1 Introduction

Drought is one of the most frequent and deadliest natural disasters in South Asia, which can directly and indirectly reduce water availability and food production, causing tremendous ramifications on the environmental and social systems (Mishra et al. 2016; Kakatkar et al. 2018; Pradhan et al. 2021). Numerous studies revealed that the drought over South Asia was closely related to weak South Asian summer monsoon (SASM) (Ratnam et al. 2010; Pradhan et al. 2021; Zachariah et al. 2023). A case in point is that the SASM circulation in 2015 is the weakest since the 2000s (Fig. 2a), as shown by a sharp decrease in Indian summer rainfall by 14% with respect to a long period average (Mishra et al. 2016; Pradhan et al. 2021; Zachariah et al. 2023). Statistically, the 2015 drought was the tenth driest during 1906–2015 (Mishra et al. 2016) and the driest during 2010–2022 in India (Monsoon Online, https://mol.tropmet.res.in/monsoon-interannual-timeseries). This severe drought was one of the South Asian deadliest natural disasters on record, leaving at least 330 million people at risk of water and food shortages (BBC India 2016; Mishra et al. 2016). Therefore, unraveling the physical processes involved in SASM variations is essential for the monsoon prediction and disaster mitigation, which has been highly concerned by the international scientific community.

The causes for the weak atmospheric circulations, decreased rainfall, and short duration of SASM in 2015 have been studied and attributed to a strong El Niño event and its related atmospheric teleconnection (Kulkarni et al. 2016; Mishra et al. 2016; Kakatkar et al. 2018; Pradhan et al. 2021; Hu et al. 2022; Zachariah et al. 2023). Specifically, warm sea surface temperature (SST) anomalies and ascending motions appeared over the equatorial central-eastern Pacific in summer 2015, which induced descending motions and rainfall deficit over the western Pacific, the Maritime Continent, the eastern Indian Ocean, and Indian landmass. Besides, associated with the El Niño event, the meridional tropospheric temperature contrast between Eurasian region and the Indian Ocean weakened, resulting in the weak SASM circulations. In addition to the physical process related to El Niño–Southern Oscillation (ENSO) (Kumar et al. 1999, 2006; Annamalai et al. 2007; Hu et al. 2022, 2023; Zhang et al. 2022b; Chen et al. 2023a), the interannual variations of the SASM can be modulated by many other processes, which include but are not limited to the processes associated with Indian Ocean SST anomalies (Behera et al. 1999; Ashok et al. 2001; Roxy et al. 2015; Wang et al. 2018; Cai et al. 2021), North Atlantic SST anomalies and the North Atlantic Oscillation (Goswami et al. 2006; Li et al. 2008; Rajeevan and Sridhar 2008), the extratropical wave train (Borah et al. 2020), the thermal and dynamical forcings of the Tibetan Plateau (Boos and Kuang 2010; Wu et al. 2007, 2012), Eurasian snow cover (Hahn and Shukla 1976), and aerosols (Lau and Kim 2006; Meehl et al. 2008; Bollasina et al. 2011; Li et al. 2016). However, no consensus has been reached regarding the relative contributions of the dynamical (nonradiative) and radiative processes to the 2015 SASM.

Based on the total energy balance, an attribution method named the coupled atmosphere-surface climate feedback-response analysis method (CFRAM) was developed (Cai and Lu 2009; Lu and Cai 2009). It can be used to decompose temperature differences between two climate states into partial temperature changes caused by individual dynamical and radiative processes, including the processes of solar irradiance, ozone, carbon dioxide, surface albedo, water vapor, clouds, surface dynamics, and atmospheric dynamics. The CFRAM has been successfully applied to quantify the attribution of the temperature anomalies associated with various climate systems, including ENSO (Deng et al. 2012; Hu et al. 2016), the East Asia winter monsoon (Li and Yang 2017), the South China Sea summer monsoon (Li et al. 2020b), and global warming (Chen et al. 2018; Hu et al. 2017, 2020; Kong et al. 2022). However, the aerosols’ effect has not been unraveled in these studies based on the CFRAM.

In the recent decades, the anthropogenic aerosol emissions over South Asia have increased due to the rapid population growth and the accompanied industrialization (Ramanathan et al. 2005; Li et al. 2020a). Previous studies revealed that the changes in different types of aerosols can lead to distinct features of land-sea thermal contrast at the surface and the atmosphere, and may even exert opposite effects on SASM variations (Mitchell and Johns 1997; Ramanathan et al. 2005; Lau and Kim 2006; Meehl et al. 2008; Bollasina et al. 2011; Jin et al. 2016; Li et al. 2016). Specifically, black carbon can induce atmospheric warming but surface cooling via absorbing solar radiation, while other types of aerosols (e.g., sulfate, organic carbon, dust, and sea salt) can induce atmospheric and surface cooling via scattering solar radiation. Recently, Zhang et al. (2022a) updated the version of the CFRAM with the effects of different types of aerosols incorporated (CFRAM-A), which made it possible to quantify the contributions of aerosols in consideration of their different types. Based on the CFRAM-A, this study provides a quantitative attribution of various dynamical and radiative processes to the 2015 SASM, practically for the aerosol process.

