U.S. Dept. of Commerce / NOAA
/ OAR / ERL
/ GFDL 
*Disclaimer
U.S. Dept. of Commerce / NOAA
/ OAR / ERL
/ GFDL
*Disclaimer
To construct mathematical models of the atmosphere and of the coupled ocean-atmosphere system which simulate the global large-scale features of climate.
To study the dynamical interaction between large-scale wave disturbances and the general circulation of the atmosphere.
To identify and elucidate the physical and dynamical mechanisms which maintain climate and cause its variation, and to examine their generality in the context of paleoclimate and the atmospheres of other planets.
1.1 OCEAN-ATMOSPHERE
INTERACTION
1.1.1
Response to CO2 and Sulfate Aerosols
J. Haywood R.J. Stouffer
S. Manabe R.T. Wetherald
V. Ramaswamy
Haywood
et al. (1473) examined the response of a coupled ocean-atmosphere model
to the estimated change of not only greenhouse gases but also anthropogenic
sulfate aerosols. Using estimates of the radiative forcing of greenhouse
gases and sulfate aerosols, they carried out an integration of the GFDL
global coupled ocean-atmosphere climate model from 1765 to 2065 AD. It
was found that the simulated warming trend of global mean surface air temperature
during the past 100 years was remarkably similar to the observed trend
(A96/P97). The model also reproduces the magnitude of the observed decadal
variability reasonably well (1359). Encouraged by these agreements, in-depth
analysis has been performed on the simulated change of the coupled ocean-atmosphere-land
surface system.
Precipitation
increases associated with global warming, particularly in high latitudes,
lead to an increased freshwater supply in the Arctic and surrounding seas.
This increase in freshwater supply contributes to the capping of the oceanic
surface by relatively low density water and inhibits the convective cooling
of the water column in the sinking region of the thermohaline circulation
(THC) in the North Atlantic Ocean. Thus, the THC begins to weaken significantly
after year 2010 (Fig. 1.1). By year 2065, the maximum value of the THC
in the Atlantic is about 10 Sv (1 Sv = 106 m3/sec),
compared to 18 Sv in the control experiment. As the
THC weakens, the northward advection of warm surface water decreases, leading to a reduced warming in the northern North Atlantic and surrounding regions.
Based
upon the results from numerical experiments, it has been suggested that
future increases in greenhouse gases will lead to a decrease in summertime
soil moisture in midcontinental regions of the Northern Hemisphere. However,
because of the reflection of insolation by sulfate aerosols, the summer
reduction of soil moisture in the present experiment is relatively small
during this century and becomes distinguishable only after the year 2010.
Eventually, the cooling of continental surface due to sulfate aerosols
is dominated by heating due to increased greenhouse gases during the 21st
century. Thus, midcontinental summer dryness in the Northern Hemisphere
becomes quite significant by the middle of the 21st century. Compared with
the soil moisture from a control experiment, the fractional reduction of
soil moisture in summer is particularly pronounced over the southern half
of North America, the region around the Mediterranean Sea, and central
Asia.
Analysis
of simulated changes in the coupled atmosphere-ocean-land surface system
will continue. Special emphasis will be placed on the oceanic change associated
with global warming.
1.1.2
Decadal Variability and Trends in the Tropical Pacific
T. Knutson S. Manabe
Possible
mechanisms for observed decadal variability and trends in tropical Pacific
SSTs over the past century have been explored using the GFDL global coupled
ocean-atmosphere climate model (er). Particular attention has been given
to the distinctive triangular-shaped pattern of warm anomalies observed
in the tropical and subtropical Pacific from the late 1970s through the
present. GFDL global climate models have been used to simulate both internal
decadal variability of this region and the response to either increased
CO2 alone or to the estimated combined forcing of CO2
and aerosols over the past century.
The
internally generated decadal variability in a medium resolution (R30) model
control run is similar in amplitude and pattern to the observed decadal
variability (and recent trends), as discussed in A96/P97. The mechanism
controlling the model's internal decadal variability appears to be similar
to the "delayed oscillator" mechanism for the shorter ENSO time
scale, based on the pattern and evolution of the model's subsurface heat
content and surface wind anomalies. The westward phase propagation of heat
content anomalies is slower for the simulated decadal variability compared
to the simulated ENSO variability, in part because of the fact that it
is also centered slightly further from the equator (~12
N
vs. 9N). The slower
westward progression may provide a partial explanation for the longer time
scale of the decadal phenomenon. The preferred time scale of the model's
internal decadal events is estimated to be ~10-15 years.
The
pattern of CO2-induced Pacific warming in the medium resolution
(R30) model shows considerable similarity to that of the simulated internal
decadal variability (correlation of 0.55), although one clear difference
is that the model's CO2-induced pattern is more spatially uniform
and is of one sign (positive) over the entire Pacific domain.
An
index of observed SST over a broad triangular region of the tropical and
subtropical Pacific (Fig. 1.2c) indicates a warming rate of +0.41C/100
yr since 1900, +1.2C/100
yr since 1949, and +2.9C/100
yr since 1971. All three warming trends are highly unusual in terms of
their duration, with occurrence rates of less than 0.5% in a 2000-year
simulation of internal climate variability using a low-resolution model
(e.g., Fig. 1.2a), even after adjusting the model's internal variability
upward to match that of a shorter, medium-resolution experiment (Fig. 1.2b).
