Using a WRF simulation to examine regions where convection impacts the Asian summer monsoon anticyclone

Abstract. The Asian summer monsoon is a prominent feature of the global circulation that is associated with an upper-level anticyclone (ULAC) that stands out vividly in satellite observations of trace gases. The ULAC also is an important region of troposphere-to-stratosphere transport. We ran the Weather Research and Forecasting (WRF) model at convective-permitting scales (4 km grid spacing) between 10 and 20 August 2012 to understand the role of convection in rapidly transporting boundary layer air into the ULAC. Such high-resolution modeling of the Asian ULAC previously has not been documented in the literature. Comparison of our WRF simulation with reanalysis and satellite observations showed that WRF simulated the atmosphere sufficiently well to be used to study convective transport into the ULAC. A back-trajectory analysis based on hourly WRF output showed that > 90% of convectively influenced parcels reaching the ULAC came from the Tibetan Plateau (TP) and the southern slope (SS) of the Himalayas. A distinct diurnal cycle is seen in the convective trajectories, with a majority of them crossing the boundary layer between 1600 and 2300 local solar time. This finding highlights the role of "everyday" diurnal convection in transporting boundary layer air into the ULAC. WRF output at 15 min intervals was produced for 16 August to examine the convection in greater detail. This high-temporal output revealed that the weakest convection in the study area occurred over the TP. However, because the TP is at 3000–5000 m a.m.s.l., its convection does not have to be as strong to reach the ULAC as in lower altitude regions. In addition, because the TP's elevated heat source is a major cause of the ULAC, we propose that convection over the TP and the neighboring SS is ideally situated geographically to impact the ULAC. The vertical mass flux of water vapor into the ULAC also was calculated. Results show that the TP and SS regions dominate other Asian regions in transporting moisture vertically into the ULAC. Because convection reaching the ULAC is more widespread over the TP than nearby, we propose that the abundant convection partially explains the TP's dominant water vapor fluxes. In addition, greater outgoing longwave radiation reaches the upper levels of the TP due to its elevated terrain. This creates a warmer ambient upper-level environment, allowing parcels with greater saturation mixing ratios to enter the ULAC. Lakes in the Tibetan Plateau are shown to provide favorable conditions for deep convection during the night.

16 Six year WWLLN lightning totals for 1500 -0400 UTC for August-September showing the lake "hot spots" in convection. Lightning

ABSTRACT
The Asian summer monsoon is a dominant feature of the global circulation. The upperlevel anticyclone (ULAC) associated with the Asian summer monsoon circulation stands out vividly in satellite observations of trace gases. The ULAC also has been diagnosed as a region that is important to troposphere-to-stratosphere transport. Therefore, understanding the role of convection in transporting boundary layer air into this upper-level circulation is important to understanding the atmospheric chemistry of the upper troposphere/lower stratosphere.
We ran the Weather Research and Forecasting (WRF) model at convective-permitting scales (4 km grid spacing) to simulate the atmosphere between 10-20 August 2012. Such highresolution modeling of the Asian ULAC previously has not been documented in the literature.
Comparison of our WRF simulation with reanalysis and satellite observations showed that WRF simulated the atmosphere sufficiently well to be used to study convective transport into the ULAC. A back-trajectory analysis showed that >90% of convectively influenced parcels reaching the ULAC came from the Tibetan Plateau (TP) and the southern slope (SS) of the Himalayas. A clear diurnal cycle is seen in the convective parcels, with their greatest impact occurring between 1600-2300 local solar time. This finding highlights the role of "everyday" diurnal convection in transporting boundary layer air into the ULAC.
WRF output at 15 min intervals was produced for 16 August to examine convection in greater detail. This high-temporal output indicated that the weakest convection occurred over the TP. However, because the TP is at 3000-5000 m MSL, its convection does not have to be as strong to reach the ULAC as in lower altitude regions. Additionally, because the TP's elevated heat source is a major cause of the ULAC, we propose that convection over the TP and the neighboring SS is geographically situated to impact the ULAC most frequently. x The vertical mass flux of water vapor into the ULAC also was calculated. Results show that the TP and SS regions dominated other Asian regions in vertically transporting moisture into the ULAC. Because convection reaching the ULAC is more widespread over the TP than nearby, we propose that the abundant convection partially explains the TP's dominant water vapor fluxes. In addition, greater outgoing longwave radiation reaches the upper levels of the TP due to its elevated terrain. This creates a warmer ambient upper level environment, allowing parcels with greater saturation mixing ratios to enter the ULAC.

