1. Introduction

The Asian monsoon is a major component of the global climate system. A number of observational studies (Yeh et al. 1957; Staff Members of the Section of Synoptic and Dynamic Meteorology, Institute of Geophysics and Meteorology, Academia Sinica 1958; Yeh and Gao 1979; Nitta 1983; Yanai and Li. 1994, and many others) have recognized that the high Tibetan Plateau plays an important role in the establishment and maintenance of the Asian summer monsoon as an elevated heat source. After the First Global Atmospheric Research Program (GARP) Global Experiment (FGGE), there have been several observational studies of heat sources over the Tibetan Plateau and surrounding areas. Nitta (1983) described the 100-day mean vertical profiles of heat and moisture sources for the four parts of the eastern Tibetan Plateau. He found that the contribution to the total heating by the sensible heat flux from the elevated surface is nearly the same as that by the condensation heating due to latent heat release.

Luo and Yanai (1984) determined the heat and moisture budgets over the Tibetan Plateau using the FGGE data for 1979. They showed the deep mixed layer at 1200 UTC on the western Tibetan Plateau (Plateau hereafter) and suggested that dry thermal convection is a key process, which is responsible for the deep tropospheric heating in the afternoon hours. They also identified the principal components of the heat source after the onset as the addition of condensation heating over the eastern Plateau.

The seasonal evolution of the monsoon circulation exhibits distinct changes during its onset phase (Li and Yanai 1996; Ueda and Yasunari 1998; Wu and Zhang 1998, and many others). He et al. (1987) showed that the summer monsoon of 1979 commenced in two transition stages. The first transition seen in middle May is characterized by the eastward intrusion of low-level southwesterlies to the west of 120°E. The second transition is the onset of the Indian monsoon that usually occurs in early June to the west of 80°E. They inferred that these abrupt changes are responses to the differential heating between the Asian continental landmass and adjacent oceans. Yanai et al. (1992, hereafter referred to as Y92) extended the heat and moisture budget analysis of He et al. (1987) to the warming process of the upper troposphere. They revealed that, during the first transition period, diabatic heating and warm horizontal advection plays a primary role in the temperature increase over the eastern Plateau. During the second transition period, adiabatic warming due to large-scale subsidence is a key process that induces a temperature increase over the western Plateau, and the Iran and Afghanistan region. Recently, the desert formation over these regions is attributed to the downward motion associated with a Rossby wave pattern induced by remote diabatic heating in the Asian monsoon region (Rodwell and Hoskins 1996).

The Global Energy and Water Cycle Experiment (GEWEX) Asian Monsoon Experiment (GAME) intensive observation period (IOP) was conducted from May to August 1998. One of the purposes was to obtain high-quality four-dimensional data assimilation (4DDA) analyses for energy and water cycle processes of the Asian summer monsoon by means of 4-times-per-day soundings with radiosondes at more than 100 stations covering Southeast Asia, the northern part of south Asia (India, Bangladesh, and Myanmar), the Tibetan Plateau, central eastern China around the Huai-He River, the South China Sea, Korea, and southwest Japan (GAME International Science Plan 1998). The GAME IOP was conducted over the area shown in Figure 1, which also denotes the areas identified as the eastern and western Tibetan Plateau.

Based on these enhanced experimental upper-air data, as well as operational observation data, the 4DDA was conducted through collaboration between the Meteorological Research Institute (MRI), Numerical Prediction Division of the Japan Meteorological Agency (JMA), and the Earth Observation Research Center of the National Space Development Agency of Japan (NASDA/EORC). The 4DDA products have hybrid vertical η-coordinates, which are identical to the conventional sigma coordinates in the lower troposphere and nearly equal to the pressure coordinate in the upper troposphere. The benefit of this coordinate is that it provides a good representation of the lower boundary condition and a reduction of finite difference errors of the pressure gradient force at the higher levels (e.g., Kasahara 1974; Simmons and Burridge 1981). Since the Asian monsoon region contains the high-altitude Tibetan Plateau, we performed the heat and moisture budget analysis on the vertical η-coordinate.

