Content uploaded by Mark W Shephard
Author content
All content in this area was uploaded by Mark W Shephard on Jan 08, 2014
Content may be subject to copyright.
eScholarship provides open access, scholarly publishing
services to the University of California and delivers a dynamic
research platform to scholars worldwide.
Lawrence Berkeley National Laboratory
Lawrence Berkeley National Laboratory
Peer Reviewed
Title:
Radiative Forcing by Long-Lived Greenhouse Gases: Calculations with the AER Radiative
Transfer Models
Author:
Iacono, Michael J.
Publication Date:
08-04-2008
Publication Info:
Lawrence Berkeley National Laboratory
Permalink:
http://escholarship.org/uc/item/83g824k8
Radiative Forcing by Long-Lived Greenhouse Gases: Calculations
with the AER Radiative Transfer Models
Michael J. Iacono, Jennifer S. Delamere, Eli J. Mlawer, Mark W. Shephard
Atmospheric and Environmental Research, Inc., Lexington, Massachusetts, USA
Shepard A. Clough
Clough Associates, Lexington, Massachusetts, USA
William D. Collins
Department of Earth and Planetary Science, University of California, Berkeley, California, USA
Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California, USA
Corresponding Author Address:
Michael J. Iacono
Atmospheric and Environmental Research, Inc.
131 Hartwell Avenue
Lexington, MA 02421 USA
E-mail: miacono@aer.com
1
Abstract1
2
A primary component of the observed, recent climate change is the radiative forcing from3
increased concentrations of long-lived greenhouse gases (LLGHGs). Effective simulation of4
anthropogenic climate change by general circulation models (GCMs) is strongly dependent on5
the accurate representation of radiative processes associated with water vapor, ozone and6
LLGHGs. In the context of the increasing application of the Atmospheric and Environmental7
Research, Inc. (AER) radiation models within the GCM community, their capability to calculate8
longwave and shortwave radiative forcing for clear sky scenarios previously examined by the9
radiative transfer model intercomparison project (RTMIP) is presented. Forcing calculations with10
the AER line-by-line (LBL) models are very consistent with the RTMIP line-by-line results in11
the longwave and shortwave. The AER broadband models, in all but one case, calculate12
longwave forcings within a range of –0.20 to 0.23 W m-2 of LBL calculations and shortwave13
forcings within a range of –0.16 to 0.38 W m-2 of LBL results. These models also perform well14
at the surface, which RTMIP identified as a level at which GCM radiation models have particular15
difficulty reproducing LBL fluxes. Heating profile perturbations calculated by the broadband16
models generally reproduce high-resolution calculations within a few hundredths K d-1 in the17
troposphere and within 0.15 K d-1 in the peak stratospheric heating near 1 hPa. In most cases, the18
AER broadband models provide radiative forcing results that are in closer agreement with high-19
resolution calculations than the GCM radiation codes examined by RTMIP, which supports the20
application of the AER models to climate change research.21
22
23
2
24
Introduction25
A primary component of recent climate change is the radiative forcing caused by changes26
in concentration of the radiatively active greenhouse gases in the atmosphere. Radiative forcing27
is “…a measure of the influence a factor has in altering the balance of incoming and outgoing28
energy in the Earth-atmosphere system and is an index of the importance of the factor as a29
potential climate change mechanism [Solomon et al., 2007]”. Between 1750 and 2005, the30
radiative forcing at the top of the atmosphere due to increases in concentration of carbon dioxide,31
methane and nitrous oxide was +1.66, +0.48, and +0.16 W m-2 respectively [Solomon et al.,32
2007]. Halocarbons such as CCl3F (CFC-11) and CCl2F2 (CFC-12) have contributed an33
additional forcing of +0.34 W m-2 since the mid-20th Century resulting in a total forcing due to34
long-lived greenhouse gases (LLGHG) of +2.64 W m-2 with a margin of uncertainty of ten35
percent. This forcing represents roughly one percent of the shortwave energy absorbed by the36
climate system and the longwave energy emitted by the Earth at the top of the atmosphere in a37
typical year, and the sign of the forcing indicates that the climate system is presently absorbing38
more energy than it releases. Accurate representation of these radiative processes in climate39
models is clearly essential to enhancing our ability to understand and to predict global climate40
change.41
42
Recent research has examined the radiative forcing calculated by several coupled43
atmosphere-ocean general circulation models (AOGCMs) that contributed to the44
Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) and by45
several reference line-by-line (LBL) radiative transfer models [Collins et al., 2006, hereinafter46
C06]. The Radiative Transfer Model Intercomparison Project (RTMIP) conducted by C0647
defined eight clear sky calculations in which the LLGHGs were specified according to various48
forcing scenarios that were performed by the AOGCMs for the AR4. This intercomparison49
