Semi-direct effect of biomass burning on cloud and rainfall over Amazon

Yan Zhang, Hongbin Yu, Rong Fu & Robert E. Dickinson

School of Earth & Atmospheric Sciences,Georgia Institute of Technology

??? III LBA Scientific Conference, July 27-29, 2004, Brasilia, Brazil

Direct effect of biomass burning aerosols

Smoke

Atmospheric Heating

• Reduce surface solar flux, cooling surface

• Absorb solar radiation in atmosphere, warming in the smoke layer

• Net effect: stabilize atmosphere

Indirect effect( through cloud microphysics)

• More aerosols more but smaller cloud droplets in clouds for a fixed water content More reflective clouds

• Suppress or delay warm rain in shallow clouds prolong clouds and promote deep/stronger convection.

Semi-direct effect:Surface solar flux :– Surface SH shallower

ABL

– LH RHABL Previously believe: suppress

convection.

However, q+WE RHABL in early afternoon may not suppress convection

The sign of semi-direct effect is uncertain

morning noon

SH LH

q+WE ,Dry air

entrainment

hei

ght

Atmospheric Boundary layer

Without aerosols

With aerosols

• What controls the semi-direct? How does semi-direct effect affect rainfall?– Much of the previous studies

have focused on indirect effect (e.g., Kaufman and Fraser, 1997; Rosenfeld, 1999; Feingold et al., 2001).

– However, the semi-direct effect can be more important (e.g., Koren et al., 2004)

Questions:

MODIS, Sept. 2002, warm clouds, southerly V-index

Warm cloud fraction decreases and cloud top become lower with increase of aerosol optical depth.

Factors that could influence the semi-direct effect:

• Soil moisture, Yu et al. 2002;

• Surface cooling (scatter) vs. atmospheric warming (absorption);

• Vertical distribution of smoke layer, Yu et al., 2002;

• Structure, diurnal development of atmospheric boundary layer.

NCAR Regional Climate Model

Aerosol forcing• Why Regional Climate Model?– Resolution: high enough to resolve

Andes– Domain: large enough to include

moisture transport from ocean to Amazon.

• RegCM consists of: Atmosphere : MM4 Land : BATS Radiation : CCM3

• Domain: 20W~~100W, 35S~~25N • Time simulated : Jun ~~ Oct , 2001.• Data used:

Landuse and Topography

USGS GTOPO30 _10 MIN data

Initial and Boundary Conditions

Reanalysis Data: NCEP Data

Derived from an integration of MODIS retrievals and GOCART simulations (Yu et al., 2003)

Experiment design:

CONTROL run: without smoke aerosol forcing

AER runs: Using MODIS + GOGART (Aug-Oct. 2001)Assumptions on the thickness of the smoke layer: – AER1: 1.5 km from the

surface– AER2: 3.5 km from the

surface (Reid et al. 1998).Same aerosol loading for AER1 & AER2.

For Aug-Oct. 1993 & 2002, respectively.

Observation from SMOCC field experiment, Sept-Nov. 2002, Andreae et al., 2004, Science.

Table 1 Comparisons ABRACOS observations and RegCM3 simulations with root depths (d) of 1.5 m and 3 m, respectively. The observations were conducted from 29 Jun to 5 July, 1993. The RegCM3 model simulation is July average for 1993.

  Rn(Wm-2)

SH(Wm-2)

LH(Wm-2)

ABRACOS Observation 137.7 23.1 113

RegCM3 (d = 1.5 m) 125.8 68.8 56

RegCM3 (d = 3 m) 124.6 32 92 

The influence of aerosols peaks in late morning (10 am LT), instead of noon when incoming solar radiation peaks—change of clouds

Diurnal cycles of the net downward surface solar flux:

Without change in cloud

Aerosol forcing

Change in surface solar flux

Sept 2002

Influences on surface net radiation, latent and sensible fluxes:

Biomass burning aerosols result in stronger reduction in surface SH.

0 0.1 0.2 0.3 0.4 0.5 0.6

150

250

350

450

550

650

750

850

950

1050

Cloud Fraction

Pre

ssur

e (

mb)

08 LST

CONTAERAER2

0 0.1 0.2 0.3 0.4 0.5 0.6

150

250

350

450

550

650

750

850

950

1050

Cloud Fraction

Pre

ssur

e (

mb)

11 LST

CONTAERAER2

0 0.1 0.2 0.3 0.4 0.5 0.6

150

250

350

450

550

650

750

850

950

1050

Cloud Fraction

Pre

ssur

e (

mb)

17 LST

CONTAERAER2

-15 -10 -5 0 5 10 15

150

250

350

450

550

650

750

850

950

1050

Cloud Fraction (%)

Pre

ssur

e (

mb)

08,11,17 LST

08 LST Diff108 LST Diff211 LST Diff111 LST Diff217 LST Diff117 LST Diff2

Cloud burning

Pre

ssu

re (

hP

a)

Top of ABLDirect effect: AerosolsSHZABL

Cloud Fraction (%)

Cloud burning can largely compensate the direct effect of aerosols on depth of ABL in early afternoon.

Obs

erva

tion

s

Model results

Biomass burning aerosols higher RHABL (model&observation).Presumably, weaker diurnal growth of ABL weaker entrainment of dry air into ABL higher RH in ABL.

The direct and semi-direct effects of aerosols on total rainfall

aerosols, 1.5 km

Without aerosols

aerosols, 3.5 km

Summary:

morning noon

SH LH

q+WE ,Dry air

entrainment

RHABL can increase, instead of decrease.

Without cloud burning

Without aerosols

Cloud burning downward solar flux compensates the direct aerosol effect

What have we learned from this study?

• The diurnal growth (esp. the entrainment) of the ABL plays a key role in determining semi-direct effect of biomass burning aerosols.

• Cloud burning can reduce the surface cooling in early afternoon, thus compensate the negative direct aerosol effect on rainfall.

• The semi-direct effect of biomass burning aerosols is highly sensitive to the vertical distribution of aerosols.

Leticia:specific & relative humidity

Manaus:specific & relative humidity

NCAR Regional Climate Model

Model domain and topography (m)

• Why Regional Climate Model?– Resolution is high enough to resolve Andes

and domain large enough to include all the key processes, especially moisture transport from ocean to Amazon.

• About the model runs:• RegCM consists of: Atmosphere : MM4 Land : BATS Radiation : CCM3• Domain selected: 20W~~100W, 35S~~25N • Time simulated : Jun ~~ Oct , 1993• Data used:

Landuse and Topography

USGS GTOPO30 _10 MIN data

Rotated Mercator: Suitable for most latitudes

Initial and Boundary Conditions

Reanalysis Data: NCEP Data

Map Projection :

Distribution of aerosol optical depth at 550nm for different months in 2001. It was derived from an integration of MODIS retrievals and GOCART simulations (Yu et al., 2003)

used in our RegCM simulations: