1. Introduction
Climate models are useful tools in studying climate variability and climate change. However, the current state-of-the-art climate models generally show large biases in the monsoon precipitation simulation, especially in the Asian monsoon region (Kang et al. 2002; Zhou et al. 2008a, 2009a,b; H. Li et al. 2010; Sperber et al. 2012). The sources of the model bias may result from either atmospheric circulations or physical parameterization schemes (J. Li et al. 2010). Sperber et al. (2012) showed the clear linkage between precipitation and circulation deficiencies in models for the Asian monsoon. They indicated that the circulation and precipitation coupling is intrinsic to the tropics. Quantifying and understanding the source of model bias in monsoon simulation are crucial for the improvement of model performance.
Reanalysis datasets have been widely used in various climate studies. The outputs of reanalysis have been classified into three major categories, depending on the relative influence of the model and the assimilated observational data (Kalnay et al. 1996). For the five reanalysis datasets we used, both the zonal and meridional wind are directly assimilated from observational data, so they belong to the most reliable class; the precipitation was derived from models rather than directly assimilated by observational data, so it belongs to the third class. As for the second class, the output is a mixture of the observational data and model results, such as the humidity and surface temperature. The reanalysis data are produced by using the most advanced operational numerical models. Because of the assimilation, the atmospheric circulation in the reanalysis dataset is the best estimate we have.
Since precipitation is one critical diagnostic variable that is sensitive to both the model physics and simulated general circulation, model performance in generating reanalysis have been evaluated in some previous studies (Janowiak et al. 1998; Bosilovich et al. 2008). Reanalysis precipitation has also been analyzed in regional areas for its using in some river discharge models, regional models, and statistical downscaling techniques (Serreze and Hurst 2000). Bosilovich et al. (2008) evaluated the global precipitation of five reanalysis datasets by dividing the globe into nine regions (North Pacific, North Atlantic, Europe, etc.). They also assessed the quality of the reanalysis in several latitudinal bands, the global ocean, and the global land. The assessment, which focused on climatology of precipitation, was done based on Taylor diagrams. In previous studies, there was little evaluation on the reanalysis data using monsoon precipitation metrics, such as the climatology of global monsoon (GM) modes, interannual variability, and long-term trend of GM precipitation, although the monsoon precipitation is a rigorous test for climate models. Thus, in this study, our first goal is to evaluate the ability of the five reanalysis in reproducing the GM precipitation climate, variability, and long-term trend. Because the atmospheric circulation in the reanalysis dataset is the best estimate we have, comparing the monsoon precipitation in reanalysis data with the observations can facilitate to quantify the enhancement of monsoon precipitation simulation in an atmospheric general circulation model (AGCM) when the circulation simulation has been improved.
The precipitation change is closely related to the processes of water vapor budget. Evaluation of the global atmospheric moisture budget in National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) was carried out by Trenberth and Guillemot (1995) and Zhou et al. (1999). The atmospheric moisture budget and its regulation of the summer precipitation variability over East Asia, the Tibetan Plateau, and the southeastern United States were studied (Zhou and Yu 2005; Feng and Zhou 2012; Li et al. 2013). However, the water vapor budget associated with the global monsoon changes is not well known. Thus another goal of our present study is to examine the major components of water budget over the global monsoon region, using five reanalysis datasets. The budget analysis of water vapor can help us to understand the reasons responsible for the long-term changes of monsoon precipitation.
The regional monsoons are restricted by the global-scale overturning circulation (Trenberth et al. 2000). Wang and Ding (2006) defined the GM precipitation domain based on the annual range of precipitation and examined the trends of the GM precipitation over land using four sets of rain gauge precipitation datasets compiled for the period 1948–2003. Results indicated that there were decreasing trends in the global monsoon index (GMI) and the Northern Hemisphere monsoon index (NHMI) across the entire 56 yr. Both of them were significant at the 95% confidence level. The Southern Hemisphere monsoon index (SHMI), however, showed no significant trend. A later study of Zhou et al. (2008b) showed evidences that both the changes of monsoon precipitation intensity and monsoon area contribute to the downward trend of global monsoon precipitation amount. The overall weakening of global land monsoon rainfall accumulation is mainly contributed by the North African and South Asian monsoon, associated with the decreasing tendencies of both rainfall intensity and monsoon coverage (Zhou et al. 2008b; Zhang and Zhou 2011). The decreasing tendencies of precipitation over South Asia may be partly induced by the large increasing trend of aerosol concentrations (Turner and Annamalai 2012). Zhou et al. (2008a) performed an ensemble AGCM simulation forced by historical sea surface temperature (SST) data and found that the weakening tendency and interannual variability of global land monsoon precipitation was reasonably reproduced. Meanwhile, the models also showed deficiencies in simulating the East Asian monsoon precipitation. They hypothesized that model physics may partly explain the deficiency of precipitation simulation. In recent decades, however, the Northern Hemisphere summer monsoon and Hadley and Walker circulations have all shown substantial intensification (Wang et al. 2013).
In this paper, we aim to examine the GM precipitation and water vapor budget derived from observations and five reanalysis datasets. Multi-reanalysis datasets are applied to avoid data dependence of the results. The study aims to answer the following questions: 1) How well can the multi-reanalysis data capture the climatology, long-term trend, and interannual variability of monsoon precipitation during the past about 30 yr? 2) How large are the differences among five reanalysis datasets in GM precipitation and general circulation? 3) Which components of water budget dominate the long-term trend and interannual variability of GM precipitation?
The remainder of the paper is organized as follows: Section 2 describes the observational and reanalysis data as well as the analyses methods. Section 3 presents the results, including the climatology, long-term trend and interannual variability of GM precipitation, along with the major components of water budget. Concluding remarks including a discussion are given in section 4.
2. Data and analysis method description
a. Data description
Two sets of observational monthly precipitation data are used: 1) Global Precipitation Climatology Project (GPCP) data (Huffman et al. 1997; Adler et al. 2003) and 2) CPC Merged Analysis of Precipitation (CMAP) data (Xie and Arkin 1997). Both of them are available in monthly and 2.5° × 2.5° spatial resolutions from 1979 to 2012. The two datasets are both derived from a mix of satellite estimates over ocean and land and rain gauge measurements from land and atolls. Although the input data have similar sources, blending methodologies of the two datasets differ considerably from each other (Adler et al. 2003; Xie et al. 2003; Yin et al. 2004). Because of the deficiencies in the CMAP dataset (Zhou et al. 2008a), GPCP is used here as the observation to evaluate the reanalysis data. CMAP is only used to measure the observational uncertainty compared to GPCP.