The most unusual observed trend is that since 1900 (96-yr duration); the
longest simulated duration of a trend of this magnitude in the 2000-year
climate model simulation is 85 years. This suggests that the observed trends
are not entirely attributable to natural (internal) variability alone,
and may have resulted in part from the sustained thermal forcing due to
greenhouse gases. To quantitatively explore the possible role of greenhouse
gases and aerosols in the observed warming trends, two low resolution (R15)
simulations (using different initial conditions) of 20th century climate
change due to the radiative forcing by these two components were analyzed.
These simulations show an accelerated warming trend (~2C/100yr)
in the triangular Pacific region beginning around 1960 (Fig. 1.2d). The
similarity between this accelerated warming trend and the more recent observed
trends (dashed lines in Fig. 1.2c) suggests that nearly all of the recent
warming in the region could be attributable to anthropogenic forcing.
On the other hand, the cooling trends in the extratropical Pacific (not shown) are more characteristic of the model's internal decadal variability than its thermally forced warming pattern. These model results indicate that neither natural variability nor greenhouse gases plus aerosol forcing alone can account for both the pronounced tropical Pacific warming trends and the extratropical Pacific cooling trends of recent decades. Both natural variability and external forcing have probably contributed to the observed trends, although their relative contributions remain uncertain.
Multicentury
integrations of the medium resolution (R30) coupled model will be examined
with regard to both internal tropical Pacific variability and the tropical
Pacific regional response to anthropogenic forcing (greenhouse gases plus
aerosols). This should provide more reliable indications of the possible
contributions of natural climate variability and anthropogenic forcing
to the recent observed trends in the Pacific.
1.1.3
Cold Ocean-Warm Land Pattern
A.J. Broccoli M.J. Nath
N.-C. Lau
The
cold ocean-warm land (COWL) pattern is obtained by decomposing variations
in spatial mean surface air temperature into a component associated with
a characteristic spatial structure function (i.e., the COWL pattern) and
a residual. This structure function, which features positive values over
the middle and high latitude Northern Hemisphere continents and negative
values over the oceans, is evident in both observed data and the output
from simulations with the GFDL coupled atmosphere-ocean model. A positive
(negative) value for this structure function indicates that land is anomalously
warm (cold) and oceans are anomalously cold (warm). Variations in the amplitude
and polarity of the structure function are an important source of variability
in spatial mean surface air temperature, with positive values of the structure
function contributing to above average spatial mean temperatures, and vice
versa.
During
the past 25 years, the structure function extracted from observed surface
air temperature data has undergone a systematic upward trend. This trend
has contributed to the accelerated warming of the Northern Hemisphere over
the same time period. To assess the importance of this upward trend and
the likelihood that it may be of anthropogenic origin, the variability
in structure function amplitude in a 1000-year integration of the GFDL
coupled atmosphere-ocean model with constant radiative forcing is used
to approximate the unforced variability of the real climate system. A comparison
of the recent observed trend in hemispheric mean temperature associated
with the structure function with trends in the same quantity simulated
by the model (Fig. 1.3) indicates that the observed trend is unusually
large compared with the trends generated internally by the coupled model.
The results of this comparison suggest that the recent positive trend in
structure function may not be purely a manifestation of unforced variability.
A
similar positive trend appears when the structure function is determined
from an integration of the coupled model with time-varying radiative forcing
based on variations in CO2 and sulfate aerosol through the year
2065. The upward trend in the structure function determined from this integration
begins toward the end of the twentieth century after a long period of little
or no systematic trend. The trend is larger in the cold season, a characteristic
shared by the observed structure function during the past 25 years.
The
unusually large magnitude of the observed trend in structure function relative
to the internally-generated variability in the coupled model, combined
with the existence of a similar trend in the coupled model integration
with anthropogenic radiative forcing suggests that the observed trend in
structure function may have an anthropogenic origin (fe).
Further
investigation will focus on the origin of the positive trend in structure
function that appears in the observed surface air temperature data. Output
from additional coupled
model simulations of the 20th and 21st centuries may be examined in an effort to produce a more statistically robust data sample.
1.1.4
Thermohaline Circulation and CO2
S. Manabe R.J. Stouffer
An
investigation of the response of the coupled ocean-atmosphere system to
the rate of increase in the atmospheric carbon dioxide concentration has
been completed using the model described in (1042). Five integrations of
the coupled model were performed in which atmospheric CO2 increases
at the rates of 0.25, 0.5, 1.0, 2.0 and 4.0%/year (compounded) until it
is doubled and remains unchanged thereafter. The corresponding time required
for the doubling of atmospheric CO2 is 280, 140, 70, 35, 17.5
years, respectively. For comparison, two equilibrium integrations were
also carried out for a period of several thousand years, with the atmospheric
CO2 concentration constant at its present day and twice present
day values, respectively, throughout the integration.
