Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Model assessment of the observed relationship between El Niño and the northern East Asian summer monsoon using the Community Climate System Model Community Atmosphere Model-Community Land Model version 3 (CAM-CLM3)
 Plausibility and reproducibility of the relationship and physical mechanisms proposed between the northern East Asian summer monsoon (NEASM) and El Niño are verified and illuminated using the Community Climate System Model (CCSM) with the Community Atmosphere Model–Community Land Model version 3 (CAM-CLM3). The climate responses over East Asia and western North Pacific to El Niño are simulated in the CCSM3 using dynamic atmosphere and land with prescribed climatological and El Niño sea surface temperature (SST) anomalies. A significantly intensified NEASM is simulated for the El Niño experiments, which validates the positive correlation between NEASM precipitation and El Niño found in recent observational analysis. Analysis of lower level wind vector and vorticity in the model experiments elucidates the physical mechanism behind the positive correlation between NEASM precipitation and El Niño, and shows that the western North Pacific anticyclone plays an important role in the connection between the NEASM and the tropical SST anomalies.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In our study, we explore the relation between the NEASM and El Niño in model simulations, and elucidate the physical mechanisms governing monsoon strength and seasonality using the Community Climate System Model (CCSM) with the Community Atmosphere Model–Community Land Model version 3 (CAM-CLM3) [Collins et al., 2006]. Previous general circulation model (GCM) simulations have poorly reproduced the observed Asian summer monsoon rainfall, particularly in East Asia [Lau et al., 1996; Kang et al., 2002; Wang et al., 2004]. However, CCSM-CAM3 is known for reasonable simulation of the major features of the global hydrological cycle [Hack et al., 2006]. Meehl et al.  examined aspects of regional monsoon regimes (but not including the NEASM) in the CCSM3 with the CAM3. In their study, they concluded that most of the major 850 hPa wind characteristics in the monsoon regimes are well represented in the CCSM3, with associated regional monsoon precipitation maxima. Here, we analyze the CCSM3 with the CAM-CLM3 in both its ability to reproduce NEASM circulation patterns and rainfall and the model response to El Niño. We are investigating the response of the coupled atmosphere and land models to forcing from El Niño SST compared to climatological SST.
2. Methods and Data
2.1. Model Description and Experiments
 The CCSM3 framework allows combinations of dynamic or prescribed data models to be used for any of the model components. In our case we are using the dynamic atmosphere and land with prescribed SSTs and sea ice distributions, which have the equivalent physics of the CAM-CLM3 model. In the model El Niño experiments the Data Ocean Model (DOCN6), with annually cycling SSTs, is used to simulate the monthly forcing from SSTs of an average El Niño compared to the standard 1949∼2001 monthly climatology. The monthly average El Niño SST anomalies are generated from the Hadley Center monthly SST data [Rayner et al., 2003], using anomalies for all months that had seasonal averaged SST in the Niño 1, 2 and 3 regions warmer than 1°C above climatology for the 1900∼2003 period (Figure 1). The monthly El Niño SST anomalies are added to the climatology SST parameters for use in the DOCN6 model. The climate responses to the El Niño SST anomalies are simulated in the CCSM3 through the CAM3, and with the CLM3 using the new Moderate Resolution Imaging Spectroradiometer (MODIS) land surface parameters from Lawrence and Chase  and using the new CLM SiB surface hydrology described by Lawrence and Chase , which is used in all experiments to avoid the problems of evapotranspiration partition and surface hydrology found with the standard release of CLM 3.0. Sea ice distributions are prescribed from the 1949∼2005 monthly climatology for all experiments using the Community Sea Ice Model version 5 (CSIM5).