concluded that while there is excellent agreement among the LBL models, there are substantial50
differences in forcing among the AOGCMs and between the AOGCMs and LBL models of order51
0.5 W m-2 or greater [Collins et al., 2006]. The largest AOGCM forcing discrepancies were seen52
at the Earth’s surface. Neglect of individual molecular absorbers (such as methane in the53
shortwave) in the AOGCMs and the radiative transfer methods utilized were both cited as54
reasons for the differences from the LBL codes.55
3
56
The RTMIP forcing results reported by C06 for the IPCC climate models do not include57
calculations with the radiation models [Clough et al., 2005] that have been developed at58
Atmospheric and Environmental Research, Inc. (AER) with support from the Department of59
Energy (DOE) Atmospheric Radiation Measurement (ARM) Program. Radiance closure studies60
are a critical feature of these models that enhances their wider use in climate applications. These61
studies provide a continual evaluation of the models with surface, aircraft and satellite62
measurements to quantify and to improve their accuracy [e.g. Turner et al., 2004]. This63
addresses the essential need to establish the accuracy of GCM radiation models through careful64
evaluation with measurements. This summary of radiative forcing results for the AER models65
following the RTMIP calculations is motivated by the wide use of the AER line-by-line model66
within the community and in particular by the increasing application of the AER correlated k-67
distribution broadband models within the GCM community. This includes the operational use of68
RRTMG_LW or SW in the ECMWF weather forecast system [Morcrette et al., 2008], the NCEP69
global forecast system (GFS) model, the Weather Research and Forecasting (WRF) model, and70
the ECHAM5 climate model [Wild and Roeckner, 2006] among other applications. RRTMG is71
also being evaluated for use in the GFDL and NCAR climate models. Details of the AER72
radiation models and the source code are available at the AER radiative transfer web site73
(www.rtweb.aer.com).74
75
Radiation Models76
LBLRTM is an accurate, efficient and highly flexible line-by-line radiative transfer77
model that continues to be extensively validated with measured atmospheric radiance spectra78
from the sub-millimeter to the ultraviolet [Clough et al., 2005; Turner et al., 2004]. LBLRTM79
includes all significant molecular absorbers and represents the effects of self-broadening and80
foreign-broadening from water vapor with the MT_CKD_v1.4 continuum model. It uses all81
parameters on the HITRAN 2004 database, and it includes the continua of carbon dioxide,82
oxygen, nitrogen, ozone as well as Rayleigh scattering extinction. The algorithmic accuracy of83
LBLRTM is 0.5 percent, and the limiting errors are generally attributable to the input line84
parameters and line shape. Integrated fluxes and heating rates are derived from LBLRTM85
radiance spectra with an independent program (RADSUM), which is also available at the AER86
4
radiative transfer web site. For the RTMIP calculations presented, LBLRTM/RADSUM utilizes87
three angles (six streams) for flux integration.88
89
CHARTS (code for high-resolution accelerated radiative transfer with scattering) is a90
monochromatic plane-parallel radiative transfer model for the line-by-line calculation of91
radiances and fluxes at a single level for thermal and solar regimes in general scattering92
atmospheres [Moncet and Clough, 1997]. The atmosphere is treated as a vertically stratified93
medium in CHARTS, and individual layers are considered homogeneous in gaseous optical94
depth and in the cloud and aerosol optical properties. Molecular optical depths for the layers are95
provided by LBLRTM. Using the adding-doubling method, CHARTS efficiently treats multiple96
scattering for reference calculations of spectral radiances at single levels. Calculations with this97
model have been validated against high-resolution spectral radiance measurements [Mlawer et98
al., 2000]. The CHARTS algorithm is applied here to the shortwave net flux forcing calculations99
at the three pressure levels considered by the RTMIP intercomparison.100
101
RRTM has been developed for both the longwave (LW) and shortwave (SW) regions as102
reference broadband radiative transfer models that closely reproduce line-by-line results [Mlawer103
et al., 1997]. Absorption coefficients for the primary and minor molecular species required for104
the correlated k-distribution method used by RRTM have been obtained from LBLRTM.105
Molecular absorbers included in RRTM are water vapor, carbon dioxide, ozone, methane, nitrous106
oxide, oxygen, nitrogen and the halocarbons in the longwave and water vapor, carbon dioxide,107
ozone, methane and oxygen in the shortwave. The water vapor continuum is based on CKD_v2.4108
in the versions of RRTM applied to this work, and molecular line parameters are based on109
HITRAN 2000 for water vapor and HITRAN 1996 for all other molecules. Extinction from110
aerosols, clouds and Rayleigh scattering are also included, and the discrete-ordinates algorithm111