It
was found that the integrations with the slowest rates of increase in the
atmospheric CO2 concentration have the largest decrease in the
strength of the thermohaline circulation, THC, in the Atlantic Ocean by
the time of CO2 doubling, as indicated in Table 1.1. This reduction
of the THC in response to increasing CO2 was discussed earlier
in (1042). They showed that, as CO2 increases, the temperature
and absolute humidity increase in the troposphere, enhancing the poleward
moisture transport. The enhanced moisture transport, in turn, causes a
marked increase in precipitation in high latitudes, capping the oceanic
surface with relatively fresh, low density water. The capping reduces the
negative buoyancy production and convective activity in the sinking region
of the THC, thereby reducing its intensity. As noted above, the slower
the rate of CO2 increase, the longer it takes for the CO2
concentration in the atmosphere to double, making more time available for
the capping of the oceanic surface and the development of a halocline in
high North Atlantic latitudes and in the Arctic Ocean. This is why the
weakening of the THC by the time of CO2-doubling is largest
in the integration using the slowest rate of CO2 increase (0.25%)
as indicated by Table 1.1.
| Experiment | 0.25% | 0.5% | 1% | 2% | 4% | 2X-Equil. | Control |
| THC (Sv) | 10 | 12 | 13 | 15 | 17 | 17 | 17 |
Table 1.1. The maximum intensity
of the THC (units: Sv. = 106 m3/sec) in the North Atlantic Ocean at the
time when atmospheric concentration of CO2 has doubled. Here, the maximum
intensity of the THC is defined as the maximum value of the stream function
which represents the meridional overturning in the North Atlantic. The
transient response experiments are identified by the percentage increase
of atmospheric CO2 per year.The
"Control" and "2X-Equil." columns indicate the maximum
intensities of the THC which are attained towards the end of several thousand
years integration with the normal and twice-normal concentrations of atmospheric
CO2, respectively.
However,
the THC begins to reintensify approximately 100 years after the CO2
stops increasing (Fig. 1.4) and eventually recovers to its original intensity
(1192), which explains why the THC in the two equilibrium integrations
have similar intensities. This is in sharp contrast to the weakening of
the THC found in the transient integrations. This similarity in the strength
of the THC between the two equilibrium integrations has been noted previously
(700). The reason for the similarity is that the meridional density gradient
which drives the THC differs very little between the equilibrium climates.
If the waters warmed uniformly everywhere in response to the increased
atmospheric CO2, the density reduction of the low latitude water
would be relatively larger than the high latitude water (because of the
nonlinear dependence of sea water density on temperature), increasing the
meridional density gradient. On the other hand, the polar amplification
of the surface warming and the decrease in surface salinity in high latitudes,
as discussed above, combine to reduce the meridional density gradient.
The net result of these opposing effects is that the meridional density
gradient, and therefore the THC, differs little between the two equilibrium
climates.
However,
on shorter time scales (hundreds of years), the weakening of the THC reduces
the warming in the northern North Atlantic and surrounding regions (1042,
1192). This simulated
dependence of the THC on the rate of CO2 increase has potentially profound implications for any future greenhouse gas mitigation strategies.
In-depth
analysis of the experiments will continue.
1.1.5
Sea Surface Temperature Variability in the Greenland
Sea
T. Delworth S. Manabe
R.J. Stouffer
Analyses
of a multi-millenial integration of a coupled ocean-atmosphere model have
revealed pronounced oscillations of ocean temperature and salinity in the
Greenland Sea (db). The oscillations, involving both the surface and subsurface
layers, have a time scale of approximately 40-80 years, with a peak around
60 years, and are associated with fluctuations in the intensity of the
East Greenland Current. The Greenland Sea temperature and salinity variations
are preceded by large-scale changes in near-surface salinity in the Arctic,
which appear to propagate out of the Arctic through the East Greenland
Current. These anomalies then propagate around the subpolar gyre into the
Labrador Sea, the central North Atlantic and the Norwegian Sea. The oscillations
are coherent with previously identified (1182) multi-decadal fluctuations
in the intensity of the North Atlantic thermohaline circulation. In addition,
the model oscillations have a distinct resemblance to oscillations detected
in the instrumental and proxy record1.
The oscillations involve the atmosphere as well, with the cold SST anomalies
and intensified East Greenland Current being associated with cold surface
air temperatures and northerly wind anomalies over the Greenland Sea.
Additional
analysis of these variations has revealed an intriguing multicentennial
scale modulation of the multidecadal variability. Wavelet analysis of SSTs
in the Greenland Sea (shown in Fig. 1.5) demonstrates that there are groups
of centuries where this variability is strong, followed by other centuries
where this variability is weak. This modulation of the multidecadal variability
has very important implications for characterizing and predicting decadal
to multidecadal variability, as well as for issues of climate change detection.
Additional
analyses will be conducted regarding the mechanisms responsible for the
large multidecadal variability signature in the coupled ocean-atmosphere
model, particularly with regard to the role of the atmosphere. The multicentennial
scale modulation, as well as the centennial variations, will also be investigated
further.
1.1.6
Modeling Study of Water Vapor Feedback
A. Hall S. Manabe
The
precise amount of warming that would result from a given increase in greenhouse
gases in the atmosphere remains highly uncertain. One source of this uncertainty
is the inability to quantify the role of feedback mechanisms in determining
the sensitivity of climate to the change in greenhouse gas forcing. Water
vapor feedback has long been considered a positive feedback mechanism (50).
The global warming of the coupled atmosphere-surface system leads to an
increase in atmospheric water vapor (a greenhouse gas), thereby reducing
further the radiative damping which would otherwise mitigate the warming.
Thus, the CO2-induced warming is enhanced by additional water
vapor feedback.
Water
vapor feedback should affect naturally-occurring temperature anomalies
in much the same way that it affects a warming induced by a change in greenhouse
gas concentrations. For example, if some completely natural forcing induced
a warm temperature anomaly in the coupled atmosphere-surface system, the
resulting increase in atmospheric water vapor and hence greenhouse trapping
would decrease the radiative damping of that anomaly and the warm anomaly
will be larger than it would be in the absence of water vapor feedback.