 We perform four experiments using 30-year equilibrium simulations throwing out the first 5 years as spin up from each simulation. The model experiments are: (1) three control runs of the CAM-CLM3 with a different initial condition and (2) a single run using average El Niño SST anomalies. Differences in initial conditions between each realization of the ensemble are set by randomly increasing or decreasing the surface temperature of all land grid cells by ±0.1°C. We combine three control runs to estimate model variability, and use the combined control run (75 years) to compare with the El Niño run (25 years). Spatial resolution of the CAM3 is a 26-level 42-wave triangular spectral truncation (T42 L26) of the atmospheric physics, but all parameter and history files are run on a rectangular geographic grid of 64 × 128 cells which equates to 2.8125° × 2.8125° (actually there is a pole offset of the grid which makes the latitude spacing slightly smaller). The CLM3 is run on the same rectangular grid as the CAM3. The DOCN6 and CSIM5 models are both run on a 180 × 360 domain which equates to 1° × 1°. For the comparisons of the model with observed results, the resolution of the model output is regridded to that of observation (i.e., 2.5° × 2.5°).
2.2. Data and Statistical Methods
 To compare the CAM-CLM3 results with observations, we use observed precipitation (mm/day) derived from the version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation [Adler et al., 2003], and mean sea level pressure (SLP (hPa)), 500 hPa geopotential height (HGT (gpm)), 150 hPa divergence (DIV (s−1)) and 850 hPa wind vector (m/s) and vorticity (s−1) calculated with mean 150 and 850 hPa, respectively, u and v winds (m/s) from the National Centers for Environmental Prediction–Department of Energy (NCEP-DOE) Atmospheric Model Intercomparison Project (AMIP-II) monthly reanalysis [Kanamitsu et al., 2002]. GPCP version-2 precipitation is a globally complete, monthly analysis of surface precipitation at 2.5° × 2.5° resolution available from January 1979 to the present. It is a merged analysis that incorporates precipitation estimates from low-orbit satellite microwave data, geosynchronous-orbit satellite infrared data, and surface rain gauge observations [Adler et al., 2003]. SLP, 500 hPa HGT, 150 hPa DIV, and 850 hPa wind vector and vorticity from the NCEP-DOE AMIP-II reanalysis are strongly influenced by observed data [Kalnay et al., 1996]. We use the NCEP-DOE AMIP-II reanalysis as fixes to known processing errors in the NCEP-NCAR (National Center for Atmospheric Research) reanalysis have been incorporated and it uses an improved forecast model and data assimilation system [Kanamitsu et al., 2002]. Spatial and temporal resolutions of the reanalysis are consistent with those of the GPCP precipitation. In this study, a 25-year climatology for 1979∼2003 is used to compare with the model results from control and El Niño simulations.
 In order to check if there is significant correlation in the spatial patterns between CAM-CLM3 control and observed climatology, we compute the spatial correlation coefficients between the model and observations on the basis each grid points in the specified region (i.e., globe or NEASM region). A Student's t-test is conducted to quantify the statistical significance of the difference of means between El Niño and normal years (i.e., climatology SST) from the simulated and observed variables (precipitation, 150 hPa DIV, SLP, and 500 hPa HGT). We use the t statistic for unequal population variances, because the F tests for the variances of the two samples show significant differences (at the 90% level; not shown here) in some regions over East Asia and the WNP (10°N∼60°N and 100°E∼190°E). Significant regions at the 90 and 95% levels are contoured. The relationship between NEASM precipitation and El Niño during the premonsoon season (December through May) is examined using a statistical forecast model of Lee et al.  (equation (1)).
where “SST in the TEP” is SST in the tropical eastern Pacific (5°S∼5°N, 170°W∼80°W), “SST in the TWP” is SST in the tropical western Pacific (0°∼15°N, 130°E∼160°E), “SST in the TIO” is SST in the tropical Indian Ocean (10°S∼10°N, 40°∼100°E), and “OHC in the TIO” is ocean heat content in the tropical Indian Ocean (20°S∼15°N, 50°E∼70°E) during the premonsoon season.
 We calculate NEASM precipitation using a statistical model where the predictors in the forecast model (SSTs and OHC) are averaged over the specified regions in the model runs during the premonsoon season. The difference between NEASM precipitation for the El Niño simulation versus the control simulation is compared with that from observed Hadley Center SST during El Niño years versus normal years. We also calculate NEASM precipitation simulated in the CAM-CLM3 runs.
