UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Caribbean sea surface temperatures and El Niño : a new outlook Lau, Justin A. 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2016_may_lau_justin.pdf [ 12.12MB ]
JSON: 24-1.0228057.json
JSON-LD: 24-1.0228057-ld.json
RDF/XML (Pretty): 24-1.0228057-rdf.xml
RDF/JSON: 24-1.0228057-rdf.json
Turtle: 24-1.0228057-turtle.txt
N-Triples: 24-1.0228057-rdf-ntriples.txt
Original Record: 24-1.0228057-source.json
Full Text

Full Text

Caribbean Sea Surface Temperatures and El Nin˜oA New OutlookbyJustin A. LauB. Science, The University of British Columbia, 2010A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Geography)The University of British Columbia(Vancouver)March 2016c Justin A. Lau, 2016AbstractMass coral bleaching in recent years has become a recurring event and was sus-pected to have a relationship with El Nin˜o events. Changes in the understanding ofwhat constitutes an El Nin˜o event prompted further research into the relationshipwith Caribbean sea surface temperatures due to their impact on corals.Multiple statistical tests were employed to profile the relationship between theindividual event types and the Caribbean. Ultimately, a bootstrapping techniquedetermined that Central Pacific El Nin˜o events bear a relationship, while EasternPacific event types do not.An attempt to hindcast El Nin˜o events in order to comment on the history ofimpacts upon the Caribbean was unsuccessful due to a lack of sufficient input data,but a model determining potential locations of data is presented.iiPrefaceThis thesis is original, unpublished, independent work by the author, Justin A. Lau,based upon data from publicly available satellite and in-situ geochemical proxydata hosted by NOAA – meticulously analyzed for your intellectual consumption,and in partial fulfillment of the degree of Master of Science in the Department ofGeography at the University of British Columbia.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Caribbean Sea-Surface Temperature Response to Central Pacific andEastern Pacific El Nin˜o Influence During the Satellite Era . . . . . . 52.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1.1 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.2 Indices of El Nin˜o . . . . . . . . . . . . . . . . . . . . . 82.2.3 Empirical Orthogonal Functions . . . . . . . . . . . . . . 102.2.4 Cross-Correlation Analysis . . . . . . . . . . . . . . . . . 10iv2.2.5 Bootstrap Model . . . . . . . . . . . . . . . . . . . . . . 112.2.6 Degree Heating Months . . . . . . . . . . . . . . . . . . 122.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 Paleo-Oceanographic Records of El Nin˜o Influenced Sea-Surface Tem-peratures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.2.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303.2.2 NINO Hindcast Model . . . . . . . . . . . . . . . . . . . 313.2.3 Drill Prediction Model . . . . . . . . . . . . . . . . . . . 323.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.3.1 Existing Data . . . . . . . . . . . . . . . . . . . . . . . . 323.3.2 Potential Data . . . . . . . . . . . . . . . . . . . . . . . . 343.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42vList of TablesTable 1.1 CP event dates . . . . . . . . . . . . . . . . . . . . . . . . . . 3Table 2.1 El Nin˜o indices . . . . . . . . . . . . . . . . . . . . . . . . . 9Table 2.2 El Nin˜o index correlations with Pacific variability . . . . . . . 9Table 3.1 Geochemical Proxy Data . . . . . . . . . . . . . . . . . . . . 31viList of FiguresFigure 1.1 Composite of seasonal SSTAS from the summer (top) beforethe event until the spring (bottom) after the event for the EPENSO (left) and CP ENSO (right). SSTA are in C (from Ban-holzer, 2012). . . . . . . . . . . . . . . . . . . . . . . . . . . 2Figure 2.1 The distribution of major stony coral reefs (source: NOAA) . . 6Figure 2.2 Sub-regions of the Caribbean as defined for this study . . . . . 11Figure 2.3 Time series of Eastern (NINO3) and Central (NINO4) Pacifictemperatures. Linear temperature trends are indicated. TheEastern region has cooled by 0.55C, and the Central regionhas cooled by 0.18C . . . . . . . . . . . . . . . . . . . . . . 14Figure 2.4 Time series of East and West Caribbean temperatures. Lin-ear temperature trends are indicated. The Eastern region haswarmed by 0.87C, and the Western region has warmed by0.40C over the study period. . . . . . . . . . . . . . . . . . 15Figure 2.5 First empirical orthogonal function of theWest and East Caribbeanregions (1982–2010) . . . . . . . . . . . . . . . . . . . . . . 16Figure 2.6 Pareto plot of the Eastern region identifying the amount ofvariability explained by each EOF mode (bars), and the sumof total explained variability for modes 1:x (line). . . . . . . . 17Figure 2.7 The lead/lag correlation plot of the East Caribbean’s leadingprincipal component against the NINO3 and NINO4 indices.The NINO3 index peaks at 7 months (July), but the NINO4peaks at 8 months (August). . . . . . . . . . . . . . . . . . . 17viiFigure 2.8 Time series of NINO3 & the lagged first principal componentsof the Caribbean regions. Both regions of the Caribbean lacksignificant correlation based upon the adjusted sample size. . . 18Figure 2.9 Time series of NINO4 & the lagged first principal componentsof the Caribbean regions. The Eastern Caribbean was corre-lated at r=0.45, but once again the Western Caribbean was in-significant. . . . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 2.10 Average probability of statistically significant differences dur-ing EP ENSO events via bootstrapping using the t-test method(a = 0.10). The entire Caribbean is not significantly differentfrom background conditions during ASO following the peak ofan EP ENSO event. . . . . . . . . . . . . . . . . . . . . . . . 21Figure 2.11 Average probability of statistically significant differences dur-ing CP ENSO events via bootstrapping using the t-test method(a =0.10). Much of the Caribbean has a high probability (>50%)of being significantly distinguishable from background ASOconditions during CP ENSO events. . . . . . . . . . . . . . . . 22Figure 2.12 EP year DHM results. The spatial distribution and intensityduring EP years lack a discernible pattern from backgroundvariability. Only 1983 does not exhibit a region of heat stress> 2C·months, but a coherent pattern of intensity in the stressdistribution is lacking. . . . . . . . . . . . . . . . . . . . . . 23Figure 2.13 CP year DHM results. During CP events, the Caribbean ex-periences consistent instances of amplified heat stress. 2005& 2010 have both the largest values of accumulation and dis-tribution, but all four years identify regions of high stress (>2C·months) . . . . . . . . . . . . . . . . . . . . . . . . . . 24Figure 2.14 CP DHM bootstrap results showing a similarity between DHMand SSTA bootstrap probabilities and distribution. Residualsbetween the two analyses does not reveal a bias, but rather, adirect relationship. . . . . . . . . . . . . . . . . . . . . . . . 25viiiFigure 3.1 Boxplots of Sr/Ca ratios reveal that none of the sites have eitherEP or CP events discernible from both other types. . . . . . . . 33Figure 3.2 Correlations between Sr/Ca core raw ratios or temperature re-constructions (blue) and ERSST temperature data (green). . . . 35Figure 3.3 Map of tropical Pacific plus Eastern Indian Ocean and CaribbeanSea with green grid cells indicating regions where reefs existand are correlated with the NINO3, but not the NINO4 region . 36Figure 3.4 Map of tropical Pacific plus Eastern Indian Ocean and CaribbeanSea with green grid cells indicating regions where reefs existand are correlated with the NINO4, but not the NINO3 region . 