Relationship between atmospheric circulation and winter precipitation

D18O in central New York State

Adam W. BurnettDepartment of Geography, Colgate University, Hamilton, New York, USA

Henry T. MullinsDepartment of Earth Sciences, Syracuse University, Syracuse, New York, USA

William P. PattersonDepartment of Geological Sciences, University of Saskatchewan, Saskatoon, Saskatchewan, Canada

Received 22 July 2004; revised 12 October 2004; accepted 1 November 2004; published 25 November 2004.

[1] Oxygen isotope values of meteoric precipitation (d18O)are strongly influenced by water vapor source andtrajectory history, and can therefore be used as a tool forthe reconstruction of atmospheric circulation. However, thisapproach requires an understanding of how differing patternsof atmospheric circulation influence precipitation d18O. Thisstudy examines the relationship between atmosphericcirculation and winter precipitation d18O in central NewYork State. Circulation back trajectories, weather maps, andd18O values for 132 precipitation samples were examined todetermine the circulation type for each event. Lake-effectprecipitation, which generated the lowest d18O values,featured low pressure over New England and northwesterlyflow over the Great Lakes. Events with the highest d18Ovalues were associated with low pressure north of New Yorkand strong southerly flow within the warm sector.Less frequent were the Atlantic coastal and warm frontaloverrunning events, both of which yielded relatively depletedprecipitation. INDEX TERMS: 3309 Meteorology and

Atmospheric Dynamics: Climatology (1620); 3344 Meteorology

and Atmospheric Dynamics: Paleoclimatology; 3354 Meteorology

and Atmospheric Dynamics: Precipitation (1854); 3364

Meteorology and Atmospheric Dynamics: Synoptic-scale

meteorology. Citation: Burnett, A. W., H. T. Mullins, and W. P.

Patterson (2004), Relationship between atmospheric circulation

and winter precipitation d18O in central New York State, Geophys.

Res. Lett., 31, L22209, doi:10.1029/2004GL021089.

1. Introduction[2] d18O values of carbonates, ice cores, and plant cellu-

lose represent an important archive that can serve asvaluable tools in paleoclimate studies. Among other things,these records provide insight on the oxygen isotope value ofmeteoric precipitation contemporaneous with their develop-ment [Anderson et al., 2002; Vinther et al., 2003]. Factorsthat control d18O values of meteoric precipitation involvefractionation processes associated with the evaporation andcondensation history of the precipitating water vapor. Thesefactors include the temperature at which condensationoccurs, the amount of precipitation, the isotope value ofthe water vapor source, and the degree to which the watervapor has traveled over land [Dansgaard, 1964; Rozanski etal., 1993; Araguas-Araguas et al., 2000]. Because these

processes are tied to the storm system dynamics associatedwith the precipitation, records of isotopic variability canbe used to study paleo-atmospheric circulation. Manyrecent paleoclimate studies using d18O values of lacustrinecarbonates have interpreted isotopic variability as a functionof changing atmospheric circulation, rather than a simple airor lake water temperature signal [Edwards and Wolfe, 1996;Teranes and McKenzie, 2001; Kirby et al., 2002]. Further-more, recent modeling efforts have shown that the interpre-tation of isotopic data using only local meteorologicalconditions, such as temperature, can lead to incorrect pale-oclimate reconstructions [Noone and Simmonds, 2002].[3] Developing an atmospheric circulation history using

d18O archives, however, requires an understanding of theways in which circulation influences the isotope value of themeteoric precipitation at the site in which the archives aredeveloped. One approach is to use modern precipitationisotope and circulation records to develop statistical rela-tionships that can be used to interpret the paleorecords.Unfortunately, modern records of precipitation d18O arelimited, with the most extensive records contained in theInternational Atomic and Energy Agency (IAEA)Global Network of Isotopes in Precipitation (GNIP) dataset.The coarse spatial distribution and short period of record formany sites in the GNIP dataset limit their use in circulationstudy. The monthly resolution of the GNIP data also masksthe influence of individual storms on the d18O record.[4] This study presents the initial findings of an ongoing

effort to understand the relationship between winter atmo-spheric circulation and d18O of individual precipitationevents for central New York State, which is an area inwhich little work of this type has been performed. Thisregion is interesting because of its location near severallakes in which paleoenvironmental records have been, andare being, developed [Mullins, 1998; Mullins et al., 2003].Also, this is a region that receives significant winterprecipitation in the form of Great Lake-effect snowfall.The impact of Great Lake derived moisture on precipitationisotopic variability and the associated sedimentary recordsof this region are undocumented.

