH profiles from Manus, and Darwin and the average of the two sites.

[22] To quantify the radiative and latent heating contributions to the vertical structure of the total diabatic heating, Figure 9 shows the tropics-wide average (15°S–15°N) radiative and latent heating profiles (left panel) and the addition of the regimes 1–4 radiative heating to the TRMM PR latent heating profile (right panel) for Manus, Darwin, and the average of the two sites. The mean radiative heating has a maximum of 0.45 K/day between 200 and 250 hPa. The heating maximum in Darwin is ~0.25 K/day greater than Manus, giving a sense of potential uncertainty concerning the assumptions of the look-up table method. The latent heating maximum is 1.7 K/day at 450 hPa, so the radiative heating maxima account for 26% of the latent heating maximum (albeit at different heights). The latent and radiative heating profiles in Figure 9 are surprisingly similar to the western Pacific profiles above 500 hPa by Sherwood et al. [1994]

[23] It should be noted that while the time and space averaged radiative heating of the upper level convective clouds represented by regimes 1–4 (i.e., deep convective clouds, precipitating stratiform clouds, nonprecipitating anvil, and cirrus) is relatively small, these clouds still retain large local radiative heating values and the cloud radiative forcing of upper tropospheric cloud has been shown to destabilize the cloud layer [Gray and Jacobson, 1977; Webster and Stephens, 1980]. Tropical convective clouds also remain an important component in the water budget of tropical convective systems [Houze et al., 1980; Frederick and Schumacher, 2008].

[24] Figure 10 shows the longitudinally varying enhancement factor [e.g., Lin and Mapes, 2004], which is the ratio between column-integrated radiative heating for mid- and upper level (i.e., above 800 hPa) clouds and latent heating, for 15°S–15°N. The radiative heating of all upper level clouds enhances latent heating by about 21%. More specifically, the enhancement of the latent heating is contributed by the radiative heating below 200 hPa, which offsets the small radiative cooling at 150 hPa (Figure 9). This enhancement of the vertical gradient of the latent heating in the upper troposphere (i.e., above 450 hPa) is important for triggering the radiative–convective instability [Yu et al., 1998, Raymond, 2001]. The large heating and near zero cooling of the MIX and CI profiles contributes significantly to this large enhancement factor. Over East Africa and the western Indian Ocean (30–60°E), the radiative heating of upper level clouds enhances latent heating by about 42% due to the lack of deep convective rain, and thus latent heating, over that region. There could be some ambiguity in representing the East Africa and western Indian Ocean heating for the MIX category, which may result in the overestimation of the LW heating for the above two regions. However, the overall enhancement factors are consistent with previous studies, which found values ranging from 4% to 20% [Bergman and Hendon, 2000; Lebsock et al., 2010; Mekonnen and Rossow, 2011].

Figure 10.

Enhancement factor (i.e., the ratio between the column-integrated ISCCP radiative heating for regimes 1–4 and TRMM PR latent heating) averaged between 15°S and 15°N using look-up tables based on radiative heating profiles from Manus and Darwin and the average of the two sites.

4 Direct Impacts of Diabatic Heating on the Large-Scale Circulation

4.1 Model

[25] This study uses an idealized version of the National Center for Atmospheric Research (NCAR) Community Atmosphere Model Version 3 (CAM3) to investigate the directly forced response of the large-scale circulation to tropical diabatic heating (i.e., convective feedbacks are not allowed). CAM3 was run at T85 with a horizontal resolution of ~1.4° × 1.4°, 26 vertical levels and a rigid lid at about 3 hPa. The boundary conditions and physical processes of the model were replaced with linear damping of temperature and wind perturbations toward a prescribed basic state to isolate the radiative and latent heating impacts. In addition, to obtain the zonally asymmetric large-scale response to the applied forcing and to prevent interaction between the zonally asymmetric response and mean flow, the zonal mean of the dynamical response (temperature, wind, surface pressure, divergence, and vorticity) was not allowed to change with time and topography, and water vapor was removed from the model. Under these conditions, the model provides a nonlinear, zonally asymmetric, quasi-steady state atmospheric response with respect to the magnitude of input heating. The same processing and choice of coefficients for Newtonian cooling and Rayleigh friction can be found in the study by Schumacher et al. [2004]. All of the following numerical experiments are based on a motionless basic state. The thermal profile bears the tropical mean static stability. For the resting basic state, the model usually reaches equilibrium after 2 weeks [similar to Schumacher et al. 2004]. To obtain a more stable solution, results from the 20th day are used to represent the steady-state solution.

