ArticlePDF Available

Abstract and Figures

Equilibrium climate sensitivity refers to the ultimate change in global mean temperature in response to a change in external forcing. Despite decades of research attempting to narrow uncertainties, equilibrium climate sensitivity estimates from climate models still span roughly 1.5 to 5 degrees Celsius for a doubling of atmospheric carbon dioxide concentration, precluding accurate projections of future climate. The spread arises largely from differences in the feedback from low clouds, for reasons not yet understood. Here we show that differences in the simulated strength of convective mixing between the lower and middle tropical troposphere explain about half of the variance in climate sensitivity estimated by 43 climate models. The apparent mechanism is that such mixing dehydrates the low-cloud layer at a rate that increases as the climate warms, and this rate of increase depends on the initial mixing strength, linking the mixing to cloud feedback. The mixing inferred from observations appears to be sufficiently strong to imply a climate sensitivity of more than 3 degrees for a doubling of carbon dioxide. This is significantly higher than the currently accepted lower bound of 1.5 degrees, thereby constraining model projections towards relatively severe future warming.
Content may be subject to copyright.
ARTICLE doi:10.1038/nature12829
Spread in model climate sensitivity
traced to atmospheric convective mixing
Steven C. Sherwood
, Sandrine Bony
& Jean-Louis Dufresne
Equilibrium climate sensitivity refers to the ultimate change in global mean temperature in response to a change in
external forcing. Despite decades of research attempting to narrow uncertainties, equilibrium climate sensitivity
estimates from climate models still span roughly 1.5 to 5 degrees Celsius for a doubling of atmospheric carbon dioxide
concentration, precluding accurate projections of future climate. The spread arises largely from differences in the
feedback from low clouds, for reasons not yet understood. Here we show that differences in the simulated strength of
convective mixing between the lower and middle tropical troposphere explain about half of the variance in climate
sensitivity estimated by 43 climate models. The apparent mechanism is that such mixing dehydrates the low-cloud layer
at a rate that increases as the climate warms, and this rate of increase depends on the initial mixing strength, linking the
mixing to cloud feedback. The mixing inferred from observations appears to be sufficiently strong to imply a climate
sensitivity of more than 3 degrees for a doubling of carbon dioxide. This is significantly higher than the currently
accepted lower bound of 1.5 degrees, thereby constraining model projectionstowardsrelatively severe future warming.
Ever since numerical global climate models (GCMs) were first developed
in the early 1970s, they have exhibited a wide range of equilibrium
climate sensitivities (roughly 1.5–4.5 uC warming per equivalent doub-
ling of CO
and consequently a broad range of future
warming projections, with the uncertainty due mostly to the range of
simulated net cloud feedback
. This feedback strength varies from roughly
zero in the lowest-sensitivity models to about 1.2–1.4 W m
in the highest
. High clouds (above about 400hPa or 8km) contribute
about 0.3–0.4 W m
to this predicted feedback because the tem-
peratures at the tops of the clouds do not increase much in warmer
climates, which enhances their greenhouse effect. Mid-level cloud
changes also make a modest positive-feedback contribution in most
Another positive feedback in most models comes from low cloud,
occurring below about 750 hPa or 3 km, mostly over oceans in the
planetary boundary layer below about 2 km. Low cloud is capable of
particularly strong climate feedback because of its broad coverage and
because its reflection of incoming sunlight is not offset by a commen-
surate contribution to the greenhouse effect
. The change in low cloud
varies greatly depending on the model, causing most of the overall
spread in cloud feedbacks and climate sensitivities among GCMs
No compelling theory of low cloud amount has yet emerged.
A number of competing mechanisms have, however, been suggested
that might account for changes in either direction. On the one hand,
evaporation from the oceans increases at about 2% K
, which—all
other things being equal—may increase cloud amount
. On the other
hand, detailed simulations of non-precipitating cloudy marine bound-
ary layers show that if the layer deepens in a warmer climate, more dry
air can be drawn down towards the surface, desiccating the layer and
reducing cloud amount
The lower-tropospheric mixing mechanism
We consider that a mechanism similar to this one, which has so far
been considered only for a particular cloud regime, could apply more
generally to shallow upward moisture transports, such as by cumulus
congestus clouds or larger-scale shallow overturning found broadly
over global ocean regions. Air lifted out of the boundary layer can
continue ascending, rain out most of its water vapour, and then return
to a relatively low altitude—or it can exit the updraught directly at the
low altitude, retaining much more of its initial vapour content. The
latter process reduces the ‘‘bulk precipitation efficiency’’ of convection
allowing greater transport of moisture out of the boundary layer for a
given precipitation rate. Such a process can increase the relative humidity
above the boundary layer
and dry the boundary layer. Unlike the global
hydrological cycle and the deep precipitation-forming circulations
however, it is not strongly constrained by atmospheric energetics
We present measures of this lower-tropospheric mixing and the
amount of moisture it transports, and show that mixing varies sub-
stantially among GCMs and that its moisture transport increases in
warmer climates at a rate that appears to scale roughly with the initial
lower-tropospheric mixing.
Mixing-induced low cloud feedback
The resulting increase in the low-level drying caused by lower-tropospheric
mixing produces a mixing-induced low cloud (MILC) feedback ofvari-
able strength, which can explain why low-cloud feedback is typically
and why it is so inconsistent among models.
In a GCM, vertical mixing in the lower troposphere occurs in two
ways (Extended Data Fig. 1). First, small-scale mixing of heat and water
vapour within a single grid-column of the model is implied by con-
vective and other parametrizations. Lower-tropospheric mixing and
associated moisture transport would depend on transport by shallow
cumulus clouds, but also on the downdrafts, local compensating sub-
sidence and evaporationof falling rain that are assumed to accompany
deeper cumulus. Second, large-scale mixing across isentropes occurs
via explicitly resolved circulations. Whether this contributes to lower-
tropospheric mixing will again depend on model parametrizations,
but in this case, on their ability to sustain the relatively shallow heating
that must accompany a shallow (lower-tropospheric) circulation. We
measure these two mixing phenomena independently, starting with
Climate Change Research Centre and ARC Centre of Excellence for Climate System Science, University of New South Wales, Sydney 2052, Australia.
Laboratoire de Me
´orologie Dynamique and Institut
Pierre Simon Laplace (LMD/IPSL), CNRS, Universite
´Pierre et Marie Curie, Paris 75252, France.
2 JANUARY 2014 | VOL 505 | NATURE | 37
Macmillan Publishers Limited. All rights reserved
the small-scale part, and show that both phenomena progressively dry
the boundary layer as climate warms.
The small-scale component of mixing
Lower-tropospheric mixing parametrized within a GCM grid cell
cannot be directly diagnosed from model output (although it contri-
butes to the convective terms in the water vapour budget; see below).
We assert, however, that an atmosphere’s propensity to generate such
mixing can be gauged by observing the thermal structure just above
the boundary layer in ascending, raining regions. As discussed above,
air there is either transported directly from the boundary layer with
minimal precipitation via lower-tropospheric mixing, or indirectly by
ascending in deeper, raining clouds and then descending. Theair would
arrive cool and humid in the former case, but warmer and drier in the
latter case owing to the extra condensation, allowing us to evaluate
which pathway dominates by observing mean-state air properties.
To do this we use an index S, proportional to the differences DT
and DR
of temperature and relative humidity between 700 hPa
and 850 hPa (Staken as a linear combination; see Methods Summary)
averaged within a broad ascending region which roughly coincides
with the region of highest Indo-Pacific ocean temperatures (the Indo-
Pacific Warm Pool; Fig. 1). Of the full set of 48 models used in this
study, those with a less negative DT
in this region consistently
show a more negative DR
there (Fig. 2a), and the variations in
each quantity are quite large. We interpret this as strong evidence that
both quantities are dominated by variations, evidently large, in the
amount of lower-tropospheric mixing in the ascent region, with higher
Sindicating stronger mixing.
Small-scale lower-tropospheric mixing of moisture is part of the
overall source of the water vapour that is associated with the para-
metrized convection, M
. This quantity is available from nine of
the models (see Methods Summary). It always exhibits strong drying
near the surface. Above about 850 hPa, it can either dry the atmo-
sphere on average or moisten it depending on the model (Extended Data
Fig. 2), reflecting the competition between drying from condensation
and moistening from lower-tropospheric mixing and from evaporat-
ing precipitation falling from higher altitudes.
Although M
does not reflect lower-tropospheric mixing alone,
we can test whether lower-tropospheric mixing (as diagnosed from S)
affects how M
responds as climate warms. The available data
confirm that, given a 14 K warming, convective drying of the plan-
etary boundary layer increases by 4–17 W m
(6–30%), compared to
a typical increase of 8% in global or tropical surface evaporation. The
drying increase is highly correlated (r520.79) with S(Fig. 2b). Thus,
convective dehydration of the planetary boundary layer outstrips the
increase in surface evaporation with warming, in all models except
those with the lowest S. Higher-sensitivity models also have higher S
(Fig. 1), suggesting that this process drives a positive feedback on climate.