INTRODUCTION AND BACKGROUND
Southeast Asia is a region of rapidly increasing population and emissions (Zhang et al. 2009).
Deep convection in the region, associated with the Asian summer monsoon (ASM), can transport surface-based emissions into the upper troposphere/lower stratosphere (UTLS; e.g., Li et al. 2005;Randel et al. 2010). In particular, the upper-level anticyclone (ULAC) associated with the ASM appears to provide a pathway for troposphere-to-stratosphere transport (TST; Gettleman et al. 2004;Park et al. 2007), making its chemical composition important to UTLS chemistry.
However, the exact role that deep convection in the ASM plays in controlling the composition of the ULAC is a subject of considerable debate. Previous studies of the region (e.g., Fu et al. 2006;Park et al. 2007Park et al. , 2009James et al. 2008;Chen et al. 2012) have reached varying, even contradictory, conclusions regarding the impacts of ASM convection on the composition at upper levels. These previous studies were limited by coarse-resolution models and by observations that were limited in both time and space.
The impact of convective detrainment on the composition of the tropical UTLS has been debated in many previous studies (e.g., Fueglistaler et al. 2009, especially sections 3.4 and 3.7, and references therein). Bergman et al. (2012) suggested three pathways that convective air can take: 1) air is detrained below the tropopause and recirculated back into the boundary layer, 2) air is detrained above the tropopause, but then is advected into a region of negative heating rates, thus leading to descent, or 3) air is detrained above the tropopause, advected into regions of positive heating, and subsequently lofted into the lower stratosphere (see their Fig. 1 for a schematic of these different pathways). However, the relative frequency of these different pathways has not been determined.
A common approach for diagnosing the pathways by which air travels to the UTLS is to perform back trajectory analyses (on a global scale) to determine the sources of air comprising the tropical tropopause layer (TTL; Gettleman and Forster 2002;Fueglistaler et al. 2009;Bonazzola and Haynes 2004;Fueglistaler et al. 2004;James et al. 2008;Tzella and Legras 2011;Bergman et al. 2012). This method allows quantitative and qualitative statistics about transport into the TTL to be inferred. However, the use of global models in these studies limited their ability to make conclusive statements about the impacts of convection. This is partially due to the coarse resolution at which the models were run, which greatly limits their ability to resolve convective scale motions. For example, Bonazzola and Haynes (2004)  become "trapped" within the anticyclonic circulation (Randel and Park 2006;Park et al. 2008). Gettleman et al. (2004) showed that ASM circulation might provide 75% of the total H 2 O flux into the tropical lower stratosphere during the summer months, indicating its importance to TST.
Air in the ULAC also is periodically subjected to long-range transport across the Pacific where it can impact the air quality of the western United States ). Randel and Park (2006) showed that anomalous mixing ratios of H 2 O and O 3 are correlated with ASM convective events, but with a 5 day lag. Additional observations revealed that H 2 O at 216 hPa is highly correlated with convection, but is displaced both temporally and horizontally at 100 hPa (Park et al. 2007). Modeling studies of the region have indicated that deep convection over India and SE Asia can account for 30% of the CO at 100 hPa (Park et al. 2009). However, a back trajectory analysis by James et al. (2008)  with radiative cooling can collide with onshore low-level monsoon flow to initiate deep convection over the southern slope of the Himalayas (e.g., Luo et al. 2010;Romatschke et al. 2010). Previous studies examining the impacts of convection on the ULAC (e.g., Park et al. 2009;Chen et al. 2012) were not able to explicitly resolve these mesoscale convective events and therefore had to rely on convective parameterizations. Parameterizations vary in their assumptions, "trigger" mechanisms, and closure techniques (e.g., Kain 2003 vs. Grell andDevenyi 2002). Therefore, varying model results sometimes can be attributed to which parameterization is chosen (e.g., Taylor 2011). To fully understand the details of deep convective transport into the ULAC, convection must be explicitly resolved, and, to our knowledge, this has not been documented previously.
The goal of this paper is to simulate ASM convection at the fine-scale resolution that previously has not been documented, and then use this high-resolution output to study the impacts of convection on the ULAC. We seek to identify the geographic regions where convection reaches the ULAC, and provide physical and dynamical reasoning for why these regions are important. Chapter 2 describes our high-resolution, convective-permitting (4 km grid spacing; Lackmann 2010) Weather Research and Forecasting (WRF; Skamarock et al. 2008) simulations, including the data used to drive the model and the simulation period that was chosen. That chapter also describes the HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT; Draxler and Hess 1998) model that was used for trajectory analyses, along with the observational data that were used to complement our WRF simulation. In chapter 3, the WRF simulation is used with HYSPLIT and observations to describe convective transport into the ULAC, diagnose regions where convection impacts the ULAC most frequently, and provide physical insight into why certain geographical regions are more important than others when considering convective impacts on the ULAC. Chapter 4 presents a summary and conclusions.