37ixGlossaryASO August-September-OctoberAVHRR Advanced Very High Resolution RadiometerCCA Cross-Correlation AnalysisCMIP5 Coupled Model Intercomparison Project Phase 5CP Central PacificDHM Degree Heating MonthDJF December-January-FebruaryENSO El Nin˜o-Southern Oscillation (used interchangeably with El Nin˜o)EOF Empirical Orthogonal FunctionERSST Extended Reconstruction Sea Surface TemperatureEP Eastern PacificITCZ Intertropical Convergence ZoneMMM Maximum Monthly MeanNINO3 A sea-surface temperature index used to identify El Nin˜os. Defined by theregion bounded by 5N–5S, 150W– 90WNINO4 A sea-surface temperature index used to identify El Nin˜os. Defined by theregion bounded by 5N–5S, 160E– 150WxNOAA National Oceanographic and Atmospheric Administration of the UnitedStates of AmericaOI-SST Optimum Interpolation Sea-Surface Temperature. A dataset from the Na-tional Oceanographic and Atmospheric Administration of the United Statesof AmericaPC1 Primary Principal ComponentRCP4.5 Intergovernmental Panel on Climate Change Representative Concentra-tion Pathway, 4.5 W/m2 scenarioSPCZ Southern Pacific Convergence ZoneSST Sea-Surface TemperatureSSTA Sea-Surface Temperature AnomalyxiAcknowledgmentsMy classwork and location of my office were at the UBC Vancouver campus –traditional unceded territories of the Musqueam people.I’d like to thank Douw Steyn, Tara Ivanochko, as well as members of the Tun-dra Ecology lab 2007–2010 for inspiring me to dive further the world of research.Thank you to NOAA for the useful databases that made my research possible.Laura Albert and Allina Tran were my closest confidants who never ceased tobelieve that this was possible even in the most difficult of times.Sandra Banholzer and the rest of the Climate and Coastal Ecosystems Lab(CCEL) team, Ian McKendry, and ESPECIALLY my supervisor Simon Donnerwho exhibited amazing patience and support amidst very difficult extenuating cir-cumstances.xiiChapter 1IntroductionClimate varies everywhere on Earth at different temporal and spatial scales. Thesevariations can have wide-ranging effects that make them both difficult and interest-ing to study. The El Nin˜o-Southern Oscillation (ENSO) is one such example of acoupled ocean-atmosphere system that affects much of the globe in many differentways. Recently, new insights into the identification and characteristics of ENSOhave arisen, which are altering the perception of potential impacts of ENSO events(Ashok et al., 2007; Yeh et al., 2009; Yu et al., 2012).Many of the effects of ENSO events exist within the tropics – a region densewith coral reefs, which are sensitive to changes in their environment. Occurrencesof mass coral bleaching in recent times have appeared to follow certain types ofENSO events, leading to questions regarding the links between them.First described by Rasmusson and Carpenter (1982), ENSO is the primary sourceof short-term climatic variability in the Pacific Ocean (McPhaden, 2004)– explain-ing 45% of Sea-Surface Temperature Anomaly (SSTA) from 1979–2004 (Ashoket al., 2007). The coupled system is composed of an atmospheric portion calledthe Southern Oscillation and the oceanic portion called El Nin˜o, but there is veryhigh correlation between the components, thus, an analysis of either component isindicative of the whole system.Both El Nin˜os and La Nin˜as (negative, or cold phase of ENSO) are phase lockedwith the seasonal cycle, with peaks occurring in the boreal winter (Wang andFiedler, 2006). Warm phases generally occur every 3-7 years and are very fre-1Figure 1.1: Composite of seasonal SSTAS from the summer (top) before theevent until the spring (bottom) after the event for the EP ENSO (left) andCP ENSO (right). SSTA are in C (from Banholzer, 2012).quently followed by La Nin˜as (Mock, 2007). The predominant interval between ElNin˜os is 4 years (Cane, 2005), but it should also be noted that events have occurredin back-to-back years as well as not arisen for more than a decade at a time.Since the early 2000’s, scientists have begun to take notice of ENSO-like eventsthat do not fit in with the traditional model identified by Rasmusson and Carpenter(1982). These discoveries were often associated with the inability of the NINO3.4index to capture what appeared to be an El Nin˜o signal in SSTAS. This new ENSOhas had various names associated with it: El Nin˜o Modoki (Ashok et al., 2007),Dateline El Nin˜o (Trenberth and Stepaniak, 2001), and Central Pacific (CP) El Nin˜o(Kao and Yu, 2009; Lee and McPhaden, 2010; Yeh et al., 2009). The classical ElNin˜o is now referred to as canonical or Eastern Pacific (EP) El Nin˜o (Figure 1.1).CP El Nin˜os exhibit a tripole pattern where a warm pool resides in the cen-tre of the Pacific, close to the International Date Line, flanked by cold SSTAS oneither side along the equator (Ashok et al., 2007; Kao and Yu, 2009; Yeh et al.,2009). The analysis by Ashok et al. (2007) established that since the last climateregime shift in 1978, CP El Nin˜o’s have been responsible for 12% of Sea-Surface2Temperature (SST) variability in the Pacific Ocean. Ashok et al. (2007), as wellas Trenberth and Stepaniak (2001), used their analysis to establish a new index fordetermining the CP El Nin˜os; however, most other studies used the NINO4 index.Ashok et al. (2007) warn against this practice as the NINO4 index is correlatedwith both El Nin˜o and El Nin˜o Modoki. This difference may account for why Yehet al. (2009) and Ashok et al. (2007) have slight differences in the years that theylabel CP/Modoki events (Table 1.1).Table 1.1: CP/Modoki events identified by 2 different studies (1985-2005)Years 1986 1990 1991 1992 1994 2001 2002 2004Ashok et al., 2007 x x x x x x xYeh et al., 2009 raw x x x x x xYeh et al., 2009 detrended x x x xYeh et al. (2009) discovered that the frequency of ENSO event types experi-enced a shift between 1854-2007. Comparing pre- and post-1990, EP El Nin˜oshave occurred 0.19 times per year versus 0.29 times per year respectively, whereasCP El Nin˜os have occurred 0.01 times per year compared to 0.29 times per year.Another analysis in the time domain found that CP ENSO events tend to have alifespan of only 8 months compared to EP ENSOS, which are 15 months on average(Kao and Yu, 2009).Due to the differences in the CP and EP El Nin˜os, the associated teleconnectionsare different (Ashok et al., 2007; Yeh et al., 2009), and need to be explored.To extend ENSO records beyond the instrumental record, there are two differentmethods: dendrochronology and sclerochronology (Mock, 2007). Sclerochronol-ogy analyzes ratios of geochemical tracers within the corals’ aragonite skeletonsas a proxy for SST. d 18O ratios have been used to establish several hundred yearsof data (Cobb et al., 2003; Mock, 2007). Positive SST are associated with negatived 18O values; however, a confounding issue with d 18O is that it is also dependent onsalinity, which can lead to complications if the salinity of the location at the samepoint in time is unknown (Pfeiffer, 2009). A proxy that only relies on temperatureis the Sr/Ca ratio in the coral aragonite, which varies predictably with temperature3due to the long residence time of Sr and Ca in the ocean (Pfeiffer, 2009). The onlyreason that more Sr/Ca data do not exist has been the associated cost, so it is fore-seeable that the scientific community will continue to rely upon d 18O primarily.ENSO cycles are known to have significant climate teleconnections. They gen-erate a response by global circulation pattern alterations that result in altered tem-perature and precipitation around many parts of the world (Cane, 2005; Mock,2007). These changes generally follow the same patterns, but occasionally theteleconnections may propagate slightly differently as ENSO events are all unique insome way (Mock, 2007).Climate often dictates the local biota, and when that climate is disturbed, thebiota are forced to acclimatize or evolve; thus, many organisms in ENSO regionshave adapted to the highly variable ENSO regime (Wang and Fiedler, 2006). Wangand Fiedler (2006) reviewed many of the biotic responses to El Nin˜o. Their re-sults suggest that the primary biotic response begins at the lowest trophic levelsof marine life due to the change of the oceanic nutricline. All the other impactsare connected via the trophic chain leading from this up to marine mammals andseabirds (Wang and Fiedler, 2006).This study aims to draw links between climatological and oceanographic vari-ability of the tropical Pacific with that of Caribbean SST. The findings will help fur-ther our knowledge about teleconnections of the recently discovered CP El Nin˜os.Additionally, coral