Thermal contrast between the Tibetan Plateau and tropical Indian Ocean and its relationship to the South Asian summer monsoon

Zhngqun Li , , Ziniu Xio , ,

a State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences,Beijing, China

b College of Earth and Planetary Sciences, University of Chinese Academy of Sciences, Beijing, China

Keywords:Tibetan Plateau Indian Ocean Thermal contrast South Asian summer monsoon Monsoon onset

ABSTRACT The land-sea thermal contrast is an important driver for monsoon interannual variability and the monsoon onset.The thermal contrast between the Tibetan Plateau and the tropical Indian Ocean at the mid-upper troposphere is proposed as a thermal contrast index (TCI) for South Asian monsoon. The authors investigate the TCI associated with South Asian summer monsoon (SASM) intensity and SASM onset. It is observed that the TCI considering the Tibetan Plateau and tropical Indian Ocean demonstrates a stronger and closer correlation with SASM intensity(0.87) than either the Tibetan Plateau (0.42) or tropical Indian Ocean ( ? 0.60) singly. It is implied that the TCI could preferably represent the impact of land-sea thermal condition on SASM activity. Further analysis reveals that the evolution of TCI is related to the SASM onset. The TCI is almost always larger in early onset years than it is in late onset years during the period before SASM onset. In addition, the change of the pentad-by-pentad increment of TCI leads the SASM variation. The correlation coefficient between the TCI increment and SASM index reaches a maximum when the TCI increment leads by 15 pentads. The results of this study show that the TCI plays an important role in SASM activities and is a potential indicator for SASM onset forecasting.

1. Introduction

The thermal contrast between the Asian continent and its adjacent oceans to the south is the basic driver of South Asian summer monsoon(SASM), with an increased land-sea thermal contrast resulting in an enhanced SASM circulation ( Li and Yanai, 1996 ). The importance of the thermal condition of the Tibetan Plateau and the Indian Ocean in driving the establishment of SASM has been recognized previously ( Yanai et al., 1992 ; Minoura et al., 2003 ). Zhang et al. (2017) explored the thermal difference between the southern Indian Ocean and the South Asian region through the atmospheric heat source. The results indicated that the outbreak time of the Indian summer monsoon lags behind the time of thermal conversion.

The Tibetan Plateau plays a role as “the world’s water tower ”, exerting significant influence on atmospheric circulation through its mechanical and thermodynamic effects ( Duan and Wu, 2005 ; Xu et al., 2008 ).More specifically, it plays an important role in the activity and evolution of the SASM through directly heating the mid-upper troposphere, which results in a strong thermal contrast between the Tibetan Plateau and the surrounding atmosphere ( Wu et al., 2007 ). Zhao et al. (2018) found that the thermal forcing of the Tibetan Plateau is altered by the Indian Ocean sea surface temperature anomaly in the late spring on an interannual timescale. The Indian Ocean sea surface temperature also plays an important role in the monsoon onset and evolution, and the regional climate ( Nayagam et al., 2013 ). Chakravorty et al. (2016) considered that the persistent tropical Indian Ocean warming induces positive precipitation anomalies in the withdrawal phase of monsoon by changing the atmospheric circulation and modulating the water vapor flux.

Many previous works have studied the effect of the land-sea thermal contrast on the East Asian summer monsoon (EASM) ( Peng et al., 2000 ;Zhu et al., 2000 ; Cheng et al., 2008 ). It has been shown that the EASM activity is closely linked to the land-sea thermal gradient between the Asian continent and its adjacent Pacific Ocean. Some studies have emphasized that the seasonal transition of the thermal difference between the East Asian continent and the West Pacific could represent the onset of the EASM ( Qi et al., 2008 ). Hu and Duan (2015) concluded through investigating the relative contributions of the summertime Indian Ocean basin mode (IOBM) and Tibetan Plateau thermal forcing to the interannual variability of the EASM circulation system that both the Tibetan Plateau thermal forcing and the IOBM help to enhance the EASM. The Tibetan Plateau is one of the most important heat sources located in the north of the South Asian monsoon region. Although the influences of both the Tibetan Plateau and the Indian Ocean have been studied in many works, the impact of the thermal contrast between the Tibetan Plateau and the Indian Ocean on South Asian monsoon is still in need of further study. In this paper, we propose a Tibetan Plateau-Indian Ocean thermal contrast index (TCI) and explore its interannual variability. Furthermore, its representativeness and potential predictability for South Asian monsoon are investigated.