DISORT [Stamnes et al., 1988] is used for multiple scattering calculations. Several optional112
cloud liquid and ice parameterizations are available that allow the specification of cloud fraction113
and cloud physical or optical properties, though the shortwave is limited to fully clear or overcast114
calculations. RRTM_LW accuracy in clear sky relative to LBLRTM for flux is 1.0 W m-2 at all115
levels, and heating rates agree to within 0.1 K d-1 in the troposphere and 0.3 K d-1 in the116
stratosphere. RRTM_SW accuracy in clear sky is 1.0 W m-2 for direct flux and 2.0 W m-2 for117
5
diffuse flux relative to line-by-line calculations. For the RTMIP calculations presented,118
RRTM_LW utilizes four angles (eight streams) and RRTM_SW/DISORT uses eight angles (16119
streams) for flux integration.120
121
With the objective of providing a radiative transfer model that can be directly applied to122
GCMs [Iacono et al., 2003; Iacono et al., 2000] with an accuracy that is traceable to ARM and123
other measurements, RRTM has been modified to produce RRTMG. The former model retains124
the highest accuracy relative to line-by-line results for single-column calculations, and the latter125
provides better efficiency with minimal loss of accuracy for GCM applications. While RRTMG126
shares the same basic physics and absorption coefficients as RRTM, it incorporates several127
modifications to improve computational efficiency, to update the code formatting for easier128
application to GCMs, and to represent sub-grid scale cloud variability. In particular, the total129
number of quadrature points (g-points) used to calculate radiances in the longwave has been130
reduced from the standard 256 in RRTM_LW (with 16 g-points in each of the 16 spectral bands)131
to 140 in RRTMG_LW (with the number of g-points in each spectral band varying from 2 to 16132
depending on the absorption in each band). In the shortwave, the number of g-points has been133
reduced from the standard 224 in RRTM_SW (with 16 in each of the 14 spectral bands) to a total134
of 112 in RRTMG_SW. In addition, DISORT has been replaced with a two-stream radiative135
transfer solver [Oreopoulos and Barker, 1999] in RRTMG_SW. Also, RRTMG has been fully136
reformatted for consistency between the longwave and shortwave and to incorporate modern137
FORTRAN90 functionality. Clear sky RRTMG_LW accuracy relative to LBLRTM for flux is138
1.5 W m-2 at all levels, and heating rates agree to within 0.2 K d-1 in the troposphere and139
generally 0.4 K d-1 in the stratosphere. RRTMG_SW accuracy in clear sky relative to140
RRTM_SW is within 3 W m-2 (about 0.3 percent) for flux at all levels, and heating rates agree to141
within 0.1 K d-1 in the troposphere and 0.35 K d-1 (about 1 percent) in the stratosphere. For the142
RTMIP calculations presented here, RRTMG_LW and RRTMG_SW utilize the single standard143
diffusivity angle (two streams) for flux integration, though RRTMG_LW also incorporates a144
small modification of the diffusivity angle in some spectral bands that varies as a function of145
total column water vapor to improve fluxes and heating rates in profiles with high water vapor146
amounts.147
148
6
Due to the complexity of representing cloud overlap in the presence of multiple149
scattering, RRTM_SW with DISORT is limited to calculations with clear or fully overcast150
conditions. This limitation has been addressed for RRTMG_LW and SW with the addition of151
McICA, the Monte-Carlo Independent Column Approximation [Barker et al., 2002; Pincus et152
al., 2003], which is a statistical technique for representing sub-grid scale cloud variability153
including cloud overlap. This method represents cloud fraction by replacing the scalar cloud154
amount with a randomly sampled binary array dimensioned on g-point. Depending on its binary155
cloud amount of zero or one, an individual g-point is treated as fully clear or fully cloudy,156
respectively. Due to the highly variable contribution of each g-point to the total absorption, this157
method introduces random noise to the cloudy calculation of radiance, but the result has been158
shown to be unbiased. This approach provides the flexibility to represent the vertical correlation159
of the clouds (i.e. cloud overlap) in some detail by imposing an assumed relation (such as160
random or maximum-random) among the binary cloud arrays across the vertical dimension.161
Cloud physical properties (ice and liquid water path and particle sizes) or cloud optical properties162
(optical depth) are also dimensioned on g-point in each layer following the arrangement of the163
binary cloud fraction arrays, and these arrays are used in the radiative transfer to calculate total164
cloudy radiance. RRTMG is especially well suited to utilize McICA due to its relatively high165
number of g-points. However, this feature is not applied in the clear cases examined here.166
167
Radiative Forcing168
The radiative transfer calculations performed with the AER models follow the eight cases169
defined by C06, in which only the concentrations of the long-lived greenhouse gases were170