4.2 Dynamical Response

[26] Schumacher et al. [2004] showed that an elevated latent heating maximum resulting from an increase in stratiform rain fraction strengthened the upper level tropical circulation and that a geographically varying latent heating profile better indicated the observed westward tilted zonal circulation characteristics at upper levels. However, Schumacher et al. also showed that if very high stratiform rain fractions (e.g., 70%, which is rarely observed climatologically) were assumed, the resulting highly elevated latent heating profile caused an extreme circulation response. Thus, it is important to understand and quantify the factors that affect the upper level heating associated with tropical convective cloud systems to determine the convective system impacts on the large-scale circulation. Since streamfunction is related to vorticity (ζ = − ∇ 2ψ), which is the direct response of heating, and it is smoother than the wind field, streamfunction is usually used to represent the large-scale response of the forcing [Ting and Yu, 1998, Schumacher et al., 2004, Hurrell et al., 2006]. This study uses the zonal stream function anomaly to represent the horizontal large-scale dynamical response to diabatic heating.

[27] Figure 11a shows the simulated streamfunction anomalies at 250 hPa forced by the TRMM PR latent heating and indicates broad quadrapole patterns centered on areas of maxima heating (such as the Maritime Continent) seen in previous work [Hartmann et al., 1984, Nigam, 1994, Ting and Yu, 1998, Schumacher et al., 2004]. When the model is forced with the ISCCP radiative heating (Figure 11b), the quadrapole patterns are smaller in scale and shift to being centered over Africa and Panama. In addition, overall streamfunction anomalies are 10–15 times smaller (the maximum anomaly for latent heating is about 15 m2s-1 and about 1.5 m2s-1 for radiative heating). Because the radiative heating-driven quadrapole patterns are offset from the latent heating-driven patterns, the anomalous circulation at 250 hPa is slightly weakened when the model is forced with both latent heating and radiative heating (Figure 11c). Figure 11d shows the streamfunction anomaly from 10 years (1998–2007) of monthly NCEP reanalysis data. The response to the latent heating (Figure 11a) shows reasonable agreement with the NCEP reanalysis within the tropics (30°S–30°N) despite the idealized nature of the model set-up. The radiative heating response does not significantly improve the comparison. Recall, however, that radiative–convective interactions are not allowed in this modeling framework.

Figure 11.

Model streamfunction anomalies (in m2 s-1) at 250 hPa forced by (a) TRMM PR latent heating, (b) ISCCP radiative heating for regimes 1–4 (i.e., CD + CC + MIX + CI), (c) TRMM PR latent heating + ISCCP radiative heating for regimes 1–4, and (d) NCEP reanalysis during 1998–2007. Solid lines indicate positive streamfunction anomalies (clockwise flow). The ISCCP radiative heating uses the look-up table based on the average profile from Manus and Darwin.

[28] For a vertical perspective, Figures 12a–12c show the mean zonal mass flux from the TRMM PR latent heating forcing, the response to the ISCCP radiative heating associated with upper level clouds, and the response to the latent and radiative heating combined. Figure 12d is the zonal mass flux for the 10-year mean NCEP data set. The zonal mass flux [Newell et al. 1974] is defined as

display math

where a is the radius of the earth and ΔΦ is the width of the latitude strip (centered at the equator) in which the zonal mass flux is calculated. In this case, ΔΦ =20°. 〈u′〉 is the 10°S−10°N averaged eddy zonal wind.

Figure 12.

Vertical cross sections of simulated zonal mass flux along 20° strip (in kg s-1) centered on the equator forced by (a) TRMM PR latent heating, (b) ISCCP radiative heating for regimes 1–4 (i.e., CD + CC + MIX + CI), (c) TRMM PR latent heating + ISCCP radiative heating for regimes 1–4, and (d) from monthly NCEP reanalysis zonal winds based on 10-year mean (1998−2007). The ISCCP radiative heating uses the look-up table based on the average profile from Manus and Darwin.