The large-scale component of mixing
We next turn to the large-scale lower-tropospheric mixing, which we
associate with shallow ascent or flows of air upward through the top of
the boundary layer that diverge horizontally before reaching the
upper troposphere. Although air ascending on large scales over warm
tropical oceans typically passes through nearly the whole troposphere,
over cooler oceans its ascent often wanes with altitude, showing that
this type of mixing indeed occurs in the Earth’s atmosphere (Fig. 3).
The associated mid-level outflows are well documented for the central
a Low sensitivity
s (unitless)
b High sensitivity
Figure 1
Multimodel-mean local stratification parameter
.The index Sis
the mean of swithin the regions outlined in white. Multimodel averages of sare
shown separately for low-sensitivity (ECS ,3.0 uC) (a) and high-sensitivity
(ECS .3.5 uC) (b) models, amongcoupled models with known ECS. The white
dots inside the S-averaging region show the locations of radiosonde stations
used to help estimate Sobservationally. A few coastal regions that are off-scale
appear white.
–11 –10 –9 –8 –7 –6
ΔT700–850 (K)
ΔR700–850 (%)
r = –0.76
0.30 0.35 0.40 0.45 0.50
+4 K change in planetary
boundary layer Msmall (W m–2)
r = –0.79
Figure 2
Basis for the index
of small-scale lower-tropospheric mixing
and its relationship to the warming response. a,DT
versus DR
each averaged over a tropical region of mean ascent (see Fig. 1), from all 48
coupled models. For reference, a saturated-adiabatic value of DTis shown by
dotted line at 27.2 K, and a dry-adiabatic value (not shown) would be about
216 K. Error bars are 2sranges. b, Change in small-scale moisture source
below 850 hPa in the tropics upon 14 K ocean warming, versus S
computed from the control run, in eight atmosphere models and one CMIP3
model. Symbol colour indicates modelling centre or centre where atmosphere
model was originally developed and symbol shape indicates model generation.
Macmillan Publishers Limited. All rights reserved
and eastern Pacific and Atlantic Intertropical Convergence Zone and
some monsoon circulations
. Although these are indeed the regions
where shallow ascent is steadiest, and hence clearest in monthly-mean
data (Fig. 3), in daily reanalysis data, shallow ascent is equally strong
outside the tropics owing largely to contributions from extratropical
storms. We also note that although we focus here on regions of
ascending air, that is because the ascending branches are where the
circulations are easiest to measure; they must, however, descend else-
where, exerting a net transport of water vapour that is upward and
towards the convective regions.
Figure 3 compares the observations with two example models. Neither
model shows as much shallow ascent (red colour) as the observation-
based estimates, but the Institut Pierre Simon Laplace (IPSL)-CM5A
model comes closer. Althoughconvective treatment in the newer IPSL-
CM5B model is more detailed and produces better results in important
, here it is seen to produce strong deep ascent of air (white
spots) where it is weaker and shallower in observations (red zones),
showing that improvement in some aspects of a simulation does not
automatically improve others.
We quantify the large-scale lower-tropospheric mixing more thor-
oughly by calculating the ratio Dof shallow to deep overturning (see
Methods Summary) in a broad region encompassing most of the
persistent shallow ascent (see Fig. 3). This index Dvaries by a factor
of four across 43 GCMs (see below). Interestingly, however, Dand S
are uncorrelated (r50.01), confirming that the two scales of mixing
are controlled by different aspects of model design.
The effective source of moisture M
LT, large
due to this shallow over-
turning (that is, large-scale, lower-tropospheric convection) and its
change upon climate warming, can be directly calculated from model
wind and humidity fields. We approximate M
LT, large
using monthly-
mean data from the ten available atmospheric models (see Methods
Summary). M
LT, large
isolates only shallow mixing, whereas M
includes the effects of all parameterized convection; yet despite this,
the profiles M
LT, large
(Fig. 4) resemble those of M
, with strong
drying in the boundary layer and weak moistening above. Not unex-
pectedly, these effects are greater in the high-Dmodels than in the
Crucially, the low-level drying also increases faster upon 14K
warming in the high-Dmodels (by about 30%, or 1.5 W m
when expressed as a latent heat flux) than in the low-Dmodels (25%,
or 0.9 W m
). Thus, the response of M
LT, large
grows with Das
grew with S; the relationship for Dis not as strong (r50.46 for
land 1ocean, r50.25 for ocean only), partly because the spread of D
happens to be somewhat narrow among the available atmosphere
models, but is still significant at 95% confidence.
Climate sensitivity
We now apply the indices Sand Dto the 43 GCMs for which an
equilibrium climate sensitivity (ECS) is available. Each index inde-
pendently explains about 25% of the variance in ECS (Fig. 5a, b).
Because the ranges of Dand Sare similar (each 0.3–0.4), as are
(approximately) those of their drying responses upon warming (see
below), we form an overall lower-tropospheric mixing index (the
LTMI) by simply adding the two: LTMI 5S1D. This LTMI explains
about 50% of the variance in total system feedback (r50.70) and ECS
(r50.68) (Fig. 5c). Thus, although our measure of lower-tropospheric
mixing does not explain all of the variations among GCMs, it does
explain a significant portion of the model spread.
This explanatory power derives primarily from low cloud feed-
backs. The correlation between LTMI and the 14 K change in short-
wave cloud radiative effect in the atmosphere models, which spans a
range of 1.8 W m
in the tropics, is 0.65 in the tropics and 0.57
in subsidence regions (equivalent values estimated from a subset of
the coupled models providing the needed output are 0.25 and 0.47
ω500 (hPa day–1)
140 0 –140
ω850 (hPa day–1)
Figure 3
The structure of monthly-meantropospheric ascent reveals large-
scale lower-tropospheric mixing in observations and models. Upward
pressure velocity vin one month (September) from the MERRA reanalysis
(a), the IPSL-CM5A model (b) and the IPSL-CM5B model (c) with values at
850 hPa shown in red and those at 500 hPa shown in green plus blue. Bright
red implies ascent that is weighted toward the lower troposphere with
mid-tropospheric divergence (see colour scale), white implies deep ascent, and
dark colours imply descent. In a, black lines outline the region in which the
index Dof large-scale lower-tropospheric mixing is computed. The Pacific and
Atlantic Intertropical Convergence Zone regions are consistently red in the
reanalyses and models, whereas isolated red patches in other areas tend to vary
with time.
2 JANUARY 2014 | VOL 505 | NATURE | 39
Macmillan Publishers Limited. All rights reserved
respectively). These correlations suggest that the predictive skill of
LTMI arises from both subsidence and other regions; further work
is needed to better assess this. Cloud amount reduces more in high-
LTMI models both at low and mid-levels (Extended Data Fig. 3),
although the greater net radiative impact of low cloud makes its effect
. Previously reported water vapour and lapse-rate feed-
are, in contrast, not correlated with the LTMI.
Is the imputed lower-tropospheric mixing impact on low clouds
strong enough to explain the approximately 1.5 W m
spread of
cloud feedbacks seen in GCMs?
One recent study
imposed increased
surface latent heat fluxes in a large region typified by shallow clouds,
finding an increase in cloud-related net cooling of about 1 W m
for a
2–3 W m
increase in the surface flux, other things held fixed. An
even larger sensitivity, nearly 1:1, has been reported in a different
model for advective changes in moisture input
. If a similar but opposite
cloud response occurred for moisture removal by lower-tropospheric
mixing, then to explain the feedback spread, the boundary-layer drying
responses would need to span a range across models of about 3 W m
per K of surface warming. This roughly matches the contributionto the
spread from M
alone (Fig. 2b). The additional drying response
from M
LT, large
was about 0.6 W m
greater in the high-Dmodels
(mean Dof 0.34) than in the low-Dones (mean 0.24), which, if rescaled
by the full spread of Din the full GCM ensemble, implies a further
source of spread in drying response of about 2 Wm
. We con-
clude that, even if not all low clouds are as sensitive as the ones exam-
ined in the cited studies, the lower-tropospheric mixing response is
strong enough to account for the cloud feedback spread and its typ-
ically positive sign
Why does moisture transport increase so strongly with warming?
The magnitude of these increases, typically 5%–7% per K of surface
warming, is roughly what would be expected if the circulations remained
similar against a Clausius–Clapeyron increase in moisture gradients
as indeed it does,at least for the large-scale part
(ExtendedData Fig. 4).
Further study is needed to understand why this is so, and to examine in
greater detail how clouds respond to changing moisture transports;
changes in low cloud amount may for example help the atmosphere
restore imbalances in boundary layer moist enthalpy such as those caused
by lower-tropospheric mixing
. Because LTMI ignores any information
on clouds, it is likely that additional measures of cloud characteristics
could explain some of the variations in low-cloud feedback not yet
explained here.