This study attempts to determine quantitatively the horizontal and vertical distributions of heat sources and moisture sinks over and around the Tibetan Plateau through budget computations of mass, heat, and moisture with the GAME reanalysis dataset archived during the GAME IOP from 1 May to 31 August 1998. The other objective of the present work is to reveal the seasonal evolution process of the heat sources and moisture sinks on a pentad timescale. We compare the results of heat and moisture budget analyses by using satellite-derived precipitation data as a resource independent of the GAME reanalysis product.

2. Data and method

The major data used in this study are the 18- and 24-h forecasted 4DDA products generated by the GAME reanalysis project on a 2.25° × 2.25° grid for 1 May–31 August 1998. Temperature (T), heating rate of longwave radiation and shortwave radiation (Q_R), zonal and meridional wind components (u, υ), and specific humidity (q) are given on 30 η-levels. Surface pressure (P_s) is also utilized to determine the vertical distribution of the pressure field.

The global spectral model, which was processed with the JMA operational global forecast/analysis cycle model (GSM9912), has an equivalent grid spacing of about 55 km (T213) horizontally and 30 η-levels (L30) vertically with a model top of 10 hPa. The convection scheme used in the model is the prognostic Arakawa–Schubert scheme (Randall and Pan 1993; Moorthi and Suarez 1992). Objective analyses are performed on the same level and Gaussian grids with a three-dimensional optimum interpolation (OI) method; the first guess field is the 6-h forecast of the global model derived from previous analysis (6-hourly forecast-analysis cycle). It is known that the forecast model needs some spinup time (∼12 h) to get consistent fields for the heat and moisture budgets in the forecasted values (Gibson et al. 1997). That is the reason why we have used 18- and 24-h forecasted 4DDA products for the analyses. The quality of 4DDA products depends on the data quality control (QC) system and OI parameters as well as the accuracy of the forecast model. In the global assimilation model, we carried out careful QC for the observation data. The QC includes self-consistency checks and comparison with first guess fields of the assimilation. In a case of radiosonde observations, a self-consistency check is conducted by considering climatological threshold, blacklisted stations, consistency check among observation elements, vertical check, solar radiation correction, bias correction, and so on. To achieve this, we did preassimilation for the whole period of GAME, and looked for statistics of departure (D-values) between observation and guess field for all stations. In a vertical check, the system examines whether lapse rate exceeds adiabatic one, whether temperature and geopotential height profiles satisfy hydrostatic equilibrium, and so on. In our assimilation system, correction of the solar radiation is made for the geopotential height and temperature of the radiosonde observations on the higher levels above 200 hPa. To make the bias correction, the preassimilation experiments were conducted for the whole GAME IOP. Based on statistics of the D-values in the preassimilation, we conducted the bias correction for the geopotential height and temperature of radiosonde observations on all levels. On the other hand, the 4DDA system has two thresholds CR and CP (CR > CP) for comparison with guess fields. Here, CR is a criterion for unconditional reject, and is understood to be the upper limit of forecast error of the model; CP is a criterion for unconditional pass, and the observations with a smaller D-value than CP are accepted for the assimilation as correct data. The observations whose D-value is larger than CR, are rejected as error data. If CR > D-value > CP, it is considered that the observations are suspicious and a spatial consistency check is necessary. That is, if neighboring observations around suspected stations have similar D-values, then it can be concluded that the observations are correct and the first guess has an error, otherwise the observations are incorrect and should be rejected. In the GAME reanalysis system, a dynamic approach is adopted in the D-values check, depending on absolute values of horizontal gradient and tendency of the first guess fields. That is, stricter comparison (smaller CR and CP) was adapted where the tendency and horizontal gradient are smaller. On the other hand, a looser comparison (greater CR and CP) was made where the tendency and horizontal gradient are larger (Onogi 1998).

