Exploiting observations to seek robust responses in global precipitation

Richard P. AllanDepartment of Meteorology, University of Reading

Thanks to Brian Soden, Viju John, William Ingram, Peter Good, Igor Zveryaev and Mark Ringer

http://www.met.reading.ac.uk/~sgs02rpa r.p.allan@reading.ac.uk

Introduction

“Observational records and climate projections provide abundant evidence that freshwater resources are vulnerable and have the potential to be strongly impacted by climate change, with wide-ranging consequences for human societies and ecosystems.” IPCC (2008) Climate Change and Water

How should the water cycle respond to climate change?

See discussion in: Allen & Ingram (2002) Nature; Trenberth et al. (2003) BAMS

Hawkins and Sutton (2010) Clim. Dyn.

• Increased Precipitation• More Intense Rainfall• More droughts• Wet regions get wetter, dry

regions get drier?• Regional projections??

Precipitation Change (%)

Climate model projections (IPCC 2007)

Precipitation Intensity

Dry Days

Pre

cip.

(%

)

Allan and Soden (2008) Science

Can we use observations to confirm robust responses?

Tropical ocean precipitation

• dP/dSST:

GPCP: 10%/K (1988-2008)

AMIP: 3-11 %/K (1979-2001)

• dP/dt trend

GPCP: 1%/dec(1988-2008)

AMIP: 0.4-0.7%/dec(1979-2001)

(land+ocean)

SSM/I GPCP

Allan et al. (2010) Environ. Res. Lett.

NCAS-Climate Talk 15th January 2010 Trenberth et al. (2009) BAMS

Physical basis: energy balance

Surface Temperature (K)

Lambert & Webb (2008) GRL

Late

nt H

eat

Rel

ease

, LΔ

P (

Wm

-2)

Rad

iativ

e co

olin

g, c

lear

(W

m-2)

Models simulate robust response of clear-sky radiation to warming (~2 Wm-2K-1) and a resulting increase in precipitation to balance (~2 %K-1)

e.g. Allen and Ingram (2002) Nature, Stephens & Ellis (2008) J. Clim

NCAS-Climate Talk 15th January 2010

Rad

iativ

e co

olin

g, c

lear

(W

m-2K

-1)

Allan (2009) J. Clim

Models simulate robust response of clear-sky radiation to warming (~2 Wm-2K-1) and a resulting increase in precipitation to balance (~2 %K-1)

e.g. Allen and Ingram (2002) Nature, Stephens & Ellis (2008) J. Clim

Trends in clear-sky radiation in coupled models

Clear-sky shortwave absorptionSurface net clear-sky longwave

Allan (2009) J. Clim

Can we observe atmospheric radiative heating/cooling?

John et al. (2009) GRL

See also discussion in Trenberth and Fasullo (2010) Science

Surface net longwave and water vapour

Allan (2009) J . Climate

ER

A40

N

CE

P

SR

B

SS

M/I

• Surface net longwave strongly dependent on column water vapour

• Increased water vapour enhances ability of atmosphere to cool to the surface

Current changes in tropical ocean column water vapour

Changing observing systems applied to reanalyses cause spurious variability.

John et al. (2009)

models

Wat

er V

apou

r (m

m)

Is the mean state important?

• Models appear to overestimate water vapour– Pierce et al. (2006) GRL;

John and Soden (2006) GRL– But not for microwave data?

[Brogniez and Pierrehumbert (2007) GRL]

• This does not appear to affect feedback strength– John and Soden (2006)

• What about the hydrological cycle?– Symptomatic of inaccurate

simulation?

Pierce et al. (2006)

GRL

Does low-level moisture rise at 7%/K?Specific humidity trend correlation (left) and time series (right)

Willett et al. (2007) Nature

Robust relationships globally.

Less coherent relationships regionally/over land/at higher altitudes?

Evidence for reductions in RH over land (Simmons et al. 2009 JGR) which are physically plausible.