The SASM can be depicted by various indices, encompassing zonal wind (Webster and Yang 1992), meridional wind (Goswami et al. 1999), India rainfall (Parthasarathy et al. 1992), out-going longwave radiation (Wang and Fan 1999), and land-sea thermal contrast (Fu and Fletcher 1985; Li and Yanai 1996; Sun et al. 2010; Dai et al. 2013). Among them, the indices based on land-sea thermal contrast have been widely used in the field of meteorology, because they not only represent the intensity of SASM circulations, but also describe monsoon formation (Li and Yanai 1996; Sun et al. 2010; Dai et al. 2013). Dai et al. (2013) compared the contribution of land-sea thermal contrast at the mid–upper troposphere with that at the mid–lower troposphere to the strength and variations of the SASM circulations, and revealed that the former was about three times larger of the latter. Therefore, we employ the land-sea thermal contrast at the mid–upper troposphere over South Asia to represent the SASM, as previous studies adopted (Li and Yanai 1996; Sun et al. 2010; Dai et al. 2013; Kong et al. 2022).

Based on the premise, we will make a quantitative attribution of mid–upper troposphere temperature associated with the 2015 SASM to various dynamical and radiative processes based on the CFRAM-A. The remainder of this paper is arranged as follows. Section 2 describes the data and methods used in this study. In Sect. 3, we present the land-sea thermal contrast features associated with the 2015 SASM, and calculate partial temperature changes associated with multiple dynamical and radiative processes and their summation. In addition, the contributions of different types of aerosols to the 2015 SASM have also been quantified. Section 4 gives a description of conclusions and discussion.

2 Data and methods

2.1 Data

The input variables required for the CFRAM-A include surface air temperature, surface sensible and latent heat fluxes, surface albedo, solar irradiance at the top of the atmosphere, air temperature, geopotential height, specific humidity, cloud amount, cloud liquid/ice water path, mixing ratios of greenhouse gases, ozone, and aerosols. Greenhouse gases consist of carbon dioxide and methane, and aerosols include five types, which are black carbon, sulfate, organic carbon, dust, and sea salt. The Modern-Era Retrospective Analysis for Research and Applications version 2 (MERRA-2) is the first long-term global atmospheric reanalysis to assimilate aerosols observations and represent aerosol-climate interactions from the National Aeronautics and Space Administration (Gelaro et al. 2017). Previous studies revealed that the MERRA-2 could well reproduce the spatiotemporal variations of Asian aerosols obtained from the ground measurements (Song et al. 2018; Zhang et al. 2020). Thus, we obtain the above monthly variables from the MERRA-2 with a horizontal resolution of 0.5° \(\times\) 0.625° and a vertical resolution of 42 pressure levels (https://gmao.gsfc.nasa.gov/GMAO_products/reanalysis_products.php; Gelaro et al. 2017). The monthly mean mixing ratios of carbon dioxide and methane are adopted from the National Oceanic Atmospheric Administration Earth System Research Laboratories (https://gml.noaa.gov/ccgg/trends). The monthly rainfall is derived from the Global Precipitation Climatology Project (GPCP) version 2.3 with a horizontal resolution of 2.5° \(\times\) 2.5° (https://psl.noaa.gov/data/gridded/data.gpcp.html; Adler et al. 2018).

To make an attribution of the 2015 SASM, the two mean states required for the CFRAM-A are the summer 2015 and the climatological summer during 2002–2022. The period of 2002–2022 is selected due to the availability of the high-quality reanalysis, and a background of relatively clean aerosols without major volcanic eruptions. Specifically, plentiful aerosols were emitted by the Philippines volcanic eruption in 1991 and sustained for several years (Minnis et al. 1993). In addition, the assimilated observations utilized in the MERRA-2 were doubled after 2002, most of which were contributed by the A-train satellites (Gelaro et al. 2017). More observations assimilated in the MERRA-2 could remarkably improve the reanalysis quality and further reduce the uncertainties of model dependent fields, such as aerosols and clouds, which are sensitive and critical to assess the magnitudes and spatial patterns in attribution analysis.

2.2 Methods

2.2.1 The CFRAM and CFRAM-A methods

The CFRAM was developed based on the total energy balance in an atmosphere-surface column including one surface layer and multi atmospheric layers (Cai and Lu 2009; Lu and Cai 2009). The difference in total energy between two climate states can be expressed as:

$$\Delta \frac{\partial E}{{\partial t}} = \Delta S - \Delta R + \Delta Q^{nonradiative}$$
(1)