We end by considering observational estimates of Sand D(see
Fig. 5). These show an Snear the middle of the GCM range, but a
Dclose to the top end, as hinted already by Fig. 3. Dmay not be well
constrained because vmust be inferred from observational reana-
lyses, although available horizontal wind observations support the
existence of strong mid-level outflows
, and the result is consistent
across both reanalyses examined. The reanalysis estimates of Sare less
consistent but this quantity can be fairly well constrained by radio-
sonde observations.
Taking the available observations at face value implies a most likely
climate sensitivity of about 4 uC, with a lower limit of about 3 uC.
Indeed, all 15 of the GCMs with ECS below 3.0 uC have an LTMI
below the bottom of the observational range. Further work may be
needed to better constrain these indices, and to test whether their
relationship to ECS is robust to design factors common to all models.
For example, this should be tested in global cloud-resolving models.
–0.8 –0.6 –0.4 –0.2 0.0 0.2
e-scale source (
–1 day–1)
Pressure (hPa)
High D
Low D
High D, +4 K
Low D, +4 K
Figure 4
Estimated water vapour source M
LT, large
due to large-scale lower-
tropospheric mixing and its response to warming. See Methods for
calculation details. Data are from ten atmosphere models, averaged from 30uS
to 30uN over oceans, with the average of the four models having the largest D
shown in magenta and the average of the four models with the smallest D
shown in blue. Dashes show results in 14 K climate. Changes at 14 K are
nearly identical whether or not land areas are included.
0.2 0.3 0.4 0.5 0.6 0.7
Climate sensitivity
r = 0.50
0.1 0.2 0.3 0.4 0.5
Climate sensitivity
r = 0.46
0.4 0.5 0.6 0.7 0.8 0.9 1.0
LTMI, (S + D)
Climate sensitivity
r = 0.68/0.70
Figure 5
Relation of lower-tropospheric mixing indices to ECS. ECS versus
S(a), D(b) and LTMI 5S1D(c) from the 43 coupled models with known
ECS. Linear correlationcoefficients rare given in each panel (r50.70 in cis the
correlation to the total system feedback). Error bars shown near panel axes
indicate 2sranges of the direct radiosonde estimate (a) and the Svalue from
radiosondes added to the Dvalue from each of the two reanalyses(c). ERAi and
MERRA are the two monthly reanalysis products.
Macmillan Publishers Limited. All rights reserved
The possibility can never be ruled out that feedbacks could exist in
nature that are missing from all models, which would change the
climate sensitivity from that suggested by our result. Nonetheless,
on the basis of the available data, the new understanding presented
here pushes the likely long-term global warming towards the upper
end of model ranges.
Although a few previous studies have already noted that higher-sensitivity
models simulate certain cloud-relevant phenomena better
, ours is
the first to demonstrate a causal physical mechanism, or to show
consistent predictive skill across so many models, or to point to pro-
cesses connecting low-cloud regions to the deep tropics. The MILC
mechanismis surprisingly straightforward. Lower-tropospheric mixing
dries the boundary layer, and the drying rate increases by 5–7% K
warmer climates owing to stronger vertical water vapour gradients.
The moisture source from surface evaporation increases at only about
2% K
. Thus as climate warms, any drying by lower-tropospheric
mixing becomes larger relative to the rest of the hydrological cycle,
tending to dry the boundary layer. How important this is depends on
how important the diagnosed lower-tropospheric mixing was in the
base state of the atmosphere. Lower-tropospheric mixing is unrealist-
ically weak in models that have low climate sensitivity.
Climate-sensitivity-related differences in lower-tropospheric mix-
ing, both at small (Fig. 1) and large scales (Fig. 3), are most detectable
in regions of tropical deep or mixed-level convection and mean upward
motion. This does not mean, however, thatthe greater low-level drying
in a warmer climate or the spread of drying among models will be
limited to these regions.
Large-scale lower-tropospheric mixing carries water vapour not only
upward but also horizontally away from subsidence regions; because both
directions of transport intensify in a warmer atmosphere
, subsidence
regions should bear the brunt of the overall boundary-layer drying.
Moreover, shallow ascent is equally strong (though more transient) in
mid-latitude storm tracks and in the tropics, suggesting that MILC
feedback may be just as important outside the tropics as within them.
As for small-scale lower-tropospheric mixing, even though there are
reasons to measure it in ascending regions (see Methods), its impact
upon warming is much more widespread and differs significantly among
models in subsiding regions (Extended Data Fig. 5). We hypothesise
that this is because models with more small-scale lower-tropospheric
mixing in ascending regions also have more in descending regions,
although we cannot confirm this directly. Overall, the behaviour is con-
sistent withpublished results showing that subsiding regions contribute
strongly to the spread of cloud feedbacks in models, with storm tracks
and tropical convective regions also playing a part
Lower-tropospheric mixing behaviour appears to result from a
competition between shallow and deep convection in situations where
either could occur. Such situations persist in many tropical regions,
notably the Intertropical Convergence Zone. Understanding and
properly representing this competition in climate models is undoubt-
edly necessary for more accurate future climate projections.
Although tested here on models used over the past decade or so, we
presume that this mechanism has been a leading source of spread in
sensitivity since the dawn of climate modelling. Finally to identify an
atmospheric process that drives variations in climate sensitivity offers an
unprecedented opportunity to focus research and model development
in ways that should lead to more reliable climate change assessments.
Data for computing Sand Dcome from control runs of 48 models: 18 from the
Coupled Model Intercomparison Project version 3 (CMIP3)
and 30 from
CMIP5 (ref. 29) (see Extended Data Tables 1 and 2). ECS was reported for all
but one CMIP3 model by the Intergovernmental Panel on Climate Change
. For
CMIP5 we employ effective climate sensitivities calculated from abrupt 4 3CO
experiments, availablefor 26 models, following a standard regression procedure
Data for M
and M
LT, large
come from ten CMIP5 atmospheremodels providing
‘amip’ (specified ocean surface temperature) control and 14 K ocean warming
runs. Eight of these models providedM
; we also included data fromthe Parallel
Climate Model (CMIP3).
Observational estimates come from radiosondes and two monthly reanalysis
products (ERAi and MERRA). Reanalyses are produced from a model con-
strained to the fullest extent possible by a variety of observations
We calculate Swithin a region where convective effects are a leading term in
thermodynamic budgets, defined by the upper quartile of the annual-mean mid-
tropospheric ascent rate where it is upward, 2v
(vthe pressure velocity). We
define S;(DR
/100% 2DT
/9 K)/2, which normalizes DR
100% humidity, DT
to the approximately 9-K range between dry and
saturated adiabatic values, and averages these two pieces of information with
equal weight to reduce noise from other factors.
To calculate M
LT, large
we compute v
(the average of vat 850 hPa and 700 hPa)
and v
(the average of vat 600 hPa, 500 hPa and 400 hPa). D5v
the local horizontal outflow in the lower troposphere above the boundary layer.
Moisture is transported upward and outward wherever D.0 and v
,0. We
restrict measurement to tropical ocean regions from 160uWto30uE (see Fig. 3).
The moisture supplied to the environment is estimated as M
LT, large
)æ, where pis the pressure, qis the specific humidity, Æ...æindicates
the mean over the restricted region, and His the step function. Finally,
Online Content Any additional Methods, Extended Data display items and Source
Data are available in the online version of the paper; references unique to these
sections appear only in the online paper.
Received 16 May; accepted 5 November 2013.
1. Hegerl, G. C. & Zwiers, F. W. Working Group I: The Physical Science Basis of Climate
Change Ch. 9, 663–745 (Cambridge Univ. Press, 2007).
2. Cess, R. D. et al. Intercomparison and interpretation of climate feedbackprocesses
in 19 atmospheric general circulation models. J. Geophys. Res. 95, 16601–16615
3. Dufresne,J. L. & Bony, S. An assessmentof the primary sources of spreadof global
warming estimates from coupled atmosphere-ocean models. J. Clim. 21,
5135–5144 (2008).
4. Soden, B. J. et al. Quantifying climate feedbacks using radiativekernels. J. Clim. 21,
3504–3520 (2008).
5. Zelinka, M., Klein, S. & Hartmann, D.Computing and partitioning cloud feedbacks
using cloud property histograms. Part I: cloud radiative kernels. J. Clim. 25,
3715–3735 (2012).
6. Hartmann, D. L. & Short, D. A. On the use of earth radiation budget statistics for
studies of clouds and climate. J. Atmos. Sci. 37, 1233–1250 (1980).
7. Bony, S. & Dufresne, J. L. Marine boundary layer clouds at the heart of tropical
cloud feedback uncertainties in climate models. Geophys. Res. Lett. 32, L20806
8. Rieck, M., Nuijens, L. & Stevens, B. Marine boundary-layer cloud feedbacks in a
constant relative humidity atmosphere. J. Atmos. Sci. 69, 2538–2550 (2012).
9. Stevens, B. On the growth of layers of nonprecipitating cumulus convection.
J. Atmos. Sci. 64, 2916–2931 (2007).
10. Yano, J. & Emanuel, K. An improved model of the equatorial troposphere and its
coupling with the stratosphere. J. Atmos. Sci. 48, 377–389 (1991).
