第七章 結論與未來研究方向
深灰色虛線表示 10 個模式之系集平均
圖 A5-2 同圖 A5-1，但為 10 個模式於 IIa 區之個別結果。
79
圖 A5-3 同圖 A5-1，但為 10 個模式於 IIb 區之個別結果。
圖 A5-4 同圖 A5-1，但為 10 個模式於 III 區之個別結果。
80
圖 A5-5 同圖 A5-1，但為 10 個模式於 IVa 區之個別結果。
圖 A5-6 同圖 A5-1，但為 10 個模式於 IVb 區之個別結果。
81
圖 A5-7 同圖 5-3，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-8 同圖 A5-7，但為 10 個模式於 IIa 區之個別結果。
82
圖 A5-9 同圖 A5-7，但為 10 個模式於 IIb 區之個別結果。
圖 A5-10 同圖 A5-7，但為 10 個模式於 III 區之個別結果。
83
圖 A5-11 同圖 A5-7，但為 10 個模式於 IVa 區之個別結果。
圖 A5-12 同圖 A5-7，但為 10 個模式於 IVb 區之個別結果。
84
圖 A5-13 同圖 5-4，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-14 同圖 A5-13，但為 10 個模式於 IIa 區之個別結果。
85
圖 A5-15 同圖 A5-13，但為 10 個模式於 IIb 區之個別結果。
圖 A5-16 同圖 A5-13，但為 10 個模式於 III 區之個別結果。
86
圖 A5-17 同圖 A5-13，但為 10 個模式於 IVa 區之個別結果。
圖 A5-18 同圖 A5-13，但為 10 個模式於 IVb 區之個別結果。
87
圖 A5-19 同圖 5-5，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-20 同圖 A5-19，但為 10 個模式於 IIa 區之個別結果。
88
圖 A5-21 同圖 A5-19，但為 10 個模式於 IIb 區之個別結果。
圖 A5-22 同圖 A5-19，但為 10 個模式於 III 區之個別結果。
89
圖 A5-23 同圖 A5-19，但為 10 個模式於 IVa 區之個別結果。
圖 A5-24 同圖 A5-19，但為 10 個模式於 IVb 區之個別結果。
90
圖 A5-25 同圖 5-6，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-26 同圖 A5-25，但為 10 個模式於 IIa 區之個別結果。
91
圖 A5-27 同圖 A5-25，但為 10 個模式於 IIb 區之個別結果。
圖 A5-28 同圖 A5-25，但為 10 個模式於 III 區之個別結果。
92
圖 A5-29 同圖 A5-25，但為 10 個模式於 IVa 區之個別結果。
圖 A5-30 同圖 A5-25，但為 10 個模式於 IVb 區之個別結果。
93
圖 A5-31 同圖 5-7，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-32 同圖 A5-31，但為 10 個模式於 IIa 區之個別結果。
94
圖 A5-33 同圖 A5-31，但為 10 個模式於 IIb 區之個別結果。
圖 A5-34 同圖 A5-31，但為 10 個模式於 III 區之個別結果。
95
圖 A5-35 同圖 A5-31，但為 10 個模式於 IVa 區之個別結果。
圖 A5-36 同圖 A5-31，但為 10 個模式於 IVb 區之個別結果。
96
圖 A5-37 同圖 5-8，但為 10 個模式於六個次區域之個別結果。不同顏色代表不 同模式，深灰色虛線表示 10 個模式之系集平均。
97
圖 A5-38 同圖 5-9，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
圖 A5-39 同圖 A5-38，但為 10 個模式於 IIa 區之個別結果。
98
圖 A5-40 同圖 A5-38，但為 10 個模式於 IIb 區之個別結果。
圖 A5-41 同圖 A5-38，但為 10 個模式於 III 區之個別結果。
99
圖 A5-42 同圖 A5-38，但為 10 個模式於 IVa 區之個別結果。
圖 A5-43 同圖 A5-38，但為 10 個模式於 IVb 區之個別結果。
100
圖 A5-44 同圖 5-10，但為 10 個模式於 I 區之個別結果。不同顏色代表不同模式，
深灰色虛線表示 10 個模式之系集平均。
101
圖 A5-45 同圖 A5-44，但為 10 個模式於 IIa 區之個別結果。
102
圖 A5-46 同圖 A5-44，但為 10 個模式於 IIb 區之個別結果。
103
圖 A5-47 同圖 A5-44，但為 10 個模式於 III 區之個別結果。
104
圖 A5-48 同圖 A5-44，但為 10 個模式於 IVa 區之個別結果。
105
圖 A5-49 同圖 A5-44，但為 10 個模式於 IVb 區之個別結果。
Evaluating the ‘‘Rich-Get-Richer’’ Mechanism in Tropical Precipitation Change under Global Warming
CHIACHOU
Research Center for Environmental Changes, Academia Sinica, and Department of Atmospheric Sciences, National Taiwan University, Taipei, Taiwan
J. DAVIDNEELIN
Department of Atmospheric and Oceanic Sciences and Institute of Geophysics and Planetary Physics, University of California, Los Angeles, Los Angeles, California
CHAO-ANCHEN
Research Center for Environmental Changes, Academia Sinica, Taipei, Taiwan
JIEN-YITU
Department of Atmospheric Sciences, Chinese Culture University, Yang-Ming Shan, Taipei, Taiwan
(Manuscript received 6 February 2008, in final form 24 October 2008)
ABSTRACT
Examining tropical regional precipitation anomalies under global warming in 10 coupled global climate models, several mechanisms are consistently found. The tendency of rainfall to increase in convergence zones with large climatological precipitation and to decrease in subsidence regions—the rich-get-richer mechanism—