Here, ∆ is the difference in summer between 2015 and 2002–2022. E depicts the total energy, and ∆\(\frac{\partial E}{\partial t}\) indicates the changes in the rate of energy storage. R (S) represents the divergence (convergence) of longwave (shortwave) radiative flux. \({\text{Q}}^{\text{nonradiative}}\) indicates the energy convergence due to the nonradiative processes at each layer. ∆\(\frac{\partial E}{\partial t}\) is negligible in the atmosphere on the seasonal mean timescale, but could be substantial in the ocean. Through linearizing the radiative energy perturbations by neglecting the effects induced by interactions among different radiative processes (e.g., water vapor, clouds, aerosols), ∆S and ∆R in Eq. (1) can be expressed as the summation of partial radiative energy flux convergence/divergence caused by individual processes (Chen et al. 2017):

$$\Delta S \approx \Delta S^{{{\text{SR}}}} + \Delta S^{{{\text{O}}_{3} }} + \Delta S^{{{\text{CO}}_{2} }} + \Delta S^{{{\text{AL}}}} + \Delta S^{{{\text{WV}}}} + \Delta S^{{{\text{CLD}}}}$$
(2)
$$\Delta R \approx \Delta R^{{{\text{O}}_{3} }} + \Delta R^{{{\text{CO}}_{2} }} + \Delta R^{{{\text{WV}}}} + \Delta R^{{{\text{CLD}}}} + \frac{\partial R}{{\partial T}}\Delta T$$
(3)

where ∆T is the vertical profile of temperature difference in summer between 2015 and 2002–2022. \(\frac{\partial R}{{\partial T}}\) is the Planck feedback matrix, which quantifies the change in the vertical profile of longwave radiative flux divergence induced by changes of the atmospheric and surface temperatures. Through replacing ∆\(S\) and ∆\(R\) in Eq. (1) with that in Eqs. (2) and (3) and multiplying both sides by \(\left( {\frac{\partial R}{{\partial T}}} \right)^{ - 1}\), we obtain:

$$ \Delta T = \left( {\frac{\partial R}{{\partial T}}} \right)^{ - 1} \left[ {\begin{array}{*{20}c} {\Delta S^{{{\text{SR}}}} + \Delta \left( {S - R} \right)^{{{\text{O}}_{3} }} + \Delta \left( {S - R} \right)^{{{\text{CO}}_{2} }} + \Delta S^{{{\text{AL}}}} + } \\ {\Delta \left( {S - R} \right)^{{{\text{WV}}}} + \Delta \left( {S - R} \right)^{{{\text{CLD}}}} + \Delta Q^{{{\text{sur\_dyn}}}} + \Delta Q^{{{\text{atmos\_dyn}}}} } \\ \end{array} } \right] $$
(4)

The temperature differences at each atmospheric layer can be attributed to the partial temperature differences associated with the changes in solar irradiance (SR), ozone (\({\text{O}}_{3}\)), carbon dioxide (\({\text{CO}}_{2}\)), surface albedo (AL), water vapor (WV), clouds (CLD), surface dynamics (sur_dyn), and atmospheric dynamics (atmos_dyn).

It is worthy to notice that \(\Delta {\text{Q}}^{{{\text{sur\_dyn}}}}\) and \(\Delta {\text{Q}}^{{{\text{atmos\_dyn}}}}\) are estimated as residuals from the radiative energy perturbations. The former mainly includes the energy originated in surface latent heat flux, surface sensible heat flux, and surface residual term. The surface residual term includes runoff, soil heat diffusion, and snow/ice melting/freezing over land/snow/ice surface, and heat storage and all-scales oceanic motions over ocean surface. The latter represents the energy transported by atmospheric turbulence, convection, and advection. Although the radiative energy perturbations can inevitably include errors from linearization and offline calculation, these errors are relatively small and negligible (Chen et al. 2017). For more specific details of the CFRAM, readers may refer to Cai and Lu (2009), Lu and Cai (2009), Deng et al. (2012), Chen et al. (2017), and Hu et al. (2017).

In the CFRAM-A, Zhang et al. (2022a) isolated the direct radiative effects of aerosols and the effects of methane (\({\text{CH}}_{4}\)) from the residuals of radiative energy perturbations in the CFRAM. The aerosols mainly consist of black carbon (BC), sulfate (SULF), organic carbon (OC), dust (DUST), and sea salt (SS). The indirect effects of aerosols are unable to be isolated, because they are included in cloud change. Reanalysis datasets so far could not provide independent cloud change without the effects of aerosols. Thus, Eq. (4) can be expressed as:

$$ \Delta T = \left( {\frac{\partial R}{\partial T}} \right)^{ - 1} \left[ {\begin{array}{*{20}c} {\Delta S^{\text{SR}} + \Delta \left( {S - R} \right)^{{}_{3}{\text{O} }} + \Delta \left( {S - R} \right)^{{\text{CO}}_{2} } + \Delta \left( {S - R} \right)^{{\text{CH}}_{4} } + \Delta S^{\text{AL}} + } \\ {\Delta \left( {S - R} \right)^{\text{WV}} + \Delta \left( {S - R} \right)^{\text{CLD}} + \Delta \left( {S - R} \right)^{\text{BC}} + \Delta \left( {S - R} \right)^{\text{SULF}} + } \\ {\Delta \left( {S - R} \right)^{\text{OC}} + \Delta \left( {S - R} \right)^{\text{DUST}} + \Delta \left( {S - R} \right)^{\text{SS}} + \Delta Q^{\text{sur\_dyn}} + \Delta Q^{\text{atmos\_dyn}} } \\ \end{array} } \right] $$
(5)

Utilizing the rapid radiative transfer method for general circulation model version 5 (Iacono et al. 2008), the offline radiative transfer calculations in the CFRAM-A are conducted separately for summer 2015 and the period of 2002–2022 in summer, to obtain the individual radiative energy perturbations in Eq. (5). For more specific details of the CFRAM-A, reader may refer to Zhang et al. (2022a).

