VariationsV6N3 12/9/08 11:56 AM Page 1

us CLIVAR U.S. CLIVAR December 2008, Vol. 6, No. 3 VARIATIONS Reigns Summer Drought and Heat Waves in by David M. Legler, Director Southern Africa: Observations and recent workshop in Lincoln, Nebraska on Drought turned Coupled Model Behavior Aout to be a tremendously interesting and unprecedented gath- Bradfield Lyon ering of the operational modeling International Research Institute for Climate and Society, and forecasting community (e.g. Columbia University, NY, NY NCEP), those who deve lop rou- tine/operational products and out- tively correlated is of course well looks (e.g. NCEP, the National uring the severe summer known with the physical linkage being Drought Mitigation Ce nter), and the drought of 1991-92 in via changes in the surface energy budg- research community who are help- southern Africa it is esti- et. Drier-than-average conditions reduce ing to provide new insight and mated that as much as 3 soil moisture favoring an increase in the deve lop new capabilities. Working Dmillion tons of grain production were surface sensible heat flux and thus high- together and in harmony, these lost in this predominately rain-fed agri- er surface air temperature. In southern groups are mining for improve d cultural region (Dilley and Heyman Africa the tendency for below-average predictability of long-term drought, 1995). The extreme high temperatures rainfall and above-average temperatures exploring the causes of drought, that accompanied the drought not only are often seen, for example, during El and discussing how best to improve contributed to the crop losses but also to Niño events which tend to be associated the products and knowledge that widespread livestock mortality with the near-synchronous occurrence can be conveyed through a suite of (Sivakumar 2006) and stresses on of deficient rainfall and elevated surface services. regional water supplies (IPCC 2001). air temperatures on the seasonal time The joint occurrence of these two cli- scale (as in 1991-92). However, the rela- This issue of Variations pro- mate extremes is of consequence in any tionship between shorter periods (i.e. vides examples of some of the find- region of the globe although research several days) of extreme high tempera- ings of the US CL IVAR Drought has largely focused on the two phenom- tures (heat waves) and the occurrence of Working Group. The model runs the ena considered in isolation. As part of meteorological drought is less clear as Working Group deve loped are now the U.S. CLIVAR-led Drought in daily maximum temperatures are also ava ilable for anyone to analyze , Coupled Models Project (DRiCOMP) a affected by variations in cloud cover, and since they are global runs, they study was undertaken to examine how proximity to large water bodies, prevail- are a potential source of new well climate models have performed in ing wind direction, thermal advection, knowledge about drought in many simulating the observed climate of the etc. Indeed, for some coastal locations regions of the world. Working 20th Century with respect to generating of southern Africa, this study finds that Group members are preparing drought, heat waves and their joint heat waves often occur in association nearly ten manuscripts describing occurrence. The models were then uti- with strong downslope flow generated lized to assess, at least within (a subset Continued on Page Two by propagating synoptic weather sys- of) coupled space, how tems which are distinct from the larger the concomitant occurrence of heat scale anomalous atmospheric circula- IN THIS ISSUE waves and drought may change under tion associated with drought (not Summer Drought and Heat Waves in Africa increasing anthropogenic forcing. The shown)...... 1 primary focus was on the subcontinent The joint occurrence of heat waves Drought Working Group Activities...... 4 of southern Africa. and drought is also confounded by the Analysis of Drought Simulations ...... 6 On monthly and seasonal time prospect of . For exam- Carribean Low Level Jet ...... 8 scales the tendency for summer rainfall ple, climate change scenarios are in gen- Calendar ...... 12 and surface air temperature to be nega- eral agreement that land surface temper-

U.S. CLIMATE VARIABILITY AND PREDICTABILITY (CLIVAR) 1717 Pennsylvania Avenue, NW, Suite 250, Washington, DC 20006• www.usclivar.org VariationsV6N3 12/9/08 11:56 AM Page 2

Continued from Page One U.S. CLIVAR their analysis indicating predictability atures will increase in the 21st Century, heat wave was defined as a period of at of major SST patterns and trends, as generally increasing the probability of least 3 consecutive days with the daily well as soil moisture. Pe gion and extreme daily temperatures (though not maximum surface air temperature (Tmax) Kumar present one such effort in this necessarily in a spatially uniform manner, exceeding the 90th percentile during the issue. Another highlight of the e.g., Meehl and Tibaldi 2004). Yet these D e c e m b e r-Fe bruary season. The T m a x Workshop was reviewing the findings same models diverge considerably in percentiles were obtained by ranking the of the Drought in Coupled Models their projected changes in precipitation, daily data over the periods 1961-2000, (DRICOM P) funded research activi- which in the case of southern Africa has 1981-2000, 2046-65 and 2081-2100. ties. They inve stigated mechanisms some showing a drier, and others wetter, Drought conditions were based on differ- important for drought. Ke rry Cook’s climate compared with the 20th century ent indicators including the three month paper in this issue focuses on the role (IPCC 2007). Therefore, in the coming Standardized Precipitation Index (SPI3), of the Caribbean Lo w Leve l Jet in century there could potentially be more a related standardized drought index regional drought. U.S .CLIVAR’s activ- heat waves even in regions with upward (WASP, Lyon and Barnston 2005), and ities in drought are now ready to be trends in rainfall (even in the absence of one based on precipitation minus evapo- documented and published. Watch drought), or decreased precipitation may ration (P-E). Observational data included for 13 manuscripts from DRI COMP lead to more frequent , favoring daily maximum temperatures for stations PIs and others who attending the the joint occurrence of heat waves via across South Africa for 1961-2000 and Drought Workshop, along with the changes in surface heat fluxes. gridded monthly precipitation analyses Working Group manuscripts, that will The models utilized in the study were obtained from the Global Precipitation be collected into a special issue of primarily the GFDL2.0, ECHAM5, and Climatology Center (GPCC) for the same the Journal of Climate. CSIRO3.5 all of which were included in period. Drought-related research is the Coupled Model Intercomparison The observational results are general- not complete though. Much more Project (CMIP-3), their output data ly well-captured by the station analysis could be done to understand the cou- archived at the Program for Climate shown in Figure 1 for Kimberley, South pled -atmosphere-land nature Model Diagnosis and Intercomparison Africa located in the central interior of the of drought and its forcings. (PCMDI) website. For comparison with country (see Figure 2c). The uncondi- Moreove r, we don’t fully know how observations the 20th Century Climate in tional probability (relative frequency of well long-term drought can be pre- Coupled Model (20C3M) runs were uti- occurrence) of a heat wave occurring dur- dicted and we have n’t yet fully con- lized. Climate change projections were ing any given month is 0.32. The proba- sidered how to develop early warn- based on output from the same models bility of drought, defined as occurring ing capabilities for drought and its forced with the A1B sce- when SPI3 is < -1.0, is 0.16. However, evo lution. There is no doubt drought nario. The selection of the three models the conditional probability of a heat wave research will remain critical for used here was based on their general abil- occurring given that drought conditions advancing our capabilities. Only by ity to generate realistic ENSO character- exist is 0.77 or about 2.4 times the uncon- working with our colleagues in istics (e.g., AchutaRao and Sperber 2006) ditional probability (the conditional prob- GEWEX, and with the operational while also having daily maximum surface ability at Kimberley is greater than at sev- and observa tions/products communi- air temperature (and other) data available eral other stations, cf. Figure 2a, but in ties will we continue to make for both the 20C3M and A1B runs. A general it exceeds the unconditional prob- advancements that translate into improved products and services.