Land OceanWillett et al. (2008) J Clim

NCAS-Climate Talk 15th January 2010

CC Wind Ts-To RHo

Muted Evaporation changes in models are explained by small changes in Boundary Layer:1) declining wind stress2) reduced surface temperature lapse rate (Ts-To)3) increased surface relative humidity (RHo)

Richter and Xie (2008) JGR

Evaporation

Physical Basis: Moisture Transport

Cha

nge

in M

oist

ure

Tra

nspo

rt,

dF (

pg/d

ay)

If the flow field remains relatively constant, the moisture transport scales with low-level moisture.Model simulation

scaling

Held and Soden (2006) J Climate

Projected (top) and estimated (bottom)

changes in Precipitation minus Evaporation d(P-E)

Held and Soden (2006) J Climate

~

First argument:P ~ Mq.

So if P constrained to rise more slowly than q, this implies reduced M

Second argument:ω=Q/σ.

Subsidence (ω) induced by radiative cooling (Q) but the magnitude of ω depends on (Гd-Г) or static stability (σ).

If Г follows MALR increased σ. This offsets Q effect on ω.See Held & Soden (2006) and Zelinka & Hartmann (2010) JGR

P~Mq

Physical Basis: Circulation response

Models/observations achieve muted precipitation response by reducing strength of Walker circulation. Vecchi and Soden (2006) Nature

P~Mq

Physical Basis: Circulation response

• Evidence for recent increased strength of tropical Hadley/Walker circulation since 1979?– Sohn and Park (2010) JGR

Walker circulation index (top) and sea level pressure anomalies (bottom) over equatorial Pacific (1948-2007)

Hadley circulation index over 15oS-30oN band

Extreme PrecipitationPhysical basis: water vapour

1979-2002• Clausius-Clapeyron– Low-level water vapour (~7%/K)– Intensification of rainfall: Trenberth et al.

(2003) BAMS; Pall et al. (2007) Clim Dyn

• Changes in intense rainfall also constrained by moist adiabat -O’Gorman and Schneider (2009) PNAS

• Does extra latent heat release within storms enhance rainfall intensity above Clausius Clapeyron?– e.g. Lenderink and van Meijgaard (2010)

Environ. Res. Lett.; Haerter et al. (2010) GRL

Changes in Extreme Precipitation Determined by changes in low-level water vapour and updraft velocity

Above: O’Gorman & Schneider (2008) J Clim

Aqua planet experiment shows extreme precipitation rises with surface q, a lower rate than column water vapour

Right: Gastineau and Soden (2009) GRL Reduced frequency of upward motion

offsets extreme precipitation increases.

• Analyse daily rainfall over tropical oceans– SSM/I v6 satellite data, 1988-2008 (F08/11/13)– Climate model data (AMIP experiments)

• Create rainfall frequency distributions

• Calculate changes in the frequency of events in each intensity bin

• Does frequency of most intense rainfall rise with atmospheric warming?

Precipitation Extremes

Trends in tropical wet region precipitation appear robust.

– What about extreme precipitation events?

METHOD

Increases in the frequency of the heaviest rainfall with warming: daily data from models and microwave satellite data (SSM/I)

Reduced frequency Increased frequencyAllan et al. (2010) Environ. Res. Lett.

• Increase in intense rainfall with tropical ocean warming (close to Clausius Clapeyron)

• SSM/I satellite observations at upper range of model range

Turner and Slingo (2009) ASL: dependence on convection scheme?

Observational evidence of changes in intensity/duration (Zolina et al. 2010 GRL)

Links to physical mechanisms/relationships required (Haerter et al. 2010 GRL)

Contrasting precipitation response expected

Pre

cipi

tatio

n Heavy rain follows moisture (~7%/K)

Mean Precipitation linked to

radiation balance (~3%/K)

Light Precipitation (-?%/K)

Temperature e.g.Held & Soden (2006) J. Clim; Trenberth et al. (2003) BAMS; Allen & Ingram (2002) Nature

Models ΔP [IPCC 2007 WGI]

Is there a contrasting precipitation response in wet and dry regions?

Rainy season: wetter

Dry season: drier Chou et al. (2007) GRL

Precip trends, 0-30oN

The Rich Get Richer?