has previously been examined in different approximations by Chou and Neelin, and Held and Soden. The effect of increased moisture transported by the mean circulation (the ‘‘direct moisture effect’’ or ‘‘thermo-dynamic component’’ in respective terminology) is relatively robust, while ‘‘thermo-dynamic feedback is poorly un-derstood and differs among models. The argument outlined states that the thermodynamic component should be a good approximation for large-scale averages; this is confirmed for averages across convection zones and descent regions, respectively. Within the convergence zones, however, dynamic feedback can substantially increase or decrease precipitation anomalies. Regions of negative precipitation anomalies within the con-vergence zones are associated with local weakening of ascent, and some of these exhibit horizontal dry advection associated with the ‘‘upped-ante’’ mechanism. Regions of increased ascent have strong positive precipitation anomalies enhanced by moisture convergence. This dynamic feedback is consistent with re-duced gross moist stability due to increased moisture not being entirely compensated by effects of tropo-spheric warming and a vertical extent of convection. Regions of reduced ascent with positive precipitation anomalies are on average associated with changes in the vertical structure of vertical velocity, which extends to higher levels. This yields an increase in the gross moist stability that opposes ascent. The reductions in ascent associated with gross moist stability and upped-ante effects, respectively, combine to yield reduced ascent averaged across the convergence zones. Over climatological subsidence regions, positive precipitation anomalies can be associated with a convergence zone shift induced locally by anomalous heat flux from the ocean. Negative precipitation anomalies have a contribution from the thermodynamic component but can be enhanced or reduced by changes in the vertical velocity. Regions of enhanced subsidence are associated with an increased outgoing longwave radiation or horizontal cold convection. Reductions of subsidence are as-sociated with changes of the vertical profile of vertical velocity, increasing gross moist stability.
Corresponding author address: Chia Chou, Research Center for Environmental Changes, Academia Sinica, P.O. Box 1–48, Taipei 11529, Taiwan.
E-mail: [email protected]
1982 J O U R N A L O F C L I M A T E VOLUME22
DOI: 10.1175/2008JCLI2471.1
106
1. Introduction
Predicting future temperature changes under global warming is a challenging task, but predicting future precipitation changes may be even more difficult. The agreement among climate model simulations on the spatial distribution of time mean precipitation changes tends to be very poor, especially at a regional scale (e.g., Cubasch et al. 2001; Allen and Ingram 2002;
Stott and Kettleborough 2002; Neelin et al. 2006;
Meehl et al. 2007). This paper aims to contribute moisture and energy budget analysis of balances and mechanisms contributing to such precipitation changes in the tropics.
Before addressing the specific questions associated with this, we note that there are a number of aspects of precipitation change under global warming that are better documented. A number of model studies indicate that increased precipitation intensity and decreased precipitation frequency will most likely occur associated with the warming (recently, Wilby and Wigley 2002;
Trenberth et al. 2003; Kharin and Zwiers 2005; Meehl et al. 2005; Barnett et al. 2006; Sun et al. 2007). Evidence for such changes has also been sought in observations, as reviewed in Trenberth et al. (2007). Despite imper-fect simulation of precipitation distributions in climate models (e.g., Dai and Trenberth 2004; Dai 2006; Wilcox and Donner 2007), these effects are thought to be robust because of a simple underlying argument. Moisture con-tent available for extreme events tends to increase at a rate roughly governed by the Clausius–Clapeyron equa-tion, while the energy available to drive convection (for averages over sufficiently large scales that transports are negligible) increases less quickly (e.g., Allen and Ingram 2002; Meehl et al. 2007).