biologists and marine management specialists will be better in-formed about what the different types of El Nin˜o will mean for the Caribbean andintegrate it into their work.1.1 Questions1. What is the correlation between different El Nin˜o types and their strength inthe Caribbean?2. What is the lag time between the peaks of the different types of El Nin˜o andtheir associated peak in the Caribbean?3. Can hindcasts be developed that differentiate between ENSO type?4Chapter 2Caribbean Sea-SurfaceTemperature Response to CentralPacific and Eastern Pacific ElNin˜o Influence During theSatellite Era2.1 IntroductionCorals reefs (Figure 2.1) are very important systems that consist of carbonaceousstructures that provide a diverse range of services. Approximately 400-2000 impe-rial tons/ha/yr of CaCO3 is laid down and added to reefs by living coral each year(Chave et al., 1972). Even though reefs only occupy a small portion of the oceanfloor, they provide habitat for ˜33% of marine fish, and account for ˜10% of fishconsumption by humans (Lough and van Oppen, 2009). Individual corals have asymbiotic relationship between host cells and algal organisms called zooxanthel-lae. Approximately 17% of the world’s coastlines are made from tropical coralreefs (Birkeland, 1997). These corals live in this region because the optimal tem-perature range is 26-28C, but the average maximum temperature is 29.5C, thus5Figure 2.1: The distribution of major stony coral reefs (source: NOAA)they are very sensitive to change (Hubbard, 1997; Sheppard et al., 2009b). If thisstress is sustained at greater than 3C above seasonal maximum for more than twodays, or greater than 1C for several weeks, the coral will begin to ‘bleach’ (Brown,1997). The term bleach comes from the fact that the coral appears very white.The change in colour occurs when the colourful zooxanthellae leave the host cells,meaning the carbonaceous skeleton beneath the translucent cells becomes visible(Lough and van Oppen, 2009; Muller-Parker and D’Elia, 1997; Sheppard et al.,2009b). In extreme cases, the magnitude of temperature change as well as persis-tence can lead to the death of the coral as it has insufficient nutrients to survive(Muller-Parker and D’Elia, 1997).Widespread bleaching was first observed during the 1982-1983 ENSO event,and has since occurred multiple times, being labeled “profound”, “unprecedented”,and a “critical threat” linked to global warming (Lough and van Oppen, 2009; Mc-Carthy et al., 2007; Oliver et al., 2009; Sheppard et al., 2009a; Wood, 2007). Thebleaching events of 2005 and 2010 were so widespread that they were considered“global bleaching events” by experts (Eakin et al., 2010).Very high levels of bleaching, disease, and mortality responses were recordedacross the Caribbean in 2005 (Eakin et al., 2010). During 2010, Tobago and6Venezuela were two examples of areas in the Caribbean Sea that experiencedbleaching and mortality worse than in 2005 amidst widespread bleaching through-out the region (Alemu and Clement, 2014; Bastidas et al., 2012) – a region whereDonner et al. (2007) placed the probability of the 2005 event at 1-in-500.Heat stress and coral bleaching events in some parts of the tropics are correlatedwith ENSO events, and have occurred at the same time as all known mass bleachingevents: 1982/1983, 1986/1987, 1997/1998, 2004/2005, 2009/2010 (Brown, 1997;Oliver et al., 2009; Veron, 1995; Wood, 2007). The ENSO events do not directlycause the bleaching; however, they increase the likelihood of positive SSTAS thatinduce the bleaching response (Eakin, 2009).In the Caribbean, these bleaching events happen after the peak of the ENSOevent. Known as the tropical atmospheric bridge, anomalous cloud cover and evap-oration patterns that alter the heat flux into the surface of the ocean due to El Nin˜ois currently the mechanism used to explain the teleconnections between PacificENSO events, and its effects around the global tropics, including the Caribbean(Klein et al., 1999). The current literature on these bleaching events predates thedifferentiation between EP and CP ENSO events, thus the associated lags and im-pacts need to be reassessed. However, both the 2005 and 2010 bleaching events inthe Caribbean followed a large CP ENSO event, suggesting that there may be a linkbetween bleaching and CP events.2.1.1 ObjectivesTo establish the links between different ENSO types and their Caribbean SST tele-connection(s).1. Are there relationships between particular El Nin˜o types and different re-gions (East versus West) of the Caribbean Sea?(a) What are their associated SST impacts?(b) What is the lag time between the peaks of the different types of El Nin˜oand their associated peak in the Caribbean?(c) Is heat stress related in the Caribbean associated with these SST im-pacts?72.2 Methods2.2.1 DataData from the National Oceanographic and Atmospheric Administration (NOAA)Optimum Interpolation Sea-Surface Temperature (OI-SST) v.2 dataset was utilized(Reynolds et al., 2007) for a 29 year period (January 1982 – December 2010),and uses a climatological baseline of 1986–2005. 1982 was the first full year ofhigh-resolution satellite SST data available, and for this reason was chosen as thestart date. The dataset consists of 14 x 14 latitude-longitude resolution gridded,daily data prepared by the National Centers for Environmental Prediction and theNational Climatic Data Center, and has been averaged into monthly units for thisstudy. The dataset is based on observations from the Advanced Very High Reso-lution Radiometer (AVHRR) instruments aboard the Pathfinder satellites and in situdata (Reynolds et al., 2007). Analysis of SST in the Caribbean focused on August-September-October (ASO), which are the summer months with the highest SST.2.2.2 Indices of El Nin˜oEl Nin˜o indices are used to determine the state of the climate, and whether a par-ticular event type exists at a given point in time (Table 2.1). It has been argued thatdue to its geographical limitations, the NINO4 index alone is insufficient for iden-tifying CP events (Li et al., 2010). These limitations lead to the mis-categorizationof events, and a lack of mathematical orthogonality to the EP indices (Table 2.2).Li et al. (2010) introduces the IEMI index, which is nearly completely orthogonal,but has a tendency to stop identifying CP events partway through boreal winter dueto warming in the Eastern region, before being classified as an event again laterin the year. This flaw occasionally forces events to not meet the minimum thresh-old of 5 consecutive months at >1 standard deviation above baseline (Banholzer,2012). As a compromise, all time-series analysis is conducted using the NINO4index. However, when identifying individual CP years for analysis, only the yearswith agreement between multiple studies with different methods were selected (Yuet al., 2012).The NINO3 and NINO4 indices were computed using SSTA data with a climato-8Table 2.1: El Nin˜o indices (adapted from Ashok et al. (2007))Index Name DefinitionNINO3 the region is bounded by (5N–5S, 150W–90W); the area-averaged SSTA over this regionis known as NINO3 index, which is a well-knownENSO indexNINO3.4 the region is bounded by (5N–5S, 170W–120W); the area-averaged SSTA over this regionis known as NINO3.4 indexNINO1+2 the region is bounded by (equator to 10S, 90W–80W); the area-averaged SSTA over this regionis known as NINO1+2 indexNINO4 the region is bounded by (5N–5S, 160E–150W); the area-averaged SSTA over this regionis known as NINO4 indexEMI [SSTA]A-0.5*[SSTA]B-0.5*[SSTA]C where thebrackets are geographically averaged SSTAs:A(165E-140W, 10S-10N), B(110W-70W,15S-5N), C(125E-145E, 10S-20N)IEMI IEMI = 3*[SSTA]A-2*[SSTA]B-[SSTA]C wherethe brackets are geographically averaged SSTAsover the same regions as the EMITable 2.2: Correlation values between the indices and the first two PC’s dur-ing 1979-2008. Maximum and minimum correlations in each column arebold. (from Li et al. (2010))Index PC1 PC2NINO3 0.97 -0.12NINO3.4 0.95 0.17NINO1+2 0.84 -0.47NINO4 0.79 0.50IEMI -0.02 0.94EMI 0.24 0.91TNI 0.04 0.909logical baseline of 1986–2005. The NINO indices are defined as the area