The reanalysis physical variables used in this study include the monthly specific humidity, the meridional and zonal wind components, and the surface pressure. Five reanalysis datasets are used:
NCEP–U.S. Department of Energy (DOE) Atmospheric Model Intercomparison Project II (AMIP-II) reanalysis (NCEP-2) (Kanamitsu et al. 2002);
40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) (Uppala et al. 2005).
Japanese 25-yr Reanalysis Project (JRA-25) (Onogi et al. 2007);
Interim ECMWF Re-Analysis (ERA-Interim), which is an interim reanalysis to replace ERA-40 (Dee et al. 2011); and
Modern-Era Retrospective Analysis for Research and Applications (MERRA), which was conducted by National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC) (Bosilovich 2008; Schubert et al. 2008).
Table 1 presents detailed information on five reanalysis datasets. The reanalysis data are produced by different numerical weather models or assimilation systems; thus, they are independent from each other. While ERA-40 and ERA-Interim are produced by the same numerical weather model, their assimilation algorithms are different. In the process of data assimilation, NCEP-2, ERA-40, JRA-25, and MERRA apply three-dimensional variational data assimilation (3DVAR), while ERA-Interim uses the most advanced 4DVAR. Since the NCEP–NCAR reanalysis dataset (NCEP-1) employs the same model and data assimilation system as NCEP-2, only NCEP-2 is analyzed. The “radiances” in Table 1 means only the radiance products in satellite data have been assimilated into reanalysis data. The word “retrievals” means both radiance and retrievals products are assimilated (e.g., wind profiles, rain rates). The rain rates over the ocean from the Special Sensor Microwave Imager (SSM/I) are assimilated in NCEP-2 and MERRA (Kanamitsu et al. 2002; Bosilovich 2008). Satellite data are incorporated in MERRA, including SSM/I rain rate and Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) rain rate (Rienecker et al. 2011).
Detailed information on the five reanalysis datasets used in this study.
All reanalysis data are interpolated onto a 2.5° by 2.5° grid by using a bilinear interpolation technique. We use the “global monsoon year” defined by Wang et al. (2012): that is, one global monsoon year is starting from 1 May to the following 30 April. Based on the original time periods covered by the five reanalysis datasets (Table 1), our study is focused on the time period from May 1979 to April 2011 for NCEP-2, JRA-25, ERA-Interim, and MERRA and from May 1979 to April 2001 for ERA-40.
b. Analysis method
1) Definition of GM domain
Following Wang and Ding (2008), the GM domain is defined by the region in which the annual range (AR) of precipitation exceeds 2.5 mm day−1 and the local summer (May–September in the Northern Hemisphere and November–March in the Southern Hemisphere) precipitation exceeds 55% of annual total amount. Here AR is defined as the difference of May–September (MJJAS) and November–March (NDJFM) mean precipitation if the monsoon region is located in in the Northern Hemisphere (NH) or the difference of the NDJFM and MJJAS mean precipitation if the monsoon region is located in the Southern Hemisphere (SH).
The GM domains defined by GPCP data are shown in Fig. 1. The six major monsoon regions are the North African (NAF), the Southern African (SAF), the Asian (ASN), the Australian (AUS), the North American (NAM), and the South American (SAM) monsoons. The North African monsoon domain covers the region of the more commonly described West African monsoon. The regions where land–ocean thermal contrasts are not involved are not regarded as monsoon regions, such as the central South Pacific region (CSP) (Liu et al. 2009).
The global monsoon precipitation domain based on the GPCP data from 1979 to 2011. The red dots indicate land points, and the green dots, oceanic points. The blue curves outline the global monsoon domain areas. There are six major regional monsoons: North African monsoon (NAF), Southern African monsoon (SAF), Asian monsoon (ASN), Australian monsoon (AUS), North American monsoon (NAM), and South American monsoon (SAM).
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The global monsoon precipitation domain based on the GPCP data from 1979 to 2011. The red dots indicate land points, and the green dots, oceanic points. The blue curves outline the global monsoon domain areas. There are six major regional monsoons: North African monsoon (NAF), Southern African monsoon (SAF), Asian monsoon (ASN), Australian monsoon (AUS), North American monsoon (NAM), and South American monsoon (SAM).
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The global monsoon precipitation domain based on the GPCP data from 1979 to 2011. The red dots indicate land points, and the green dots, oceanic points. The blue curves outline the global monsoon domain areas. There are six major regional monsoons: North African monsoon (NAF), Southern African monsoon (SAF), Asian monsoon (ASN), Australian monsoon (AUS), North American monsoon (NAM), and South American monsoon (SAM).
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
2) Measurement of monsoon precipitation
Three methods are used to measure the variation of monsoon precipitation (Zhou et al. 2008a; Wang and Ding 2006; Wang et al. 2012, 2013).
Following Wang et al. (2012, 2013), the first method measures the mean monsoon precipitation variation over Northern Hemisphere monsoon regions, Southern Hemisphere monsoon regions, and global monsoon regions. The area-weighted averaged MJJAS monsoon precipitation in NH and the area-weighted averaged NDJFM monsoon precipitation in SH are used to measure the strength of the NH and SH summer monsoon precipitation [NH summer monsoon index (NHSMI) and SH summer monsoon index (SHSMI)]. The sum of NHSMI and SHSMI, termed the global summer monsoon index (GSMI), is used to quantify the global summer monsoon strength.
The second approach measures the coherent pattern of the change in GM precipitation. Since the values of AR reflect the local monsoon intensity, the leading EOF mode of AR represents the major spatial variability of monsoon intensity and the corresponding PC1 represents the major interannual variability of monsoon intensity. In our analysis, PC is termed as the annual range index (ARI).
The third method measures the statistical significance of the AR trend for each grid point within the GM domain. Both the trend-to-noise ratio (T2N) and Mann–Kendall rank statistics (MK) (Sneyers 1990) are used to test the significance of linear trends.
