decadal climate and global change research · decadal climate and global change research 5.1...

2001 CDC Science Review 71

CHAPTER 5

Decadal Climate and

Global Change Research

5.1 Modeling studies of low frequencyatmospheric variability and climatechange

5.1.1 Atmospheric circulation changeand the role of tropical SST warming

A significant component of global sur-face temperature trend since 1950 is

linked with planetary scale changes inatmospheric wind and pressure patterns,and our analysis based on atmosphericGCM experiments indicates that the lat-ter have been forced by changes in tropi-cal SSTs. During northern winter, thetrend in 500-mb heights since 1950 con-sists of a uniform tropical and subtropi-cal height increase, and a wavy pattern of

The Climate Diagnostics Center contributes to understanding the variations in Earth's climate system ondecadal to centennial time scales. Research includes process and model simulation studies to elucidate therelation between changes in the atmosphere and those in the ocean. CDC scientists seek to provide physicaland dynamical understanding of observed long term climate variations and change through analysis of hier-archies of designed GCM experiments. These include atmospheric models forced by SSTs, ocean modelsforced by wind stress, and coupled ocean-atmosphere GCMs, including runs forced by greenhouse gases.

Our investigations have focused on determining fundamental processes responsible for decadal climate vari-ability and change, and assessing whether the latter are due to human influences or natural variability.According to the Third Assessment Report of the IPCC, it is now very likely that global temperatures duringthe 1990s were the highest since 1861. The same appears to be true for tropical sea surface temperatures,and the areal coverage of the so-called oceanic warm pool (SSTs>28.5°C) (Fig. 5.1, top). CDC scientists arediagnosing relationships between this tropical ocean warming, the global atmospheric circulation and recentclimate change.

CDC is engaged in understanding how slow changes in climate affect interannual variability. One key ques-tion is determining whether the warm pool change over the equatorial west Pacific impacts the statistics of ElNiño/Southern Oscillation (ENSO) in the eastern Pacific. It is evident that the strongest El Niño events in theinstrumental record have occurred in recent decades (Fig. 5.1, middle); we are assessing if this is a signatureof climate change or merely random fluctuations. The global impacts of ENSO have also changed in recentdecades. A fundamental question being pursued at CDC is whether the ENSO teleconnections diagnosedfrom historical data of the 19th and 20th centuries will be accurate depictions of ENSO impacts in this new,unique century of human induced climate change.

CDC scientists are also studying the origin and climatic impact of midlatitude ocean changes. Most dramaticamong these is the multi-decadal variability in SSTs over the Pacific poleward of 30°N (Fig. 5.1, bottom), anindex of which has been termed the Pacific-Decadal Oscillation (PDO). The apparent long time scale of thisoceanic behavior is quite different from that of the ENSO time series. Nonetheless, our analysis shows astrong relation between the two on interannual time scales, and an intriguing question is the extent of theircoupling on multi-decadal scales. Likewise, the low frequency variations of North Pacific SSTs since 1950have atmospheric counterparts, including changes in the oceanic storm track and the strength of the uppertropospheric westerly jet. CDC scientists are studying the nature of air-sea interaction over the North Pacific,and assessing to what extent the diagnostic relations mentioned above entail predictability.

_____________________________

CHAPTER 5 Decadal Climate and Global Change Research

72 2001 CDC Science Review

stationary wave change in higher lati-tudes highlighted by lower pressure overthe North Pacific and North Atlantic(Fig. 5.2, top). The change in the latterregion projects strongly on the NorthAtlantic Oscillation (NAO) structure ofmonthly variability. Recent hypothesesfor the North Atlantic climate changeinclude a positive feedback resultingfrom coupling with North Atlantic SSTs,and the possibility that the slow varia-tions are nothing more than samplingartifacts of a random stationary process.Earlier studies of North Pacific climatechange have also argued for forcing fromthe slow variations in extratropical(North Pacific) SSTs (see Fig. 5.1),

together with forcing from changes in thetropical Pacific ocean. We have analyzeddata from a 12-member ensemble ofatmospheric simulations with NCAR'sCCM3, using global SST variations since1950, and these confirm that theobserved circulation pattern trends areconsistent with global air-sea interac-tions (Fig. 5.2, middle). The role of trop-ical SSTs is revealed from another GCMensemble in which monthly SST varia-tions are prescribed over only the 30°N–30°S band. That the observed trend iscaptured by the tropically forced simula-tions (Fig. 5.2, bottom) alone suggeststhat the gradual warming of those waterssince 1950 is forcing NH climate change.