WRF Model Description
We used WRF Version 3.4.1 (Skamarock et al. 2008) to make simulations over three domains (Fig. 2). From outer to inner, the domains used 36 km, 12 km, and 4 km grid spacing, respectively. Meteorological initial and boundary conditions came from NCEP/GFS final analyses with 1° × 1° grid spacing. Temperature, horizontal winds, and moisture in the outer two domains (d01 and d02) were nudged to the GFS analyses every 6 h. This nudging occurred only on vertical levels above the boundary layer, and was used to keep the large-scale features consistent with observations. Nudging was not employed in the innermost domain (d03).
However, because d03 receives its boundary conditions from d02, the nudging occurring on d02 will ultimately influence the results of d03. All analyses presented in this paper are from d03.
All three domains had 70 terrain-following vertical sigma levels that spanned the surface to 10 hPa. Diffusive damping was applied in the topmost 5 km of the model to prevent gravity waves from reflecting off the upper boundary of 10 hPa. Meteorological variables were integrated forward in time using Runge-Kutta 3 rd order time integration. All moisture variables and other scalars were advected using a positive-definite scheme to ensure mass conservation.
The outer two domains of the simulation used an updated version of the Grell-Devenyi (Grell and Devenyi 2002) cumulus parameterization. The innermost domain with 4 km grid spacing did not use a convective parameterization, i.e., convection was explicitly resolved. We use the phrase "explicit resolution" loosely because, even at 4 km grid spacing, the model is unable to resolve all convective motions (e.g., individual cumulus clouds). Furthermore, using explicit resolution at 4 km grid spacing will inherently limit the minimum convective updraft width to this scale. For example, an updraft in nature having a width of 1 km would be overproduced in the WRF simulation, which could lead to anomalous latent heating and unrealistic updraft strengths. This caveat should be considered while interpreting the results of the WRF simulation. Cloud physics were computed using the 2-moment Thompson et al. (2008) bulk parameterization which accounts for graupel, ice, and snow processes. We used the Yonsei University (YSU) planetary boundary layer scheme with non-local closure (Hong et al. 2006).
The YSU scheme has been used in previous studies of the region (e.g., Romatsche et al. 2010).
Short wave radiation came from the Goddard scheme (Chou et al. 2001), while the Rapid

Simulation Period
The WRF model was used to simulate atmospheric conditions over SE Asia (Fig. 2) during the 10 day period between 0000 UTC 10 August -0000 UTC 20 August 2012. This particular 10 day simulation period also was chosen because the ULAC changed location, size, and shape (Fig. 3). Although this is not unusual, previous climatological studies of this region (e.g., Chen et al. 2012) were not able to account for these changes. And, because we are examining transport specifically into the ULAC, this variability ensured that our results would not be biased toward a particular geographical region, e.g., if the ULAC were stagnant, then one region might appear to dominate another.

HYSPLIT Model Description
Forward and backward trajectories were computed using the HYSPLIT model (Draxler and Hess 2010). HYSPLIT interpolates WRF data into its own terrain-following sigma coordinate system without compromising the vertical resolution of the native WRF output.
Advection takes place in a Lagrangian framework, i.e., following the parcel. The time step for the integration of the parcel's position varies based on the maximum velocity at a given grid point (see Draxler and Hess 1998 for full details). This trajectory framework has been used extensively (e.g., Draxler 1996) and provides a way to follow individual parcels experiencing deep convection to gain a better understanding of the transport occurring during the ASM.

