Longitudinal variation of tides in the MLT region: 2. Relative effects of solar radiative and latent heating
Abstract
[1] Part 2 of our study examines the relative importance of radiative heating and latent heating in accounting for vertically propagating tides that impose longitude variability on mesosphere-lower thermosphere (MLT) winds, temperatures, and densities. Our results are based upon numerical simulations using the Global-Scale Wave Model (GSWM) and new tidal heating rates derived from International Satellite Cloud Climatology Project (ISCCP) radiative fluxes (see part 1), Tropical Rainfall Measuring Mission (TRMM) latent heating profiles, and TRMM rainfall rates. Contrary to previous results and general perceptions, we demonstrate that radiative heating is more important than latent heating in accounting for MLT longitude variability due to tides although latent heating causes some large nonmigrating tidal oscillations such as DE3. Through comparison with TIMED SABER temperature measurements, the model results are shown to approximate many observed features of this longitude variability.
1. Introduction
1.1. Research Contexts
[2] In part 1 of this study [Zhang et al., 2010], we reported on Global-Scale Wave Model (GSWM) results obtained with net radiative heating rates derived from International Satellite Cloud Climatology Project (ISCCP) radiative fluxes (referred to as GSWM-ISCCP) and demonstrated that radiative heating in the troposphere can result in considerable longitudinal variability due to nonmigrating tides that propagate into the mesosphere-lower thermosphere (MLT) region. When they interfere with migrating tides, these nonmigrating tides account for more than half of the longitude tidal variation observed in SABER temperatures in the MLT region. The part 1 study is done in the absence of latent heating which is a considerable energy source for atmospheric tides. For instance, Forbes et al. [1997] analyzed 7 years of global cloud imagery (GCI) data, forced the GSWM using the annual-mean migrating tidal heating rates due to latent heat release, and reported that maximum migrating tidal amplitudes of order 10–20 m/s and 5–8 K resulted in the MLT as compared with the total tidal fields of order 20–60 m/s and 10–40 K known to characterize this region. Hagan et al. [1997] demonstrated that a latent heat release tidal source must be pursued to account fully for diurnal tidal variability. Hagan and Forbes [2002] reported that “the aggregate effects of the latent heating source vary from month to month and produce measurable diurnal amplitude variations (∼10–20m/s and 5–15°K) at select low- and middle-latitude locations in the upper atmosphere.” Additional research history on both radiative and latent heating effects on atmospheric tidal oscillations was introduced in part 1 of this study. In part 2, we drive the GSWM using updated latent heating rates in addition to radiative heating to verify these conclusions and to better understand the role of latent heating compared with that of radiative heating in achieving tidal longitude variability in the MLT region. We refer to the combined response to these sources as “GSWM-09” results. In the rest of this section, we introduce the basics of latent heat forcing of tides. In section 2, we review the history of using observed rainfall rates to infer latent heating rates, especially with respect to the study of atmospheric tides. This provides adequate background to understand how and why we use available data products to develop the latent heating algorithm used in the current study. In section 3, we present and interpret the new GSWM-09 results driven by these updated heating rates and compare them with temperature tidal signatures observed by SABER. Details of the GSWM model and SABER observations can be found in part 1 of this study.
1.2. Latent Heat Energy
[3] The energy associated with latent heat is carried in the water vapor that evaporated from the Earth's surface where radiative energy is absorbed. Carried upward by convection, the water vapor condenses in convective or stratiform clouds, so latent heat is released and the atmosphere is heated while at the same time precipitation may be induced. According to the diagram of Earth's energy budget credited to NASA's Earth Radiation Budget Experiment (ERBE, http://asd- www.larc.nasa.gov/erbe/components2.gif), the amount of total latent heat energy is about 70% of that of radiative heat in the troposphere. The energy going into tidal excitation is a fraction of the total energy of either radiative heating or latent heating and can be further divided into migrating (i.e., Sun-synchronous) and nonmigrating tidal energy. The nonmigrating tidal energy depends on the longitude variation of tidal heating. With more longitudinal variability, one might presuppose that latent heating may have a greater nonmigrating portion than that of radiative heating if we think about the factors that affect the latent heating: First, about half of the radiative heating is directly related to Earth's topography, while almost all the latent heating is directly related to the land-sea difference and albedo distribution, etc. of Earth's surface. Second, the tidal phase of latent heating (i.e., local time of maxima) is far from globally uniform as compared with that of radiative heating. In contrast to the well-scheduled processes of shortwave and longwave radiative transfer, the time to develop a convective or stratiform precipitation system generally depends on where the precipitation system is developed, on land or on sea. Since its amplitude and its phase are zonally varying, we do not expect the latent heating response to be dominantly characterized by migrating components as is the case for radiative heating. The following study on Tropical Rainfall Measuring Mission (TRMM) latent heating rates quantitatively confirms our general understanding of latent heat energy in the above context.
