Simulation of the Indian monsoon and its variability during the last millennium

Introduction Conclusions References


Introduction
Based on long-term climate reconstructions derived from various well-dated proxy data (e.g.Borgaonkar et al., 2010;Fleitmann et al., 2007;Liu et al., 2009;Ponton et al., 2012;Prasad et al., 2006) the last Millennium is the best documented historical climate epoch.It can be divided into two major climate periods: the Medieval Warm Period (ca.900-1350 AD) and the Little Ice Age (ca.1500-1850 AD) (e.g.Graham et al., 2010;Lamb, 1965;Grove, 1988).Variations in solar insolation coupled with the remote impact of the internal dynamics of climate modes in the oceans like ENSO and Indian Ocean Dipole are some of the major drivers leading to long-term fluctuations in global temperature conditions during the last 1200 yr with the occurrence of the characteristic warm and cold periods.These thermal changes show a strong impact on global and regional climate phenomena like monsoons (Polanski et al., 2012), which are primarily formed due to seasonal and latitudinal differences in the incoming solar radiation with effects on the land-sea thermal contrast (Gadgil, 2003;Webster et al., 1998).As consequence, large-scale pressure gradients are generated including strong low-level atmospheric wind circulations (Dallmeyer et al., 2012).Monsoon systems are characterised by a strong spatiotemporal variability from multi-millennial to intraseasonal time scales (Ding, 2007;Lau et al., 2000;Wang, 2006).The Asian Monsoon System is the strongest monsoon system of the world (Clift and Plumb, 2008), which is divided into two strongly non-linear interacting subsystems: the East Asian Monsoon and the Indian Monsoon (Wang et al., 2001) affecting the livelihood of more than 2.5 billion people especially by the increased occurrence and frequency of extreme monsoon rainfall like deficient rainfall events with impacts on meteorological and agricultural drought conditions in recent times (e.g.Krishnan et al., 2009;Shaw and Nguyen, 2011;Ummenhofer et al., 2012).Therefore, the understanding of the Introduction

Conclusions References
Tables Figures

Back Close
Full large-scale mechanisms and regional spatiotemporal variations leading to extended past monsoon failures (megadroughts) is crucial for improving the Indian Monsoon forecasting and to develop proper mitigation strategies for the future.In this context a combined multi-proxy-climate model approach, as presented in HIMPAC ("Himalaya: Modern and Past Climates"; http://www.himpac.org)and CADY ("Central Asian Climate Dynamics"; http://www.cady-climate.org)projects, can be used to analyse the past monsoon climate in India within the last Millennium.Several paleoclimate studies have been already carried out to investigate drought conditions in Asia during the last Millennium as well as in present-day (e.g.Cook et al., 2010;Krishnan et al., 2009;Sinha et al., 2011b) using different multiproxy archives, observation and reanalysis data as well as climate model simulations.For instance based on new tree-ring width reconstructions, Sinha et al. (2011b) postulate episodic and widespread reoccurrences of monsoon megadroughts continued throughout the LIA, which are not clearly related to ENSO variability in the tropical Pacific Ocean as seen in present-day.Moreover, intraseasonal monsoon variability with a "break" spell dominated mode favors reduced precipitation over India along with southward retreat of ITCZ (Sinha et al., 2011b) and intensified rainfall over the Himalayan region (Gadgil, 2003).Other studies suggest that the occurrence of wet (India-Pakistan region) and dry (Bay of Bengal region) monsoon signals are in response to shifting atmospheric circulation patterns due to changes in SST and local convection (Mujumdar et al., 2012).The combination of statistical data analysis and/or climate simulations with paleoclimatic reconstruction methods has been performed in several studies (e.g.Hind et al., 2012;Jones et al., 2009;Kleinen et al., 2011;Rehfeld et al., 2012;Sundberg et al., 2012). of external forcing parameters like solar variations, volcanoes or land cover changes derived from reconstructions.General circulation models of the atmosphere (AGCMs) are often used to simulate the large-scale atmospheric circulation (e.g.Kaspar et al., 2010) and the related monsoon circulation.Owing to the coarse horizontal resolution, global AGCMs are not able to capture regional-scale atmospheric circulation and precipitation processes, which strongly depend on the realistic simulation of the local topography (Polanski et al., 2010).Therefore, in this study a higher resolved AGCM is used to better resolve the rainfall patterns in orographically induced regions.In addition the model simulations will be compared with paleoclimatic reconstructions based on different proxies and archives in India.In this context the reconstructed climate of the well-dated Lonar record in Central India plays an important role (Anoop et al., 2012;Prasad et al., 2012).Due to its special location in the sensitive Core Monsoon Zone, the site is probably influenced by both monsoon branches: from the Arabian Sea and the Bay of Bengal.Particularly with regard to the long continuously chronology of the last 11 000 yr, the Lonar site presents a unique possibility for a comparison of longterm climate time series.Therefore, here the evaluation with climate model simulations is introduced and highlighted for the first time.
After a short introduction of the global climate model simulations, the reconstructed relative moisture index and the observation data are presented as well as different climate indices used for this study (Sect.2).In Sect. 3 the modern spatiotemporal summer monsoon climatology is validated with observation data.Section 4 presents the simulation results for the last Millennium, in which the driving mechanisms for anomalous dry monsoon years are analysed.Further, the Palmer Drought Severity Index is calculated to quantify the monsoon failures in model data in comparison with tree-ring based drought conditions.In addition, the calculated relative annual rainfall anomalies are compared with paleoclimatic reconstructions based on the moisture index.Finally, concluding remarks summarise the main results (Sect.5).Introduction