11. Sherwood, S. C. & Meyer, C. L. The general circulation and robust relativehumidity.
J. Clim. 19, 6278–6290 (2006).
12. Allen, M. R. & Ingram, W. J. Constraints on future changes in climate and the
hydrologic cycle. Nature 419, 224–232 (2002).
13. Zhang, C. D., Nolan, D. S., Thorncroft, C. D. & Nguyen, H. Shallow meridional
circulations in the tropical atmosphere. J. Clim. 21, 3453–3470 (2008).
14. Nolan, D. S., Powell, S. W., Zhang, C. D. & Mapes, B. E. Idealized simulations of the
Intertropical Convergence Zone and its multilevel flows. J. Atmos. Sci. 67,
4028–4053 (2010).
15. Rio, C., Hourdin, F., Grandpeix, J. Y. & Lafore, J. P. Shifting the diurnal cycle
of parameterized deep convection over land. Geophys. Res. Lett. 36, L07809
16. Soden, B. J. & Vecchi, G. A. The vertical distribution of cloud feedback in coupled
ocean-atmosphere models. Geophys. Res. Lett. 38, L12704 (2011).
17. Soden, B. J. & Held, I. M. An assessment of climate feedbacks in coupled ocean-
atmosphere models. J. Clim. 19, 3354–3360 (2006).
18. Webb, M. & Lock, A. Coupling between subtropical cloud feedback and the local
hydrological cycle in a climate model. Clim. Dyn. 41, 1923–1939 (2013).
19. Brient, F. & Bony, S. Interpretation of the positive low-cloud feedback predicted by
a climate model under global warming. Clim. Dyn. 40, 2415–2431 (2013).
20. Held, I. M. & Soden, B. J. Robust responses of the hydrological cycle to global
warming. J. Clim. 19, 5686–5699 (2006).
21. Zahn, M. & Allan, R. P. Climate warming-related strengthening of the tropical
hydrological cycle. J. Clim. 26, 562–574 (2013).
22. Klein, S. A. et al. Are climate modelsimulations of clouds improving? An evaluation
using the ISCCP simulator. J. Geophys. Res. 118, 1329–1342 (2013).
2 JANUARY 2014 | VOL 505 | NATURE | 41
Macmillan Publishers Limited. All rights reserved
23. Volodin, E. M. Relationbetween temperature sensitivity to doubled carbondioxide
and the distribution of clouds in current climate models. Izv. Atmos. Ocean. Phys.
44, 288–299 (2008).
24. Clement, A. C., Burgman, R. & Norris, J. R. Observational and model evidence for
positive low-level cloud feedback. Science 325, 460–464 (2009).
25. Fasullo, J. T. & Trenberth, K. E. A less cloudy future: the role of subtropical
subsidence in climate sensitivity. Science 338, 792–794 (2012).
26. Webb, M. J., Lambert, F. H. & Gregory, J. M. Origins of differences in climate
sensitivity, forcingand feedback in climatemodels. Clim. Dyn.40, 677–707 (2013).
27. Vial, J., Dufresne, J.-L. & Bony, S. On the interpretation of inter-model spread in
CMIP5 climate sensitivity estimates. Clim. Dyn. 41, 3339–3362 (2013).
28. Randall, D. A. & Wood, R. A. Working Group I: The Physical Science Basis of Climate
Change Ch. 8, 589–662 (Cambridge University Press, 2007).
29. Taylor, K. E., Stouffer, R. J. & Meehl,G. A. An overview of CMIP5 and the experiment
design. Bull. Am. Meteorol. Soc. 93, 485–498 (2012).
30. Gregory, J. M. et al. A new method for diagnosing radiative forcing and climate
sensitivity. Geophys. Res. Lett. 31, L03205 (2004).
31. Forster, P. M. et al. Evaluatingadjusted forcing and model spreadfor historical and
future scenarios in the CMIP5 generation ofclimate models. J. Geophys. Res. 118,
1139–1150 (2013).
32. Uppala, S. M. et al. The ERA-40 re-analysis. Q. J. R. Meteorol. Soc. 131, 2961–3012
33. Rienecker, M. M. et al. MERRA: NASA’s modern-era retrospective analysis for
research and applications. J. Clim. 24, 3624–3648 (2011).
Acknowledgements This work was supported by the FP7-ENV-2009-1 European
project EUCLIPSE (number 244067). We acknowledge the World Climate Research
Programme’s Working Group on Coupled Modelling, which is responsible for CMIP,
and we thank the climate modelling groups for producing and making available their
model output, especially the participants contributing additional CFMIP2experiments
and diagnostics crucial to our study. The US Department of Energy’s Program for
Climate ModelDiagnosis and Intercomparison providescoordinating supportfor CMIP
and led the development of software infrastructure in partnership with the Global
Organisation for Earth System Science Portals. We also thank the National Center for
Atmospheric Research and the Earth System Grid Federation for providing access to
PCM output, the Australian National Computational Infrastructure, and the IPSL
Prodiguer-Ciclad facility for providing a convenient archive of CMIP data. Finally, we
thank B. Stevens, C. Bretherton and G. Schmidt for comments on early versions of the
Author Contributions S.C.S. led the study and the writing of the paper, and did the
calculations of LTMI and related diagnostics.S.B. computed cloud radiative effect and
assisted in interpreting results and writing the paper. J.-L.D. computed ECS and
assisted in interpreting results and writing the paper.
Author Information Reprints and permissions information is available at The authors declare no competing financial interests.
Readers are welcome to comment on the onlineversion of the paper. Correspondence
and requests for materials should be addressed to S.C.S. (
Macmillan Publishers Limited. All rights reserved
Data for computing Sand Dcome from 48 models: 18 from the CMIP3 (Coupled
Model Intercomparison Project version 3)
, the first two years of each ‘‘picntrl’’
run, and 30 models from theCMIP5 (ref. 29), the first two years of each ‘‘1pctCO2’’
run. Two years of datais sufficient to specify Sand Dto within 0.02 or betterof their
long-term values. CMIP3 data were obtained from the Australian National
Computational Infrastructure node, and CMIP5 data including the ‘amip’ and
‘amip14K’ runs were obtained on 14 September 2012 and 22 October 2012 from
the IPSL Ciclad repository. ECS values for CMIP3 were reported for all but one
model bythe Intergovernmental Panel on ClimateChange
. For CMIP5 we employ
effective climate sensitivities calculated from abrupt 4 3CO
experiments, avail-
able for 26 of the 30 CMIP5 models, followinga standard regression procedure
Data for M
and M
LT, large
come from ten CMIP5 atmosphere models pro-
viding ‘amip’ (specified ocean surface temperature) control and 14 K ocean
warming experiments. A key advantage of this experiment setup is that inter-
annual ocean variability is the same in the control and warming runs, and changes
in the sea surface temperature pattern—which could complicate interpretation,
especially for circulation changes—are avoided. Data are from 1989–98, except
for IPSL-CM5A, in which some of these years were corrupted and alternative
years were used. Results from individual years were similar to those for the ten-
year averages. Eight of these models provided M
; we also included data from
the PCM CMIP3 1%-per-year-to-quadrupling experiment, with changes rescaled
to the 14 K equivalent (actualchange 3.3 K). PCM M
data come from tenyears
near the beginning and ten years near the end of the 1%-per-year-to-quadrupling
experiment, obtainedfrom the National Center for AtmosphericResearch node of
the Earth System Grid.
The shortwave cloud radiative effect is obtained by differencing the all-sky and
clear-sky top-of-atmosphere shortwave fluxes for each model run. To calculate
cloud feedback we first composite the sensitivity of the shortwave cloud radiative
effect to sea surface temperature in dynamical regimes defined by vertical-mean
vertical velocity, and then we compute the sum (weighted by the probability
distribution function of v) over regimes (or only subsidence regimes defined
by v.0)
. For coupled models, the warming-induced change is obtained from
abrupt CO
-quadrupling experiments, after removing the instantaneous change
associated with rapidadjustment to higher CO
estimated from thefirst 12 months
after quadrupling. Only one realization is used per model. For atmosphere-only
models it is simply the difference between the 14K and the control simulations.
Observational estimates come from radiosondes and from two monthly reana-
lysis products (ERAi and MERRA), years 2009–10. The reanalyses are produced
from a model constrainedto the full extent possible by a variety of observations
MERRA reanalysisdata from 1 September 2009 were used to compare Dinside and
outside the tropics, but monthly data were used otherwise. Radiosonde data were
obtained from the Integrated Global Radiosonde Archive and subjected to simple
quality-controlchecks for outliers. The ten stations sited in the relevant region and
meeting the criteria describedby a previous study
were used, and the mean taken
over the 2 years. The radiosonde network sampling bias, as determined from
station-sampled reanalysis output, was relatively small compared to the overall
reanalysis biases.
