Download North American 2008 Cooling Attributed to Natural

Oceans and Drought
Martin Hoerling
NOAA Earth System Research Laboratory, Boulder Colorado
SUMMARY
Three features of the relationship between oceans and drought are highlighted. The first
is the link between the seasonal pulse of monsoons and the seasonality of sea surface
temperatures (SST). It is shown that the north-south migration of monsoon rains is
strongly tied to the ocean’s response to the annual march of the Sun. The second is the
link between interannual-to-decadal droughts and anomalous ocean states. It is shown
that SST variability enhances drought variability over land areas within 40°N-40°S. The
effect of ocean states on drought is demonstrated for select historical events, and the
ocean observing system needs for drought early warning are discussed. A third feature
concerns the link between the oceans and drought in a warming climate. Ocean
temperatures have risen over the warmest portions of the tropical oceans, likely due to the
effect of increasing greenhouse gases. The effect of such sea surface temperature trends
on regional rainfall trends, including recent droughts over southwest North America and
Australia, are discussed.
Earth’s Dry and Wet Seasons
The annually recurring dry and wet seasons represent the most dramatic pulse of Earth’s
climate. Notable are the tropical monsoons with wet phases developing in tandem with
the poleward migration of solar radiation. The global pattern of boreal summer minus
winter rainfall captures the essence of this pulse, with wet conditions prevailing over the
Sahel, India, Southeast Asia, and Central America, while southern Africa, northern
Australia and the Amazon simultaneously experience their dry phases (Fig. 1).
Figure 1. The seasonal cycle of rainfall (July/Aug/Sept- Jan/Feb/March) based on
simulations of a climate model using realistic observed seasonally varying sea surface
temperatures. Green (red) indicates regions of wetter (drier) boreal summer (winter)
climate.
The ocean’s response to the march of the seasons is a key factor driving the monsoons.
During boreal summer, the NH oceans warm compared to the SH oceans, and monsoon
rains develop over the NH continents. The seasonal cycle of rainfall simulated in a
climate model in which specified sea surface temperatures have no seasonal cycle is
compared to that occurring in a parallel run using realistic seasonally varying SSTs (Fig.
2). A large component of the monsoon rainfall cycle is attributable to the seasonal cycle
of SSTs according to these experiments (cf. Fig. 1 and 2). These experiments further
indicate that the oceans play a major role in the monsoon’s pulse, and that the annually
recurring wet and dry seasons and not mere slaves to the seasonal cycle of the warming
and cooling of land temperatures.
Figure 2. The difference in seasonal cycle of rainfall (July/Aug/Sept- Jan/Feb/March)
between simulations of a climate model using realistic observed seasonally varying sea
surface temperatures and parallel simulations in which SSTs have no seasonal cycle..
Green (red) indicates regions where the SST seasonal cycle drives a wetter (drier) boreal
summer (winter) climate compared to runs having no SST seasonal cycle.
Interannual and Decadal Droughts
Interannual to decadal drought events, particularly over tropical and subtropical latitudes,
are sensitive to anomalous sea surface temperatures. Climate model simulations indicate
that the effect of SST variability (i.e., fluctuations relative to the climatological seasonal
cycle) significantly enhance the variance of precipitation, and thus contribute to the
occurrence of severe droughts. For instance, precipitation variance is more than doubled
in such regions as the Maritime Continent, equatorial South America and portions of
India (Fig. 3). These are among the regions that have particularly strong sensitivity to the
El Niño/Southern Oscillation (ENSO) phenomenon. Likewise, many other continental
regions between 40°N-40°S experience an increase in precipitation variability on the
order of 10%-25% due to the SST variability.
Figure 3. The ratio of variance of annual standardized precipitation in a climate
simulation forced with observed interannual SST variations of 1901-2000 to that
occurring in the same model using climatological seasonally varying SSTs only. Colored
areas indicate where precipitation variance is enhanced, with red shades denoting a
more than a doubling of variance.