We also used the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) data (Xie and Arkin 1996) on 2.5° grids derived from five kinds of satellite estimates [Geostationary Operational Environmental Satellite (GOES) Precipitation Index (GPI), outgoing longwave radiation (OLR)-based precipitation index (OPI), Special Sensor Microwave Imager (SSM/I) scattering, SSM/I emission, and Microwave Sounding Unit (MSU)].

The apparent heat source Q₁ and apparent moisture sink Q₂ (e.g., Yanai 1961; Nitta 1972; Yanai et al. 1973) are calculated from the thermodynamic and moisture budget equations in the η-coordinates (Japan Meteorological Agency 1997) as follows:

View Expanded

where T is temperature; q, the mixing ratio of water vapor; V, the horizontal wind; R_d and c_p, the gas constant and the specific heat at constant pressure of dry air; p, the pressure; η̇, vertical η-velocity; and L_c, the latent heat condensation.

The pressure on half levels is

View Expanded

where k is the full-level number and k + 1/2 is the half-level number that increases with altitude. Here, k_max = 30, which is the number of model levels, and Δp_k is vertical thickness, indicating the difference between p_k and p_k−1. The relationship between full level and half level is

p_k+1/2A_kB_kp_s(4)

where A_k and B_k are parameters that define the η-coordinates. In Eq. (4), A_k is zero at the surface, so that the lowest level becomes parallel to the ground surface.

The quantity η̇(∂T/∂η) is computed using Eqs. (5) and (6).

View Expanded

As shown by Yanai et al. (1973), vertically integrating (1) and (2) from P_T to P_s, we obtain

View Expanded

where

View Expanded

Q_R is radiative heating rate; P, S, and E are, respectively, the amount of precipitation, the supply of sensible heat, and the rate of evaporation at the surface. The value Q_R used in Eq. (7) is derived from radiative processes of the forecast model, which provides first guess fields in our global assimilation. Net radiative forcing consists of shortwave heating and longwave heating (cooling). In this model, the shortwave is defined as radiation with wavelength shorter than 5 μm and mostly results from solar emission. A physical concept of parameterization of the shortwave radiation in the global model is based on Lacis and Hansen (1974). Incident solar radiation to the earth's atmosphere is partly absorbed by ozone in the stratosphere, and absorbed by water vapor and absorbed (scattered) by clouds and aerosol in the troposphere. After absorption and scattering, some of the radiation is reflected back off the earth's surface depending on albedo and the remainder is absorbed by the surface. On the other hand, longwave radiation, which is defined as wavelength longer than 5 μm is evaluated by the wide-band model (Luther and Fouquart 1984). The longwave radiation is absorbed and/or emitted by water vapor, ozone, carbon dioxide, and so on. A computational scheme of longwave heating in this global model is based on Katayama (1972). Both shortwave and longwave heating and/or cooling are monitored in the global model, and saved as a physical file for an evaluation of model performance.

The integrated heat budget [Eq. (7)] is interpreted as the total large-scale apparent heat source and consists of radiative cooling, net condensation heating, and heat supply from the surface. Also as discussed by Yanai et al. (1973), the eddy vertical flux F can be derived from

View Expanded

where the horizontal averages are denoted by ( ). Deviations from the horizontal averages are shown by primes. In Eq. (10), g is gravity, ω is vertical η-velocity, and h is moist static energy defined by

hc_pTgzL_cq.(11)

The value F indicates the vertical eddy transport of total heat and has been widely used as a useful measure of the activity of cumulus convection (Yanai and Johnson 1993). However, it should be noted here that the eddy heat flux can apply to any eddies such as dry convection and moist convection.

Before discussing the results obtained from budget computations, we shall explain the physical interpretation of Q₁ and Q₂ and their relationship (Yanai et al. 1973; Yanai and Tomita 1998).

First, positive (negative) Q₁ indicates the presence of apparent heating (cooling). In contrast, positive (negative) Q₂ corresponds to regions of apparent moisture sinks (sources). In other words, positive (negative) Q₂ implies an apparent condensation (evaporation) region. If the vertical profiles of Q₁ and Q₂ are similar to each other, there may be stratiform clouds. However, if the heating process is mainly due to the release of latent heat associated with cumulus convection, the vertical distributions of Q₁ and Q₂ will differ (e.g., Thompson et al. 1979). In such a case, there should be significant eddy transport between the peak level of Q₁ and Q₂.

