SEARCH

SEARCH BY CITATION

Keywords:

  • Asia monsoon;
  • atmospheric general circulation model;
  • precipitation;
  • water isotopes

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] The stable isotopic composition of water has been used as a paleoproxy to reconstruct past climates over the Asian monsoon region, but the main controls on the variability of isotopes of water in precipitation have not been characterized quantitatively in this region. Therefore, we used an atmospheric general circulation model incorporating stable water isotope physics to quantitatively estimate the relative contributions to isotope variability in precipitation falling in the Asian monsoon region. As in previous research, we identified two primary factors controlling the interannual variability of δ18Oprecip (defined as (Rsample/RVSMOW − 1) × 1000, where RVSMOW is the 18O ratio in Vienna Standard Mean Ocean Water) and its correlation with El Niño–Southern Oscillation (ENSO) events: the amount of precipitation at the observation site, and distillation during transport from source regions. Two sensitivity experiments revealed that distillation during transport from source regions was the dominant controlling factor; at Bangkok, Bombay, and Hong Kong, the amount of local precipitation contributed 27%, 33%, and 25% while distillation processes contributed 70%, 60%, and 70%, respectively. Similarly, distillation processes accounted for 80%, 82%, and 83% of observed differences in δ18Oprecip between El Niño and La Niña years at these three cities, respectively. Therefore, interannual variability of δ18Oprecipat the three stations primarily reflects distillation during transport from source regions, and it is also governed by the large-scale tropical variability (ENSO).

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] In natural systems, the proportions of HDO and H218O in precipitation vary in both space and time. In his pioneering paper, Dansgaard [1964] demonstrated that the stable isotopic composition of water in precipitation is strongly and positively correlated with surface air temperature at high latitudes, a phenomenon known as the “temperature effect.” This empirical relationship has been applied in analysis of isotopes of water preserved in ice cores of Greenland and Antarctica, for the reconstruction of paleoclimates [Cuffey, 2004; Steffensen, 2008; Sime et al., 2009]. Conversely, Dansgaard [1964] demonstrated that at low latitude sites, particularly on tropical islands, the isotopic composition of water in precipitation is negatively correlated with the amount of precipitation, a relationship typically referred to as the “amount effect.”

[3] In addition, the proportion of H218O in precipitation at low latitudes is closely related to large-scale atmospheric variability, especially the El Niño–Southern Oscillation (ENSO), the South American monsoon, and the Asian monsoon [Cole et al., 1993; Araguas-Araguas et al., 1998; Thompson et al., 2000; Vuille et al., 2003a, 2003b, 2005; Vuille and Werner, 2005; Bradley et al., 2003; Aggarwal et al., 2004; Brown et al., 2009; Ichiyanagi and Yamanaka, 2005; He et al., 2006]. These relationships have been used to reconstruct past climates from isotopes of water preserved in ice cores from low latitude sites such as the Tibetan plateau [Thompson et al., 1989; Yao et al., 1996; Thompson et al., 1997; Duan et al., 2004; Davis et al., 2005], the Himalayas [Thompson et al., 2000], and the Andes mountains [Thompson et al., 1985, 1998; Hoffmann, 2003; Hoffmann et al., 2003], and in speleothems from low-latitude caves in China [Y. J. Wang et al., 2001; Yuan et al., 2004; Wang et al., 2005; Dykoski et al., 2005; Overpeck and Cole, 2008; Wang et al., 2008; Cheng et al., 2009].

[4] However, the use of isotopes of water as a paleoproxy to reconstruct past climates is insufficiently validated because such reconstructions have been based on empirical relationships between isotopes of water and climatic factors, as described above. For example, Zhang et al. [2008] reconstructed the history of the Asian monsoon over the past 1810 years by using changes in the proportion of H218O in speleothems from Wanxiang Cave, China. Based on this reconstruction, he discussed the relationship between Asian monsoon variability and the decline of Chinese dynasties during this period. However, their reconstruction of the Asian monsoon was based on a straightforward application of the empirical relationship between the H218O composition and amount of precipitation, and the index δ18OH2O (amount of 18O in water expressed as the per mil departure from standard mean ocean water (SMOW)) in the Wanxiang Cave speleothems may not be spatially representative over a wide enough area to serve as a basis for the reconstruction of Asian monsoon history. Similar reconstructions of low latitude paleoclimate variables have also not been sufficiently validated in other studies. To validate reconstructions of past climate over the entire Asian monsoon region, we need to understand the main factors causing water isotope variability, in a quantitative way.