In light of the increased rainfall variance resulting from sea surface temperature
variability, a key question is whether ocean observing systems can harvest the potential
predictability implied in Fig. 3. Towards addressing this question we consider several
major droughts and explore the ocean’s role more specifically.
Two iconic U.S. drought events of the last century are the Dust Bowl during the 1930s
and the comparably severe and sustained drought during the 1950s. These affected
different geographical regions (Fig. 4), and the question is whether early warning
capability for such droughts now exists given current ocean observing and prediction
systems (e.g., such as TAO/TRITON and ENSO prediction).
Figure 4. The observed standardized precipitation index (SPI) averaged for 1932-1939
(left) and for 1946-1956 (right). Red (blue) shaded regions indicate reduced (increased)
precipitation during these epochs. Note that the Great Plains droughts of the 1930s and
1950s occurred over different geographical regions.
There is compelling evidence that ocean observations over the tropical Pacific and
associated predictions of ENSO can forewarn of southern U.S. Plains drought, such as
occurred during the 1950s This is indicated by the high statistical correlation between an
index of annual precipitation over the southern Plains and annual sea surface
temperatures during the 20th Century (Fig. 5, right panel). Yet, that same statistical
analysis fails to indicate a coherent relation between drought over the Northern Plains,
the epicenter for 1930s drought, and the state of global ocean temperatures (Fig. 5, left
panel).
Figure 5. The observed correlation between annual sea surface temperatures and the
standardized precipitation index (SPI) averaged over the Northern Plains (left) and the
Southern Plains (right) during 1895-2007. Geographic regions are shown by gray
hatching. Note that Southern Plains drought correlates with cold phases of ENSO, but
that there is no appreciable correlation of tropical SSTs with Northern Plains drought.
The correlations themselves are not proof (or lack thereof) that ocean conditions caused
the major 20th Century US droughts. A common strategy to test the efficacy of the
ocean’s role, and to verify a cause-effect link, is to force atmospheric climate models
with the observed SSTs that occurred during the drought period. Such experiments
conducted using SSTs spanning the 1930s and 1950s are unable to generate the extensive
dry conditions over the central and northern Plains of the 1930s, but yield a realistic
simulation of the 1950s drought pattern (Fig. 6). The Southern Great Plains thus resides
within an epicenter of potentially predictable drought for which an ocean observing
system would appear to also constitute a drought early warning system.
Figure 6. The simulated standardized precipitation index (SPI) averaged for 1932-1939
(left) and for 1946-1956 (right). Red (blue) shaded regions indicate reduced (increased)
precipitation during these epochs. Results are based on a 40-run average of atmospheric
GCM simulations using three different models. Note the very similar patterns of Great
Plains droughts of the 1930s and 1950s occurring in the GCM simulations.
As another case study, Sahelian drought events of seasonal duration often occur in
concert with anomalous sea surface temperature conditions. Yet, the prolonged multidecadal drying of the Sahel that commenced in the 1970s and persisted until the mid1990s raises new questions whether long-lived ocean conditions can likewise be
important factors. Can the existing global ocean observing system forewarn of protracted
Sahelian drought in the future?
Statistical analysis of the relation between Sahel rainfall and an index of the
interhemispheric contrast in SSTs reveals a strong relation on annual to decadal time
scales. The prolonged drying of the Sahel occurred in unison with a cooling of NH SSTs
compared to the SH SSTs, and subsequent climate model simulations using observed
SSTs have confirmed a causal link. In particular, a prolonged drying trend of the Sahel
during the 1950-1999 period was forced by cooling of the subtropical North Atlantic
together with a warming of the Indian Ocean. An open question is the predictability of
such ocean conditions, and whether existing ocean observations are adequate to meet the
requirements to harvest predictability.