2.2.2 Thermodynamic energy diagnosis

The thermodynamic energy equation or the temperature tendency equation is utilized to investigate the contributions from horizontal advection, adiabatic heating and diabatic heating to the change in air temperature (Cheung et al. 2013; Li et al. 2019), which is written as:

$$\frac{\partial T}{{\partial t}} = \underbrace { - V \cdot \nabla T}_{advT}\underbrace {{ - W\frac{T}{\theta }\frac{\partial \theta }{{\partial p}}}}_{dTdp} + \underbrace {{\frac{{\dot{Q}}}{{C_{p} }}}}_{DH}$$
(6)

where advT represents the horizontal advection of temperature, dTdP indicates the adiabatic heating caused by vertical motions, and DH depicts the diabatic heating. T is the air temperature, V is the horizontal winds, W is the vertical velocity, θ is the potential temperature, \(\dot{Q}\) is the diabatic heating rate per unit mass, and \({\text{C}}_{\text{p}}\) (1004 J K−1 kg−1) is the specific heat at constant pressure.

3 Results

3.1 Atmospheric circulation and rainfall features associated with 2015 SASM

Figure 1 shows the summertime features of atmospheric temperature, circulation, and rainfall in 2015 relative to 2002–2022 climatology. The 500–200-hPa mean thermal contrast between the vicinity of the Tibetan Plateau and the Indian Ocean in 2015 is negative, resulting in the weakened SASM circulations (Fig. 1a). Correspondingly, upper-level westerly anomalies and lower-level easterly anomalies appear over South Asia, which is not conducive to water vapor transported from the tropical oceans to South Asia, resulting in severe drought over India (Fig. 1b). The 2015 drought posed great threatens to water and food security and affected the lives of 330 million people (BBC India 2016; Mishra et al. 2016).

Fig. 1
figure 1

Composite differences in summer a 500–200-hPa mean temperature (contour; K), geopotential height (shading; gpm), and 200-hPa winds (vectors; m s−1), and b 850-hPa winds (vectors; m s−1) and rainfall (shading; mm day−1) between 2015 and 2002–2022 climatology. Climatological temperature anomalies (relative to the average temperature over the domain of 15°S–45°N/40°–120°E, shading; K) and winds (vectors; m s−1) in summer during 2002–2022 at c 200 hPa and d 850 hPa. The yellow boxes outline the regions for defining the LSTCI

Full size image

Northeasterly wind is observed at the upper troposphere over South Asia in summer during 2002–2022, along with southwesterly wind at the lower level (Fig. 1c, d). Such a SASM-related atmospheric circulation is accompanied by a remarkable thermal contrast between the vicinity of the Tibetan Plateau and the tropical Indian Ocean, which is more pronounced at the upper level than that at the lower level (Fig. 1c, d). As in previous studies (Li and Yanai 1996; Sun et al. 2010; Dai et al. 2013), we employ the land-sea thermal contrast index (LSTCI) defined as the difference in 500–200-hPa mean temperature between southern Eurasia (SE, 20°–40°N, 60°–100°E) and the tropical Indian Ocean (TIO, 10°S–10°N, 60°–100°E) to measure the variation of SASM circulation.

To verify the robustness of our results, another two widely used SASM indices representing the land-sea thermal contrast are also employed: the meridional geopotential height contrast index (MZI), defined as the difference in geopotential height thickness at 500–200 hPa between SE and the TIO (Dai et al. 2013), and the Webster-Yang index (WYI), defined as the difference in zonal wind between 850 and 200 hPa averaged over 0°–20°N/40°–110°E (Webster and Yang 1992). Figure 2a shows the normalized time series of the LSTCI, MZI, and WYI during 2002–2022. It is apparent that the LSTCI is significantly correlated with the MZI (r = 0.84, p < 0.01) and the WYI (r = 0.73, p < 0.01). Furthermore, the three indices display a similar interannual variation and show that the 2015 SASM is the weakest during 2002–2022.