Variations Published three times per year U.S. CLIVAR Office 1717 Pennsylvania Ave., NW Suite 250, Washington, DC 20006 (202) 419-3471 [email protected] Staff: Dr. David M. Legler, Editor Cathy Stephens, Assistant Editor and Staff Writer © 2008 U.S. CLIVAR Permission to use any scientific material (text and fig - Figure 1. Times series of SPI3 (bars) and occurrence of heat waves (red dots) for ures) published in this Newsletter should be obtained individual months of the DJF seasons 1960-2000 at Kimberley, South Africa. The from the respective authors. Reference to newsletter materials should appear as follows: AUTHORS, year. solid line represents the SST anomaly averaged over the Nino3.4 region. The Title, U.S. CLIVAR Newsletter, No. pp. (unpublished correlation between Nino3.4 SST anomaly and SPI3 is -0.53. The Pr(Heat Wave) manuscript). This newsletter is supported through contributions to = 0.32; Pr(Drought SPI < -1) = 0.16, Pr(Heat Wave | Drought) = 0.77. The the U.S. CLIVAR Office by NASA, NOAA—Climate base period climatology used for both drought and heat waves is 1960-2000. Program Office, and NSF. Page 2 VariationsV6N3 12/9/08 11:56 AM Page 3

VARIATIONS observational period. However, relative to a 1981-2000 base period climatology the probability of a heat wave (Figure 3b) in the models shows a marked increase associated with a general upward trend in summer temperature in the models. Figure 3c shows that rela- tive to the 1960-2000 base period drought conditions become more likely Figure 2. The relative frequency of occurrence of the conditional probability of a across much of southern Africa during heat wave given drought conditions during the period 1960-200 based on a) sta- tion observations and b) coupled models (20C3M runs). Figure 3c shows the 2046-65. This result, however, is highly spatial domains used for the stations and models. The location of the Kimberley, model-dependent as can be seen by the South Africa station is shown by the arrow. dashed line in Figure 3c which indicates a tendency for wetter than average con- ditions when the analysis is based on ability across most stations). The tem- the 20C3M runs for the 1961-2000 peri- output from the CRNM model. Indeed, poral correlation between SPI3 and the od. The associated frequency distribu- there is not a consensus among all Nino 3.4 SST index is -0.53, indicating tions of the conditional probabilities are CMIP-3 models for drier conditions dur- that El Niño (La Niña) events are often shown in Figures 2a and 2b, respective- ing summer in the coming century as associated with drought (unusually wet) ly. The observations and models show suggested in the IPCC AR4 report. conditions. As such, the likelihood of a generally similar distributions with an However, modeled droughts are general- heat wave during an El Niño event is enhanced likelihood of heat waves dur- ly exacerbated when surface evaporation substantially enhanced relative to the ing drought relative to the unconditional is taken into account (i.e. using drought unconditional probability while during probabilities, which (as seen for indices based on P-E, not shown) in the La Niña it is markedly reduced. Figure Kimberley) are typically around 0.3. A1B runs. 1 also indicates that there has been a The joint occurrence of heat waves Overall, the joint occurrence of heat clear increase in the number of heat and drought for the models forced with waves and drought is generally similar in waves at Kimberley during the past two the A1B greenhouse gas scenario is sum- 20C3M runs and observations (condi- decades. marized in Figure 3a for model years tional probabilities typically between 0.5 The conditional probability of a heat 2046-65. A similar analysis was also and 0.7 for both sets of data). While this wave given the existence of drought performed for the model years 2081- behavior is similar in A1B runs when the conditions was computed for all 21 sta- 2100 (not shown). Heat waves are based two extremes are computed relative to tions located within the inner box on Tmax percentiles obtained from the future climate base states, from a practi- shown in Figure 2c (station data was DJF seasons 2046-65. The drought cal viewpoint it is likely more meaning- only available for South Africa), and index is also computed relative to this ful to consider how they vary relative to across all model grid points in the larg- new base period climatology. The joint the current climate. From the drought er, boxed domain (roughly 70-80 grid behavior (Figure 3a) does not show side, this includes consideration of points depending on model) based on notably different behavior than for the changes in atmospheric demand for

Figure 3. . a) The relative frequency of occurrence of the conditional probability of a heat wave given drought conditions computed from the model grid points within the larger domain in Figure 2c for model runs forced with the A1B greenhouse gases scenario. b) As in (a) but for the unconditional probability of a heat wave during the model years 1981-2000 (light shading) and 2046-65 (dark shading) using percentile rankings from 1981-2000. c) The relative frequency of drought index values for the model period 2046-65 using a 1960-2000 base period climatology. The solid line shows a normal distribution for the index and the dashed line is for the CRNM model.