2. From Rainfall Rates to Latent Heating Profiles
[4] The objective of this paper is to use observation-based latent heating rates to drive GSWM to study the influence of this source mechanism on the MLT tidal components. However, the latent heating rate measurements are not sufficiently time-resolved to extract the tidal components. On the other hand, measured rainfall rates, which are a proxy for latent heating, are sufficiently time-resolved to study the tidal components. In sections 2.1–2.4 we describe our technique of combining measurements of daily-averaged latent heating rates and local time-resolved rainfall rates to drive the GSWM and to study the MLT tidal response to latent heating. In section 2.1, we review the Global Cloud Imagery (GCI) rainfall rates and latent heating rates used by GSWM-02; in section 2.2, we introduce TRMM rainfall rates and compare them with GCI rainfall rates; in section 2.3, we introduce the product of TRMM latent heating profiles by the Mesoscale Atmospheric Processes Branch at NASA's Goddard Space Flight Center; and in section 2.4, we introduce the method by which we combine TRMM tidal rainfall rates with TRMM daily latent heating profiles to get TRMM tidal latent heating profiles. Then we show the features of the new TRMM latent heating and compare the tidal spectra it generates with those of ISCCP net radiative heating.
2.1. GCI Rainfall Rate/Latent Heating Rate
[5] As mentioned in section 1, latent heat is released when water changes phase between vapor and small liquid or frozen water particles. Like other energy processes, latent heating cannot be directly detected by current remote sensing or in situ instruments. Since latent heating is closely related to clouds and precipitation, current retrieval schemes depend heavily on cloud-resolving models (CRM) while rainfall rate has been a proxy of latent heating for a long time [e.g., Tao et al., 2006].
[6] For the purpose of studying tides and other global-scale waves, the global/zonal coverage for latent heating or rainfall rates are best obtained by satellite measurements since ground-based rainfall observations are still too sparse. The latent heating rates used in GSWM-02 [Hagan and Forbes, 2002, 2003] were derived from cloud top temperatures observed by ISCCP's Global Cloud Imagery (GCI) project. The GCI synoptic images of the global cloud field were created from infrared measurements taken aboard four geostationary and two polar-orbiting platforms simultaneously observing the Earth. The rainfall rates and hence the latent heating rates derived from the composite GCI as described by Forbes et al. [1997] have 2.5° horizontal resolution between ±40° latitude, good enough to provide diurnal and semidiurnal latent heating rates as needed by the GSWM when the globally uniform vertical heating profile shape is specified [Forbes et al., 1997]. Figure 1 shows the diurnal and semidiurnal GCI rainfall rate distributions for the 7 year climatological January averaged over 1988–1994. We see that the major rainfall cells are all sea breeze convergence zones [i.e., Simpson, 2007] with the South America and Africa Zones being fairly strong and the South Asia and South Pacific Convergence Zone (SPCZ) being weaker. The Intertropical Convergence Zone (ITCZ) is not so obvious. Note that there is missing data in the meridional band over the Indian Ocean, because the geostationary satellite over that region was not operative.
2.2. TRMM Rainfall Rates
[7] The TRMM project provides tropical/subtropical three-hourly rainfall rate data and daily latent heating profiles, both with fine horizontal resolution. The orbit of the TRMM satellite has an inclination of 35° with respect to the Equator. This orbit provides extensive coverage in the tropics and allows each location in the same zonal circle to be covered at almost the same local time each day on either ascending or descending legs, but the local time precesses day by day at the rate of 24 hours/46 days/leg. This kind of sampling enables analysis of the diurnal cycle of precipitation but the full diurnal cycle of local time coverage takes up to 46 days of observation depending on latitude within ±37°. TRMM Microwave Imager (TMI) and Precipitation Radar (PR) are two of the six instruments onboard the TRMM satellite. TMI is a multichannel passive microwave radiometer that provides information on the integrated column precipitation content, cloud liquid water, cloud ice, rain intensity, and rainfall types (e.g., stratiform or convective). PR, the first of its kind in space, is an electronically scanning radar that measures the 3-D rainfall distribution over both land and ocean, and defines the layer depth of the precipitation.
[8] Measurements from TRMM/TMI microwave radiometer are physically most closely related to precipitation and are thus considered to provide more accurate rainfall measurements. However, to get synoptic maps of tropical rainfall of the type needed to generate heating rate inputs to the GSWM, TRMM/TMI measurements still need to be combined with those from infrared sensors flying on the international constellation of geosynchronous satellites (geo-IR). The accurate but sparse TMI estimates are used to provide spatially and temporally varying calibration for the plentiful but imprecise geo-IR precipitation estimates [Huffman et al., 1997, 2001]. Three precipitation products are provided: TRMM-calibrated microwave estimates (three-hourly accumulations) by merging all available microwave radiometers; a geosynchronous infrared estimate which is calibrated by the merged-microwave data (hourly estimates); and a combination of the first two fields (three-hourly accumulations). The TRMM 3B42 product we use in this study is the third of these and is shown in Figure 2 with the diurnal and semidiurnal TRMM rainfall rate distributions similar to those in Figure 1 but for the 2002–2006 climatological average in January. Comparing TRMM with GCI, we see that all the major sea breeze convergence rainfall cells including the SPCZ measured by TRMM have similar amplitudes, while the GCI amplitudes in SPCZ are weaker and in African and South American continents are much stronger. Also the ITCZ is obvious in TRMM though weaker than those of the sea breeze convergence cells, while there is no obvious ITCZ in GCI.
[9] Comparing Figure 1 and Figure 2 further, we see a global GCI phase shift to later times as compared with TRMM. While most of the TRMM diurnal phases over ocean are in the morning, most of the GCI diurnal phases are in the afternoon with phases over continents in late night.