[5] Many parameters that correlate with changes in the isotopic composition of water in precipitation have been identified for the low latitudes. Examples include the local precipitation amount [Dansgaard, 1964], the moisture source of the precipitation [Cole et al., 1999; Bhattacharya et al., 2003], distillation during transport [Hoffmann and Heimann, 1997; Araguas-Araguas et al., 1998; Yoshimura et al., 2003; Vuille and Werner, 2005; Vuille et al., 2005], the recycling of water by land surface exchange processes [Gat and Matsui, 1991], and re-evaporation of precipitation in unsaturated air [Worden et al., 2007]. However, the relative importance of these parameters has not been quantitatively investigated yet. Even though relationships between isotopic composition of precipitation water and large-scale patterns of atmospheric variability, such as ENSO or monsoons, are frequently acknowledged and even applied to reconstruct past atmospheric variability, the actual causes of these relationships remain unclear. As a result of insufficient understanding of the causes of variability inδ18O (defined as (Rsample/RVSMOW − 1) × 1000, where RVSMOW is the 18O ratio in Vienna SMOW) in precipitation (hereafter, δ18Oprecip) in low latitudes, water isotopes are inadequately validated for use as paleoproxies in reconstructing paleoclimates.

[6] Some studies have explained the apparent relationship between δ18Oprecipat low latitudes and large-scale atmospheric variability by invoking the amount effect. For instance,Vuille et al. [2005] reported that variations in the precipitation at observation sites explain, to first order, the relationship between δ18Oprecip and the Asian monsoon and Indian monsoon circulation. Similarly, Ichiyanagi and Yamanaka [2005] concluded that the amount of precipitation was the main influence on the observed relationship between interannual variations of δ18O in precipitation at Bangkok and the ENSO index in May and October. In contrast, Vuille and Werner [2005] found that at many locations in South America under the influence of its monsoon, the moisture transport history and the amount of rainfall along the upstream pathway may be more important controls on δ18Oprecip than the amount of precipitation at the observation sites. At low latitude sites, these two controls mentioned above have also been linked to the relationship between δ18Oprecipand large-scale interannual variability of the atmosphere.

[7] The objective of this study was to quantitatively estimate the relative contributions of the main controls on interannual variability of the isotopic composition of precipitation in the Asian monsoon region, and determine which of them explain the relationship between the isotopic composition of precipitation and large-scale atmospheric variability. To do this, we used an atmospheric general circulation model (AGCM) capable of representing isotopes, and observations ofδ18Oprecip, surface air temperature, and precipitation from the Global Network of Isotopes in Precipitation (GNIP) database of the International Atomic Energy Agency (IAEA) (http://www-naweb.iaea.org/napc/ih/IHS_resources_gnip.html), which contains records of isotopic ratios in precipitation from over 50 years of measurements. We first validated the model results and identified the main factors controlling interannual variability in δ18Oprecip and the temporal relationship between δ18Oprecip and ENSO events. Then, we evaluated the proportional contribution of each factor (as a ratio) to the interannual changes in δ18Oprecip, to select the dominant contribution to its interannual variability in the Asian monsoon. Finally, we investigated which factor was most responsible for the apparent relationship between δ18Oprecipand large-scale atmospheric variability in this region.

2. Methods and Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

2.1. Model and Observations

[8] Water isotope tracers have been incorporated into several AGCMs [Joussaume et al., 1984; Jouzel et al., 1987; Hoffmann et al., 1998; Noone and Simmonds, 2002; Mathieu et al., 2002; Schmidt et al., 2005; Lee et al., 2007; Yoshimura et al., 2008; Risi et al., 2010; Tindall et al., 2010; Kurita et al., 2011]. For the present study, we similarly incorporated stable isotope tracers into CCSR/NIES/FRCGC AGCM5.7b [Hasumi and Emori, 2004] by using the similar methods of Kurita et al. [2011] to model the physical processes of the water isotopes. The modeling of isotopic ratios in precipitation at cold temperatures must address the switch from vapor–liquid transitions to vapor–ice transitions at the onset of snow formation. In our model, this was performed for mixed clouds, which can contain both liquid droplets and ice crystals, by evaluating the supersaturation following the method of Ciais and Jouzel [1994], but we did not consider the influence of drop size on water isotopes in precipitation because CCSR/NIES/FRCGC AGCM5.7b does not calculate drop size.

[9] The model features T42 spectral truncation horizontally (a resolution of approximately 2.8° on an equivalent grid) and 20 vertical layers. In this study, it was driven by monthly global sea surface temperature (SST) fields provided by the Atmospheric Model Intercomparison Project 2 (AMIP2) data set for the period 1981–2004, and we focus on the results during that period here. All model outputs are monthly averages, and we emphasize precipitation amount, surface air temperature, ratios of stable isotope in precipitation, and SST. We used the AMIP2 data set to calculate the modeled ENSO index. We used the observed monthly averaged SST anomaly in the Niño 3.4 region (170°–120°W, 5°S–5°N), obtained from the Climate Prediction Center (CPC), as the observed ENSO index. We obtained the spatial patterns of observed precipitation from the CPC Merged Analysis of Precipitation (CMAP) [Xie and Arkin, 1997] database. The GNIP database contains not only precipitation isotopes but also surface air temperature and precipitation amount, and we obtained the observed precipitation isotopes, surface temperature, and precipitation at various station sites from that database.