Similarly, since cold air holds less water vapor, a negative temperature
anomaly would decrease the greenhouse effect of the atmosphere through
reduced water vapor mixing ratios, forcing the cold anomaly to be even
colder. A positive feedback mechanism such as water vapor feedback should
therefore impact the radiative damping of naturally-occurring temperature
anomalies in a manner similar to the way greenhouse gases induce global
warming.
To
study the role of water vapor feedback in natural temperature variability,
the GFDL R15 coupled ocean-atmosphere model was integrated for 1000 years
in two configurations; the first with a completely interactive water vapor
budget and the second with the water vapor distribution constrained to
climatological values for the purposes of radiation calculations, thus
effectively disabling the model's water vapor feedback. As shown in the
top panel of Fig. 1.6, the integration which included water vapor feedback
was observed to possess higher surface temperature variability on all spatial
and temporal scales, implying that water vapor feedback is, indeed, positive
in the context of the model's natural variability. The standard deviation
of global mean surface temperature is 0.144C
in the integration with water vapor feedback, and only 0.092C
in the integration without water vapor feedback. In addition, water vapor
feedback affects surface temperature anomalies more as time and spatial
scales increase. Finally, surface temperature anomalies over the ocean
are more affected by water vapor feedback than their land counterparts.
In
order to better understand the role of water vapor feedback in a global
warming scenario, two analogous 500-year integrations were performed in
which CO2 was doubled. Water vapor feedback was observed to
be positive in the enhanced-CO2 global warming case as well.
In fact, the model's water vapor feedback has an even larger impact on
surface warming in response to a doubling of CO2 than in naturally-occurring,
low-frequency, global mean surface temperature anomalies. The standard
deviation of the equilibrium global
warming at the surface in the model with water vapor feedback was 3.78C,
but only 1.05C
for the case of climatologically prescribed water vapor (see the bottom
panel of Fig. 1.6). The strength of the feedback, therefore, appears to
depend on the type of temperature anomaly it affects. One factor determining
the effectiveness of the feedback is the altitude to which these temperature
anomalies typically penetrate. The higher the anomaly penetrates, the stronger
the feedback.
Analysis
of these experiments will continue and the role of water vapor feedback
in well-known climate fluctuations (such as ENSO) will be investigated.
In addition, the role of water vapor feedback in inducing precipitation
anomalies will be explored.
1.1.7
Coupled Model Development
T. Delworth K. Dixon
M. Spelman R. Stouffer
Substantial
effort has been expended over the last year toward improving a medium (R30)
resolution version of the coupled ocean-atmosphere-land surface model.
This model has a horizontal grid size of approximately 250 km, with 14
atmospheric levels and 18 oceanic levels. Previous integrations of similar
models have suffered from substantial climate drift which severely compromised
the utility of those models for some scientific applications.
A
new version of this model has been developed with somewhat stronger horizontal
diffusion. Specifically, the background horizontal mixing coefficient was
increased to 0.75 x 106 cm-2 s-1. Preliminary
results with this model demonstrate that the time-mean simulation is quite
realistic, and that the climate drift in this model has been substantially
reduced from previous versions. (The drift in global mean sea surface temperatures
over the first 80 years of the integration is less than 0.1C.)
The success of this latest version of the medium resolution coupled model
is a vital step towards studies of a host of climate issues over the next
several years. In particular, the much improved simulation of many aspects
of the atmospheric circulation in this model, including the simulation
of transient eddies, will allow a more robust examination of many critical
issues in climate research. It is anticipated that this model will have
a substantially improved simulation of ENSO, the Asian monsoon, and decadal
variability, as well as other features.
In
addition, substantial changes were made to the engineering aspects of model
execution and data storage. The model output is now routinely stored in
netCDF (a data storage protocol common to the meteorological and oceanographic
research community). This allows easy exchange of model output with colleagues
both within and external to GFDL, as well as facilitating the use of many
existing analysis packages. A considerable effort was also made in completely
rewriting the system for controlling model execution and data post-processing.
This system allows for a very flexible and highly automated process of
model control.
The
control run of the medium resolution coupled model will be extended indefinitely
(more than a century). The natural variability in this extended integration
of the coupled model will be examined and documented. A number of integrations
will be conducted using this model to explore the response of the climate
system to altered radiative forcings.
1.2 CONTINENTAL
HYDROLOGY AND CLIMATE
1.2.1
Greenhouse-Induced Changes in Extremes of River Discharge
K.A. Dunne* R.T. Wetherald
P. C.D. Milly*
*U. S. Geological Survey
In
a continuation of earlier work on temporal variability of runoff and river
discharge, the response of river-discharge extremes (floods and droughts)
to an equilibrium quadrupling of atmospheric carbon dioxide has been computed.
Analyses were all focused on a set of geographic areas corresponding to
several major gauged river basins of the world. The analysis was based
on monthly climate-model runoff fields computed in earlier studies (1192,
1201). Observations have been provided by national hydrological services
and are typically of 50-100 years duration.
Extremes
of river discharge are highly sensitive to water storage in river basins.