[10] Figure 3 is based on TRMM 3G68 rainfall rate products which are derived purely from TRMM instruments. The TRMM monthly or climate rainfall products, referred to as the level 3 products, are derived primarily from accumulations of the TRMM level 2 products, or instantaneous rainfall estimates corresponding to the original satellite snapshot views. 3G68 is an hourly gridded product containing TRMM 2A12 [Kummerow et al., 2001], 2A25 [Iguchi et al., 2000], and 2B31 [Haddad et al., 1997a, 1997b] rain estimates. The 3G68 product includes 24 hourly 0.5 × 0.5 degree grids into a single daily file. Although there is no synoptic map, TRMM instruments can provide up-to-46-day mean continuous diurnal local time coverage with fine latitude resolution (less than 1°) and relatively sparse longitude resolution (∼24°). Figure 3 depicts multiyear-mean (January 1998 to December 2007) rainfall rates from the Web site of Goddard Earth Sciences Data and Information Services Center at http://daac.gsfc.nasa.gov/precipitation/trmm_apps/trmm_diurnal_3g68.shtml. All the plots show that the rainfall rates over the oceans peak in the morning while the rainfall rates over the land peak in the afternoon. As a result, the zonal mean rainfall rates have two peaks in local time with similar magnitudes. This means that the migrating diurnal component is almost absent from the zonal mean and the migrating semidiurnal component is dominant. This suggests rainfall rates and hence latent heating rates do not have a significant migrating diurnal component but a relatively strong migrating semidiurnal component and should be evident in the TRMM latent heating rate spectra. We will return to this discussion later in the report.
[11] The phenomenon that rainfall over land peaks in the afternoon can be explained by the fact that most of the tropical major rainfall cells on land are caused by sea breeze convergences which are usually well developed after the land is heated by radiative heating after noon. On the other hand, storms over the ocean are generally weaker and caused by land breeze convergences that are usually well developed after the land is cooled down after midnight [e.g., Simpson, 2007]. The phases in the southeast Pacific Ocean and South America in Figure 2 (top right) are consistent with this interpretation. Physically, the rainfall rates provided by TRMM instruments are more accurate, especially those by TRMM Precipitation Radar which has the highest spatial resolution (4 km), the best vertical resolution (80 levels), and the most directly measured surface rainfall estimates. The consistency between TRMM 3G68 rainfall rates in Figure 3 which are purely by TRMM measurements and the TRMM 3B42 multisatellite rainfall rates in Figure 2 with respect to land-sea tidal phases means that the TRMM 3B42 rainfall rates employed by this study will lead to an improvement over the GCI rainfall rates which formed the basis for GSWM-02 latent heating rates. We will enhance this conclusion in the discussion of Figure 7.
2.3. TRMM CSH Latent Heating Profiles
[12] As mentioned in section 2.2, TRMM 3B42 rainfall rates are improved over the GCI rainfall rates that GSWM-02 employed. We want to use TRMM 3B42 rainfall rates because we really want latent heating rates with local time dependence. As detailed below, the latent heating rate data products unfortunately can only be used to construct daily mean profiles. However, we will combine the daily-average latent heating rates with TRMM 3B42 local time-resolved rainfall rates to obtain latent heating rate profiles as a function of local time. In the remainder of this section, we will demonstrate that (1) the TRMM PCSHb (refer to the following paragraph) latent heating rate data does not have the same local time coverage as the TRMM satellite and (2) our method of combining the TRMM PCSHb daily-average latent heating rate data based on TRMM 3G68 with the local time-dependent 3B42 rainfall rate data will lead to significant differences in latent heating rates compared to earlier GSWM model studies.
[13] Applying their Convective-Stratiform Heating (CSH) algorithm [Tao et al., 2006] to the TRMM 3G68 gridded PR rainfall product and a separate data set containing TRMM PR echo top height data in order to screen for shallow heating, the team at the Mesoscale Atmospheric Processes Branch of NASA's Goddard Space Flight Center has constructed daily latent heating profiles (PCSHb version) extending from 1 January 1998 to 31 December 2007 at 0.5° horizontal resolution and 1 km vertical resolution based on the TRMM 3G68 data set which contains gridded values of the total pixels, rainy pixels, mean rain rate, and percentage of convective rain from the PR algorithm and the TMI algorithm [Kummerow et al., 2000]. One might think, considering the local time precession of the TRMM satellite, that the local time variation of latent heating profiles might be obtainable from these data as well. (That is, each day, the data at the same zonal circle were measured at a certain local time and the next day at a different local time. So every precession cycle, e.g., 39 days, the local time was shifted 24 hours. As a result, a diurnal local time cycle is covered by every precession cycle.) In an attempt to retrieve the local time information which the PCSHb latent heating profiles may contain, we binned the local time of TRMM satellite over the course of a day into the 0.5° × 0.5° grids in the PCSHb version of latent heating rates profiles. Some of the grids overpassed by both ascending and descending legs of the satellite may include data from both legs at different local times. These data were excluded from our analysis since we do not know which local time of the two legs was used by the PCSHb product. Now the PCSHb latent heating profiles in the remaining grids/bins each has a different local time for different days. Then we bin 39 days of TRMM PCSHb data into twenty-four 1 hour local time grids for every 3° latitude × 24° longitude bin, sliding forward one day at a time. Recall that the TRMM satellite's local time precession rate is 24 hour/46-day/leg. At the equator, where the local times of the two legs are 12 hours apart, we need only 23 days of data to have 24 hours local time coverage. At 35° latitudes, the local times of the two legs are much closer, so we need 39 days of data to have 24 hours of local time coverage. We invoked the 39 day window needed at higher latitudes in order to have a uniform UT window length for all latitudes within ±37°. By fitting over the 24°x15 longitudes and the 24 hours in local time, we have 39 day mean tidal components of TRMM PCSHb latent heating profiles every day within ±37° latitude. Unfortunately, the resulting tidal components are noisy and do not have reasonable spatial continuity. So we conclude that the local time differentiation in the PCSHb products was ignored in the adopted algorithm, and we use the PCSHb products only to obtain the daily vertical distribution of the latent heating while using TRMM 3B42 rainfall rate product to obtain the local time coverage to retrieve tidal components. The algorithm to do this is introduced in section 2.4.
