Analysis of climate trends in tropical rainfall C.-H. Sui - PowerPoint PPT Presentation

Analysis of climate trends in tropical rainfall C.-H. Sui 隋中興 隋中興 隋中興 隋中興 J.-J. Sun 孫既仁 孫既仁 孫既仁 孫既仁 Dept. of Atmospheric Sci., National Taiwan Univ Peng Liu 刘鹏 刘鹏 刘鹏 刘鹏 Nanjing Univ. of Information Science and Technology

Observed climate variations Analysis of climate trends in tropical rainfall (forced component) and (free component) (by GHS, aerosols, SW), (due to coupled climate processes) The two components can be further divided as thermodynamically dynamically induced induced Forced Dynamical residual Contribution Free I. Global-Scale Features of Multidecadal Trends II. Climate Trends in Rainfall

EOF (Low-pass filtered, detrended SSTA) 1 st mode 2 nd mode HadISST HadISST 2 nd mode 1 st mode ERSST ERSST Pacific Decadal Oscillation Atlantic Multidecadal Oscillation PDO ( ~ 55 yr , ~ 20 yr) AMO (~ 65 yr) Primarily a Pacific oscillation Primarily an Atlantic oscillation 60 yrs 60 yrs 1 20 yrs Fig. 3 1890 1910 1930 1950 1970 1990 1890 1910 1930 1950 1970 1990

Combined EOF WOD, NCEP (U, V, w , Z, T, Q) at 8 pre. levels ERA40

Linear trend 1980-2009 (K/30yr) a

Regressed T 2m on the 1 st PC (the PDO-like) from HadISST EOF analyses PDO phase Walker circulation strengthening +tive ↘ -- Regressed T 2m on the 2 nd PC (the AMO-like) from HadISST EOF a analyses Hadley cell strengthening ↘ Northward Atm AMO phase & Northward shift of ITCZ Heat Transport -- ↗ +tive S N ↗ Northward Ocean Heat Transport

stream func./NCEP Contour (1980-2007 mean) Shadings (2007-1994) minus (1993-1980) [unit:10 10 kg/s]

Liu and Sui (2013) ST gm GISS [SAT GHCN, HadISST1 1880-Nov 1981, NOAA OISSTv2, Dec 1981-present]

The Partial Least Squared (PLS) Regression Analysis Predictors: X ij ( j denotes grids in space) Predictant: Y i ( i denotes time) The PLS-regression method seeks to predict Y with the predictors (a linear combinations of X) and to maximize (1) variance in Y , and (2) correlation between X and Y

Centennial trends in original (R 0 ) & PLS-regressed (R PLSR ) rainfall time series at four stations in Taiwan [ ERSST (1900~2011) is used in PLS regression ] Taipei Tainan Hentsung Taidon Mean 5.10 5.28 6.63 5.42 (mm/d) R 0 12.6 -14.2 -28.6 -13.7 (% /100yr) R PLSR 7.4 0.1 0.1 -3.5 (% /100yr)

SSM/I rainfall data for the period of 1988-2009 Mean rainfall Trend in “ rich → richer, Rainfall amount poor → poorer” (mm/day per decade) Trend in overall increasing Rainfall Intensity tendency (% per decade)

Trends in rainfall amount over tropical oceans 1988-2009 SSMI GPCP SSMI Trends in original data (R o ) Trends in original data (R o ) SSMI GPCP E s +( ∇• ∇• qv) b E s ( ∇• ∇• qv) b ∇• ∇• ∇• ∇• Mean 3.62 2.92 3.72 4.49 -0.77 (mm/d) trend original 1.9 0.53 2.47 2.38 1.9 (% /decade) trend PLSR -0.8 -0.73 -0.1 -0.23 -1.25 (% /decade) Trends in PLS regressed data (R o ) Trends in PLS regressed data (R o )

Decadal trends in rainfall amount over tropical oceans (30 o S-30 o N)

� � � � = ⋅ ∆ = ⋅ ∆ + ∆ ⋅ p p n dr p p n dr p n dr r r r r r r � � � � = ⋅ ∆ = ⋅ ∆ + ∆ ⋅ p p n dw p p n dw p n dw w w w w w w (1995-2010) − (1979-1994)