2. Data

The data used in this study include daily-mean wind at 850 hPa and 200 hPa, and air temperature from 500 hPa to 200 hPa, which are obtained from the European Centre for Medium-Range Weather Forecasts(ECMWF) interim reanalysis (ERA-Interim), with a 1°×1° horizontal resolution and covering the period from 1979 to 2017 ( Dee et al., 2011 ).In addition, the monthly outgoing longwave radiation (OLR) data on a 2.5° square grid, obtained from the National Center for Atmospheric Research, ranging from 1979 to 2017, are used ( Liebmann and Smith,1996 ). The OLR is used in this paper as an indicator of convective activity.

We applied the SASM index definition used by Webster and Yang (1992) (hereinafter referred to as the WYI) in this study. The WYI describes the differences between 850 hPa (

U

) and 200 hPa(

U

) zonal winds averaged over (0°-20°N, 40°-110°E), expressed as WYI =

U

?

U

(0°-20°N, 40°-110°E). For the SASM onset date (hereinafter referred to as SASMOD), the criteria defined by Wang et al.(2009) was used in this study. Based on the 850 hPa zonal winds averaged over (5°-15°N, 40°-80°E) as an onset circulation index (OCI), the SASMOD is defined as the first day when OCI exceeds 6.2 m s, with the provision that the OCI in the ensuing consecutive six days also exceeds 6.2 m s( Wang et al., 2009 ).The confidence level of the lead-lag correlation is evaluated using the two-tailed Student’s

t

-test, and the effective degrees of freedom(

N

) is evaluated based on the work of Davis (1976) , Chen (1982) , and Bretherton et al. (1999) .For simultaneous correlation (i.e., lag-0), the

N

is calculated by the following formulas ( Davis, 1976 ; Chen, 1982 ):

where

n

is the sample size.

T

is the integral time scale.

τ

is the time interval and

R

and

R

are the autocorrelations of two time series. An asterisk () indicates normalization.For lead-lag correlation, the

N

is evaluated following Bretherton et al. (1999) :

where

n

is the sample size and

r

and

r

are the lag autocorrelations of two time series, respectively.

3. Results

3.1. Definition of the new TCI

The SASM is primarily a tropical summer monsoon. As a direct dynamic response to the diabatic heating, the difference between upper and lower-layer winds can be closely linked to the strength of the heat source. Studies have shown that the upper-layer thermal contrast is more important for the SASM ( Sun and Ding, 2011 ). Dai et al. (2013) also found that summer thermal structure shows a larger land-sea thermal gradient in the upper than in the lower troposphere. Zhao et al.(2015) proposed an upper-level circulation index, which can successfully monitor and predict the EASM. It implies that the upper troposphere may play a predominant role in driving the Asian summer monsoon circulation. Thus, it is valuable to find an index to conveniently explore the thermal influence on South Asian monsoon variability. In the following, the temperature at the mid-upper troposphere (500-200 hPa) level is employed for analysis of the thermal contrast.