[14] We use Figure 4 to show that TRMM latent heating can lead to significant differences from the earlier GSWM-02 model studies. Figures 4 (top) and 4 (middle) show the distribution of daily mean latent heating rates in the tropical/subtropical region at the peak altitude 6.75 km from the TRMM-CSH PCSHb product compared with that of the daily mean rainfall rate on the surface from the TRMM 3B42 product in January. We see that the magnitude from south Asia to the SPCZ area is larger in Figures 4 (top) and 4 (middle) compared with the amplitude distributions of tidal components in Figure 1 or even Figure 2. Figure 4 (bottom) shows the distribution of ratio of latent heating rate at the peak altitude to rainfall rate on the ground. We see this ratio varies globally while the ratio used to convert rainfall rate to latent heating rate used in the GSWM-02 is a constant,.44 K/mm. This difference, in addition to the differences between GCI rainfall rates and TRMM 3B42 rainfall rates in both amplitudes and phases, guarantee a difference between TRMM latent heating and the latent heating rates used in the GSWM-02. Earlier in this report, we discussed that this difference is an improvement based upon the physical interpretation of the rainfall phases. In section 2.4, we strengthen this claim by showing tidal responses in the upper atmosphere due to TRMM latent heating that are closer to the SABER observations than the GSWM-02 outputs based upon GCI-based latent heating.
2.4. Combined TRMM Latent Heating Profiles: New Latent Heating Input
[15] Now we know what we need are TRMM CSH PCSHb daily latent heating profiles and TRMM 3B42 three-hourly rainfall rate data. By combining these two data sets, we can generate latent heating profiles that do vary with local time and geographic location. Basically, the 5 year (i.e., 2002–2006) climatological monthly mean TRMM 3B42 rainfall rates within ±37° latitudes are decomposed into diurnal and semidiurnal amplitudes and phases, then the amplitudes are normalized by the daily mean. Next, the normalized amplitudes are multiplied by the 5 year climatological monthly mean TRMM CSH PCSHb daily latent heating profiles. The resultant January latent heating distributions are shown in Figure 5 in altitude-averaged forms in order to compare to the distributions shown in Figure 2. We see that the phases of the combined latent heating follow those of the TRMM 3B42 rainfall rates and the amplitudes are the combination of TRMM rainfall rates and TRMM-CSH latent heating.
[16] Figure 6 shows the spectra of the combined TRMM latent heating compared with that of ISCCP net radiative heating (see part 1) in September. In order to better see the nonmigrating components, the migrating radiative heating components are omitted since they are very dominant. The migrating components (DW1 and SW2) are included in the latent heating spectra. As anticipated (section 2.2), the diurnal migrating component is small but the migrating semidiurnal component is significant. Considering that both amplitudes and phases of latent heating vary significantly with longitude, it is not surprising that the latent heating nonmigrating signatures are stronger than those of radiative heating with the DE3, D0, DW2, DW3 and DW5 nonmigrating components being obvious in both the radiative and latent heating spectra. There are also some similarities in the radiative and latent heating semidiurnal spectra with SE2 being a very consistent one. The similarities of the two spectra suggest that both latent heating and radiative heating are affected by the same topography although in different ways. One of the differences between the radiative and latent heating spectra is that latent heating is confined within a narrower latitude band around the equator. As discussed in part 1 of this study, the atmosphere in the mean state works as a filter to tides and other waves. Below we will discuss which components in the tropospheric heating spectra can propagate to the MLT region by exciting the full heating spectrum in the GSWM and examining the MLT response.
[17] Before showing GSWM results, we illustrate vertical heating profiles for major TRMM latent heating tidal components. Figure 7 shows the altitude-latitude distributions of the TRMM latent heating amplitudes for DE3, DW1 and SW2 components in September. We see that the DE3 is the most significant component, and that SW2 is more significant than DW1. The latter is consistent with Figure 3 which verifies the technique of combining TRMM/TMI with geo-IR to arrive at the 3B42 data. Both DE3 and SW2 have smooth and consistent amplitude structures as a function of latitude while DW1 does not. The vertical profiles of DE3 and SW2 peak at 5–8 km and in the northern hemisphere during fall equinox.