2.2. Estimating Transport Pathways From Source Regions With Backward Trajectories

[10] We calculated backward trajectories from Bangkok, Bombay and Hong Kong to estimate the precipitation along the transport pathways from source regions (see section 5). We calculated 10-day backward trajectories by using data taken at 6-h intervals from the model. Because most water vapor in the atmosphere exists at low levels, we performed the calculations for three isentropic surfaces, the 303, 306, and 309 K surfaces, which in summer at Bangkok correspond to barometric surfaces of approximately 900, 850, and 800 hPa.

[11] In this study, we defined the precipitation along the transport pathway from a source region to Bangkok (Ptotal) as follows:

  • display math

where P305K,t, P310K,t, and P315K,t denote the precipitation at the grid points where data on the respective isentropic surfaces were archived at time step t (every 6 h for 10 days) in the calculation of the backward trajectories. Thus, Ptotal is the average of the total precipitation along the backward trajectories on the three isentropic surfaces.

2.3. Sensitivity Experiments

[12] We carried out two sensitivity experiments with CCSR/NIES/FRCGC AGCM5.7b to estimate the relative contributions to interannual variability in δ18Oprecip of two effects: the local amount effect, and the distillation effect in transport from the source regions, at Bangkok, Bombay, and Hong Kong (see section 5). In the first sensitivity experiment, we estimated the contribution of the local amount effect, and in the second, we estimated the combined contributions of both effects.

[13] Dansgaard [1964] suggested that the amount effect can be explained by the isotopic fractionation that occurs during three processes. In the first, such fractionation occurs during condensation of water vapor as isotopically heavy water is preferentially removed. This leaves the isotopic composition of the remaining vapor increasingly lighter. As a result, water isotopes in precipitation become increasingly lighter during episodes of heavy precipitation.

[14] Second, isotopic equilibration between raindrops and isotopically enriched water vapor below the cloud base is more complete when the raindrops are small. In general, water vapor near the surface has a heavier isotopic composition because of the influence of surface evaporation and little condensation. Thus, as the proportion of small raindrops becomes higher, the isotopic composition of the precipitation becomes heavier [Lee and Fung, 2008].

[15] The third important process is evaporation from precipitation below the cloud base. Low humidity in the atmosphere results in isotopically heavy precipitation because of the preferential removal of light water isotopes by evaporation from rain. Because the model used here does not take into account the influence of drop size on water isotopes in precipitation, the amount effect reproduced by the model is due to the first and third processes, which are incorporated into the model.

[16] We set up the first sensitivity experiment to estimate the relative influences of these two fractionation processes assumed to be responsible for the amount effect. We removed the influence of the local amount effect on interannual variability of δ18Oprecip at Bangkok, Bombay, and Hong Kong by assuming the isotopic fractionation coefficients in the two processes to be 1.0 in the grid that contained the station and the surrounding eight grids.

[17] Distillation during transport from the source regions also results in isotopic fractionation by the same processes along the transport pathways. Thus, in the second sensitivity experiment, we removed the influences of both the local amount effect and distillation during transport by turning off the two processes not only in the region around the station, as in the first sensitivity experiment, but also along the transport pathways estimated from the backward trajectories. In this study, we defined the transport pathways from the source regions as those grids through which the air parcels passed on the three isentropic surfaces during a minimum of 50 time steps in the entire 23-year period.

3. Model Evaluation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[18] We evaluated the model's ability to reproduce δ18Oprecip and the relationships between δ18Oprecip and the precipitation, surface air temperature, and the ENSO index. In section 3.1 we evaluate the simulated spatial distribution of annual mean δ18Oprecip. In section 3.2, we evaluate the simulated monthly variation of δ18Oprecip at Bangkok, New Delhi, Hong Kong, and Bombay. We selected the first three stations because they are among the few in the Asian monsoon region for which sufficiently long and continuous isotope records are available [Araguas-Araguas et al., 1998], and there is also a relatively long record of δ18Oprecip variation at Bombay. In section 3.3, we evaluate the modeled interannual variability of δ18Oprecip at the four stations, and in section 3.4, we examined the simulated relationships of δ18Oprecip with surface temperature, precipitation, and the ENSO index in the Asia monsoon region by comparing the records at the four stations.

[19] Because the Asian monsoon region is very large, these four stations may not represent the entire region. However, the four stations are strongly influenced by the two major convective regimes of the region: a regime over the Bay of Bengal, Indian Ocean, and the Arabian Sea, and another over the South China Sea and the Philippine Sea. These two convective heat sources drive the Indian and southeast monsoons [B. Wang et al., 2001], and the interannual variability of the East Asian monsoon is also strongly influenced by the interannual variability in the convection over the Philippine Sea [Nitta, 1987; B. Wang et al., 2001]. Therefore, these four stations reflect most of the Asian monsoon region rather well.

