The climatology of planetary boundary layer height in China derived from radiosonde and reanalysis data

The important roles of planetary boundary layer (PBL) in climate, weather and air quality 15 have long been recognized, but little is known about the PBL climatology in China. Using the fineresolution sounding observations made across China and reanalysis data, we conducted a comprehensive investigation of the PBL in China from January 2011 to July 2015. The boundary layer height (BLH) is found to be generally higher in spring and summer than that in fall and winter. The comparison of seasonally averaged BLH derived from observations and reanalysis shows good 20 agreement. The BLH derived from threeor four-times-daily soundings in summer tends to peak in the early afternoon, and the diurnal amplitude of BLH is higher in the northern and western sub-regions of China than other sub-regions. The meteorological influence on the annual cycle of BLH is investigated 4 as well, showing that BLH at most sounding sites is negatively associated with the surface pressure and lower tropospheric stability, but positively associated with the near-surface wind speed and temperature. This indicates that meteorology plays a significant role in the PBL processes. Overall, the key findings obtained from this study lay a solid foundation for us to gain a deep insight into the fundamentals of PBL in China, which helps to understand the roles of PBL playing in the air pollution, weather and 5 climate of China.


Introduction
The planetary boundary layer (PBL), the lowest layer of troposphere, is directly influenced by the 10 Earth's surface, at a response timescale of about an hour or less (Garratt, 1994;Stull, 1988). The accurate parameterization scheme with regard to how the PBL evolves is critical to prediction of weather, climate and air quality, which heavily relies on the high resolution observations of atmospheric profiles (Hu et al., 2010;Wood, 2012;Xie et al., 2012). Through the PBL, the exchanges of momentum, water, heat and air pollutants take place between the Earth's surface and the free troposphere. To 15 characterize the structure of PBL, the concept of boundary layer height (BLH) is commonly used (Seibert, 2000;Seidel et al., 2010), since the BLH determines the vertical extent of turbulent mixing, vertical diffusion, and convective transport within PBL.
The development of PBL is dominated by complex surface forcings, including sensible heat flux, frictional drag, evaporation, transpiration, and terrain induced flow modification (Stull, 1988). As a 20 result, the BLH is quite variable both in time and space. During a diurnal cycle, the BLH is typically shallow (a few hundred meters) at night due to the strong near-surface stability, and the PBL can be well developed and reach several kilometers in the afternoon.
Traditionally, the BLH is diagnosed using the height-resolved observations of temperature, humidity, and wind from radiosondes (Seibert, 2000;Seidel et al., 2010). The vertical resolution of these profiles are not sufficient to estimate BLH, since most soundings only provide a few records below 500 hPa 5 (Liu and Liang, 2010). Partly due to the lack of fine-resolution soundings, there have been rather limited investigations of the BLH climatology over specific locations and regions using atmospheric sounding data (Liu and Liang, 2010;Norton and Hoidale, 1976;Seidel et al., 2012).
In addition to the traditional radionsonde-based BLH algorithms, other methods relying on new data sources such as ground-based lidar (Hennemuth and Lammert, 2006;Sawyer and Li, 2013), sodar 10 (Beyrich, 1997), ceilometer (Eresmaa et al., 2006), aircraft sounding (Dai et al., 2014), and space-borne lidar (Chan and Wood, 2013;Zhang et al., 2016) have been put forward to determine the PBL structure and its processes. However, the multiple types of observations tend to reveal different 7 This study of BLH climatology has important implications for better understanding the physical mechanisms, seasonal variation, and spatial distribution of severe air pollution in China (H. Zhang et al., 2015;R. Zhang et al., 2015), since the PBL structure and processes play important roles in modulating the air quality (Guo et al., 2009;Guo et al., 2010;Hu et al., 2014;Miao et al., 2015;Lee et al., 2015).
During the haze events the presence of aerosol can somehow modify the PBL meteorology and 5 exacerbate pollution (Wang et al., 2013;Quan et al., 2013;Ding et al., 2016;Miao et al., 2016). For instance, the effects of black carbon particles on PBL have been adequately discussed in an early important publication (Zhang et al., 2008). Guo et al. (2014) found a lack of a diurnal variation, but a cycle of 4-7 days in the aerosol properties, indicating a reduced PBL diurnal trend during polluted periods. Furthermore, another recent observation study (Peng et al., 2016) has demonstrated rapid aging 10 of black carbon particles in China, leading to considerably enhanced light absorption, which suggests a large impact of atmospheric stabilization by black carbon particles. However, the main goals of this study are to investigate the BLH climatology in China, and to use it to explore its potential association with meteorology.
