The impact of atmospheric dynamics on vertical cloud overlap over the Tibetan Plateau

The accurate representation of cloud vertical overlap in atmospheric models is particularly significant for predicting the total cloud cover and for the calculations related 25 to the radiative budget in these models. However, it has received too little attention due to the limited observation, especially over the Tibetan Plateau (TP). In this study, 4 years (2007–2010) of data from the CloudSat cloud product and collocated ERA-Interim reanalysis product were analyzed to examine the seasonal and zonal variations of cloud overlap properties over the TP region, and evaluate the effect of atmospheric dynamics on 30 cloud overlap. Unique characteristics of cloud overlap over the TP have been found. The statistical Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2017-642 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 1 August 2017 c © Author(s) 2017. CC BY 4.0 License.


Introduction
Clouds can cause considerable changes in the Earth's radiation budget, the global 50 hydrological cycle and large-scale atmospheric circulations via changes in their various macrophysical (e.g., cloud cover, height, and thickness) and microphysical properties (e.g., cloud phase and droplet and crystal size) (Rossow and Lacis, 1990;Hartmann et al., 1992;Stephens, 2005;Kawamoto and Suzuki, 2012;Yan et al., 2012;Wang et al., 2010).However, our incomplete understanding of their underlying physical processes makes the 55 representation of clouds in climate models still unreliable, which keeps clouds as the largest uncertainty when estimating and interpreting changes in the Earth's energy budget (Boucher et al., 2013).
The Tibetan Plateau (TP), which is also known as the "roof of the world" or the "world water tower", plays a significant role in determining global atmospheric 60 circulations, in addition to its strong influence over Asia via its thermal and dynamic forcings (Yanai et al., 1992;Ye and Wu, 1998;Duan and Wu, 2005;Xu et al., 2008;Wu et al., 2015).Specifically, the TP has experienced distinct climate changes over the past measurement is the best choice for identifying the vertical overlap properties of clouds 95 and thereby, better quantifying the radiative budget at the surface and its related climate change impacts (e.g., glacier degradation and frozen soil ablation) over the TP region.
Fortunately, the millimeter-wavelength cloud profiling radar (CPR) launched on CloudSat (Stephens et al., 2002) and the cloud-aerosol lidar with orthogonal polarization (CALIOP) (Winker et al., 2007) launched on CALIPSO (Cloud-Aerosol Lidar and 100 Infrared Pathfinder Satellite Observation) provide an unprecedented opportunity to investigate vertical cloud overlaps on a global scale (Barker et al., 2008;Kato et al., 2010;Li et al., 2011;2015;Tompkins and Di Giuseppe, 2015;Di Giuseppe and Tompkins, 2015).Based on two months of cloud mask profile information from the CloudSat and CALIPSO satellites, Barker (2008) quantified the properties of cloud overlap on a global 105 scale and identified a latitudinal dependence.Recently, Di Giuseppe and Tompkins (2015) further evaluated the impact of wind shear on the global-scale cloud overlap and identified an empirical relationship between the cloud overlap and wind shear for use in models by using 6 months of CloudSat-CALIPSO data.However, the related question of the cloud overlapping over the TP region has received too little attention, as has the 110 question of how the unique thermo-dynamical environment (e.g., the atmospheric stability) of the TP region affects cloud overlap.As such, this study mainly focuses on the impacts of various atmospheric states and large-scale atmospheric dynamics on cloud overlap over the TP region by combining the cloud cover profile information from the 2B-GEOPROF-LIDAR dataset (Mace et al., 2009) and the meteorological parameters 115 from the ERA-Interim reanalysis datasets (Dee et al., 2011).This paper is organized as follows: a brief introduction to all the datasets and methods used in this study is given in Section.2. Section 3.1 outlines the seasonal and zonal variations of the cloud overlap parameters over the TP region.Further analyses of the impacts of the variations in the atmospheric state and large-scale atmospheric dynamics 120 on cloud overlap are provided in Sections.3.2 and 3.3.Finally, the conclusions and discussion are presented in Section.4.