Fig. 5.1. (a) Warm pool index that describes the time series of areal coverage (expressed as the number of 2° by 2°grid boxes) of SST>28.5°C within the region (30°N–30°S, 50°E–120°W). (b) Cold tongue index of ENSO thatdescribed the time series of area-averaged SST anomalies within the region (5°N–5°S, 160°E–80°W). (c) NorthPacific SST index for the area average within the region (30°N–50°N, 150°E–150°W). All time series are for Janu-ary 1895–April 2001, and have been smoothed with a 13-month running mean. The warm pool index is derivedfrom GISST 2.3b data. The ENSO and North Pacific SST indices use the Kaplan data for 1895–1949, Smith andReynolds reconstructed SST data from 1950–1981, and the Reynolds OI data after 1982. Anomalies are computedrelative to the entire 107 year period.

Decadal Climate and Global Change Research CHAPTER 5


In fact, further analysis suggests that thesecular warming within the oceanicwarm pool region itself is most relevantfor the simulated climate change, partic-ularly over the distant North Atlantic sec-tor.

Tropical SST warming since 1950 hasaltered tropical rainfall, the likely imme-diate cause for the simulated atmosphericcirculation changes. The modeled rain-fall trend is consistent with that of theunderlying SSTs, with increased precipi-tation throughout the entire equatorialIndo-Pacific region (Fig. 5.3, top). Thisleads to the question of the origin of thetropical SST change itself (Fig. 5.3, bot-tom). Analysis of coupled ocean-atmo-sphere experiments indicate that thewarming trend is beyond the range ofnatural variability. The changes doappear consistent with anomalous green-house forcing, however, insofar as thepattern and amplitude of warming duringthe past half century is similar to thatpredicted by such models when forcedby observed greenhouse gas changes.Our current working hypothesis is thatthe spatial pattern of NH winter climatechange, and the regional change over theNorth Atlantic/European sector espe-cially, is being forced by a tropical oceanwarming, and implicitly reflects an emer-gent anthropogenic signal.

5.1.2 Global warming and atmosphericangular momentum

The recent trends in wintertime NHheight are strongly annular from the sur-face to the lower stratosphere. Heightshave been increasing in the tropics but

Fig. 5.2. The linear trend of the winter season (Decem-ber to February) 500-mb height field based on observa-tions (top), 12-member CCM3 ensemble forced byglobal SST variations (middle), and 12-member CCM3ensemble forced by tropical SST variations over 1950-1999 (bottom). The model results have been multipliedby a factor of 2. Height increases (decreases) are indi-cated by solid (dashed) contours, and the contourincrement is 20 m per 50 years. (Based on results ofHoerling, Hurrell and Xu, 2001, Science).



decreasing near the poles, and associatedwith this has been an increase in theatmospheric angular momentum (AAM)(Fig. 5.4). Much of the increase reflects achange in relative angular momentumthat is due to a broad westerly windincrease within 30°N–30°S. We havebeen seeking an explanation for thischange using GCM simulations sub-jected to various forcings. First, thechange of AAM was investigated in athree-member ensemble of coupledocean-atmosphere model simulations

with increasing atmospheric greenhousegases and sulfate aerosol loading for theperiod 1900–2100. A highly significantincrease in total AAM was found tooccur in those runs, with an indicationthat the forced change emerges above thenoise of natural variability by the late20th Century.

We hypothesize that the AAM isresponding to SST changes and accom-panying tropical convection, rather thanto direct changes in radiative forcing. Inthe model, the AAM is found to accom-pany an increase in tropical SST with asensitivity of ~1 AMU/°C (1 AMU =1025 kg m2 s-1). A similar sensitivity isfound in the observed AAM response toNiño 3.4 SSTs during El Niño. To furtherexamine the role of SST changes only,we analyzed a 12-member ensemble ofNCAR CCM3 simulations forced withthe history of global SSTs during 1950–99. The simulated AAM from these runscompares well with the time history ofthe (single realization) observed AAM,and with the observed trend in Niño 3.4SSTs (see Fig. 5.4). The AAM time

Fig. 5.3. The linear trend of the winter season (Decem-ber to February) total precipitation from the 12-memberCCM3 ensemble forced by tropical SST variations(top), and the observed sea surface temperatures (bot-tom). (Based on results of Hoerling, Hurrell and Xu,2001, Science).