Page 3 VariationsV6N3 12/9/08 11:56 AM Page 4

U.S. CLIVAR water in a warmer climate. The full study References Group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change is described in a manuscript currently AchutaRao, K., and K.R. Sperber, 2006: [Houghton, J.T. et al., (eds.)]. Ch. 10.2.1.4. ENSO simulation in coupled ocean-atmos - being prepared for submission to the Cambridge University Press, United Kingdom phere models: are the current models better? Journal of Climate. and New York, NY. Clim. Dyn., 27, 1-15. Lyon, B., and A.G. Barnston, 2005: ENSO Acknowledgements D i l l e y, M., and B.N. Heyman, 1995: and the Spatial Extent of Interannual This work was supported by a small grant ENSO and Disaster: Droughts, Floods and El P rec ipitation Extremes in Tropical Land for exploratory research from the Niño/Southern Oscillation Warm Events. A reas. Jo u rnal of Climat e, 1 8 , 5 0 9 5 - National Science Foundation, NSF-ATM Disasters, 19, 181-193. 5109. 07-39256. Daily maximum temperature IPCC, 2007: Climate Change 2007, The Meehl, G.A., C. Tebaldi, 2004: More Physical Science Basis. Working Group I data for stations across South Africa were Contribution to the Fourth Assessment Report I n t e n s e, M o re Fre q u e n t , and Longe r kindly provided by the South African Lasting Heat Waves in the 21st Century. of the Intergovernmental Panel on Climate Science, 305, 994-997. Weather Service. The comments of Tony Change. Ch. 11. Cambridge University Press, Barnston on this work are also appreciat- NY, NY. S i va k u m a r, M . V. K . , 2 0 0 6 : C l i m at e prediction and agriculture: current status ed. IPCC, 2001: Climate Change 2001: The and future challenges. Clim. Res., 33, 3- Scientific Basis. Contribution of Wo r k i n g 17. Overview of the Drought Working Group Activities

Siegfried Schubert and David Gutzler, co-chairs

h t t p : / / w w w. u s c l i v a r. o rg / O rg a n i z a t i o n / d r in which a number of land surface mod- he U.S. CLIVAR Drought ought-wg.html), there are two areas that els were driven for 10 years with the Working Group was established deserve particular attention. The first is same observations-based meteorological in November 2006 to “facilitate T the work done to develop drought indices forcing. A key result of the study was the progress on the understanding and pre- of relevance to model prediction and demonstration that, while the raw soil diction of long-term (multi-year) drought evaluation. The second is the effort to moisture output from the various models over North America and other drought- design, coordinate and implement a new shows considerable differences, the prone regions of the world, including an set of climate model simulations that results become much more similar after assessment of the impact of global address some of our key uncertainties they are suitably normalized. T h e change on drought processes (Gutzler about the role of sea surface temperature results are highlighted in Figure 1, which and Schubert 2007)”. The specific tasks forcing and land-atmosphere feedbacks shows that the different land surface of the working group were to 1) propose in the development and maintenance of models produce very similar information a working definition of drought and drought. about the temporal variability of soil related model predictands of drought, 2) moisture (and therefore drought) in many coordinate evaluations of existing rele- The work on drought indices (Koster parts of the world when the soil moisture vant model simulations, 3) suggest new et al 2008), while acknowledging the his- values from each model are normalized experiments (coupled and uncoupled) torical importance of traditional indices by their respective climatological means designed to address outstanding uncer- such as the Palmer Drought Severity and variances. This indicates for exam- tainties in the nature of drought, 4) coor- Index, focused attention on developing ple that one can use the current genera- dinate and encourage the analysis of new indices that take advantage of infor- tion of Land Surface Models to reliably observational data sets to reveal mation derived from land surface models monitor drought in the United States antecedent linkages of multi-year and data assimilation techniques. The Great Plains region. droughts and 5) organize a community effort in particular addressed the robust- workshop to present and discuss the ness of soil moisture in the current gen- The second major effort of the work- results. eration of land surface models as an indi- ing group was the coordinated (multi- cator of drought and more generally, institutional and multi-model) develop- This paper highlights the working hydro-climatic variability. This work ment of a set of idealized simulations group activities during its two-year provides a consistent assessment across designed to improve our understanding tenure. While the working group has models regarding drought prediction and of the physical mechanisms that link SST made progress on all five of the tasks simulation, and it supports the use of variations to regional drought, including listed above (the reader can for example land surface models in drought monitor- an assessment of the role of land-atmos- find summaries and links to existing ing activities. phere feedbacks (Schubert et al. 2009). model simulations and relevant observa- Specific questions addressed by the tional datasets at: The study took advantage of the GSWP-2 project (Dirmeyer et al. 2006) experiments include: What are mecha- Page 4 VariationsV6N3 12/9/08 11:56 AM Page 5