Datasets and methodology
In this study, 4 years (2007)(2008)(2009)(2010) (Mace and Zhang, 2014).For this study, we 135 only use the "CloudFraction" parameter from the aforementioned dataset.This parameter reports the fraction of the lidar volume within the radar resolution volume that contains hydrometeors (Mace et al., 2009;Mace and Zhang, 2014) and describes the cloud mask profile.Several previous studies have identified a cloudy atmospheric layer based on different thresholds of the lidar-identified cloud fraction, including a 99% (Barker, 2008; 140 Di Giuseppe and Tompkins, 2015) or 50% threshold (Haladay and Stephens, 2009;Verlinden et al., 2011).Although a threshold of 99% is used in this study, we also test the sensitivity of cloud overlap to the cloud fraction threshold.In addition, the ECMWF-AUX dataset (Partain, 2004), which is an intermediate dataset that contains the set of ancillary ECMWF state variable data interpolated across each CloudSat CPR bin, 145 are also used to provide the pressure and height information of each vertical bin in the cloud mask profile.The vertical and horizontal resolutions of these products are 240 m and 1.1 km, respectively.

Meteorological reanalysis dataset
The daily 6-hourly dataset from the ERA-Interim reanalysis (Dee et al., 2011), which 150 has a grid resolution of 0.25°×0.25°, is used to characterize the atmospheric dynamics over the TP.For each cloud mask profile in the 2B-GEOPROF-LIDAR dataset, the vertical profiles of the zonal wind u, meridional wind v, relative humidity rh, specific humidity sh and atmospheric temperature T most closest to the cloud observations in both space and time are extracted and further interpolated in the vertical direction to match the 155 bin numbers and sizes of the cloud mask profile.Then, we follow the method used by Di Giuseppe and Tompkins (2015) to project the u and v winds of every vertical bin onto the satellite overpass track, averaging in the along-track direction for all profiles in the selected CloudSat data segment (see section 2.3) to derive the scene-average, along-track horizontal wind V and the corresponding wind shear.Here, we define the wind shear 160 , ij dV dz between the layers i and j, which is used to calculate the cloud overlap parameter as follows: , , max{ ; } min{ ; } where V i and V j are the horizontal winds at layers i and j, respectively, and D i,j is the layer separation distance.For the CloudSat overpass track (Fig. 1a), Di Giuseppe and where is the potential temperature, L v is the latent heat of vaporization, r s is the saturation mixing ratio, C p is the specific heat capacity at a constant pressure, and T is the atmospheric temperature.The smaller the , es i j z , the more unstable the atmosphere. 175 Additionally, the scene-averaged vertical velocity at 500 hPa is extracted from the ERA-Interim reanalysis to analyze the impact of vertical motion on cloud overlap.mask over the TP region derived from the 2B-GEOPROF-lidar dataset.This cloud mask includes eight, four, two and one segments with lengths of 25, 50, 100 and 200 km, respectively.For each segment, the average cloud cover profile is derived first.Then the vertical overlap between any two atmospheric layers in this profile is calculated when the 185 cloud coverages (C i and C j ) of both layers exceed 0. Following the examples of previous studies (Hogan and Illingworth, 2000;Di Giuseppe and Tompkins, 2015), we consider nonadjacent layers to be a continuous cloud pair when all layers between them are classified as cloudy.Otherwise, these layers are classified as a discontinuous cloud pair (Hogan and Illingworth, 2000;Di Giuseppe and Tompkins, 2015).Using various 190 assumptions of overlap, the total cloud coverages of any two cloud layers are given by ,