GPCP: (1995-2010) − − − − (1979-1994) SSMI V7: (2000-2011) − − − − (1988-199 20N 10N 0 10S 20S

(1995-2010) − − (1979-1994) − −

Conclusions - I Global-Scale Features of Multidecadal Trends Two dominant modes : PDO/IPO and AMO. PDO-like oscillation : ENSO-like and hemispheric symmetric AMO : hemispheric asymmetric In recent decades since 1980 The two oscillations contribute significantly to the increasing trends in ST gm [almost half of the warming ] (Liu and Sui 2013)

Conclusions - II In recent years 1988-2009 Decadal oscillations contribute significantly to an increasing general circulation. 1. Trends in PLS regression analysis: “dynamic part of rain amount ” major cause of “rich → → → richer” → “residual rain amount ” weakening 2. Trends in w-rain combined analysis: “thermodynamic part of heavy rain” overall increasing Caution: reliability limited by the short record of data

As the climate warms: Mass exchange between the boundary layer & free troposphere − 1 δ = δ + αδ α ≡ ≈ P / P M / M T , d ln e /dT 0 . 07 K s Held and Soden (2006) radiative cooling ↗ → subsidence ↗ ∆ ∆ R W Stephens and Ellis (2008) ≈ b ( b ~ 0 . 6 ) R W Good global constraints , How about tropics where deep and shallow clouds respond to climate forcing differently ? Obs. Rainfall record , short, mix with multiscale oscillations

CRM : The Physic of Precipitation － 300K --- 305K CWV BWV BWV 8%/K 5%/K~1%/0.2K~1%/decade 4%/K 99.9% 0 99.99% The earth’s surface temperature warmed by 0.19 ± 0.04K/decade according to the Global Historical Climatology Network between 1987-2006 (C. J. Muller & P. A. O’Gorman 2011)

More Precipitation Extreme? (D. M. Romps , 2010)

Time series of boreal summer (MJJAS) mean indices (mm·d −1 ) averaged over the entire NH monsoon rainfall domain (r = 0.85) (U 850 -U 200 ) ave (0 ° –20 ° N, 120 ° W–120 ° E) Max. |cross-equatorial, zonal mean meridional mass stream function (10 10 kg·s −1 )| U850 averaged over the equatorial Pacific (10 ° S–10 ° N, 140 ° E–120 ° W) . Wang et al. 2013 PNAS ERAI and GPCP, version 2.2, 1979–2011

Detrended SST by 30S-30N mean trend Domain: 30S-30N 30S-30N Mean: 3.6175(mm/day)

2.92 0.53 % -0.73 % (10y) -1 3.62 1.9 % -0.8 % (10y) -1 (a) Time series of the averaged values over tropical oceans (30 o S-30 o N) for the original GPCP rainfall (R0, red), and the residual rainfall after a partial least squared regression analysis (PLSRA) applied to SST and rainfall fields to suppress the natural oscillations in precipitation data (after the first and second PLSRA, R1 in green and R2 in blue). The dashed lines indicate the corresponding linear trends. (b) The same as (a) except for SSM/I Units in mm d -1 .

GPCP Q bdy EVP Qconv mean 2.92 3.72 4.49 -0.77 R0 0.53 2.47 2.38 1.9 R1 0.29 0.39 0.36 -1.1 R2 -0.73 -0.1 -0.23 -1.25 % (10y) -1

Decadal trends in rainfall amount over tropical oceans SSMI GPCP SSMI Trends in original data (R o ) Trends in original data (R o ) SSMI GPCP E s +( ∇• ∇• qv) b E s ( ∇• ∇• qv) b ∇• ∇• ∇• ∇• Mean 3.62 2.92 3.72 4.49 -0.77 (mm/d) trend original 1.9 0.53 2.47 2.38 1.9 (% /decade) trend PLSR -0.8 -0.73 -0.1 -0.23 -1.25 (% /decade) Trends in PLS regressed data (R o ) Trends in PLS regressed data (R o )

1.9 0.53 %DY

GPCP: (1995-2010) − − − − (1979-1994) SSMI V7: (2000-2011) − − − − (1988-199 20N 10N 0 10S 20S