Fig. 5.4. Time series of global angular momentum from 1950–1999 based on NCEP/NCAR reanalysis (blue), 12-member CCM3 ensemble forced with global SST variations (black), Canadian Centre for Climate Modeling cou-pled GCM forced by greenhouse gases and aerosols (green). The standard deviation among the CCM3 ensemblemembers shown in shading. The time series of Niño 3.4 SSTs is shown in red. All curves smoothed with a 5-yr run-ning mean. (Based on results of Huang, Weickmann and Hsu, 2001, J. Climate, in press).



series in the greenhouse forced experi-ment also shows an increasing trend,though it appears to underestimate theobserved low frequency variability. Allfour curves show a trend during 1950–99of approximately the same magnitude,although the coupled run has the smallestand the observed AAM has the largesttrend. The curves cluster together after1980, while before that time the AMIPensemble lies between the coupledensemble and the observed curve. Theimplication is that the observed AAMtrend is partially the result of globalwarming, through its effect on the tropi-cal heat sources, although natural vari-ability (and NCEP reanalysis errors)probably also contribute. In any case, theanthropogenic forced change through1999 is as yet modest, especially com-pared to the AAM change predicted to

occur by this model over the next 100years (~ 4 AMU).

5.1.3 Decadal variations in ENSO andits global impact

Observations reveal that the globalimpacts of ENSO vary substantially ondecadal time scales. One example is thebreakdown in the ENSO-Indian monsoonrelationship in the last quarter century. Inaddition to such changes in tropical tele-connections, ENSO's extratropicalimpacts have also shown strong decadalvariations. For example, ENSO hasexplained a much higher fraction of theseasonal atmospheric variability in thePNA-sector since 1977 compared to theprior quarter century. The February-Aprilseasonal correlation of 500-mb heightswith a SST index of ENSO has roughly

Fig. 5.5. Correlation pattern of seasonally averaged February–April 500 hPa geopotential height with the ENSOindex of the middle panel of Fig. 5.1 calculated for two sub-periods: 1948–1976 (left panel), and 1977–1999 (rightpanel). Results are shown for the Northern Hemisphere for the area extending north of 20 degrees latitude. (Basedon results of Diaz, Hoerling and Eischeid, 2001, Int. J. Climat., in press).



doubled during 1977–99 compared to1948–76 (Fig. 5.5). The correlation,averaged across the southern US between120°W–70°W, increases from roughly –0.2 to –0.6 in the recent period, andpoints to a substantial increase in thelate-winter potential predictability.

One might suppose that this change inpredictability is due to the strongerENSO forcing as measured by theincreased interannual variance of tropicalPacific SSTs since 1977 (see Fig. 5.1).Yet, this increase is only about 20 per-cent, and it is not reconcilable with themuch greater fractional increase in vari-ance of the remote ENSO signal. Recentchanges in the life cycle of tropicalPacific SST anomalies during warm

events may be a key factor. The so-calledcanonical warm event evolution prior to1977, using the cases of Rasmusson andCarpenter, had peak warming in Decem-ber, followed by rapid decay (Fig. 5.6,left panel contours). We have calculatedthe difference in SST lifecycles betweenthe post-1977 and pre-1977 warm events(Fig. 5.6, left panel shading), and foundthat recent cases prolong their warminginto spring of year+1, and also deferSouth American coastal warming untilspring of year (+1). This change is rele-vant for the teleconnections because therecent warm events peak closer to thepeak in climatological warming of theeast equatorial Pacific ocean, so that thetotal SST is actually now maximized inspring of year+1 rather than in early win-

Fig. 5.6. Hovmuller diagrams illustrating the change in the composite life cycle of SST and rainfall anomaliesalong the equatorial Pacific during 1948–1999. In the left panel, contours illustrate the composite El Niño SSTanomaly based on pre-1976 events, whereas the shading illustrates the difference in El Niño SST composites forevents occurring after versus before 1976. In the right side panel, the contours illustrate the change in El Niñorainfall anomaly composites for events occurring after versus before 1976, as derived from climate simulations. Theshading repeats the SST change of the left panel. (Based on results of Diaz, Hoerling and Eischeid, 2001, Int. J. Cli-mat., in press).



ter. Using output from AMIP-style simu-lations spanning 1950–99, we have founda large increase in rainfall for the post-1977 events relative to the pre-1977events (Fig. 5.6, right panel contours)that corresponds closely with the changein SSTs. The change in composite rain-fall represents a doubling relative to thepre-1977 cases, suggesting that therecent teleconnection strength is at leastqualitatively consistent with a secularchange in tropical forcing, though otherfactors may also be involved.

