The present study used a globally coupled ocean-atmosphere-sea-ice-land-surface climate model named the Community Climate System Model version 4 (CCSM4) developed at the National Center for Atmospheric Research (NCAR) (Gent et al., 2011). The atmospheric model of the Community Atmosphere Model version 4 (CAM4) is employed at T31 resolution of approximately 3.75° × 3.75° and with 26 vertical levels. The land model, run on the same horizontal grid as the atmosphere, is the Community Land Model version 4 (CLM4). Both the ocean model (POP2) and sea-ice component (CICE4) use a 3° horizontal grid with 60 levels in the vertical direction. More details on CCSM4 are available online at http://www. cesm.ucar.edu/models/ccsm4.0/, and further information can be found in Gent et al. (2011).

Following the protocol of the Paleoclimate Modeling Intercomparison Project, the major differences in boundary conditions between the mid-Holocene (6 ka) and the pre-industrial (0 ka) period lie in the Earth's orbital parameters (Berger, 1978) and the atmospheric concentrations of greenhouse gases. To separate the contribution of ocean feedback from total climate change between the two periods, four experiments were undertaken: two atmosphere-ocean coupled (AOGCM) experiments (AO (0 ka) and AO (6 ka)) and two atmosphere-only (AGCM) simulations (A (0 ka) and A (6 ka)). The same configurations of the 0 ka experiments were performed by both the AOGCM and AGCM, and thus the differences between them, i.e. (AO (6 ka)-AO (0 ka)) and (A (6 ka)-A (0 ka)) allows us to discuss the role of ocean feedback during the mid-Holocene. Further details about the simulations are given in Tian and Jiang (2013).

The monsoon climate is characterized not only by annual reversal of prevailing surface winds, but also by a marked contrast between a wet summer and dry winter (Webster et al., 1998; Wang and Ding, 2006). Following the definition given by Liu et al. (2009) and Wang et al. (2012), GMA is defined here as the region in which the local summer-minus-winter precipitation rate exceeds 2 mm d^-1 and the local summer rainfall exceeds 55% of annual total precipitation. Summer denotes May through September (MJJAS) in the NH and November through March (NDJFM) in the SH. Similar to the spatial pattern shown in Wang et al. (2012), CCSM4 can reasonably reproduce the present monsoon regions over land in North, East, and South Asia, northern and southern Africa, northern Australia, and North and South America, as well as over the oceans of the tropical western Pacific and western Indian Ocean (Fig. 1). As in Zhou et al. (2008), Hsu et al. (2012), and Wang et al. (2012), GMP is defined as the sum of total summer rainfall in the monsoon area, and the GMP intensity (GMPI) is measured by area-averaged mean summer rainfall in the GMA.

Figure 1

	Figure Option ViewDownloadNew Window
	Figure 1 Global monsoon areas for the two (a) AGCM and (b) AOGCM experiments. Grey indicates the common areas between 6 ka and 0 ka, and blue (red) indicates expanded (reduced) monsoon areas at 6 ka relative to 0 ka.

In the AGCM simulations (Fig. 1a), the mid-Holocene monsoon area expanded slightly over most land monsoon regions as well as the Indian Ocean, whereas it reduced over the tropical western Pacific and Atlantic with respect to the pre-industrial period. At the regional scale, land monsoon area increased by 0.8× 10⁶ km² (12.1%) in northern Africa (0-25° N, 20° W-50° E), 2.3× 10⁶ km² (33.5%) in Asia (5-70° N, 70-150° E), and 0.3× 10⁶ km² (3.1%) in South America (0-30° S, 35-85° W), but decreased by 0.3× 10⁶ km² (8.5%) and 0.5× 10⁶ km² (7.9%) in North America (0-30° N, 55-120° W) and southern Africa (0-35° S, 10-40° E). In the AOGCM simulations (Fig. 1b), however, the monsoon area changes between the mid-Holocene and pre-industrial period were different from those of the AGCMs. As compared to A (0 ka), monsoon regions additionally registered over India and the southeastern Pacific, increased over the western Indian Ocean, but decreased over the tropical western Pacific and disappeared over the eastern Indian Ocean and tropical West Atlantic in AO (0 ka). In AO (6 ka), monsoon areas expanded over most land monsoon regions as well as the western Indian Ocean and southeastern Pacific, but retreated over southern Africa, South America, and other ocean monsoon regions relative to AO (0 ka). At the regional scale, land monsoon area changes between the two periods were 2.5× 10⁶ km² (51.4%), 1.0× 10⁶ km² (6.5%), -0.5× 10⁶ km² (-9.9%), -0.5× 10⁶ km² (-6.5%), and -0.8× 10⁶ km² (-7.8%) in northern Africa, Asia, North America, southern Africa, and South America, respectively. Taking all simulations into account, ocean feedback showed a substantial impact on GMA changes between the two periods. Regionally and qualitatively over land, dynamic ocean induced an increase in monsoon area over northern Africa but a decrease over Asia as well as North and South America, and had no significant effect over southern Africa and northern Australia (10-30° S, 110-155° E). On the other hand, the ocean monsoon area increased in the western Indian Ocean and southeastern Pacific but reduced in the eastern Indian Ocean, tropical western Pacific, and tropical West Atlantic due to ocean feedback.

