The simulation of water vapor transport in East Asia using a regional air–sea coupled model

Authors

  • Suxiang Yao,

    Corresponding author
    1. Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science & Technology, Nanjing, China
    • Corresponding author: S. Yao, Department of Atmospheric Science, Nanjing University of Information Science & Technology, Nanjing, China. (yaosx@nuist.edu.cn)

    Search for more papers by this author
  • Qian Huang,

    1. Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science & Technology, Nanjing, China
    Search for more papers by this author
  • Yaocun Zhang,

    1. Department of Atmospheric Science, Nanjing University, Nanjing, China
    Search for more papers by this author
  • Xu Zhou

    1. Key Laboratory of Meteorological Disaster of Ministry of Education, Nanjing University of Information Science & Technology, Nanjing, China
    Search for more papers by this author

Abstract

[1] A regional air–sea coupled model (RegCM3-POM) is used, and the ability of the model to simulate June, July, and August (JJA) moisture transport is verified. The coupled model (RegCM3-POM) has better performances than the uncoupled model (RegCM3) in the simulation of the JJA mean water vapor characteristics from 1979 to 2002 and its interannual variations. The differences of water vapor between RegCM3-POM and RegCM3 are mainly caused by different winds in the middle and low troposphere. The simulated circulation pattern of RegCM3-POM is more similar to the Climate Forecast System Reanalysis (CFSR) data than that of the uncoupled model. The simulated geopotential height and the gradient of geopotential height in the coupled model are more accurate than those in the uncoupled model. The real-time and lead–lag relationship between geopotential height and sea surface temperature (SST) in RegCM3-POM is consistent with that between the height of CFSR data and HadISST (Hadley Center Sea Ice and Sea Surface Temperature data). Thus, the simulated water vapor in the lower troposphere and the vertically integrated moisture transport are more accurate in the coupled model than in RegCM3. In addition, the correlation coefficients between moisture transport and SST in RegCM3 are positive in South China Sea, corresponding to SST forcing; the correlation coefficients in RegCM3-POM are negative, which is consistent with that between the vapor transport in CFSR data and HadISST. Thus, the air–sea coupled model represents more accurately than the uncoupled model the relationship between SST and moisture transport.

1 Introduction

[2] East China experiences a typical monsoon climate. During the summer, the bulk of central and eastern Asia is under the influence of southwest surface winds accompanied by large amounts of water vapor. Water vapor transport is one of the most important components of the East Asian monsoon system, and a large amount of water vapor is carried from the adjacent oceans to the East Asian monsoon regions by large-scale monsoon circulations [e.g., Park and Schubert, 1997; Zhou and Yu, 2005; Zhou and Wang, 2006; Zhao et al., 2007]. The characteristics of the transport and budget of water vapor as well as its association with precipitation in the monsoonal regions of East Asia have been documented in a number of studies [e.g., Xu et al., 2004; Wei et al., 2005; He et al., 2007; James et al., 2008; Liu et al., 2009; Sun et al., 2011].

[3] In addition to statistical methods, many studies on the East Asian monsoon have used regional climate models (RCMs) with high resolution, and that takes into account a relatively complete number of the physical processes involved. Recently, efforts have been made to construct an air–sea coupled regional model for East Asia. Ren and Qian [2005] used a p-σ RCM coupled with the revised Princeton Oceanic Model [POM; Mellor, 2003] to simulate the summer monsoon. Yao and Zhang [2010] used a regional air–sea coupled climate model based on RegCM3 [Pal et al., 2007] and POM to simulate summer precipitation over China. Li and Zhou [2010] used a regional air–sea coupled model based on RegCM3 and the Hybrid Coordinate Ocean Model [Bleck, 2002] and found that the regional air–sea coupled model was more suitable for simulating the East Asia monsoon. In a model, an accurate representation of moisture transport is a prerequisite for the realistic simulation of precipitation [Wang et al., 2003].

[4] Because almost all of the water vapor comes from the ocean, the main difference between the air–sea coupled model and the uncoupled model is the inclusion of a different forcing process in the air–sea interface. These studies have shown that to evaluate the performance of a regional air–sea coupled model in East Asian monsoon regions, it is necessary to analyze the modeling of the water vapor transport. In this paper, a regional air–sea coupled model (RegCM3-POM) based on RegCM3 and POM is used to simulate the climate in East Asia from 1979 to 2002, and the simulated water vapor transport is analyzed. This research reveals the influences of the air–sea interaction on the regional climate of East Asia to some extent.