Increased precipitation at high latitudes and decreased precipitation in the subtropics is a common feature among climate models, while in the annual average the deep tropics tend to have a precipitation increase (Meehl et al.
2007). Examining precipitation change averaged over latitudinal bands, Zhang et al. (2007) argued for detect-able human impacts on precipitation, with larger ampli-tudes in observations than in model simulations. Even at regional scales, a few areas exhibit precipitation change that is consistent among model simulations (Christensen et al. 2007). For instance, relatively consistent projections are noted for twenty-first-century decreasing trends of precipitation in southern Europe (Rowell and Jones 2006), southwestern North America (Milly et al. 2005;
Seager et al. 2007), parts of Southeast Asian dry seasons (Li et al. 2007), and the Caribbean–Central America re-gion in summer (Neelin et al. 2006) where an observed trend is also noted.
In the tropics, at large scales on the annual average, Held and Soden (2006, hereafter HS06) sought robust features among climate models and found a weakening of the tropical circulation, which tends to compensate the effect of the increased atmospheric moisture on tropical precipitation in convergence zones. Chou et al. (2007) studied hemispherical averages of tropical precipitation and found a widening of the seasonal precipitation range between wet and dry seasons to be common among models. For large spatial averages, Neelin et al. (2006) used a measure of the amplitude for the increase and decrease of regional precipitation of each model and found these amplitudes to agree relatively well among climate models even if the locations of the strong changes differed.
Despite these areas where some agreement among climate models and observations can be found, very sub-stantial differences in the geographic distribution of pre-cipitation changes are typical not only at the regional scale but even for relatively large spatial averages (e.g., Allan and Soden 2007, 2008; Chou et al. 2007; Trenberth and Dai 2007; Trenberth et al. 2007; Wentz et al. 2007; Zhang et al.
2007) and in the tropical examples to be examined here.
Among mechanisms for regional change of mean tropical precipitation, Chou and Neelin (2004, hereafter CN04) and HS06 proposed closely related approaches, both of which are built on arguments by Manabe and coworkers (Manabe and Wetherald 1975; Knutson and Manabe 1995; Wetherald and Manabe 2002). The term
‘‘rich-get-richer mechanism’’ is appropriate for both variants, since the essence of the mechanism is to tend to enhance precipitation in regions that already have strong moisture convergence and precipitation. We will attempt to clarify the relationship between the two studies as we examine the mechanism in more detail. At its most basic, in a warmer climate, the atmospheric moisture tends to increase, governed by the Clausius–
Clapeyron expansion with relative humidity that tends to change less strongly (often approximated as con-stant). If one ignores changes in the flow, regions with climatological convergence/divergence will tend to have an increased moisture convergence/divergence, which tends to yield precipitation increases in tropical con-vergence zones and decreases in descent regions. How-ever, additional factors immediately enter. CN04 noted that moisture increases at low levels tend to reduce the gross moist stability (Yu et al. 1998) unless there is a compensating increase in the dry stability component.
Reduced gross moist stability would yield an enhanced ascent in ascending regions, providing a dynamic feed-back that can potentially increase the precipitation.
However, increased dry static energy due to warming or increased depth of convection can oppose this effect.
The lack of theory for changes in the gross moist stability
15 APRIL2009 C H O U E T A L . 1983
yields uncertainties in the dynamic convergence feed-back, which locally can be large (Chou et al. 2006).
The moisture increase is also not uniform—moisture within the convergence zones tends to increase through a deeper layer because of the convection, creating horizontal gradients of the moisture anomalies. This yields another mechanism that operates in certain regions of the convec-tive margin (Neelin et al. 2003; CN04), the ‘‘upped-ante’’
mechanism. Dry advection associated with inflow from the less-moistened subsidence regions into the convergence zones tends to suppress convection in the convective margin regions. The upped-ante mechanism is thus one means by which the margin of the convergence zone can be shifted inward (noting that because the margin has a finite width, the anomalies tend to be spread out over a region larger than the shift in a particular contour of precipitation or moisture convergence). Precipitation in the convective margins is reduced and the convergence feedback weakens the ascent, which weakens the tropical circulation. We note that the horizontal gradient of the moisture anomalies can be smoothed out slightly by hor-izontal transport, which tends to yield a more constant relative humidity (Soden et al. 2005).