[18] The results illustrated in Figure 6 are not quite consistent with the GCI spectra reported by Forbes et al. [1997], which were characterized by a strong DW1 component resulting in annual-average migrating diurnal temperature amplitudes of the order 5–10 K in the MLT region. The differences in the TRMM and GCI heating rates are primarily caused by differences in the observational techniques. The GCI rainfall rates were generally overestimated based on IR brightness temperature. That is, the GCI diurnal migrating rainfall rates were most likely aliased with the cloud top DW1 temperature perturbations. In contrast, the rainfall rates obtained from TRMM are measured by microwave radiometer which provides measurements that are physically related to the precipitation content in convective systems.
[19] With both the radiative and latent heating rates in hand, we run the GSWM first using TRMM latent heating only, and then we run the GSWM with both radiative heating and latent heating combined (the latter being GSWM-09). By comparing the tidal responses due to latent heating with those due to radiative (described in part 1, GSWM-ISCCP) and total heating, we can ascertain them due to each heat source, as well as the aggregate tidal longitudinal variability due to the combined effects of both heat sources. These results are described in section 3.
3. Results
[20] We run the GSWM with latent heating and total (latent plus radiative) heating, respectively, and refer to the former as GSWM/TRMM and the latter as GSWM-09. Consistent with the results driven by radiative heating, the excitations from O3 and O2 [Hagan et al., 1999] in the stratosphere and mesosphere are included (part 1 of this study) and all parameters are kept the same except for background (i.e., zonal mean) winds/temperature and troposphere heat sources.
[21] Figure 8 shows the GSWM-TRMM and GSWM-09 DW1 temperature amplitude and phase results compared with those observed by SABER in March when DW1 has its seasonal peak amplitude [Forbes et al., 2008]. As expected, we see that the GSWM/TRMM DW1 does not achieve significant amplitudes. However, it does constructively interfere with the GSWM/ISCCP DW1 component (Figure 6 of part 1 of this study). The GSWM-09 amplitudes are larger and closer to the SABER observations. Both have the distinctive pattern of the symmetric propagating component with maxima of ∼22K at 70–110 km, although the observed two-layer altitude structure is still not simulated as discussed in part 1 of this study. The GSWM-09 amplitudes in the stratosphere-mesosphere are generally smaller than the SABER observations. This indicates that GSWM-09 may underestimate the ozone heating that forces a trapped (in situ) response in the stratosphere. Figures 8g and 8h illustrate GSWM-02 DW1 amplitudes that are about 1.5 times larger than the SABER observations due to overestimates in the DW1 GCI latent heating based on cloud top brightness temperature. As mentioned earlier, this may be attributable to an aliasing relationship between rainfall rate and thus latent heating rate with the cloud top brightness temperature.
[22] Figure 9 shows a comparison similar to Figure 8 but for the SW2 tide in May. We see that the GSWM/TRMM SW2 amplitudes in the MLT region have the same dominant antisymmetric behavior attributable to the (2,3) mode as that of SABER. A comparison with Figure 7 in part 1 of this study suggests that SW2 latent and radiative heating responses constructively interfere, since the GSWM-09 amplitudes are larger than either the GSWM/ISCCP or GSWM/TRMM results. The GSWM-09 SW2 amplitudes do agree better with SABER than the GSWM/ISCCP results except the amplitude peak at 40°S/120 km. Similar to DW1, GSWM-09's SW2 amplitudes in the stratosphere-mesosphere region are also smaller than those measured by SABER. Besides in the stratosphere-mesosphere region, the SW2 response in the MLT may be strongly influenced by ozone heating, plus by in situ chemical heating, and thus the precise phasing between these contributions and the responses to the tropospheric heating may not be correctly captured in our simulation. In this respect, modeling SW2 is more challenging than modeling DW1.
[23] The nonmigrating tidal temperature spectra illustrated in Figure 10 correspond to GSWM-09 (Figure 10, top), GSWM-TRMM (Figure 10, middle), and SABER (Figure 10, bottom) results at 95 km. We see the effects of atmospheric filtering and characteristic dissipation when we compare these results with the heating rate spectra illustrated in Figure 6. The GSWM and SABER DE3 are relatively much stronger at 95 km than the heat sources, while at 95 km D0 is relatively weaker and the DW5 is missing in GSWM-09 responses at 95 km although it is still evident in the SABER spectra and in the GSWM spectra at other altitudes (not shown). In the semidiurnal spectra, we see that the GSWM-09 SW4 and SW6 at 95 km are comparatively stronger than their respective heating rates, while SE2 is relatively weaker and SW1 is missing. In contrast, SABER shows a SW1 and a relatively stronger SW3, which dominate over the SW4 and SW6 temperature responses. This may reflect the absence of other SW1 and SW3 forcing mechanisms in the GSWM, such as nonlinear wave-wave interactions [Angelats i Coll and Forbes, 2002; Forbes and Wu, 2006). Notably, the 95 km GSWM/TRMM and GSWM-09 temperature spectra are much more similar than TRMM latent and ISCCP radiative heating spectra both in the relative strength and latitudinal distribution of each wave component. The MLT tides due to both latent and radiative heating are similarly spread out in latitude although the sources' latitudinal distributions are not so alike. The maximum values of these heating rates suggest that latent heating has a more dominant influence on nonmigrating tidal responses in the MLT region due to its greater contribution to nonmigrating tidal forcing in the troposphere, at least in September.