3) The calculation of PCC and RMSE
The quality of reanalysis data is measured by pattern correlation coefficient (PCC) and root-mean-square error (RMSE) with reference to the observational data.
Pattern correlation statistics come in two types: centered and uncentered. The centered (uncentered) statistic measures the similarity of two patterns after (without) removal of the global mean. In our analysis, we applied the centered PCC, using area-weighted PCC and RMSE.
4) Major water budget components
In this study, our focus is mainly on the global monsoon region, which is large enough that the runoff can be ignored. The sum of water vapor terms is generally equal to the spatial pattern of precipitation (not shown), demonstrating that the contribution of runoff is negligible.
The total water vapor transport includes stationary and transient components (Trenberth and Guillemot 1995; Seager et al. 2010; Feng and Zhou 2012; Li et al. 2013), the stationary component is predominant in monsoon area (Simmonds et al. 1999; Zhou and Yu 2005), and thus we focus on this dominant part in our analysis.
3. Results
In the following analysis, first the quality of reanalysis data in reproducing the climatological GM precipitation and GM domain is assessed. Then the long-term trend and interannual variability of GM precipitation revealed by observation and different reanalysis data are examined. Finally, the climatology, long-term trend and interannual variability of evaporation, ΔCWV, and water vapor convergence are measured to identify which water budget component dominates the GM precipitation variation.
a. Climatology of GM precipitation
The climatological mean (May 1979–April 2001 for ERA-40 and May 1979–April 2011 for the other datasets) of annual precipitation and the differences between reanalysis datasets and GPCP are shown in Fig. 2. The quality of reanalysis dataset is measured by PCC and RMSE with reference to the GPCP. In the observation (Fig. 2a), the precipitation tends to be maximized at the tropical areas and is generally equatorially symmetric. The PCC and RMSE in Fig. 2a are calculated between GPCP and CMAP, measuring the observational uncertainty. Five reanalysis datasets reasonably reproduce the observational patterns. PCCs (RMSEs) are 0.88, 0.86, 0.87, 0.92, and 0.90 (1.52, 1.87, 1.47, 1.89, and 0.99) for NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA, respectively. The intensity of intertropical convergence zone (ITCZ) precipitation is ~4 mm day−1 larger than observations in NCEP-2, ERA-40, and ERA-Interim and ~3 mm day−1 lower than observations over South America in JRA-25 and MERRA.
The long-term mean of (a) observed precipitation (1979–2011). The precipitation differences between observed and the reanalysis datasets: (b) NCEP-2 (1979–2011), (c) ERA-40 (1979–2001), (d) JRA-25 (1979–2011), (e) ERA-Interim (1979–2011), and (f) MERRA (1979–2011). Units are millimeters per day.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The long-term mean of (a) observed precipitation (1979–2011). The precipitation differences between observed and the reanalysis datasets: (b) NCEP-2 (1979–2011), (c) ERA-40 (1979–2001), (d) JRA-25 (1979–2011), (e) ERA-Interim (1979–2011), and (f) MERRA (1979–2011). Units are millimeters per day.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The long-term mean of (a) observed precipitation (1979–2011). The precipitation differences between observed and the reanalysis datasets: (b) NCEP-2 (1979–2011), (c) ERA-40 (1979–2001), (d) JRA-25 (1979–2011), (e) ERA-Interim (1979–2011), and (f) MERRA (1979–2011). Units are millimeters per day.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Monsoon climate is characterized by a contrasting rainy summers and dry winters (Webster 1987). The annual cycle of monsoon precipitation is primarily presented by the solstitial mode and the equinoctial asymmetric mode in spatial pattern (Wang and Ding 2008). The solstitial and equinoctial modes represent the spatial pattern of first and second multivariable EOFs (MV-EOFs) modes of the annual cycle of precipitation and 850-hPa winds, respectively. The solstitial mode is represented by the precipitation difference pattern between June–September (JJAS) and December–March (DJFM). Five reanalysis datasets reasonably reproduce the precipitation spatial patterns of the solstitial mode, with PCCs of 0.88, 0.88, 0.90, 0.93, and 0.91 and RMSEs of 2.07, 1.83, 1.63, 1.70, and 1.40, for NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA, respectively. The precipitation over Asian–Australian monsoon region is slightly overestimated in five reanalysis data (Figs. 3b–f).
(left) The solstice mode (Jul–Sept minus Dec–Mar average precipitation) and (right) the equinox asymmetric mode (Apr–May minus Oct–Nov average precipitation) obtained from reanalysis datasets: (a),(g) GPCP; (b),(h) NCEP-2; (c),(i) ERA-40; (d),(j) JRA-25; (e),(k) ERA-Interim; and (f),(l) MERRA . Units are millimeters per day. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(left) The solstice mode (Jul–Sept minus Dec–Mar average precipitation) and (right) the equinox asymmetric mode (Apr–May minus Oct–Nov average precipitation) obtained from reanalysis datasets: (a),(g) GPCP; (b),(h) NCEP-2; (c),(i) ERA-40; (d),(j) JRA-25; (e),(k) ERA-Interim; and (f),(l) MERRA . Units are millimeters per day. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(left) The solstice mode (Jul–Sept minus Dec–Mar average precipitation) and (right) the equinox asymmetric mode (Apr–May minus Oct–Nov average precipitation) obtained from reanalysis datasets: (a),(g) GPCP; (b),(h) NCEP-2; (c),(i) ERA-40; (d),(j) JRA-25; (e),(k) ERA-Interim; and (f),(l) MERRA . Units are millimeters per day. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The equinoctial asymmetric mode is represented by the average of April–May (AM) precipitation minus October–November (ON) precipitation pattern. In the observational data we can clearly see the pattern and location of the spring and fall ITCZ precipitation. The precipitation patterns of equinoctial asymmetric mode in five reanalysis data resemble the observation (Figs. 3h–l). As demonstrated by PCCs, the solstitial mode is simulated better than the asymmetric mode in five datasets.