Retrieval of the overlap parameter
Further, we use the definition of Hogan and Illingworth (2000) to quantify the degree of cloud overlap using the "overlap parameter" , which is calculated using the following formula: The overlap parameter can thus be written as follows: , , max , , Eq. 4 indicates that the parameter ranges from 0 (random) to 1 (maximum) when the total observed cloud cover falls between the values derived using the maximum and 200 random overlap assumptions.However, may be negative when the degree of cloud overlap is lower than that predicted by the random overlap assumption.pairs.However, this sensitivity decreases when the cloud fraction threshold exceeds 50%.
Given these results, a cloud fraction threshold of 99% is used in the following sections.
Using this cloud fraction threshold (99%), we test the sensitivity of to the spatial sampling scale for both discontinuous and continuous cloud pairs (Fig. 3a and Fig. 3c).
Many previous studies have used ground-and space-based radar to verify the validity of 210 the random overlap assumption for the study of vertically discontinuous clouds (Hogan and Illingworth, 2000;Mace at al., 2002;Naud et al., 2008;Di Giuseppe and Tompkins, 2015).Our results show that the degree of cloud overlap of the discontinuous clouds over the TP region is lower than that of random overlap.In Fig. 3a, almost all the values are negative at the given cloud fraction threshold (99%) and fall between -0.2 and 0. Thus, 215 the total cloud coverage is still slightly underestimated for discontinuous cloud pairs when using the random overlap assumption.Assuming a cloud layer distance of less than 9 km, increases as the spatial sampling scale increases (e.g., from 25 km to 200 km).
This correlation indicates that a maximum overlap is more common for a larger domain, which is consistent with the results of previous studies (Hogan and Illingworth, 2000; 220 Oreopoulos and Khairoutdinov, 2003;Oreopoulos and Norris, 2011).The obvious dependence of on the spatial sampling scale is also observed for continuous cloud pairs when they are separated by a distance of less than 9 km (Fig. 3c).For a continuous cloud pair, we found that the value of decreased from 0.92 to -0.35 and then slightly increased with increasing distance.When the cloud layer distances exceeded 1.5 km, the 225 degree of cloud overlap was even lower than that seen for random overlap.Our overlap parameters for the TP are clearly different from the globally averaged values derived by Di Giuseppe and Tompkins (2015) although the same datasets (with only 6 months of data) were used.Previous studies have reported the existence of a negative when the cloud layer separation distances are greater than 2km, as derived using ground-based 230 radar at different sites (Naud et al., 2008).This result is most often seen for discontinuous cloud layers.The related results also indicated that and its corresponding decorrelation length are very sensitive to the spatial sampling scale (Hogan and Illingworth, 2000;Mace at al., 2002;Naud et al., 2008;Oreopoulos and Norris, 2011).
However, it is obvious that the more negative values seen in our statistical results are 235 related to the degree of cloud overlap over the TP, which is more prone to minimal overlap as the cloud layer separation distances exceed 1.5 km.Tompkins and Di Giuseppe (2015) provided some heuristic discussions of the negative values and sampling scale dependence and attributed these phenomena to a data truncation error.Simply, these authors theorized that some overcast or single cloud layers will be removed 240 from the samples when the spatial sampling scale is smaller than the cloud system scale, thus biasing and its decorrelation length.Given a spatial sampling scale of 50 km, the ratio of the sampling scale to the cloud system scale decreases strongly from the equator to the poles because many of the frontal cloud systems of the middle and high latitudes are larger than the convective cloud systems over the tropics.Thus, regional atmospheric 245 models should account for the typical cloud system scales when they are applied at a fixed horizontal resolution.Based on these arguments, Tompkins and Di Giuseppe (2015) suggested a simple filter for the continuous cloud pairs, so that only measurements with cloud coverages smaller than a given threshold of 50% are retained to reduce the inversion errors of and its corresponding decorrelation length caused by data resolution for the CPR measurements, the along-track scale (S in km) of a cloud system is S=N×1.1 (Zhang et al., 2014;Li et al., 2015).Thus, the probability of a small-scale cloud system decreases with increasing height.On average, the horizontal scale of a cloud system at a height of 15 km (mean value: 66.4 km) is almost ten times greater than 260 that at a height of 2 km (mean value: 6.7 km).From the statistical results shown in Fig. 3b, we can infer that a spatial sampling scale of 50 km will result in larger retrieval errors of at higher atmospheric heights above the lower troposphere.Thus, to reduce the uncertainty caused by data truncation, we follow the suggestion of Tompkins and Di Giuseppe (2015) and apply a data filter, retaining only the cloud samples with cloud 265 coverages of less than 50%.Given the cloud fraction threshold (99%), spatial sampling scale (50 km) and upper limit of cloud cover (50%), the number of available samples is at least one million, thus ensuring statistical significance.both exhibit obvious seasonal variations as well.For layer distances below 4 km, the atmospheric stability and wind shear gradually decrease from January to August and then steadily increase (see Figs. 4b,4c,4e and 4f).These trends may be associated with the unique topographical forcing and thermo-dynamical environment of the TP.The TP is usually considered to be an atmospheric heat source during the summer (Wu et al., 2015).(Hogan and Illingworth, 2000;Mace and Benson-Troth, 2002).Naud et al. (2008) found that radar measurements 310 from multiple sites and the corresponding NCEP reanalysis dataset indicated that a more unstable atmosphere tends to favor a maximum overlap over a random one.Indeed, we also find that an unstable atmospheric stratification over the TP and its corresponding weak wind shear often lead to a greater when the layer distance is less than 4 km.