Second, if the values of 〈Q₁〉 and 〈Q₂〉 are similar and vertical distributions of Q₁ and Q₂ differ, then the heating is due to cumulus convection. If the vertical distributions are similar and 〈Q₁〉 ∼ 〈Q₂〉, then mesoscale or stratiform heating is dominant. Furthermore, if the sensible heat supply from the surface is of primary importance, the heat source does not accompany the moisture sink. In this case, the horizontal distributions of 〈Q₁〉 and 〈Q₂〉 will be different (Y92).

3. The mean flow features and precipitation

To illustrate the mean large-scale circulation, we present, in Figs. 2a,b, the sequence of monthly horizontal and vertical wind fields. Also plotted are vertical η-velocity at the 10th level (roughly the lower to middle troposphere). The flow pattern clearly reveals the orographic and thermal effect of the Tibetan Plateau. During May and June, there appears to be a contrasting feature between the western and eastern plateau. We observe upward motion over the western plateau especially along with the Himalaya Mountains, while the obvious subsidence motion is found over the eastern plateau between 30°–40°N to the east of 100°E. The westerly zonal flow curves both northward and southward in the downstream direction along the topographic contours. In contrast, the northwesterly and southwesterly wind, after flowing around the elevated topography, join on the eastern plateau.

In the summer monsoon season (July and August), the most remarkable seasonal change in the flow pattern is the weakening of the westerly jet stream over the whole plateau region. We do not find distinct vertical motion over the western and eastern plateau in comparison with the premonsoon season.

Before discussion of the result of condensation heating over the plateau, we examine satellite-derived precipitation, which is independent of our budget computations. Figure 3a displays the mean rain rate of CMAP analysis (in mm day⁻¹) in May 1998 over and around the Tibetan Plateau. Interestingly, the precipitation is located in the Himalaya, Karakorum, and Tienshan Mountains and the Pamir Plateau. The averaged rain rate is 1 ∼ 2 mm day⁻¹ over the western plateau, and roughly corresponds to 30 ∼ 60 W m⁻², which is a little smaller than our estimates, the rest accounted for by dry convective heating (see section 4). The most important point is that there is premonsoon rainfall in the western plateau with substantial upward motion, particularly in the steep-slope regions. These results implies that the contribution of the condensation heating may be important in the heat balances during the preonset phase of the summer monsoon over the western plateau, which will be examined in the following section.

4. Contrasting features of heat and moisture budgets

5. Total heat and moisture balance

For the two key regions of the Tibetan Plateau, the monthly mean values of 〈Q₁〉 and 〈Q₂〉 are deduced by vertically integrating from 10 hPa to the P_s. The value L_cP is obtained from the model's surface precipitation. Tables 1 and 2 show all terms in the total heat and moisture balances obtained in this study over the western and eastern plateau. Also listed are the heat source and moisture sink estimated by Y92 for the western and eastern plateau. Y92 computed monthly 〈Q₁〉 and 〈Q₂〉 based on the combination of FGGE IIb and QXPMEX data from December 1978 to August 1979. As described in the previous section, the sensible and latent heat fluxes at the surface are evaluated from 〈Q₁〉, 〈Q₂〉, 〈Q_R〉 and precipitation using integrated budget Eqs. (7) and (8). Thus, the sensible heat flux and evaporation are estimated as residuals.

In the western plateau during May (Table 1), the atmosphere is cooled by radiation (−40 W m⁻²) but heated by both latent heat release (81 W m⁻²) and sensible heat flux from the ground surface (62 W m⁻²), indicating contribution by the sensible heat supply and latent heat release are nearly equivalent to each other.