The AR mode of precipitation is measured by the difference of MJJAS and NDJFM mean precipitation (Fig. 4). The major monsoon rainy regions tend to reside on each side of the equatorial perennial precipitation regions. Five reanalysis datasets have reasonable performances in reproducing the observed major monsoon rainy regions over the South Asian, the Southern African, the Australian and the South American monsoon regions. The PCCs (RMSEs) are 0.85, 0.85, 0.87, 0.91, and 0.88 (1.81, 1.53, 1.38, 1.42, and 1.22 mm day−1) for NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA, respectively. The western Pacific monsoon precipitations are weakly reproduced by the five datasets. They all overestimate the AR intensity in this region.
Annual range of precipitation (color shading, mm day−1) derived from (a) observations, and reanalysis datasets (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA. Black curves outline the global monsoon domain captured in each datasets. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim, and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Annual range of precipitation (color shading, mm day−1) derived from (a) observations, and reanalysis datasets (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA. Black curves outline the global monsoon domain captured in each datasets. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim, and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Annual range of precipitation (color shading, mm day−1) derived from (a) observations, and reanalysis datasets (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA. Black curves outline the global monsoon domain captured in each datasets. The time periods are from May 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim, and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
We further examine the quality of five reanalysis datasets in capturing the GM domain, as highlighted by black curves in Fig. 4. The GM domains revealed by five reanalysis datasets are generally consistent with those in GPCP. However, the oceanic monsoon regions in the Northwest and South Pacific are not well reproduced. Most reanalysis datasets failed to capture the monsoon regime in the North American and East Asian sectors. Because the two regions are areas where the zonally land–ocean thermal contrast and meridional hemispheric thermal contrast coexist (Zhou et al. 2008a,b), the quality of reanalysis dataset needs to be improved in these domains.
The PCC and RMSE values of solstitial modes, equinoctial asymmetric modes, and AR modes in five reanalysis datasets between 45°S and 45°N with reference to the combined GPCP are showed in Table 2. We first calculate the PCC and RMSE between GPCP and CMAP to quantify the observational uncertainty. There is remarkable agreement between GPCP and CMAP in the spatial pattern of climatological mean. The reanalysis datasets reproduce the spatial pattern in the GPCP well, with PCCs larger than 0.85 and RMSEs less than 2.0 mm day−1. Among five reanalysis datasets ERA-Interim (NCEP-2) is the best (worst) in reproducing the observational spatial distribution.
The PCC and RMSE of annual range modes: JJAS minus DJFM (solstice) mode and AM minus ON (equinox asymmetric) mode for the five reanalysis datasets between 45°S and 45°N with reference to GPCP. The first row shows the PCC and RMSE between GPCP and CMAP to quantify the observational uncertainty.
Another important climatological characteristic of GM precipitation is the accumulated summer precipitation distribution as function of its intensity, which is shown in Fig. 5. The interval for rain rates is 1 mm day−1. The observed distribution shows a maximum accumulation at the intermediate range of 8–9 mm day−1 and tails off toward light rain very fast. The distributions of five reanalysis datasets show reasonable overall shape. However, NCEP-2, ERA-40, JRA-25, and ERA-Interim slightly overestimate (underestimate) the precipitation with the intensity larger than ~12 mm day−1 (smaller than ~12 mm day−1). ERA-Interim slightly overestimates the moderate rain (10–20 mm day−1), which resulted from its frequency bias rather than intensity. NCEP-2, ERA-40, and JRA-25 overestimate the precipitation with the intensity larger than 10 mm day−1. MERRA is the best among five datasets in reproducing the overall shape and magnitude, except it slightly underestimates the accumulated amount of precipitation between 5 and 15 mm day−1.
The distribution of accumulated global monsoon summer (MJJAS in the Northern Hemisphere and NDJFM in the Southern Hemisphere) precipitation amount as function of its intensity for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. The interval for intensity is 1 mm day−1. The time period is from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The distribution of accumulated global monsoon summer (MJJAS in the Northern Hemisphere and NDJFM in the Southern Hemisphere) precipitation amount as function of its intensity for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. The interval for intensity is 1 mm day−1. The time period is from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The distribution of accumulated global monsoon summer (MJJAS in the Northern Hemisphere and NDJFM in the Southern Hemisphere) precipitation amount as function of its intensity for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. The interval for intensity is 1 mm day−1. The time period is from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
b. Long-term trend and interannual variability of GM precipitation
1) Mean intensity of GM precipitation
The normalized time series of the Northern Hemisphere, Southern Hemisphere, and GM mean precipitation (NHSMI, SHSMI, and GSMI, respectively) in local summer are shown in Fig. 6. All indices are derived from five reanalysis datasets, along with GPCP. As shown in Table 3, the observational climate mean values of NHSMI, SHSMI, and GSMI are 7.10, 6.21, and 6.72 mm day−1. The corresponding values of five reanalysis datasets are a little larger than GPCP, except MERRA. Although the interannual variation in ERA-Interim resembles observation, it has a larger climate mean value for NHSMI, SHSMI, and GSMI (8.42, 7.46, and 8.02 mm day−1) than observations (Table 3).
The normalized time series of the (a) Northern Hemisphere, (b) Southern Hemisphere, and (c) global monsoon precipitation strength in the local summer (MJJAS for Northern Hemisphere and NDJFM for Southern Hemisphere) for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The normalized time series of the (a) Northern Hemisphere, (b) Southern Hemisphere, and (c) global monsoon precipitation strength in the local summer (MJJAS for Northern Hemisphere and NDJFM for Southern Hemisphere) for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The normalized time series of the (a) Northern Hemisphere, (b) Southern Hemisphere, and (c) global monsoon precipitation strength in the local summer (MJJAS for Northern Hemisphere and NDJFM for Southern Hemisphere) for observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The correlation coefficients between observed data and the five reanalysis datasets as well as the long-term trend and climatological mean values about the normalized Northern Hemisphere monsoon index, the Southern Hemisphere monsoon index, and the global monsoon index.
The observational time series indicates an increasing trend in the GSMI, NHSMI, and SHSMI across the entire period, especially after 1988 (Fig. 6). The similar results are demonstrated by Wang et al. (2012). The increasing trends of GSMI after 1988 in these reanalysis datasets are slightly stronger than that in the GPCP, except ERA-Interim (Table 3). The weaker increasing trend of GSMI in ERA-Interim may due to its negative trend of NHSMI. The upward trends of GSMI based on NCEP-2, JRA-25, ERA-40, and MERRA data are statistically significant at 99% confidence level. The downward trend of NHSMI in ERA-Interim is different from observation and other reanalysis datasets, which may induce insignificant trend of GSMI.