WWLLN Data
Lightning data have proven to be a reliable indicator of deep convection during summer months over continental regions (e.g., Ávila 2010). In addition, in regions of complex terrain and sparse observations, such as the ASM area, lightning data can serve as a high resolution proxy for convection (e.g., Deierling and Petersen 2008   WRF again predicts similar geopotential heights at 700 hPa compared to the MERRA reanalysis (~7 m difference averaged over the entire domain and simulation period). The use of a common color bar makes some differences stand out more than actually occurs. However, one minor 14 and WRF (right column) geopotential height (m) a ), 1200 UTC 15 August (c and d), and 1200 UTC 1 ) at 700 hPa for C 19 August (e and difference is 9.5 days into the WRF run (Fig. 4f) when WRF produces a weak, secondary region of low pressure in the eastern portion of the domain (~98° E, 26° N) that is not apparent in the MERRA reanalysis.
We next compare our WRF simulated outgoing longwave radiation (OLR; a proxy for convection) to Fengyun (FY)-2D gridded satellite-derived brightness temperatures to confirm that WRF is simulating convection in the correct locations. The gridded FY-2D data are provided by the Center for Environmental Remote Sensing, Chiba University, on a 4 km 4 km grid (available at ftp://fy.cr.chiba-u.ac.jp/grided/). Figure 5 shows snapshots of these parameters at 1200 UTC on 12, 15, and 18 August 2012. Because we are comparing "apples to oranges," Because WRF closely simulates the general synoptic features of the upper and lower troposphere, and places convection in locations that generally agree with satellite observations (albeit with some differences), we are confident that WRF is simulating the real atmosphere sufficiently well to investigate convective transport into the ULAC at a resolution that previously has not been documented.

Back Trajectory Analysis
To examine convective transport into the ULAC, we performed a back trajectory study using HYSPLIT and our hourly WRF output. We defined the ULAC as the 14430 m geopotential height contour at 150 hPa, and the 16830 m geopotential height contour at 100 hPa.
These values were chosen by finding height contours that remained closed, within our domain, throughout the entire simulation. However, because the ULAC changed location throughout the runtime (Fig. 3), this criterion could not always be met. Other values were tested to ensure that our results were not compromised by the choice of these specific values. The GFS FNL data indicated that the lapse-rate derived tropopause (WMO definition) was located between 105-95 hPa during our runtime. Therefore, the two chosen pressure surfaces correspond to just below (150 hPa) and very near (100 hPa) the tropopause level. At each time step of the WRF simulation, we recorded the latitude, longitude, and geopotential height of each gridpoint within the defined ULAC. Six hour back trajectories then were released from these locations at each time step between 0000 UTC 11 August and 2300 UTC 19 August. This method accounted for the movement and change in shape and/or area of the ULAC that was seen in Fig. 3 Park et al. 2007 for an analysis that differentiates tropospheric air from stratospheric air based on CO and O 3 concentrations).