3.1. Simulated Annual Mean Over the Asian Monsoon Region

[20] We compared the annual mean δ18Oprecip as simulated by our model with observations made at IAEA stations, based on at least three years of recorded data (Figure 1). In low latitudes of the Asian monsoon region, the observed δ18Oprecip values are higher over the regions between 70°E and 100°E and between 170°E and 160°W than over the region between 100°E and 170°E. Aggarwal et al. [2004]indicated that this pattern is related to the three different types of monsoon circulation: the continental monsoon trough (CMT), the ocean monsoon trough, and the trade-wind trough (TWT), respectively [Chan and Evans, 2002]. Over the CMT region (70°E–100°E), moisture is evaporated from the Indian Ocean and transported eastward by strong low-level (850 hPa) westerly flows, whereas in the TWT region (170°E–160°W) moisture is derived from the Pacific Ocean and transported westward by trade winds.

image

Figure 1. Annual mean δ18Oprecip (‰) as simulated by CCSR/NIES/FRCGC AGCM5.7b for the period 1982–2004 (contours and mapped colors). Numbers and colors within circles are values observed at IAEA stations for the period 1961–2002.

Download figure to PowerPoint

[21] In the western part of the CMT region, δ18Oprecip is higher than in the eastern part because heavy water isotopes are removed from water vapor by distillation during transport from the source region, and it is higher in the eastern than the western TWT region for the same reason. At middle latitudes, precipitation becomes gradually more depleted in H218O with increasing latitude (“temperature effect”); precipitation is more depleted in H218O over inland areas than over coastal areas at the same latitude (“continental effect”); and precipitation is more depleted in H218O over mountainous regions such as the Tibetan plateau than over lowlands (“altitude effect”). Our model simulation reproduced these characteristics over the Asian monsoon region, although it may have overestimated the amount of depletion in the middle latitudes, compared with observations.

3.2. Simulated Monthly Variations Compared With Observed Local Precipitation at Bangkok, Bombay, New Delhi, and Hong Kong

[22] The monthly variations in the simulated amount of precipitation in the CCSR/NIES/FRCGC AGCM5.7b model did not reproduce the observed precipitation at Bangkok, Bombay, New Delhi, and Hong Kong very well, particularly in the summer months (Figure 2). For example, at Bangkok the model tended to overestimate the amount of precipitation; indeed, in July, the predicted amount was double the observed amount. Conversely, the model underestimated the amount of precipitation at Hong Kong.

image

Figure 2. Monthly variations in δ18Oprecip (circles) and precipitation (bars) according to the CCSR/NIES/FRCGC AGCM5.7b AGCM simulation (black bars and circles) and GNIP observations (gray bars and circles) at (a) Bangkok, (b) Bombay, (c) Hong Kong, and (d) New Delhi.

Download figure to PowerPoint

[23] Although the precipitation at the four observation sites was not reliably reproduced, the simulated monthly variation of δ18Oprecip reproduced the observed variation relatively well (Figure 2). Observations show that δ18Oprecip gradually decreased as the Asian monsoon evolved, and reached a minimum only after the peak precipitation had been reached during the monsoon; the model approximately captured both this pattern and the general seasonal variations in δ18Oprecip.

3.3. Simulated Interannual Variations at Bangkok, Bombay, New Delhi, and Hong Kong

[24] We next compared the modeled and observed values of δ18Oprecip, as well as their variances, during the four summer months (JJAS) at Bangkok, Hong Kong, and New Delhi (Table 1). At Bangkok, the correlation between modeled and observed δ18Oprecip was statistically significant, but the correlations at Hong Kong and New Delhi were not significant. The modeled and observed variances were similar at both Bangkok and Hong Kong (Table 1).

Table 1. Correlations (JJAS) Between Modeled and Observed δ18Oprecip and Variances of δ18Oprecip in the Model and Observations (After Removing Seasonality)
 BangkokBombayHong KongNew Delhi
  • a

    Statistically significant findings at the α = 0.05 level.

Correlation (δ18Oprecip)0.65a-0.330.48
Variances of model (δ18Oprecip)1.61.30.81.3
Variances of observations (δ18Oprecip)1.20.41.22.8

3.4. Simulated Relationships With Surface Temperature, Precipitation, and ENSO Index

[25] We examined the correlations of δ18Oprecip with air temperature and precipitation at three observation sites, and with the ENSO index, in both the model simulation and the observations (Table 2). Although the model was not skillful in reproducing the interannual variability in δ18Oprecip, precipitation, or surface temperature, the correlation coefficients are relatively similar between the model simulation and the observations. This result suggests that the model can capably reproduce the interannual relationships between δ18Oprecip and precipitation, surface temperature, and ENSO.

Table 2. Correlations (in JJAS) of δ18Oprecip With Precipitation and Temperature at Bangkok, Bombay, Hong Kong, and New Delhi (Seasonality Removed), and With the ENSO Index, in the Model and Observations
 BangkokBombayHong KongNew Delhi
  • a

    Statistically significant correlations at the α = 0.05 level.