5.2 Modeling studies of fundamentalocean processes

From the various atmospheric GCMexperiments described above, twoaspects of tropical ocean change in the20th Century have been implicated asforcing observed atmospheric change; i)the warming of the warm pool region andii) the nonstationarity of the ENSO timeseries. Our research on fundamentalocean processes leads to the hypothesisthat these two oceanic changes are them-selves coupled, and in particular that therecent increase in El Niño amplitude isconsistent with the increase in warm pooltemperature.

5.2.1 Decadal ENSO variability and therole of warm pool SST

From detailed analysis of the 1986–87 ElNiño event, we find that El Niño repre-sents a mechanism by which the equato-rial Pacific transports heat poleward, aresult subsequently confirmed from amore extended study using NCEP datafor the last 20 years. In particular, a sys-

tematic relationship between the oceanheat content in the western Pacific andthe magnitude of El Niño warming wasdiagnosed for six events since 1980—thehigher the heat content in the westernPacific, the stronger the subsequent ElNiño warming (Fig. 5.7, top). The occur-rence of warm pool heat content maxima,

Fig. 5.7. (a) Zonal distribution of upper ocean heatcontent (0–260 m) in the equatorial belt (5°S–5°N)when the western Pacific heat content reaches its pre-El Niño peak. Upper ocean heat content used for thisfigure was smoothed in time using a Hanning windowwith a width of 13 months. (b) The correspondingdepth of the 20°C isotherm. The ocean temperatureused for calculating the 20°C isotherm depth wassmoothed in time using a Hanning window with awidth of 13 months. (Based on results from Sun 2001,J. Climate, submitted).

b

a

Longitude

Dep

th (

m)

Hea

t C

onte

nt (

x 10

10 J

m-2

)



at which time the zonal cross sections ofFig. 5.7 were made, precede the maximain Niño 3 SST anomalies by 12–24months.

Our interpretation is that higher heat con-tent in the western Pacific is achievedthrough a deepening of the local ther-mocline, thus linking the heat content inthe western Pacific to the potentialenergy of the ocean and thereby with thestability of the coupled ocean-atmo-sphere system (Fig. 5.7, bottom). Aswarm-pool SST initially increases, thezonal SST contrast also increases,strengthening the trade winds. The stron-ger Walker circulation then interacts withclouds and water vapor, allowing moresolar radiation to reach the ocean's sur-face over the east Pacific equatorial cold-tongue, and at the same time, reducingthe surface evaporative cooling over thatregion. This is so because the impact ofchange in the gradient term of the latentheat flux formula exceeds the impact ofincreased wind speed. Through non-localocean wave dynamics and transports,heat content increases in the equatorialupper ocean of the western Pacific warmpool. The resulting steeper tilt of theequatorial thermocline is hypothesized todestabilize the coupled system which isfollowed by energy release through astronger El Niño.

To test this hypothesis, we constructed acoupled model. The atmospheric modelis statistical, with the equatorial surfacewinds proportional to the zonal SST gra-dients. The ocean component is a primi-tive equation model and thereforeexplicitly calculates the heat budget ofthe entire equatorial upper ocean. The

model produces ENSO-like variations.The evolution of the subsurface oceantemperature over the life cycle of themodel El Niño resembles that of obser-vations (Fig. 5.8). In response to anincrease in warm pool SST, the modelhas a stronger El Niño. Similar to theobservational results, this simple modelshows that an increase in warm pool SSTstrengthens the zonal SST contrast dur-ing ENSO's cold phase, which leads toan effective increase in the upper oceanheat content in the warm pool. StrongerEl Niño warming then follows, whichacts as a poleward heat pump. Of course,other processes can operate to increasewarm pool SSTs. In regard to theobserved recent climate change (see Fig.5.1), it is reasonable that the warm poolhas increased due to local radiative forc-ing related to the increase in anthropo-genic gases. It is hypothesized that thisexternally forced change may be influ-encing the statistics of ENSO through themechanisms described above.