Conclusions References
Tables Figures

Back Close
Full 2 Model and data

Global climate model simulations
The general framework (Fig. 1) of the climate model simulations within the HIMPAC and CADY project for the last Millennium encompasses a three step hierarchy of model runs starting from a fully coupled atmosphere-ocean Full  (Marsland et al., 2003): the GCM for the ocean.The dynamical core of ECHAM5 is based on spectral weather prediction models, which have been developed at the ECMWF and modified for climate research at MPI.The MPIOM consists of thermodynamic sea-ice processes and simulates the ocean circulation in response to the atmospheric forcing.ECHAM5 and MPIOM are coupled via the OASIS coupler.Furthermore the ocean biogeochemistry module HAMOCC5 (HAMburg Oceanic Carbon Cycle Model, version 5) (Wetzel et al., 2006) and the land surface scheme JSBACH (Joint Scheme for Biosphere Atmosphere Coupling in Hamburg) (Raddatz et al., 2007) are considered in the MPI-ESM.The time slice experiments with the atmosphere-alone component ECHAM5 have been run with a higher spectral horizontal resolution of T63 (ca.1.8 • × 1.8 • = ca.180 × 180 km) and 31 vertical model levels.The SST and SIC data have been taken from the coarse resolved model and have been spatially interpolated onto the T63 horizontal grid to initialize the model runs.The same set of full forcing has been also applied for the ECHAM5 simulations to get consistent model runs.to the highest temporal correlation of ensemble member mil0014 (0.67) for AIMR compared to the ensemble mean, three 200-yr long time slices for Medieval Warm Period (MWP; 900-1100 AD), Little Ice Age (LIA; 1515-1715 AD) and Recent Climate (REC; 1800-2000 AD) have been chosen to simulate them with the higher resolved atmosphere-only ECHAM5 model to identify the response of different horizontal resolutions of the atmospheric model component on the simulation of extreme monsoon rainfall events on multidecadal to centennial time scales and to investigate the dynamics and teleconnections leading to these events.The selection of the time slices within the prominent climatic periods for the MWP and LIA has been also done in comparison with the thermal conditions during these periods to identify the most dominant climatic signals within the last 1000 yr.

Observational and reanalysis data
The validation of modern summer monsoon rainfall patterns over land surface regions between ECHAM5 model and gridded observations has been carried out with GPCC6 reanalysis data from Global Precipitation Climatology Centre (Schneider et al., 2011) and the APHRODITE Monsoon Asia (60 Set (Yatagai et al., 2012) for a present-day period from 1951-2000.Both observation data sets have a spatial horizontal resolution of 0.5 • .

Climate indices
In order to quantify the long-term relationship between global teleconnection patterns and the AIMR in summer (JJAS) and winter (DJF) the temporal correlation between different climate indices and the AIMR has been calculated.

Total Solar Irradiance
The Total Solar Irradiance (TSI) is defined as the amount of the shortwave radiation emitted by the sun with a distinctive range of wavelengths (spectra) determined by the sun temperature.Total means that the solar flux has been integrated over all wavelengths (http://www.ipcc.ch/pdf/glossary/tar-ipcc-terms-en.pdf;http://science.jrank.org).