Precipitation amount (model)−0.68a−0.350.12−0.57a
Precipitation amount (observation)−0.44a−0.57−0.07−0.48
Surface air temperature (model)0.23−0.380.090.53a
Surface air temperature (observation)0.010.030.250.09
ENSO index (model)0.83a0.85a0.52a0.78a
ENSO index (observation)0.53a0.65a0.58a0.36

4. Relationship With Climate Over the Asian Monsoon Region

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[26] Stable water isotopes have often been interpreted as reflecting local variations of surface temperature or precipitation. In addition, many recent studies have suggested that large-scale changes in atmospheric circulation induced by ENSO or monsoons are also linked toδ18Oprecip variations over the Asian monsoon region [Vuille et al., 2005; He et al., 2006; Ichiyanagi and Yamanaka, 2005]. As we showed in section 3, the model can capture the relationships in time between δ18Oprecip and local climatic factors, and the link between δ18Oprecipand ENSO events, associated with large-scale changes in atmospheric circulation in this region. Thus, we further investigated these temporal relationships to better understand the factors controlling changes inδ18Oprecip.

4.1. Correlation With Surface Temperature and Precipitation

[27] We compared the temporal correlations of δ18Oprecip with the modeled and observed precipitation (Figure 3) and temperature (Figure 4) during June through September (JJAS). We examined the correlations between δ18Oprecip and the precipitation by using GNIP data from stations for which there were at least 7 years of δ18Oprecip records during JJAS between 1961 and 2002. The temporal correlations between δ18Oprecip and the modeled local precipitation were statistically significant in many regions, including the southern Indian subcontinent, Indochina, the maritime continent, and regions around the Philippines (Figure 3a). The magnitudes of the correlation coefficients between δ18Oprecip and precipitation were high in the GNIP observations at Bombay and Singapore (Figure 3b), but the number of observations was insufficient to establish statistical significance. The correlation coefficient between δ18Oprecip and the precipitation at Jakarta was not statistically significant because at that site the precipitation is not high in the boreal summer.

image

Figure 3. Coefficients (×100) for the correlation of δ18Oprecip with the precipitation in JJAS in (a) the model and (b) GNIP observations. Seasonality was removed in the correlations. In the model results, the correlations are statistically significant at the α = 0.05 level when the absolute value of the correlation coefficient is >0.42. Crosses within the circles at the observation sites indicate statistically significant correlations.

Download figure to PowerPoint

image

Figure 4. Correlation coefficients (×100) for the correlation of δ18Oprecip with surface temperature in JJAS in (a) the model and (b) GNIP observations. In the model results, the correlation is statistically significant at the α = 0.05 level when the absolute value of the correlation coefficient is >0.42. Crosses within the circles designating the observation sites denote statistically significant correlations. Seasonality was removed in the correlations.

Download figure to PowerPoint

[28] We found statistically significant correlations in the GNIP observations between δ18Oprecip and observed precipitation at Bangkok and New Delhi, as well as over New Guinea and Australia, even though the last two places are not in the Asian monsoon region. At Bombay and Singapore, the correlation of δ18Oprecip with precipitation was not statistically significant because of the low number of direct measurements. In both the model results and the observations, δ18Oprecip was not significantly correlated with the precipitation during this season over most of China.

[29] In contrast, correlations between local temperature and δ18Oprecip over most land areas of the Asian monsoon region were not significant (Figure 4). In the GNIP observations, the correlation of δ18Oprecip with temperature was statistically significant only at New Delhi, where δ18Oprecip was negatively correlated with the precipitation, and moreover, where temperature and precipitation were also significantly and negatively correlated among themselves (data not shown). Thus, at New Delhi, the correlation between δ18Oprecip and temperature might reflect that between δ18Oprecip and precipitation.

4.2. Correlation With ENSO

[30] In the model, over broad areas of the Indian subcontinent, the Indochina peninsula, the maritime continent, the Philippines, south China, and the equatorial Indian Ocean, δ18Oprecip in JJAS showed a significant positive correlation with ENSO (i.e., the Niño3.4 SST anomaly; Figure 5a). This result corresponds well to the significant δ18Oprecip–ENSO correlations in the GNIP observations in the western Pacific region (at Bangkok, Bombay, Singapore, and Hong Kong; Figure 5b). In contrast, over the central equatorial Pacific, δ18Oprecip showed a significant negative correlation with ENSO in the model, but this relationship cannot be properly validated by using observations because observational data are lacking for this region.

image

Figure 5. Correlation coefficients (×100) for the correlation of δ18Oprecip with the Niño3.4 SST anomaly in JJAS in (a) the model and (b) GNIP observations. In the model results, the correlations are statistically significant at the α = 0.05 level when the absolute value of the correlation coefficient is >0.42. Crosses within the circles designating the observation sites denote statistically significant correlations. Seasonality was removed in the correlations.

Download figure to PowerPoint

[31] The correlation of the precipitation at each location with ENSO variation in the model was significantly negative during JJAS over broad areas around the Indian subcontinent and the eastern Indian Ocean and Indonesia (Figure 6a), which can be explained by the suppression of large-scale convection associated with a shift of the Walker circulation [Webster et al., 1998; Kumar et al., 1999; Goswami and Xavier, 2005].

image

Figure 6. Correlation coefficients (×100) of the correlation of the Niño3.4 SST anomaly with the precipitation in JJAS in each grid in (a) the model and (b) CMAP observations. Correlations are statistically significant at the α = 0.05 level if the absolute value of the correlation coefficient is >0.42 for the model and >0.44 for CMAP. Seasonality was removed in the correlations.