5.2.2 North Pacific decadal ocean vari-ability and the role of the tropics

As mentioned earlier, the time series ofthe ENSO index is correlated with that ofNorth Pacific SSTs (see Fig. 5.1), despitetheir different time scales of variation.This reflects in part the well-known factthat ENSO influences the North Pacificcirculation, which in turn forces NorthPacific SSTs, on interannual time scales.The question we have pursued is to whatextent this atmospheric bridge betweenthe tropics and extratropics contributes tothe decadal variability over the NorthPacific, including the Pacific DecadalOscillation? We have addressed this



question by comparing the observed andsimulated leading pattern (EOF 1) andassociated principal component timeseries (PC) of wintertime North Pacificdecadal SST variability (Fig. 5.9). Themodel results are obtained from theensemble average of 16 50-year GFDLR30 AGCM simulations in which

observed SSTs are specified in the tropi-cal Pacific over the period 1950–1999and a mixed layer model (MLM) is cou-pled to the AGCM elsewhere over theglobal oceans. The EOFs are based onthe monthly SST anomalies that werefirst low-pass filtered to retain periodsgreater than ~10 years and then the fil-

Fig. 5.8. The observed (left panels) and coupled model simulated (right panels) ocean temperature anomalies dur-ing an El Niño event's life cycle. Vertical sections are from surface to 330 m. Warm (cold) anomalies in red (blue).Observations based on El Niño composite during 1980–1999 based on NCEP ocean analysis. (Based on resultsfrom Sun 2001, J. Climate, submitted)

Initial Stage Initial Stage

Developing Stage Developing Stage

Mature Stage Mature Stage

Decaying Stage Decaying Stage

Observations Model

Dep

th (

m)

Dep

th (

m)

Dep

th (

m)

Dep

th (

m)



tered values from November to Marchwere averaged together. The observedand MLM EOFs resemble each other inseveral respects: they both explain abouthalf of the variance, and they are rela-tively well correlated in space and time,with a spatial (temporal) correlation of0.71 (0.69). The patterns in Fig. 5.9 arevery similar to those based on unfiltereddata which has conventionally been usedto define the PDO. The observed andsimulated PCs are well correlated withthe filtered ENSO index time series, withcorrelation values of 0.77 and 0.90,respectively. In addition, maps of SSTdifferences centered on 1976 (e.g., 1977-1988 minus 1970–1976, and 1977-1998minus 1951-1976; not shown) indicatethat the “abrupt climate transition” in themodel and observations are similar andresemble the leading EOF but the ampli-tude of the differences is approximatelyhalf as large in the MLM. Overall, ourmodel results suggest that a significantfraction of the variance of the dominantpattern of low frequency SST variabilityin the North Pacific is associated with theatmospheric bridge.

5.2.3 Subduction and Rossby wavedynamics: mechanisms for decadalocean variability

Dynamic ocean processes likely play afundamental role in climate variabilityon decadal timescales. Rossby wavepropagation can introduce multi-yeardelays in the oceanic response to changesin the atmospheric forcing. Subduction,where surface waters enter and flowwithin the permanent pycnocline, pro-vide a link between the extratropical andtropical oceans over an ~8 year period. It

has been conjectured that when the sub-ducted anomalies reach the equator, theyalter the equatorial SSTs and affect theNorth Pacific Ocean via the atmosphericbridge, completing a circuit that enablesdecadal oscillations. CDC has beeninvolved in observational and modelingstudies that examine subduction andRossby waves in the Pacific Ocean.

The standard deviation of the depth ofthe 25.5 sθ isopycnal surface, obtainedfrom a global OGCM forced by observedsurface fluxes, indicates that there arethree major centers of variability, includ-

Fig. 5.9. EOF 1 of the low-pass filtered (> ~10 years)SST anomalies during November–March from (a)observations and (b) the MLM. (c) The first principalcomponent (time series associated with EOF 1) of thelow-pass SST anomalies from observations (greenline), the MLM (blue line) and low-pass ENSO index(black line). The correlations (r) between the three timeseries are given above (c). (Based on the results ofAlexander, Blade, Newman, Lau, and Lanzante, 2001,J. Climate, submitted).