averagedSSTA values within their respective regions as outlined in Table 2.1.Classification criteria for different ENSO events vary. Event years were identi-fied by comparing results from papers studying CP & EP ENSO via different meth-ods (Ashok et al., 2007; Banholzer, 2012; Lee andMcPhaden, 2010; Li et al., 2010;Wang and Fiedler, 2006; Yeh et al., 2011), and the four most prominent events ofeach kind that were agreed up between studies, were selected. As many of thestudies were completed prior to the end of the study period, the methods werereplicated by this study to extend the results for a more representative comparisonin the latter years. The years in this study agree with analysis conducted by Yuet al. (2012), who put together a “consensus” list of CP and EP events.2.2.3 Empirical Orthogonal FunctionsIn order to determine the primary modes of SST variability, and the effect of spa-tial scale on the variability signal, Empirical Orthogonal Function (EOF) analysiswas conducted on the SSTA data within two different regions in the Caribbean:‘west’ and ‘east’ Figure 2.2. As a data compression technique, this analysis math-ematically defines a new set of theoretical eigenvectors designed to maximallycapture variability in a single dimension, orthogonal to subsequent eigenvectors(Preisendorfer and Mobley, 1988). In this case, the technique identifies the mathe-matical function of SST along a theoretical dimension within both space and time.2.2.4 Cross-Correlation AnalysisCross-Correlation Analysis (CCA) was employed in order to determine the lengthof time between the peak of the ENSO signal, and the peak SST response in theCaribbean Ocean. The Primary Principal Component (PC1) for each region andeach index time series were ran through the CCA with a maximum lag of 12 monthsbetween event the event peak as determined by the NINO indices, and the peak ofthe regional PC1 in question. Normalized correlations were computed for eachtime step in the range, leading to 25 results per test with the assumption that if anidentifiable effect is to be felt from an ENSO event, it would occur within the next12 months.10WestRegionEast RegionFigure 2.2: Sub-regions of the Caribbean as defined for this study. The‘West’ region is 10–25N, 75–90W and the ‘East’ region is 10–20N,50–75W2.2.5 Bootstrap ModelIn traditional statistical tests, low population members and/or degrees of freedomare less likely to lead to statistical significance being established, thus potentiallyleading to a Type II statistical error. In this case, n= 4 for both the EP & CP events.In order to counteract this, a bootstrap technique (Efron and Tibshirani, 1991) wasadapted from Kim et al. (2009). For both EP and CP events, four ASO compositesare chosen at random from the 29 year (1982–2010) time series of ASO SST datato generate a time-series for which a t-test is conducted between it and the four EPor CP events. The result was a lat-long matrix of statistical significance (0 or 1)at the a = 0.10 level. This process was repeated 10000 times per event type, thenthe average of the result matrices was taken, resulting in a single lat-long matrixoutlining the probability of statistical significance per event type.112.2.6 Degree Heating MonthsDegree Heating Month (DHM) measurements are a method of determining heatstress accumulation in the surface ocean. The method of computing DHM em-ployed in this study was developed by Donner et al. (2005), and is based on theNOAA Coral Reef Watch degree-heating week real-time prediction method forcoral bleaching and stress. A DHM value of >2C·months suggests thermal stresssufficient enough to cause severe mass coral bleaching, and possibly coral mortal-ity (Donner, 2009). The DHM values only accumulate when the SST exceeds theMaximum Monthly Mean (MMM), which is calculated by identifying the warmestmonthly SST within the 1985–2000 monthly climatology. The DHM value for anymonth is then given by Equation 2.1 (Donner et al., 2005).DHMmonth =3Âm=0(SSTmMMM)> 0 (2.1)In the Caribbean, DHM values accumulate during summer, so the maximumannual DHM is calculated by selecting the greatest single monthly DHM value ina given year, resulting in a 29 value time-series.The bootstrap analysis was re-conducted using this DHM maximum time-seriesdata instead of SST data for a comparison with the SST bootstrap results.2.3 ResultsThis section summarizes the results of the various analyses conducted using theReynolds et al. (2007) SST data. As the analysis was conducted in an exploratoryfashion, it is organized in a manner that lays out a path of results that were nec-essary for the next piece of analysis to take place. Thus, a discussion of generaltrends in the Caribbean and Pacific Oceans precedes EOF results for the Caribbean,then lag derivation between the EOF results and Pacific indices allowing for time-series correlations that necessitate bootstrap analysis and DHM work to come todefinitive conclusions.To consider the background trends, the entire SSTA time-series was considered.In the Pacific Ocean, the East and Central regions have changed temperatures inopposition (Figure 2.3). The EP region, as defined by the NINO3, has cooled by120.55C during the study period. However, the Central Pacific, within the NINO4region, has experienced a cooling of only 0.18C.In the Caribbean, the SST regime follows a Northern Hemisphere seasonalcycle with temperatures fluctuating by > 2C annually. In both of the definedCaribbean sub-regions (Figure 2.2) temperatures have been increasing. In the East,SST has increased by 0.87C, but in the West, the increase has only been 0.40Cduring the study period (Figure 2.4).In the ‘West’ region, PC1 accounts for 46.1% of the variability in SST. Thevariability has an even distribution throughout most of the Caribbean Sea, but isslightly less active in the northwest part of the region, which reaches into the Gulfof Mexico (Figure 2.5).In the ‘East’ region, the primary centre of variability exists slightly to the Eastof the Lesser Antilles, but also small intense regions of variability along the Northcoast of South America, close to the mouths of the Magdalena and Orinoco Rivers.In this area, PC1 accounts for 76.3% of the variability(Figure 2.6).Cross-correlation could not render a statistically significant result, but showedthat the difference between the peak of an EP ENSO event and its response in the‘East’ region of the Caribbean Sea is seven months. The same analysis conductedfor the CP signal reveals a peak of 8 months (Figure 2.7). In the ‘West’ region, theredoes not appear to be a detectable response to either the EP or CP ENSO signals,evident by a lack of a peak in correlation within the 12 month window.Using a Durbin-Watson test, it was found that the PC1 time-series exhibits highfirst-order autocorrelation (DW=0.19). To account for this, all correlation testswere tested for significance using an adjusted sample size as outlined by Santeret al. (2000). In the East, the adjusted sample size was 16 (ne = 16.54).Linearly detrended SST data for EOF analysis produced results similar to thenon-detrended results being presented, and are therefore not presented.The correlations between the NINO3 region and the lagged Eastern CaribbeanPC1 was not significant at the a = 0.1 level (r = 0.412, p=0.113) (Figure 2.9),suggesting there may not be a strong link between EP events and the EasternCaribbean. The correlation between the NINO4 region and the lagged EasternCaribbean PC1 was significant at the a = 0.1 level (r = 0.457, p=0.0751) (Fig-ure 2.9), suggesting linkage between CP events and this region of the Caribbean.131982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 20112324252627282930Year°C  East Pacific (NINO3) Central Pacific (NINO4)Figure 2.3: Time series of Eastern (NINO3) and Central (NINO4) Pacific temperatures. Linear temperature trends areindicated. The Eastern region has cooled by 0.55C, and the Central region has cooled by 0.18Cover the study period.141982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 20112525.52626.52727.52828.52929.530Year°C  East Caribbean West Caribbean East Trend West TrendFigure 2.4: Time series of East