Download figure to PowerPoint

[32] During an El Niño event, the tropical convection normally located in the western Pacific shifts to the central and eastern Pacific. As a result, an anomalous subsidence extends from the western Pacific to the Indian subcontinent. This subsidence due to the shift of the tropical convection suppresses convection and precipitation over the western Pacific and Indian subcontinent [Kumar et al., 1999].

[33] Over the western Pacific and Indian subcontinent, the amount effect caused by precipitation anomalies induced by ENSO may cause a positive correlation between δ18Oprecip and ENSO. However, over southern China and northern India, precipitation is not necessarily correlated with ENSO, whereas δ18Oprecipis. Because a statistically significant correlation between precipitation and ENSO exists over the southern Indian subcontinent, the maritime continent, and Indochina thought to be upstream of southern China and northern India, changes in ENSO-induced upstream precipitation may be responsible for the positive correlation betweenδ18Oprecip and ENSO over the downstream regions. Thus, during El Niño events, decreased precipitation may produce isotopically enriched water vapor over the upstream regions, resulting in the enrichment of precipitation with heavier isotopes over the downstream regions. Conversely, during La Niña events, isotopically depleted water vapor resulting from increased precipitation over the upstream regions may lead to depletion of heavier isotopes in precipitation over the downstream regions.

[34] In the observations as well (Figure 6b), variability in the precipitation over the upstream regions is an important controlling factor between δ18Oprecip and ENSO events. At Bangkok, Bombay, and Hong Kong, the observed precipitation did not correlate with ENSO during JJAS, although δ18Oprecip did (Figure 5b). Moreover, although the modeled precipitation around Indochina and the southern Indian subcontinent showed a significant correlation with ENSO events (Figure 6a), the observations in these regions did not (Figure 6b). Therefore, over both the upstream and downstream regions, the model might overestimate the relationship between δ18Oprecip and ENSO.

5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[35] In this section, we examine the quantitative contributions to the interannual variability of water isotopes in precipitation over this region, and attempt to establish which factor controls the relationship between δ18Oprecip and ENSO at Bangkok, Bombay, and Hong Kong.

[36] In the Asian monsoon region, only three stations (Bangkok, New Delhi, Hong Kong) have more than 10 years of continuous isotopic records during 1982–2002. The skill of the model and its variances on an interannual timescale were superior at Bangkok, and moreover, Bangkok is located between the two major convective heat sources that drive the Asian monsoon – the convection over the Bay of Bengal, India and Arabian Sea, and that over the South China Sea and the Philippine Sea [B. Wang et al., 2001]. Therefore, to understand the variability of δ18Oprecip over the entire Asian monsoon region it is very important to first understand its variability at Bangkok.

[37] At Bangkok, as most of the annual precipitation occurs between May and October (Figure 2a), we investigated interannual correlations during this part of the year (MJJASO), examining the relationships between the precipitation along the backward trajectories of air masses and δ18Oprecip at Bangkok. Then, we performed two sensitivity experiments to estimate the quantitative contributions of the main factors to the interannual variability of water isotopes in precipitation over the Asian monsoon region, and which factors were responsible for the relationship between δ18Oprecip and ENSO at Bangkok. We conducted the same two sensitivity experiments at Bombay and Hong Kong because the model indicated a strong correlation between δ18Oprecip and ENSO at those stations as well. We do not discuss the New Delhi data here because the results of backward trajectories were not reliable at New Delhi (data not shown).

5.1. The Relationship With Precipitation Amount During the Transport of Moisture From Source Regions to Bangkok

[38] We examined results for the modeled backward trajectories for the years with the greatest δ18Oprecip enrichment (year 1982; Figures 7b–7d) and depletion (year 1988; Figures 8b–8d) in Bangkok, because we expected the characteristics in these years to be most distinctive. First, we examined the precipitation anomaly between the year of greatest enrichment (1982) and the long-term mean from 1982 to 2004 (Figure 7a).

image

Figure 7. (a) Precipitation anomalies between the precipitations in the year of greatest δ18O enrichment (1982) and the long-term mean in MJJASO during 1982–2004. Ten-day backward trajectories for three isentropic surfaces: (b) 303 K, (c) 306 K, and (d) 309 K in the year of greatest enrichment (1982). Trajectories are colored differently according to month.

Download figure to PowerPoint

image

Figure 8. (a) Precipitation anomalies in the year of greatest δ18O depletion (1988) from the long-term mean in MJJASO during 1982–2004. Ten-day backward trajectories for three isentropic surfaces: (b) 303 K, (c) 306 K, and (d) 309 K in 1988. Trajectories are colored differently according to month.

Download figure to PowerPoint

[39] For the lowest isentropic surface, most air parcels in 1982 arrived at Bangkok from the southwest, from the equatorial Indian Ocean (Figure 7b). In contrast, for the other two isentropic surfaces, most air parcels originated in the western Indian Ocean and were transported by the Somali jet (Figures 7c and 7d). The precipitation along some parts of the inferred transport pathways was less than the long-term precipitation mean.