Seasonal and zonal variations of the overlap parameter for continuous clouds
However, Figs. 4e and 4f also reveal an inconsistency in the relationship of layer distance 315 and the corresponding wind shear and atmospheric stability.For example, we can see that the wind shear for a 2-km layer distance is greater than that for a 3-km distance, but the atmosphere is also more unstable.This difference is probably because two cloud layers with the same separation but different altitudes are sorted into the same statistical group.
Generally, the seasonal variation of is correlated with the concurrent large-scale 320 dynamics and atmospheric states.
Additionally, Fig. 5 shows the zonal variations of , es z and dV dz over the TP.Figs.5a and 5d clearly indicate that and the related layer distances are larger to the south of the TP and smaller to the north.Their zonal variations can be attributed to the zonal differences of the atmospheric dynamics of the region (see Fig. 5b, 5c, 5e and 5f).

325
Convective activity is more prevalent over the southern part of the TP than over the northern part (Fujinami and Yasunari, 2001).Thus, larger values of can be found to the south.Based on the statistical results shown in Figs. 4 and 5, we confirm the importance of the impact of atmospheric dynamics on cloud overlap.Although is different than the multilayered cloud fraction that describes the ratio of the number of 330 multilayered cloud profiles to the number of total sample profiles in a given grid box (Li et al., 2015), we found that the meteorological factors have similar effects on the  reaches 2 km.Given the same layer distances (e.g., 1.5 km), we observe differences of more than 0.5 in the under unstable and stable atmospheric conditions.For discontinuous clouds, the variations of with changes in layer distances do not follow the same inverse exponential model; however, the impact of atmospheric stability is obvious.
When the upper limit threshold of cloud cover is changed from 100% to 50% based on 370 the suggestion from Tompkins and Di Giuseppe (2015), the sensitivity of to changes in the large-scale atmospheric dynamics is significantly reduced, especially for changes in (Fig. 6b).Figs.6b and 6d clearly show that increases when the upper limit of cloud cover is reduced to 50%, especially given a stable atmosphere ( 5 It is possible that this increase results from the weakening of the dynamic factors on the sensitivity of to the 500 hPa vertical velocity (see the Fig. s3 in the supplementary material).Although the seasonal cycle of is related to the local vertical velocity, we found that the sensitivity of is rather weak for all vertical velocity bins.Naud et al.

395
(2008) also found a weak sensitivity of to vertical velocity, which they attributed to the uncertainty in the reanalysis dataset.Even when using the lower cloud cover threshold (50%), the proportion of negative values is still high (increasing to 41% from the 39.8% for the cloud cover threshold of 100%) for cloud samples with smaller layer distances (<3.5 km).Such a high proportion means that the contribution of negative to 400 the cloud overlap properties over the TP cannot be ignored in the parameterization of .

Treatment of negative overlap parameters and parameterization
As we know, the overlap parameter for continuous cloud layers can be modeled as an inverse exponential function: / DL e (6) where D is the layer separation distance and L is the decorrelation length scale that characterizes the transition from the maximum to random overlap assumption.In previous studies, it is found that L depends on the vertical and temporal resolutions of the lidar data or the spatial scale of the models, and the values of L have varied across many studies (Hogan and Illingworth, 2000;Mace and Benson-Troth, 2002;Pincus et al., 2005).