Oceanic Ni ño Index
The oceanic part of ENSO phenomenon has been analysed by calculating the Oceanic Ni ño Index (ONI) from NOAA/NWS/CPC (2011) based on the SST anomalies in the Introduction

Conclusions References
Tables Figures

Back Close
Full

Dipole Mode Index
The Dipole Mode Index (DMI) has been calculated by the difference of SST anomalies between the tropical western Indian Ocean (50  (Clark et al., 2003).

North Atlantic Oscillation
The winter (DJF) station-based NAO Index (Hurrell, 1995) is estimated by the difference of normalized SLP between Lisbon, Portugal and Stykkisholmur/Reykjavik on Iceland.Positive (negative) values are associated with stronger zonal (meridional) stream in mid-latitudes, more intense (weaker) weather systems and warmer (cooler) temperatures over Western Europe.

Palmer Drought Severity Index
The Palmer Drought Severity Index (PDSI) developed by Palmer (1965) is one of the best climatological indicators of the prolonged and abnormal periods of moisture deficiency, which can quantify the drought severity over different climates (Wells et al., 2004).Monthly PDSI has been calculated to quantify long-term drought conditions within the ECHAM5 simulations for the period from 1856 to 2000 AD overlapping with the Kaplan SST V2 data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA from their web site at http://www.esrl.noaa.gov/psd/for further investigation of correlation between Sea Surface Temperature Anomalies (SSTA) and droughts over Monsoon Asia.The climatological dataset of available water content from global soil texture and derived water-holding capacities (Webb et al., 2000) has been used for PDSI calculations.

Modern summer monsoon climatology and variability
In order to evaluate the skill of the ECHAM5 simulated spatiotemporal summer monsoon patterns, in a first step the rainfall simulations have been compared with the coarse resolved MPI-ESM of Millennium Experiment (Jungclaus et al., 2010) as well as with two gridded observed rainfall data sets (Schneider et al., 2011;Yatagai et al., 2012) for a recent time period from 1951-2000.The focus is on analysing the reliability of the climate simulations with respect to different horizontal resolutions for past climate studies.Introduction

Conclusions References
Tables Figures

Back Close
Full

Spatial rainfall patterns (JJAS)
Compared to GPCC6 the agreement in the climatology and interannual variability of the spatial ISM rainfall patterns over land surface regions, simulated by the higher resolved ECHAM5 model for the period from 1951-2000 (not shown), is much better than for the coarse resolved MPI-ESM, which is primary related to the more detailed representation of orographic features in the higher resolved GCM.The coarse model shows a homogeneous rainfall distribution, whereas ECHAM5 T63L31 simulates topography induced rainfall belts more realistically.Due to the characteristic modern ISM circulation with deep convection and orographic uplifting processes, intensive and long-term extended rainfall events occur especially in the windward of the Western Ghats at the Malabar Coast, at the southern slopes of the Himalayas and in the northern Bay of Bengal, which are relatively well captured by ECHAM5 even if the amount of rainfall is less than in GPCC6.Therefore, the realistic simulation of orography and small scale rainfall patterns respectively is crucial to get a better understanding of changes in rainfall distribution over short horizontal distances in a highly structured relief, which has been achieved by the higher resolved ECHAM5 model.

Temporal rainfall climatology (JJAS)
The Taylor diagram (Taylor et al., 2001) has been used to compare the simulated and observed temporal summer monsoon rainfall climatology from 1951-2000 covering a box over India (0 • N-50 • N and 60 • E-12 • E).First the model output and the observations have been interpolated to a coarser grid before analysing the data.Figure 4 shows a good temporal correlation and a small RMSE between the two observations as well as between the two models.The coarser resolved MPI-ESM (D) represents a slightly better agreement with GPCC6 (B) and APHRODITE (A) than ECHAM5 (C) with the increased spatial resolution.The standard deviation is also closer to GPCC6.It is believed that higher spatial resolved simulations can better capture the internal climate variations which would also limit the potential agreement between model simulations Introduction

Conclusions References
Tables Figures

Back Close
Full due to interpolation errors, e.g.here we used the bilinear interpolation to decrease the spatial resolutions to the lowest resolved model (MPI-ESM).The temporal correlation of the two climate models and GPCC6 is about 0.6 and the centered RMSE is about 0.5 for MPI-ESM and 0.7 for ECHAM5.
In general the ECHAM5 model is able to successfully simulate the modern spatiotemporal patterns of summer monsoon rainfall climatology and variability with respect to observation data, which has been already shown in other state-of-the-art global monsoon modeling studies (e.g.Dallmeyer et al., 2010).Therefore, it can be used for further interpretation of past climate changes in the Indian Monsoon.
4 Climatic change of Indian monsoon during last millennium