VARIATIONS

the results from all 8 combinations (warm and cold phases) of the Pacific and Atlantic patterns for the 5 AGCMs. A key result is that all the models agree that the combination of a cold Pacific and warm Atlantic (PcAw) tends to pro- duces the largest precipitation deficits, while the combination of a warm Pacific and cold Atlantic (PwAc) tends to pro- duce the largest precipitation surpluses. The various experiments and infor- mation on access to the output from the various model runs may be found at: f t p : / / g m a o f t p . g s f c . n a s a . g o v / p u b / d a t a / c l i 2 Figure 1. Average r between land surface model (LSM) root zone soil moisture v a r _ d r o u g h t _ w g / R E A D M E / w w w / i n d e x time series, for LSMs driven with the same meteorological forcing. (At a given .html location, an r2 value was computed for each pairing of LSMs in the study; the average of the r2 values is plotted.) The plotted values show the degree to which References the LSM products from the different models contain the same information on the Dirmeyer, P. A., X. Gao, M. Zhao, Z. time variability of soil moisture (from Koster et al. 2008). Guo, T. Oki and N. Hanasaki, 2006: The Second Global Soil Wetness Project (GSWP- 2): Multi-model analysis and implications nisms that maintain drought across the adjustment technique to nudge the cou- for our perception of the land surface. Bull. seasonal cycle and from one year to the pled model towards the imposed SST Amer. Meteor. Soc., 87, 1381-1397. next? What is the role of the different forcing patterns. Gutzler, D. and S. Schubert: The U.S. ocean basins, including the impact of El While the full set of experiments C L I VAR Working Group on Long-Te r m Nino/Southern Oscillation (ENSO), the covers a wide range of SST forcing pat- Drought,” U.S. CLIVAR Variations (Spring 2007, volume 5, No. 1). Pacific Decadal Oscillation (PDO), the terns and includes runs that examine the Atlantic Multi-decadal Oscillation Koster, R. D., Z. Guo, R. Yang, P. A. impact of land-atmosphere feedbacks, Dirmeyer, and K. Mitchell, 2008: On the (AMO), and warming trends in the glob- the baseline set of experiments examine Nature of Soil Moisture in Land Surface al ? What is the role of the land? drought response to the three leading Models. Submitted to J. Climate. To what extent can droughts develop patterns of annual mean SST variability. Schubert, S. and the extended USCLIVAR independently of ocean variability due drought working group, 2009: A USCLIVAR These consist of a global trend pattern, a P roject to Assess and Compare the to year-to-year memory that may be Pacific ENSO-like pattern, and an Responses of Global Climate Models to inherent to the land? Atlantic pattern that resembles the D rought-Related SST F o rcing Patterns: Overview and Results. To be submitted to J. A number of groups contributed Atlantic Multi-decadal Oscillation Climate . model runs to the project. NASA’s (AMO). Figure 2 shows, for example, Global Modeling and A s s i m i l a t i o n Office (GMAO) contributed runs made with version 1 of the NASA Seasonal-to- Interannual Prediction Project (NSIPP- 1) AGCM. NOAA’s Climate Prediction Center, with support from the Climate Test Bed, contributed runs made with the Global Forecast System (GFS) AGCM. NOAA’s Geophysical Fluid Dynamics Laboratory (GFDL) con- tributed runs made with the A M 2 AGCM, while the Lamont-Doherty Figure 2. The annual and continental United States mean responses for a) precipi- Earth Observatory of Columbia tation and b) surface temperature for all 8 combinations of the Pacific and Atlantic University contributed runs made with SST forcing patterns, for all 5 participating AGCMs. The colors refer to the forc- the NCAR CCM3.0 AGCM. NCAR ing patterns consisting of combinations of the Pacific (ENSO-like) and Atlantic contributed runs made with the CAM3.5 (resembling the AMO) SST patterns. For example, PcAw indicates an SST forcing AGCM. An additional set of runs was consisting of the cold phase of the Pacific pattern and the warm phase of the made by COLA/University of Miami Atlantic, while PcAn indicates the cold phase of the Pacific pattern and the neutral with the coupled (atmosphere-ocean) phase of the Atlantic pattern (climatological SSTs). CCSM3.0 model employing a novel Page 5 VariationsV6N3 12/9/08 11:56 AM Page 6

U.S. CLIVAR

Shmakin 2002), CAM3.5, and the GFS (Campana and Caplan 2005). These five Analysis of the multi-model models have diverse developments histo- ries, and comprise of both spectral and U.S. CLIVAR Drought Working Group grid-point models, different physical simulations parameterizations and spatial resolutions. The AGCM's were forced with a set Philip J. Pegion and Arun Kumar of idealized anomalous SST forcing. The SST anomalies were produced from a the atmospheric circulation that lead to rotated empirical orthogonal function he U.S. CLIVAR Drought work- drought. (REOF) analysis of the annual mean ing group formed in December HADISST (Rayner et al. 2003) for the The six modeling groups were NASA- T2006. Some of the goals of the years 1901-2004, these three leading GSFC, LDEO, GFDL, UMD/NCAR, working group were to coordinate evalu- REOFs together explain over half of the UM/COLA, NCEP/CPC. Five of the ations of existing model simulations, and inter-annual variance. The loading pat- groups, with the exception of UM/COLA to also coordinate new experiments terns and principal components are preformed the experiments with a global designed to address some of the out- shown in Fig. 1. The three patterns are atmospheric general circulation model standing uncertainties related to drought referred to as the Trend, Pacific, and (AGCM) forced with the identical SSTs. variability and predictability. Six model- North Atlantic patterns. The purpose of UM/COLA did the experiments with a ing groups joined in the effort and pro- the idealized experiments was to isolate coupled model. The models were the duced a wealth of model simulation data the influence of each pattern separately NSIPP1 AGCM (Bacmeister et al. 2000, that will be analyzed for many years. as due to the short historical record of Schubert et al. 2004), CCM3 (Kiehl et al. The primary goal of model experiments SST observations, it is difficult to isolate 1998, Seager et al. 2005), GFDL AM2.1 was to quantify the role sea surface tem- the impact of each SST mode in AMIP (Delwoth e al 2006, The GFDL Model perature (SST) variability on changes to type simulations or in observations. Development Team 2004, Milly and The sets of experiments consisted of each AGCM run with a repeating season- al cycle of climatological SSTs, and also with fixed anomalies added to the repeat- ing seasonal cycle. To assess the role each of SST patterns has on climate vari- ability, the patterns are scaled by +/- the standard deviation of the principal com- ponent. The trend is scaled by 1 standard deviation, which roughly represents a period of 1901-1942 for the negative scaling, and 1965-2004 for the positive period. The other EOFs are scaled by twice the standard deviation to empha- size their atmospheric influence. The anomaly patterns are then added to a repeating climatological seasonal cycle of the HADISST data, with the climatol- ogy based on the 1901-1999 period. The treatment of sea-ice was left to the indi- vidual modeling groups, but an ice extent climatology was specified. To isolate the role of each of the SST patterns, the anomalies for the Pacific pattern were set to zero in the Atlantic Basin, and con- versely, the anomalies in the Pacific and Indian oceans were set to zero in the Figure 1. The three leading rotated EOFs of HADISST annual mean SST for North Atlantic forcing pattern. 1901-2004. Combined these three EOFs explain over 52% of the inter- annual variance. Each modeling center ran many long simulations with the idealized SST pat- Page 6 VariationsV6N3 12/9/08 11:56 AM Page 7