Drought in a Warming Climate
According to the Fourth Assessment Report of the Intergovernmental Panel on Climate
Change, surface temperatures averaged over the NH during the second half of the 20th
Century were very likely higher than during any other 50-year period since 1500. Most
of the warming of both land and sea surface temperatures during the last half-century is
very likely the consequence of human-induced emissions of greenhouse gases and
anthropogenic aerosols. Since 1977, virtually all oceans have warmed, with the notable
exception of the tropical and subtropical east Pacific (Fig. 7, top). It is in this latter
region where the simulated SST change due to GHG and aerosol forcing deviates most
appreciably from observations.
Figure 7. The 1977-2006 trend in annual observed sea surface temperatures (top) and
CMIP3 simulated sea surface temperatures. Trend toward warmer (colder) SSTs are
indicated by red (blue) shades. The difference in 30-yr trends (OBS minus CMIP) is
shown in the lower panel. Simulations are based on the average of 21 CMIP3 models that
were forced by the observed GHG and aerosol variations from 1977-1999, and by the
SRES A1B emissions scenario for 2000-2006.
A key challenge in the emerging area of decadal prediction is to understand the
relationship between slow and sustained changes in oceans and regional climates. One
example was already discussed regarding the ocean’s role in Sahel drying. To what
extent are trends in observed SSTs since 1977 contributing to regional trends in rainfall,
and to what extent is that relationship indicative of an anthropogenic influence?
Most notable have been the drying trends during 1977-2006 observed over the Maury
Darling Basin and over a broad region of the eastern Pacific and adjacent continents
including southwestern North America (Fig, 8, top). The latter features are consistent
with an influence of the oceans, as revealed by the results of atmospheric GCM
simulations forced by the observed SSTs (Fig. 8, middle). However, this ocean influence
is evidently not a feature of human-induced climate change, in so far as coupled oceanatmosphere GCMs forced by GHG and aerosols fail to generate this wide spread-dryness.
Figure 8. The observed (top), AMIP ensemble mean simulated (middle), and CMIP
ensemble mean simulated (bottom) 30-year trends in annual precipitation during 19772006. AMIP simulations are based on an ensemble of 36 integrations of 4 different
atmospheric models forced by the observed monthly varying SSTs, but using
climatological GHGs. CMIP simulations are based on the average of 21 CMIP3 models
that were forced by the observed GHG and aerosol variations from 1977-1999, and by
the SRES A1B emissions scenario for 2000-2006.
Oceans Observations and Drought
Drought as a natural hazard often has severe, and long lasting societal impacts. This
presentation has given several examples of the impact of oceans on drought. It has
attempted to show where ocean information and related observing systems could be used
for societal benefit relating to early warning for and possible mitigation of drought
impacts. It is not an exhaustive survey of the potential benefits of ocean information, and
the talk focused solely on the sea surface temperature information.
It was shown that ocean observations in the tropical east Pacific, information that is vital
to support predictions of ENSO, are tantamount to drought early warning over vast areas
of the tropics and select midlatitude regions. Regarding the latter, the talk emphasized
that ocean information related to ENSO would provide early warning for drought over
the U.S. Southern Great Plains, although there was less indication that ocean information,
nor skillful predictions of SSTs, would serve to provide advance warning for drought in
the U.S. Central to Northern Plains, a region devastated by drought during the 1930s.
Sahel drought of multi-decadal duration in the latter half of the 20th Century is now
known to have been strongly forced by global ocean conditions. Research is only
beginning to explore the capability to predict the relevant ocean states and their impacts
on such long time scales. Regardless of the outcome of the prediction enterprise, there is
considerable societal benefit in knowing the causes for drought conditions. A particular
example was given of the trends in rainfall during 1977-2006, and the evidence that
drying over the tropical Pacific and the adjacent Americas has resulted in part from sea
surface temperature trends, though ones apparently unrelated to the anthropogenic
forcing of the oceans.
Climate-quality ocean observations are essential for distinguishing natural from humaninduced ocean conditions, and therefore are key in the accurate attribution of the causes
of climate trends. Of particular importance is ensuring that natural variability, when
occurring, is not misunderstood to indicate that climate change is either not happening or
is happening more intensely than the true human influence.