Simulated rainfall changes and comparison with reconstructions
Changes in the external forcing (e.g.Total Solar Irradiance (TSI), volcanoes or orbital parameters) and in the diabatic heating source have an impact on the circulation and rainfall patterns (Dallmeyer et al., 2010).In of local convection related to a decreasing moisture convergence, weaker upward motion of moisture and a decreased instability in the vertical column of the troposphere (Polanski et al., 2010).In LIA summer (b) the dipole pattern disappears compared to REC and a slight intensification of ISM is simulated with increasing rainfall over India, the Tibetan Plateau and over the regions influenced by EASM.The LIA is characterised by a summer cooling trend over tropics and a slight warming over northeastern Asia.Therefore the rainfall changes are more embedded in the large-scale circulation driven by the meridional temperature gradient between the cooler Indian Ocean with the surrounding land surface regions and the warmer northeastern Eurasian continent (not shown).The comparison in (c) verifies the characteristic regional intensification and weakening trends in summer monsoon rainfall between the warmer and cooler time slice.
In the winter months (lower panel) the MWP (a) and LIA (b) show similar spatial rainfall anomalies compared to REC with more simulated rainfall over Indian subcontinent, Tibetan Plateau, the Bay of Bengal and East Asia and less rainfall northern of 35 • N. The wetter winter conditions during LIA (b) are connected to an enhancement of Westerlies inducing more precipitation events at Himalayan slopes and southern parts.The comparison between MWP and LIA (c) illustrates the drier winter in MWP over India and the surrounding ocean basins and wetter winter conditions eastern of that region.These changes are related to an intensification of Westerlies over Tibetan Plateau going along with the northern shift and weakening of the Siberian High (not shown).
In general the climatological summer and winter monsoon comparison between MWP and LIA indicates that less rainfall is simulated over India and southwestern regions and more rainfall is calculated over Central Asia during the warmer climate, which are determined by regional and large-scale dynamics.In order to quantify the dynamical relationship between AIMR in summer and winter months with external solar forcing and internal feedbacks based on different climate modes (see Sect. for the three simulated time slices.Before applying the calculation for the ECHAM5 data, the different indices have been validated for present-day with observation data (not shown).The solar forcing, defined by the TSI, is the main driver for summer monsoon rainfall with a strong anticorrelation of −0.95 (−0.94 and −0.95) for MWP (LIA and REC), which is statistically significant at a 95 % confidence level.A positive (negative) solar irradiance leads to a weakening (intensification) of summer monsoon rainfall.
The causal mechanisms between solar forcing and rainfall events with a focus on dry monsoon years will be described in Sect.4.2.Especially in the summer months, the external solar forcing is coupled and overlain by internal climate modes of the ocean (ENSO and IOD) with asynchronous intensities and lengths of periods shown by anticorrelations of −0.61 (−0.57and −0.65) for the ONI in MWP (LIA and REC) and −0.68 (−0.57and −0.70) for the DMI in MWP (LIA and REC).Further, the correlation analysis showed, that the influence of the climate indices on the winter monsoon rainfall is much weaker, except the NAO Index has a weak positive correlation to the AIMR in winter of 0.33 (0.24 and 0.28) for MWP (LIA and REC).Whereas PDO shows a median but not significant dependency of 0.50 (0.54 and 0.54) on the AIMR in summer, the AMO mode has no correlation to the simulated AIMR.Moisture changes during the last millennium have been described by paleoclimatic reconstructions based on a relative moisture index using different archives and proxies (Anoop et al., 2012;Bhattacharya et al., 2007;Chauhan et al., 2000;Denniston et al., 2000;Ely et al., 1999;Kar et al., 2002;Ponton et al., 2012;Prasad et al., 2012) (Fig. 6).The figure compares the ECHAM5 simulated relative annual rainfall anomalies and the reconstructed paleoclimatic moisture index between MWP, LIA and REC.Since the driving model simulations of the coarse resolved Millennium Experiment show a significant long-term increasing trend in the solar irradiance and the Meridional Overturning Circulation (MOC) in the Atlantic Basin for the last 100 yr (Zanchettin et al., 2010), the ECHAM5 simulated rainfall time series for REC had been detrended before the comparison with the rainfall anomalies of the other time slices to eliminate the long-term climate signal.Introduction