2. Data and Methods

[5] This study uses d18O and dD values derived from 132winter (November–March) precipitation samples collected

GEOPHYSICAL RESEARCH LETTERS, VOL. 31, L22209, doi:10.1029/2004GL021089, 2004

Copyright 2004 by the American Geophysical Union.0094-8276/04/2004GL021089$05.00

L22209 1 of 4

daily near Colgate University (42.8�N, 75.6�W) between1999 and 2003. These samples are part of an effort todevelop a precipitation isotope record of sufficient lengthand resolution to study the impact of individual storm typeson precipitation isotopic variability and aid in the interpre-tation of paleo-isotopic records. Samples were collectedusing a funnel attached to a 500 ml bottle mounted on a1.5 m post in an open setting. The bottles were replaceddaily at 7:30 am local time. Our goal was to collect allwinter precipitation; however, some days were missedbecause bottles were not deployed properly. These caseswere limited and not linked to weather-related conditionsthat might bias the isotope data. Each sample representedthe precipitation that had fallen over the previous 24-hours.[6] Liquid samples were poured into a 60 ml bottle and

sealed tightly to prevent evaporation. Snow samples, mostof which collected within the funnel, were covered in plasticwrap, melted at room temperature, and stored in a sealedcontainer. Samples were analyzed at the University ofSaskatchewan Isotope Laboratory using a GC Pal LiquidAutosampler mounted on a Thermo Finnigan TC/EA cou-pled to a Thermo Finnigan Delta Plus XL mass spectrom-eter via a Conflo III interface. Stable isotopes are measuredrelative to VSMOW using internal standards calibrated withinternational standards. Sample precision is determined tobe ±0.39% for d18O and ±3% for dD (1s, n = 45).[7] The atmospheric circulation during each precipitation

event was evaluated using a 48-hour back trajectory ap-proach originating near Colgate University. This techniquehas been used to a limited extent by others and can provideinsight into the source and meteorological history of theprecipitating water vapor [Lee et al., 2003]. Back trajecto-ries were calculated using the HYSPLIT model provided bythe National Oceanographic and Atmospheric Administra-

tion’s Air Resource Laboratory. Each trajectory was initiatedat 500 meters above ground level using archived data fromthe ETA 80 km data assimilation system. Vertical motionswere modeled dynamically within HYSPLIT. Trajectorydata were coupled with a qualitative assessment of weatherconditions for each precipitation event using synoptic-scalesurface and middle-tropospheric weather maps. A circula-tion classification was constructed in which each precipita-tion event was assigned to a particular synoptic weatherpattern.

3. Results

[8] The circulation analysis produced four dominantprecipitation weather types, which were associated with115 (87%) of the events. The remaining 17 days representedless frequent weather types and were bundled into an‘‘other’’ category. The most common precipitating weathertype was the lake-effect pattern (42%). These events, whichaverage 0.16 cm (s = 0.20 cm) of liquid precipitation, asmeasured using hourly records from nearby Syracuse, NY,are associated with high pressure over the central andMidwestern states, lower pressure over New England andnorthwesterly surface winds over the Great Lakes(Figure 1a). The northwesterly flow associated with theseevents can be seen in the trajectory summary (Figure 2a).Aloft, lake-effect events feature deep troughing over NewEngland, which provides the cold air necessary for lake-effect instability.[9] Nearly 25% of the weather types were classified as

warm sector events and are characterized by an average of0.56 cm (s = 0.61 cm) of liquid precipitation and surfacelow pressure north of the study region. Conditions aloft

Figure 1. Surface circulation composites associated with:(a) lake-effect, (b) warm sector, (c) Atlantic coastal, and(d) warm frontal overrunning precipitation.

Figure 2. 48-hour back trajectories associated with:(a) lake-effect, (b) warm sector, (c) Atlantic coastal, and(d) warm frontal overrunning precipitation.