410
For example, Barker (2008) found a wide range of L values, with a median value of 2 km, when analyzing the global CloudSat and CALIPSO datasets.Oreopoulos and Norris (2011) derived L based on radar measurement taken over the Southern Great Plains (SGP) of the USA.Their results indicated that the value of L ranges from 2 to 4.5 km across different seasons and that smaller spatial scales correspond with smaller L values.In 415 other studies, L was usually a function of latitude or total cloud cover (Shonk et al., 2010;2014;Yoo et al., 2014).In our study, however, an obvious drawback of Eq. ( 6) is that the negative values of cannot be parameterized in the inverse exponential modeling.
Oreopoulos and Norris (2011) also noted that the negative values of in their dataset were too frequent for the use of inverse exponential modeling to be robust.

420
Our dataset show that the proportion of negative values decreases as the averaged cloud height of the two cloud layers increases, except for clouds with a mean height exceeding 15 km (Fig. 7).Thus, the overlap parameter for the lower atmosphere of the As with Eq. ( 5), new can also be written as: As new is always larger than 0, Eq. ( 8) allows its simple parameterization through inverse exponential modeling.Table 1 lists several parameterization schemes for the decorrelation length scale L. Schemes 4, 5 and 6 are derived by considering L as a function of wind shear or atmospheric stability based on both a multiple linear regression 440 method and our statistical dataset.The regression formula of L and L 1 can be written as: Here, L , 1 L , b1 ,b2, c1 and c2 are the fitting parameters.In Table 1, scheme 7 is the previously mentioned scheme from Di Giuseppe and Tompkins (2015), i.e., using the global CloudSat-CALIPSO cloud data and ECMWF reanalysis dataset to parameterize L 445 as a function of wind shear.The method of Shonk et al. (2010) was also used, which proposed an empirical linear relationship between L and latitude (see scheme 8).For detailed information about each scheme, see the descriptions in Table 1.Note that the R-squared values (R 2 ) for schemes 4, 5 and 6 are 0.98, 0.44 and 0.94, respectively.Fig. 8 presents the zonal differences between the calculated and observed cloud 450 covers for the aforementioned schemes and indicates that the maximum (scheme 2) and minimum overlap assumptions (scheme 3) result in large cloud cover biases at most of the layer distances, especially for layer distances greater than 1 km and less than 2 km (where the bias exceeds 5%).The bias of the random overlap assumption (scheme 1) is greater (>5%) when the layer distance is less than 1 km, although a lower bias may be 455 found for greater distances.As in scheme 2 (the maximum overlap assumption), scheme 7 significantly underestimates the cloud cover for layer distances above 1 km, whereas the obvious underestimation (e.g., >5%) of scheme 8 only exists for layer distances between 2 and 3 km.Notably, the large bias of scheme 7 may occur because the scheme is based on the global CloudSat-CALIPSO measurements and ECMWF reanalysis dataset for a 460 short period (January-July 2008); as such, some obvious regional or seasonal cloud overlap properties are easily obscured by global averaging.Another possible cause is that the other dynamic factors (e.g., the atmospheric stability) were not considered in the parameterization of this scheme.If we replace the negative values with zero and parameterize L as a function of wind shear and atmospheric stability (see scheme 4),

465
scheme 4 still results in some underestimations, especially for layer distances between 2 and 3 km, as seen in Fig. 8.However, when we consider the true cloud overlap as the result of the linear combination of the maximum and minimum overlaps (schemes 5 and 6) and the influence of wind shear, the bias of scheme 5 is reduced to±2%.In comparison, scheme 6 has a relatively lower bias at large layer distances but higher 470 R-squared values (R 2 =0.94).These results indicate that the effects of both wind shear and atmospheric stability on cloud overlap should be accounted for in the parameterization of the overlap parameter over the TP, and the use of Eq. ( 8) can considerably reduce the cloud cover bias caused by the negative values.