In addition to the long-term trend, the observational GSMI also shows a strong interannual variability (Fig. 6c). Similar variations are seen in the NHSMI and SHSMI time series (Figs. 6a,b). The year-by-year variations of GSMI derived from the five reanalysis datasets are consistent with GPCP, as evidenced by the correlation coefficients listed in Table 3 (0.59, 0.58, 0.67, 0.53, and 0.70 for NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA, respectively), which are all statistically significant at 99% confidence level. Among the five datasets, the NHSMI time series derived from ERA-40 and ERA-Interim show the lowest correlation coefficients (0.34 and 0.29) with observations because of their slightly positive and negative trend compare to strong increasing trend in the observations. If we remove the long-term trend, the correlation coefficient with observation will increase to 0.47 and 0.47 in ERA-40 and ERA-Interim, reaching the 99% statistically significant level. The correlation coefficients in the Southern Hemisphere are usually higher than those in the Northern Hemisphere, which may be induced by the unreasonably reproduction for North African monsoon precipitation.
2) EOF pattern of GM precipitation
To reveal the coherent pattern of the change in global monsoon intensity, Fig. 7a shows the coherent spatial pattern of the leading EOF mode of the AR, which is based on GPCP. Since AR is defined by the local summer minus winter precipitation, its values reflect the local monsoon intensity. So the leading EOF mode of AR represents the major spatial variability of monsoon intensity, and the corresponding PC1 represents the major interannual variability of monsoon intensity. Results derived from five reanalysis datasets are shown in Figs. 7b–f. All five reanalysis can reproduce the observed positive anomalies in the Australian monsoon region and the northern part of the Asian region. All five reanalysis datasets, except for ERA-40 and MERRA, show the observed negative anomalies in southern part of Asian monsoon region. Only ERA-Interim can reasonably reproduce the observed anomalous negative pattern in the western part of North American monsoon region. In the North Africa, all five reanalysis datasets fail to reproduce the observed positive anomalies. The EOF pattern in most area is positive in observation but negative in all reanalysis data.
The leading-EOF mode spatial pattern of the normalized annual range anomalies over the global monsoon regions in (a) the observations; and the (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA datasets. Black curves outline the boundaries of the monsoon domain. The time periods are from March 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The leading-EOF mode spatial pattern of the normalized annual range anomalies over the global monsoon regions in (a) the observations; and the (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA datasets. Black curves outline the boundaries of the monsoon domain. The time periods are from March 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The leading-EOF mode spatial pattern of the normalized annual range anomalies over the global monsoon regions in (a) the observations; and the (b) NCEP-2, (c) ERA-40, (d) JRA-25, (e) ERA-Interim, and (f) MERRA datasets. Black curves outline the boundaries of the monsoon domain. The time periods are from March 1979 to April 2011 for GPCP, NCEP-2, JRA-25, ERA-Interim and MERRA; and from May 1979 to April 2001 for ERA-40.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The corresponding observational ARI shows increasing tendency for 1979–2011 (Fig. 8a). All five reanalysis datasets show similar but stronger increasing trends than the observation. ARI of NCEP-2, ERA-Interim, and MERRA show relatively high correlation coefficients (0.61, 0.73, and 0.57) with the observations, which are all statistically significant at 95% confidence level. ERA-Interim shows the highest correlation coefficient (Table 4). After linear detrending, the results of five reanalysis data are obviously improved (Table 4) with nearly identical phase with the observation in interannual variation (Fig. 8b). As is shown in Table 4, the ERA-Interim exhibits the highest skill in reproducing the observed interannual variability of GM precipitation.
(a) The time series of the ARI, the principle component corresponding to the EOF mode in Fig. 7 and (b) the ARI after detrending for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(a) The time series of the ARI, the principle component corresponding to the EOF mode in Fig. 7 and (b) the ARI after detrending for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(a) The time series of the ARI, the principle component corresponding to the EOF mode in Fig. 7 and (b) the ARI after detrending for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The correlation coefficients between observed data and the five reanalysis datasets about the ARI before and after detrending.
3) Long-term trend of GM precipitation
To facilitate the comparison with observational analysis of Wang and Ding (2006), two methods, T2N and MK, are used to test the significance of linear trends (figure not shown). The significance spatial patterns detected by two methods are consistent. Similar to Fig. 3a in Wang and Ding (2006), a strong observational increasing trend in monsoon rain intensity is found in the North African, Australian, and the central part of the Asian monsoon regions. All these trends are statistically significant at the 99% level. A decreasing trend in monsoon rainfall is seen over the South Asian and South American land monsoon regions. The patterns of trends derived from five reanalysis datasets are consistent with the observation over most of GM domain, except for North Africa. In the observations, there is an increasing trend over North Africa. However, in the five reanalysis datasets, there are decreasing trends over North Africa. The feature is also evident in Figs. 7 and 9g.
Normalized time series of area-average annual precipitation range in (a) North African monsoon, (b) Asian monsoon, (c) North American monsoon, (d) Southern African monsoon, (e) Australian monsoon, and (f) South American monsoon for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. Lines are 3-yr filtered. (g) The AR trends reflected by observations and each reanalysis dataset. The trends are calculated from 1979 to 2011.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Normalized time series of area-average annual precipitation range in (a) North African monsoon, (b) Asian monsoon, (c) North American monsoon, (d) Southern African monsoon, (e) Australian monsoon, and (f) South American monsoon for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. Lines are 3-yr filtered. (g) The AR trends reflected by observations and each reanalysis dataset. The trends are calculated from 1979 to 2011.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Normalized time series of area-average annual precipitation range in (a) North African monsoon, (b) Asian monsoon, (c) North American monsoon, (d) Southern African monsoon, (e) Australian monsoon, and (f) South American monsoon for: observations, and reanalysis datasets NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA. Lines are 3-yr filtered. (g) The AR trends reflected by observations and each reanalysis dataset. The trends are calculated from 1979 to 2011.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
In terms of the regional monsoon areas (NAF, ASN, NAM, SAF, AUS, and SAM), the time series of normalized AR in these regions are shown in Figs. 9a–f. These results suggest that the increasing trend over GM region after 1988 is mainly contributed to regional precipitation trends over the Asian, North American, Southern African, and Australian monsoon regions. Five reanalysis datasets do not show consistent trends with observations in the North African, North American, and Australian monsoon regions (Fig. 9g), which are also seen in SST forced AGCM simulation (Zhou et al. 2008a). In North Africa only, all five reanalysis datasets are contrary to the observations (Fig. 9g).
c. The variation of water budget components
To better understand the variability and long-term trend of GM precipitation, we should assess the other terms of water budget. Through an intercomparison of five reanalysis datasets, the measurement can be regarded more reliable by avoiding single dataset dependence.