ing: i) the Kuroshio region (30°N,160°E), ii) along the outcrop line at 35°Nbetween 180°–140°W, and iii) the tropicsbetween 10°N–15°N. The variability inthe Kuroshio region reflects changes inthe ocean thermal structure resultingfrom the basin-wide changes in thestrength of the westerly winds thatoccurred in the late 1970s and late 1980s.The thermocline changes lag changes inthe basin-wide wind stress curl forcingby 4–5 years, consistent with the times-cale of oceanic adjustment throughRossby wave propagation. The secondcenter of variability is associated withsubduction, where it has been proposed

that thermal anomalies produced at thesurface primarily by anomalous heatfluxes, propagate equatorward alongisopycnals by the mean currents. Analy-ses of the OGCM and observations hasallowed us to track thermal anomaliesfrom their source region 25°N–35°N,140°W–170°W southwestward to ~18°Nover a period of ~8 years. South of thislatitude thermocline variability appearsto be driven by local wind forcing.

The isopycnal depth changes in the sub-tropics of both hemispheres are associ-ated with large thermocline temperaturevariations in both the OGCM and obser-

Fig. 5.10. (a) Evolution of the thermocline depth along 13.6°N (from east to west) as computed using a first-modebaroclinic Rossby wave equation forced with the Ekman pumping derived from the NCEP-NCAR reanalyses overthe period 1958–1997. The equation was solved using the method of characteristics. Contour interval is 10 m. Neg-ative values (shallower thermocline) are shaded in blue, while positive values (deeper thermocline) are shaded inred. (b) Same as in (a), but computed from the NCAR OGCM. The 25.5 isopycnal has been used as a proxy forthermocline depth. (c) Evolution of the depth of the 25.5 isopycnal along 130.8°W from 13.6°N to the equator. (d)Evolution of the depth of the 25.5 isopycnal along the equator, from east to west. In the OGCM anomalies originat-ing along 13.6°N can be tracked all the way to the equator and along the equator. (Based on results of Capotondiand Alexander, 2001, J. Phys. Oceanogr., in press).



vations. We have examined the variabil-ity at 10°N–15°N by comparing theOGCM results with those obtained froma simple Rossby wave model forced bythe same winds used in the OGCM simu-lation. The evolution of thermoclinedepth is remarkably similar in the simplemodel and OGCM (Fig. 5.10), indicatingthat a substantial portion of the variabil-ity in the 10°–15°N latitude band resultsfrom wind-forced baroclinic Rossbywaves. Spectra and Hövmoller diagramsbased on low-pass filtered OGCM outputindicate that most of the thermoclinevariability occurs at periods longer than~7 years. East of the dateline, subtropicalEkman pumping anomalies exhibit vari-ability over decadal periods and propa-gate westward at speeds close to thephase speed of first baroclinic modeRossby waves. Thus, the spectral charac-teristics of the forcing may be responsi-ble for the enhanced oceanic response atlow frequencies. The thermocline signalthat propagates across the basin at 13°N,

moves southward along the westernboundary and then eastward along theequator (see Fig. 5.10). The low-fre-quency variations of the thermoclinedepth along the equator may modulatethe amplitude and period of ENSOevents on decadal timescales, an out-come we plan to explore in the nearfuture.

5.3. Empirical studies of decadal vari-ability and climate change

5.3.1 Observed change in the globalhydrologic system

Interest in the potential impacts of cli-matic change in high elevation regionshas grown in the past decade, as informa-tion from glacial monitoring sites andfrom fieldwork has demonstrated thatsignificant melting and glacial recessionhas been occurring. Fifty-year trends innear-surface temperature (1948–2000)averaged over 5 different mountainous

Fig. 5.11. The 1948–2000 linear trend in surface area above the freezing level surface (FLS) for the tropics, NH,SH, and globe for each calendar month. Data is the NCEP/NCAR reanalysis. (Based on a study by Diaz, Eischeid,Duncan and Bradley 2001).



regions are similar in magnitude to glo-bal temperature trends (+0.4–0.6 °C/50yrs). We have been studying the relationbetween such behavior and changes inthe land surface area located above thefreezing level. The freezing level surface(FLS) is defined to be the 0°C isotherm,and the land area inside the roughly 2.5°lat/lon grid boxes of the NCEP/NCARReanalysis data that exceeds the FLSheight was calculated monthly during1948–2000. The most striking result isthe strong seasonal dependency of trendin the land surface area above the freez-ing level, with largest decreases in springconsistent with a warming, but nearlyequal increases in fall indicative of cool-ing (Fig. 5.11). Overall, the annualchange reveals a small net decrease.