VARIATIONS

terns, and a control run with the climato- logical SSTs. The runs varied between 36 years (NCEP GFS) to 51 years for the other models with the first year discarded as spin-up. Both the positive and nega- tive polarity of each pattern was used to force the AGCM. In addition, various combinations of these patterns were used in additional experiments. In total, there were 15 different experiments run with the SST patterns. The results presented in this paper are for the model response to each of the individual SST patterns, and does not address how the atmospher- ic response to the SST patterns interplay with each other, or the relative role of the tropical vs. extra-tropical SSTs. This analysis looks at agreement between models responses the different SST forcing patterns; as such an agree- ment enhances our confidence in AGCM based results. The response is defined as the multi-model mean of the positive polarity minus the negative polarity experiments. The annual mean response Precip in temperature and precipitation for the 3 leading patterns is shown in Fig. 2. Temp Shading indicates model agreement, on the sign of the anomalous response. The Figure 2. The multi-model response to the three patterns, precipitation (left) and precipitation response to the trend SST surface temperature (right). Contours show the model mean response, and shad- patterns is mainly confined over the ing indicated where at least 4 out of the 5 models (3 out of 4 for the Trend pat- ocean, and coastal regions. The response terns, as CCM3.5 did not complete a negative Trend experiment) produced the includes an increase in precipitation over same sign response. Units are mm/day for precipitation and oK for temperature. the southern tip of Indian, extending east- ward through the maritime continent, North Atlantic SST pattern shows an with a decrease of precipitation over the 105oW. The probability density func- increase over the regions of positive central equatorial pacific. There is also tion (PDF) of annual mean precipitation SST anomalies, and reduced precipita- an increase in precipitation over Central anomalies for the Great Plains is shown tion for most of the globe, with the America and a warming over most of the in Fig. 3. Although the mean precipita- exception over the Indonesian region. globe. tion response to the Trend is fairly small The surface temperature response is a (0.04 mm/day), it does have a role in The Pacific SST pattern shows a large scale warming that extends across shifting the odds of having long term more robust response, with increased northern Africa and into southern Asia. droughts. As a comparison, the Dust precipitation over most of the equatorial There is also a signal of increased sur- Bowl's precipitation anomalies over the Pacific, a decrease over the maritime face temperature over the southern US great plains average for the years continent. There is also a decrease in pre- Great Plains, and Mexico, and central 1932-1938 was only -0.2mm/day based cipitation over the Amazon and extend- South America. ing across the tropical Atlantic Ocean on GHCN gridded precipitation data. The impact of SST patterns over the into Africa, and an increase in precipita- The PDFs for the Atlantic and Pacific United States is dominated by the tion over North America and the southern SST experiments have greater separa- Pacific SST pattern, but upon closer portion of South America. The tempera- tion. inspection, the precipitation response to ture response shows warming over the In order to quantify the change in the the Trend and Atlantic SST forcing are tropical land masses and over northwest- odds of having long term drought, all of also statically significant at the 5% level ern Canada into Alaska, and a cooling the models were pooled together and re- when average over United States Great over the central United States. sampled to calculate the odds of having Plains, which is defined by the region 3-years with below normal precipitation The precipitation response to the within 30oN to 50oN, and 95oW to Page 7 VariationsV6N3 12/9/08 11:56 AM Page 8

U.S. CLIVAR

(GFDL) for making the AM2.1 runs avail- able, the National Center for Atmospheric Research (NCAR) for making the CAM3.5 runs available, and the Center for Ocean Land Atmosphere (COLA) and the University of Miami's Rosenstiel School of Marine and Atmospheric Science for making the CCSM3.0 coupled model runs available. References Bacmeister J., P. J. Pegion, S. D. Schubert, and M. J. Suarez, 2000: An atlas of seasonal means simulated by the NSIPP 1 atmospheric GCM. Vol. 17. NASA Te c h . Memo. 104606, Goddard Space Flight Center, Greenbelt, MD, 194 pp. Campana, K. and P. Caplan, Editors, 2005: Technical Procedure Bulletin for T382 Global Forecast System ( h t t p : / / w w w. e m c . n c e p . n o a a . g o v / g c _ w m b / D o c u m e n t a t i o n / T P B o c t 0 5 / T 3 8 2 . T P B . F I N A L . h t m). Delworth et al., 2006. GFDL's CM2 Figure 3. Probability density function of annual mean precipitation over the US global coupled climate models - Part 1: Great Plains (95W-105W, 30N-50N) for all of the models combined (235 years Formulation and simulation characteristics, of simulation for the Trend forcing (top) and Pacific forcing (middle), and 185 J. Climate, 19, 643-674. years for the Atlanticpattern (bottom). Solid curve is from the warm (positive) The GFDL Global Atmospheric Model forcing, and the dashed is for the negative pattern. Units are mm/day. Development Team, 2004. The New GFDL Global Atmosphere and Land Model AM2- LM2: Evaluation with Prescribed SST in a row. The odds were calculated ran- caveat that these are all AGCMs with Simulations. J. Climate, 17, 4641-4673. domly selecting 3 years of precipitation are just responding to SST anomalies Kiehl, J.T., J.J. Hack, G. Bonan, B.A. with replacement, and counting how and the possible influence of coupled Boville, D. Williamson and P. Rasch, 1998. many times the sample contained all 3 air-sea interaction is ignored. That being The National Center for A t m o s p h e r i c years of negative precipitation anom- said, we have shown that the Pacific pat- R e s e a rch Community Climate Model: alies. The re-sampling was done 10,000 tern is the dominant forcing on precipi- CCM3. J. Climate, 11, 1131-1149. times for each pattern. If a distribution is tation variability globally, which is not Milly, P. C. D., and A. B. Shmakin, 2002. normally distributed, the expected prob- Global modeling of land water and energy surprising since that pattern also has the balances. Part I: The land dynamics (LaD) ability for 3-years in a row of below nor- largest amplitude. The Atlantic and model. J. Hydrometeorology, 3, 283-299. mal precipitation is 12.5%. The Monte Trend SST anomalies do also produce R a y n e r, N. A., D. E. Parker, E. B. Carlo estimate for the warm Pacific pat- significant changes in precipitation, just Horton, C. K. Folland, L. V. Alexander, D. P. tern shows only a 0.1 % chance of hav- more localized, but the temperature Rowell, E. C. Kent, and A. Kaplan, 2003: ing 3 years in a row of below average response to the Atlantic and Trend does Global analyses of sea surface temperature, precipitation, compared to 79% when the have worldwide impacts. sea ice, and night marine air temperature Pacific is cool. The Atlantic forces a since the late nineteenth century, J. Geophys. Acknowledgement Res., 108(D14), 4407. shift in probability from 36.7% chance of Schubert, S.D. M.J. Suarez, P.J. Pegion, having 3 dry years in the warm phase This work was carried out as part of a U.S. C L I VAR drought working group activity R. D. Koster, and J.T. Bacmeister, 2004: versus a 3.9% chance when it is cool. supported by NASA, NOAA, and NSF to Causes of long-term drought in the U.S. The Trend shifts the chance to 24.0% for coordinate and compare climate model sim- Great Plains,. J. Climate, 17, 485-503. the warm pattern compared to 7.0% ulations forced with a common set of ideal- Seager, R., Y. Kushnir, C. Herweijer, N. when the SSTs were cool. The climatol- ized SST patterns. The authors would like to Naik and J. Velez, 2005: Modeling tropical ogy runs for all of the models pooled thank NASA's Global Modeling and forcing of persistent droughts and pluvials: 1856-2000. J. Climate, 18, 4068-4091. together give a chance of 13.7 percent Assimilation Office (GMAO) for making the chance of having 3 dry years. The devi- NSIPP1 runs available, the Lamont-Doherty ation from 12.5% is due to the skewness Earth Observatory of Columbia University of precipitation. for making their CCM3 runs available, NOAA's Climate Prediction Center Because these experiments were done (CPC)/Climate Test Bed (CTB) for making with many different AGCMs, more con- the GFS runs available, NOAA's fidence can be put in the results, with the Geophysical Fluid Dynamics Laboratory Page 8 VariationsV6N3 12/9/08 11:56 AM Page 9