Some forms of storage (soil water, snowpack) are represented in climate
models, but others (ground water, rivers and lakes) are not. For this study,
a simple linear storage model was introduced to represent these neglected
forms of storage. For each month, model runoff was aggregated by river
basin and the resulting total runoff was routed through the linear reservoir
to form a synthetic river discharge. The linear reservoir is characterized
by a mean residence time of water in the basin. The residence time was
estimated from the frequency spectrum of observed river discharge of each
basin under the assumption that river discharge is a red-noise process
that results when a basin filters essentially white-noise runoff. Most
of the inferred residence times were on the order of 1-3 months.
The
resulting time series of modeled monthly river discharge were analyzed
statistically for annual high and low values, and the results are expressed
in terms of event return periods. (The return period of a flood F is the
mean time between floods having magnitude equal or greater than F; drought-flow
return periods can be similarly defined.) A wide variety of responses to
greenhouse warming has been noted in preliminary analyses. For the Mississippi
at Vicksburg (Fig. 1.7), the general decrease in annual runoff within that
basin leads to increased return periods for flooding and greatly decreased
return periods for low flows. In contrast, the basins that experience a
general increase in flow commonly show the opposite effects. Analyses of
trend detectability suggest that changes in extreme flows typically become
evident at about the same time as changes in annual mean flows. In the
case of the Mississippi at Vicksburg, these trends are not statistically
significant until after the time of doubling of atmospheric carbon dioxide.
The
assumption of white-noise runoff cannot be directly tested, but analysis
of spectra of modeled and observed precipitation would help to establish
its plausibility or to develop needed modifications. Consequently, plans
call for analysis of available precipitation
observations and comparisons with model outputs. Return-period analyses of hydrologic extremes will be continued in more detail, and an attempt will be made to interpret the results for their practical implications for contemporary climate change.
1.2.2
Trend Analysis of Observed River Discharge
K.A. Dunne* P.C.D. Milly*
* U. S. Geological Survey
Previous
trend analyses of 20th-century river discharge have continued, with datasets
extended in both time and space. The time series of observed annual discharge
from nine large river basins, tropical and extratropical, were analyzed.
Best-fit linear trends of annual discharge for 1904-93 ranged from -9%/century
(Northern Dvina River in European Russia) to +30%/century (St. Lawrence
River). Trend significance was assessed in the context of a Markov (lag-one
autoregressive) model of river runoff. For only two of the rivers were
the trends significant at the 95% level (upper Amazon River, +8%/century;
St. Lawrence River).
Preliminary
estimates of expected anthropogenic trends in river flow have also been
made on the basis of various climate-model experiments and assessments
of global water-resource development and land-use changes. These estimates
suggest that the anthropogenic signal in river discharge due to the considered
processes (climate change due to increased atmospheric carbon dioxide and
sulfate aerosols, stomatal closure due to increased atmospheric carbon
dioxide, direct water-balance effects of deforestation) is not sufficiently
large to explain the trends noted in the Amazon and St. Lawrence basins.
Possible explanations include: 1) failure of the statistical assumptions
in the analysis of trend significance; 2) underestimation of the magnitude
of the anthropogenic effects that were considered; 3) neglected anthropogenic
or natural forcing functions external to the climate system; or 4) improbable,
chance natural variations of discharge.
Analyses
will continue, with a focus on extraction of statistically robust information
from the observed time series and interpretation of the observed trends.
The possible role of aerosols from biomass burning will also be investigated.
1.2.3
Soil-Moisture Predictability and Associated Climate
Predictability
P.C.D. Milly* C.A. Schlosser
*U. S. Geological Survey
Ensemble
forecast experiments were performed with the GFDL climate model to explore
the predictability of soil moisture and the impact of initial soil moisture
information on monthly to seasonal climate forecasts. Initially, the model
was run for 100 years, in conjunction with climatological average ocean-surface
temperatures, to establish a model climatology. In the process, instantaneous
global fields of all prognostic variables, including soil moisture, were
saved at the beginning of each month. These fields were used, in various
combinations, as initial conditions for ensemble forecast experiments.
Preliminary analysis has focused on ensemble forecasts initialized at the
beginning of the northern-hemisphere summer months (i.e., June-August).
For each initialization date, eight ensemble forecasts were run with 10
members per ensemble (for a total of 80 forecast runs). For an ensemble
forecast, each member of the ensemble was provided the same initial condition
of soil moisture, while the initial atmospheric state was randomized between
members. Each ensemble was assigned a different initial soil moisture state
and each forecast was run for one year.
A
"relative inter-ensemble variance" is used as a measure of predictive
skill for soil moisture, surface-air temperature, and precipitation. This
quantity is defined as the ratio of the inter-ensemble variance to the
total variance among all of the forecast runs (80 of them) starting on
a common date. The relative inter-ensemble variance increases (i.e., tends
to a value of 1) with stronger influence of the initial soil moisture information.
If the initial soil moisture information has no impact on the ensemble
forecast means (i.e., all of the forecasts in an ensemble are essentially
independent of each other), then the relative inter-ensemble variance tends
to a lower value (dependent on ensemble size and number of ensembles).
Figure
1.8 shows a representative example of the initial analysis for the July
1 forecasts. Soil moisture is predictable with an exponential decay time
period on the order of one month. The associated predictive skill for near-surface
air temperature is smaller, but its time scale does not differ significantly
from that for soil moisture. So far, soil-moisture initialization appears
to yield no significant predictive skill for precipitation. In general,
larger predictive skill is seen for area-averaged and/or time-averaged
soil moisture and air temperature (shown in Fig. 1.8) than for their respective
instantaneous, grid-point values.