We calculate Sin ascending regions, where convective effects are a leading term
in thermodynamic budgets; in subsidence regions humidity is sensitive to irrel-
evant non-local factors and even to numerical resolution
, perhaps explaining
why it is less informative for our purposes. The calculation region is defined by the
upper quartile of the annual-mean mid-tropospheric ascent rate in ascending
regions, 2v
(where vis the pressure velocity). We define S;(DR
100% 2DT
/9 K)/2, which normalizes DR
to 100% humidity and
to the approximately 9 K range between the dry and saturated adiabatic
values, and then averages these two pieces of information with equal weight. Such
averaging should reduce the noise from other factors that influence one quantity
or the other. Varyingthe weighting of the two terms does notstrongly affect results.
To calculate M
LT, large
, we first compute v
(the average vat 850 hPa and
700 hPa) and v
(the average vat 600 hPa, 500 hPa and 400 hPa). The difference
then measures the local horizontal outflow in the lower troposphere
above the boundary layer. Moisture is transported upward and outward at this
level wherever D.0 and v
,0. We restrict measurement to tropical ocean
regions from 160uWto30uE (see Fig. 3). The moisture supplied to the envir-
onment is then estimated as M
LT, large
)æ, where qis the
specific humidity, Æ...æindicates the mean over the restricted calculation region,
and His the step function. The index Dis computed as D;ÆDH(D)H(2v
Values of Dand Sare similar over ten years of data or one year, and are similar
whether individual months or long-term means for each month of the year are
used. These indices capture over 25% of the ECS variance even if computed from
only a single month of data from each model. Thus, long records are unnecessary
for deducing the strength of lower-tropospheric mixing.
The reason for restricting calculation of Dto the cooler tropical longitudes is
that a few climate models erroneously place much of the shallow ascent over
warm oceans, where it does not seem to contribute as much to low-cloud feed-
back. In observations, and in most models, the restriction has little effect because
most of the shallow ascent persistent enough to appear in monthly-mean data is
already located in the specified region. We speculate that the location of the ascent
matters because the associated shallow descent is more relevant if it occurs over,
or upstream of, regions of radiatively important low cloud.
Both lower-tropospheric mixing indices retain statistically significant correla-
tions with ECS for all alterations to their definitions that we tried. Specifically, the
correlation of Swith ECS (r
) is similar with v
percentiles of 0.25 or 0.5,
but drops with looser thresholds, which begin to pick up parts of the resolved
lower-tropospheric mixing region. Tighter thresholds reduce the spread in S
between models, reducing r
. The correlation r
is somewhat weaker
(as low as 0.3) if the longitudinal restriction for Dis removed, or if other defini-
tions of v
and v
are used.
34. Sherwood, S. C., Meyer, M. L., Allen, R. J. & Titchner, H. A. Robust tropospheric
warming revealed by iteratively homogenized radiosonde data. J. Clim. 21,
5336–5352 (2008).
35. Sherwood, S. C., Roca, R., Weckwerth, T. M. & Andronova, N. G.
Tropospheric water vapor, convection and climate. Rev. Geophys. 48, RG2001
Macmillan Publishers Limited. All rights reserved
Extended Data Figure 1
Illustration of atmospheric overturning
circulations. Deep overturning strongly coupled to the hydrological cycle and
atmospheric energy budget isshown by solid lines; lower-tropospheric mixing
is shown by dashed lines. The MILC feedback results from the increasing
relative role of lower-tropospheric mixing in exporting humidity from the
boundary layer as the climate warms, thus depleting the layer of water vapour
needed to sustain low cloud cover.
Macmillan Publishers Limited. All rights reserved
Extended Data Figure 2
Small-scale moisture source
profile averaged over all tropical oceans, for two selected climate models (see
legend) with very differentwarming responses, in present-day (solid) and 14K
(dashed) climates.
Macmillan Publishers Limited. All rights reserved
Extended Data Figure 3
Response of cloud fraction to warming. Profile of
average change in model cloud fractional cover at 14 K in the four atmosphere
models with largest (magenta) and smallest (blue) estimated 14K increases in
planetary-boundary-layer drying, averaged from 30uSto30uN (dashed) or
60uSto60uN (solid). The drying estimate is obtained by adding the explicitly
computed change in M
LT, large
to the change in M
estimated from Svia the
relationship shown inFig. 2a. The typical mean cloud fraction below 850 hPa is
about 10% to 20%, and the changes shown are absolute changes in this fraction,
so are of the order of 10% of the initial cloud cover.
Macmillan Publishers Limited. All rights reserved
Extended Data Figure 4
Response of large-scale lower-tropospheric
mixing to warming. Profiles of mean vertical velocity in regions of shallow
ascent, in control and 14 K climates. The similarity of dashed and solid lines
indicates that mass overturning associated with these regions is roughly the
same in the warmer simulations, on average.
Macmillan Publishers Limited. All rights reserved
Extended Data Figure 5
Response of small-scale, low-level drying to
warming. Change in convective moisture source M
below 850 hPa upon a
14 K warming in eight atmosphere models and one CMIP3 coupled model;
units are W m
, with negative values indicating stronger drying near the
surface. Zero contours are shown in white (a few off-scale regions also appear
white). The models used for calculating M
are the eight shown here plus two
for which M
data were unavailable: CNRM-CM5 and FGOALS-g2.
Macmillan Publishers Limited. All rights reserved
Extended Data Table 1
List of CMIP5 coupled models used
Centre acronyms used to identify them in scatter plots are also shown. The derived forcing, total feedback, and equilibrium climate sensitivities are given for models with abrupt 4 3CO
Macmillan Publishers Limited. All rights reserved
Extended Data Table 2
List of CMIP3 coupled models used
Centre acronyms used to identify them in scatter plots are also shown, as are feedback values given by
ref. 28.
Macmillan Publishers Limited. All rights reserved
... In particular, low-cloud feedbacks have been shown to contribute to the simulated increase in equilibrium climate sensitivity in the latest generation of coupled general circulation models (GCMs; Zelinka et al., 2020). A key physical process modulating low clouds is shallow convective mixing between the boundary layer and the free troposphere, which may explain up to half of the variance in climate sensitivity estimates from a collection of CMIP3 and CMIP5 models (Sherwood et al., 2014). Shallow convective mixing transports moisture upward, moistening the lower troposphere and drying the boundary layer; this potentially suppresses further developments of low-level clouds, allowing shortwave radiation to pass through the atmosphere and warm the Earth's surface. ...
... A number of methods can potentially be applied to estimate shallow convective mixing over large spatial scales. For example, Sherwood et al. (2014) used the difference of temperature and relative humidity between 700 and 850 hPa to estimate shallow convective mixing in climate models. However, shortwave and longwave radiation influence temperature and relative humidity, potentially obfuscating the mixing signal. ...
... All satellite retrievals are binned and monthly averages are computed over 5° × 5° grid boxes, which have measurement uncertainty no larger than 8-10‰. Finally, we use the ERA5 reanalysis (Hersbach et al., 2020) to compare the δD-diagnosed convective mixing estimated in this study to the estimates defined by Sherwood et al. (2014). ...
Full-text available
Low‐cloud feedbacks contribute large uncertainties to climate projections and estimated climate sensitivity. A key physical process modulating low‐cloud feedbacks is shallow convective mixing between the boundary layer and the free troposphere. However, there are challenges in acquiring observational constraints of shallow convective mixing with global coverage. To this end, we propose a novel approach to constraining convective mixing using stable water vapor isotope profiles from satellite retrievals. We demonstrate that the vertical gradient of water vapor δD between the boundary layer and free troposphere can be used to track shallow convective mixing. Analyzing isotopes in water vapor alongside low‐cloud properties from satellite retrievals, we find that low‐cloud fraction appears largely insensitive to convective mixing in shallow cumulus regions. Our results also suggest that strong shallow convective mixing is associated with the moistening of the free troposphere. The new estimate of shallow convective mixing and its relationship with low‐cloud properties offers a potential constraint on simulations of low‐cloud feedbacks and estimates of climate sensitivity.
... Convective parameterizations are often rightly considered as first among many sources of atmosphere model uncertainty (e.g., Sanderson et al., 2008;Sherwood et al., 2014). Deep convection is not the only challenge: unsaturated turbulence in the subcloud layer and elsewhere, shallow cumulus convection, and even grid-resolved overturning circulations also qualify as "convection." ...
... Cloud impacts on radiative transfer comprise a challenging set of processes at both climatic and intraseasonal time scales, but often are causally downstream of convection. For instance, Sherwood et al. (2014) found that the differences in climate sensitivity among 43 climate models, 10.1029/2021MS002826 4 of 30 ascribable to the low-cloud radiative feedback, can largely be further attributed to differences in the simulated strength of "mixing" between the lower and middle tropical troposphere, by both parameterized and resolved motions. Recognizing these subtleties of the term "convection," in this study we find that many SCAM behaviors are traceable mainly to deep convection, handled by a single scheme or algorithm (the Zhang-McFarlane scheme (G. ...