During July (Table 2), the eastern Tibetan Plateau exhibits heat and moisture balances similar to those in the western plateau but their values are relatively larger. The amount of moisture sink (115 W m⁻²) is larger than half of the heat sources (188 W m⁻²). Considering the precipitation amount (169 W m⁻²) and the radiative cooling (−58 W m⁻²), it is estimated that over the eastern plateau the latent heat release (169 W m⁻²) is the dominant factor for the total heating rather than sensible heat supply (77 W m⁻²) from the elevated mountain surface in the mature phase of the monsoon.

The monthly mean heat and moisture budget components for the two contrasting regions are compared with those obtained by Y92 in Tables 1 and 2. It should be noted here that the result of Y92 is derived from 1979 FGGE data and the direct comparison between our results and Y92 is difficult. Therefore, we shall discuss the relative relationship between 〈Q₁〉 and 〈Q₂〉 over the plateau region.

In the western plateau during May, our estimated values of the apparent heat source (103 W m⁻²) are slightly smaller than that obtained by Y92 (129 W m⁻²), but the values are generally comparable. In contrast, the moisture sink obtained by our computations (60 W m⁻²) is much larger than those given by Y92 (18 W m⁻²). It is noteworthy that the latent heat release (76 W m⁻²) and sensible heat flux from the land surface (74 W m⁻²) are nearly equivalent. These features suggest that the western plateau in May can be viewed as a hybrid of convective rainfall and dry thermal convection.

For the eastern plateau during July, calculated values of 〈Q₁〉 in this study (188 W m⁻²) are much larger than those obtained by Y92 (94 W m⁻²). The estimated moisture sink in Y92 (48 W m⁻²) is also about half that of this study (115 W m⁻²). It should be mentioned here that these differences might be due, in part, to interannual variability. As described in section 3, our analysis region of the eastern plateau is larger than those of Y92 and includes the steep slope in its eastern boundary where substantial rains are observable in Fig. 3b. Moreover, this area was affected by anomalous precipitation caused by the combined effect of extremely high sea surface temperature over the southeastern Indian Ocean and La Niña conditions (Shen et al. 2001). This may have caused the large amplitude in the diabatic heating obtained in the present study.

6. Possible heating mechanism above the Tibetan Plateau

The question before us now is, what is causing the heating over the western plateau? Evidence has been presented in the recent studies of dry thermal convection (e.g., Luo and Yanai 1984; Yanai and Li 1994) or regional circulation (Kuwagata et al. 2001).

To examine the conditional instability, we computed the equivalent potential temperature, which is given by

View Expanded

where T is the temperature at the lifting condensation level (LCL). The LCL is determined in the following manner. First, consider the ratio of the mass of water vapor to the mass of dry air (ε ≡ m_υ/m_d). We compute partial pressure of water vapor (e) at the surface:

View Expanded

The experimental relationship between vapor pressure and the temperature is

e^{a(T_d−273.15)/(b+T_d−273.15)}(14)

where a is 17.3 and b is 237.3 above the water surface (Alduchov and Eskridge 1996). Using Eqs. (13) and (14), we obtain the dewpoint temperature (T_d) and estimate the height of the LCL above the surface to be

dz_lclTT_d(15)

Figure 17 shows the vertical distribution of the potential temperature θ (K) and the equivalent potential temperature θ_e(K) over the western plateau based on daily mean data. If we compare these vertical profiles in May with those in July, we recognize a more stable condition for dry convection with a more unstable stratification for moist convection. The seasonal time series of LCL tends to become lower toward August (Fig. 18), which indicates the surface air during premonsoon season is relatively dry, compared with the postmonsoon onset. It should be noted here that the LCL is relatively lower in early May. This may be associated with higher humidity due to precipitation and suggests that the premonsoon heating is characterized by the hybrid nature of wet and dry processes. To this point, further study should be focused on the intraseasonal variation. In contrast, condensation heating may be an important player during the wet season over the Tibetan Plateau.