To start with, according to the WYI defined by Webster and Yang(1992) , the SASM index (SASMI) for the area (5°-25°N, 70°-100°E) was computed, which covers the region from the Indian subcontinent to the Bay of Bengal and east to the Indochina peninsula, as indicated by the red box in Fig. 1 (a). During the SASM season, the southwesterly flow through the region to the Chinese continent significantly impacts the weather and climate there. Secondly, the 500-200 hPa temperature is averaged by the air temperature of 500 hPa, 450 hPa, 400 hPa, 350 hPa,300 hPa, 250 hPa, 225 hPa, and 200 hPa. Then, the correlation between the 500-200 hPa average air temperature and the SASMI in June-July-August (JJA) is calculated, as shown in Fig. 1 (a). A distinct positive center of large values of correlation coefficient and a negative one could be found located south and north of the SASM region, respectively ( Fig.1 (a)). It is indicated that the thermal conditions of the two regions are important and sensitive to the change of SASM activity. Taking the difference between the mean 500-200 hPa average air temperature over the north region (25°-38°N, 65°-95°E) and that over the south region(5°S-8°N, 65°-95°E) to represent the thermal contrast intensity, a new index, TCI, is proposed and calculated as follows:

where Avg represents the regional average and

T

is the mean 500-200 hPa air temperature. The highly correlated region (25°-38°N, 65°-95°E) over the Tibetan Plateau is called

T

and the region (5°S-8°N,65°-95°E) over the tropical Indian Ocean is called

T

hereinafter.

3.2. Interannual variation features represented by the TCI

Fig. 1. (a) Correlation map of JJA-averaged 500-200 hPa air temperature with reference to the JJA zonal wind vertical shear (SASMI) between 850 hPa and 200 hPa averaged over (5°-25°N, 70°-100°E) (red box) for the 39-yr period from 1979 to 2017. The black dots indicate the 95% confidence level. The solid black lines are the north region over the southern Tibetan Plateau (25°-38°N, 65°-95°E), and the dashed black one is the south region over the tropical Indian Ocean (5°S-8°N,65°-95°E). (b) Time series of JJA 500-200 hPa air temperature over the Tibetan Plateau (orange, units: °C) and Indian Ocean (green, units: °C), together with their linear fit (dashed lines). The correlation coefficients between the Tibetan Plateau/Indian Ocean and the SASMI are shown in the top left-hand corner of (b). The linear trends of the Tibetan Plateau and Indian Ocean JJA air temperature and Mann-Kendall probability levels are given in the top right-hand corner of (b). (c)Normalized JJA-mean SASMI (solid black line) and TCI (bar) for the period 1979-2017. The correlation coefficient between the SASMI and TCI is given in the top left-hand corner of (c).

The interannual variations of

T

and

T

during the period from 1979 to 2017 are presented in Fig. 1 (b). Both

T

and

T

show significant upward trends with rates of 0.18 °C/10 yr and 0.16 °C/10 yr,respectively. All these trends are statistically significant at the 1% level,using the Mann-Kendall test. It is noticed that

T

and

T

both have similar linear trends but are different to each other on the interannual timescale. After the

T

,

T

, and TCI were calculated and compared with SASMI, it was noticed that TCI has the strongest relationship with SASMI. The correlation coefficient is 0.42 between the SASM intensity and

T

, while it is ? 0.60 between the SASM intensity and the

T

. Both of them tested above the 0.01 level of statistical significant. As we know,the positive correlation between

T

and the SASM intensity suggests that a warm (cold) anomaly of

T

is accompanied by an anomalous strong (weak) SASM. Meanwhile, the negative correlation between

T

and the SASM intensity indicates that an anomalous cold (warm)

T

is accompanied by a strong (weak) SASM. But what about the TCI? Is the SASM more sensitive to the intensity of the TCI? To answer this question, the variations of the normalized TCI and SASM index for the period 1979-2017 are shown in Fig. 1 (c). It is seen that the thermal contrast of the Tibetan Plateau and the Indian Ocean is closely related with the SASM intensity. The correlation coefficient between them even reaches 0.87. The result indicates that the TCI is a better measure of the SASM than both

T

and

T

.