Full-text available
To explore the interactions among column processes in the Community Atmosphere Model (CAM), the single‐column version of CAM (SCAM) is integrated for 1000 days in radiative‐convective equilibrium (RCE) with tropical values of boundary conditions, spanning a parameter or configuration space of model physics versions (v5 vs. v6), vertical resolution (standard and 60 levels), sea surface temperature (SST), and some interpretation‐driven experiments. The simulated time‐mean climate is reasonable, near observations and RCE of a cyclic cloud‐resolving model. Updraft detrainment in the deep convection scheme produces distinctive grid‐scale structures in humidity and cloud, which also interact with radiative transfer processes. These grid artifacts average out in multi‐column RCE results reported elsewhere, illustrating the nuts‐and‐bolts interpretability that SCAM adds to the hierarchy of model configurations. Multi‐day oscillations of precipitation arise from descent of warm convection‐capping layers starting near the tropopause, eventually reset by a burst of convective deepening. Experiments reveal how these oscillations depend critically on an internal parameter that controls the number of neutral buoyancy levels allowed for determining cloud top and computing dilute convective available potential energy in the deep convection scheme, and merely modified a little by disabling cloud‐base radiation (heating of cloud base). This strong dependence of transient behavior in 1D on this parameter will be tested in the second part of this work, in which SCAM is coupled to a parameterized dynamics of two‐dimensional, linearized gravity wave, and in the 3D simulations in future study.
... To reduce the uncertainty in the future drought projections, previous studies [3,6] suggested to use better performing GCMs [high-skill GCMs or best GCMs, which capture the critical monsoon features and show less bias during the historical climate; [3,6]] for the future drought projections; as the majority of GCMs are unable to capture the summer monsoon features (i.e., seasonality, onset timing of monsoon, intensity) and show a significant bias [7,66,77]. Other than the model selection approach, further improvements in the physics of GCMs are also needed for the better representation of monsoon precipitation, cloud formation, and convective precipitation [7,11,77,89]. In addition, proper representation of human activities (i.e., land use land cover) in Global climate models will improve the feedback and interaction between land and atmosphere, which will reduce the uncertainty in the drought assessment for the future climate [12]. ...
Full-text available
Drought is one of the complex and deleterious natural hazards that poses severe challenges to water security, food production, ecosystem, and socio-economic condition in India. Using efficient drought monitoring and assessment, the severe impacts of drought can be reduced. However, drought monitoring and assessment are associated with large uncertainty due to different datasets, methods, drought indices, and modeling approaches. Here, we examine the sources of uncertainty in drought assessment using multiple observational and future projections datasets, methods, and hydrological models. Moreover, we discuss potential ways to overcome the challenges associated with drought assessment in India. The drought assessment without considering uncertainty may cause overestimation or underestimation of risk in the observed and projected future climate, which further affects the planning and management of water resources. Therefore, a thorough understanding of these challenges is essential to improve the existing drought monitoring and assessment approaches.
... We use three WRF simulations driven by Community Climate System Model 4 (CCSM4, Gent et al., 2011), the Geophysical Fluid Dynamics Laboratory Earth System Model 2 (GFDL-ESM2G, Donner et al., 2011), and the Hadley Centre Global Environment Model version 2 (HadGEM2-ES, Jones et al., 2011). These three GCMs represent a range of climate sensitivities that encompasses most of the coupled model intercomparison project phase 5 (CMIP5) GCMs when projecting future temperature changes (Sherwood et al., 2014). For more details on these simulations, see ; Zobel et al. (2018a,b). ...
This study develops a statistical conditional approach to evaluate climate model performance in wind speed and direction and to project their future changes under the representative concentration pathway 8.5 scenario over inland and offshore locations across the Continental United States. The proposed conditional approach extends the scope of existing studies by characterizing the changes of the full range of the joint wind speed and direction distribution. Directional wind speed distributions are estimated using two statistical methods: a Weibull distributional regression model and a quantile regression model, both of which enforce the circular constraint to their resulting estimates of directional distributions. Projected uncertainties associated with different climate models and model internal variability are investigated and compared with the climate change signal to quantify the statistical significance of the future projections. In particular this work extends the concept of internal variability to the standard deviation and high quantiles to assess the relative magnitudes to their projected changes. The evaluation results show that the studied climate model capture both historical wind speed, wind direction, and their dependencies reasonably well over both inland and offshore locations. In the future, most of the locations show no significant changes in mean wind speeds in both winter and summer, although the changes in standard deviation and 95th-quantile show some robust changes over certain locations in winter. The proposed conditional approach enables the characterization of the directional wind speed distributions, which offers additional insights for the joint assessment of speed and direction.
... Since the radiative forcing caused by different climate change agents can be estimated from radiative transfer calculations (Clough & Iacono, 1995;Collins et al., 2006), our results imply that the adjustment responses of different schemes to these agents can be compared by simple linear algebra calculations using their LRF matrices, without having to import different schemes into the same host model. This could potentially be helpful in climate change research, where parameterizations are a major contributor to intermodel spread in climate sensitivity predictions (e.g., Geoffroy et al., 2017;Ringer et al., 2014;Sherwood et al., 2014;Webb et al., 2013). Moreover, our results also show that when convection is disorganized in MCM (hence high SCM-MCM comparability), the LRF matrices can be constructed using SCMs, thereby drastically reducing the computing overhead, yet still ensuring adequate representation of the parameterization behavior in a 3D setting. ...
Full-text available
Single‐column models (SCMs) simulations are sometimes used to evaluate model physics and aid parameterization development. However, few studies have systematically compared SCM behavior—where column boundary conditions are specified—with that of corresponding 3D models, where columns interact dynamically. Here we address this by comparing forced responses of an SCM in radiative‐convective equilibrium (RCE) with those of a multi‐column model (MCM) where the model domain is in RCE but individual columns are not, examining what factors affect the models' comparability. We find that convective organization in the MCM depends at least as much on the convection scheme as on other mechanisms known to organize convection (e.g., radiative feedback). Moreover, convective organization emerges as a robust factor affecting SCM–MCM comparability, with more aggregated states in 3D associated with larger behavior deviations from the 1D counterpart. This is found across five convection schemes and applies to simulated mean states, linear responses to small tendency perturbations, and adjustments to doubled‐CO2 forcing. Nevertheless, we find that even when convection is organized, behavior differences between pairs of schemes in the SCM are largely preserved in the MCM. This indicates that when model physics produces accurate behavior in a 1D setup, it will be more likely to do so in a 3D setup. However, our idealized RCE framework implies that these conclusions may not apply to situations with strong large‐scale forcing or encountered over land. Lastly, we demonstrate the practical value of linear responses by showing that they can accurately predict an SCM's tropospheric adjustment to doubled‐CO2 forcing.
... Meteorologists have attempted to represent unresolved surface driven convection in atmospheric models since the very beginning of computational atmospheric modeling. While many methods have been developed and applied successfully, shortcomings in convective parametrizations still cause uncertainty among numerical climate simulations (Sherwood et al., 2014;Vial et al., 2016), as well as biases in the onset of continental precipitation in numerical weather prediction (Grabowski et al., 2006). The most popular and widespread class of convective parametrizations use the mass-flux approach, which was developed decades ago (Arakawa & Schubert, 1974;Yanai et al., 1973) and is still an active field of research and development (e.g., Cohen et al., 2020;Lopez-Gomez et al., 2020). ...
Full-text available
A segmentation algorithm is applied to high resolution simulations of shallow continental convection to identify individual convective 3D objects within the convective boundary layer, in order to investigate which properties of the objects vary with the object width. The study analyses the geometry of the objects, along with their profiles of vertical velocity and total water, to assess various assumptions often used in spectral mass‐flux convection schemes. The methodology of this paper is unique in that we use (a) a novel application of the watershed algorithm to detect individual objects in the constantly evolving continental boundary layer efficiently, and (b) an unprecedentedly large number of simulations being analyzed. In total, 26 days of LASSO simulations at the Atmospheric Radiation Measurement Southern Great Plains site are analyzed, yielding roughly one million objects. Plume‐like surface‐rooted objects are found to be omnipresent, the vertical extent of which is strongly dependent on the object width. The vertical velocity and moisture anomaly profiles of the widest objects are roughly consistent with the classic buoyancy‐driven rising plume model. The kinematic and thermodynamic properties of the objects vary with object width. This width dependence is strongest above cloud base, but much weaker below. Finally the impact of neglecting the contribution of covariances to the vertical moisture flux, which is commonly used in mass‐flux parameterizations, is investigated. The average effect of neglecting covariances increases linearly with object width, leading to a 20% flux underestimation for 2 km wide objects. Implications of the results for spectral convection scheme development are briefly discussed.