2 The Model and Data

[5] In this study, we selected the regional air–sea coupled model RegCM3-POM. The coupled model is based on RegCM3 which was developed by the International Center for Theoretical Physics (Version 3.0; released in 2005) and POM (Version 2 K). The POM has already been used in studies on the regional air–sea interaction [e.g., Zou and Zhou, 2012]. The coupled model, RegCM3-POM, simulates East Asian rainfall well [e.g., Huang et al., 2012]. In the coupled model, fluxes are exchanged at the air–sea interface. The atmospheric module provides the wind speed, precipitation, evaporation, sensible heat flux, net solar radiation, and net long-wave radiation flux over the sea surface used to force the ocean module; the ocean module provides the sea surface temperature (SST) to the atmosphere module. In the coupled system, the ocean adjusts to a new state based on the thermal and mechanical forcing of the atmosphere, and, in turn, the atmosphere is forced by ocean thermal forcing. At the air–sea interface, the coupled model accounts for the air–sea interaction process.

[6] In this paper, RegCM3-POM is used to simulate summer (June, July, and August (JJA)) climatic characteristics over China from 1979 to 2002, and the same experiments are conducted using the uncoupled model RegCM3 as a control experiment (CTR) to validate the accuracy of the coupled run (CPL).

[7] In the atmospheric module of the coupled model and the uncoupled model, the Lambert conformal projection is used; the grid spacing is 60 km × 60 km; there are 18 layers in the vertical direction, and the top level is 5 hPa; the planetary boundary layer scheme is that developed by Holtslag [Holtslag et al., 1990]; a Modified-Kuo scheme [Anthes, 1977] is chosen as the convective precipitation scheme; both the CTR and CPL use the National Center Atmospheric Research (NCAR) CCM3 radiation scheme [Kiehl et al., 1996]; at the ocean surface, the Zeng scheme [Zeng et al., 1998] is chosen to calculate the ocean flux. For the ocean module of the coupled system, a grid spacing of 0.5° × 0.5° is used, the σ coordinate is chosen in the vertical direction with 16 levels, and the radiation boundary condition is chosen as the lateral boundary condition.

[8] The initial underlying conditions of the coupled model are provided by the SST simulated by POM each year from 1979 to 2002. The initial conditions of the uncoupled model are provided by the HadISST [Rayner et al., 2003] data, and the uncoupled model is forced by HadISST data as the underlying surface. The initial condition and the lateral boundary condition data are supplied by NCEP/NCAR [National Centers for Environmental Prediction and National Center Atmospheric Research; Kalnay et al., 1996; Kistler et al., 2001] reanalysis for both the atmosphere module of the coupled model and the uncoupled model.

[9] It is necessary to point out that the CPL experiment design has some shortcomings. Because of the characteristics of the RCM and the limit of calculation conditions, the 24 year simulation is not a continuous run. The coupled model started from the SST simulated by the oceanic model POM each year from 1979 to 2002. This type of run may be very different from a continuous 24 year run of a coupled model, and the interannual variation in CPL is different from the continuous run that driven by internal model dynamics. However, the main aim of the work is to evaluate the impact of air–sea interaction in RCM, so the experiment design can achieve such a purpose.

[10] The NCEP Climate Forecast System Reanalysis daily data [CFSR data; Saha et al., 2010] is used to test the simulation results. The grid spacing of the CFSR data is approximately 0.5° × 0.5°, and the starting year of CFSR data is 1979. The CPC Merged Analysis of Precipitation data [CMAP; Xie and Arkin, 1997] are used to test the simulated precipitation.

3 The Simulation Results

[11] The SST is closely associated with atmospheric circulation and its variation; therefore, the distribution of the summer mean (JJA) simulated SST is first analyzed. The HadISST data are looked as the observation dataset to compare with the simulated SST results in the coupled model. In the western Pacific and South China Sea, the pattern of the simulated JJA SST (Figure 1b) is consistent with HadISST (Figure 1a); the simulated SST is higher than HadISST in the eastern Bay of Bengal, the South China Sea, and the western tropical Pacific, the deviation is about 0.4°C (Figure 1c); the temporal correlation characteristics between the simulated SST and HadISST from 1979 to 2002 are shown in Figure 1d, and the coefficients in most areas are higher than 0.5. In general, the simulated SST bias is less than 1°C, and high correlation coefficients also prove that the variation of the simulated SST is reasonable.