VARIATIONS

relatively strong all year as the tropical easterly flow funnels between the north The Caribbean Low-Level Jet: Regional coast of South America and the islands Dynamics and its Relationship to of the Greater Antilles (Cuba, Puerto Rico, Jamaica, and Hispaniola). Precipitations here is a distinct difference in the large-scale context of the CLLJ between Kerry H. Cook and Edward K. Vizy the summer and winter seasons. During Department of Geological Sciences and Institute for Geophysics the months of October through April, the Jackson School of Geosciences jet flow continues westward over the The University of Texas at Austin Caribbean Sea, finally obtaining a south- ward component in meeting the orogra- The Caribbean low-level jet nitude of the meridional SLP and tem- phy of Central America and entering the (CLLJ) is an easterly jet located perature gradients in the NARR, eastern Pacific basin. In contrast, during Tover the Caribbean Sea off the explaining that the CLLJ is, to first the months of May through September, northern coast of South America, south order, geostrophic between 925 and 800 the CLLJ splits into two branches over of Hispaniola (Haiti and the Dominican hPa. The u-momentum balance is not at the central Caribbean Sea. One branch Republic). It is present throughout the all geostrophic, with advection and fric- proceeds westward and enters the Pacific year and transports large amounts of tion similar in magnitude to Coriolis and basin while the northern branch flows moisture, forming a primary connection pressure gradient accelerations. northward connecting the CLLJ with moisture transport throughout the between the Gulf of Mexico and the A complete analysis of the horizontal tropical Atlantic Ocean. For most of the Caribbean and Gulf of Mexico, and with momentum balance confirms that the the climatological GPLLJ and water year the jet core is located very close to seasonal cycle of the CLLJ is primarily the surface, at 925 hPa, but in the fall the vapor transport into the central U.S. due to seasonal variations in the positive Differences between these two states, jet core lifts higher, up to 800 hPa in meridional geopotential height gradient September and October. shown in Figure 2, indicates that the dif- between far northern South America and ference in the circulation essentially The CLLJ is depicted near the edge the central Caribbean. In boreal sum- amounts to a northerly wind anomaly of the North American Regional mer, this gradient is strong because of that is related to the development of pos- Reanalysis (NARR; Mesinger et al. the position of the North Atlantic sub- itive zonal height gradients associated 2006) domain (Fig. 1), providing an tropical high (NASH) to the north of the with summer heating over Central opportunity for a high-resolution (32 km CLLJ. The secondary jet speed maxi- America and southern Mexico. x 32 km) investigation of the regional mum in boreal winter occurs because dynamics and, since precipitation is low geopotential heights develop to the Dynamics of the CLLJ: Diurnal assimilated in the NARR, an analysis of south as part of the South America mon- Variations the relationship between the CLLJ and soon system, and this also generates Unlike the climatological GPLLJ, which rainfall over the central U.S, Mexico, strong positive meridional gradients is composed primarily of individual low- Central America, and the Caribbean. across the CLLJ region. Minima in the level jet events that develop over the Note that, at 925 hPa, the highest wind CLLJ magnitude in the transition sea- central U.S. at night in the summer speeds in the NARR domain are those of sons reflect the absence of these months, the diurnal cycle of the CLLJ is the CLLJ. enhancements to the positive meridional not very dramatic. There are two wind Dynamics of the CLLJ: Seasonality gradient that is nonetheless present and speed minima during each 24 hour peri- The CLLJ is present all year, with a Figure 1. peak monthly-mean magnitude in July of Zonal wind about 12 m/s. The jet is also strong in speed at 925 December at January, with wind speeds hPa in July of 8-9 m/s. Minima of 6-8 m/s occur in over the May and October. This semiannual sea- NARR sonal cycle is in contrast to that of the domain for climatological Great Plains low-level jet the 1979- (GPLLJ), which forms only in boreal 2007 clima- summer (if one takes a wind speed of 3 tology. m/s as the threshold velocity to define a Contour inter- “jet”). Muñoz et al. (2008) relate sea- val is 2 m/s. sonality in the magnitude of the CLLJ to a similar semiannual cycling of the mag- Page 9 VariationsV6N3 12/9/08 11:56 AM Page 10