The earlier results are derived from daily mean values and may be responsible for the lower LCL compared to the daytime analysis (Luo and Yanai 1984). It is important to note that the plateau domain has salient diurnal variations (Staff Members of the Section of Synoptic and Dynamic Meteorology, Institute of Geophysics and Meteorology, Academia Sinica 1958; Yeh and Gao 1979; Luo and Yanai 1984; Y92; Yanai and Li 1994) especially in the premonsoon season. Moreover, it has been revealed that the precipitable water in July exhibits remarkable differences between the evening and morning hours over the central, the northeast, and along the southern periphery of the Tibetan Plateau (Yatagai 2001). Ueno (1998) has estimated the plateau-scale distribution of precipitation for the year 1993 by use of Geostationary Meteorological Satellite (GMS) data, showing the presence of a distinct diurnal cycle of precipitation. It has been revealed that water vapor significantly decreases in the afternoon over the whole plateau (Yanai and Li 1994), which can be attributed to the plateau-scale circulation (Yanai and Li 1994) or regional thermally induced circulation (Kuwagata et al. 2001). Endo et al. (1994) has examined the planetary boundary layer (PBL) over the central plateau in the premonsoon season during the fair weather period of 1993. They found that the PBL exhibits remarkable diurnal variation and its top reaches 7000–8000 m. Thus, the role of the diurnal variations in view of regional-scale and large-scale interactions over the complex topography in the boundary and mixed layer would be important issues to be revealed in the near future. Of course all the various boundary layer convection processes can not be simply categorized with only two prototype regimes. A more detailed approach needs to include consideration of diurnal variation and intraseasonal variability.

7. Conclusions and remarks

We have used GAME 4DDA products from the GAME reanalysis joint project between MRI, JMA, and NASDA/EORC to study spatial and temporal variations of heat sources and moisture sinks over and around the Tibetan Plateau for the period from 1 May to 31 August 1998. We have mainly focused on the heat and moisture budgets over the western and eastern portion of the plateau, including the bases of the southern and eastern high-elevation mountains. The main findings of the present study may be summarized as follows.

1. During the premonsoon period of May, the results of heat and moisture budget analyses indicate that the western Tibetan Plateau plays an important role as a center of the heat source in the first transition of the Asian summer monsoon, while the eastern Tibetan Plateau remains a heat sink in the same period.

2. During July, corresponding to the mature phase of the summer monsoon, there is a large heat source in the eastern plateau with amplitude about 2 times those over the western plateau in May. In contrast, the heating over the western plateau becomes weaker in July compared to May.

3. Throughout May and June, we observe strong upward motion along the western and southwestern slopes of the western plateau, while the obvious subsidence motion is found over the eastern plateau. This paired circulation becomes weaker in July and August.

4. The condensation heating generated by convective rainfall and the sensible heat supply from the ground surface is nearly equal during May over the western plateau. A similar relationship, but a relatively larger contribution of latent heat release to the total heating, can be recognized over the eastern plateau in July.

5. The analyses of the static stability and the lifting condensation level indicate that the total heating can be attributed to both the convective rainfall and the dry thermal convection in the premonsoon period (May). On the other hand, the release of latent heat relevant to moist convection is a dominant factor for tropospheric heating after the monsoon onset.

In conclusion, we have shown that the western plateau in the premonsoon period can be characterized by the hybrid nature of “wet” processes due to condensation heating and “dry” processes associated with the dry thermal convection. Finally we should note here that the analysis period of the present study corresponds to a peculiar year. From May through August 1998, sea surface temperatures were higher than normal in the whole Indian Ocean, and the Asian summer monsoon was much more modulated (Matsumoto et al. 1999; Shen et al. 2001). This is due to combined effects of termination of a dipole mode in the equatorial Indian Ocean and the La Niña phase in the Pacific Ocean (Ueda and Matsumoto 2000). The Asian monsoon is considered to have large interannual variability that may be responsible for the different magnitudes of budget results among several studies. In this respect, the ability to accurately observe the heat and moisture balances in the Tibetan Plateau over a complete annual cycle, as well as variability on longer timescales, would significantly promote progress in monsoon prediction.