[40] In the year of greatest depletion (1988; Figure 8), the trajectories calculated for the three isentropic surfaces are very similar to those in 1982, the year of greatest enrichment. In 1988, over some parts of the trajectories, the precipitation was higher than its long-term mean, although for other parts the two did not differ significantly. The correlation ofδ18Oprecip at Bangkok with precipitation along the transport pathways was statistically significant (r = −75), and moreover, it was stronger than the local correlation between the precipitation and δ18Oprecip at Bangkok (Table 2). In addition, the correlation of the precipitation at Bangkok with precipitation along the transport routes was also statistically significant (r = 0.49). We obtained similar results at Bombay and Hong Kong (not shown). We also used backward trajectories to calculate the transport pathways from the source regions to New Delhi, but the results were not reliable. Therefore, we conducted sensitivity experiments only for Bangkok, Bombay, and Hong Kong.

5.2. Sensitivity Experiments

[41] In the first sensitivity experiment, the average values of δ18Oprecip at Bangkok, Bombay, and Hong Kong were equal to the average values of δ18O values in water vapor in each respective region and significantly lower than those obtained in the standard experiment (Figure 9). These results imply that in the first sensitivity experiment isotopic fractionation was properly turned off as intended. The variances of δ18Oprecip in the first sensitivity experiment at the three sites were respectively 27%, 33%, and 25% smaller than the variances in the standard experiment (Table 3). These results suggest that the amount effect did not dominantly control the interannual variability of δ18Oprecip at Bangkok, Bombay and Hong Kong.

image

Figure 9. Interannual variability of δ18Oprecip (a) at Bangkok during MJJASO and at (b) Bombay and (c) Hong Kong during JJAS in the standard experiment (white circles), the first sensitivity experiment (black circles), and the second sensitivity experiment (red circles).

Download figure to PowerPoint

Table 3. Average Values and Variances of δ18Oprecip at Bangkok in MJJASO and at Bombay and Hong Kong in JJAS in the Standard Experiment, the First Sensitivity Experiment, and the Second Sensitivity Experiment
 Standard ExperimentSensitivity Experiment 1Sensitivity Experiment 2
Average value (Bangkok)−7.7‰−16.4‰−6.6‰
Variances (Bangkok)1.10.80.03
Average value (Bombay)−3.4‰−12.8‰−6.6‰
Variances (Bombay)1.20.80.08
Average value (Hong Kong)−7.9‰−17.1‰−6.4‰
Variances (Hong Kong)0.80.60.04

[42] We designed the second sensitivity experiment to remove the influences of both the amount effect and of distillation during transport on the interannual variability in δ18Oprecip at Bangkok, Bombay and Hong Kong. The results showed that at both Bangkok, Bombay and Hong Kong, most of the interannual variability in δ18Oprecip was removed. The variances of δ18Oprecip in the second sensitivity experiment were approximately 97%, 93%, and 95% smaller than they were in the standard experiment at Bangkok, Bombay, and Hong Kong, respectively. If we assume that the differences between the variances in the first and the second sensitivity experiments represent the contribution of distillation processes during transport to the interannual variability of δ18Oprecip at the three stations, then the contributions from distillation are 70%, 60%, and 70%, at Bangkok, Bombay, and Hong Kong, respectively. Thus, the contribution of distillation is about three times as large as the local amount effect at Bangkok and Hong Kong, and about twice as large at Bombay.

[43] In summary, the results of the two sensitivity experiments indicate that the dominant control on interannual variability of δ18Oprecip at Bangkok, Bombay, and Hong Kong is not the local amount effect but rather is distillation during transport from source regions.

5.3. Relationship With Large-Scale Atmospheric Variability

[44] Interannual variations of δ18Oprecip anomalies in the standard experiment and the first sensitivity experiment at Bangkok, Bombay, and Hong Kong tended to be strongly positive in El Niño years and strongly negative in La Niña years (Figure 10). The differences in average δ18Oprecip values between El Niño and La Niña years at Bangkok, Bombay, and Hong Kong were only 0.6‰ (or 20%), 0.6‰ (or 18%), and 0.2‰ (or 17%) smaller, respectively, in the first sensitivity experiment than in the standard experiment (Table 4). Since the results of the second sensitivity experiment at the three sites showed that effects other than the local amount effect and distillation effect were very small (Figure 9 and Table 3), the contribution of distillation during transport to δ18Oprecip at the three stations was approximately 2.4‰ (or 80%), 2.7‰ (or 82%), and 1‰ (or 83%), respectively. Therefore, the local amount effect does not dominate the relationship between δ18Oprecip and ENSO at the three sites; instead, most of the difference between El Niño and La Niña years is caused by the distillation that occurs along the transport pathways from the source regions. Therefore, distillation is the dominant control on the relationship between δ18Oprecip and ENSO.

image

Figure 10. Interannual variability of δ18Oprecip anomalies (a) at Bangkok during MJJASO and at (b) Bombay and (c) Hong Kong during JJAS in the standard experiment (white circles), in the first sensitivity experiment (black circles), and interannual variability of the Niño3.4 SST index (ENSO) (red circles). Red and blue bars represent El Niño and La Niña years, respectively.