CN04 emphasizes dynamic feedbacks, such as the convergence feedback, enhancing thermodynamic ef-fects because of the increase in moisture and moisture gradients. HS06 emphasizes the robust aspects of the thermodynamic effects, especially in the annual aver-age, ensemble averaver-age, and zonal average. Under the hypothesis that many of the regional discrepancies among climate models are associated with dynamic feedbacks, we use moist static energy (MSE) diagnostics informed by the hypothesized mechanisms of CN04 to examine processes affecting regional tropical precipi-tation change. We analyze 10 coupled general circula-tion model (CGCM) simulacircula-tions in order to seek con-sistent budgetary balances associated with mechanisms that induce regional change of mean tropical precipita-tion within and outside the convergence zones. Changes in rainfall distribution characteristics, though important, are not discussed here. In section 2, we briefly describe the data and the moisture and MSE budgets, clarifying the relationship between CN04 and HS06 in terms of these budgets. For reasons outlined below, the analysis carried out separately for tropical convergence zones and subsidence regions, respectively, in sections 3 and 4. The discussion and conclusions are in section 5.
2. Data and formula a. Data
Data are from the climate model simulations in the archive supported by the Program for Climate Model
Diagnosis and Intercomparison for the Fourth Assess-ment Report (AR4) of the IntergovernAssess-mental Panel on Climate Change (IPCC). The A2 scenario for anthro-pogenic emissions is used. Since the A2 scenario is at the higher end of the IPCC emission scenario, the warming is larger than other scenarios, such as the A1B scenario, but spatial patterns and mechanisms are expected to be similar no matter what scenario is used. One realization from each of the 10 models is used. The models that we chose are based on the data availability. The NCAR PCM1 and GISS-ER models Table 1 do not have all components of surface heat fluxes, so the corresponding net energy into the atmospheric column Fnetis calcu-lated only in 8 models; other variables are calcucalcu-lated in all 10 models. We define the period of 1961–90 as cur-rent climate and use the period of 2070–99 for future climate. All anomalies shown in this study are differ-ences between these two periods. The climate model simulations used here are briefly described in Table 1 and a detailed description can be found online (http://
www-pcmdi.llnl.gov/ipcc/model_documentation/ipcc_
model_documentation.php).
b. Moisture and moist static energy budgets
To understand mechanisms that induce tropical pre-cipitation anomalies, the vertically integrated moisture budget,
P9 5 hv›pq9i hv9›pqi hv $qi9 1 E9 1 residual q, (1) is diagnosed first, where () denotes climatology in 1961–90, which is defined as current climate. Here ()9 represents the departure from the current climate. The precipitation P is in energy units (W m22) which, di-vided by 28, becomes mm day21. The E is evaporation, vis pressure velocity, v is horizontal velocity, and the specific humidity q is in energy units by absorbing the latent heat per unit mass, L. Vertical integralh i denotes a mass integration through the troposphere with pTas the depth of the troposphere:
hXi 5 g1 ðpspT
ps
Xdp, (2)
where g is gravity and psis surface pressure. Because tropical convection affects layers above the convection heating (Holloway and Neelin 2007), the vertical inte-gral from 1000 to 30 hPa is used in this study, much higher than the tropopause. In (1), precipitation anomalies are roughly balanced by anomalous vertical moisture trans-port associated with mean flow (v) and anomalous flow
1984 J O U R N A L O F C L I M A T E VOLUME22
(v9), anomalous horizontal moisture transport, and evap-oration anomalies. The residual_q term includes transient and nonlinear terms, such ashv9›pq9i.
The second term on the right of (1) is a dynamical feedback associated with anomalous circulation v9. In-sight into the terms that balance v9 can be obtained from the vertically integrated MSE budget, which can be written as
hv9›phi 5 h v›ph9i hv =(q 1 T)i9 1 Fnet9
1residual h, (3)
where T is atmospheric temperature in energy units, that is, absorbing the heat capacity at constant pressure Cp. The MSE is h 5 q 1 s, and the dry static energy is s 5 T 1 f, with f being the geopotential. The net
where T is atmospheric temperature in energy units, that is, absorbing the heat capacity at constant pressure Cp. The MSE is h 5 q 1 s, and the dry static energy is s 5 T 1 f, with f being the geopotential. The net