Figure 1.

The summer (JJA) mean SST (unit: °C) from 1979 to 2002 in (a) HadISST, (b) CPL, (c) the bias between CPL and HadISST, and (d) the temporal correlation coefficients between the simulated SST in CPL and HadISST (shaded areas exceed 0.5; P(|r| > 0.49) = 0.01).

[12] The simulation of precipitation is an important index to illustrate the simulation ability of a climate model, and the results are shown in Figure 2. The distribution of the simulated precipitation in CPL (Figure 2b) is more similar to the CMAP precipitation (Figure 2c) than that of the precipitation in CTR (Figure 2a). Precipitation is abundant over the ocean in both the CPL model and CMAP data, while the rainfall centers in the uncoupled model are on land. The temporal correlation coefficients between the rainfall in the CTR and the rainfall of CMAP data (Figure 2d) are not significant in the Bay of Bengal and the South China Sea. Comparing with the results of the coefficients between the CTR and CMAP, the correlation coefficients between the rainfall in CPL and the rainfall of CMAP (Figure 2e) are relatively higher in the Bay of Bengal (10–22°N, 90–100°E), South China Sea (7–15°N, 115–120°E), the eastern Philippines (7–20°N, 125–140°E), the north-western Pacific (25–35°N, 130–140°E), and the Yangtze River Basin in China (25–32°N, 110–120°E).

Figure 2.

The summer mean rainfall averaged from 1979 to 2002 in (a) CTR (unit: mm d−1), (b) CPL (unit: mm d−1), (c) CMAP (unit: mm d−1), and the temporal correlation between the simulation and the CMAP (d, between CPL and CMAP; e, between CTR and CMAP; P(|r| > 0.39) = 0.05). Figure 2f shows the five areas in Tables 1 and 2.

Table 1. The JJA Mean Precipitation Averaged From 1979 to 2002 in Different Areas (Unit: mm d−1)
Region Number12345
NameBay of BengalSouth China SeaEastern PhilippinesNorth-western PacificThe Yangtze River Basin
Definition(90–100°E,(115–120°E,(125–140°E,(130–140°E,(110–120°E,
10–22°N)7–15°N)7–20°N)25–35°N)25–32°N)
CTR7.064.845.167.006.30
CPL11.298.017.038.405.43
CMAP10.578.827.636.326.13
Table 2. The Correlation Coefficients Between the Simulation Precipitation and CMAP Precipitation
Region Number12345
NameBay of BengalSouth China SeaEastern PhilippinesNorth-western PacificThe Yangtze River Basin
Definition(90–100°E,(115–120°E,(125–140°E,(130–140°E,(110–120°E,
10–22°N)7–15°N)7–20°N)25–35°N)25–32°N)
CTR and CMAP0.230.400.17340.540.34
CPL and CMAP0.320.500.420.610.69

[13] In order to analyze the rainfall simulation ability of the CPL quantitatively, the above five areas are further studied (Figure 2f; Table 1). In the eastern Bay of Bengal, South China Sea, and the eastern Philippines, the rainfall in CTR is much weaker than the rainfall of the CMAP, and the precipitation in CPL is consistent with that of CMAP data to some extent; in the north-western Pacific, the simulated rainfall bias in CPL is higher than that in CTR; in the Yangtze River Basin of China, the simulated precipitation of CTR is stronger than that of CMAP data. Considering both Figure 2 and Table 1, the precipitation of the uncoupled model is heavier on land and weaker on the ocean than that of CMAP.

[14] Table 2 shows the temporal correlation coefficients between the simulated precipitation and CMAP precipitation in summer from 1979 to 2002 in these five areas. In all the five areas, the coefficients between CTR and CMAP are lower than those between CPL and CMAP.

[15] The lateral boundary conditions in the uncoupled model (CTR) and the air–sea coupled model (CPL) are the same; therefore, the water vapor from the lateral boundary in the two models is also the same. However, different sea surface conditions will affect both the atmospheric circulation and the water vapor transport considering the different air–sea interaction processes. The water vapor transport is inline image, where inline image is the wind speed, and q is the specific humidity.