andWest Caribbean temperatures. Linear temperature trends are indicated. The Easternregion has warmed by 0.87C, and the Western region has warmed by 0.40C over the study period.15Figure 2.5: First empirical orthogonal function of the West and EastCaribbean regions (1982–2010). Analysis was conducted on the indi-vidual regions (East and West), but is being presented simultaneously.The Eastern Caribbean region has more intense levels of variability, andless of a uniform distribution than the West.For both the NINO3 & NINO4, the Western Caribbean did not have significantcorrelations.The bootstrap analysis indicates that in the Caribbean, the probability thatSSTA variability is statistically attributable to EP ENSO events are extremely low(Figure 2.10). The highest probability within either of the defined East or WestCaribbean regions does not exceed 25%. In the entire analyzed region, the maxi-mum probability is under 60%, and is centred at 28N, 68W.In contrast, CP ENSO events have very high likelihoods of affecting SST in theCaribbean (Figure 2.11). The maximum probability lies eastward of the regions ofconcern with a probability of >85%. However, the probabilities within the regionsremains high, with a majority in excess of 50%.When the bootstrap model was run with scalar area averages of the two regions,resultant likelihoods were extremely low or nil for both the EP and CP events, ne-16  1   2   3   4   5   6   7   8   9  1001020304050607080901000%10%20%30%40%50%60%70%80%90%100%Figure 2.6: Pareto plot of the Eastern region identifying the amount of vari-ability explained by each EOF mode (bars), and the sum of total ex-plained variability for modes 1:x (line).−12 −9 −6 −3 0 3 6 9 12−0.08−0.06−0.04−0.0200. (months)Correlation  NINO3NINO4Figure 2.7: The lead/lag correlation plot of the East Caribbean’s leading prin-cipal component against the NINO3 and NINO4 indices. The NINO3 in-dex peaks at 7 months (July), but the NINO4 peaks at 8 months (August).171981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011−1−0.500.51  1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011−1−0.500.51  NINO3 IndexW Caribbean PC1NINO3 IndexE Caribbean PC1Figure 2.8: Time series of NINO3 & the lagged first principal components of the Caribbean regions. Both regions ofthe Caribbean lack significant correlation based upon the adjusted sample size.181980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012−1−0.500.51  1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012−1−0.500.51  NINO4 IndexW Caribbean PC1NINO4 IndexE Caribbean PC1Figure 2.9: Time series of NINO4 & the lagged first principal components of the Caribbean regions. The EasternCaribbean was correlated at r=0.45, but once again the Western Caribbean was insignificant.19cessitating the use of high-resolution analysis. If detrended SSTA data is utilizedwithin the bootstrap model, results are similar.A time-series of annual maximum DHM values show regions within the Caribbeanthat have accumulated heat stress. For instance, 1998 accumulated>3.5C·monthsthan 1983 in certain areas, as shown in the image of the four strong EP years’ heatstress, Figure 2.12. Between the four EP years, there does not seem to be a consis-tent pattern of intensity or distribution. Similarities only exist between the intensitylevels experienced in 1998 and 2007 events; however, this is not the case with thefour CP years. In all four of the CP years, large regions of high stress accumulation>3.5C·months, as well as spatial pattern similarities (Figure 2.13). These resultscorroborate the results produced by Eakin et al. (2010).The re-application of the bootstrap model to the maximum DHM data identi-fies that the DHM results are very likely to be both significant and insignificantwith comparable magnitudes and distributions to that of the original SSTA boot-strap results. The results are not identical down to grid-cell resolution, but exhibitsimilarities that are visually indistinguishable, and whose residuals appear to bestochastic. As in the case of the original bootstrap analysis, the EP years haveextremely low probabilities of being significant within the Caribbean, and the CPprobabilities are generally high, and widespread Figure DiscussionMultiple statistical methods were employed to determine the characteristics ofCaribbean SST teleconnections from Pacific ENSO events. EOF analysis identifiedsub-regions of variability, cross-correlation analysis identified the lag period fromevent peaks in the Pacific to the regional modes of variability in the Caribbean, cor-relation with an adjusted sample size for the overall time series identified what wasregionally significant while accounting for autocorrelation, and bootstrap analysisidentified whether the events are significant on a much finer resolution and targetedtime frame.Regionally, the East Caribbean is not significantly related to EP ENSO events;however, CP ENSO is (a = 0.1). This differentiation is based on comparing theentire time-series, and not only important subsets of the data. If the mechanism of20EP Years  100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30N00. 2.10: Average probability of statistically significant differences during EP ENSO events via bootstrapping usingthe t-test method (a = 0.10). The entire Caribbean is not significantly different from background conditionsduring ASO following the peak of an EP ENSO event.21CP Years  100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30N00. 2.11: Average probability of statistically significant differences during CP ENSO events via bootstrapping usingthe t-test method(a = 0.10). Much of the Caribbean has a high probability (>50%) of being significantlydistinguishable from background ASO conditions during CP ENSO events.22Maximum DHM 1983100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 1988100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 1998100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 2007100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30N  00.511.522.533.544.55Figure 2.12: EP year DHM results. The spatial distribution and intensity during EP years lack a discernible pattern frombackground variability. Only 1983 does not exhibit a region of heat stress > 2C·months, but a coherent patternof intensity in the stress distribution is lacking.23Maximum DHM 1995100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 2003100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 2005100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30NMaximum DHM 2010100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30N  00.511.522.533.544.55Figure 2.13: CP year DHM results. During CP events, the Caribbean experiences consistent instances of amplified heatstress. 2005 & 2010 have both the largest values of accumulation and distribution, but all four years identifyregions of high stress (> 2C·months)24CP event significance probabilities of DHMs100W 90W 80W 70W 60W 50W 40W 30W 20W10N15N20N25N30N 2.14: CP DHM bootstrap results showing a similarity between DHM and SSTA bootstrap probabilities and distri-bution. Residuals between the two analyses does not reveal a bias, but rather, a direct relationship.25teleconnection is only related to certain periods within the year (i.e. December-January-February (DJF)), the signal may be lost by comparing the entire time-series, hence the necessity of bootstrapping. Bootstrapping refines the spatial scale,as well as targets subsets of data as best representations of the events, and comparesit to all other states of the system within the time series of ASO data.The motivation behind the bootstrap analysis is that it avoids the problemsassociated with small sample sizes. The probability of a statistically significantt-test result in a grid-cell is represented in the results of both bootstrap tests. Theresults of this analysis identify that not all ENSO event types are the same. CPevents significantly affect the SSTA intensity and distribution in the Caribbean, butEP events do not.Due to the distribution and intensity of DHM also being linked to only CP ENSOevents, it is important to be able to identify the category of El Nin˜o event in