The remainder of this paper is organized as follows. Section 2 describes the data and method used, 15 followed by the uncertainty analysis for the BLH retrievals in section 3. Section 4 presents the climatology of BLH in China, as well as the associations with atmospheric variables. Finally, the key findings are summarized in section 5.

20
已删除: The main goals of this study are to investigate the BLH climatology in China, and to use it to explore the potential influence of meteorology on the climatological BLHs. The radiosonde network of L-band sounding system dated back to 2002 when the China Meteorological Administration (CMA) began upgrading its radiosonde system. As of the beginning of 2011, deployment of the L-band sounding systems have been expanded to 120 operational radiosonde stations (Fig. 1). The GTS1 digital electronic radiosonde, one of key components of the L-band sounding system, is now widely used in operational radiosonde stations in China, providing fine-5 resolution profiles of temperature, pressure, relative humidity, wind speed and direction twice a day at 0000 (0800) and 1200 UTC (2000 BJT) (Fig. 2).
Previous intercomparison studies (e.g., Bian et al., 2011) indicate that the two types of radiosondes, i.e., Vaisala RS80 and GTS1, agree very well in the profile measurements in the troposphere (including PBL), albeit the large biases in the upper atmospheric levels. Therefore, the data from radiosondes of 10 GTS1 are good enough to derive BLH for the superior performance in PBL.
The sounding data collected from January 1, 2011 to July 31, 2015 are used here to obtain the BLH at widely scattered radiosonde network across China. In total, we have made use of 391,552 profiles across China, including 1,578 profiles at 0200 BJT,190,027 profiles at 0800 BJT, 10,313 profiles at 1400 BJT, and 189,634 profiles at 2000 BJT. In summer (the monsoon season), up to two additional 15 soundings are launched occasionally at 0600 UTC (1400 BJT) and 1800 UTC (0200 BJT) during certain intensive observing periods at selected stations. The fine-resolution sounding observations provide essential information to investigate the diurnal structure of PBL.
In addition to the CMA sounding observations, the simultaneous profiling data of temperature and wind from ERA-Interim reanalysis (Dee et al., 2011) are also used to derive BLH in this study. The 20 ERA-Interim reanalysis assimilates a variety of measurements, including radiosonde observations at highly discrete levels (the so-called standard sounding levels), into weather prediction models in a physically consistent manner. The horizontal resolution of ERA-Interim reanalysis is 0.75 degree. In the vertical, there are 60 layers using hybrid coordinates that extend from the surface to the top of 0.1 hPa level, with 21 layers between the surface and 5 km above ground level (AGL). The vertical resolution is about 20 m near the surface, and gradually decreases to about 200 m at 900 hPa level and 500 m at 500 5 hPa level.
Similar to the methods used by Seidel et al. (2012) to investigate the characteristics of BLH climatology in the United States and Europe, the bulk Richardson number (Ri) method (Vogelezang and Holtslag, 1996) was taken to simultaneously estimate the BLH from CMA soundings and ERA-Interim data. The Ri method has been proven to be one of the best methods for BLH climatology analysis, since 10 it is suitable for both stable and convective boundary layers and can be applied to large amount of radiosonde and reanalysis data (Seidel et al., 2012). Ri is defined as the ratio of turbulence associated with buoyancy to that induced by mechanical shear, which is expressed as where z denotes height above ground, s the surface, g the acceleration due to gravity, θ v virtual potential 15 temperature, u and v are the component of wind speed, and u * the surface friction velocity. u * can be ignored here due to the much smaller magnitude compared with bulk wind shear term in the denominator (Vogelezang and Holtslag, 1996). Note that the Ri is dimensionless, and has nothing to do with the intensity of turbulence, but it can determine whether the turbulence exists or not. Previous theoretical and laboratory studies (e.g., Stull, 1988) suggested that when Ri is smaller than the critical 20 value (~0.25), the laminar flow becomes unstable. Thus, the lowest level z at which interpolated Ri crosses the critical value of 0.25 is referred to as BLH in this study, similar to the criteria used by Seidel respectively. Unless noted otherwise, all the values of BLH are presented with reference to the height above ground level, rather than that above sea level, to rule out the impact of topographic variation.