[29] The response to the latent heating (Figure 12a) shows good agreement with the NCEP tropical zonal circulation (Figure 12d), which is consistent with Schumacher et al. [2004]. As was the case with the streamfunction anomalies in Figure 11, the radiative heating only response (Figure 12b) shows a similar pattern to the latent heating response but is much weaker and exhibits more elevated circulation centers. However, the addition of the radiative heating to the latent heating (Figure 12c) only marginally brings the latent heating response closer to NCEP.

[30] The structure of high clouds may be different across the tropics [Kubar et al. 2007], and the cloud heating profiles for each ISCCP regime show variations between Manus and Darwin. Therefore, we forced the model by ISCCP radiative heating for Manus and Darwin data separately to examine the sensitivity of the direct dynamical response to the different heating structures. The radiative heating induced large-scale circulation shows some differences in the horizontal and vertical circulation anomaly. However, the direct radiative heating induced dynamical impact is still small compared to latent heating induced one.

5 Conclusions

[31] Radiative heating profiles for the four tropical ISCCP cloud regimes composed of significant upper level clouds were derived using multi-year MMCR-MPL retrievals from the DOE ARM sites at Manus, Papua New Guinea and Darwin, Australia. Each regime shows maximum mean radiative heating near 12 km and maximum cooling near 15 km, with variations in the gradient of upper level heating dependent on the ISCCP cloud regime and location. For example, MCSs show the strongest radiative heating gradient while mixed convective clouds show the weakest gradient. In addition, upper level gradients are stronger over Darwin compared to Manus. The zonal distribution of radiative heating associated with deep convective systems maximizes over Africa, the Indian Ocean and West Pacific, and South America. However, the radiative heating from mixed convective cloud and cirrus extends further into the ocean ITCZ regions. The tropics-wide average radiative heating profile of the ISCCP upper level cloud regimes based on the combined Manus and Darwin data sets shows a maximum close to 0.45 K/day between 200 and 250 hPa, compared to a TRMM PR latent heating maximum of 1.7 K/day near 450 hPa. The column-integrated enhancement of latent heating by radiative heating processes above 800 hPa is about 20% and is most geographically pronounced in the Indian Ocean and Maritime Continent.

[32] The dynamical core of the NCAR CAM3.0 was utilized to simulate the tropical dynamical response to the diabatic heating of tropical convective systems. When run as a separate forcing, the climatological mean upper level radiative heating (i.e., the local heating weighted by the occurrence frequency) slightly weakened the latent heating model response at upper levels (cf. Figures 11a and 11c). The modest circulation changes suggest that the direct dynamical impact of radiative heating profiles associated with tropical convective clouds is not important in terms of the large-scale Walker and Hadley Circulation variations. Convective interactions appear to be required to make upper level clouds felt by the large-scale circulation. We note this is consistent with the conclusions of Bergman and Hendon [2000] who suggest that the indirect impact of cloud radiative forcing on the large-scale circulation (i.e., via radiative-convective interactions) might be more important than the direct cloud radiative forcing. However, the different patterns and signs in streamfuction anomaly driven by the direct radiative heating suggests that the height of the radiative heating maxima and the gradient of the radiative heating profile are important because they could suppress the latent heating-driven horizontal response at certain heights and regions of the tropical atmosphere (although weakly in this idealized modeling case).

[33] This study used an observationally based estimate of cloud radiative heating rather than relying on the full physics of the model to produce the upper level cloud radiative heating profiles. While the dynamical results were reasonable (suggesting that this radiative heating may be used to analyze model cloud radiative heating fields), interactions between the cloud radiative forcing, deep convection, and large-scale circulation cannot be examined in this framework. Upper level clouds can feed back on convection via destabilizing cloud-top cooling, stabilizing upper level warming, or differential heating between cloudy and clear regions [Stephens et al., 2008, and references therein]; thus, more quantitative feedback between observed tropics-wide upper level cloud radiative forcing and the large-scale circulation will have to be evaluated by a more interactive modeling framework. Finer grid resolution and separation of cloud type of the ISCCP data will also help further evaluation of the climate impact of the radiative heating of the tropical convective clouds.

Acknowledgments

[34] The authors thank Dr. Gang Hong for helpful discussions and Justin Stachnik for assistance with the ISCCP cloud regime data. The authors would also like to thank the anonymous reviewers, whose comments significantly strengthened this paper. This research was supported by the following grants: ARM-DOE Grant DE-FG02-06ER64174.

Advertisement