varied. Temperature, water vapor and ozone were specified by the standard mid-latitude summer171
profile, and calculations were performed over 40 vertical layers from the surface to a height of172
0.01 hPa. Neither aerosols nor clouds were considered in the RTMIP calculations reported here173
or by C06. Radiative forcings were calculated as net flux differences between the C06 cases, and174
Table 1 lists the specific LLGHG concentration changes for each of the seven case differences175
examined. For example, the result for case 2a-1a is the forcing from increasing CO2 from its176
value of 287 ppmv in 1860 AD to its value of 369 ppmv in 2000 AD, while case 2b-1a provides177
the forcing from doubling CO2 from its 1860 value to 574 ppmv. Case 3b-3a shows the forcing178
from increasing CO2, CH4 and N2O from their values in 1860 to 2000 as well as increasing the179
7
CFCs from zero to their present values. Other differences for cases 3a through 3d involve various180
changes in the LLGHGs. Case 4a-2b includes the effect of increasing water vapor in the whole181
column by 20 percent in the presence of doubled CO2 as a proxy for the predicted increase in182
H2O in a warmer climate. The shortwave forcing calculations with the AER models depart from183
the C06 RTMIP specification in that the former were performed over the 820-50000 cm-1 (0.2-184
12.2 µm) spectral range with a solar constant of 1368.2 W m-2. The C06 calculations used a185
spectral range 0.2-5.0 µm and a reduced solar constant of 1360 W m-2 to accommodate this186
spectral limit in the GCM radiation models studied by C06. The contribution of the 5.0-12.2 µm187
region to the total solar forcing for the cases examined is 0.01 W m-2 or less, except for a 0.05 W188
m-2 contribution to the total forcing in case 4a-2b at the surface.189
190
While the results for the IPCC GCMs and several LBL models were presented by C06 as191
inter-model averages and standard deviations, the radiative forcings for the AER codes are192
presented here for each model individually. The longwave radiative forcing based on net flux193
calculations at the top of the atmosphere, at 200 hPa, and at the surface is listed in Table 2 for194
LBLRTM_v11.1, RRTM_LW_v3.2 and RRTMG_LW_v4.4. These can be compared with Table195
8 in C06, which summarizes the GCM and line-by-line results for the longwave models196
examined in that study. For each case and at each of the three vertical levels, the radiative197
forcing calculated by LBLRTM is within 0.06 W m-2 of the mean LBL model result of C06. In198
most cases, the LBLRTM forcing is slightly higher than the C06 mean LBL forcing. The199
exception is case 3a-1a (the forcing from increasing the CH4 and N2O concentrations from zero200
to their 1860 values), for which the LBLRTM result is slightly lower than the C06 mean LBL201
result. For all cases and at all three levels, the difference between the LBLRTM forcing and the202
mean C06 LBL result is generally less than the standard deviation of the C06 LBL forcing203
values. Radiative forcing differences at the top of the atmosphere between the AER broadband204
longwave models and LBLRTM range from -0.20 to 0.23 W m-2 depending on the case with the205
exception of case 3a-1a for which RRTM_LW and RRTMG_LW are about 0.55 W m-2 lower206
than the line-by-line result. The later difference results from known minor deficiencies in the207
methane and nitrous oxide spectroscopic treatment in the broadband models, which is being208
addressed for future versions of these models. Although these differences are consistent with the209
differences between the mean GCM and LBL results (-0.35 to 0.17 W m-2) shown by C06, it210
8
should be noted that the standard deviation of radiative forcing at the top of the atmosphere for211
the C06 GCM models varies from 0.13 to 0.82 W m-2 among the seven forcing cases. This212
suggests that some of the individual GCM model differences in the C06 study are considerably213
larger than the mean difference. At 200 hPa, the AER broadband model differences from214
LBLRTM range from -0.17 to 0.22 W m-2 in all cases except for a forcing difference of 0.35 W215
m-2 for case 3a-1a. The mean C06 GCM model forcings at this level have a range of –0.42 to216
0.06 W m-2 in difference from the mean C06 LBL result and a range in forcing standard217
deviation of 0.15 to 0.73 W m-2. At the surface, the AER broadband models perform especially218
well relative to LBLRTM with differences of 0.11 W m-2 or less in all cases, except for a219
difference of 0.37 W m-2 between RRTMG_LW and LBLRTM in case 4a-2b (the forcing from220
increasing water by 20 percent with doubled CO2). By contrast, the difference in mean forcing221
for the GCM and LBL models examined by C06 varies from –0.52 to 0.67 W m-2 at the surface222
across the seven cases. The standard deviation of surface longwave forcing among the C06 GCM223
models ranges from 0.13 to 0.87 W m-2. The impact of the water vapor continuum change224
between LBLRTM (MT_CKD_1.4) and RRTM/RRTMG (CKD_2.4) on case 4a-2b is 0.01 W m-
225
2 or less at the top of the model and at 200 hPa, and it reduces the forcing by about 0.1 W m-2 at226