Back Trajectories from 150 and 100 hPa ULAC
With  through the evening and early nighttime (Fig. 9). The large diurnal cycle seen in Figs. 8-9, with the peak in convective impact occurring between 1000-1700 UTC (1600-2300 LST), is consistent with previous observational studies of convection over the TP and SS regions (e.g., Liping et al. 2002;Barros et al. 2004;Hirose et al. 2005;Yaodong et al. 2008;Luo et al. 2010;Romatschke et al. 2010). Our WRF simulation and HYSPLIT back trajectories emphasize the importance of this diurnal convection in pumping boundary layer air into the ULAC. The results of the ULAC back trajectory analysis highlighted the importance of convection over the Tibetan Plateau in affecting the composition of the ULAC (Fig. 7). To further illustrate this point, Fig.   10 shows the number of convective trajectories reaching the ULAC at 150 hPa (black line; see   ubiquitous throughout the study domain on this day (Fig. 11) Figure 12 displays the locations where convective forward trajectories crossed the 150 hPa surface, binned into 0.1° × 0.1° cells, for the TP (Fig. 12a), SS (Fig. 12b), BOB (Fig.   12c), and IND regions (Fig. 12d). The vertical velocity of each trajectory was determined by dividing the vertical distance between the PBLH and 150 hPa by the time (to within 15 min) required to reach that altitude. The SS region exhibited the strongest, although not as frequent, convection (maximum updrafts of ~23 m s -1 ; Fig. 12b), followed by the BOB (maximum updrafts of ~21 m s -1 ; Fig. 12c), and finally the IND and TP regions (maximum updrafts of ~20  Romatschke et al. (2010). Furthermore, the updraft strengths in each region (Fig. 12) are consistent with those of deep convection occurring in nature (Markowski and Richardson 2010), ensuring that the convective-permitting WRF is simulating the deep convection realistically. One should note that although vertical motions over the TP region are weaker than in the other three regions, its trajectories still reach the 150 hPa surface more frequently than those from the other regions (Fig. 12a). Because the height of the Tibetan Plateau is 3000-5000 m above MSL, the highest of the four regions, its convection does not have to be as deep to reach the 150 hPa surface. The 150 hPa surface also is elevated over the TP (e.g., approximately 95 m higher over the TP than over the BOB throughout the runtime).
However, this slight elevation of the 150 hPa surface is small in comparison to the height of the TP. This allows diurnally-driven convection over the TP (Figs. 8 and 9) with weaker updrafts (Fig. 12a) to impact the UTLS more frequently, whereas convection in other regions must be deeper to penetrate into the UTLS.
Based on the above mentioned results, we hypothesize that convection in the TP and SS regions is most favorable for impacting the ULAC because of their geographic locations. The ASM upper-level anticyclone is a result of the elevated heating of the Tibetan Plateau and diabatic heating from ASM convection (e.g., Hoskins and Rodwell 1995;Duan and Wu 2005;Wu et al. 2012). Although the ULAC does oscillate (Krishnamurti et al. 1973;our Fig. 3), the ULAC is always thermally linked to the TP due to its elevation, and convection occurring over the Tibetan Plateau and neighboring southern slope will thus be favorable for affecting the composition of the ULAC. This idea is supported by the results in Figs. 7 and 12. Specifically,   12 shows that deep convection certainly is occurring in other locations besides the TP and SS regions; however, it is precisely the geographic locations of TP and SS regions that make them so important in impacting the ULAC.