3. Comparisons of CAM-CLM3 Control Run With Observations
Figures 2a and 2b show the zonal and spatial distributions of annual precipitation from the CAM-CLM3 control simulations and 1979∼2003 GPCP climatology. In order to check if there is significant correlation between the distributions, we calculate the spatial pattern correlation between model and observational precipitation over the globe (6048 data points). Annual precipitation from the CAM-CLM3 is significantly correlated with that from GPCP (r = 0.79; p value < 0.01). The tropical precipitation in the intertropical convergence zone (ITCZ) is generally well captured, which is shown in two peaks, although there is a more exaggerated ITCZ in the CAM-CLM3 than in GPCP (Figure 2a). This is due to overestimated precipitation in the tropical Indian and Pacific Oceans between 30°S∼30°N expect for equatorial region. (Figure 2b, bottom; horizontal distributions and differences of seasonal mean precipitation are shown in Figure S1 in the auxiliary material). Subtropical precipitation minima (subtropical High) are generally displaced poleward about 10°, and the secondary precipitation maxima (midlatitude frontal precipitation) are shifted equatorward in the southern hemisphere and poleward in the northern hemisphere (Figure 2a).
 Comparisons of CAM-CLM3 with observed variables over East Asia and the WNP are shown in Figure 3. During June through August (JJA), precipitation from the control simulations is overestimated over the subtropical North Pacific (i.e., 10°N∼30°N and 150°E∼170°W) and underestimated over the NEASM region (Figure 3a) relative to that from observations. CAM-CLM3 cannot reproduce the observed rainband in the NEASM region, which is due to an unrealistic dry zone over the Korean peninsula and Japan extending to the northwestern Pacific. This problem has been mentioned in previous GCM intercomparison studies [e.g., Lau et al., 1996; Kang et al., 2002] indicating the general model difficulty in simulating regional precipitation. Observed precipitation data may also not be highly reliable owing to its high spatial variability, the lack of rain gauges, especially over oceanic and unpopulated land areas, and the uncertainties and inhomogeneity in the satellite observations [e.g., Lau et al., 1996; Xie and Arkin, 1997; Adler et al., 2003]. Therefore, we use observed variables which are smoother, better observed and therefore more reliable than precipitation for comparison with model results. These include 150 hPa DIV, SLP, 500 hPa HGT, and 850 hPa wind and vorticity, which are class A variables (i.e., most reliable) from reanalysis [Kalnay et al., 1996].
 JJA upper level divergence zone is used as an NEASM index, which coincides with the lower level convergence zone of the summer monsoon [e.g., Lee et al., 2008]. JJA 150 hPa DIV from the control runs reproduces the observed monsoon band in the NEASM region, though it is overestimated over the subtropical North Pacific and underestimated over the southern EASM region (20°N∼30°N and 110°E∼145°E) (Figure 3b). The spatial distributions of averaged JJA SLP, 500 hPa HGT, and 850 hPa wind and vorticity are shown in Figures 3c, 3d, and 3e. Simulated spatial patterns of SLP, 500 hPa HGT, and 850 hPa vorticity over the NEASM region (135 data points) are significantly correlated with those from NCEP-DOE AMIP-II (r = 0.90, 0.96 and 0.37, respectively; all p values < 0.01). Vorticity is calculated with u and v winds using finite differencing at each grid point, which tends to make it less smooth than SLP and HGT. It might be one of the reasons why the correlation value of 850 hPa vorticity is smaller relative to those of SLP and 500 hPa HGT. Simulated SLP shows high-pressure center over the North Pacific, which is one of the important dynamic mechanisms in the NEASM circulation [e.g., Wang et al., 2000], although its magnitude is overestimated in comparisons with NCEP-DOE SLP (Figure 3c). Relatively low pressure during JJA over East Asia occurs in both model and observation. The spatial pattern of 500 hPa HGT from the CAM-CLM3 shows good agreement with that from observation (r = 0.96; Figure 3d). However, the 500 hPa HGT in the control model simulation over the WNP is shifted northward and pressure is too high. Simulated 850 hPa wind vector and vorticity, which are used to elucidate the physical mechanisms between the NEASM and El Niño in the next section, also generally capture the lower level wind and vorticity patterns over East Asia and the WNP (Figure 3e). The strong negative vorticity in the WNP (i.e., WNP anticyclonic circulation) is displaced poleward about 10° in the model, but is of similar magnitude to the observed circulation. This displacement affects the northern extent of the climatological southwesterly airflow into the NEASM region from the South China Sea and tropical western Pacific shifting precipitation to north. This shift partially explains the weaker model precipitation than observed precipitation in the NEASM region (Figure 3a).