the surface (though the impact is as large as 0.3 W m-2 in the middle troposphere). Finally, it is227
notable that the AER accelerated broadband model, RRTMG, performs well relative to RRTM in228
the longwave with differences generally less than 0.05 W m-2 except for case 4a-2b, where229
differences are 0.1 W m-2 at the tropopause and above and 0.37 W m-2 at the surface.230
231
The shortwave radiative forcing results are summarized in Table 3 for CHARTS_v4.03,232
RRTM_SW_v2.7 and RRTMG_SW_v3.5. The CHARTS calculations used optical depths233
derived from LBLRTM_v11.1. These results can be compared with Table 9 in C06, which234
summarizes the GCM and line-by-line results for the shortwave models examined in that study.235
For each case and at each of the three vertical levels, the radiative forcing calculated by236
CHARTS is within 0.05 W m-2 of the mean LBL model result of C06, except for case 4a-2b. In237
the latter case, CHARTS forcings are 0.08 W m-2 lower at 200 hPa and 0.37 W m-2 lower at the238
surface than the mean C06 line-by-line result. CHARTS calculations were also performed for the239
truncated 0.2-5 µm spectral range used by the RTMIP shortwave models, and for this range240
CHARTS is within 0.06 W m-2 at 200 hPa and 0.32 W m-2 at the surface of the mean C06 LBL241
9
result in case 4a-2b. At the top of the atmosphere, radiative forcing differences between the AER242
broadband models and CHARTS range from -0.10 to 0 W m-2 in all cases. Since RRTM_SW and243
RRTMG_SW include methane and nitrous oxide, which were not treated in the C06 GCM244
shortwave models, the former models provide an improved result in cases that included these245
gases. At 200 hPa, the AER broadband model forcing differences from CHARTS range from246
–0.16 to 0.04 W m-2 over all cases. The difference in the C06 mean GCM and mean LBL model247
forcings at this level range from –0.13 to 0.35 W m-2 and a forcing standard deviation range of248
0.09 to 0.28 W m-2. Finally, at the surface the AER broadband model forcing differences from249
CHARTS range from 0.02 to 0.38 W m-2 over all cases except for a 0.59 W m-2 difference in250
case 3a-1a. The mean model forcing differences between the C06 GCM and LBL results at this251
level range from –0.50 to 0.98 W m-2, and the forcing standard deviation ranges from 0.46 to252
1.40 W m-2. This illustrates the significant improvement in shortwave forcing provided by the253
AER broadband models for the C06 RTMIP cases, especially at the surface, though further254
improvements are being investigated.255
256
Heating Rates257
Heating rate profiles are also affected by changes in LLGHGs through the impact on the258
net flux divergence. For example, doubling the concentration of carbon dioxide increases259
absorption of upwelling longwave radiation and reduces the upward longwave flux at the260
tropopause by about 4 W m-2. Emission of downward longwave radiation increases by as much261
as 4 W m-2 in the middle troposphere and by about 2 W m-2 at the tropopause and surface. This262
results in a decrease in net flux divergence in the lower troposphere and an increase in longwave263
radiative heating of up to 0.1 K d-1 at these levels. The change in the longwave heating rate264
profile from doubling the CO2 concentration from its 1860 amount is shown in the left panels of265
Figure 1 for the AER line-by-line and broadband radiation models. In this and subsequent266
figures, results are shown from the surface to 0.1 hPa in the upper panels and from the267
troposphere up to 100 hPa in the lower panels. In the stratosphere, the net flux divergence is268
increased, which results in a substantial reduction in longwave heating (enhancement in269
longwave cooling) at the 1 hPa level. In general, the AER broadband models very closely270
reproduce the changes in heating rate calculated by LBLRTM throughout the column. The271
largest departures of about 10 percent occur at and just below the stratospheric peak in longwave272
10
cooling. Changes in the shortwave heating profile for the doubled CO2 forcing case are shown in273
the right panels of Figure 1 for the two AER broadband models only. Shortwave heating rate274
calculations with CHARTS for the full profile were not performed due to the excessive275
computational effort required. RRTM and RRTMG generate very similar results with the276
exception of differences of a few hundredths K d-1 just below the tropopause and near the 1 hPa277
peak in stratospheric heating. The ability of the AER broadband models to reproduce the line-by-278
line forcing in radiative heating throughout most of the column is noticeably better than the C06279
GCM radiation models. As shown in Figure 10 of C06, the GCM models studied in RTMIP280
generate results that vary considerably with some closely reproducing the line-by-line profile of281
heating rate perturbations while others oscillate around it by as much as 0.1 K d-1. Similar282
conclusions can be drawn from the other RTMIP cases.283
284
Perturbations in the heating rate profiles from increasing the LLGHG concentrations285
from their 1860 to 2000 values (case 3b-3a) as calculated by the AER models are shown in286