A
more extensive analysis of the experiments will be completed, with a focus
on identification of scale-dependence of predictability. The results will
be interpreted in the context of previous works on stand-alone soil-moisture
predictability and atmospheric sensitivity to surface processes.
1.3.1
Tropical Cooling at Last Glacial Maximum
A. J. Broccoli
Investigation
of tropical temperature changes in a coupled atmosphere-mixed layer ocean
model simulation of the climate of the last glacial maximum has continued.
The forcing for this simulation was specified by the Paleoclimate Modeling
Intercomparison Project (PMIP), and consists of a change in orbital parameters
to their values at 21,000 years before present, a reduction in sea level
by 105 m, the imposition of continental ice sheets as reconstructed by
a geophysical inverse calculation, and an approximately 25% reduction in
atmospheric carbon dioxide.
Work
during the past year has focused on understanding some of the regional
variations in simulated glacial cooling at low latitudes. Of particular
interest are regions of relatively small cooling over the subtropical oceans
of the Southern Hemisphere (Fig. 1.9, top).
Identifying the physical mechanism responsible for these low sensitivity regions has proven to be very difficult, but some additional insight has come from a doubled CO2 integration (2xCO2) conducted using the same climate model used for the PMIP glacial simulation. Patterns of temperature change over the tropical and subtropical oceans in the 2xCO2 integration bear some resemblance to those in the glacial simulation, but with reversed polarity (Fig. 1.9, bottom). An area of relatively low sensitivity extends from west-northwest to east-southeast across the subtropical South Pacific in each integration, while a belt of relatively high sensitivity extends from just off the west coast of North America west-southwest across the North Pacific. In both integrations there appears to be some relationship between the patterns of temperature change and changes in the prevailing surface winds. These changes in atmospheric circulation may be the result of the interhemispheric asymmetry in the warming (cooling) at middle and high latitudes of the doubled CO2 (glacial) integration. The doubled CO2 integration has proven valuable because it provides a simpler framework for investigating the relationship between the circulation and temperature changes, since the large changes in atmospheric circulation associated with the expanded continental ice sheets of the last glacial maximum do not represent an additional source of complexity.
A
more complete understanding of the mechanisms that produce spatial variations
in tropical cooling will be the primary research goal. In addition, the
analysis of the simulation of the last glacial maximum will be extended
beyond the tropics, and a comparison of the GFDL model results with those
of other research groups contributing to PMIP will also be undertaken.
1.4.1
Tropical Intraseasonal Oscillations
D.G. Golder Y. Hayashi
Tropical
intraseasonal oscillations (TIOs) simulated by the R30L9 GFDL spectral
climate model (1118) consist of 40-50 and 25-30 day oscillations. The two
oscillations, however, have comparable magnitudes, being contrary to observed
TIOs which are dominated by a 40-50-day TIO. To improve the simulation
of TIOs, a 20-level R30 spectral model, having the original scheme of moist
convective adjustment, was run for 20 years with predicted clouds ("variable
clouds") and observed climatological zonal-mean clouds ("fixed
clouds") in the radiation scheme. It was found that the tropospheric
zonal velocity of the variable-cloud model exhibits a pronounced 40-50
day spectral peak (Fig. 1.10a) corresponding to the 4050 day TIO, in agreement
with observed spectra. This feature is a substantial improvement over the
fixed-cloud model, which simulates 40-50 and 25-30 day TIOs of comparable
magnitude (Fig.1.10b). These
figures also indicate that cloud feedback enhances the 40-50 day peak while
reducing the 25-30 day peak.
In
order to explain the above results, the radiative damping coefficient was
estimated as the space-time spectral regression coefficient of radiation
which is assumed to be, in part, linearly related to temperature. The coefficient
indicates that radiation acts to reduce both
the 40-50 and 25-30 day TIOs in the upper troposphere (100-200 mb), but acts to enhance them in the middle troposphere (200-500 mb). Both the positive and negative values of the coefficient are substantially enhanced by cloud feedback. Since the 40-50 day TIO is more confined to the troposphere than the 25-30 day TIO, the 40-50 and 25-30 day TIOs are probably sensitive to middle-tropospheric radiative forcing and upper-tropospheric radiative damping, respectively. If so, the radiative forcing and damping may explain the above results.
There
is also the possibility that the discrepancy between the climatology of
predicted clouds and the observed climatological cloudiness results in
substantially different model climatologies, which in turn affect the TIOs.
In particular, reduced tropospheric mean static stability enhances the
growth rate of the 40-50 day TIO, but reduces that of the 25-30 day TIO
through the evaporation-wind feedback (EWF) mechanism (ba). Increased near-surface
easterly mean flow also increases the growth rate of the 40-50 day TIO
more efficiently than the 2530 day mode through EWF instability (ba). This
possibility, however, is unlikely, since analyses of the climatology of
the temperature and velocity fields indicate that the tropospheric static
stability is not reduced by cloud feedback but rather slightly enhanced
(i.e., the tropopause temperature is increased, becoming more realistic)
and that the near-surface easterly mean flow is only marginally enhanced.