[16] Water vapor transport in the lower troposphere (850 hPa) is analyzed first. The water vapor (JJA mean) from 1979 to 2002 is shown in Figure 3. In CTR (Figure 3a), water vapor is weak in the South China Sea and the western Pacific, and the maximum water vapor center is located in the southwest of China. In the Bay of Bengal and Indo-China, zonal (eastward) water vapor (which is carried by the Indian monsoon) is weaker than that in CPL (Figure 3b) and CFSR data (Figure 3c). In CTR, the western Pacific subtropical high is not simulated well, and the meridional water vapor in the south-western China is much stronger than that in the CFSR data. The coupled model can simulate the climatic distribution of water vapor well for the Bay of Bengal, Indo-China, and South China Sea regions.

Figure 3.

The summer mean water vapor flux averaged from 1979 to 2002 at 850 hPa (contoured according to the value of the vector; the value of the shaded area exceeds 0.06; In the gray areas, the surface pressure is less than 850 hPa; unit: 10−3 kg kg−1 m s−1) simulated by the (a, CTR) uncoupled model, (b, CPL) the coupled model, and (c) the water vapor flux in CFSR reanalysis data.

[17] The spatial correlation coefficients between the simulation and the CFSR data for water vapor flux (850 hPa) are given in Figure 4. The coefficients between the CPL and CFSR reanalysis data (broken line) are always much higher than those between the CTR and CFSR data (solid line). In CPL, the simulated zonal water vapor flux and total water vapor (inline image) are more consistent with the CFSR data, and the coefficients are higher than 0.5 (Figures 4a and 4c); for the meridional vapor transport, the coefficients are between 0.4 and 0.6 (Figure 4b). In CTR, the coefficients for both zonal and meridional vapor transport are less than 0.5.

Figure 4.

The spatial correlation coefficients between CFSR and the simulation results (the solid line corresponds to CTR; the broken line corresponds to CPL; the straight line is at 0.04, P(|r| > 0.04) = 0.01) for (a) summer mean zonal vapor transport, (b) meridional vapor transport, (c) and total (the values of the vector) vapor transport from 1979 to 2002.

[18] The interannual variability is further analyzed. The mean square deviation is used to represent the interannual variability calculated by inline image, where Qt is the summer mean water vapor in year t, inline image is the average from 1979 to 2002, and Qmsd is the interannual variability. n is 24.

[19] The interannual variability of zonal and meridional water vapor transport (850 hPa) is shown in Figure 5. The maximum variability centers for zonal transport are located in the western Pacific and the South China Sea for CFSR data (Figure 5e). The simulation results indicate that in CTR (Figure 5a), the variability is much weaker than CFSR data over ocean, and the spatial distribution and center strength in CPL (Figure 5c) are more similar to the CFSR data than that in CTR. The simulated variability of the meridional water vapor transport (Figure 5b) over land is similar to the CFSR data (Figures 5b, 5d, and 5f). The uncoupled model underestimates the interannual variability in the north-western Pacific. The maximum variability centers occur in regions of eastern China, and both the coupled model and the uncoupled model can simulate the relevant characteristics.

Figure 5.

The interannual variability of zonal water vapor flux from 1979 to 2002 (a, CTR; c, CPL; and e, CFSR), and the interannual variability of meridional water vapor flux from 1979 to 2002 (b, CTR; d, CPL; and f, CFSR) (unit: 10−3 kg kg−1 m s−1) (in the gray areas, the surface pressure is less than 850 hPa).

[20] The differences in water vapor transport between the coupled model and the uncoupled model are further analyzed. The differences in water vapor transport are separated into two parts. For example, Δ(qu) = qΔu + uΔq, where qΔu indicates the part caused by the zonal wind, and uΔq indicates the part caused by the specific humidity.

[21] The differences between the zonal water vapor transports (850 hPa) of the two models are analyzed in Figure 6. The simulated results (1979–2002) in CTR (Figure 6a) show that the atmospheric model underestimates the zonal water vapor over the ocean, especially in the Bay of Bengal and the South China Sea, compared with the CFSR data (Figure 6b). The zonal water vapor flux differences between the coupled model and the uncoupled model are shown in Figure 6c. The zonal water vapor fluxes in the coupled model over areas from the Bay of Bengal to the South China Sea are higher than that in CTR, and the deviation center is consistent with the CFSR data. The differences caused by specific humidity (uΔq; Figure 6d) are much weaker than those caused by zonal wind (qΔu; Figure 6e). The zonal water vapor differences between the coupled model and the uncoupled model averaged from 1979 to 2002 are mainly determined by the differences of zonal wind.

Figure 6.