[24] Figure 11 is analogous to Figure 8 except for the DE3 in September and with the altitude range extended to 200 km. Comparison with Figure 9 in part 1 of this study suggests that the DE3 component excited by latent heating is more significant than that by radiative heating. This is also evident in the comparisons between the spectra of tidal temperature at 95 km and the spectra of the heat sources. In addition, we see that the DE3 response due to both the TRMM and ISCCP heat sources combine to produce temperature perturbations that are very close to the SABER observations. The GSWM-09 and SABER DE3 temperatures have similar altitude-latitude distributions below 120 km, and the magnitude of the peak amplitudes are very close although the GSWM-09 response is shifted slightly southward and is more confined than the SABER peak. The GSWM-09 and SABER phase patterns are also consistent. The GSWM-02 DE3 temperature amplitudes illustrated in Figures 11g and 11h depict a response that is about 1.5 times larger than that of SABER due to the same reason we discussed above for the DW1 response. That is, the GCI latent heating may be overestimated due to aliasing between rainfall rate and IR temperature measurements. The GSWM-02 peak is a little bit too high at 110 km and is not as obvious and detached as observed by SABER and modeled by GSWM-09. On the other hand, the latitudinal structure of the GSWM-02 DE3 amplitude between 70–110 km is closer to that observed by SABER. Both the GSWM-02 and GSWM-09 results show a primary DE3 peak above the SABER domain around 130–140 km over the equator. The GSWM-02 peak is comparatively higher in altitude.
[25] Figure 12 shows amplitude comparisons similar to those in Figure 11 except for the DE2 in January and July, which are comparable to Figure 10 in part 1 of this study. We see that, unlike the behavior of the DE3, the DE2 latent heating response is not as significant as the radiative response and the temperatures driven by the two heat sources are not in phase. The combined GSWM-09 results are not as large as the GSWM/ISCCP response, although the latitude structure of the amplitude patterns is similar for both responses. In July, the GSWM-09 results are closer to SABER's except that the peak is higher and smaller.
[26] Figure 13 shows comparisons similar to Figure 12, except for DW2 and SW3 in September. We see that both of these tidal components are about equally excited by tropospheric latent heating and radiative heating by comparing these results with those illustrated in Figure 11 in part 1 of this study. GSWM-09 temperatures driven by both DW2 sources achieve a stronger DW2 in the MLT region that has a similar latitude structure as the observed amplitudes but with a comparatively weaker maximum and at a higher altitude. A comparison of the SW3 amplitudes suggests that the responses attributable to the two energy sources are not phase coherent (i.e., they destructively interfere).
[27] We now turn to a discussion of the aggregate effects of all the different wave components superimposed upon each other. Figures 14a–14d show the latitude versus longitude distribution of diurnal and semidiurnal amplitudes of ISCCP radiative heating (Figures 14a and 14b) and of TRMM latent heating (Figures 14c and 14c) at 6.75 km in September. Diurnal amplitudes are in the same contour levels for both radiative heating and latent heating while semidiurnal amplitudes of both heat sources are at half of the levels for diurnal amplitudes. We see that the diurnal peak amplitude of latent heating is about 60% of that of radiative heating while the semidiurnal peak amplitude of latent heating is about 1.2 times of that radiative heating. These two ratios are generally higher than the average ratios of latent heat tidal energy over radiative heat tidal energy since we are talking about the peak amplitudes. Latent heat energy peaks around 5–8 km in altitude and concentrates in a relatively narrower latitudinal band. Both diurnal and semidiurnal distributions of radiative heating have background amplitudes which indicate the existence of migrating tides. In terms of the interference effects [Forbes et al., 2003], with the same longitude variation and a much weaker migrating tide, the diurnal nonmigrating latent heating rate must be much larger than the radiative heating rate in September. This is confirmed by examining the tidal spectra (compare Figure 9 in part 1 and Figure 10 in this paper). The semidiurnal heating amplitudes generally follow the corresponding diurnal patterns while the magnitude of semidiurnal radiative heating is about one third of the diurnal rates and the magnitude of semidiurnal latent heating is about two thirds of the diurnal rates. We also see that, although the amplitude distributions of latent heating are different from those of radiative heating, they share considerable similarities. This is not surprising since physically both of these heating sources depend on the clouds distribution and radiative energy absorbed by Earth's surface.
[28] Figures 14e–14j show the GSWM tidal responses due to the combined effects of tidal components from E6 to W6 for the diurnal and semidiurnal tides at 95 km in September. Figures 14e and 14f illustrate GSWM-09 temperature perturbations, Figures 14g and 14h show the GSWM/TRMM response, and Figures 14i and 14j illustrate SABER observations. We compare Figures 14e and 14f to Figures 14i and 14j to develop a discussion similar to the one that we conducted for Figure 12 in part 1 of this study. However, in this case we compare the combined radiative and latent heat forcing and responses with SABER observations. The longitudinal variation of the GSWM-09 diurnal temperature amplitude at 95 km in the tropical/subtropical region is very close to that observed by SABER, not only in salient structure but also in magnitude with maximum about 25K. The similarity comes from GSWM-09's good simulation of DW1 and DE3 components that dominate the longitude structure. Both GSWM-09 and SABER amplitudes have subtropical bands between 30–40°s. This reflects the consistency in DW1. But the detailed differences within the subtropical bands indicate the not-as-good GSWM-09 simulation of other diurnal nonmigrating tides and the relative phasing between these to produce the illustrated interference pattern. The corresponding comparison for semidiurnal is not as good. The results in Figures 14f, 14h, and 14j for the semidiurnal tide provide evidence of the effects of filtering of Hough modes since the latitude distributions of amplitudes in these three panels are totally different from the corresponding diurnal ones although the latitude patterns of both ISCCP and TRMM semidiurnal heating are similar to those of diurnal heating as shown in Figures 14a–14d.