In
the R30 spectral model, the near-surface mean wind is easterly (-2 m/s)
in the equatorial western Pacific "warm-pool" region (where eastward-moving
TIOs and superclusters amplify), contrary to recent observations that indicate
westerlies in this region. On the other hand, the 40-level GFDL SKYHI model
having high horizontal resolution indicates westerlies in this region,
in agreement with observations. To examine the sensitivity of TIOs to the
near-surface mean wind, numerical experiments were conducted using an idealized
nine-level R21-spectral model with moist convective adjustment. The model
prescribes globally uniform distributions of sea surface temperatures and
insolation conditions. It can also prescribe the zonal-mean component (U)
of the near-surface zonal velocity in the parameterized surface fluxes
of latent and sensible heat, while allowing the deviation from the zonal
mean to fluctuate. As found in previous studies (1452), the 40-50 and 25-30
day TIOs, which occur for U = -2 m/s (easterly), are profoundly weakened
by the elimination of the EWF mechanism. It was found that, as U shifts
from -6 m/s (easterly) to +2 m/s (westerly) in the presence of the EWF
mechanism, the dominant period of the TIOs shifts from 25-30 days eastward
to 170-190 days westward, the magnitude being much weaker for westerly
U than easterly U.
The
above idealized model experiments indicate that only weak westward-moving
TIOs are generated for westerly zonal wind. Moreover, the westerly mean
flow in the warm-pool region implies that eastward-moving TIOs are stabilized
in this region through the EWF mechanism. To explain why eastward-moving
TIOs amplify in the warm-pool region of westerlies, the united mechanism
for the generation of TIOs and superclusters is proposed as follows. In
the idealized model, TIOs are destabilized for easterly U by evaporation-wind
feedback through the EWF mechanism. On the other hand, superclusters are
generated through the saturation-triggering mechanism (1452) by the intermittent
onset of moist convection, upon saturation, to neutralize any pre-existing
conditionally unstable stratification. In the realistic model, eastward-moving
TIOs are destabilized by the EWF mechanism in the region of easterlies
east of the warm-pool region. They are further amplified in the warm-pool
region of westerlies through the modulation of the superclusters generated
by the saturation-triggering mechanism.
In
order to confirm that the difference between the tropical intraseasonal
oscillations simulated with variable and fixed clouds is not essentially
due to the slight difference in model climatologies, cloud feedback will
be suppressed in such a way that the climatology of model cloudiness remains
the same. The united mechanism for the generation of tropical intraseasonal
oscillations and superclusters will also be examined by the use of a realistic
model.
1.4.2
Baroclinic Instability, Geostrophic Turbulence, and
Extratropical Dynamics
I. Held P.
Phillipps
P. Kushner T. Schneider
O. Pauluis
1.4.2.1
Wave-Mean Flow Interaction in Zonally Asymmetric Flows
While
the theory of wave-mean flow interaction on zonally symmetric atmospheric
flows is one of the most significant achievements in atmospheric dynamics
of the past 20 years, the generalization of this theory to zonally asymmetric
basic states has proven difficult. More specifically, there does not yet
exist a good theoretical framework within which to study the interaction
between the storm track eddies in midlatitudes and larger scale quasi-stationary
planetary waves. Motivated by recent work on the problem of parameterizing
eddy fluxes in ocean models, a formalism has been developed in which one
focuses attention on the mass between two PV contours on an isentropic
surface and the ways in which this mass is redistributed by eddy activity.
This new perspective has clarified the connections between previous work
on this problem and provides a simple picture of how eddies can modify
the mean flow even when they are steady, adiabatic, and inviscid, a special
case of particular interest because, in this limit, this interaction is
a direct consequence of the zonal asymmetries of the background flow.
1.4.2.2
The Near Surface Branch of the Meridional Overturning
Circulation in the Troposphere
When
one computes the north-south mass transport within isentropic layers in
the atmosphere, one finds a relatively simple overturning circulation,
with poleward flow in the upper troposphere and return flow near the surface.
Focusing on the surface branch of this return flow, a simple argument is
presented for the range of the potential temperatures in which this equatorward
flow occurs: at a particular latitude, the return flow is primarily confined
to isentropic layers which are cold enough that they intersect the ground
at some longitudes. In contrast, those layers that are warm enough so as
to be typically uninterrupted by the surface carry a poleward flow. In
order to analyze this overturning circulation in height or pressure coordinates,
the concept of a "residual mean circulation" has historically
been very valuable, but it is poorly defined at the surface. An alternative
definition of the residual circulation has been developed that provides
a cleaner depiction of the equatorward return flow, motivated by recent
work on oceanic eddy flux closure. Calculations with an idealized atmospheric
model emphasize that the surface branch of this circulation can have important
consequences in that it can advect cold temperatures equatorward if not
prevented from doing so by air-mass modification, creating a surface inversion
throughout midlatitudes.
1.4.2.3
Studies of the Tibetan High in the Northern Summer
Using Idealized GCMs
The
anticyclone in the upper troposphere over the Asian monsoon region in Northern
summer can be thought of as driven by the vortex compression resulting
from the upward motion associated with latent heat release. But the simplest
models of this process show that the resulting anticyclone will spread
westward due to a special form of Rossby wave propagation that oceanographers
have termed the "beta plume". The observed anticyclone does spread
westward as far as Africa. The question then arises as to the factors that
control the extent of this westward expansion. Idealized models in which
the strength of midlatitude eddy activity is varied suggest that potential
vorticity mixing due to baroclinic waves is a key limiting factor. Analysis
of these experiments is in progress.
1.4.2.4
Studies of the Tropospheric Lapse Rate in Extratropical
Latitudes Using Idealized GCMs
It
is known from idealized GCM simulations that baroclinic eddies alone, in
the absence of moist convection, can maintain a stratification comparable
to that observed in midlatitudes, but even a qualitative theory for this
stratification remains elusive. A series of integrations have been performed
to determine how the stratification varies in idealized GCMs as different
parameters, such as the north-south temperature gradient or the height
of the tropopause, are systematically altered. Work is in progress on analyzing
these integrations.