1) The climatology of water budget components
The climatology of June–August (JJA) and December–February (DJF) precipitation, evaporation, wind convergence term, moisture advection term, and local moisture term are shown in Fig. 10. Since the respective spatial patterns derived from reanalysis data are similar, we use their ensemble mean here. For the Northern Hemisphere, the wet (dry) season means JJA (DJF), whereas, for the Southern Hemisphere, the wet (dry) season means DJF (JJA). The precipitation pattern in wet season over GM domain highlighted by black curve is dominated by wind convergence term. The spatial patterns of evaporation, ΔCWV, and the moisture advection term do not show large differences between wet and dry seasons. In other words, the GM precipitation characteristics are mainly driven by wind convergence term in the moisture budget. The ΔCWV is negative over most of GM regions in both local summer and winter, implying the local column water vapor term in monsoon regions does negative contribution to the monsoon precipitation. Besides, the magnitude of ΔCWV is much smaller than the other terms. The moisture advection term has a negative (positive) contribution to precipitation over the region of 30°S–30N° (in middle and high latitudes): that is, thermodynamic component of water vapor transport has no notable difference between the wet and dry seasons. However, there is notable difference between wet and dry seasons with respect to dynamic component (wind convergence term) in monsoon regions. It has a positive (negative) contribution in wet (dry) season.
The ensemble means of the climatological water budget components in (left) JJA and (right) DJF derived from the five reanalysis datasets (NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA) from May 1979 to April 2001: (a),(f) precipitation, (b),(g) evaporation, (c),(h) wind convergence term, (d),(i) moisture advection term, and (e),(j) local column water vapor. The units are millimeters per day. The scale in (e),(j) covers a much smaller range than the scale for the other water budget components.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The ensemble means of the climatological water budget components in (left) JJA and (right) DJF derived from the five reanalysis datasets (NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA) from May 1979 to April 2001: (a),(f) precipitation, (b),(g) evaporation, (c),(h) wind convergence term, (d),(i) moisture advection term, and (e),(j) local column water vapor. The units are millimeters per day. The scale in (e),(j) covers a much smaller range than the scale for the other water budget components.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The ensemble means of the climatological water budget components in (left) JJA and (right) DJF derived from the five reanalysis datasets (NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA) from May 1979 to April 2001: (a),(f) precipitation, (b),(g) evaporation, (c),(h) wind convergence term, (d),(i) moisture advection term, and (e),(j) local column water vapor. The units are millimeters per day. The scale in (e),(j) covers a much smaller range than the scale for the other water budget components.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The area-averaged precipitation as well as other terms of water budget in wet season over the GM domain is shown in Fig. 11a. The ΔCWV is not shown in the remainder of this study because of its small magnitude. All five reanalysis datasets show that the total monsoon precipitation in wet season is mainly brought by local evaporation and wind convergence term, as these two terms are both positive and almost equal to each other. The average moisture advection term contributes a slightly negative amount to total precipitation in all reanalysis datasets but MERRA.
(a) The area-average precipitation and other water budget terms (mm day−1) over the GM domain in the wet season and (b) the differences between the wet and dry seasons. In the Northern Hemisphere, the wet season is JJA and the dry season is DJF; and vice-versa in the Southern Hemisphere. All values in this figure are averaged from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(a) The area-average precipitation and other water budget terms (mm day−1) over the GM domain in the wet season and (b) the differences between the wet and dry seasons. In the Northern Hemisphere, the wet season is JJA and the dry season is DJF; and vice-versa in the Southern Hemisphere. All values in this figure are averaged from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
(a) The area-average precipitation and other water budget terms (mm day−1) over the GM domain in the wet season and (b) the differences between the wet and dry seasons. In the Northern Hemisphere, the wet season is JJA and the dry season is DJF; and vice-versa in the Southern Hemisphere. All values in this figure are averaged from May 1979 to April 2001.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
Figure 11b shows the GM precipitation differences between wet season and dry season [averaged JJA (DJF) precipitation minus DJF (JJA) precipitation in the Northern (Southern) Hemisphere], along with other water budget terms. Obviously, the wind convergence term dominates the precipitation difference between wet and dry season. The magnitudes of all other terms are much smaller than wind convergence term in the five reanalysis datasets. That is, the contrasting rainy summer and dry winter in GM regions is induced by the annual reversal of wind convergence. Our findings here support the definition of the GM domain based on precipitation, which essentially reflects the seasonal change of prevailing wind. That the change in GM seasonal precipitation reflects the change in wind on the global scale is also consistent with the global wind index outlined in Wang et al. (2013).
2) The long-term trend of water budget components
The long-term trend of precipitation, evaporation, wind convergence and moisture advection terms over all regional monsoon areas (NAF, SAF, ASN, AUS, NAM, and SAM) from 1979 to 2011 are shown in Fig. 12. In the North African and Australian monsoon regions, precipitation and the wind convergence term show an obvious trend in all five reanalysis data, indicating that change of precipitation is dominated by wind convergence. There are also notable trends of the precipitation and wind convergence in NCEP-2 and ERA-40 in the North and South American monsoon regions: that is, compared to the thermodynamic term, the dynamic component of water vapor transport dominates the summer monsoon precipitation. In North (Southern) Africa, the trends of precipitation and wind convergence term in five reanalysis datasets are negative (positive). Similarly, in North (South) America, the trends of precipitation and wind convergence term almost show positive (negative) trends in reanalysis data. The evaporation derived from all five reanalysis datasets generally shows upward trend over monsoon regions, probably resulting from global warming.