We are studying the impacts of climatechange on a number of natural systems,including alpine hydrology and ecology,and the relevance for water resourcesmanagement. A prominent feature of lowfrequency variability in streamflow hasbeen the systematic change in timing ofpeak runoff over most of the globe since

1945 (Fig. 5.12). Peak flows are occur-ring earlier now than a half-century ago.This is particularly pronounced in riverbasins with a high fraction of theirstreamflow supplied by snowmelt, andreflects the springtime warming trend inthese regions.

5.3.2 Secular change in North Pacificcyclone activity

The picture emerging from our GCManalyses of wintertime atmospherictrends since 1950 is of a relationshipbetween changes in tropical SSTs andtrend patterns in zonally averaged flowand planetary waves, in the sense that theformer is forcing the latter. Given that thelarge scale circulation controls the statis-tics of sub-seasonal variations, such phe-nomena as the storm tracks and thetransients that define them should alsoexhibit secular change. As one example,we have documented an increase in thefrequency and strength of intense wintercyclones (minimum central pressurelower than 975 hPa) in the North PacificOcean since 1948 (Fig. 5.13). The time

0˚

60˚E

120˚

E

180˚

120˚W

60˚W

KENDALL’S TAU

tau = -0.50 tau = +0.25

0˚

60˚E

120˚

E

180˚

120˚W

60˚W

Fig. 5.12. Global trends in the time of streamflow during 1945–1993 as measured by the flow-weighted averageday of flows in extratropical rivers. Red circles denote retrogression of the annual hydrograph. (Based on a studyby Dettinger and Diaz, 2001).



series of December-March averagedcounts of cyclones having a minimumpressure below 975 mb shows a substan-tial increase since 1948.

Associated with these changes areupward trends in extreme surface windspeeds between 25°–40°N, and anincrease in significant wave heights. It issurmised that the cooling trend of NorthPacific SSTs since 1950 (see Fig. 5.1)has in part been driven by the anomalouslatent heat fluxes associated with thisenhanced storminess. We also postulatethat increasing sea surface temperaturesin the western Pacific warm pool region(see Fig. 5.1) is a cause of the observedcyclone changes. The NCAR CCM3simulations give some support for thispremise, insofar as changes in the west-ern Pacific warm pool and ENSO ampli-tude on decadal-scales impact themidlatitude stationary waves via telecon-nection processes.

5.3.3 Decadal variations in summertimemonsoons

Large summertime changes haveoccurred in tropical monsoon circula-tions since 1948 that are no less dramaticthan the aforementioned low frequencyvariations in the wintertime ocean/atmo-sphere system. We already alluded to thesecular change in ENSO's interannualimpacts on Indian summertime monsoonrainfall. We have examined the meanchange in the summertime monsoonssince 1948 using the NCAR/NCEP re-analysis circulation data, together withstation rainfall data where available. Aprominent change has occurred in thedivergent mass circulation describing the

summer monsoons of both western andeastern hemispheres (Fig. 5.14, top). Themaxima in the change map representroughly 20 percent of the climatologicalmean. Station rainfall data allow us toverify that drying has indeed occurredover the Sahel in recent decades, consis-tent with the trend toward strong lowlevel divergence (red shading) andimplied sinking motion in the re-analysisdata. We have yet to establish the realismof the re-analysis changes in the othersummertime monsoons of Asia and theAmericas, and a key hurdle in makingsense of this picture is determining thefidelity in the re-analysis data itself,which is known to be biased by somespurious trends during 1948–2000.

Fig. 5.13. Time series of the frequency of North Pacificwinter cyclones having minimum sea level pressure <975 mb during 1948–1997. The dark curve is thesmoothed 5-yr point average, and the dark line is thelinear trend. Data is the NCEP/NCAR reanalysis.(Based on results from Graham and Diaz, 2001, Bull.Amer. Meteor. Soc., in press).