Conclusions References
Tables Figures

Back Close
Full Especially for MWP (a) there is a good spatial agreement between the model simulation and the reconstructions.The dry (wet) rainfall signal over Central India (Himalayas) can be detected from both.However, the model and the reconstructions disagree in the LIA (b), where the model simulates relatively more rainfall over Central India and Himalayas while the reconstructions indicate a marginally drier anomaly, but the signal is not very clear and pronounced.Therefore, more sites and higher resolved climate model simulations have to be considered in the analysis to better compare the model results with respect to the inhomogeneities in the spatial rainfall distribution.The study from Dallmeyer et al. (2012) further suggests the importance of extending the analysis period of monsoon climate change to other seasons since the paleoclimate moisture reconstructions cannot simply be interpreted as indicator for summer monsoon intensity and related precipitation.It is assumed that the orbital variations lead to seasonal insolation distribution and affected the rainfall in all seasons.Therefore, the reconstructed moisture signals should not simply be correlated and explained with variations in summer insolation (Dallmeyer et al., 2012).

Simulated rainfall anomalies
In order to investigate the interannual summer monsoon (JJAS) variability of the last Millennium with a focus on extreme rainfall events, three 30-yr-long strong (wet) and weak (dry) monsoon composites for every 200-yr time slices of MWP, LIA and REC have been selected and compared with each other regarding their characteristic rainfall anomalies and regional and large-scale dynamical drivers.The focus of the following analysis is on dry rainfall events leading to significant monsoon failures (droughts).The selection of the anomalous dry summer monsoon years (MWP: 992-1022 AD, LIA: 1633-1663AD and REC: 1913-1943 AD) has been done by calculation of ECHAM5 time series averaged over AIMR region (not shown).(upper panel) and 2 m-air-temperature (lower panel) for "MWP minus REC" (a), "LIA minus REC" (b) and "MWP minus LIA" (c).During weak monsoon years of MWP and LIA the spatial large-scale pattern of the precipitation differences are quite similar.This is an indication that the driving mechanisms of extreme dry summer monsoon rainfall are similar in both periods compared to recent climate.However, the interpretation of rainfall patterns in LIA has to be considered carefully due to the disagreement in the model-proxy-comparison (see Sect. 4.1).Particularly, similar regional patterns of rainfall anomalies are simulated over the Himalayas and Tibetan Plateau (wetter), central, southern and northeastern India including northern Bay of Bengal (drier).In contrast, northwestern India and the northern Arabian Sea show opposite summer rainfall signals between MWP (wetter) and LIA (drier), which are related to changes in atmospheric circulation and temperatures.These changes can be attributed to a similar shift of the West Pacific Subtropical High during 2010 flooding in Pakistan with suppressing of convection around Bay of Bengal and sustaining the rainfall activity over northwestern India and Pakistan, strongly influenced by internal modes of the ocean like La Ni ña and IOD (Mujumdar et al., 2012).In MWP the temperature gradient between Indian Ocean and Eurasian land surface is higher due to warmer temperature anomalies especially over central India and the continental regions north of Arabian Sea (+0.5 K) compared to LIA (see Fig. 7c).The SST changes over the Indian Ocean are another important driver for the modified atmospheric circulation and rainfall patterns in dry monsoon years between MWP and LIA (not shown), which also influence the local convection.The positive anomalies in vertical velocity at 500 hPa (see Fig. 7, upper panel) represent the regions with strongest convective activity, which are associated with a strong moisture convergence as well as a strong upward motion in the middle troposphere.This relationship is approved for MWP and LIA respectively.In MWP the strongest convective activity occurs over India and Arabian Sea, whereas the convection is reduced over these parts in LIA as a result of shifted atmospheric circulation patterns.Changes in the distribution of rainfall anomalies can be attributed to shifting regional atmospheric circulation patterns driven by modified regional SST anomalies Introduction