In order to explore the representation of the TCI, composite analysis is employed to investigate the spatial structure of monsoon circulation variability. A year is defined as a typical large index year when the standard deviation of TCI is larger than 0.8, while a typical small one is when the standard deviation of TCI is less than ? 0.8. Following these criteria, 10 years are selected as typical large TCI years (1981, 1984,1985, 1994, 1999, 2000, 2001, 2008, 2012, 2013) and 7 years as typical small TCI years (1979, 1983, 1987, 1992, 1997, 2009, 2015). For simply addressing the characteristics, the composite differences between the typical large and small TCI years are given in Fig. 2 , which can simply be referred to as the feature anomalies in large TCI years.

During a large TCI year, the convection anomaly increases over the region of the SASM including the Indian subcontinent, Bay of Bengal,and Indochina Peninsula (as in Fig. 2 (a)). The composite OLR difference attains a 95% confidence level over the Indian subcontinent, Maritime Continent, and tropical western Pacific. Corresponding to the increased convection of the SASM, significant monsoon circulation anomalies are found in Fig. 2 (b, c). At 850 hPa, significant anomalous cross-equatorial southerlies along the East African coast are apparent, and the westerly extending from the East African coast to the Bay of Bengal is also enhanced. To the east of them is the enhanced South Asian monsoon trough( Fig. 2 (b)). At the upper level of 200 hPa, both the South Asian high and the tropical easterly jet to the south of it are significantly enhanced ( Fig.2 (c)). The composite 850 hPa wind result shows quite similar features to that explored based on all-Indian summer rainfall by Annamalai et al.(1999) . The composite results of both 200 hPa wind and OLR are similar to the composite results based on the Indian monsoon index defined by 850 hPa zonal wind ( Wang et al., 2001 ). Accordingly, the thermal contrast represented by the TCI can demonstrate the SASM intensity very well.

Fig. 2. Composite difference of summer (JJA) (a) OLR (units: W m ? 2 ), (b) 850 hPa wind, and (c) 200 hPa wind (units: m s ? 1 ) between the large and small thermal contrast years with respect to TCI. The black vectors in (b) and (c) denote the wind differences passing the 95% confidence level. Stippling in (a) denotes OLR differences over the 95% confidence level.

In order to investigate the superiority of the TCI, the OLR and wind fields are compared based on the results regressed onto the TCI,

T

, and

T

( Fig. 3 ). The linear trends have been subtracted before regression.Although the main convection and circulation features associated with the TCI and the

T

are quite similar, there are some obvious differences in related convection and circulation anomalies. The

T

obtains a good relationship with OLR and lower- and upper-level wind fields( Fig. 3 (d-f)). But the TCI addresses its impact on convection over the western Pacific warm pool and SASM general circulation compared with

T

. The main features of the 850 hPa anomalous winds associated with the TCI ( Fig. 3 (b)) and

T

( Fig. 3 (e)) reveal the zonal band of westerlies stretching from eastern Africa to the Indian subcontinent. But the anomalous westerly extends more eastward and more significantly over the southern Bay of Bengal for that associated with the TCI ( Fig. 3 (b)),which is more like the SASM circulation. This means that the TCI has a more remarkable impact on the region around the Bay of Bengal and even the South China Sea. Another interesting difference is the location of anomalous southerly flow over eastern China. The significant anomalous southerly appears over North China for the TCI but South China for

T

. Checking the general circulation in detail, it is noticeable that the southeasterly associated with the TCI over eastern China originates from the north part of the anomalous cyclone over the South China Sea and its adjacent region ( Fig. 3 (b)). But the southwesterly associated with

T

appearing over eastern China is due to the anomalous anticyclone over the South China Sea and its adjacent region ( Fig. 3 (e)). Fig. 3 (c, f) shows the 200 hPa winds associated with the TCI and

T

, respectively. They both result in a significant South Asian high over the Tibetan Plateau.But the easterly over the southern Bay of Bengal corresponding to the TCI is more prominent than that corresponding to the