Figure 2. Once again, the broadband models closely reproduce the change in heating calculated287
by LBLRTM in the longwave (left panels) through most of the vertical regime. Notable288
deviations include a roughly 10 percent overestimate of the increase in longwave cooling at the 1289
hPa level by the broadband models and small differences of 0.01 K d-1 or less in the troposphere.290
In the shortwave (right panels of Figure 2), the broadband models produce nearly identical291
changes in heating throughout the profile with the largest difference of 0.01 K d-1 occurring just292
below the tropopause.293
294
Heating rate profile changes that result from increasing water vapor by 20 percent in the295
presence of doubled carbon dioxide are shown for the AER models in Figure 3. In the longwave296
(left panels), RRTM very closely reproduces the LBLRTM result in the lower troposphere, with297
RRTMG differing from the line-by-line calculation by no more than 0.05 K d-1 in the layers near298
surface. Above the middle troposphere, the two broadband models are nearly identical with299
differences from LBLRTM of a few hundredths K d-1 or less. The increase in water vapor300
enhances longwave cooling at all levels. In the shortwave, the two broadband models produce301
essentially identical results and increase shortwave heating in the troposphere by just under 0.1 K302
d-1 with generally smaller heating increases in the stratosphere. By contrast, Figure 13 in C06303
11
shows that the heating perturbations for this case among the RTMIP GCM models differ by up to304
0.1 K d-1 at various levels in the troposphere in the longwave and shortwave. Slight differences305
of a few hundredths K d-1 in heating rate are also apparent among the C06 line-by-line models. In306
the shortwave, the AER broadband heating rate calculations in Figure 3 are in excellent307
agreement with the C06 solar line-by-line model results.308
309
Conclusions310
This paper summarizes the result of producing the clear sky radiative forcing calculations311
of the Radiative Transfer Model Intercomparison Project (RTMIP) of Collins et al. [2006] with312
the AER radiation models. The wide use by the community of the line-by-line model, LBLRTM,313
and the increasing use of the broadband correlated k-distribution models RRTM and RRTMG for314
single-column and GCM applications, respectively, has motivated this comparison of the AER315
models to the line-by-line and IPCC AR4 models examined in the RTMIP experiment. Radiative316
forcing by individual long-lived greenhouse gases since the mid-18th Century is on the order of 1317
W m-2 or less. Therefore, highly accurate radiative transfer in GCMs is essential to model318
effectively the radiative contribution of LLGHGs to global climate change.319
320
In general, LBLRTM and CHARTS calculate radiative forcings that are in close321
agreement with the line-by-line models studied by C06, and in most cases the AER broadband322
models calculate forcings that are in better agreement with the line-by-line results than the mean323
of the C06 GCM radiation models. Among the seven cases and three levels studied (top of the324
atmosphere, 200 hPa, and surface), LBLRTM longwave forcings differ by 0.06 W m-2 or less325
from the mean C06 line-by-line model result, and the differences are generally less than the326
standard deviation of the latter models within each case. In all but one case, the LBLRTM327
forcing is slightly higher than the C06 mean line-by-line model forcing. In the shortwave,328
CHARTS forcings are within 0.05 W m-2 of the mean C06 line-by-line result, except for slightly329
larger differences in the case with increased water vapor.330
331
In all cases except one, the AER broadband models calculate longwave forcings that are332
within a range of –0.20 to 0.23 W m-2 of LBLRTM, and more than half of the results are within333
0.10 W m-2. The primary exception is the case in which the concentrations of methane and334
12
nitrous oxide were increased from zero to their 19th century values, in which RRTM and335
RRTMG produce 0.55 W m-2 less forcing than the line-by-line result at the top of the atmosphere336
and 0.35 W m-2 less forcing at the tropopause. Although the cause of this discrepancy in RRTM337
and RRTMG is being investigated and corrected, it should be noted that this case among those338
considered by RTMIP is the scenario that is least relevant to analyses of recent climate change.339
By contrast, the mean longwave forcing calculated by the C06 GCM models differs from the340
line-by-line calculations over a range from –0.52 to 0.67 W m-2, with the largest discrepancies at341
the surface, and about one-third of the mean results are within than 0.10 W m-2. In the shortwave,342
since the RTMIP GCM shortwave models did not include methane and nitrous oxide as noted by343
C06, the AER broadband models, which do include these gases, provide forcing calculations that344
are more comparable to the higher resolution models in most cases. In all cases except one, the345
AER broadband models calculate shortwave forcings within a range of –0.16 to 0.38 W m-2 of346