orderto predict the impacts. As identified, by Kim et al. (2009), hurricane developmentis impacted during the development phase of ENSO, but Caribbean SST are alsoaffected on an eight month lag.The years 1997/1998 and 2006/2007 were both EP years that experienced ther-mal stress in the Caribbean that may have caused bleaching, but as an event type,there is no significantly detectable difference from background conditions. The2005 bleaching event in the Caribbean was one of the largest mass bleaching eventson record Eakin et al. (2010). The extent and intensity of heat stress in 2010 waslarger in some areas than that of the 2005 event (Alemu and Clement, 2014; Basti-das et al., 2012); however, reports of bleaching were not as widespread, possiblybecause of a different response by reefs, or a lack of reporting. Both 1994/1995 and2002/2003 were the other CP years considered and also had high DHMmax values,indicating areas likely to have been bleached. Sustained bleaching conditions arenot tolerated by corals, thus the most extremely impacted regions are likely to haveexperienced reef mortality.Since the 1990s, there has been a shift towards more CP events (Lee andMcPhaden, 2010) related to a shift in mean state of the Pacific Ocean (Banholzerand Donner, 2014). Concurrent with this shift has been an increase in the fre-quency, its associated impacts, which in this case are anomalously warm regionsof the Caribbean. Results from Coupled Model Intercomparison Project Phase265 (CMIP5) suggest a shift towards more CP events under the IntergovernmentalPanel on Climate Change Representative Concentration Pathway, 4.5 W/m2 sce-nario (RCP4.5) will continue (Kim and Yu, 2012), thus corals and their associatedecosystems will suffer stress or mortality if they cannot acclimatize or adapt.Results from the bootstrap analysis indicate variability further east of the Caribbeanthan the regions of interest, but may be of interest due to its existing within the“Main Development Region” for hurricanes(10–20N; 20–85W). This may haveimplications for hurricanes in the year following an event, not just in the develop-ment phase of ENSO events as Kim et al. (2009) identified. Investigation appearswarranted.The tropical aspect of the atmospheric bridge concept for ENSO teleconnec-tions was introduced by Klein et al. (1999) in which they identified correlationsbetween ENSO events and weakening of the Hadley cell. Around the CaribbeanSea, anomalous southwesterlies weaken the trade winds, but the primary changein SST appears to be related to changes in evaporation rather than cloud cover. Inaddition, Latif and Gro¨tzner (2000) found a contribution from the positive phaseof the annual cycle to lag time associated with ENSO. Both studies identify anapproximate six month lag.These and related studies address EP ENSO teleconnections until the spring ofthe year after the ENSO peak. As noted by Alexander et al. (2004) in their studyhighlighting Indian and Western Pacific Ocean teleconnections, most teleconnec-tion studies relating to ENSO examine the summer preceding the peak, through tothe following spring only. This practise neglects to properly address ocean andocean-coupled teleconnections that are only fully effective beyond periods greaterthan their adjustment period (Liu and Alexander, 2007). The results indicate thatthere is warming before the peak correlation found by this study, thus the tele-connection is apparent in other studies, but not necessarily fully expressed in theSSTAS. The warming teleconnection found by other studies may continue to pro-pogate as SST in the Caribbean is autocorrelative in nature, and it is also likelythat the mechanism of teleconnection for CP events is a coupled ocean-atmosphereprocess, rather than purely atmospheric pushing the observed peak impact later inthe year, which would account for the difference of 6 months from these studies tothe results of this one.27Since the peak SST influence from EP events do not occur at the same timeas the seasonal cycle causes background SST maxima, this could explain the lackof a pronounced thermal stress impact. However, a delayed teleconnection to theCaribbean during CP events may explain the heat accumulation, and the associatedeffects on coral reefs.With time, this study can be reconducted with additional data for EP and CPyears. Additional events may increase the power of the CCA to the point of statisti-cal significance, which is a challenge with the sample size adjusted for autocorre-lation. Until many more events occur, it is likely that bootstrap analysis will still berequired, but the accuracy of the tests will only be improved by having more data.Currently, reporting of coral bleaching is voluntary and not sensible in realtime, thus DHM analysis is the best proxy for bleaching available, but high qualitybleaching data would improve the value of the analysis for marine biologists.28Chapter 3Paleo-Oceanographic Records ofEl Nin˜o Influenced Sea-SurfaceTemperatures3.1 IntroductionCurrently, ENSO reconstructions that differentiate between EP and CP events onlyextend to the late 1800s, but are weakly supported due to the lack of availability anddata quality pre-1950. These reconstructions are based on satellite data, ship logs,and statistical interpolation of the data (Smith et al., 2008). ENSO reconstructionsthat do not differentiate between event types are currently available for much longertime periods using fossilized coral d 18O records, tree ring measurements, and lakevarves (McGregor et al., 2013).In order to extend the records that differentiate event types, this chapter at-tempts to create a hindcast model for both event types by combining multiplein-situ geochemical proxy temperature records from the NOAA paleoclimate dataarchive for coral cores in the tropical Pacific and Atlantic Oceans. A second modelpredicts the locations of suitable coral reefs that could be geochemically analyzedwith the intention of strengthening the results of the hindcast model. The results ofthis work intends to answer two questions: whether a combination of geochemical29temperature proxy data from the tropical Pacific and Atlantic Oceans be used tomodel the occurrence of past EP and CP ENSO events, and if a model can be devel-oped that predicts where to attain additional data that may strengthen the power ofthe hindcast model.A lack of input data from appropriate locations and timeframes prevented thehindcast model producing practical results. Because of this, the results of the sec-ond model became more important, and will be of use to those conducting paleo-climate work in the future hoping to answer similar questions about ENSO.3.2 Methods3.2.1 DataReef presence data was obtained from the United Nations Environment Programfor coral reef coverage between 1954–2009, which is currently the most compre-hensive dataset for tropical reefs. The reef data was analyzed at 14 resolution tomatch the satellite data (IMaRS-USF, 2005; IMaRS-USF and IRD, 2005; Spaldinget al., 2001; UNEP-WCMC et al., 2010).Extended Reconstruction Sea Surface Temperature (ERSST) v3b (Smith et al.,2008) provided SST data for the models. The data is spatially resolved at 2x2,and temporally at monthly intervals from 1854–2011. This dataset was developedfrom in situ measurements, and statistical reconstructions in regions of sparse data.Geochemical DataGeochemical proxy data were used to create historical SST reconstructions. Sr/Caratios are a function of temperature, while d 18O is a function of both temperatureand salinity. Due to the difference in cost between Sr/Ca analysis versus that ofd 18O, Sr/Ca ratios are less available in high resolution. The NOAA paleoclimatol-ogy database was searched for the following criteria: monthly resolution, Sr/Caor d 18O, located within the tropical Pacific or Atlantic Oceans. When both val-ues were available, individual raw chemistry data took precedence over compositegeochemical values and modelled SST/SSTA values. The data utilized is listed inTable 3.1.30Table 3.1: Geochemical Proxy DataProxy TypeLocation Sr/Ca d 18O Contributor(s)Line Islands (Palmyra,Fanning, and Christmas)x xNurhati et al. (2010)Amedee