1.4.2.5
Geostrophic Turbulence and Eddy Flux Parameterization
Theoretical
developments on the fundamentals of quasi-geostrophic turbulence generated
by baroclinic instability have been described earlier (1337, 1362, 1369,
bm). Besides contributing to the fundamental understanding of atmospheric
dynamics, these developments have implications for the problem of parameterizing
the effects of mesoscale baroclinic eddies in ocean climate models. Work
is continuing on three fronts: 1) the analysis of an eddy resolving two-layer
quasi-geostrophic model of a mid-ocean subtropical gyre; 2)integrations
which focus on the manner in which surface damping halts the inverse energy
cascade in the two-layer homogeneous turbulence model; and 3) a study of
barotropic instability of a point jet. This work is aimed at bridging the
gap between studies of homogeneous turbulence and inhomogeneous flows of
geophysical interest.
Work
will continue on problems related to storm track dynamics, geostrophic
turbulence, and eddy flux parameterization using a variety of idealized
models.
1.4.3
NOAA/University Joint Study of the Maintenance of
Regional Climates and Low Frequency Variability in GCMs
I. Held
P. Phillipps
P. Kushner Q. Zhang
N.-C. Lau J. Zhang
M. Nath
A
collaboration among GFDL, NOAA/Climate Diagnostics Center, MIT, the Universities
of Washington, Chicago, and Illinois, and the Lamont Doherty Earth Observatory
has continued its study of the interrelated problems of stationary waves,
storm tracks, low-frequency variability, and the response of the atmosphere
to perturbations in boundary forcing. The group has collaborated in designing
experiments to be performed at GFDL. Further information about this project
has been provided in the annual reports from previous years. Described
here are those projects related to consortium efforts in which the GFDL
contribution is predominant.
1.4.3.1
Stochastic Models of Storm Tracks
A
recent development in atmospheric dynamics has been the construction of
approximations to atmospheric models consisting of a linear operator forced
by white noise. One approach consists of linearizing about some mean climatic
state, and then adding dissipation so as to stabilize this operator. Although
a theoretical understanding of the appropriate form and strength of the
dissipation is lacking, one can make progress in simulating the midlatitude
storm tracks by using the simplest possible linear damping and treating
the strength of the damping as a tuning parameter. Using this approach,
eddy statistics generated by an atmospheric GCM have been reproduced with
surprising fidelity, as shown in Fig. 1.11. Using this stochastic model
with different basic states, the midwinter suppression of the Pacific storm
track has been simulated in a qualitative way. This phenomenon, in which
the eddy activity over the Pacific is stronger in November and March than
in January, is a stringent test of the current understanding of the storm
tracks. This study has shown that the suppression should be understandable
in terms of differences in the background flows during the different winter
months.
1.4.3.2
Effective Linear Operators
In
an alternative approach to linear stochastic modeling, suggested in a series
of papers by Farrell, Del Sole, and Branstator and closely related to the
classic work on "fluctuation-dissipation" relations by Leith,
the linear operator is not obtained by linearizing the equations of motion
about some more or less arbitrarily chosen basic state, but rather by statistical
fitting to the time-variability of the atmosphere itself, or a model of
the atmosphere. The resulting operator is referred to as an "effective
linear operator", or ELO. The ELO is constructed using the data from
an R30 GCM with a zonally symmetric climate, and have used it not only
to simulate storm tracks with stochastic forcing. but also as an effective
Green's function with which one can attempt to compute the climatic response
to thermal or mechanical forcing.
1.4.3.3
Idealized GCM integrations
In
collaboration with S. Lee at Penn State, the analysis of a series of GCM
integrations with zonally symmetric (all ocean) boundary conditions has
continued, using the 14-level R30 model. The climatic responses to zonally
symmetric perturbations in the SST distribution are being examined. A variety
of circulation indices of interest (the strength of the surface westerlies,
the eddy kinetic energy in the storm tracks, and the strength of the Hadley
cell) show remarkably linear climatic responses. The response to the sum
of two SST anomaly patterns is quite close in many cases to the sum of
the responses to the individual anomalies. The implication is that one
can think of the general climatic response as being decomposed into a small
number of "canonical" responses -- such as the response to tropical
warming of SSTs and the response to increased north-south temperature gradient
-- at least in this zonally symmetric framework. One can then attempt to
develop a thorough dynamical interpretation of the climatic responses in
this small number of canonical sensitivity experiments. The response to
the warming of tropical surface temperatures in this idealized model has
been found to be quite similar to the changes in Southern Hemisphere circulation
seen in the R15 coupled model global warming scenario integrations.
Stochastic
models of storm tracks and the use of "effective linear operators"
to study atmospheric climatic responses to various kinds of perturbations
will be pursued. In addition, a new call for proposals has been issued
by the NOAA Office of Global Programs for collaborative work in the next
three years within this project. Work will continue to focus on the utilization
of atmospheric GCMs to improve our understanding of stationary waves, storm
tracks, and the response of the atmosphere to perturbations in the surface
boundary conditions.
G.P. Williams
[ Omitted at author's request ]
*Portions of this document contain material that has not yet been formally published and may not be quoted or referenced without explicit permission of the author(s).