BLHs computed using different critical Ri can be hardly recognized, although the BLH of Ri = 0.3 seems slightly higher.
In the spirit of overall evaluation of the ensemble BLH estimations, scatter plots have made to show CMA-BLHs computed using 0.2 and 0.25 as critical Ri, versus using 0.25 and 0.3 as critical Ri. As expected, the CMA-BLHs of Ri = 0.3 (0.2) are generally higher (lower) than those of Ri = 0.25, 5 however, the CMA-BLH based on various Ri is significantly correlated with each other (R = 0.99).
Figures 3c-3d present the uncertainties as a function of BLHs (Ri = 0.25). It is found that the 50th and 75th percentile values of the absolute uncertainties are < 0.05 km and < 0.1 km, respectively; and the 50th and 75th percentile values of the associated relative uncertainties are both < 5% for BLH (Ri = 0.25) > 2 km, and < 15 % for BLH (Ri = 0.25) < 2 km. As such, the uncertainty caused by the selection 10 of critical value is quite small.

Results and discussion
This section presents the basic BLH climatologies from the CMA radiosondes and ERA-Interim, including seasonal and diurnal variations. Intercomparisons of climatological results from radiosonde 15 observations with the ERA-Interim are made to better elucidate the BLH characteristics in China.
Finally we will explore the potential influences of meteorology on the observed annual cycle of BLH. provide information about the transition state, and the soundings of 0200 BJT capture the nocturnal boundary layer. In the next morning (around 0800 BJT), after sunrise the PBL experiences another 10 transition to the convective boundary layer.

Seasonal variation of CMA-BLH and ERA-BLH and their intercomparison
Since only the soundings of 0800 and 2000 BJT are launched conventionally, the seasonal variation of BLH is mainly investigated at these two moments (Fig. 5). At 2000 BJT, the CMA-BLH generally 15 follows a downward trend from spring to winter (Fig. 5b), that is, spring > summer > fall > winter. The climatologically strongest near-surface wind speed in most of China in spring (Guo et al., 2011;Zhao et al., 2009) and the intense solar radiation in summer (Miao et al., 2012(Miao et al., , 2015, which favor the development of boundary layer during these two seasons. In spring, the mean value of CMA-BLH is ~0.71 km, which is slightly higher than that of summer (~0.65 km) but can occasionally exceed 2 km (95th percentile value) in spring and summer.
In contrast, in fall and winter, the mean values of CMA-BLH are merely ~0.32 km, which are significantly lower than those of spring and summer. Such a huge difference of the seasonally averaged CMA-BLH between the warm seasons (spring and summer) and cold seasons (fall and winter) at 2000 5 BJT is well captured by the ERA-BLHs, although the ERA-BLHs tend to underestimate the springtime BLHs, and overestimate that of summer (Fig. 5b). The scatter plots of seasonally averaged CMA-BLHs and ERA-BLHs (Fig. 6), demonstrate the close correspondence between these two kinds of BLH, and give confidence that the seasonally averaged ERA-BLHs can be used as an alternative when the CMA soundings are unavailable. It is interesting to note that at 0800 BJT the CMA-BLH of spring is not 10 generally higher than that of summer although the near-surface wind speed is stronger in spring for both 0800 BJT and 2000 BJT (Figs. S1-S2 in Supplement). This is most likely due to the stronger nearsurface stability in spring (Fig. S3), which to some extent inhibits the development of PBL.
In terms of the spatial distributions of BLHs, there exists solar radiance difference received at various stations or grid points across China due to the large spatial range. For a sounding observation at a given 15 time (e.g., 0200, 0800, 1400, 2000 BJT), zonal evolution of BLHs have to be considered from the perspective of actual solar radiance received at the given area.