4. El Niño Sensitivity Simulations
Figure 4 shows the simulated differences of precipitation, 150 hPa DIV, SLP, and 500 hPa HGT between El Niño and control runs (Figure 4, left), and the observed differences of those variables between the 5 years of highest SST anomalies in the tropical eastern Pacific covering the Niño 1, 2, and 3 regions, which is consistent with the El Niño definition in the model, and normal year (25-year mean) (Figure 4, right). Simulated pattern in Figure 4a (left) demonstrates that JJA precipitation for the El Niño run is higher than that for the control run in the NEASM region with significant positive precipitation over the monsoon band in the NEASM region. The observed pattern (Figure 4a, right) reveals more precipitation over the NEASM region during the years of high SST in the tropical eastern Pacific (i.e., El Niño year), consistent with the simulated pattern (Figure 4a, left). On the other hand, JJA precipitation in the southern EASM region for the El Niño simulation is significantly less than that for the control simulation. As mentioned in section 3, summer precipitation from the CAM-CLM3 control cannot reproduce the observed summer rainfall in the NEASM region. Thus, we examine the impact of El Niño on the NEASM using 150 hPa DIV, SLP and 500 hPa HGT, and check if there is the consistency between the results from precipitation and more reliable variables. The simulated difference of 150 hPa DIV between El Niño and control simulations shows a significant positive value in the NEASM region over 30°N∼45°N (Figure 4b, left), consistent with the observed 150 hPa DIV difference in the NEASM region (Figure 4b, right). The simulated distributions of SLP (Figure 4c, left) and 500 hPa HGT (Figure 4d, left) show negative differences in the NEASM region when subtracting the values of the control run from the El Niño run. Negative values of SLP and 500 hPa HGT in the NEASM region are also shown in the observed differences (Figures 4c, right, and 4d, right). These model and observational results using both precipitation and more reliable variables support a stronger NEASM during El Niño.
 To examine the positive correlation between NEASM precipitation and El Niño during the premonsoon season, we calculate NEASM precipitation using a statistical forecast model (equation (1) in section 2.2) using El Niño and control SST parameters in the model. El Niño year is defined using the averaged SST anomalies in Niño 1, 2 and 3 regions for the premonsoon season that are warmer than 1°C above climatology SST for the 1900∼2003 period. NEASM precipitation is calculated from SSTs in the tropical eastern Pacific, tropical western Pacific and tropical Indian Ocean, and OHC in the tropical Indian Ocean during the premonsoon season. Because the CAM-CLM3 simulations were run with climatological SSTs, OHC is not available for our analysis. We, therefore, use area-averaged SST over the same area as a proxy for OHC, which is significantly correlated with area-averaged OHC (r = 0.60; p value < 0.01). Using model SST parameters during the premonsoon season, NEASM precipitation during El Niño increases relative to the control (5.14 versus 5.06 mm/d). The difference (+0.08 mm/d) is comparable to that from observations (+0.14 mm/d). The difference in NEASM precipitation in the dynamic CAM-CLM3 between El Niño and control simulations is +0.05 mm/d. Therefore, these results are consistent with the previous observational analysis, which showed that a significant positive correlation between NEASM precipitation and El Niño during the premonsoon season [Lee et al., 2008]. They also support the conclusion that model SST parameters effectively produce ocean forcing for the statistical model of NEASM precipitation, and CAM-CLM3 dynamics reproduce precipitation result established with the statistical model.