The long-term trend [mm day−1 (23 yr)−1] of the major water budget components from 1979–2001 over all regional monsoon areas for the NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA datasets. The red, blue, green, and yellow bars indicate precipitation, evaporation, wind convergence, and the moisture advection term, respectively.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The long-term trend [mm day−1 (23 yr)−1] of the major water budget components from 1979–2001 over all regional monsoon areas for the NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA datasets. The red, blue, green, and yellow bars indicate precipitation, evaporation, wind convergence, and the moisture advection term, respectively.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
The long-term trend [mm day−1 (23 yr)−1] of the major water budget components from 1979–2001 over all regional monsoon areas for the NCEP-2, ERA-40, JRA-25, ERA-Interim, and MERRA datasets. The red, blue, green, and yellow bars indicate precipitation, evaporation, wind convergence, and the moisture advection term, respectively.
Citation: Journal of Climate 27, 3; 10.1175/JCLI-D-13-00215.1
3) The interannual variability of water budget components
The first part of Table 5 shows the correlation coefficients of reanalysis evaporation, wind convergence, and moisture advection term with reanalysis precipitation. It is obvious that wind convergence term has the highest correlation coefficients with precipitation in all reanalysis data: that is, the interannual variability of monsoon precipitation is dominated by the wind convergence term.
The correlation coefficients of each term in water budget with reanalysis precipitation and observed precipitation.
The second part of Table 5 shows the correlation coefficients of reanalysis evaporation, wind convergence, and the moisture advection term with observed precipitation. Correlation coefficients between the wind convergence term and observational precipitation (see Table 5) are similar or higher than the values between reanalysis precipitation and observational precipitation (see Table 3). That is, although wind convergence dominates the interannual variability of precipitation in reanalysis data, it may be not a major cause to the precipitation bias in reanalysis compared to observation. The precipitation bias in reanalysis datasets may be attributed to other reasons, such as evaporation and moisture advection.
4. Conclusions and discussion
a. Conclusions
With the motivation to reveal whether a reasonable prediction of atmospheric circulation would lead to a successful prediction of monsoon precipitation, the climatology, interannual variation, and long-term trend of GM precipitation derived from five sets of reanalysis data are assessed and compared to GPCP data. To better understand the variability and long-term trend of GM precipitation, the major components of water budget are examined, including evaporation, water vapor convergence, and the change in local water vapor storage, based on five reanalysis datasets. The conclusions are as follows:
1) The performance of five reanalysis datasets
Climatology: The multi-reanalysis datasets reasonably reproduce the climatology of GM precipitation, with PCCs higher than 0.85 and RMSEs less than 2 mm day−1. The ERA-Interim is the best while NCEP-2 is the worst among the five reanalysis datasets in revealing climatological spatial pattern. However, ERA-Interim obviously overestimates the moderate rain between 10 and 20 mm day−1.
Long-term trend: The observed summer monsoon precipitation show increasing trends in Northern Hemisphere, Southern Hemisphere, and global area, which can be reasonably reproduced by five reanalysis datasets. Nevertheless, all five datasets fail in reproducing the increasing tendency of North African monsoon precipitation.
Interannual variability: The observed major features in the interannual variability of GM precipitation are reasonably captured by the five reanalysis datasets, among which ERA-Interim is the best.
2) Moisture budget analysis
The contrasting rainy summer and dry winter over GM region is induced by annual reversal of the wind convergence term (or dynamic component) in the water budget. The long-term trend and interannual variability of GM precipitation is dominated by the wind convergence term. The dynamic components of water vapor transport dominate the variation of monsoon rainfall.
b. Discussion
As shown in Table 1, the atmospheric forecast model, assimilation algorithm, vertical and horizontal resolution, and the methods by which satellite data assimilated in five reanalysis data are different. This kind of difference may cause difference in precipitation: 1) The data assimilation method of ERA-Interim is 12-h 4DVAR, which is the most advanced assimilation method. Better performance of ERA-Interim precipitation may result from better assimilation technology. 2) Both vertical and horizontal resolution of ERA-Interim and MERRA (T255 and L60 for ERA-Interim and 1/2° latitude, 2/3° longitude, and L72 for MERRA) are the highest among the five reanalysis datasets. In our results, ERA-Interim and MERRA have the highest ability in reproducing the climatology and interannual variability of monsoon precipitation, indicating higher resolution may lead to better performance. 3) Whether the retrievals products are used (e.g., wind profiles, rain rates) also influences the performance of reanalysis data in reproducing precipitation. Satellite data are incorporated in the MERRA precipitation product, including SSM/I rain rate and TMI rain rate (Rienecker et al. 2011). 4) Finally, data quality control, variational bias correction of satellite radiance data, and other improvements in bias handling of ERA-Interim compared to ERA-40 may also be important for its improvement (Dee et al. 2011).
Acknowledgments
This work was supported by the National Program on Key Basic Research Project (2010CB951904) and the National Natural Science Foundation of China under Grants 41125017 and 41330423. The contribution of Yun Qian in this study was supported by the Office of Science of the U.S. Department of Energy as part of the Earth System Modeling Program. The Pacific Northwest National Laboratory is operated for DOE by Battelle Memorial Institute under Contract DE-AC06-76RLO 1830.
REFERENCES
Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167.
Bosilovich, M. G., cited 2008: NASA’s modern era retrospective-analysis for research and applications: Integrating Earth observations. EarthZine. [Available online at http://www.earthzine.org/2008/09/26/nasas-modern-era-retrospective-analysis/.]
Bosilovich, M. G., J. Chen, F. R. Robertson, and R. F. Adler, 2008: Evaluation of global precipitation in reanalyses. J. Appl. Meteor. Climatol., 47, 2279–2299.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597.
Feng, L., and T. Zhou, 2012: Water vapor transport for summer precipitation over the Tibetan Plateau: Multidata set analysis. J. Geophys. Res.,117, D20114, doi:10.1029/2011JD017012.
Huang, R., Z. Zhang, G. Huang, and B. Ren, 1998: Characteristics of the water vapor transport in East Asian monsoon region and its difference from that in South Asian monsoon region in summer. Chin. J. Atmos. Sci, 22, 469–479.
Huffman, G. J., and Coauthors, 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc.,78, 5–20.