The spatial distributions of seasonally averaged CMA-BLHs and ERA-BLHs at 0800 BJT are presented in Fig. 7. At 0800 BJT, due to the stronger solar radiation in eastern China than that of western regions, higher BLHs can be distinctly seen over the eastern regions (Fig. 7). This BLH spatial 20 pattern is more prominent in spring and summer than in fall and winter . Comparing with the BLHs at 2000 BJT, it is found that the BLHs of 0800 BJT are generally lower (Figs. 4a and 5), and the seasonal variation of BLH is weaker. Despite these differences, the BLHs of 0800 BJT show a similar seasonal variation, that is, higher BLHs are in spring and summer. It is also intriguing to note that the highest ERA-BLH at 0800 BJT over the northern Tibet Plateau (TP) region occurs in winter (Fig. 7), which may be caused by the strongest near-surface wind over the northern TP region (Fig. S4). Since 5 there is no available sounding observation in the northern TP region, the seasonal variation of ERA-BLH in this region cannot be characterized in this study, which needs further field campaign.
As illustrated in Fig. 8, both CMA-BLHs and ERA-BLHs at 2000 BJT demonstrate a strong east-towest BLH gradient in the warm seasons (spring and summer), with higher BLHs over the western China ( Figs. 8a-b). The BLH spatial pattern may be caused by the different solar radiation in the west 10 (stronger radiation at an earlier Local Sidereal Time) and east of China at 2000 BJT.
In fall, it is interestingly to note that the east-to-west gradient of BLH is less prominent (Fig. 8c).
Likewise, the BLHs during winter in most regions of China are less than 0.3 km, except for the Tibet Plateau, in which the BLHs can exceed ~0.5 km at 2000 BJT.
Since the soundings of 0200 and 1400 BJT are only launched in summer, the seasonal variation of 15 BLH at these two moments cannot be investigated by using the CMA-BLH alone. Therefore, the seasonal variations of BLH at 0200 and 1400 BJT are roughly evaluated by using the seasonally averaged ERA-BLHs. It is found that at 1400 BJT the ERA-BLHs in spring and summer are higher than those of fall and winter ( Fig. 9). At 0200 BJT, most plain regions of China are covered by a relatively shallow boundary layer, without significant seasonal variation (Fig. 10).

Diurnal variation of BLH
In summer, four times per day (i.e. 0200, 0800, 1400, and 2000 BJT) soundings are available at most of the radiosonde sites, which allows us to investigate the summertime diurnal variation of BLH (Fig.   11a). With the intense solar radiation in summer, the boundary layer is more fully developed during the daytime. At 1400 BJT, more than half of the seasonally averaged CMA-BLHs are higher than 1 km (c.f., 5 Figs. 9b and 11a). Both the CMA-BLH and ERA-BLH demonstrate a south-to-north gradient of BLH, with higher BLH over the dry northern regions (Fig. 9b), implying that the hydrologic factors such as humidity, in addition to solar radiation, may play a role in modulating the spatial distribution of daytime BLH (Seidel et al., 2012).
After sunset, the convective boundary layer undergoes a transition to the nocturnal stable boundary 10 layer. Since the LST is earlier in the west of China than the east (15° longitude is equivalent to a onehour change in time), the evening transition of boundary layer structure is well manifested by the eastto-west gradient of BLH at 2000 BJT (Fig. 8b), although the effects of elevation on the spatial distribution of BLH cannot be ignored. During the nighttime, the BLH is less than 0.3 km over most China (Figs. 10b and 11a). Comparing the BLH at 0200 BJT with that of 1400 BJT, it is found that 15 diurnal amplitude of BLH is stronger in the north and west of China.
Similar to the spatial distribution of BLH at 2000 BJT, the BLH at 0800 BJT demonstrates the transition from nocturnal boundary layer to convective boundary layer (Fig. 7b), characterized by a west-to-east BLH gradient.
With the ground-based cloud cover observations obtained simultaneously at the same sounding sites, 20 the effects of cloud cover on the diurnal evolution of BLH in summer are investigated as well. As illustrated in Fig. 11b, during the daytime the development of PBL is typically suppressed due to less solar radiation received at the surface under cloudy condition. As a result, at 1400 BJT, the 25th, 50th and 75th percentile values of BLHs decrease by around 0.3 km. The higher mean and 75th percentile values of BLHs under clear condition could also be observed at 2000 BJT and 0200 BJT, which may be relevant to the larger heat storage of land surface. In contrast, a slightly thicker boundary layer could be 5 formed in the early morning under cloudy conditions.
Using the ERA-BLH as an alternative, the diurnal variations of BLH in the other three seasons are presented in the supplementary materials (Figs. S5-S7). It is found that the diurnal variation of BLH in spring (Fig. S5) is comparable to that of summer, while the diurnal variation of BLH in fall (Fig. S6) is less prominent. In winter (Fig. S7), due to the weak solar radiation, the development of BLH over the 10 plains of eastern China is severely suppressed, leading to the weakest diurnal variation of BLH during the four seasons over the plains regions.