Island, NewCaledoniax xQuinn and Sampson (2003) &Stephans et al. (2005)New Ireland, Papua NewGuineaxAlibert and Kinsley (2009)Rarotonga xLinsley et al. (2000b)Palmyra xCobb and Charles (2001)Clipperton xLinsley et al. (2000a)Guam xAsami et al. (2005)3.2.2 NINO Hindcast ModelAll input timeseries of geochemical temperature proxies were independently cor-related with both the NINO3 and NINO4 time periods that overlap with the ERSSTdata.Calibration required the development of event identification criteria from cor-relating ERSST data with temperature proxy data. Each criterion response wasintended to be binary in nature based upon individual cores, leading to a decisiontree that would identify an event based upon a set of conditions being met, ratherthan a direct quantitative correlation.To determine which datasets were useful in differentiating between event types,and thus useful for hindcasting, comparisons of DJF data distribution from knownENSO event years were compared with each other and neutral conditions from thesame dataset using boxplots. By finding locations where an event type would beisolated in it’s distribution would suggest that it may be suitable for significancetesting.Each dataset of geochemical proxy measurements was correlated with both31the NINO3 and NINO4 regions independently (Pearson’s Correlation Coefficient,a=0.05), then plotted side by side. As with the boxplots, correlations were onlyanalyzed for subsets of the data containing DJF of that region’s event type (NINO3(NINO4) = EP (CP)). Each dataset’s value was determined by whether the correla-tion values with the NINO3 were significantly different from that of the NINO4.3.2.3 Drill Prediction ModelIn order to determine ideal drilling locations for geochemical data that can discernbetween CP and EP events, correlations between the NOAA OI-SST DJF SSTA time-series of each grid cell between 20 N and 20 S latitude was correlated with theDJF time-series of the scalar NINO3 and NINO4 regions with zero lag (a = 0.01n = 81). Grid cells were then filtered for reef presence, and whether significantcorrelations existed between the individual cell and only one of the NINO regions(i.e. correlated with the NINO3, but not the NINO4 region). Those that met theconditions were then presented as final output.3.3 Results3.3.1 Existing DataBoxplots of Sr/Ca data reveal that current data cannot clearly discern all three states(Neutral/EP/CP) of ENSO from one another (Figure 3.1). It is seen that in all cases,the interquartile ranges of at least two of the three states of each location displayoverlap. Inclusion of d 18O to try and find sufficiently discriminatory data yieldsthe same results (not pictured). For all sites with available geochemical records, thecondition of having a single ENSO type (represented by a NINO region) differingfrom both the other ENSO type and neutral conditions was not met.In spite of the lack of differentiation in the data distribution, Pearson’s Correla-tion Coefficients were derived for each of the Sr/Ca datasets. During EP years, noneof the analyzed Sr/Ca data are significantly correlated (a=0.05) with the NINO3 re-gion. However, during CP years, the Palmyra and Amedee (2004) cores are bothsignificantly correlated (a=0.05) with the NINO4 region (Figure 3.2). Interestingly,the Christmas Island core is significantly correlated with the NINO3 region during32Figure 3.1: Boxplots of Sr/Ca ratios reveal that none of the sites have either EP or CP events discernible from both othertypes.33these times too, but not during EP years. At the a=0.01 level, none of the cores aresignificantly correlated for either region during either event type.3.3.2 Potential DataGrid cells containing potential data sources are only listed for areas significantlycorrelated (Pearson’s Correlation Coefficient, p<0.01) with the NINO3 but notNINO4 (EP events), and NINO4 but not NINO3 (CP events) and test positive forreef presence. In the analysis, there are a total of 230400 cells, with 176616 beingoceanic, and 6518 containing coral reefs. There are 492 cells (7.55% of possible)with potential data available for identifying EP ENSO, and 890 (13.65% of possible)for CP ENSO.Data for Identifying Eastern Pacific EventsPossible data that could lead to the hindcasting of EP ENSO events may exist in:Fiji, Northern Marshall Islands, French Polynesia, Lamon Bay in the Philip-pines, Spratly Islands, the islands surrounding Batam in Indonesia, Paracel islands,and South East Celebes Sea off the coast of East Kalimantan (Figure 3.3).Data for Identifying Central Pacific EventsPossible data that could lead to the hindcasting of CP ENSO events may exist in:Tuvalu and southeast to American Samoa, Southern Marshall Islands, Mi-cronesia, Solomon Islands, Louisiade Archipelago of Papua New Guinea, Bis-marck Archipelago, the islands at the north of Cendrawasih Bay in Papua NewGuinea, the islands at the north of Helmahera Sea in Papua New Guinea, Palau,entire Southern Philippines excluding Palawan, west coast of the north half ofSumatra in Indonesia including the islands nearby, and Nicobar Islands of India(Figure 3.4).3.4 DiscussionThe paleoclimate records of EP and CP ENSO currently do not pre-date observa-tional records (1950). Without differentiating between EP and CP events, recordsof ENSO exist back to 930 CE (with gaps between at several points) (Kim et al.,341940 1960 1980 20009.49.59.6n=10R=−0.402p=0.2491960 1970 1980 19908.748.768.788.8n=3R=−0.679p=0.5251940 1960 1980 20008.68.899.2n=10R=−0.195p=0.5901975 1980 1985 1990 1995262830n=4R=0.979p=0.0211970 1980 1990 2000 2010262830n=6R=0.614p=0.1951975 1980 1985 1990 19952627282930n=4R=0.769p=0.2311960 1970 1980 19909.079.0759.089.0859.099.095n=3R=0.998p=0.03728293028.628.82929.22828.52929.528.52929.528.52929.528.628.82929.229.428.728.828.92929.129.2Figure 3.2: Correlations between Sr/Ca core raw ratios or temperature reconstructions (blue) and ERSST temperaturedata (green).35Figure 3.3: Map of tropical Pacific plus Eastern Indian Ocean and Caribbean Sea with green grid cells indicatingregions where reefs exist and are correlated with the NINO3, but not the NINO4 region36Figure 3.4: Map of tropical Pacific plus Eastern Indian Ocean and Caribbean Sea with green grid cells indicatingregions where reefs exist and are correlated with the NINO4, but not the NINO3 region372009), however, this unsuccessfully provides indications of historical impacts onthe Caribbean (Chapter 2). The difficulty in separating correlations between proxydata and only one of the NINO3 or NINO4 regions is the limiting factor in creatinghindcast output. Even with this complication, there is useful information offeredby this research that could lead to a valuable hindcast.Ashok et al. (2007) identifies a tripole in the second EOF of the Pacific Ocean,which they identify as characterizing CP ENSO. The Western Pacific pole is notas intense as the Central and Eastern poles, but it is still notable around EasternIndonesia, and Papua New Guinea with an inverse signature. The Eastern Pacifichas a strong inverse signal, which is strongest at the Eastern boundary of the Oceanwhere it meets land. It is apparent that CP ENSO events have relationships with theentire equatorial Pacific.From the results, it is interesting to note that in the Pacific Ocean, data appearsto be available outside 5N–15S for EP, but within that region for CP. This mayhave a relationship with the Intertropical Convergence Zone (ITCZ) and the South-ern Pacific Convergence Zone (SPCZ), as previous studies have drawn links be-tween ENSO and these climatological features, but without differentiation betweenCP and EP events (Folland et al., 2002; Juillet-Leclerc et al., 2006). Referring backto the EOF analysis of the Pacific Ocean by Ashok et al. (2007), the intensity of EPENSO signature is higher in the equatorial region, but the relationship is latitudi-nally narrower than that of the CP events. Along with the difference in latitudinalsignal thickness, wind stress anomalies indicate that the Western arm of the SPCZexperiences northerlies during development until it peaks, and then the eastern armexperiences southerlies. A