T

. Combined with the lower- and upper-level wind fields, the anomalous circulations lead to major convergence over the Maritime Continent. Several other convergence centers appear over the Indian subcontinent, Indochina Peninsula, and South China Sea ( Fig. 3 (a, d)), while for the TCI the convective activity is more evident over the northern Bay of Bengal and the South China Sea ( Fig. 3 (a)) than that for the

T

( Fig. 3 (d)).But how do the wind anomalies over the South China Sea and its adjacent region become so different with ( Fig. 3 (b)) or without ( Fig. 3 e)the air temperature over the Indian Ocean? To answer this question, the regression fields of OLR and winds at 850 hPa and 200 hPa against

T

are shown in Fig. 3 (g-i). Fig. 3 (g) indicates that

T

is responsible for the suppressed convection in the western Pacific warm pool. It can be seen that weak heating over the eastern Indian Ocean is almost symmetric about the equator. According to the Gill response ( Gill, 1980 ), the westerly occurs in the upper level ( Fig. 3 (i)), the maximum low-level easterly occurs over the equator, and an anticyclonic flow is obtained away from the equator ( Fig. 3 (h)). The anticyclonic flow suppresses convective activity over the western Pacific Ocean, including the South China Sea and its adjacent region. The anticyclonic flow associated with

T

( Fig. 3 h)is stronger than that associated with

T

( Fig. 3 e). Thus, the results lead to an anomalous cyclone associated with the TCI located over the South China Sea and its adjacent region ( Fig. 3 b).In summary, the TCI can better depict the overall picture of the SASM circulation and contain more information about the influence of the SASM on climate than the

T

.

3.3. Seasonal and subseasonal characteristics revealed by the TCI

The onset of the SASM is the most important characteristic of its evolution. In fact, the onset of the SASM has been defined by various methods. Wang et al. (2009) used the 850 hPa zonal wind averaged over the South Arabian Sea (5°-15°N, 40°-80°E) as an onset circulation index to define the SASM onset date (SASMOD). SASMOD is well correlated with the onset date defined by the India Meteorological Department (correlation coefficient of 0.81 for the period from 1948 to 2007). According to the discussion above, SASMOD is proposed as a suitable reference for SASM onset dates. Based on the onset dates from SASMOD, the relationship between TCI and SASM onset is explored in the following analysis.

Fig. 3. Regression of JJA (a, d, g) OLR (units: W m ? 2 ), (b, e, h) 850 hPa winds, and (c, f, i) 200 hPa winds (units: m s ? 1 ) against the (a-c) TCI, (d-f) T TP , and (h-i)T IO . The black dots in (a, d, g) and the black vectors in (b, c, e, f, h, i) indicate the 95% confidence level.

As we know, the interannual variation of SASM onset is significant and is an important factor impacting local economy and society. In this paper, taking the onset date in a year as the dataset, a typical early SASM onset year is defined when the onset date is less than ? 0.8 standard deviations and a late onset year is obtained when it is larger than 0.8 standard deviations. Following these criteria, 9 early onset years(1985, 1989, 1990, 1999, 2000, 2001, 2004, 2006, 2016) and 10 late onset years (1979, 1983, 1992, 1995, 1997, 1998, 2003, 2007, 2014,2015) are obtained based on SASMOD. The average for these early onset dates is 20 May, and that for the 10 late onset dates is 9 June. The daily evolutions of average TCI for early and late onset years are presented in Fig. 4 a as blue and red lines, respectively. The blue (red) star indicates the average onset date of the SASM early (late) onset years.The difference between these early and late onsets is significant. TCI is almost always larger in early onset years than it is in the late onset years during the period before SASM onset, especially around the onset date.Also, the onset date almost always appears when the TCI slope is large.This might imply that the larger the Tibetan Plateau-Indian Ocean thermal contrast is accumulated, the earlier the SASM onset. In other words,a strong Tibetan Plateau-Indian Ocean thermal contrast is beneficial to the outbreak of the SASM.