The summer mean zonal water vapor flux at 850 hPa (1979–2002 average; unit: 10−3 kg kg−1 m s−1; a, CTR; b, CFSR; c, the difference between CPL and CTR, Δ(qu); d, uΔq; e, qΔu).

[22] The differences between the meridional water vapor transport values (850 hPa) of the two models are shown in Figure 7. The maximum meridional water vapor flux centers are located in South China (Figures 7a and 7b). The meridional water vapor flux of CTR over the west of South China is much higher than that of the CFSR data. The differences between CPL and CTR (Figure 7c) demonstrate that the simulated results of CPL are lower for the west of South China than the results of CTR. The differences caused by specific humidity (vΔq; Figure 7d) are much weaker than those caused by meridional wind (qΔv; Figure 7e). The meridional water vapor differences between the coupled model and the uncoupled model averaged from 1979 to 2002 are determined by the differences of meridional wind.

Figure 7.

The same as Figure 6 but for meridional water vapor flux.

[23] Simulated winds are closely associated with the distribution geopotential height. The spatial coefficients between the simulation and CFSR data are calculated for the summer mean geopotential height, zonal wind, meridional wind, and specific humidity at difference isobaric surfaces. Compared to the CFSR data, both the uncoupled model (solid line) and the coupled model (broken line) can simulate the spatial distribution of the geopotential height effectively, and the coefficients exceed 0.9 in the middle and upper troposphere (Figure 8a). The results of CPL for zonal wind (Figure 8b) are more similar to those of the CFSR data, in the middle and lower troposphere. The maximum coefficients for meridional wind (Figure 8c) appear in the middle troposphere, and from 1000 hPa to 100 hPa, the coefficients in CPL are all higher than those in CTR. In the middle and lower troposphere, the coefficients between the simulated results and the CFSR data for specific humidity (Figure 8d) exceed 0.6, and the results of the coupled model are similar to the results of the uncoupled model.

Figure 8.

The spatial correlation coefficients (CTR: solid line; CPL: broken line) between simulated results and the CFSR data for the (a) summer mean geopotential height, (b) zonal wind, (c) meridional wind, (d) and specific humidity at different isobaric surfaces.

[24] Considering the previous results, the performance of the coupled model simulation appears to be better than that of the uncoupled model, and the air–sea interaction affectscirculation, especially in the lower and middle troposphere. The improvement of the simulated water vapor flux in CPL is related to the simulation of the winds, which are determined by the geopotential height.

[25] The distribution of the simulated geopotential height in CPL (Figure 9b) is more similar to that described by the CFSR data (Figure 9c); the geopotential height simulated by the uncoupled model (Figure 9a) over the north-western Pacific is stronger than that of the CFSR data; the geopotential height bias in CPL experiments compared to CFSR is much lower (Figure 9e) than that in CTR experiments (Figure 9d).

Figure 9.

The summer mean geopotential height (850 hPa) from 1979 to 2002 in (a) CTR, (b) CPL,(c) CFSR, (d) the bias between CTR and CFSR, and (e) the bias between CPL and CFSR (unit: dagpm; in the gray areas, the surface pressure is less than 850 hPa).

[26] The longitudinal and zonal gradient of the geopotential height is shown in Figure 10. The zonal gradient (inline image, contoured) and the meridional wind (shaded) are shown in Figures 10a, 10c, and 10e; the longitudinal gradient (inline image, contoured) and the zonal wind (shaded) are shown in Figures 10b, 10d, and 10f; and h is the geopotential height at 850 hPa. The meridional wind in CTR is stronger in the western South China than that in CPL or the CFSR data, corresponding to a strong center of inline image. Both models underestimate the strength of inline image in areas to the south of 20°N, and the simulated zonal wind is weaker than shown in the CFSR data; however, the pattern of the zonal wind and inline image in CPL is more similar to that of CFSR data than is the CTR.

Figure 10.

The zonal gradient of geopotential height at 850 hPa (inline image, contoured; unit: 10−5 gpm m−1) and the meridional wind (shaded; unit: m s−1) in (a) CTR, (c) CPL, and (e) CFSR data; the meridional gradient of geopotential height (inline image, contoured; unit: 10−5 gpm m−1) and the zonal wind (shaded; unit: m s−1) in (b) CTR, (d) CPL, and (f) CFSR data (in the gray areas, the surface pressure is less than 850 hPa).