475
The "roof of the world", i.e., the Tibetan Plateau, has experienced obvious climate changes over the past three decades, and these climate changes are partially related to the cloud variations over the TP region.Vertical cloud overlap, which is one of the most important macro-physical properties of clouds, is particularly significant for predicting the total cloud cover and for the calculations related to the radiative budget in weather Our results show that a random overlap assumption slightly underestimates the total cloud coverage of discontinuous cloud layers over the TP, whereas the overlap parameter for continuous cloud layers has sharply transitions from a maximum to random overlap 490 with an increase of cloud layer distance, eventually trending towards a minimal overlap (e.g., negative values) as the cloud layer separation distance exceeds 1.5 km.As seen in previous studies (Hogan and Illingworth, 2000;Mace at al., 2002;Naud et al., 2008;Oreopoulos and Norris, 2011), the overlap parameter over the TP is sensitive to the spatial sampling scale.However, recent research has attributed some of these phenomena 495 (e.g., the negative values and sampling scale-dependence) to data truncation errors, which result from using a sampling scale smaller than the cloud systems.To reduce the errors in and its corresponding decorrelation length, we followed the suggestion of Tompkins and Di Giuseppe (2015) and applied a simple filter for the instances of continuous cloud layers, retaining only the cloud samples with an upper limit of cloud

525
Previous studies have pointed out that precipitation may bias the cloud overlap statistics toward maximum overlap (Mace et al., 2009;Di Giuseppe and Tompkins).
However, the present study doesn't eliminate the influence of precipitation on the overlap parameter, which means that overlap parameter will be smaller (or more negative) if cloud samples with precipitation are removed from our analysis.In addition, the future 530 work would be to use the regional climate models to evaluate the impact of new scheme on the radiative budget compared with existing overlap parameterization in models.
Cloud overlap parameter is significantly controlled by atmospheric dynamics, therefore the long-term variations of meteorological factors are bound to affect the trend of cloud overlap and the following total cloud cover and radiation budget.Indeed, a recent study 535 has shown that rapid warming and an increase of atmospheric instability over the TP leads to more frequent deep clouds, which are responsible to the reduction of solar radiation over the TP (Yang et al., 2012).In view of the decreasing trend of total cloud cover over this region (You et al., 2014), we may infer that the degree of cloud overlap over the TP possibly has an increasing trend.However, as stated by the Tompkins and Di 540 Giuseppe (2015), true cloud overlap parameter will be biased at a given spatial sampling scale if one doesn't consider the cloud system size.Thus whether the trend is true or   4), but replaces the negative with 0; Meantime, decorrelation length scale L is parameterized as a function of atmospheric stability and wind shear; 2 Scheme 5 and 6 both are based on Eq. ( 7), and decorrelation length scale L is parameterized as a function of atmospheric stability or wind shear; 3 Scheme 7 is based on Eq. ( 4), and decorrelation length scale L is parameterized as a function of wind shear; 4 Scheme 8 is based on Eq. ( 4), and decorrelation length scale L is parameterized as a function of latitude; (please see section 3.3)         1) and its variation with layer distance.
of data from CloudSat 2B-GEOPROF-LIDAR, ECMWF-AUX and the daily 6-hour ERA-Interim reanalysis were collected to analyze 125 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License. the impacts of atmospheric dynamics on the cloud overlap over the TP (27°N-39°N;78°E-103°E) region as demonstrated in Fig. 1a.2.1 Satellite datasets Radar signals can penetrate the optically thick layers that attenuate lidar signals significantly, and lidar signals may sense the optically thin hydrometeor layers that are 130 below the detection threshold of radar signals.Thus, by combining the unique complementary capabilities of the CPR on CloudSat and the space-based polarization lidar (CALIOP) on the CALIPSO satellite, the 2B-GEOPROF-LIDAR dataset can produce the most accurate quantitative descriptions of the locations of the hydrometeor layers in the atmosphere on the global scale Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.