CHARTS. As in the longwave, the exception occurs for the case in which methane and nitrous347
oxide are increased from zero to 19th century values where the broadband model forcings are348
0.59 W m-2 lower than the line-by-line result. By contrast, the mean shortwave forcing calculated349
by the C06 GCM models differs from the line-by-line calculations over a range from –0.50 to350
0.98 W m-2. Finally, the surface was identified by C06 as the level at which the largest351
discrepancies occur between the RTMIP AOGCMs and LBL models, and that study noted that352
particular attention should be given to model evaluation at this level. This analysis shows that the353
AER broadband models perform especially well at the surface relative to line-by-line354
calculations in the longwave. Although some discrepancies remain in the shortwave that are355
being investigated, significant improvement is noted relative to the C06 GCM shortwave models.356
357
Changes in LLGHGs impact radiative heating profiles as well as fluxes, and the AER358
broadband models calculate heating perturbations for the RTMIP cases that are generally very359
consistent with LBL model calculations. In most cases in the longwave, differences between the360
broadband and line-by-line model heating rates are within a few hundredths K d-1 in the361
troposphere and 0.15 K d-1 in the stratospheric peak. Shortwave heating profiles were not362
calculated with the AER LBL model for this study, but the AER broadband heating rate363
perturbations are in excellent agreement with the C06 solar LBL model calculations. The AER364
radiation models undergo continual validation with measurements or evaluation against higher365
13
resolution models, and this intercomparison of radiative forcing calculations demonstrates their366
consistency, their accuracy relative to other widely used models, and their suitability for an367
extensive range of research and climate change applications.368
369
Acknowledgments370
The authors would like to thank Andrew Conley (NCAR) for providing the atmospheric profiles371
and information necessary to reproduce the RTMIP forcing experiments with the AER models.372
We also acknowledge the helpful comments of two anonymous reviewers. This research was373
supported by the Office of Biological and Environmental Research of the U.S. Department of374
Energy under Grant No. DE-FG02-92ER61549 as part of the Atmospheric Radiation375
Measurement Program. This work was also supported by the Director, Office of Science, of the376
U.S. Department of Energy under Contract No. DE-AC02-05CH11231 and by the National377
Science Foundation Center for Multiscale Modeling of Atmospheric Processes.378
379
14
References380
381
Barker, H. W., R. Pincus, and J.-J. Morcrette (2002), The Monte Carlo Independent Column382
Approximation: Application within large-scale models, paper presented at the GCSS-ARM383
Workshop on the Representation of Cloud Systems in Large-Scale Models, Kananaskis,384
Alberta, Canada, 20-24 May.385
386
Clough, S. A., M. W. Shephard, E. J. Mlawer, J. S. Delamere, M. J. Iacono, K. Cady-Pereira, S.387
Boukabara, P. D. Brown (2005), Atmospheric radiative transfer modeling: A summary of the388
AER codes, J. Quant. Spectrosc. Radiat. Transfer, 91, 233-244.389
390
Collins, W. D., et al. (2006), Radiative forcing by well-mixed greenhouse gases: Estimates from391
climate models in the Intergovernmental Panel on Climate Change (IPCC) Fourth392
Assessment Report (AR4), J. Geophys. Res., 111, D14317, doi:10.1029/2005JD006713.393
394
Iacono, M. J., J. S. Delamere, E. J. Mlawer, and S. A. Clough (2003), Evaluation of upper395
tropospheric water vapor in the NCAR community climate model (CCM3) using modeled396
and observed HIRS radiances, J. Geophys. Res., 108(D2), 4037, doi:10.1029/2002JD002539.397
398
Iacono, M. J., E. J. Mlawer, S. A. Clough, and J.-J. Morcrette (2000), Impact of an improved399
longwave radiation model, RRTM, on the energy budget and thermodynamic properties of400
the NCAR community climate model, CCM3, J. Geophys. Res., 105, 14,873-14,890.401
402
Mlawer, E. J., P. D. Brown, S. A. Clough, L. C. Harrison, J. J. Michalsky, P. W. Kiedron, and T.403
Shippert (2000), Comparison of spectral direct and diffuse solar irradiance measurements and404
calculations for cloud-free conditions, Geophys. Res. Lett., 27, 2653-2656.405
406
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough (1997), Radiative407
transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the408
longwave, J. Geophys. Res., 102, 16,663-16,682.409
410
Moncet, J.-L., and S. A. Clough (1997), Accelerated monochromatic radiative transfer for411
scattering atmospheres: Application of a new model to spectral radiance observations, J.412
Geophys. Res., 102, 21,853-21,866.413
414
Morcrette, J.-J., H. W. Barker, J. N. S. Cole, M. J. Iacono, and R. Pincus (2008), Impact of a new415
radiation package, McRad, in the ECMWF Integrated Forecasting System, Mon. Weather416
Rev., in press.417
418
Oreopoulos, L., and H. W. Barker (1999), Accounting for subgrid-scale cloud variability in a419