Fig. 2
figure 2

a Time series of the normalized LSTCI (black solid curve), MZI (red solid curve), and WYI (blue solid curve) during 2002–2022. Regression patterns of summer b 500–200-hPa mean temperature (contour; K), geopotential height (shading; gpm), and 200-hPa winds (vectors; m s−1), and c 850-hPa winds (vectors; m s−1) and rainfall (shading; mm day−1) against the normalized LSTCI during 2002–2022. The stippled areas and the black vectors in b, c denote the significant values exceeding the 90% confidence level based on the Student’s t-test. The yellow boxes outline the regions for defining the LSTCI

Full size image

The atmospheric circulation and rainfall regressed onto the LSTCI are given in Fig. 2b, c. The LSTCI can well capture the land-sea temperature and geopotential height contrast at the mid–upper troposphere associated with the SASM (Fig. 2b). According to the thermal-wind relationship, lower-level westerly anomalies and upper-level easterly anomalies appear over South Asia during the positive LSTCI years, and vice versa (Fig. 2b, c). In the positive phase of the SASM, high-pressure anomalies control the Tibetan Plateau–northern India, resulting in stronger-than-normal upper-level suctions, accompanied by ascending motions and enhanced rainfall in India (Fig. 2c). The above analysis shows that the LSTCI can well represent the SASM in terms of the variations of associated atmospheric circulations, rainfall, and land-sea thermal contrast.

3.2 Quantifying the contributions of multiple physical processes to 2015 SASM-related land-sea thermal contrast

Here, we conduct a quantitative attribution analysis of the 2015 SASM depicted by the LSTCI. Based on the CFRAM-A, the mid–upper tropospheric temperature anomalies in summer 2015 are decomposed into partial temperature anomalies induced by individual dynamical and radiative processes. Figure 3 displays the composite differences in partial temperatures caused by solar irradiance, ozone, carbon dioxide, methane, surface albedo, water vapor, clouds, aerosols, surface dynamics, and atmospheric dynamics between 2015 and 2002–2022 climatology (Fig. 3c–l), and their summation (Fig. 3b). The difference in 500–200-hPa mean temperature between 2015 and 2002–2022 climatology obtained from the MERRA-2 (Fig. 3a) shares a similar spatial pattern with the summation of partial temperature changes derived from the CFRAM-A analysis (Fig. 3b), presenting nearly identical land-sea thermal contrast features. Furthermore, the value of 2015 LSTCI derived from the MERRA-2 is the same as that in the CFRAM-A (− 0.65 K). That is, the linearization and offline radiative calculations in the CFRAM-A are reasonable approximations.

Fig. 3
figure 3

a Summer 500–200-hPa mean temperature differences (K) between 2015 and 2002–2022 climatology obtained from the MERRA-2. b The sum of summer 500–200-hPa mean partial temperature differences (K) derived from the CFRAM-A. Summer 500–200-hPa mean partial temperature differences (K) due to the changes in c solar irradiance, d ozone, e carbon dioxide, f methane, g surface albedo, h water vapor, i clouds, j aerosols, k surface dynamics, and l atmospheric dynamics. The number on the upper right of each panel denotes the value of 2015 LSTCI (K). The gray boxes outline the regions for defining the LSTCI

Full size image

The partial temperature anomalies of 2015 LSTCI caused by solar irradiance, ozone, carbon dioxide, methane, and surface albedo are less than 0.03 K (Fig. 3c–g), which are small and negligible (Fig. 4a), and will not be discussed in the following study. Overall, the partial temperature anomalies of 2015 LSTCI resulting from water vapor, atmospheric dynamic, cloud, and aerosol processes are − 0.38 K, − 0.12 K, − 0.1 K, and − 0.1 K (Fig. 3h, l, i, and j), which separately contribute about 58%, 18%, 15%, and 15% (Fig. 4a). These positive contributions are partially offset by the negative effect of surface dynamic process, whose contribution is about -14% (0.09 K, Figs. 3k and 4a). Thus, water vapor process plays a dominate role in temperature anomalies in 2015 LSTCI. It is worthy to notice that the magnitude of temperature anomalies induced by the atmospheric dynamic process over South Asia is the greatest among all processes (Fig. 3); however, it only plays a secondary role in 2015 LSTCI due to its similar effects on the temperature anomalies over both SE and the TIO (Fig. 4).

Fig. 4
figure 4

Summer 500–200-hPa mean temperature differences (K) in 2015 a LSTCI (SE–TIO), b SE, and c TIO derived from the CFRAM-A due to the changes from different physical processes and their sum. The abbreviations ‘SR’, ‘O3’, ‘CO2’, ‘CH4’, ‘AL’, ‘WV’, ‘CLD’, ‘AER’, ‘SUR’, ‘ATM’, and ‘SUM’ represent the processes of solar irradiance, ozone, carbon dioxide, methane, surface albedo, water vapor, clouds, aerosols, surface dynamics, atmospheric dynamics, and their sum, respectively

Full size image

As illustrated above, the LSTCI is defined based on the thermal contrast between SE and the TIO. The anomalous cooling over SE (− 0.33 K) and the anomalous warming over the TIO (0.32 K) respectively account for approximately 51% and 49% of the temperature anomalies associated with 2015 LSTCI (Fig. 4). Making quantitative attribution of the two key regions of the LSTCI may provide a new insight into understating the physical processes involved in the SASM.