[27] The improvements when using the CPL model are expected mainly in the low troposphere. Hence, the correlation coefficients between the geopotential height at 850 hPa and the local SST field are analyzed (Figure 11). The maximum positive coefficients are observed for the South China Sea and the eastern Bay of Bengal (10–20°N, 90–120°E). The correlation coefficients between geopotential height and SST in the uncoupled model are also positive in the South China Sea (Figure 11a), but the coefficients are much less than those between the CFSR data and HadISST. In the coupled model (Figure 11b), the coefficients in South China Sea and the Bay of Bengal are higher than 0.5, which agree with those between HadISST and CFSR geopotential height.

Figure 11.

The temporal correlation coefficients between the JJA mean geopotential height at 850 hPa and SST (a, HadISST and CTR geopotential height; b, the simulated SST and geopotential height in CPL; c, HadISST and the CFSR geopotential height). The shaded area exceeds 0.4, and P(|r| > 0.39) = 0.05.

[28] In general, warm SSTs cause local thermal lows, which represent the thermal forcing of the ocean to the atmosphere; the coefficients should therefore be negative. The nature of local air–sea interaction can be understood from the evolution of lead–lag correlation between the atmospheric variables and SST [Von Storch, 2000; Wu et al., 2006]. Figure 12 shows the lead–lag correlation between the geopotential height at 850 hPa and the local SST. The results indicate that the coefficients are the same between the SST-lead and SST-lag correlation in the control experiment. In the tropical ocean, especially the South China Sea, the SST-lead correlation coefficients are very different from the SST-lag correlation coefficients according to the results between HadISST and CFSR Geopotential height (Figures 12c and 12f) or the coupled experiment (Figures 12b and 12e). The uncoupled model cannot accurately simulate the forcing of the atmosphere to the ocean, which can be seen from the results of SST-lag correlation (Figure 12d). In regions where the influences of the atmosphere on the ocean are noticeable, the simulation results of the uncoupled model differ from the observed results.

Figure 12.

Lead–lag correlation between geopotential height at 850 hPa and SST. (a, between HadISST in June and CTR geopotential height in July; b, between the SST in June and geopotential height in July simulated by the coupled model; c, between HadISST in June and CFSR geopotential height in July; d, between HadISST in August and CTR geopotential height in July; e, between the SST in August and geopotential height in July simulated by the coupled model; f, between HadISST in August and CFSR geopotential height in July). The shaded area exceeds 0.4, and P(|r| > 0.39) = 0.05.

[29] Circulations in the middle and lower troposphere are more accurately simulated by the coupled model than the uncoupled model, and integrated water vapor transport is further analyzed. The integrated (from the surface to 300 hPa) water vapor transport is shown in Figure 13. The results of CTR (Figure 13a) show that the water vapor transport belt lies from the Bay of Bengal to the southern edge of the Tibetan Plateau, and thence to eastern China. The simulation of water vapor transport in CPL (Figure 13b) is more accurate than in the CTR, and the transport paths from the Bay of Bengal, the Philippine Islands, and the north-western Pacific are similar to those in the CFSR data (Figure 13d). The differences between CTR and CPL (Figure 13c) indicate that the improvement of the coupled model is presented as the simulated water vapor flux over the Bay of Bengal and south-western China.

Figure 13.

Integrated summer water vapor transport (a: CTR; b: CPL; c: the differences between CPL and CTR; d: CFSR; unit: kg m−1 s−1, shaded areas denote absolute values greater than 150).

[30] Because of the same lateral conditions, the water vapor differences between CTR and CPL are the results of different underlying surface forcing. The air–sea interaction process first affects circulation and then influences water vapor transport.

[31] The relationship between SST and water vapor flux is analyzed to study the influence of air–sea interactions on circulation and water vapor transport. The combined empirical orthogonal function (EOF) decomposition for the three variables, the zonal vapor flux, meridional vapor flux, (integrated from the surface to 300 hPa) and SST, is calculated (Figure 14) to analyze the relationship between water vapor transport and SST. The variance contribution of the first EOF mode is 92% for the CFSR data, 83% for CTR, and 89% for CPL. First, the SST distribution of the combined EOF decomposition in the CTR run (Figure 14b) is similar to the HadISST (combined EOF results of CFSR moisture transport and HadISST; Figure 14f). In the CFSR data, the zonal water vapor center is positive in the Bay of Bengal and negative in the western Pacific (Figure 14e); this result is consistent with the results of CPL (Figure 14c). However, the results of the uncoupled model (Figure 14a) are conflict with the CFSR data in the South China Sea.