U.S. CLIVAR the monthly-mean 925 hPa wind speed in an area bounded by 75-70°W and 12- Figure 2. Differences in 15°N, and standard deviations from the 925 hPa geopotential 1979-2007 monthly mean zonal wind heights (contour interval is 5 speed are calculated. Strong and weak gpm) and winds (vector CLLJ months are selected based on these scale is in m s-1) between standard deviations, requiring that a the May – September and strong jet exceeds the mean by one stan- October – April averages dard deviation and that a weak jet is at for 1979-2007 from the least one standard deviation below the NARR. Shading denotes mean. Of the 348 months (29 years) ana- negative geopotential height differences. lyzed, 16% (58 months) are identified as strong CLLJ months, and 14% (50 months) as weak CLLJ months. The dis- tribution skews a bit toward strong CLLJ events because of an apparent trend after od throughout the year, at about 4 AM related on sub-seasonal time scales sug- 2000. Correlations between the standard and at 4 PM. A progressive weakening gests that interannual rainfall variations deviations of the CLLJ index and rainfall of the jet through the afternoon is more – including, perhaps, the occurrence of were calculated over the NARR time pronounced than the nighttime weaken- drought in the central U.S., Central period (not shown) and composite rain- ing, so the 4 PM minimum is more dis- America, and Mexico – may be associat- fall distributions for strong and weak tinct. This afternoon minimum is ed with CLLJ variations. CLLJ months were assembled. explained by considering meridional wo approaches were taken to under- Conclusions draw from these two analy- accelerations, which couple to the zonal stand if there is a relationship between ses, presented below, are consistent. flow through Coriolis accelerations. The the CLLJ and drought. One was to ask if To increase the numbers of events meridional acceleration is southward the CLLJ tends to be anomalous during composited and focus on diff e r e n c e s during the day and northward at night, times of drought (e.g., individual dry between the two states of the CLLJ-relat- and one cause of this diurnal sign change years). For the time period covered by ed flow, results are presented for the is variations in the meridional height the NARR climatology (1979-2006), summer (MJJAS), when the CLLJ has a gradient which strengthens from 4 AM three summers emerge as exceptionally northward component, and for the rest of to 4 PM, and weakens from 4 PM to 4 dry in the central U.S (100-90°W and the year (the ONDJFMA mean) when it AM. This is caused by the diurnal cycle 30-50°N), namely, 1980, 1988, and flows to the west. This places 24 month- of heating over northern South America. 2007, consistent with other data sets and ly-mean events in the “strong CLLJ” The center of the daytime heating is over studies. The average CLLJ for these composite and 25 in the “weak CCLJ” the northern Andes, presumably in asso- three drought years is relatively weak, composite. ciation with sensible heating over the especially in June and August. But an elevated surface. So the meridional examination of the individual years Figure 3a displays contours of the wind variation is a land/sea breeze (sole- shows that the jet was only weak in MJJAS mean precipitation (mm day-1) noidal circulation), with anomalous 1988, and this dominates the small-sam- and the vertically-integrated water vapor southward acceleration during the day ple-size composite. There is no consis- transport (?105 kg m-1) from the NARR and anomalous northward flow at night, tent signal among the 3 years. climatology. Falsely low rainfall rates, and this contributes to the weak diurnal for example over Cuba, Hispaniola, and ver Central America and southern variation of the CLLJ. along the Texas/Mexico border, are relat- Mexico (100-80°W and 10-20°N), only ed to either a lack of reported rainfall Relationship of the CLLJ to one year (2003) is identifiable as a observations or the transition from the Precipitation: A Drought Connection? drought year in the NARR period. Since dense observing network over the U.S. Magaña et al. (1999) relate seasonal the purpose of this study is to understand From a visual examination of the verti- changes in the CLLJ to the occurrence of the CLLJ’s relationship to drought in a cally-integrated moisture flux climatol- the so-called “midsummer drought”, general, climatological sense, this ogy in Fig. 3a it appears that the north- which is a normal feature of the southern approach was abandoned. Instead, a sec- ward branch of the CLLJ and the GPLLJ Mexico and Central America seasonal ond approach was pursued in which we are part of the same system (see also cycle defined by lower rainfall rates in focus on times with strong and weak Mestas-Nuñez et al. 2007), suggesting July and August as compared with June CLLJs, and ask if and where precipita- that they may vary in concert. But the and September/October. Even though tion is anomalous at the same time. moisture transport and precipitation this feature is not a drought in the sense To explore connections between the anomalies from the composites of strong of being an anomalous or extreme event, strength of the CLLJ and regional rain- and weak CLLJ events (Figs. 3b and c, the fact that the CLLJ and rainfall are fall within the NARR domain we first respectively) show clearly that the north- define a CLLJ index as the magnitude of Page 10 VariationsV6N3 12/9/08 11:57 AM Page 11