165
Tompkins (2015)  indicated that the cross-track shear of the zonal wind u has little statistical significance.Similarly with the wind shear, we calculate the vertical gradient of the saturation of the equivalent potential temperature ( the same two layers to quantify the dependence of the cloud overlap on the degree of the conditional instability of the Fig. 2 tests the sensitivity of cloud overlap to the chosen cloud fraction threshold by using discontinuous and continuous cloud pairs with a 50-km spatial sampling scale.The value of decreases as the cloud fraction threshold increases, especially for discontinuous cloud 205 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.
Fig. 3b depicts the probability distribution functions (PDFs) of the horizontal scales of the along-track cloud systems at different heights over the TP region.Here, the horizontal scale of a cloud system at a given height along the CALIPSO/CloudSat track is determined by calculating the number of continuous cloud profiles (N) when each profile Fig. 3d shows the number of samples and the percentages at different cloud layer distances for both discontinuous and continuous clouds.These results clearly show that the proportion of cloud samples with 270 Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.smaller layer distances (e.g., 4 km) accounts for 95% of all samples of continuous clouds.Notably, cloud fraction and cloud cover are different variables in our study."Cloudfraction"reports the fraction of lidar volumes in the radar resolution volume that contains hydrometeors and is used to identify a cloudy based on the chosen threshold, which is 99% in this paper.However, the segment-averaged cloud coverages at certain 275 heights represent the percentage of clouds in a given spatial sampling scale.Given the 1.1 km along-track resolution of the CPR measurements and a spatial sampling scale of 50 km, the segment-averaged cloud coverage implies that the cloud cover is resolvable to approximately 2%(Di Giuseppe and Tompkins, 2015).In addition, to considering the impact of cloud overlap on the total cloud cover and solar radiation and to minimize the 280 uncertainty of the statistical results from the surface contamination of the CPR, only datasets from 1 km above the TP surface during the daytime were used to perform the analyses in this investigation.

Figure 4a ,
Figure 4a, 4b and 4c show the seasonal variations for the continuous cloud-pair related pentad-averaged , the degree of conditional instability of the moisture convection

300
Additionally, at the start of the Asian summer monsoon, abundant water vapor converges, Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.which intensifies the atmospheric instability of the TP region when combined with the enhanced surface heating (Taniguchi and Koike, 2008).This process further promotes the development of convective clouds.Although similar seasonal variations of es z occur at different layer distances, Fig. 4e clearly shows a trend of increasing atmospheric 305 instability between the two cloud layers with decreasing layer distance.Using ground-based radars, previous studies have shown that the cloud overlap parameters at different locations exhibit obvious seasonal variations.These studies have attributed the seasonal variations to the effects of meteorological conditions Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.multi-layered cloud fractions.That is, an unstable atmosphere favors both a greater and a greater multilayered cloud fraction, such that the percentage of multilayered cloud systems is also higher over the southern part of the TP.Yuan et al. (2013) provided a 335 detailed analysis of the global characteristics of the overlap of low and high altitude clouds through the use of combined active satellite datasets.These authors found that approximately 30% of the global distribution of low altitude clouds is overlapped by high altitude clouds, and the overlap rate (i.e., the multilayered cloud fraction) exhibits a strong anti-correlation with the 700 hPa and 500 hPa vertical velocities over the ocean 340 (with a correlation coefficient of -0.94).For comparison, we analyzed the seasonal and zonal variations of the 500 hPa vertical velocity over the TP (see the Figs.s1 and s2 in the supplementary material).As with the atmospheric stability, strong updrafts are more likely to occur over the southern part of TP (except the region north of 36°N), especially during the summer season, with the opposite seasonal and zonal changes as those 345 associated with .3.2 Impacts of large-scale dynamics on the cloud overlap over the TP regionIn this section, we investigate the variations of with changes in layer distance for different meteorological conditions to facilitate the parameterization of for cases of continuous clouds.Here, each meteorological factor over the TP region is grouped into 350 one of four bins based on its values.In the present study, the four bins for statistically significant number of samples fall within each bin (i.e., at least 355 one hundred thousand samples per bin).