Vertical Fluxes of Water Vapor Mass
An important aspect of ASM deep convection is its role in transporting water vapor into the ULAC and thus its role in affecting the H 2 O maximum observed in Fig. 1. Water vapor is a greenhouse gas that can influence the radiation budget at upper levels, thus causing a positive feedback on climate (Held and Soden 2000;Gettleman and Fu 2007). Moreover, if water vapor enters the lower stratosphere, it can deplete ozone in what is referred to as a "catalytic loss cycle" (Jacob 1999;Stenke and Grewe 2005). Briefly, the following reactions occur: The net result of these reactions is the depletion of ozone, leaving (products of R2 and R3) to continue depleting ozone by way of R2 and R3. Gettleman et al. (2004) showed that the ULAC could account for 75% of the total upward water vapor flux at tropopause levels in the tropics. Thus, it is important that regional and global climate studies understand the processes controlling ULAC water vapor content.
To analyze transport into the ULAC, we calculated vertical fluxes of water vapor passing through the 100 and 150 hPa surfaces similar to the method of Halland et al. (2009): where M v is the vertical water vapor mass flux in metric tons per square kilometer per timestep, MC is the mass concentration of water vapor (kg m -3 ), w is the vertical velocity (m s -1 ), A is the area of the model grid box (16000 m 2 ), and t is the timestep (3600 s). We summed these fluxes over all ULAC grid points and normalized them by the area (km 2 ) of the ULAC to account for the change in size that occurred at each time step. The ideal gas law was used to calculate MC from our WRF output: where P L is the pressure level at which the computation is made (Pa), M is the molecular weight of water vapor (18.01528 g mol -1 ), C is the model-calculated water vapor mixing ratio (ppbv), R is the universal gas constant (8.3144 J mol -1 K -1 ), and T is the model-calculated temperature (K). vapor is constantly being transported into the ULAC by diurnal convection (Fig. 13b). At this pressure altitude, the TP region is the dominant source region, followed by the southern slope.
At 100 hPa (Fig. 13a), the convective diurnal cycle still is prominent; however, the TP and SS regions make similar contributions to the ULAC water vapor content. Referring back to Fig.   12b, these results agree with the convective characteristics of the SS region, i.e., the strong updrafts observed in the SS region are favorable for convection containing large amounts of moisture to reach 100 hPa. Fu et al. (2006) also found that convection over the Tibetan Plateau and southern slope was dominant in controlling the water vapor content at upper levels. These results (Fig. 13) also are consistent with a global water vapor flux analysis by Lelieveld et al. (2007) who proposed that the Tibetan Plateau is a favorable region for stratospheric moistening.
Conversely, the BOB and IND regions transport a negligible amount of water vapor mass flux to the ULAC at 100 and 150 hPa. This is consistent with the back trajectory analysis previously presented (e.g., Fig. 7), which showed that less than 1% of the convective trajectories entering hPa (b). The each defined the ULAC originated from these regions. Overall, the water vapor flux calculations (Fig. 13) again highlight the role of diurnal convection in transporting water vapor into the ULAC.
Similar to Fu et al. (2006), we argue that the thermodynamic environments of the TP and SS make them favorable for convection to moisten the upper levels. Compared to the other regions, the elevated surface of the Tibetan Plateau produces warmer temperatures in the UTLS. Figure 14 shows the WRF-simulated area average temperature in each region at 100 hPa ( Fig.   14a) and 150 hPa (Fig. 14b). The TP and SS regions are several degrees warmer than the BOB and IND regions at both pressure altitudes. Because of this warmer upper level ambient environment, parcels entering the ULAC in these regions will have a greater saturation mixing ratio, potentially allowing them to make a greater contribution to the ULAC water vapor content.
The water vapor flux results in Fig. 13 highlight the importance of this warmer upper level ambient environment over the Tibetan Plateau. Specifically, magnitudes of vertical moisture flux depend on both the vertical velocity and the water vapor mixing ratio (Eq. 1). We already have shown that vertical velocities over the Tibetan Plateau are weaker (Fig. 12) than in the other regions. This suggests that the greater flux values over the TP are the result of larger water vapor content at upper levels over this region. Furthermore, because convection affecting the ULAC is more widespread than in other regions (e.g., Fig. 6), this also contributes to the greater TP flux observed in Fig. 13.
The warmer temperatures over the Tibetan Plateau (Fig. 14) also might enhance the "cirrus lofting" mechanism proposed by Corti et al. (2006). More outgoing long wave radiation reaches the upper levels over the Tibetan Plateau because of its high altitude, allowing remnant cirrus clouds to warm (due to this increased OLR) and produce regions of ascent. Thus, TST within the ULAC might be more efficient over the TP due to this enhanced cirrus lofting. Referring back to Fig. 6b, convection reaching the ULAC at 150 hPa is ubiquitous over the Tibetan Plateau. If pollution were to be advected horizontally into this region (e.g., from northern India or mid-eastern China), TP convection could transport that pollution into the UTLS via direct convective injection, gradual ascent within the ULAC, or the proposed enhanced cirrus lofting.

Lake-Influenced Convection over the Tibetan Plateau
Convection over the Tibetan Plateau mostly is the result of diurnal heating and orographic lift (Liping et al. 2002;Kurosaki and Kimura 2002;Yaodong et al. 2008;Romatschke et al. 2010). The strong surface heating and subsequent rising motion (e.g., dry thermals) over the TP region lead to low-to-mid level horizontal convergence, drawing moisture from the surrounding regions (Wu et al. 1997;Carrico et al. 2003;Fu et al. 2006). This moisture, the instability generated by diurnal heating, and mesoscale trigger mechanisms such as orographic lift all combine to support the development of the observed diurnal convection in this region.
However, we hypothesize that the Tibetan lakes also provide a mesoscale source of moisture, instability, and convergence to support deep convection. Lake-effect convection usually occurs when a cold air mass is advected over a relatively warm body of water. This warm water surface provides heat and moisture fluxes, which can destabilize the air mass and ultimately produce convection (Markowski and Richardson 2010). Lake-effect convection is common downwind of the North American Great Lakes (e.g., Niziol et al. 1995). However, to our knowledge, lakeinfluenced convection over the Tibetan Plateau has not previously been documented.