3.3 Contributions of multiple physical processes to temperature anomalies over land associated with 2015 SASM

The mid–upper tropospheric temperature anomalies display an overall negative distribution over SE (− 0.33 K, Figs. 3b and 4b), which is mainly attributed to the cloud process (− 0.43 K), followed by the water vapor (− 0.29 K), surface dynamic (− 0.28 K), and aerosol (− 0.06 K) processes (Fig. 4b). These positive contributions are partially counteracted by the atmospheric dynamic process, which causes partial temperature anomalies of 0.65 K. Diagnostic analysis is further conducted to understand how the partial temperature anomalies related to the above five processes are generated over SE.

Based on the thermodynamic energy equation, the positive temperature anomalies associated with the atmospheric dynamic process over SE in summer 2015 (Figs. 3l and 4b) are determined by horizonal warm advections and adiabatic descending motions (Figs. 5a, b, and 7f). The decreased diabatic heating induced by descending motions and decreased rainfall (Fig. 1b), results in a relatively small atmospheric cooling over SE (Fig. 5c).

Fig. 5
figure 5

Composite differences in summer 500–200-hPa mean temperature tendency (10−6 K day−1) of a horizontal advection, b adiabatic heating, and c diabatic heating between 2015 and 2002–2022 climatology. The gray boxes outline the regions for defining the LSTCI

Full size image

Consistent with the descending motions, cloud fraction decreases over SE (Fig. 7c), and then less longwave (shortwave) radiation is trapped (reflected) by the atmosphere, resulting in atmospheric cooling (warming) (Fig. 7a, b). Thus, the cooling effect of cloud process (Figs. 3i and 4b) is dominated by the longwave effect. As a greenhouse gas, water vapor process generally exhibits a warming effect on temperature. This process results in negative partial temperature anomalies over SE (Figs. 3h and 4b), indicating that the water vapor decreases over this region, which is consistent with the negative anomalies of specific humidity induced by horizonal divergences of water vapor and descending motions (Fig. 7d–f). The overall positive effect of surface dynamic process is mainly determined by the changes in surface latent heat flux (− 0.38 K, Fig. 6c) and surface residual term (− 0.24 K, Fig. 6d), while surface sensible heat flux provides a negative contribution (0.34 K, Fig. 6b). These surface energy changes cause perturbations in mid–upper troposphere temperature through upward longwave radiation by multiplying the inverse of the Planck feedback matrix \(\left( {\frac{\partial R}{{\partial T}}} \right)^{ - 1}\). The contribution of aerosol process to SE cooling (18%) and TIO warming (13%) will be discussed in detail in Sect. 3.5.

Fig. 6
figure 6

Summer 500–200-hPa mean partial temperature differences (K) derived from the CFRAM-A due to the changes in a surface dynamics, b surface sensible heat flux, c surface latent heat flux, and d surface residual. The number on the upper right of each panel denotes the value of 2015 LSTCI (K). The gray boxes outline the regions for defining the LSTCI

Full size image

3.4 Contributions of multiple physical processes to temperature anomalies over oceans associated with 2015 SASM

The anomalous TIO warming (0.32 K) is primarily determined by the atmospheric dynamic process (0.77 K), followed by the water vapor (0.09 K) and aerosol (0.04 K) processes (Fig. 4c). The effects of surface dynamic and cloud processes are negative, generating area-averaged partial temperature anomalies of − 0.36 K and − 0.33 K, respectively (Fig. 4c).

Specifically, the positive temperature anomalies associated with atmospheric dynamic process are dominated by the adiabatic descending motions over the western and southeastern TIO, which result in atmospheric warming in summer 2015 (Figs. 3l, 5b, and 7f). The decreased diabatic heating associated with descending motions causes a relatively small atmospheric cooling over the western and southeastern TIO (Fig. 5c). Horizontal advections are small over the TIO region (Fig. 5a). The warming effect of water paper over the TIO is closely linked to the humid mid–upper troposphere over the northern and southwestern TIO (Fig. 7d), which is induced by ascending motions and horizonal convergences of water vapor over the region (Fig. 7e, f). In addition, the humid lower troposphere associated with increased rainfall over the TIO (Fig. 1b) can also facilitate the mid–upper tropospheric warming via its upward longwave radiation. The overall warming effect of water vapor over the TIO is partly compensated by a weak cooling effect of negative water vapor anomalies mainly induced by the descending motions over the southeastern TIO (Figs. 1b, 7d, and f).

Fig. 7
figure 7

Summer 500–200-hPa mean partial temperature differences (K) derived from the CFRAM-A due to the changes in a cloud longwave radiation and b shortwave radiation. Composite differences in summer 500–200-hPa mean c cloud fraction (10−2), d specific humidity (10−4 kg kg−1), e water vapor flux (vectors, 10−3 kg m kg−1 s−1) and its divergence (shading, 10−9 kg m kg−1 s−1), and f vertical velocity (10−2 Pa s−1) between 2015 and 2002–2022 climatology. The number on the upper right of a, b denotes the value of 2015 LSTCI (K). The red boxes outline the regions for defining the LSTCI