Zhang et al. 2007 Nature

Detection of zonal trends

Contrasting wet/dry precipitation responses

• Large uncertainty in magnitude of change: satellite datasets and models & time period

TRMM

GPCP Ascent Region Precipitation (mm/day)

John et al. (2009) GRL

• Robust response: wet regions become wetter at the expense of dry regions. Is this an artefact of the reanalyses?

Contrasting precipitation response in wet and dry regions of the tropical circulation

Allan et al. (2010) Environ. Res. Lett.

descent

ascentModelsObservations

Pre

cipi

tatio

n ch

ange

(%

)

Sensitivity to reanalysis dataset used to define wet/dry regions

• Sample grid boxes:– 30% wettest– 70% driest

• Do wet/dry trends remain?

Avoid reanalyses in defining wet/dry regions

Allan et al. (2010) Environ. Res. Lett.

Current trends in wet/dry regions of tropical oceans

• Wet/dry trends remain– 1979-1987 GPCP

record may be suspect for dry region

– SSM/I dry region record: inhomogeneity 2000/01?

• GPCP trends 1988-2008

– Wet: 1.8%/decade– Dry: -2.6%/decade– Upper range of model

trend magnitudes

Models

DR

Y

WE

T

Allan et al. (2010) Environ. Res. Lett.

Binning by regime(a) P (mm/day); % area (b) Model–GPCP P (%) (c) 2080-99 – 1980-99 P (%/K); (d) ω

Vert

ical m

oti

on

(ω

)

perc

en

tile

s

ascen

t

descen

t

Precipitation binned in percentiles of vertical motion (0-5% bin is strongest ascent) and temperature (95-100% bin is warmest) for the HadGEM1 model (top) and an ensemble of 10 CMIP3 models (bottom). (a) Mean precipitation and % area enclosed in each contour, (b) model – GPCP precipitation and model change in (c) precipitation (scaled by temperature change) and (d) vertical motion for 2080-99 minus 1980-99. e.g. see also Emori and Brown (2005) GRL

Andrews et al. (2009) J Climate

Transient responses

Transient responses• CO2 forcing experiments

• Initial precip response supressed by CO2 forcing

• Stronger response after CO2 rampdown

HadCM3: Wu et al. (2010) GRL

CMIP3 coupled model ensemble mean: Andrews et al. (2010) Environ. Res. Lett.

Degree of hysteresis determined by forcing related fast responses and linked to ocean heat uptake

Forcing related fast responses

Andrews et al. (2010) GRL

Total Slow

• Surface/Atmospheric forcing determines “fast” precipitation response

• Robust slow response to T• Mechanisms described in Dong

et al. (2009) J. Clim

• CO2 physiological effect potentially substantial (Andrews et al. 2010 Clim. Dyn.; Dong et al. 2009 J. Clim)

• Hydrological Forcing:

HF=kdT-dAA-dSH

(Ming et al. 2010 GRL; also Andrews et al. 2010 GRL)

Pre

cip

itatio

n re

spo

nse

(%

/K)

Can NWP help? e.g. Sean Milton/PAGODA

Outstanding issues

• Satellite estimates of precipitation, evaporation and surface flux variation are not reliable.

• Are regional changes in the water cycle, down to catchment scale, predictable?

• How well do models represent land surface and physiological feedbacks.

• How is the water cycle responding to aerosols?• Linking water cycle and cloud feedback issues

• Robust Responses– Low level moisture; clear-sky radiation

– Mean and Intense rainfall

– Observed precipitation response at upper end of model range?

– Contrasting wet/dry region responses

• Less Robust/Discrepancies– Moisture at upper levels/over land and mean state

– Inaccurate precipitation frequency/intensity distributions

– Magnitude of change in precipitation from satellite datasets/models

• Further work– Decadal changes in global energy budget, aerosol forcing effects

and cloud feedbacks: links to water cycle?

– Precipitation and radiation balance datasets: forward modelling

– Separating forcing-related fast responses from slow SST response

– Boundary layer changes and surface fluxes

Conclusions