VARIATIONS ward flow across the Gulf of Mexico is Atlantic warm pool and the central U.S. ited, as seen in Fig. 3e, it is clear that a not enhanced (diminished) when the (e.g., as suggested by Wang 2007) when strong CLLJ during these non-summer CLLJ is strong (weak) during the sum- the CLLJ is used as a selector. months is accompanied by the formation mer months. Instead, strong (weak) According to the NARR, rainfall of a meridional CLLJ branch that is sim- CLLJ wind speeds are associated with rates are enhanced off the east and west ilar to the summer climatological pattern enhanced (reduced) moisture transport coasts of Panama and Costa Rica and (Fig. 3a). During these strong jet events, into the Caribbean from the northeast, reduced over land when the CLLJ is moisture is transported across the Gulf high (low) rainfall rates over Hispaniola, strong. It is not exactly clear how or if of Mexico and into the southern Great and strong (weak) easterly moisture this signal is related to the findings of Plains and the southeastern U.S. advection across the western Caribbean. Magaña et al. (1999), who suggest that a Positive rainfall anomalies centered in The resulting moisture diverg e n c e strong CLLJ is associated with high oro- Louisiana and Texas occur as a result. anomalies over the central and western graphic rainfall in eastern Panama and When the ONDJFMA CCLJ is weak Caribbean are associated with negative Costa Rica and drying to the west due to (Fig. 3f), there is southwestward anom- (positive) rainfall anomalies in the a rainshadow effect. We examined the alous moisture transport across the strong (weak) CLLJ case. Note that response as in Figs. 3b and c at high res- southern Great Plains and low rainfall. there is no evidence of a dipole pattern olution, taking advantage of NARR’s 32 Closing Remarks of precipitation anomalies between the grid spacing, and did not find rainfall Despite the apparent relationship enhancement over between the CLLJ and the GPLLJ, we land that would found no systematic connection between indicate the oro- variations in the CLLJ and summer graphic mecha- drought in the Great Plains in examining n i s m / r a i n s h a d o w the NARR for 1979-2007. However, the e f fect. However, most extreme dry year of the period for the rainfall assimi- the central U.S. was 1988, and the CLLJ lated into the was quite weak that year so a case-study NARR in these approach might be useful for uncovering regions has a coars- a drought mechanism relevant to other er resolution than time periods, past or future. Further the grid onto which work on the non-summer months would it is interpolated, so also be useful, for example, for fall and perhaps the rela- early spring in isolation to understand tionship emerg e s the dynamics of seasonal transitions that when station data is may be informative about the dynamics used. of how drought periods begin and are Figure 3d dis- broken. plays contours of References the ONDJFMA Magaña, V., J.A. Amador, and S. Medina, mean precipitation 1999: The midsummer drought over Mexico (mm day-1) and the and Central America. J. Climate, 12, 1577- 1588. v e r t i c a l l y - i n t e g r a t- M e s i n g e r, F., and Coauthors, 2006: ed water vapor North American regional reanalysis. Bull. transport (?105 kg Amer. Meteor. Soc., 87, 343-360. m-1) from the Mestas-Nuñez, A.M., D.B. Enfield, and NARR climatology. C. Zhang, 2007: Water vapor fluxes over the In contrast to the intra-America sea: Seasonal and interannual MJJAS climatology variability and associations with rainfall. J. Figure 3. May – September (a) climatological mean precip- (Fig. 3a), there is no Climate, 20, 1910-1922. itation (mm day-1) and vertically integrated water vapor evident connection Muñoz, E., A. Busalacchi, S. Nigam, and A. Ruiz-Barradas. 2008: Winter and summer transport (?105 kg m-1), and (b) strong and (c) weak between the CLLJ Caribbean low-level jet composite minus climatological structure of the Caribbean low-level jet. J. and the transport of Climate, 21, 1260-1276. mean precipitation and vertically integrated water vapor moisture into the transport difference. Additionally, October – April (d) cli- Wang, C., 2007: Variability of the central U.S. and Caribbean low-level jet and its relations to matological mean precipitation and vertically integrated Mexico during climate. Climate Dynamics, 29, 411-422. water vapor transport, and (e) strong and (f) weak these months. But Caribbean low-level jet composite minus climatological when strong CLLJ mean precipitation and vertically integrated water vapor events are compos- transport difference. Page 11 VariationsV6N3 12/9/08 11:57 AM Page 12

U.S. CLIVAR Calendar of CLIVAR and CLIVAR-related meetings Further details are available on the U.S. CLIVAR and International CLIVAR web sites: www.usclivar.org and www.clivar.org

89th AMS Annual Meeting Workshop on El Nino and Climate 11-15 January 2009 Change CLIVAR Working Group on Ocean Phoenix, Arizona 25-26 March 2009 Modeling: Mesoscale Eddies Workshop Attendance: Open Perth, Australia 27-29 April 2009 Contact: http://www.ametsoc.org Attendance: Open Exeter, UK Contact: http://www.greenhouse2009.com Attendance: Open U.S. CLIVAR Western Boundary Contact: Current Workshop CLIVAR Pacific Implementation Panel http://www.metoffice.gov.uk/confer- 15-17 January 2009 Meeting ence/mesoscale_workshop/ Phoenix, Arizona 27-28 March 2009 Attendance: Open Perth, Australia 1st US Atlantic Meridional Overturning Contact: Attendance: Invited Circulation Meeting http://www.usclivar.org/WBCWorkshop2 Contact: http://www.clivar.org 4-6 May 2009 009.php Annapolis, Maryland 3rd ARGO Science Meeting - The Attendance: Open CLIVAR Working Group on Seasonal Future of ARGO Contact: to Interannual Prediction 25-27 March 2009 http://www.AtlanticMOC.org/AMOC200 12-14 January 2009 Hangzhou, China 9.php Miami, Florida Attendance: Open Attendance: Invited Contact: World Ocean Conference Contact: http://www.clivar.org http://www.argo.ucsd.edu/ASW3.html 11- 15 May 2009 Manado, Indonesia AMMA-Ocean/TACE/PIRATA meeting WCRP Joint Scientific Committee Attendance:Open 2-6 February 2009 Meeting Contact: Toulouse, France 6-9 April 2009 http://www.woc2009.org/home.php Attendance: Limited College Park, Maryland Contact: http://amma-international.org Attendance: Invited NASA Ocean Vector Wind Science Contact: http://www.clivar.org Team Meeting Ninth International Conference on 18-20 May 2009 Southern Hemisphere Meteorology European Geophysics Union General Woods Hole, MA and Oceanography Assembly Attendance: Invited 9-13 February 2009 19-24 April 2009 Contact: http://coaps.fsu.edu/scatterome- Melbourne, Australia Vienna, Austria try/meeting/ Attendance: Open Attendance: Open Contact: http://9icshmo.org Contact: http://www.egu.org CLIVAR/GOOS Indian Ocean Panel 3-5 June 2009 NOAA Climate Predictions CLIVAR Working Group on Ocean Le Reunion, France Applications Workshop Modeling: Panel Meeting Attendance: Invited 24-27 March 2009 30 April - 1 May 2009 Contact: http://www.clivar.org Norman, Oklahoma Exeter, UK Attendance: Open Attendance: Invited Contact: http://climate.ok.gov/cpasw/ Contact: http://www.clivar.org

U.S. CLIVAR OFFICE 1717 Pennsylvania Avenue, NW Suite 250 Washington, DC 20006 U.S. CLIVAR contributes to the CLIVAR Program and is a member of the World Climate Research Programme Subscription requests, and changes of address should be sent to the attention of the U.S. CLIVAR Office ([email protected])