L22209 BURNETT ET AL.: PRECIPITATION OF d18O AND CIRCULATION L22209

2 of 4

feature southwest flow along a ridged pattern over theAtlantic Coast. This pattern places the study region withinan open warm sector dominated by southerly wind flow(Figures 1b and 2b). Precipitation is driven by cold frontallifting operating on the relatively humid warm sector airoriginating from the Gulf of Mexico and nearby Atlantic.[10] Atlantic coastal systems, which produce the largest

average liquid precipitation amounts (ave = 0.86 cm, s =0.56 cm), comprised 15% of the events and feature a deepsurface low moving northward along the coast (Figure 1c).These events are largely associated with easterly trajectories(Figure 2c), although many trajectories begin over the GreatLakes before entering central New York from the east.Aloft, coastal systems feature deep troughing over theMid-Atlantic and riding over the west. The least commonweather type (6%) was the warm frontal overrunningcondition. These events are driven by troughing over theMidwest, which carries surface low pressure south of thestudy region. These events yield an average liquid precip-itation of 0.80 cm (s = 0.50 cm) and are associated withsoutherly surface flow and broad lifting over a warm frontalboundary south of the region (Figure 1d). Although some ofthe warm frontal overrunning trajectories originate from thewest, most eventually flow into central New York from thesouth (Figure 2d).[11] A scatter plot showing the d18O and dD values for

each sample, coded by weather type, is shown in Figure 3.Also shown in Figure 3 is the local meteoric water line(LMWL), which was calculated as the best fit regressionline through the sample data, and the global meteoric water(GMWL) line as reported by Rozanski et al. [1993]. Acomparison of the water lines shows that the LMWL is mostdifferent from the GMWL in the most negative sector, butconverges as the sample values increase.[12] The d18O of lake-effect precipitation is most nega-

tive, with an average of �17.9% (s = 4.3%). Coastal andwarm frontal overrunning events had slightly higher values,with average d18O of �16.1% (s = 3.0%) and �16.6% (s= 3.9%) respectively. Highest d18O values are associatedwith the warm sector events (ave = �8.2%, s = 3.1%).

Mean d18O for the ‘‘other’’ weather type category, whichcaptures a wide range of less frequent types, falls betweenthe warm sector and coastal events (ave = �12.5%, s =4.3%). A statistical comparison of the mean d18O valuesamong the weather types was performed using a t-test.Among the types associated with low d18O precipitation,lake-effect values are significantly more negative thancoastal precipitation (P-value = 0.043). By contrast, themean d18O of the warm frontal overrunning weather type isnot statistically different from the lake-effect or coastal d18Ovalues (P-values = 0.409 and 0.765 respectively). This lackof statistical difference may be influenced by the smallnumber of warm frontal overrunning events in the dataset.The mean d18O for warm sector events is significantlydifferent from all other weather types, with P-values of0.001 or less.[13] An examination of the weather types associated with

each precipitation day showed that 14 samples representedsingle storm systems lasting multiple days. Using a pairedt-test, we assessed the possibility that system longevity mightalter isotope values. Results showed that the precipitationd18O did not differ significantly from the previous day’svalue in these 14 cases (P-value = 0.671). We also assessedthe degree to which each weather type occurred equallythroughout all months of the winter season. Events thatoccur more frequently within different parts of the winterseason may carry a temperature related d18O bias. Using aChi-squared test, we compared the observed monthly dis-tribution of events to an equal distribution. In all cases, theobserved monthly distribution did not differ significantlyfrom the equal distribution.[14] A comparison of d18O values and the mean surface

temperature recorded at Syracuse, NY yielded a statisticallysignificant positive relationship. However, an r2 value of0.25 indicates that 75% of the precipitation d18O variabilitycannot be explained by surface air temperature alone. Usingthe derived temperature-d18O equation, (d18O = 0.285T �23.46), the average residual d18O value was calculated foreach circulation type. The most extreme average residualvalues were associated with the lake-effect (�3.3%) andwarm sector (+3.3%) events, indicating that these eventsyield precipitation that is more negative and positiverespectively than would be expected through surface tem-perature alone. The coastal and warm frontal overrunningevents exhibited slightly more negative values (�1.5% and�0.7%) than predicted by temperature.