To be sure, the re-analysis mass circula-tion changes are consistent with the re-analysis rainfall changes (compare topand middle panels of Fig. 5.14), and it isthe origin of the latter which requires anexplanation. We have begun to pursuethe possibility that such apparent tropi-cal-wide, low frequency changes in theatmosphere are reacting to a lowerboundary SST change, perhaps akin tothe wintertime change in tropical forcingdiscussed in earlier sections. The lowerpanel of Fig. 5.14 shows the change inrainfall as derived from a multi-GCM

ensemble of AMIP-style simulations of1950–1999. Rainfall has increased in thesimulations over oceanic regions (redshading), especially the warm pool of thewestern Pacific and the Indian Ocean.This is in broad agreement with the re-analysis. Note also that the general dry-ing (blue shading) over North Africa,Indonesia, and the Caribbean in re-analy-sis data emerges as a response pattern toSST. Our assessment is not complete,and we are pursuing various hypothesesmotivated from our empirical study.

EPILOGUE

It has become increasingly important toprovide attribution for low frequencyvariations and change in Earth's climate.Whether this is for improved scientificunderstanding, predictability assess-ment, or to better inform societal plan-ning and decision making, CDC isdedicating increasing resources to cli-mate change research. At CDC, we seekto offer dynamical explanations forobserved low frequency variations andchange, thereby drawing strongly uponour expertise on seasonal to interannualvariability, especially regarding air-seainteractions and teleconnective influ-ences.

A key challenge that we will pursue inCDC is to understand and anticipate theregional characteristics of climatechange. While there is now little questionthat the climate has changed in a globallyand annually averaged sense, it is unclearwhat the local manifestations of this are,nor do we appreciate their seasonaldependencies. Beyond its relevancy to

NCEP-Reana. 850 mb Velocity Potential (Chi): 1968-99 minus 1948-67 JJA

0 60 120 180 240 300 360-60

-40

-20

0

20

40

60

NCEP-Reana. Precip.: 1968-99 minus 1948-67 JJA

0 60 120 180 240 300 360-60

-40

-20

0

20

40

60

4-model Ave. Precip.: 1968-99 minus 1950-67 JJA

0 60 120 180 240 300 360-60

-40

-20

0

20

40

60

-110 -90 -70 -50 -30 -10 10 30 50 70 90 110

Fig. 5.14. Interdecadal change in the (top) 850mbvelocity-potential (5 x 106 m2 s-1), (middle) precipita-tion (mm month-1) of the NCEP/NCAR reanalysis,and (bottom) precipitation of the average of four atmo-spheric general circulation models (0.3 mm month-1).(Based on results of Quan, Diaz, and Fu, 2001, J. Cli-mate, submitted).



long term planning, this problem is ofhigh relevance to seasonal climate pre-dictions. The fact is that the leadingsource of US winter temperature skill inthe 1990's is due to the so-called optimalclimate normals (OCN) tool, which weunderstand to be essentially a trend pre-diction. It is necessary that a physicalexplanation for such trends be given, andthat they be clearly distinguished fromlow frequency climatic variations. Mostapparent of these trends is the US winter-time surface warming, but other seasonsshow a more complicated pattern fortemperature and rainfall change. Webelieve that progress can be made byimproving our understanding of theregional responses to the slow, system-atic changes in tropical oceans such asillustrated in Fig. 5.1, and we expect thatmuch is to be gained from our existingknowledge of the interannual impacts oftropical forcing.

The change in the oceans itself is a prob-lem that will focus future CDC decadalclimate research. The mean change inocean temperatures is a question that willrequire increased analysis of coupledocean-atmosphere models. We expect to

partner with GFDL, NCAR and otherinterested scientists to diagnose andunderstand the variability in coupledmodel simulations, both natural andforced. We are especially interested inthe sensitivity of ENSO to climatechange, both its statistical properties andits interannual global impacts. A relatedchallenge is to understand whether theyear-to-year predictability of climate willchange appreciably under the influenceof human-induced mean change. WillENSO as an oceanic phenomena becomemore predictable? Is it possible also thatnew regions will begin to have usefulENSO-related climate predictability?Likewise, we would like to understandwhether the seasonal cycle of predict-ability will change due to an alteredmean climate. These questions, amongothers, cut across time scales, and thegreatest payoff in solving them may infact be to advance key problems onshorter time scales, such as interannualprediction.

Contributed by: M. Alexander, A. Capo-tondi, H. Diaz, M. Hoerling, H. Huang,X. Quan, D. Sun, and K. Weickmann.

decadal climate and global change research · decadal climate and global change research 5.1...

Documents