Janowiak, J. E., A. Gruber, C. Kondragunta, R. E. Livezey, and G. J. Huffman, 1998: A comparison of the NCEP–NCAR reanalysis precipitation and the GPCP rain gauge-satellite combined dataset with observational error considerations. J. Climate, 11, 2960–2979.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83, 1631–1644.
Kang, I. S., and Coauthors, 2002: Intercomparison of the climatological variations of Asian summer monsoon precipitation simulated by 10 GCMs. Climate Dyn., 19, 383–395.
Li, H., A. Dai, T. Zhou, and J. Lu, 2010: Responses of East Asian summer monsoon to historical SST and atmospheric forcing during 1950–2000. Climate Dyn., 34, 501–514.
Li, J., R. C. Yu, H. M. Chen, and W. H. Yuan, 2010: Evaluation and analysis of summer rainfall over mainland China in three reanalysis datasets (in Chinese). Meteor. Mon.,36, 1–9.
Li, L., W. Li, and A. P. Barros, 2013: Atmospheric moisture budget and its regulation of the summer precipitation variability over the southeastern United States. Climate Dyn., 41, 613–631.
Liu, J., B. Wang, Q. Ding, X. Kuang, W. Soon, and E. Zorita, 2009: Centennial variations of the global monsoon precipitation in the last millennium: Results from ECHO-G model. J. Climate, 22, 2356–2371.
Onogi, K., and Coauthors, 2007: The JRA-25 reanalysis. J. Meteor. Soc. Japan, 85, 369–432.
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648.
Schubert, S., and Coauthors, 2008: Assimilating Earth system observations at NASA: MERRA and beyond. Extended Abstracts, Third WCRP Int. Conf. on Reanalysis, Tokyo, Japan, WCRP, V1-103. [Available online at http://wcrp.ipsl.jussieu.fr/Workshops/Reanalysis2008/Documents/V1-104_ea.pdf.]
Seager, R., N. Naik, and G. A. Vecchi, 2010: Thermodynamic and dynamic mechanisms for large-scale changes in the hydrological cycle in response to global warming. J. Climate, 23, 4651–4668.
Serreze, M. C., and C. M. Hurst, 2000: Representation of mean Arctic precipitation from NCEP–NCAR and ERA reanalyses. J. Climate, 13, 182–201.
Simmonds, I., D. Bi, and P. Hope, 1999: Atmospheric water vapor flux and its association with rainfall over china in summer. J. Climate, 12, 1353–1367.
Sneyers, R., 1990: On the statistical analysis of series of observations. WMO Tech. Note 415, 192 pp.
Sperber, K. R., H. Annamalai, I.-S. Kang, A. Kitoh, A. Moise, A. Turner, B. Wang, and T. Zhou, 2012: The Asian summer monsoon: An intercomparison of CMIP5 vs. CMIP3 simulations of the late 20th century. Climate Dyn., 41, 2711–2744.
Trenberth, K. E., and C. J. Guillemot, 1995: Evaluation of the global atmospheric moisture budget as seen from analyses. J. Climate, 8, 2255–2272.
Trenberth, K. E., D. P. Stepaniak, and J. M. Caron, 2000: The global monsoon as seen through the divergent atmospheric circulation. J. Climate, 13, 3969–3993.
Turner, A. G., and H. Annamalai, 2012: Climate change and the South Asian summer monsoon. Nat. Climate Change, 2, 587–595.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012.
Wang, B., and Q. Ding, 2006: Changes in global monsoon precipitation over the past 56 years. Geophys. Res. Lett., 33, L06711, doi:10.1029/2005GL025347.
Wang, B., and Q. Ding, 2008: Global monsoon: Dominant mode of annual variation in the tropics. Dyn. Atmos. Oceans, 44, 165–183.
Wang, B., J. Liu, H.-J. Kim, P. J. Webster, and S.-Y. Yim, 2012: Recent change of the global monsoon precipitation (1979–2008). Climate Dyn., 39, 1123–1135.
Wang, B., J. Liu, H.-J. Kim, P. J. Webster, S.-Y. Yim, and B. Xiang, 2013: Northern Hemisphere summer monsoon intensified by mega-El Niño/Southern Oscillation and Atlantic multidecadal oscillation. Proc. Natl. Acad. Sci. USA, 110, 5347–5352.
Webster, P. J., 1987: The elementary monsoon. Monsoons, J. S. Fein and P. L. Stephens, Eds., Wiley, 3–32.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 2539–2558.
Xie, P., J. E. Janowiak, P. A. Arkin, R. Adler, A. Gruber, R. Ferraro, G. J. Huffman, and S. Curtis, 2003: GPCP pentad precipitation analyses: An experimental dataset based on gauge observations and satellite estimates. J. Climate, 16, 2197–2214.
Yin, X., A. Gruber, and P. Arkin, 2004: Comparison of the GPCP and CMAP merged gauge-satellite monthly precipitation products for the period 1979–2001. J. Hydrometeor., 5, 1207–1222.
Zhang, L., and T. Zhou, 2011: An assessment of monsoon precipitation changes during 1901–2001. Climate Dyn., 37, 279–296.
Zhou, T.-J., and R.-C. Yu, 2005: Atmospheric water vapor transport associated with typical anomalous summer rainfall patterns in China. J. Geophys. Res., 110, D08104, doi:10.1029/2004JD005413.
Zhou, T.-J., X. Zhang, and S. Wang, 1999: The air-sea freshwater exchange derived from NCEP/NCAR reanalysis data. Acta Meteor. Sin., 57, 264–282.
Zhou, T.-J., R. Yu, H. Li, and B. Wang, 2008a: Ocean forcing to changes in global monsoon precipitation over the recent half-century. J. Climate, 21, 3833–3852.
Zhou, T.-J., L. Zhang, and H. Li, 2008b: Changes in global land monsoon area and total rainfall accumulation over the last half century. Geophys. Res. Lett., 35, L16707, doi:10.1029/2008GL034881.
Zhou, T.-J., B. Wu, and B. Wang, 2009a: How well do atmospheric general circulation models capture the leading modes of the interannual variability of the Asian–Australian monsoon? J. Climate, 22, 1159–1173.
Zhou, T.-J., and Coauthors, 2009b: The CLIVAR C20C project: Which components of the Asian–Australian monsoon circulation variations are forced and reproducible? Climate Dyn., 33, 1051–1068.