SUMMARY AND CONCLUSIONS
The Asian summer monsoon (ASM) is a dominant feature of the global circulation. Satellite observations of the upper troposphere/lower stratosphere (UTLS) highlight the importance of the ASM upper-level anticyclone (ULAC) in controlling the composition of these levels (Fig. 1).
Previous studies have shown that troposphere-to-stratosphere transport is enhanced within the ULAC, emphasizing its importance to UTLS chemistry. The processes controlling the composition of the ULAC are important, although not fully understood nor agreed upon. Deep convection generally is accepted as an important mechanism for transporting surface-based emissions into the ULAC. However, previous studies that have examined convective transport into the ULAC have reached varying, often conflicting conclusions as to which geographic region contains the convection that most impacts the ULAC. These differing conclusions partially are a result of using coarse resolution numerical models and observations that are limited in both time and space.
We have used a convective-permitting (4 km grid spacing) WRF simulation to examine convective transport into the ULAC. To our knowledge, such high resolution modeling has not been used previously to study convective transport in the ASM region. We ran WRF from 0000 UTC 10 August to 0000 UTC 20 August 2012. Comparisons with independent reanalyses and satellite observations revealed that the WRF simulation closely captured the synoptic-scale features of the upper and lower atmosphere, and generally placed convection in regions that agreed with observations. Therefore, the simulation was suitable for more detailed analyses.
We defined the ULAC as the 14430 m geopotential height contour at 150 hPa and the 16830 m contour at 100 hPa. Then, using our WRF output and HYSPLIT, we released 6 h back trajectories from every grid point within the ULAC at every time step of the WRF simulation.
This method allowed us to account for the observed changes in position and shape of the ULAC during the simulation period (Fig. 3). We defined convective trajectories as those that reached the boundary layer, starting from the ULAC, at some point during their 6 h backtrack. The back trajectories revealed that more than 90% of the convective parcels comprising the ULAC came from either the Tibetan Plateau or Southern Slope regions. These results were consistent with Li et al. (2005), Fu et al. (2006), and Wright et al. (2011). However, they were somewhat different from the findings of Park et al. (2007Park et al. ( ,2009 and Chen et al. (2012) who argue that convection over the Bay of Bengal is more important than that of the Tibetan Plateau. Further analysis revealed a large diurnal cycle in the number of convective trajectories reaching the ULAC, with the greatest number occurring during the late afternoon and into the early night (1600-2300 LST). The observed timing of the convective trajectories was consistent with observational studies of convection over the TP and SS regions (e.g., Yaodong et al. 2008;Luo et al. 2010).
The trajectories highlighted the importance of this "everyday" diurnal convection in transporting boundary layer air into the ULAC and pointed to the TP as the dominant region where this convection occurs.
A large increase in the number of convective trajectories occurred on 16 August 2012 ( Fig. 8a). We produced WRF output at 15 min intervals to examine this day in detail. A forward trajectory analysis was conducted using the high temporal resolution output. The results indicated that the weakest convection occurred over the Tibetan Plateau region, while the strongest was located over the Southern Slope. However, parcels originating from the Tibetan Plateau reached the 150 hPa surface more frequently than those originating over the other three regions. Because the elevation of the Tibetan Plateau is ~3-5 km, its convection does not have to be as deep to reach the 150 hPa surface.
Vertical mass fluxes of water vapor were calculated to determine which regions transported the most moisture into the ULAC. At 150 hPa, the TP region contributed the most water vapor to the ULAC (Fig. 13b). At 100 hPa, the TP and SS regions transported similar amounts. Although convection is weakest over the TP, the warmer ambient environment at upper levels, leading to higher parcel saturation mixing ratios, might allow convection occurring there to transport more moisture into the ULAC. This added warmth is due to the elevated base of the Tibetan Plateau, which allows more outgoing long wave radiation to be absorbed at higher altitudes.
Although not a focus of our study, the elevated surface of the Tibetan Plateau also might enhance the "cirrus lofting" mechanism proposed by Corti et al. (2006). Because more outgoing long wave radiation will reach the upper levels over the elevated TP, cirrus clouds over this region will absorb more energy, creating mesoscale regions of ascent. This mechanism might enhance troposphere-to-stratosphere transport over the Tibetan Plateau.
Results showed that convection over the Tibetan Plateau most frequently impacts the ULAC. Observed WWLLN lightning during the study period showed that lightning over the Tibetan Plateau agreed closely with the convective trajectories impacting the ULAC, further validating this finding. Close analysis of the WWLLN lightning climatology showed that Tibetan lakes might provide a mesoscale triggering mechanism to initiate convection in the region. The role of the Tibetan lakes in triggering convection that ultimately impacts the ULAC requires further study.