4. Discussion

[15] This study provides a first step toward an understand-ing of how different winter circulation types in the north-eastern United States influence precipitation d18O andunderscores the importance of circulation when interpretingd18O archives. Lake-effect events are especially interesting inthat they represent the impact of water recycling on precip-itation d18O. The northwesterly trajectories that are associatedwith lake-effect precipitation bring cold, dry air across therelatively warm lake surfaces [Burnett et al., 2003]. The d18Oof the Great Lakes surface water, as reported by Gat et al.[1994], is in the �6.8% to �8.8% range. Vertical fluxesof heat and recycled moisture destabilizes the overlyingatmosphere, thus producing the lake-effect precipitation.

Figure 3. Scatterplot showing d18O and dD values for eachprecipitation event coded by circulation type. The localmeteoric water line is based on a linear regression throughthe observations, whereas the equation for the globalmeteoric water line comes from Rozanski et al. [1993].

L22209 BURNETT ET AL.: PRECIPITATION OF d18O AND CIRCULATION L22209

3 of 4

The degree to which lake-effect precipitation records GreatLake moisture recycling was examined in isotope terms byGat et al. [1994] using the deuterium excess (d = dD � 8 �d18O). For the data presented in this study, deuterium excesscan be visually assessed through a comparison of theLMWL and GMWL in Figure 3. The larger differencesbetween the LMWL and the GMWL for the lake-effectsamples reflect larger deuterium excess and recycled mois-ture. By contrast, the warm sector events, with their strongGulf of Mexico and Atlantic moisture inflow, yield anenriched precipitation with little continental rainout andmoisture recycling. These conclusions are consistent withfindings reported by Gedzelman and Lawrence [1982] forwarm sector precipitation.[16] Coastal and warm frontal overrunning systems ex-

hibit relatively low d18O values that we associate withbroad-scale frontal lifting and an associated altitude effect.This condition is especially pronounced in the warm frontaloverrunning events where warm, humid air is transportednorthward and gradually lifted over a warm frontal bound-ary. As this air rises, cools, and condenses, the resultingprecipitation d18O decrease. Lawrence et al. [1982] presenta similar argument for dD variability at Mohonk Lake, insoutheastern New York State, in which they associatenegative dD value events with more southerly and easterlystorm tracks. The coastal systems, with their low d18O, mayalso reflect warm frontal overrunning along the easterlyflow on the north side of the low. Gedzelman and Lawrence[1990] preformed an isotopic analysis of two coastal stormsand found the lowest d18O from stratiform precipitationwithin the cold sectors of each storm. This low d18O valuestratiform precipitation is consistent with the easterly over-running and depleted Atlantic coastal precipitation observedin the current study. As the coastal systems move north andeast, they can create northwesterly flow and lake-effectprecipitation over the study area, adding to lower d18Ovalues. Gedzelman and Lawrence [1990] also found asignificant amount effect with coastal systems, in whichd18O values decreased as precipitation totals increased. Onaverage, the coastal events in this study exhibit the largestaverage liquid precipitation and may be influenced by anamount effect.

5. Conclusions[17] The distinctiveness of precipitation d18O derived

from different winter storm types supports the ideathat d18O records, such as those stored in lacustrine carbo-nates, can be used to reconstruct regional-scale patterns ofpaleo-circulation. However, the results of this study remainlimited. In the case of many lake systems, lacustrinecarbonate d18O values are influenced by factors other thanwinter circulation and precipitation. The degree to which aparticular lake system derives its water from winter andsummer precipitation, the lake residence time, and summerevaporation all influence the lake water d18O. Furthermore,one must understand the significance of lake water temper-ature on the d18O values, which may serve to conflate thelake water d18O-atmospheric circulation signal. Recogniz-

ing these constraints, we hope that this ongoing effort todocument circulation-isotope relations in central New YorkState for all seasons will provide an important tool forpaleoclimate interpretation throughout this region andothers.

[18] Acknowledgments. This research was supported by funding fromthe Colgate University Research Council and by NSF grant BCS-0418012.

ReferencesAnderson, W. T., M. Saurer, F. Schweingruber, S. M. Bernasconi, and J. A.McKenzie (2002), Model evaluation for reconstructing the oxygen iso-topic composition in precipitation from tree ring cellulose over the lastcentury, Chem. Geol., 182, 121–137.