[29] Note also that in Figures 14g and 14h, which illustrate the GSWM/TRMM results, the contour levels are half of those for the GSWM-09 and SABER results. They are also half the levels for GSWM-ISCCP. Although latent heating in September generally excites more significant diurnal nonmigrating components in the MLT region than radiative heating, we see that the diurnal longitude variation at 95 km excited by latent heating alone is much less than that excited by radiative heating, and thus less than half of that excited by the combined sources since the latent heating and radiative heating have positive interference for both DE3 and DW1 components that dominate the longitude structure. Further, the diurnal longitudinal peak structure does not quite agree with that observed by SABER. The reason is that the nonmigrating components excited by latent heating must interfere with the large migrating tidal component due to radiative heating in order to yield the large longitudinal variability that is observed. In the real world, tidal components excited by all heating sources interfere with each other. Typically and primarily, the DE3 interferes with the DW1 in GSWM-09 diurnal simulation (Figures 14e and 14f) and the SABER observation (Figures 14i and 14j). The GSWM/TRMM longitude variability (Figures 14g and 14h) is relatively small because the tidal components are predominantly driven by latent heating and have a small DW1 component.
[30] Figure 15 summarizes the relative effects of latent heating and radiative heating, including migrating and nonmigrating components for both diurnal and semidiurnal tides, in the MLT region as a function of month. Figure 15 (top) depicts the heat sources averaged up to 15 km altitude, while Figure 15 (bottom) shows the temperature responses averaged over the MLT region from 80 to 120 km and between ±30° latitude. Note that the plotted curves are amplitude ratios and the total means the scalar sum of amplitudes.
[31] Figure 15 (top left) shows that the migrating tidal heating rate is primarily attributable to the radiative heat source. The diurnal migrating component (DW1) dominates the diurnal tidal response (from DE6 to DW6), while the semidiurnal migrating component (SW2) dominates the semidiurnal response (from SE6 to SW6). Note also that the DW1 is more dominant than SW2. On the other hand, the diurnal migrating latent heat source is just about equivalent to the remainder of the diurnal heating rates (i.e., all the DE6 to DW6 components) while the semidiurnal migrating heating rate is 2–3 times the average, assuming that the average ratio for each wave component is about 8%.
[32] From Figure 15 (top right), we see that both diurnal and semidiurnal nonmigrating latent heating rates are comparable and smaller than radiative heating rates. Nonmigrating latent heat energy is about 60–70% of nonmigrating radiative heat energy and migrating latent heating rates are less than 20% of migrating radiative heating rates with diurnal migrating latent heating at just 5% of diurnal migrating radiative rates. What needs to be pointed out is that the peak amplitude of the DE3 (i.e., the dominant nonmigrating latent heating component) is more than that driven by radiative heating. But, as seen in the heating spectra (Figure 6), the nonmigrating components of radiative heating have more peaks which are spread out between ±30° latitudes. If we concentrate on the Equator instead of the ±30° average, the blue curves in Figure 15 (top right) are much larger (∼150%).
[33] The MLT tidal temperature excited by radiative heating has much more seasonal variation than their sources in the troposphere. This is evidence of the seasonal variation in mean wind filtering and dissipation effects. The two solid red curves in Figure 15 (left) show that the diurnal radiative migrating tidal response is very suppressed around the solstices as are the migrating semidiurnal tides (the two dotted red curves in Figure 15 (left)). The two solid blue curves show that the seasonal variation of the migrating diurnal latent heating is seasonally out-of-phase to that of the migrating diurnal radiative heating. Even though the diurnal latent heat source is smaller than the semidiurnal component, it propagates up relatively more efficiently, especially during solstices, resulting in a strong seasonal variation in MLT diurnal temperature. The solid and dotted black curves in Figure 15 (bottom left) represent the diurnal and semidiurnal migrating parts of the SABER tidal temperature oscillations, respectively. Both the semiannual and annual variations in the SABER diurnal response are similar to those of the dominant diurnal migrating radiative energy represented by the red curve in Figure 15 (top left), and are approximately parallel to the average of the red and blue solid curves in Figure 15 (bottom left). But this relationship does not hold for the semidiurnal component. This suggests that the semidiurnal migrating component in the MLT region has energy sources other than radiative and latent heat energy in the troposphere. Some important ones are the absorption of ultraviolet radiation in the stratosphere and mesosphere [e.g., Hagan, 1996], in situ EUV heating and exothermic chemical heating.
[34] Note that the ratios illustrated in Figure 15 (left) are the migrating over the total (migrating + nonmigrating) heating rates. There is more radiative heat energy forcing the migrating tides and resulting in dominant migrating response in the MLT region. Conversely, the predominant latent heat forcing is in the nonmigrating component and this results in significant nonmigrating tidal temperatures in the MLT region.