Fig. 6
Fig. 6 illustrates the variations of with layer distance for different large-scale atmospheric dynamics and different upper limit thresholds of cloud cover.When considering the proportion of cloud samples with small layer distances (<3.5 km), which account for 92% of all samples with continuous clouds, we only provide the statistical

375
scale of the cloud system for instances with lower cloud cover.As shown in Figs.6b and 6d, the effects of the dynamic factors on are obvious, although the sensitivity of is reduced relative to that seen in Figs.6a and 6c.Compared with es z , the dependence of on dV dz seems weak, especially in the presence of discontinuous clouds.Naud et al. (2008) tested the sensitivity of to wind shear at three sites and 380 found that wind shear slightly affects when the layer distance is larger than 2 km.However, in a recent study, Di Giuseppe and Tompkins (2015) demonstrated the important effect of wind shear on the global cloud overlap by using a combination of the CloudSat-CALIPSO cloud data and the ECMWF reanalysis dataset.These results and our own results demonstrate that variations in cloud overlap are closely related to 385 atmospheric dynamics, but their relationship is not stable and has spatial variations.Shortly, the effect of the atmospheric stability in cloud overlap may be more important over convective regions (e.g., the intertropical convergence zone and TP during summer season), while the effect of wind shear may be dominant over the mid-latitudes.Given continuous cloud layers over the TP, Figs.6b and 6d clearly illustrate that the effects of 390 wind shear and atmospheric stability on cloud overlap should both be accounted for in the parameterization of over the TP.In addition to these two factors, we also test the es z Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.

480
Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.forecasting and climate models.However, the study of cloud overlap over the TP region has received less attention than necessary because of the limited regional cloud observations and our incomplete knowledge of the physical processes of clouds.In this study, we collect 4 years (2007-2010) of data from the CloudSat 2B-GEOPROF-LIDAR dataset, the ECMWF-AUX dataset and ERA-Interim daily 6-hourly reanalysis to analyze 485 the seasonal and zonal variations of cloud overlap over the TP region and to further evaluate the effects of atmospheric dynamics on cloud overlap.

500
scheme for quantifying the degree of cloud overlap through a linear combination of the maximum and minimum overlaps, and further parameterize L as a function of the wind shear and atmospheric stability.Compared with other parameterizations, this new scheme considerably reduces the bias in the predicted cloud coverage over the TP.Although the parameterization method derived in our study focuses on only the TP and may therefore Atmos.Chem.Phys.Discuss., https://doi.org/10.5194/acp-2017-642Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 1 August 2017 c Author(s) 2017.CC BY 4.0 License.significant over different parts of TP still needs further quantified because related cloud system scales over these regions are possibly changed as well.

Figure 3 .
Figure 3. (a), (c) The sensitivity of to the spatial sampling scale for non-continuous

765Figure 4 .
Figure 4. (a),(b) and (c) The seasonal variations of the pentad-averaged cloud overlap parameter , degree of conditional instability to moist convection es z and wind shear dV dz for the continuous clouds over the TP ; (d), (e) and (f) The seasonal variations of the pentad-averaged , es

Figure 6 .
Figure 6.The variation of overlap parameter with layer distance under different large-scale dynamics and upper limit thresholds of cloud cover (100%) for (a) and (c); (50% cloud cover threshold) for (b) and (d). 780

Figure 7 .
Figure 7.The proportions of negative at the different layer distances and averaged cloud heights of two cloud layers for the continuous clouds.

Figure 2 . 845 Figure 3 .
Figure 2. The sensitivity of cloud overlap parameter to cloud fraction threshold for non-continuous (left) and continuous (right) cloud pairs on a 50 km spatial sampling scale.

Figure 4 .
Figure 4. (a),(b) and (c) The seasonal variations of the pentad-averaged cloud overlap parameter , degree of conditional instability to moist convection es z and wind shear dV dz for the continuous clouds over the TP ; (d), (e) and (f) The seasonal

Figure 6 .
Figure 6.The variation of overlap parameter with layer distance under different large-scale dynamics and upper limit thresholds of cloud cover (100%) for (a) and (c); (50% cloud cover threshold) for (b) and (d). 875

Figure 7 .
Figure 7.The proportions of negative at the different layer distances and averaged cloud heights of two cloud layers for the continuous clouds.

Figure 8 .
Figure 8.The zonal difference of cloud cover between calculated and observed for different schemes (see the Table1) and its variation with layer distance.

Table 1 .
Several parameterizations of decorrelation length scale L from the exponential fit 740 as a function of atmospheric stability es z , wind shear dV dz or latitudeΦ
) and its variation with layer distance.885Atmos.