Araguas-Araguas, L., K. Froehlich, and K. Rozanski (2000), Deuterium andoxygen-18 isotope composition of precipitation and atmospheric moist-ure, Hydrol. Processes, 14, 1341–1355.

Burnett, A. W., M. E. Kirby, H. T. Mullins, and W. P. Patterson (2003),Increasing Great Lake-effect snowfall during the twentieth century: Aregional response to global warming?, J. Clim., 16, 3535–3542.

Dansgaard, W. (1964), Stable isotopes in precipitation, Tellus, 16, 436–468.

Edwards, T. W. D., and B. B. Wolfe (1996), Influence of changing atmo-spheric circulation on precipitation d18O-temperature relations in Canadaduring the Holocene, Quat. Res., 46, 211–218.

Gat, J. R., C. J. Bowser, and C. Kendall (1994), The contribution of eva-poration from the Great Lakes to the continental atmosphere: Estimatebased on stable isotope data, Geophys. Res. Lett., 21, 557–560.

Gedzelman, S. D., and J. R. Lawrence (1982), The isotopic composition ofcyclonic precipitation, J. Appl. Meteorol., 21, 1385–1404.

Gedzelman, S. D., and J. R. Lawrence (1990), The isotopic composition ofprecipitation from two extratropical cyclones, Mon. Weather Rev., 118,495–509.

Kirby, M. E., H. T. Mullins, W. P. Patterson, and A. W. Burnett (2002), Lateglacial-Holocene atmospheric circulation and precipitation in the north-east United States inferred from modern calibrated stable oxygen andcarbon isotopes, Geol. Soc. Am. Bull., 114, 1326–1340.

Lawrence, R., S. D. Gedzelman, J. C. White, D. Smiley, and P. Lazov(1982), Storm trajectories in the eastern US D/H isotopic compositionof precipitation, Nature, 296, 638–640.

Lee, K. S., A. J. Grundstein, D. B. Wenner, M. S. Choi, N.-C. Woo, andD. H. Lee (2003), Climate controls on the stable isotopic composition ofprecipitation in northeast Asia, Clim. Res., 23, 137–148.

Mullins, H. T. (1998), Environmental change controls of lacustrinecarbonate, Cayuga Lake, New York, Geology, 26, 443–446.

Mullins, H. T., A. W. Burnett, M. F. Hillfinger IV, and M. E. Kirby (2003),Millennial-scale holocene lake level fluctuations at Green Lakes StatePark, New York: Implications for atmospheric circulation over the north-eastern United States, Northeast. Geol. Environ. Sci., 25, 197–205.

Noone, D., and I. Simmonds (2002), Associations between d18O of waterand climate parameters in a simulation of atmospheric circulation for1979–95, J. Clim., 15, 3150–3169.

Rozanski, K., L. Araguas-Araguas, and R. Gonfiantini (1993), Isotopicpatterns in modern global precipitation, Climate Change in IsotopicRecords, Geophys. Monogr. Ser., vol. 78, edited by P. K. Stewart et al.,pp. 1–36, AGU, Washington, D. C.

Teranes, J. L., and J. A. McKenzie (2001), Lacustrine oxygen isotoperecord of 20th-century climate change in central Europe: Evaluation ofclimatic controls on oxygen isotopes in precipitation, J. Paleolimnol., 26,131–146.

Vinther, B. M., S. J. Johnson, K. K. Anderson, H. B. Clausen, and A. W.Hansen (2003), NAO signal recorded in the stable isotopes of Greenlandice cores, Geophys. Res. Lett., 30(7), 1387, doi:10.1029/2002GL016193.

�����������������������A. W. Burnett, Department of Geography, Colgate University, Hamilton,

NY 13346, USA. (aburnett@mail.colgate.edu)H. T. Mullins, Department of Earth Sciences, Syracuse University,

Syracuse, NY, USA.W. P. Patterson, Department of Geological Sciences, University of

Saskatchewan, Saskatoon, Saskatchewan, Canada.

L22209 BURNETT ET AL.: PRECIPITATION OF d18O AND CIRCULATION L22209

4 of 4