Figure 14.

The first combined EOF mode for water vapor flux (vector; a, c, and e) and SST (b, d, and f). Figures 14a and 14b are the CTR results; Figures 14c and 14d are the CPL results; Figures 14e and 14f are the CFSR data and HadISST results.

[32] The correlation coefficients between the magnitude of the vertically integrated water vapor flux (inline image) and SSTs are shown in Figure 15. Here, the coefficients in Figure 15c are the results between HadISST and the CFSR water vapor flux. In CTR (Figure 15a), the coefficients are positive between the integrated vapor flux and the SST in the South China Sea and the western Pacific. In CPL (Figure 15b), the coefficients in the Bay of Bengal and the South China Sea are negative, in agreement with those described by CFSR data (Figure 15c). The positive (negative) coefficient means that the positive (negative) SST anomaly corresponds to increased water vapor. To a certain extent, the uncoupled model includes only the SST forcing process, and water vapor is more plentiful over the warmer ocean; thus, the coefficients are positive. The atmosphere also has an effect on the ocean, and water vapor may remove heat energy from the ocean; therefore, negative coefficients may reflect atmospheric forcing. The air–sea coupled model includes the air–sea interaction process; therefore, the relationship between SST and water vapor flux is consistent with that found in the CFSR data.

Figure 15.

The correlation coefficients between magnitude of integrated water vapor flux and SST (a: CTR; b: CPL; and c: CFSR water vapor flux and HadISST; the absolute values of the shaded areas exceed 0.4, and P(|r| > 0.39) = 0.05).

4 Discussion and Conclusions

[33] A regional air–sea coupled model, which is based on the RegCM3 RCM and the POM regional oceanic model, is used to simulate the climate in East Asia from 1979 to 2002.

[34] The results show that the coupled RegCM3-POM model can effectively simulate water vapor transport in East Asia. In CPL and the CFSR data, zonal water vapor centers (850 hPa) are located in the Bay of Bengal and the South China Sea, and meridional water vapor centers are located in the northern Bay of Bengal and western South China. The interannual variability of zonal water vapor transport is strong in the western Pacific and the South China Sea, and the interannual variability of meridional vapor transport is strong in eastern China. The ability of the coupled model to simulate moisture transport is improved, while the uncoupled model cannot effectively simulate moisture transport and its interannual variability in the lower troposphere.

[35] The differences between the water vapor flux of the coupled model and the uncoupled model are mainly caused by different zonal and meridional winds. Analysis of the wind fields and geopotential height shows that the spatial correlation coefficients between the coupled model and the CFSR data are higher than those between the uncoupled model and the CFSR data in the middle and lower troposphere. The coupled model can effectively simulate the gradient of the geopotential height. Thus, the results of the simulation for zonal wind, meridional wind, and water vapor in the lower troposphere and, based on these simulations, the integrated moisture transport are more accurate when using the coupled model than the uncoupled model.

[36] The relationship among geopotential height, moisture transport, and SST is further analyzed. The real-time and lead–lag relationship between geopotential height (850 hPa) and SST is accurate in the coupled model. The uncoupled model RegCM3 forced by the given SST cannot reproduce the real-time and SST-lag correlation between SST and the geopotential height. The advantage of the coupled model is the capability to simulate how the state of the atmosphere is changed due to the interaction with the surface of the ocean (turbulent transport of heat, momentum, and moisture), while the ocean adjusts to a new state with the thermal and mechanical forcing of the atmosphere. The modified state of the ocean has then a further effect on the atmosphere

[37] An analysis of the combined EOF decomposition for the three variables of zonal water vapor flux, meridional water vapor flux and SST indicates that the coupled model more accurately simulates the leading EOF modes than the uncoupled model. The correlation coefficients between the magnitude of vertical water vapor flux and SST are positive in the uncoupled run, corresponding to SST forcing. The correlation coefficients in the coupled model are negative, which is consistent with those between the CFSR data and HadISST. The air–sea interaction process modeled by the coupled model simulates more accurately the relationship between SST and the atmospheric variables considered.

Acknowledgments

[38] This study is sponsored by: the National Natural Science Foundation of China under 40805047, 41105058, and 40805039; KLME1202; the Priority Academic Program Development of Jiangsu Higher Education Institutions.