 In order to elucidate the physical mechanisms between the NEASM and El Niño shown in the observational analysis [Lee et al., 2008], we examine mean JJA differences in 850 hPa wind vector and vorticity between the El Niño and control simulations (Figure 5, left). For the El Niño run (i.e., warm SST anomalies in the tropical eastern Pacific), there are cold SST anomalies in the tropical western Pacific, WNP, and western South Pacific (a horseshoe pattern (see Figure 1)). Cold SST anomalies in the WNP lead to weaker local atmospheric convection and thus a stronger WNP anticyclonic circulation. The stronger WNP anticyclonic circulation further leads to intensified southwesterly flows into the NEASM region which lead to an anomalously strong monsoonal circulation in the region (Figure 5, left), because the southwesterly flows originating in the Indian monsoon region and northern Australia and its neighboring sea regions are the dominant heat and moisture sources for the NEASM [e.g., Yihui, 1994]. This results in more NEASM precipitation (Figure 4a, left). This result supports the positive relation between the NEASM and El Niño shown in the model simulations as well as observational analyses (Figure 4). This result also supports the idea that the WNP anticyclone plays an important role in the connection between the NEASM and tropical Pacific SSTs [e.g., Wang et al., 2000; Chang et al., 2000a, 2000b; Wang and Li, 2004; Lee et al., 2005; Lee et al., 2008]. Simulated spatial pattern in Figure 5 (left) is significantly correlated (p value < 0.01) with observational composite differences of mean JJA 850 hPa wind vector and vorticity for the 5 years of both highest and of lowest of SST anomalies in the tropical eastern Pacific shown by Lee et al.  (Figure 5, right). Thus, this result validates a conclusion that oceanic heat sources in the tropical Pacific can impact the NEASM through the intensity of WNP anticyclonic anomalies proposed by Lee et al. .
 CAM-CLM3 is used to assess the physical relationship between the NEASM and SST anomalies during El Niño observed by Lee et al. . The spatial patterns of the differences of precipitation, 150 hPa DIV, SLP, and 500 hPa HGT between the El Niño and control simulations validate the positive correlation between the NEASM and El Niño supporting a stronger NEASM during an El Niño year. We elucidate the physical linkages between the NEASM and tropical SST anomalies proposed in the observational analysis that a stronger NEASM is related to above-normal WNP anticyclonic anomalies and thus more cyclonic anomalies in the NEASM region due to the Pacific–East Asian teleconnection, which is connected to tropical SST anomalies: During El Niño year, there are cold SST anomalies in the tropical western Pacific, WNP, and western South Pacific. The cold SST anomalies in the WNP lead to a weaker local atmospheric convection and thus a stronger WNP anticyclonic circulation. The stronger WNP anticyclonic circulation further leads to intensified southwesterly flowing into the NEASM region, which lead to an anomalously strong monsoonal circulation in the region. This results in more NEASM precipitation.
 The intensity of cyclonic circulation in the NEASM region can also affect that of WNP anticyclone, because the two regional circulations are linked by anomalously stronger westerly airflows around 30°N during El Niño (Figure 5) through a plausible positive feedback. So, what extent NEASM affects the WNP anticyclone needs to be investigated in future study using observational and modeling studies.
 CAM-CLM3 does a good job at representing the spatial patterns of precipitation over the globe and regional atmospheric dynamic fields (i.e., 150 hPa DIV, SLP, 500 hPa HGT, and 850 hPa vorticity) over the NEASM region, although with some differences from observations. However, CAM-CLM3 poorly reproduces summer precipitation over East Asia similar to other models as shown in previous intercomparision studies. We therefore used broad circulation statistics as our comparison metric.
 We wish to thank the anonymous reviewers for valuable suggestions. The first author is thankful to Department of Geography, University of Colorado for the Gilbert F. White Doctoral Fellowship. Partial supports of this work by National Science Foundation via ATM-0437538 and ATM-0639838 grants are also thankfully acknowledged.