With reference to the pre-industrial period, the mid-Holocene ocean feedback led to more monsoon precipitation over the globe and global ocean areas but less over global land areas, with an average change of 10.9× 10⁹, 15.8× 10⁹, and -4.8× 10⁹ m³ d^-1, respectively. As compared to between the AGCM (Fig. 2a) and AOGCM (Fig. 2b) simulations, Fig. 2c shows that ocean- induced changes in monsoon precipitation were unevenly distributed worldwide. Monsoon precipitation generally increased in northern and southeastern Africa, the southwestern Indian Ocean, the tropical western and central North Pacific, North America, and the tropical West Atlantic, but decreased in East and South Asia, the southeastern Indian Ocean, northern Australia, and the tropical eastern Pacific. The former increase resulted from the ocean-induced convergence of water vapor flux in those regions, whereas the divergence held true for the latter regions (Fig. 2d). At the hemispheric scale, land monsoon precipitation decreased by 2.8× 10⁹ m³ d^-1 in the NH and 2.0× 10⁹ m³ d^-1 in the SH, while ocean monsoon precipitation increased by 13.9× 10⁹ m³ d^-1 in the NH and 1.8× 10⁹ m³ d^-1 in the SH, showing consistent change between the NH and SH land or ocean areas but opposite change between the land and ocean areas in the NH or SH. At the regional scale over land, monsoon precipitation reduced by 7.4× 10⁹, 2.7× 10⁹, 5.9× 10⁹, and 1.0× 10⁹ m³ d^-1 in Asia, North and South America, and northern Australia, respectively, while it increased by 5.1× 10⁹ and 1.9× 10⁹ m³ d^-1 in northern and southern Africa due to ocean feedback, respectively. Thus, the dynamic ocean-induced increase in GMP mainly resulted from more monsoon precipitation over the NH ocean areas, whereas less precipitation over land monsoon areas had a negative contribution.

Figure 2

Figure Option ViewDownloadNew Window

Figure 2 Changes in summer monsoon precipitation (units: mm d-1) for the (a) AGCM and (b) AOGCM, and (c) due to ocean feedback. (d) Changes in summer vertically integrated water vapor flux (arrows; units: kg m-1 s-1) and its divergence (shading; units: mm d-1) due to ocean feedback. The vertical integration of the water vapor flux was performed from the surface to 300 hPa, and the impacts of topography have been removed using the surface pressure. The units of divergence have been converted to mm d-1 assuming the density of liquid water as 1 g cm-3.

On the whole, the mid-Holocene GMPI decreased due to ocean feedback. The intensity of monsoon precipitation decreased by 0.11 and 0.15 mm d^-1 for the globe and global land areas, respectively, but increased by 0.13 mm d^-1 for the global ocean. At the hemispheric scale, monsoon precipitation weakened over the NH land and ocean areas but strengthened over the SH monsoon areas, revealing consistent change in monsoon precipitation intensity between land and ocean in the NH or SH but opposite change between the NH and SH land or ocean. Regionally, monsoon precipitation intensity decreased in northern Africa, Asia, North America, and northern Australia but increased in southern Africa and South America due to ocean feedback. Thus, the dynamic ocean-induced decrease in GMPI resulted from the weakened monsoon precipitation over the NH land monsoon areas.

The influence of the dynamic ocean on monsoon precipitation occurs mainly through changes in evaporation from the ocean and the large-scale atmospheric circulation. In response to the mid-Holocene changes in orbital forcing and atmospheric concentrations of greenhouse gases, ocean feedback led to significant changes in summer surface temperature relative to the pre-industrial period (Fig. 3). In the NH, the mid-Holocene ocean was warmer at the mid- and high-latitudes but colder at the lower latitudes. Accordingly, summer temperature generally rose north of approximately 50° N but decreased in the south, except in East and South Asia, western Europe, the northeastern Pacific, and the western Atlantic, leading to an overall decrease in meridional temperature gradient through the lower to mid and high latitudes. This was beneficial to the occurrence of anomalous southerly winds, which brought additional moisture from the tropics in the lower troposphere to compensate for warming-induced loss of atmospheric mass due to intensified ascending motion or weakened descending motion at the mid and high latitudes. In the SH, however, summer temperature change was very weak, regionally and zonally. At the same time, ocean-induced change in summer temperature was stronger in the land than in the ocean areas, resulting in an overall decrease (increase) in the land-sea thermal contrast in the NH, except for at approximately 60° N (in the SH), which was unfavorable (favorable) to inland winds and moisture transport, and thus monsoon precipitation intensity, over the land monsoon areas. Regionally, summer temperature increased in East and South Asia but decreased in the adjacent oceans, inducing an increase in both zonal and meridional land-sea thermal contrast, which was favorable to anomalous southerly winds and thus monsoon precipitation over East and South Asia.

Figure 3

	Figure Option ViewDownloadNew Window
	Figure 3 Summer (MJJAS for the NH and NDJMF for the SH) temperature changes (units: K) due to ocean feedback, with the corresponding changes in the zonal mean of summer temperature (units: K; blue line) and land minus ocean surface temperature (units: K; red line) in the right panel.

However, the summer vertically integrated water vapor flux diverged over those monsoon regions (Fig. 2d), corresponding to decreased precipitation over the East and South Asian monsoon regions (Fig. 2c).