multi-layer 1-D solar radiative transfer algorithm, Q. J. R. Meteorol. Soc., 125, 301-330.420
421
Pincus, R., H. W. Barker, and J.-J. Morcrette (2003), A fast, flexible, approximate technique for422
computing radiative transfer in inhomogeneous cloud fields, J. Geophys. Res., 108(D13),423
4376, doi:10.1029/2002JD003322.424
425
15
Solomon, S., et al. (2007), Technical Summary, In: Climate Change 2007: The Physical Science426
Basis. Contribution of Working Group I to the Fourth Assessment Report of the427
Intergovernmental Panel on Climate Change, edited by S. Solomon, D. Qin, M. Manning, Z.428
Chen, M. Marquis, K. B. Averyt, M. Tignor and H. L. Miller, Cambridge Univ. Press,429
Cambridge, United Kingdom and New York.430
431
Stamnes, K., S. C. Tsay, W. Wiscombe, and K. Jayaweera (1988), A numerically stable432
algorithm for discrete-ordinate-method radiative transfer in scattering and emitting layered433
media, Appl. Opt., 27, 2502-2509.434
435
Turner, D. D., et al. (2004): The QME AERI LBLRTM: A closure experiment for downwelling436
high spectral resolution infrared radiance, J. Atmos. Sci., 61, 2657–2675.437
438
Wild, M., and E. Roeckner (2006), Radiative fluxes in the ECHAM5 general circulation model,439
J. Climate, 19, 3792-3809.440
441
442
443
444
16
Figure captions445
446
Figure 1. Heating rate profile changes from doubling the CO2 concentration from 287 to 574447
ppmv using the standard mid-latitude summer profile for the longwave (left) and shortwave448
(right) from the surface to 0.1 hPa (top) and the troposphere (bottom) as calculated by the AER449
radiation models.450
451
Figure 2. Heating rate profile changes from increasing the concentrations of CO2, CH4, N2O,452
CFC-11 and CFC-12 from 1860 to 2000 values using the standard mid-latitude summer profile453
for the longwave (left) and shortwave (right) from the surface to 0.1 hPa (top) and the454
troposphere (bottom) as calculated by the AER radiation models.455
456
Figure 3. Heating rate profile changes from increasing the water vapor concentration in the457
standard mid-latitude summer profile in all layers by 20 percent with doubled CO2 (574 ppmv) in458
all calculations for the longwave (left) and shortwave (right) from the surface to 0.1 hPa (top)459
and the troposphere (bottom) as calculated by the AER radiation models.460
461
462
463
464
465
466
467
17
468
Table 1. Changes in Atmospheric Constituents for the Radiative Forcing Calculations [from469
Collins et al., 2006]470
471
Case
CO2 (ppmv)
CH4 (ppbv)
N2O (ppbv)
CFC-11 (pptv)
CFC-12 (pptv)
H2Oa
2a-1a
287 → 369
---
---
---
---
---
2b-1a
287 → 574
---
---
---
---
---
3b-3a
287 → 369
806 → 1760
275 → 316
0 → 267
0 → 535
---
3a-1a
---
0 → 806
0 → 275
---
---
---
3b-3c
---
---
275 → 316
0 → 267
0 → 535
---
3b-3d
---
806 → 1760
---
0 → 267
0 → 535
---
4a-2b
---
---
---
---
---
1.0 → 1.2
472
a Value listed for H2O is the change in the multiplier applied to the water vapor mixing ratio in473
the reference MLS profile.474
18
475
Table 2. Longwave Radiative Forcinga
476
477
478
a Units are in W m-2.479
480
481
Forcing Cases
Level
Field
2a-1a
2b-1a
3b-3a
3a-1a
3b-3c
3b-3d
4a-2b
TOM
F RRTMG_LW
1.10
3.07
2.07
3.07
0.56
0.78
3.77
TOM
F RRTM_LW
1.10
3.05
2.09
3.04
0.58
0.80
3.87
TOM
F LBLRTM
1.03
2.84
2.15
3.60
0.54
0.98
3.79
200 hPa
F RRTMG_LW
2.04
5.74
2.98
3.10
0.47
0.76
4.63
200 hPa
F RRTM_LW
2.05
5.76
3.01
3.10
0.49
0.78
4.72
200 hPa
F LBLRTM
1.98
5.54
3.06
3.45
0.46
0.93
4.52
Surface
F RRTMG_LW
0.57
1.79
1.05
1.15
0.35
0.42
11.92
Surface
F RRTM_LW
0.56
1.73
1.00
1.12
0.33
0.39
11.55
Surface
F LBLRTM
0.57
1.68
1.10
1.08
0.31
0.48
11.55
19
482
Table 3. Shortwave Radiative Forcinga
483
484
485
a Units are in W m-2.486
487
488
489
490
491
492
Forcing Cases
Level
Field
2a-1a
2b-1a
3b-3a
3a-1a
3b-3c
3b-3d
4a-2b
TOM
F RRTMG_SW
0.02
0.06
0.07
0.05
0.00
0.05
0.75
TOM
F RRTM_SW
0.02
0.07
0.07
0.04
0.00
0.05
0.75
TOM
F CHARTS
0.04
0.12
0.13
0.14
0.00
0.08
0.77
200 hPa
F RRTMG_SW
-0.31
-0.92
-0.52
-0.29
0.00
-0.21
0.47
200 hPa
F RRTM_SW
-0.26
-0.77
-0.42
-0.24
0.00
-0.16
0.47
200 hPa
F CHARTS
-0.26
-0.76
-0.44
-0.40
-0.02
-0.16
0.43
Surface
F RRTMG_SW
-0.21
-0.57
-0.53
-0.31
0.00
-0.32
-6.14
Surface
F RRTM_SW
-0.21
-0.59
-0.54
-0.31
0.00
-0.33
-6.19
Surface
F CHARTS
-0.31
-0.95
-0.87
-0.90
-0.02
-0.54
-6.24
20
493
494
Figure 1. Heating rate profile differences from doubling the CO2 concentration from 287 to 574 ppmv using the495
standard mid-latitude summer profile for the longwave (left) and shortwave (right) as calculated by the AER496
radiation models.497
498
499
21
500
501
Figure 2. Heating rate profile differences from increasing the concentrations of CO2, CH4, N2O, CFC-11 and CFC-502
12 from 1860 to 2000 values using the standard mid-latitude summer profile for the longwave (left) and shortwave503
(right) as calculated by the AER radiation models.504
22
505
506
Figure 3. Heating rate profile differences from increasing the water vapor concentration in the standard mid-latitude507
summer profile in all layers by 20 percent with doubled CO2 (574 ppmv) in all calculations for the longwave (left)508
and shortwave (right) as calculated by the AER radiation models.509
510