Full size image

The cooling effect of surface dynamic process over the TIO (− 0.36 K, Fig. 6a) is mainly determined by the changes in surface residual term (− 1.08 K, Fig. 6d), which is partially counteracted by the changes in surface sensible heat flux (0.11 K, Fig. 6b) and surface latent heat flux (0.60 K, Fig. 6c). Specifically, the cooling effect of surface dynamic process over the western and southeastern TIO is mainly determined by the surface residual term and surface sensible heat flux over the western TIO (Fig. 6b and d) and surface latent heat flux over the southeastern TIO (Fig. 6c). Consistent with the monsoon-associated vertical motions, decreased cloud fraction appears over the western and southeastern TIO, while increased cloud fraction appears over the northern and southwestern TIO (Fig. 7c and f). Still, the longwave effect dominates the total effect of cloud (Figs. 2i and 7a, b), which induces overall negative partial temperature anomalies over the TIO (Fig. 4c).

3.5 Contributions from different types of aerosols

The total effect of aerosols (-0.1 K, Fig. 8a) is dominated by the change in black carbon (− 0.06 K, Fig. 8b) and sea salt (− 0.05 K, Fig. 8e). The effect of sulfate provides a small positive contribution (− 0.01 K, Fig. 8d), while the changes in organic carbon and dust display small negative contributions (0.01 K and 0.01 K, Fig. 8c and f). The land-sea thermal contrast induced by aerosol process is characterized by negative partial temperature anomalies over SE (− 0.06 K) and positive partial temperature anomalies over the TIO (0.04 K) (Figs. 4b, c, and 8a).

Fig. 8
figure 8

Summer 500–200-hPa mean partial temperature differences (K) derived from the CFRAM-A due to the changes in a total aerosol and its different types: b black carbon, c organic carbon, d sulfate, e sea salt, and f dust. The number on the upper right of each panel denotes the value of 2015 LSTCI (K). The gray boxes outline the regions for defining the LSTCI

Full size image

As discussed in the introduction, the strong absorption of solar radiation by black carbon leads to atmospheric warming, while the scattering of solar radiation by other types of aerosols (i.e., organic carbon, sulfate, sea salt, and dust) provides anomalous cooling for the atmosphere. In addition, according to Eq. (5), the atmospheric temperature anomalies can be induced by the direct effect of same-level aerosols and the feedback of the lower-level aerosols through multiplying the inverse of the Planck feedback matrix \(\left( {\frac{\partial R}{{\partial T}}} \right)^{ - 1}\). Thus, horizontal distributions and vertical profiles of the changes in different types of aerosols are displayed in Figs. 9 and 10.

Fig. 9
figure 9

Composite differences in summer 500–200-hPa mean a black carbon (10–11 kg kg−1), b organic carbon (10–11 kg kg−1), c sulfate (10–11 kg kg−1), and d sea salt (10–11 kg kg−1) between 2015 and 2002–2022 climatology. The gray boxes outline the regions for defining the LSTCI

Full size image
Fig. 10
figure 10

Composite differences in the vertical cross-section averaged over 60°–100°E in summer a black carbon (10–11 kg kg−1), b organic carbon (10–11 kg kg−1), c sulfate (10–11 kg kg−1), and d sea salt (10–11 kg kg−1) between 2015 and 2002–2022 climatology

Full size image

The partial temperature anomalies induced by aerosols display an overall cooling pattern over SE (− 0.06 K, Fig. 8a), which is primarily attributed to the effects of sea salt (− 0.03 K, Fig. 8e), black carbon (− 0.02 K, Fig. 8b), and sulfate (− 0.02 K, Fig. 8d). Specifically, the central-northern SE cooling is closely linked to black carbon process (Fig. 8b), which is induced by the decreased black carbon in the whole troposphere (Figs. 9a and 10a). The southern SE cooling is mainly related to sea salt and sulfate processes (Fig. 8d, e), which is caused by the increased sulfate at 1000–200 hPa and sea salt at 500–400 hPa and 1000–800 hPa (Figs. 9c, d and 10c, d).

The partial temperature anomalies caused by aerosol process averaged over the TIO is 0.04 K, which is primarily driven by the black carbon (0.04 K, Fig. 8b) and sea salt (0.02 K, Fig. 8e) processes. The organic carbon (− 0.01 K, Fig. 8c) and sulfate (− 0.01 K, Fig. 8d) processes provide negative contributions to the TIO warming. Specifically, the black carbon increases at the troposphere over the TIO, inducing 500–200-hPa mean atmospheric warming (Figs. 9a and 10a). The partial temperature anomalies induced by sea salt process exhibits a northern warming and southern cooling pattern over the TIO, which is caused by decreased (increased) sea salt over the northern (southern) TIO at the lower levels (Fig. 10d). The increased organic carbon at the whole troposphere and sulfate at 500–200 hPa and 1000–850 hPa result in 500–200-hPa mean atmospheric cooling over the TIO (Figs. 9b, c and 10b, c).