Since the thermal difference indicated by the TCI is significantly different before SASM onset, it is necessary to further explore the possibility that the TCI can be used as a precursor signal of the monsoon onset. It is found that the increment of the TCI demonstrates interesting evolution characteristics before SASM onset. The pentad-by-pentad increment of the TCI (black line) and the pentad-averaged SASMI (red line) are shown in Fig. 4 (b). It is clear from Fig. 4 (b) that, before the 30th pentad, the pentad-by-pentad increment of the TCI increases rapidly ahead of the SASMI and reaches its maximum peak leading the SASMI. To see the pentad-by-pentad increment of the TCI’s relationship with the SASMI in detail, the lead-lag correlation between them was calculated ( Fig. 4 (c)).The pentad-by-pentad increment of the TCI leading the SASMI has significant correlation, and the correlation coefficient reaches a maximum with a 15-pentad lead. The result indicates the pentad-by-pentad increment of the TCI could serve as a signal for SASM forecasting.

The SASM onset is around the 30th pentad. As mentioned above, the pentad-by-pentad increment of the TCI reaches its maximum around the 30th pentad, which corresponds to the SASM onset time. This indicates that the SASM onset could be directly related to the pentad-bypentad increment of the TCI. To further examine the relationship between them, we computed the pentad-by-pentad increment of the TCI for typical early and late onset years according to the onset dates defined by SASMOD. The earliest onset in SASMOD is the 27th pentad.For simplicity, the increment of TCI in the first 25 pentads (hereinafter referred to as TCI25) was calculated for analysis. The result revealed that TCI25 is larger in early onset years than it is in late onsets. The difference in TCI25 between early and late onsets is statistically significant at the 0.01 level, using a two-sided Student’s

t

-test. The correlation coefficient between TCI25 and SASMOD (

r

= ? 0.34) presents a significant negative relationship at the greater than 95% confidence level. This result suggests that TCI25 has good skill in predicting SASM onset.

4. Concluding remarks

In this study, the relationship between the mid-upper tropospheric thermal contrast and the SASM was investigated. The TCI defined between the Tibetan Plateau and tropical Indian Ocean demonstrates very good ability in monitoring the SASM variability. As a basic driving factor of the SASM, the cooperative thermal effect containing both the Tibetan Plateau and the Indian Ocean may contribute more to the SASM than that just the Tibetan Plateau or Indian Ocean separately.

Fig. 4. (a) Daily evolution of average TCI for early (blue line) and late (red line) SASM onset years (units: °C). The horizontal axis is the days of one year. The average date of early (late) onsets is indicated by a blue (red) star. The SASM onset date is defined using the 850 hPa zonal wind ( Wang et al., 2009 ). (b) Pentad-by-pentad increment of TCI (black line, units: °C) and the pentad-averaged SASMI (red line, units: m s ? 1 ). (c) Lead-lag correlations between the pentad-by-pentad increment of TCI and the SASMI. The blue solid dots in (c) indicate the 99% confidence level and the blue circles denote the 95% confidence level.

In addition, the TCI is not only significantly correlated with the interannual variation of SASM intensity, but also closely correlated with the evolution of the SASM at the subseasonal scale. The pentad-by-pentad increment of the TCI increases rapidly before the onset of monsoon and leads the increase of the SASMI. Also, the TCI increment has good predictability with respect to annual variability. When the TCI increment is anomalously strong, the SASM onset will be earlier. The TCI can be used as precursor signal of SASM outbreak.

Funding

This work was supported jointly by the Strategic Priority Research Program of the Chinese Academy of Sciences [Grant number XDA20060501] and the National Natural Science Foundation of China[Grant numbers U1902209 and 91637208 ].

Acknowledgments

The discussion with Liang ZHAO is appreciated.