Full-text available
Extreme drought occurs on every continent, negatively impacting natural systems and the built environment. Realized and anticipated future warming affects global hydrology, influencing the severity and frequency of both extreme precipitation events and precipitation deficits. Understanding future drought conditions is essential for risk aware water management strategies and to protect food security for a growing human population, while safeguarding natural capital critical to limiting further warming. Here we quantify socioeconomic and ecological exposure to extreme drought. We focus on global, regional, and national scales at increasing levels of climate warming, from today’s 1.0 °C world to 4.0 °C of warming. Drought is quantified using the self-calibrated Palmer drought severity index calculated from globally mosaiced regional climate simulation (REMO2015). Exposure to extreme drought increases monotonically with warming level. For every 0.5 °C warming increase up to 3.0 °C, an additional 619 million people live in areas with 25% likelihood of annual extreme drought, in addition to the 1.7 billion people (25% of 2020 global population) exposed in today’s 1.0 °C world. Spatially, global drying is amplified in the tropics, where drought frequency increases at twice the global rate. Per 0.5 °C increase in warming, extreme drought annual likelihoods increase 1.5 times greater in forested than non-forested areas, jeopardizing climate regulation associated with forested biomes. Cropland exposure to 50% likelihood of annual extreme drought in two of the highest producing countries, China and Brazil, increases 4× and 13× between 1.0 °C and 2.0 °C, spanning a third of national cropland by 3.0 °C. At 1.5 °C (4.0 °C), 16% (39%) of global hydroelectric generating capacity will be exposed to at least a 50% likelihood of annual extreme drought, up from 5% in today’s 1.0 °C world. Given the near-term likelihood of surpassing 1.5 °C, high resolution drought exposure assessments must inform risk aware development and resilience efforts.
Realistic simulation of the Earth's mean-state climate remains a major challenge, and yet it is crucial for predicting the climate system in transition. Deficiencies in models' process representations, propagation of errors from one process to another, and associated compensating errors can often confound the interpretation and improvement of model simulations. These errors and biases can also lead to unrealistic climate projections and incorrect attribution of the physical mechanisms governing past and future climate change. Here we show that a significantly improved global atmospheric simulation can be achieved by focusing on the realism of process assumptions in cloud calibration and subgrid effects using the Energy Exascale Earth System Model (E3SM) Atmosphere Model version 1 (EAMv1). The calibration of clouds and subgrid effects informed by our understanding of physical mechanisms leads to significant improvements in clouds and precipitation climatology, reducing common and long-standing biases across cloud regimes in the model. The improved cloud fidelity in turn reduces biases in other aspects of the system. Furthermore, even though the recalibration does not change the global mean aerosol and total anthropogenic effective radiative forcings (ERFs), the sensitivity of clouds, precipitation, and surface temperature to aerosol perturbations is significantly reduced. This suggests that it is possible to achieve improvements to the historical evolution of surface temperature over EAMv1 and that precise knowledge of global mean ERFs is not enough to constrain historical or future climate change. Cloud feedbacks are also significantly reduced in the recalibrated model, suggesting that there would be a lower climate sensitivity when it is run as part of the fully coupled E3SM. This study also compares results from incremental changes to cloud microphysics, turbulent mixing, deep convection, and subgrid effects to understand how assumptions in the representation of these processes affect different aspects of the simulated atmosphere as well as its response to forcings. We conclude that the spectral composition and geographical distribution of the ERFs and cloud feedback, as well as the fidelity of the simulated base climate state, are important for constraining the climate in the past and future.
Full-text available
Lightning-induced fire is the primary disturbance agent in boreal forests. Recent large fire years have been linked to anomalously high numbers of lightning-caused fire starts, yet the mechanisms regulating the probability of lightning ignition remain uncertain and limit our ability to project future changes. Here, we investigated the influence of lightning properties, landscape characteristics, and fire weather on lightning ignition efficiency – the likelihood that a lightning strike starts a fire - in Alaska, United States of America, and Northwest Territories, Canada, between 2001 and 2018. We found that short-term fuel drying associated with fire weather was the main driver of lightning ignition efficiency. Lightning was also more likely to ignite a wildfire in denser, evergreen forest areas. Under a high greenhouse gas emissions scenario, we predicted that changes in vegetation and fire weather increase lightning ignition efficiency by 14 ± 9 % in Alaska and 31 ± 28 % in the Northwest Territories per 1 ℃ warming by end-of-century. The increases in lightning ignition efficiency, together with a projected doubling of lightning strikes, result in a 39 to 65 % increase in lightning-caused fire occurrence per 1 ℃ warming. This implies that years with many fires will occur more frequently in the future, thereby accelerating carbon losses from boreal forest ecosystems.
Precipitation over tropical oceans rapidly increases when the environmental column saturation fraction (CSF) increases past a critical value of ∼0.7. Past studies suggested that increased stratiform rainfall greatly contributes to the rapid rainfall enhancement. In this study, the sequential roles of non-deep convection, deep convection, and mesoscale convective system (MCS) in precipitation-moisture interactions are examined using 19 years of satellite observations. When CSF is below ∼0.5, non-deep convection dominates total rainfall, and predominantly contributes to moistening of the environment. Between the CSF range of 0.5–0.7, transition to deep convective rainfall begins. Meanwhile, MCS contribution to total rain rapidly increases, and the environment is further moistened. MCS becomes the major rainfall type above the critical CSF value (∼0.7), with the rapid increase of total rain mostly explained by the rapid increase in MCS rain area. Rainfall reduction at high CSF values is jointly contributed by MCS and non-deep convection.
Full-text available
The present study provides an intercomparison and interpretation of climate feedback processes in 19 atmospheric general circulation models. This intercomparison uses sea surface temperature change as a surrogate for climate change. The interpretation of cloud-climate interactions is given special attention. A roughly threefold variation in one measure of global climate sensitivity is found among the 19 models. The important conclusion is that most of this variation is attributable to differences in the models' depiction of cloud feedback, a result that emphasizes the need for improvements in the treatment of clouds in these models if they are ultimately to be used as reliable climate predictors. It is further emphazied that cloud feedback is the consequence of all interacting physical and dynamical processes in a general circulation model. The result of these processes is to produce changes in temperature, moisture distribution, and clouds which are integrated into the radiative response termed cloud feedback.
Full-text available
This study proposes a novel technique for computing cloud feedbacks using histograms of cloud fraction as a joint function of cloud-top pressure (CTP) and optical depth (τ). These histograms were generated by the International Satellite Cloud Climatology Project (ISCCP) simulator that was incorporated into doubled-CO 2 simulations from 11 global climate models in the Cloud Feedback Model Intercomparison Project. The authors use a radiative transfer model to compute top of atmosphere flux sensitivities to cloud fraction perturbations in each bin of the histogram for each month and latitude. Multiplying these cloud radiative kernels with histograms of modeled cloud fraction changes at each grid point per unit of global warming produces an estimate of cloud feedback. Spatial structures and globally integrated cloud feedbacks computed in this manner agree remarkably well with the adjusted change in cloud radiative forcing. The global and annual mean model-simulated cloud feedback is dominated by contributions from medium thickness (3.6 < τ ≤ 23) cloud changes, but thick (τ > 23) cloud changes cause the rapid transition of cloud feedback values from positive in midlatitudes to negative poleward of 50°S and 70°N. High (CTP ≤ 440 hPa) cloud changes are the dominant contributor to longwave (LW) cloud feedback, but because their LW and shortwave (SW) impacts are in opposition, they contribute less to the net cloud feedback than do the positive contributions from low (CTP > 680 hPa) cloud changes. Midlevel (440 < CTP ≤ 680 hPa) cloud changes cause positive SW cloud feedbacks that are 80% as large as those due to low clouds. Finally, high cloud changes induce wider ranges of LW and SW cloud feedbacks across models than do low clouds.
Full-text available
This study diagnoses the climate sensitivity, radiative forcing and climate feedback estimates from eleven general circulation models participating in the Fifth Phase of the Coupled Model Intercomparison Project (CMIP5), and analyzes inter-model differences. This is done by taking into account the fact that the climate response to increased carbon dioxide (CO2) is not necessarily only mediated by surface temperature changes, but can also result from fast land warming and tropospheric adjustments to the CO2 radiative forcing. By considering tropospheric adjustments to CO2 as part of the forcing rather than as feedbacks, and by using the radiative kernels approach, we decompose climate sensitivity estimates in terms of feedbacks and adjustments associated with water vapor, temperature lapse rate, surface albedo and clouds. Cloud adjustment to CO2 is, with one exception, generally positive, and is associated with a reduced strength of the cloud feedback; the multi-model mean cloud feedback is about 33 % weaker. Non-cloud adjustments associated with temperature, water vapor and albedo seem, however, to be better understood as responses to land surface warming. Separating out the tropospheric adjustments does not significantly affect the spread in climate sensitivity estimates, which primarily results from differing climate feedbacks. About 70 % of the spread stems from the cloud feedback, which remains the major source of inter-model spread in climate sensitivity, with a large contribution from the tropics. Differences in tropical cloud feedbacks between low-sensitivity and high-sensitivity models occur over a large range of dynamical regimes, but primarily arise from the regimes associated with a predominance of shallow cumulus and stratocumulus clouds. The combined water vapor plus lapse rate feedback also contributes to the spread of climate sensitivity estimates, with inter-model differences arising primarily from the relative humidity responses throughout the troposphere. Finally, this study points to a substantial role of nonlinearities in the calculation of adjustments and feedbacks for the interpretation of inter-model spread in climate sensitivity estimates. We show that in climate model simulations with large forcing (e.g., 4 × CO2), nonlinearities cannot be assumed minor nor neglected. Having said that, most results presented here are consistent with a number of previous feedback studies, despite the very different nature of the methodologies and all the uncertainties associated with them.