Download figure to PowerPoint

Table 4. Differences in the Average Value of δ18Oprecip at Bangkok in MJJASO and at Bombay and Hong Kong in JJAS Between El Niño and La Niña Events in the Standard Experiment and the First Sensitivity Experiment
 Standard ExperimentSensitivity Experiment 1
Bangkok3‰2.4‰
Bombay3.3‰2.7‰
Hong Kong1.2‰1.0‰

6. Summary and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[45] Stable water isotopes in precipitation have been used as a proxy for paleoclimate over Asian monsoon regions, but their use involves many uncertainties because the contributions of the main factors causing their variability have not yet been investigated quantitatively. Therefore, in this more quantitative study, we estimated the contributions of the main factors controlling interannual variability of δ18Oprecip over the Asian monsoon region.

[46] As indicated previously [e.g., Vuille et al., 2005], on interannual timescales δ18Oprecip is significantly correlated with the observed precipitation in the Asian monsoon region, particularly at low latitudes, although a temperature effect is seen at a few places. In addition, time variations in δ18Oprecipare significantly correlated with atmospheric changes associated with ENSO events. By examining 10-day backward trajectories, we also showed that the precipitation along transport pathways from source regions also controls both the interannual variability ofδ18Oprecip and the relationship between δ18Oprecip and ENSO events. Thus, we identified two effects, a local amount effect and distillation along transport pathways, as the two main factors controlling the interannual variability of δ18Oprecip.

[47] By performing two sensitivity experiments, we showed that the dominant control on δ18Oprecip at Bangkok, Bombay and Hong Kong is distillation occurring during transport from source regions. The contributions of the local precipitation amount and distillation during transport were 27% and 70%, respectively, at Bangkok; 33% and 60%, respectively, at Bombay; and 13% and 85%, respectively, at Hong Kong. Although previous studies had identified distillation along transport pathways as being important, they had done so only qualitatively [e.g., Araguas-Araguas et al., 1998; Vuille and Werner, 2005; Vuille et al., 2005].

[48] We also showed that the dominant control on the relationship between δ18Oprecip and ENSO events was the variability in the precipitation along the transport pathways from source regions, although some previous studies have suggested that changes in precipitation at the observation site induced by ENSO may be the dominant factor in this relationship. The contribution of distillation during transport to the differences in δ18Oprecip between El Niño and La Niña years was 80%, 82%, and 83% at Bangkok, Bombay, and Hong Kong, respectively.

[49] Our results indicate that the precipitation variability caused by ENSO along the transport pathways tends to synchronize variability in δ18Oprecip over a wide area around the Western Pacific Ocean. For this reason, the spatial extent of the correlation of δ18Oprecip in precipitation with ENSO events is much more widespread than the extent of correlations with the local surface temperature and precipitation. Therefore, we conclude that δ18Oprecipat Bangkok, Bombay, and Hong Kong can be appropriately used as a large-scale index reflecting distillation along transport pathways from the source regions.

[50] Our findings in this study are based on simulations under modern climate conditions and on interannual timescales. Future studies should examine whether our conclusions hold for different timescales and under past climate conditions, such as those prevailing during the Last Glacial Maximum and the mid-Holocene. In addition, the influences of interannual changes in transport pathways should be investigated in other regions where the expected pathways are distinctly different.

[51] Some problems are associated with the use of backward trajectories. For example, backward trajectories assume an adiabatic process, although this assumption may not be reasonable in the Asian monsoon region, where latent heat release is large because of the formation of precipitation. “Tagged water experiments,” as described by Koster et al. [1986], Numaguti [1999], Bosilovich and Schubert [2002], and Yoshimura et al. [2004], can overcome these problems and should be performed in the future.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[52] This work was supported by a Grant-in Aid for Scientific Research (KAKENHI) (S)(19106008) from the Japan Society for the Promotion of Science, the Innovative Program of Climate Change Projection for the 21st Century of the Ministry of Education, Culture, Sports, Science and Technology (MEXT) and the Japan Society for the Promotion of Science (JSPS) grant 23686071. Comments by five anonymous reviewers and the editor are highly appreciated.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods and Data
  5. 3. Model Evaluation
  6. 4. Relationship With Climate Over the Asian Monsoon Region
  7. 5. Quantitative Determination of the Dominant Control of δ18Oprecip at Bangkok, Bombay, New Delhi, and Hong Kong
  8. 6. Summary and Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
jgrd18117-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
jgrd18117-sup-0002-t02.txtplain text document1KTab-delimited Table 2.
jgrd18117-sup-0003-t03.txtplain text document1KTab-delimited Table 3.
jgrd18117-sup-0004-t04.txtplain text document0KTab-delimited Table 4.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.