In summary, the thermal contrast between SE and the TIO induced by the black carbon and sea salt processes is more pronounced than by other aerosol processes (Fig. 8b and d). The decreased black carbon over the central-northern SE may originate from local sources through suppressed vertical transports (Fig. 7f). On the other hand, the increased black carbon over the TIO may be explained by the increased biomass burning over Southeast Asia. Previous studies elucidated that the aerosol emissions from biomass burning over Southeast Asia are mainly caused by forest fires (Shi et al. 2014, 2015). In summer 2015, less rainfall occurs over Southeast Asia (Fig. 1b), which provides a favorable condition for forest fires, leading to an increase in the emissions of black carbon. The sea salt process is dominated by the changes in lower-level sea salt, which is positively correlated with wind speed. As displayed in Fig. 1b and d, the anomalous lower-level wind speed in summer 2015 increases over the southern TIO and southern SE and decreases over the northern TIO, which are consistent with the sea salt anomalies (Fig. 10d).

4 Conclusions and discussion

Associated with the anomalously belated and weak SASM in 2015, prolonged droughts occurred over the broad swathes of the Indian peninsula, leading to tremendous water and food shortages and economic loss (BBC India 2016; Mishra et al. 2016). The causes of this substantially weak SASM case have been studied extensively, mainly attributed to the atmospheric and oceanic processes associated with a strong El Niño event in 2015 (Chen et al. 2016; Kulkarni et al. 2016; Mishra et al. 2016; Pradhan et al. 2021; Zachariah et al. 2023). In addition, the Arctic Oscillation and North Pacific atmospheric variability are also in their extreme phase in 2015 (Chen et al. 2023b). Studies have shown that Arctic Oscillation can have significant impacts on ENSO, SST anomalies in the TIO and temperature anomalies over SE (Chen et al. 2014; Gong et al. 2017; Cheng et al. 2023). Therefore, the formation of the temperature anomalies over SE and the TIO associated with the 2015 weak SASM may be partly due to the extreme Arctic Oscillation. However, no consensus has been reached so far regarding the relative contributions of dynamical and radiative processes to the 2015 weak SASM.

The CFRAM-A is an efficient offline attribution method, which can decompose temperature differences between two climate states into partial temperature changes caused by individual dynamical and radiative processes, including the processes of solar irradiance, ozone, carbon dioxide, methane, surface albedo, water vapor, clouds, aerosols, surface dynamics, and atmospheric dynamics. Based on this method, the current study provides a quantitative attribution analysis of the 2015 SASM, which is characterized by the difference in 500–200-hPa mean temperature between SE and the TIO. The contributions of individual dynamical and radiative processes to the temperature anomalies over SE and the TIO have been further intercompared. The main results demonstrate that the temperature anomalies associated with the 2015 SASM can be mainly attributed to water vapor feedback effect (58%), followed by atmospheric dynamic process (18%). The processes of clouds and aerosols contribute nearly equally (15%). These positive contributions are partially offset by the negative contribution from surface dynamic process (-14%). The contributions from solar irradiance, ozone, carbon dioxide, methane, and surface albedo to the 2015 SASM are much smaller.

As two important components, the SE cooling and TIO warming respectively account for approximately 51% and 49% to the temperature anomalies associated with the 2015 SASM. The processes of water vapor and aerosols contribute positively to both SE and the TIO, and the effect of water vapor is closely linked to the decreased (increased) specific humidity over SE (TIO) via its greenhouse effect. The cooling effect of aerosols over SE is induced by increased sea salt, decreased black carbon, and increased sulfate. The warming effect of aerosols over the TIO is highly related to increased black carbon and decreased sea salt. The atmospheric dynamic process exerts a negative impact on SE and a positive impact on the TIO. The former is caused by horizonal warm advections and adiabatic descending motions, while the latter is mainly induced by adiabatic descending motions. The contributions from cloud and surface dynamic process are positive over SE, while they are negative over the TIO. The cooling effect of cloud over both SE and the TIO is closely linked to decreased cloud fraction via its longwave effect. However, the shortwave effect of cloud is more important than its longwave effect for the 2015 weak SASM. The surface dynamic process is mainly determined by the changes in surface latent heat flux and surface residual term over SE, while the change in surface residual term plays a dominant role over the TIO.

This study reveals two novel features. First, previous analyses to the 2015 weak SASM are mostly qualitative. Here, we quantify the relative contributions of various dynamical and radiative processes to the weak monsoon with the aid of the CFRAM-A, particularly the contributions of different types of aerosols. Considerable studies have elucidated that aerosols can exert significant impacts on SASM variations (Lau and Kim 2006; Meehl et al. 2008; Bollasina et al. 2011; Li et al. 2016). In the recent decades, the anthropogenic aerosol emissions increased over South Asia, and especially, the fossil fuel black carbon increased about 6 folds from the 1930s to the 2000s (Ramanathan et al. 2005; Li et al. 2020a). Quantifying the contributions of aerosols is essential for pollution mitigation policies. Secondly, we make a comprehensive attribution of temperature anomalies over SE and the TIO in summer 2015, and compare the effects of individual dynamical and radiative processes on SE with those on the TIO. This result can enhance our understanding of the physical processes in the land and oceans responsible for SASM variations.