Full-text available
The response of low-level clouds to climate change has been identified as a major contributor to the uncertainty in climate sensitivity estimates among climate models. By analyzing the behaviour of low-level clouds in a hierarchy of models (coupled ocean-atmosphere model, atmospheric general circulation model, aqua-planet model, single-column model) using the same physical parameterizations, this study proposes an interpretation of the strong positive low-cloud feedback predicted by the IPSL-CM5A climate model under climate change. In a warmer climate, the model predicts an enhanced clear-sky radiative cooling, stronger surface turbulent fluxes, a deepening and a drying of the planetary boundary layer, and a decrease of tropical low-clouds in regimes of weak subsidence. We show that the decrease of low-level clouds critically depends on the change in the vertical advection of moist static energy from the free troposphere to the boundary-layer. This change is dominated by variations in the vertical gradient of moist static energy between the surface and the free troposphere just above the boundary-layer. In a warmer climate, the thermodynamical relationship of Clausius-Clapeyron increases this vertical gradient, and then the import by large-scale subsidence of low moist static energy and dry air into the boundary layer. This results in a decrease of the low-level cloudiness and in a weakening of the radiative cooling of the boundary layer by low-level clouds. The energetic framework proposed in this study might help to interpret inter-model differences in low-cloud feedbacks under climate change.
[1] The annual cycle climatology of cloud amount, cloud-top pressure, and optical thickness in two generations of climate models is compared to satellite observations to identify changes over time in the fidelity of simulated clouds. In more recent models, there is widespread reduction of a bias associated with too many highly reflective clouds, with the best models having eliminated this bias. With increased amounts of clouds with lesser reflectivity, the compensating errors that permit models to simulate the time-mean radiation balance have been reduced. Errors in cloud amount as a function of height or climate regime on average show little or no improvement, although greater improvement can be found in individual models.
The mechanisms that govern the response of shallow cumulus, such as found in the trade wind regions, to a warming of the atmosphere in which large-scale atmospheric processes act to keep relative humidity constant are explored. Two robust effects are identified. First, and as is well known, the liquid water lapse rate increases with temperature and tends to increase the amount of water in clouds, making clouds more reflective of solar radiation. Second, and less well appreciated, the surface fluxes increase with the saturation specific humidity, which itself is a strong function of temperature. Using large-eddy simulations it is shown that the liquid water lapse rate acts as a negative feedback: a positive temperature increase driven by radiative forcing is reduced by the increase in cloud water and hence cloud albedo. However, this effect is more than compensated by a reduction of cloudiness associated with the deepening and relative drying of the boundary layer, driven by larger surface moisture fluxes. Because they are so robust, these effects are thought to underlie changes in the structure of the marine boundary layer as a result of global warming.
The Modern-Era Retrospective Analysis for Research and Applications (MERRA) was undertaken by NASA’s Global Modeling and Assimilation Office with two primary objectives: to place observations from NASA’s Earth Observing System satellites into a climate context and to improve upon the hydrologic cycle represented in earlier generations of reanalyses. Focusing on the satellite era, from 1979 to the present, MERRA has achieved its goals with significant improvements in precipitation and water vapor climatology. Here, a brief overview of the system and some aspects of its performance, including quality assessment diagnostics from innovation and residual statistics, is given. By comparing MERRA with other updated reanalyses [the interim version of the next ECMWF Re-Analysis (ERA-Interim) and the Climate Forecast System Reanalysis (CFSR)], advances made in this new generation of reanalyses, as well as remaining deficiencies, are identified. Although there is little difference between the new reanalyses in many aspects of climate variability, substantial differences remain in poorly constrained quantities such as precipitation and surface fluxes. These differences, due to variations both in the models and in the analysis techniques, are an important measure of the uncertainty in reanalysis products. It is also found that all reanalyses are still quite sensitive to observing system changes. Dealing with this sensitivity remains the most pressing challenge for the next generation of reanalyses. Production has now caught up to the current period and MERRA is being continued as a near-real-time climate analysis. The output is available online through the NASA Goddard Earth Sciences Data and Information Services Center (GES DISC).
In HadGEM2-A, AMIP experiments forced with observed sea surface temperatures respond to uniform and patterned +4 K SST perturbations with strong positive cloud feedbacks in the subtropical stratocumulus/trade cumulus transition regions. Over the subtropical Northeast Pacific at 137°W/26°N, the boundary layer cloud fraction reduces considerably in the AMIP +4 K patterned SST experiment. The near-surface wind speed and the air-sea temperature difference reduces, while the near-surface relative humidity increases. These changes limit the local increase in surface evaporation to just 3 W/m2 or 0.6 %/K. Previous studies have suggested that increases in surface evaporation may be required to maintain maritime boundary layer cloud in a warmer climate. This suggests that the supply of water vapour from surface evaporation may not be increasing enough to maintain the low level cloud fraction in the warmer climate in HadGEM2-A. Sensitivity tests which force the surface evaporation to increase substantially in the +4 K patterned SST experiment result in smaller changes in boundary layer cloud and a weaker cloud feedback in HadGEM2-A, supporting this idea. Although global mean surface evaporation in climate models increases robustly with global temperature (and the resulting increase in atmospheric radiative cooling), local values may increase much less, having a significant impact on cloud feedback. These results suggest a coupling between cloud feedback and the hydrological cycle via changes in the patterns of surface evaporation. A better understanding of both the factors controlling local changes in surface evaporation and the sensitivity of clouds to such changes may be required to understand the reasons for inter-model differences in subtropical cloud feedback.
We diagnose climate feedback parameters and CO2 forcing including rapid adjustment in twelve atmosphere/mixed-layer-ocean (“slab”) climate models from the CMIP3/CFMIP-1 project (the AR4 ensemble) and fifteen parameter-perturbed versions of the HadSM3 slab model (the PPE). In both ensembles, differences in climate feedbacks can account for approximately twice as much of the range in climate sensitivity as differences in CO2 forcing. In the AR4 ensemble, cloud effects can explain the full range of climate sensitivities, and cloud feedback components contribute four times as much as cloud components of CO2 forcing to the range. Non-cloud feedbacks are required to fully account for the high sensitivities of some models however. The largest contribution to the high sensitivity of HadGEM1 is from a high latitude clear-sky shortwave feedback, and clear-sky longwave feedbacks contribute substantially to the highest sensitivity members of the PPE. Differences in low latitude ocean regions (30°N/S) contribute more to the range than those in mid-latitude oceans (30–55°N/S), low/mid latitude land (55°N/S) or high latitude ocean/land (55–90°N/S), but contributions from these other regions are required to account fully for the higher model sensitivities, for example from land areas in IPSL CM4. Net cloud feedback components over the low latitude oceans sorted into percentile ranges of lower tropospheric stability (LTS) show largest differences among models in stable regions, mainly due to their shortwave components, most of which are positive in spite of increasing LTS. Differences in the mid-stability range are smaller, but cover a larger area, contributing a comparable amount to the range in climate sensitivity. These are strongly anti-correlated with changes in subsidence. Cloud components of CO2 forcing also show the largest differences in stable regions, and are strongly anticorrelated with changes in estimated inversion strength (EIS). This is qualitatively consistent with what would be expected from observed relationships between EIS and low-level cloud fraction. We identify a number of cases where individual models show unusually strong forcings and feedbacks compared to other members of the ensemble. We encourage modelling groups to investigate unusual model behaviours further with sensitivity experiments. Most of the models fail to correctly reproduce the observed relationships between stability and cloud radiative effect in the subtropics, indicating that there remains considerable room for model improvements in the future.
[1] We utilize energy budget diagnostics from the Coupled Model Intercomparison Project phase 5 (CMIP5) to evaluate the models' climate forcing since preindustrial times employing an established regression technique. The climate forcing evaluated this way, termed the adjusted forcing (AF), includes a rapid adjustment term associated with cloud changes and other tropospheric and land-surface changes. We estimate a 2010 total anthropogenic and natural AF from CMIP5 models of 1.9 ± 0.9 W m−2 (5–95% range). The projected AF of the Representative Concentration Pathway simulations are lower than their expected radiative forcing (RF) in 2095 but agree well with efficacy weighted forcings from integrated assessment models. The smaller AF, compared to RF, is likely due to cloud adjustment. Multimodel time series of temperature change and AF from 1850 to 2100 have large intermodel spreads throughout the period. The intermodel spread of temperature change is principally driven by forcing differences in the present day and climate feedback differences in 2095, although forcing differences are still important for model spread at 2095. We find no significant relationship between the equilibrium climate sensitivity (ECS) of a model and its 2003 AF, in contrast to that found in older models where higher ECS models generally had less forcing. Given the large present-day model spread, there is no indication of any tendency by modelling groups to adjust their aerosol forcing in order to produce observed trends. Instead, some CMIP5 models have a relatively large positive forcing and overestimate the observed temperature change.