Conclusions References
Tables Figures

Back Close
Full in the Indian Ocean due to changes in external forcing and internal feedbacks of the surrounding oceans (see also Sect. 4.1).Figure 8 schematise the ECHAM5 estimated influence and strength of the external forcing versus internal feedbacks of the ocean on the AIMR in the summer monsoon months (JJAS) for dry (wet) monsoon years in the MWP time slice.As already discussed in Sect.4.1, the dominant driver of summer monsoon rainfall variability are variations in the Total Solar Irradiance with even higher temporal anticorrelations of −0.98 (−0.95) for dry (wet) monsoon years on interannual than on multidecadal time scale, overlapped by nonlinear feedbacks of the oceanic internal climate modes.The ENSO phenomenon has a stronger internal feedback on the interannual rainfall anomalies than the DMI with a temporal anticorrelation of −0.70 (−0.69) in dry (wet) monsoon years, whereas the correlation between AIMR and DMI tends towards zero.Even if the same hierarchy of forcing for dry and wet monsoons is obvious, the physical mechanisms and effects leading to the formation of the characteristic interannual rainfall anomalies are reverse.These processes can be summarised as follows: positive (negative) solar irradiance leads to warming (cooling) effects of the Indian Ocean with a weakening (enhancement) of the meridional temperature gradient between the Eurasian land mass and the ocean and a weakening (intensification) of the lower-tropospheric heat low.Therefore, the moisture advection towards the continent embedded in a reduced (enhanced) southwestern low-level jet over Arabian Sea and Bay of Bengal is weaker (stronger).Finally, a weaker (stronger) convective activity going along with a lower (higher) occurrence and frequency of summer rainfall events can be simulated.Further it can be identified, that the convection is driven by large-scale moisture advection from the Indian Ocean and a strong upward motion of water vapor over land related to enhanced instability and moisture convergence, which has been already analysed in previous studies (e.g.Bhalme et al., 1980;