Association of BLHs with related meteorological variables
As a preliminary effort to explore the potential influence of meteorology on the observed annual cycle 15 of BLH, we turned to ground-based weather observations and soundings that are simultaneously obtained from sites shown in Fig. 1. In particular, the association of CMA-BLH with four other atmospheric variables was evaluated using correlation analyses, including surface pressure, 10-m wind speed, near-surface temperature, and lower tropospheric stability (LTS). The LTS is defined as the difference in potential temperature between 700 hPa and the surface (Slingo, 1987), which can be used 20 to describe the thermodynamical state of the lower troposphere (Guo et al., 2016).
For instance, Figure 12 compares the annual cycles of CMA-BLH with these four meteorological variables in Beijing. The annual cycles of CMA-BLHs at 0800 and 2000 BJT are quite similar. In other words, both CMA-BLHs peak in May, and reach the minimum in October. The annual cycle of CMA-BLH is anti-correlated with the annual cycle of surface pressure (Fig. 12a), implying that the seasonal shift of large-scale high/low pressure systems can suppress/facilitate the development of boundary layer 5 through the associated large-scale descending/ascending motion (Liu et al., 2013;Medeiros et al., 2005).
The annual cycle of BLH is positively correlated with the annual cycle of near-surface wind speed and temperature (Figs. 12b-12c). In addition, the annual cycle of CMA-BLH is anti-correlated with the variation of LTS (Fig. 12d). Comparing with the near-surface temperature, the variation of LTS is in closer correspondence to the annual cycle of BLH. 10 Extending this analysis to the full set of radiosonde stations analyzed, Figure 13 presents the correlations among these variables at 0800 and 2000 BJT. Overall, results obtained at most sites in China are consistent with those at Beijing. In particular, the annual cycle of CMA-BLHs is anticorrelated with the variation of surface pressure and LTS (Figs. 13a, 13b, 13g, and 13h), and positively correlated with near-surface wind speed and temperature (Figs. 13c-13f). Generally, the correlations are 15 closer at 2000 BJT, and the correlations are weaker at some sites close to the southern coasts and the east of Tibet Plateau, probably due to the modulation of local geographical effects (Stull, 1988). The positive/negative correlations derived from the annual cycle of CMA-BLHs and these four meteorological parameters, can also be obtained based on the monthly mean values of CMA-BLHs and meteorological parameters (Figs. S8-S9). 20

Concluding remarks
As the first effort to obtain a sounding-based BLH climatology in China, we have made an extensive use of sounding data from the L-band radiosonde network operated and maintained by CMA from January 2011 to July 2015. The CMA-BLHs were also compared with coincident ERA-BLHs using the same bulk Richardson number method. 5 The mapping of BLH for the first 41 months of new-generation radiosonde network provides considerable insight into the large-scale spatio-temporal variations of boundary layer characteristics across China with continuous radiosonde data studied to date. Previously unknown facts about the boundary layer across whole China were first discovered. A complex pattern of the boundary layer has been revealed, exhibiting strong geographical, diurnal, and seasonal variations, corroborated by BLHs 10 derived from ERA-Interim. There are substantial influences by elevated terrain features and major landwater boundaries. The association of BLH with solar radiation is a compelling relationship when examining the results for the first 4 years of operation of radiosonde network.
The sounding observations at 0800 and 2000 BJT show that the BLH of spring and summer is generally higher than that of fall and winter. This annual cycle of BLH at most sites is found to be anti- The climatological diurnal cycle of BLH, revealed by the four-times-daily sounding in summer, shows 5 that BLH peaks in the afternoon and diurnal amplitude of BLH is higher in the north and west of China.
Comparing with other seasons, it is found that the diurnal variation of BLH is typically stronger in spring and summer than in fall and winter.
The climatological features of BLH in China revealed here have important implications for better understanding the formation mechanism of aerosol pollution, cloud and even extreme precipitation in 10 China. In the near future, other algorithms for deriving BLH merit a more comprehensive test based on the radiosonde network of China, apart from further scrutiny for the sources of uncertainty with regard to the data and methods applied in this study.