[35] Comparisons between Figures 15 (top right) and 15 (bottom right) show that the ratios of latent/radiative temperature responses (Figure 15, bottom right) are much larger than the latent/radiative heating ratios (Figure 15, top right), suggesting that zonal mean wind filtering does not significantly affect the tidal components forced by latent heating. This is not surprising, since these components are largely confined to a narrow latitudinal band around the equator where zonal mean winds and their gradients are weak. As a result, in the MLT region, the total diurnal nonmigrating tidal oscillation excited by latent heating is comparable to that excited by radiative heating while the total semidiurnal nonmigrating oscillation excited by latent heating is generally a little bit less than that excited by radiative heating. During equinox, the diurnal nonmigrating tidal oscillation excited by latent heating is even stronger than that excited by radiative heating. Consistent with the behavior of their sources, the migrating tidal temperatures in the MLT region resulting from latent heating are much smaller than those resulting from radiative heating, except in May to July when the diurnal migrating oscillation is small. Although latent heating is a significant source of the DE3 oscillation in September (comparison between Figure 11 and Figure 6 in part 1 of this study), the MLT nonmigrating tidal temperature oscillations attributable to total nonmigrating (i.e., W6 to E6) latent heating (Figure 15) is comparable to the response excited by radiative heating during all months.
[36] As a highlight of this study, we present the latitude-seasonal variation of DE3 amplitudes in Figure 16 as a result of our demonstrated ability to model this component with the GSWM. We see that GSWM-09 DE3's latitude-seasonal variation is similar to that of SABER with peak amplitude in August-September and in the equatorial region with a bias toward the Southern Hemisphere at 110 km. This is a significant improvement over the SABER/GSWM-02 comparison in Figure 6 of Zhang et al. [2006], where the DE3 of GSWM-02 peaks in November. From Figure 16, we also see that the larger and main-featured contribution to DE3 comes from TRMM latent heating although GSWM-ISCCP DE3's peak amplitude is about a half of GSWM-09 DE3's. During August-September, GSWM-TRMM and GSWM-ISCCP constructively interfere to reach an amplitude peak, while in November the interference of GSWM-TRMM and GSWM-ISCCP is more destructive. The degree of interference during other months is intermediate between these extremes.
4. Summary and Conclusions
[37] A new version of the Global-Scale Wave Model, GSWM-09, is a major result of this work. GSWM-09 includes updated and more realistic background (i.e., zonal mean) winds and temperatures based on SABER observations, along with radiative and latent heating rates based on ISCCP and TRMM data, respectively. These updates produce significant improvements over the longitudinal variability in tidal temperature predicted by GSWM-02, as evidenced by comparisons with SABER diagnostics in the MLT region. New findings that result from this work are as follows:
[38] 1. The diurnal cycle of radiative heating in the troposphere is as important as latent heating for producing the longitude dependence of the tidal response in the MLT region. This is consistent with the fact that the nonmigrating tidal latent heating rates are about 70% of those of radiative sources.
[39] 2. Latent heating in the troposphere accounts for much less of the MLT migrating tidal response than does radiative heating since the migrating diurnal latent heating component is weak. The semidiurnal migrating latent heat source is comparatively stronger but produces an MLT response that is weaker than that of the radiative source.
[40] 3. Troposphere latent heating alone accounts for less than half of the longitudinal variation in tidal temperature perturbations at 95 km, a typical altitude for both migrating and nonmigrating tides in the MLT region. At 110 km, where DE3 peaks and DW1 remains large, latent heating accounts more for the longitudinal variation in tidal temperature although the magnitude of the longitudinal variation is about the same as that at 95 km.
[41] 4. Longitudinal variability of the diurnal tide in the MLT region is dominated by DE3 and DW1, and thus the seasonal variations of DE3 and DW1 mainly determine the seasonal variation of longitude variability of diurnal tide. One of the major accomplishments of GSWM-09 is the improvement of the seasonal variation of DE3 over that of GSWM-02.
[42] 5. TRMM latent heating rates are characterized by a dominant DE3 component, which is comparatively larger than the ISCCP DE3 radiative heating source. In August-September, during peak DE3 activity, tropospheric latent heating contributes at least twice as much as radiative heating to the DE3 tidal response at 110 km (Figure 16). During its seasonal peak, DE3 is one of the few tides that experiences constructive interference between the components excited by radiative and latent heating.
[43] 6. GSWM-09 simulations of monthly diurnal and semidiurnal winds and temperatures will be made available over the existing GSWM web site (http://www.hao.ucar.edu/modeling/gswm/gswm.html), for the community to utilize for model-measurement comparisons.
Acknowledgments
[44] This work was supported under grant ATM-0346218 from the NSF Aeronomy Program and grant NNX08AF22G from the NASA Guest Investigation to the University of Colorado. The National Center for Atmospheric Research is funded by the National Science Foundation. We appreciate TRMM 3B42 and 3G68 data products. Thanks to W.-K. Tao and S. Lang in the Mesoscale Atmospheric Processes Branch of NASA's Goddard Space Flight Center for providing us with the TRMM CSH latent heating data. We also appreciate the JGR reviewers' thoughtful and thorough comments, which helped to strengthen this paper.
[45] Amitava Bhattacharjee thanks the reviewers for their assistance in evaluating this paper.