Conclusions References
Tables Figures

Back Close
Full In addition, a qualitative comparison of the simulated annual rainfall changes with paleoclimatic reconstructions based on a moisture index has been carried out for the MPW and LIA according to REC.In this context, the well-dated and long chronology of the Lonar record, has been highlighted and compared for the first time with model data.For MWP there is a good spatial agreement between the model simulation and the reconstructed moisture signal.The dry (wet) rainfall signal over Central India (Himalayas) can be detected from both.However, the model and the reconstructions disagree in the LIA, where the model simulates relatively more rainfall over Central India and Himalayas while the reconstructions indicate a marginally drier anomaly, but the signal is not very clear and pronounced.Therefore, more sites and higher resolved climate model simulations have to be considered in the analysis for a better model-proxy comparison with respect to the inhomogeneities in the spatial rainfall distribution and the reconstruction-inherent uncertainties of proxy data.
In order to investigate the interannual summer monsoon (JJAS) variability of the last Millennium with a focus on extreme rainfall events, three 30-yr-long strong (wet) and weak (dry) monsoon composites for every 200-yr time slices of MWP, LIA and REC have been selected and compared with each other regarding their characteristic rainfall anomalies and regional and large-scale dynamical drivers.The focus has been on dry rainfall events leading to significant monsoon failures (droughts).During weak monsoon years of MWP and LIA the spatial large-scale pattern of the precipitation differences are quite similar.This is an indication that the driving mechanisms of extreme dry summer monsoon rainfall are similar in both periods compared to recent climate.Particularly, similar regional patterns of rainfall anomalies are simulated over the Himalayas and Tibetan Plateau (wetter), central, southern and northeastern India including northern Bay of Bengal (drier).In contrast, northwestern India and the northern Arabian Sea show opposite summer rainfall signals between MWP (wetter) and LIA (drier), which are related to changes in atmospheric circulation and temperatures.In MWP the temperature gradient between Indian Ocean and Eurasian land surface is higher due to warmer temperature anomalies especially over central India and the continental regions north of Arabian Sea (+0.5 K) compared to LIA.The positive anomalies in vertical velocity at 500 hPa represent the regions with strongest convective activity, which are associated with a strong moisture convergence as well as a strong upward motion in the middle troposphere.In MWP the strongest convective activity occurs over India and Arabian Sea, whereas the convection is reduced over these parts in LIA as a result of shifted atmospheric circulation patterns.Changes in the distribution of rainfall anomalies can be attributed to shifting regional atmospheric circulation patterns driven by modified regional SST anomalies in the Indian Ocean due to changes in external forcing and internal feedbacks of the surrounding oceans.
The dominant drivers of summer monsoon rainfall variability are variations in the Total Solar Irradiance with high temporal anticorrelations of −0.98 (−0.95) for dry (wet) monsoon years, overlapped by nonlinear feedbacks of the oceanic internal climate modes.
The ENSO phenomenon has a stronger internal feedback on the interannual rainfall anomalies than the DMI with a temporal anticorrelation of −0.70 (−0.69) in dry (wet) monsoon years.Even if the same hierarchy of forcing for dry and wet monsoons is obvious, the physical mechanisms and effects leading to the formation of the characteristic interannual rainfall anomalies are reverse.These processes can be summarised as follows: positive (negative) solar irradiance leads to warming (cooling) effects of the Indian Ocean with a weakening (enhancement) of the meridional temperature gradient between the Eurasian land mass and the ocean and a weakening (intensification) of the lower-tropospheric heat low.Therefore, the moisture advection towards the continent embedded in a reduced (enhanced) southwestern low-level jet over Arabian Sea and Bay of Bengal is weaker (stronger).Finally, a weaker (stronger) convective activity going along with a lower (higher) occurrence and frequency of summer rainfall events can be simulated.
Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | The following study describes the monsoon variability and its large-scale physical mechanisms on different time scales with a focus on monsoon failures within the last Millennium for the Medieval Warm Period, the Little Ice Age and the recent climate respectively by an analysis of global climate model simulations with the general circulation model ECHAM5.To get a consistent view of the past climate, paleoclimatic model simulations describe the historical climate on different spatiotemporal scales by consideration Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 2
Figure 2 illustrates annual rainfall time series averaged over the All-India-Monsoon-Region from 800-2005 AD for the five different ensemble members of the Millennium experiment (Jungclaus et al., 2010) using the coarse resolution MPI-ESM.According 709 encompassing also the Core Monsoon Zone in Central India as well as the northern Himalayan region (see Fig.3).The annual relative moisture signal from the multiproxy investigations has been translated into a qualitative moisture index for a three-part scale: minus (plus) values indicate drier (wetter) conditions in each 200 yr time slice compared to the reference period."No changes" are marked with zero values.Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Experiment.The thresholds for El Ni ño and La Ni ña are defined by the following criteria: El Ni ño: at least 5 consecutive seasons with an ONI ≥ 0.5 • C. La Ni ña: at least 5 consecutive seasons with an ONI ≤ 0.5 • C.
Discussion Paper | Discussion Paper | Discussion Paper | 2.3.6Pacific Decadal Oscillation The Pacific Decadal Oscillation (PDO) is a SST variability mode of about 20-30 yr in the northern Pacific Ocean (120 • E-240 • E and 20 • N-70 • N).During a warm (cold) phase the Pacific becomes cool (warm) Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Fig. 5 the climatological summer (JJAS; upper panel) and winter monsoon (DJF; lower panel) relative rainfall anomalies as well as differences in mean sea level pressure are illustrated for the ECHAM5 simulations during the MWP, LIA and REC time slices.Compared to REC, for the MWP (a) summer months, ECHAM5 simulates less rainfall over most parts of western Central Asia, the northern Arabian Sea, India and the Bay of Bengal and increasing rainfall northeast of this region including the Tibetan Plateau.This dipole precipitation pattern is related to regional changes in temperature and circulation.The overall warming signal in MWP is more pronounced north of 30 • N and a slight cooling is simulated in the tropics.Therefore, the intensity of temperature gradient and the position of the ITCZ are shifted to the east going along with an enhancement of EASM (not shown).Drier conditions over northern Arabian Sea and Bay of Bengal are connected to lower SSTs in the ocean basins going along with a reduction Discussion Paper | Discussion Paper | Discussion Paper | 2.3), the temporal correlation coefficients have been calculated among the different indices Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Figure 7 illustrates ECHAM5 simulated anomaly composites of summer monsoon rainfall and vertical velocity at 500 hPa Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Drought Severity Index and global SST anomaliesFurther the characteristic rainfall patterns in dry monsoon years have been correlated to global SST anomalies by using Empirical Orthogonal Function (EOF) Analysis to quantify the relationship between monsoon failures (droughts) on interannual time scale and global internal variability modes of the oceans.According to that, the EOF/PC1 and EOF/PC4 of the PDSI (see alsoSect.2.3.6)have been calculated within the Monsoon Asia region for the 200-yr-long ECHAM5 present-day simulation (REC) from 1800-2000 AD and later correlated with global SST anomalies from Kaplan SST V2 data for pre-and summer monsoon (MAMJJA) and winter monsoon (DJF) season between 1856 and 2000 AD.The simulation of the broad-scale monsoon failure events with robust long-term modes of spatiotemporal variability can provide deeper insights into the dynamics of Indian monsoon variability(Cook et al., 2010).Two of the EOF modes are highlighted in Fig.9.The leading pattern accounts for 15.75 % of the total field variability.Dominant same-sign loadings are detected over India, South and Southeast Asia with opposite sign loadings over Tibetan Plateau, northern Pakistan, Pamir and Tien-Shan Mountains, which are in well agreement with the DEOF1 of the reconstructed PDSI fromCook et al. (2010).It indicates the simulated dipole rainfall patterns in dry monsoon years between India and Tibetan Plateau seen especially in LIA.Moreover the temporal expansion of this mode (PC1) correlated well with the hemispherically SST anomalies of interdecadal variability in the Pacific Ocean, seen also inCook et al. (2010).The late-20th century trend toward drier conditions and weaker monsoons over India and Southeast Asia (see Fig. 9) can be explained by similar mechanisms leading to droughts in MWP and LIA.The EOF4 pattern accounts for 4.5 % of the total field variance with strong positive loadings over China and northeastern Asia and negative loadings over India.The PC4 of this mode is positively correlated with Indian Ocean SST anomalies, which are related to the DMI and tends to more positive PDSI values during the last 200 yr.The ENSO-PDSI relationship in Discussion Paper | Discussion Paper | Discussion Paper |the ECHAM5 simulation is relatively weak (not shown), whereas the internal dynamics of IOD is more pronounced.5ConclusionsThe general circulation model ECHAM5 has been used to simulate the Indian Monsoon Variability within the last Millennium.The focus has been on 200-yr-long time slices of the Medieval Warm Period (900-1100 AD), the Little Ice Age (1515-1715 AD) and the recent climate(1800( -2000 AD) AD).The evaluation of spatiotemporal monsoon patterns with present-day observation data is in agreement with other state-of-the-art monsoon modeling studies.Due to the better horizontal resolution of the ECHAM5 model, the rainfall patterns especially in the orographic regions can be captured more realistically.However, high resolution regional climate models are necessary as an additional tool to improve the skill for the simulation of small scale processes in comparison with observed data.The multidecadal rainfall changes have been simulated and compared for MWP and LIA according to the recent climate.ECHAM5 calculated a weakening (enhancement) in summer (winter) monsoon rainfall due to colder (warmer) SSTs in the Indian Ocean.Variations in the external solar insolation are the main drivers for these SST anomalies, shown by very strong temporal anticorrelations of −0.95 (−0.94 and −0.95) for MWP (LIA and REC) between the Total Solar Irradiance and the All-India-Monsoon-Rainfall in the summer monsoon months.A positive (negative) solar irradiance leads to a weakening (intensification) of summer monsoon rainfall.The external solar forcing is further coupled and overlain by internal climate modes of the ocean shown by anticorrelations of −0.61 (−0.57and −0.65) for the ONI in MWP (LIA and REC) and −0.68 (−0.57and −0.70) for the DMI in MWP (LIA and REC) with asynchronous intensities and lengths of periods.Moreover, the correlation analysis depicted a weaker influence of the climate indices on the winter monsoon rainfall.Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Furthermore
the characteristic rainfall patterns during years of weak monsoon activity have been correlated to global SST anomalies by using EOF Analysis to quantify the relationship between droughts on interannual time scale and global internal variability modes of the oceans for the present-day simulation from 1800-2000 AD.Dominant Figures Back Close Full Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

2.1.1 Max Planck Institute for Meteorology Earth System Model (MPI-ESM)
All climate model simulations used in this study are based on the analysis of a set of five ensemble members of the fully coupled MPI-ESM (Jungclaus et al., 2010), which were conducted in the framework of the Integrated Project Millennium (http://www.mad.zmaw.de/service-support/consortium-model-runs/millennium-experiments/index.html) at the Max Planck Institute for Meteorology (MPI).
(Roeckner et al., 2003)erent components, which are coupled with each other: ECHAM5 (European Centre, HAMburg, version 5)(Roeckner et al., 2003): the general circulation model for the atmosphere (AGCM) and MPIOM (Max Planck Institute Ocean Model) Introduction • E-70 • E and 10 • S-10 • N) and the tropical southeastern Indian Ocean (90 • E-110 • E and 10• S-0 • S).The index oscillates between positive and negative values.A positive (negative) DMI event is characterized by warmer (cooler) than normal SSTs in the western and cooler (warmer) SSTs in the eastern basin

Table 1 .
Paleoclimatic records from India used in this study.Records are listed from north to south.