differentiation may reveal a closer relationship betweena particular ENSO event type and the SPCZ as found in Chapter 2, or just a changein the characteristics of the SPCZ based on the differences in wind stress anomaliesand relationships SSTA.Existing research indicates that monsoon conditions are coupled to at least oneENSO type (Krishnamurthy and Kirtman, 2003; Kug and Kang, 2006; Qu et al.,2005). Unfortunately, the research does not have an indication of which type ofENSO as their definition is based off of undifferentiated events. The South ChinaSea and Indian Oceans are both involved in the ENSO–Monsoon system coupling,hence their appearance in the list of potential data sites. However, it is unfortunate38that there is not more of an abundance of reefs in the region from which to collectdata. Qu et al. (2005) found changes in wind stress due to ENSO weakened theinter-ocean heat flux through the South China Sea and into the Indian Ocean via theIndonesian Throughflow. It is conceivable that a reanalysis differentiating betweenENSO types would find that different heat transport pathways would correspondwith channels containing the reefs outlined in both Figure 3.3 and Figure 3.4.39Chapter 4ConclusionsRecent changes in the understanding of ENSO dynamics created the necessity fornew investigations into its teleconnections. In this study, implications for theCaribbean were investigated prompted by the appearance of severe coral bleachingevents following CP ENSO events. By understanding the links between differenttypes of ENSO and their affect on SST in the Caribbean, as well as when they havehappened historically, a better understanding of the effects on coral reefs as wellas other topics of regional importance (e.g. hurricane development) can be investi-gated.Though the timing of the lagged response in the Caribbean could not be deter-mined with statistical significance, it appears that there is a 7 month delay betweenan ENSO event, and a response in the Eastern region of the Caribbean. The boot-strap model determined that CP ENSO activity has a strong probability of affectingSST in the Caribbean(> 50% in a majority of the region), and are also related tothe DHM responses.When investigating the possibility of a hindcast model, which would be usefulin understanding historical impacts upon the Caribbean, a lack of sufficient highresolution proxy data became apparent. In lieu of available data, a model suggest-ing locations to obtain data that can be used in the hindcast model discovered thatthere are many potential sites distributed throughout the tropics. The sites of mostimportance appear to be related to the SPCZ, ITCZ, and asian monsoons, warrantingfurther investigation into the difference between ENSO types and their relationships40with those features.The limited availability of known EP and CP ENSO events necessitated the useof the bootstrapping technique in the model. With time, additional ENSO eventswill be identified expanding the input data that can be run through the model, butthe need for bootstrapping will still exist. Nevertheless, the confidence of the modelwill be stronger with a larger sample size.When it comes to obtaining new high resolution core analyses, the results ofthis work ought to be considered. This may be sufficient from a pure scientific view,but geopolitical concerns are a reality of any research, thus an important consider-ation about where to drill for data. Luckily, the potential sites for drilling appearto be plentiful, providing sufficient options even if certain sites are not consideredfeasible.Globally, there are many teleconnections of ENSO events that can be inves-tigated. While previous schools of thought believed that there were connectionsbetween ENSO events and SST in the Caribbean, they did not have an understand-ing of the different types of ENSO. This thesis provides evidence for a relationshipwith CP ENSO, and in doing so, shows that there is insufficient evidence to es-tablish a relationship with EP ENSO. Additionally, it highlights the necessity forfurther data collection in order to understand the history of different event types,and makes suggestions on where to drill in order do so.41ReferencesJ. B. Alemu and Y. Clement. Mass coral bleaching in 2010 in the southerncaribbean. PLoS one, 9(1):e83829, 2014. ! pages 7, 26M. A. Alexander, N.-C. Lau, and J. D. Scott. Broadening the atmospheric bridgeparadigm: Enso teleconnections to the tropical west pacific-indian oceans overthe seasonal cycle and to the north pacific in summer. Earths Climate: TheOcean-Atmosphere Interaction, Geophys. Monogr, 147:85–103, 2004. !pages 27C. Alibert and L. Kinsley. Sr/ca and ba/ca coral record from new ireland, papuanew guinea, for the past 170 years, 2009. IGBP PAGES/World Data Center forPaleoclimatology Data Contribution Series # 2009-142 NOAA/NCDCPaleoclimatology Program, Boulder CO, USA. ! pages 31R. Asami, T. Yamada, Y. Iryu, C. Meyer, and T. Q. G. Paulay. Guam coral oxygenisotope data for 1790 to 2000, 2005. IGBP PAGES/World Data Center forPaleoclimatology Data Contribution Series #2005-051. NOAA/NGDCPaleoclimatology Program, Boulder CO, USA. ! pages 31K. Ashok, S. Behera, S. Rao, H. Weng, and T. Yamagata. El nin˜o modoki and itspossible teleconnection. J. Geophys. Res, 112(10.1029), 2007. ! pages 1, 2, 3,9, 10, 38S. Banholzer. The central pacific el nin˜o and its impact on weather and forest firepatterns in western north america. Master’s thesis, University of BritishColumbia, 2012. ! pages vii, 2, 8, 10S. Banholzer and S. Donner. The influence of different el nio types on globalaverage temperature. Geophysical Research Letters, 41(6):2093–2099, 2014.ISSN 1944-8007. doi:10.1002/2014GL059520. URLhttp://dx.doi.org/10.1002/2014GL059520. 2014GL059520. ! pages 2642C. Bastidas, D. Bone, A. Croquer, D. Debrot, E. Garcia, A. Humanes, R. Ramos,and S. Rodrı´guez. Massive hard coral loss after a severe bleaching event in2010 at los roques, venezuela. Revista de Biologı´a Tropical, 60:29–37, 2012.! pages 7, 26C. Birkeland. Introduction, pages 1–12. Life and death of coral reefs. 1997. !pages 5B. E. Brown. Disturbances to reefs in recent times, pages 354–379. Life anddeath of coral reefs. 1997. ! pages 6, 7M. A. Cane. The evolution of el nin˜o, past and future. Earth and PlanetaryScience Letters, 230(3-4):227–240, 2/15 2005. ! pages 2, 4K. E. Chave, S. V. Smith, and K. J. Roy. Carbonate production by coral reefs.Marine Geology, 12(2):123–140, 2 1972. ! pages 5K. M. Cobb and C. Charles. Palmyra island monthly coral oxygen isotope data,2001. IGBP PAGES/World Data Center for Paleoclimatology DataContribution Series #2001-043. NOAA/NGDC Paleoclimatology Program,Boulder CO, USA. ! pages 31K. M. Cobb, C. D. Charles, H. Cheng, and R. L. Edwards. El nino/southernoscillation and tropical pacific climate during the last millennium. Nature, 424(6946):271–276, Jul 17 2003. LR: 20061115; JID: 0410462; 0 (OxygenIsotopes); CIN: Nature. 2003 Jul 17;424(6946):261-2. PMID: 12867962;2003/03/17 [received]; 2003/05/30 [accepted]; ppublish. ! pages 3S. Donner. Coping with commitment: projected thermal stress on coral reefsunder different future scenarios. Plos One, 4(6):e5712, 2009. ! pages 12S. Donner, W. Skirving, C. Little, M. Oppenheimer, and O. Hoegh-Guldberg.Global assessment of coral bleaching and required rates of adaptation underclimate change. Global Change Biology, 11(12):2251–2265, 2005. ! pages 12S. D. Donner, T. R. Knutson, and M. Oppenheimer. Model-based assessment ofthe role of human-induced climate change in the 2005 caribbean coralbleaching event. Proceedings of the National Academy of Sciences, 104(13):5483–5488, 2007. ! pages 7C. Eakin. Climate, weather and coral bleaching, page 41. Coral Bleaching -Patterns, Processes, Causes and Consequences. Springer-Verlag EcologicalStudies: Berlin, Heidelberg, 2009. ! pages 743C. Eakin, J. Morgan, S. Heron, T. Smith, G. Liu, L. Alvarez-Filip, B. Baca,E. Bartels, C. Bastidas, C. Bouchon, et al. Caribbean corals in crisis: recordthermal stress, bleaching, and mortality in 2005. PloS one, 5(11):e13969, 2010.! pages 6, 20, 26B. Efron and R. Tibshirani. Statistical data analysis in the computer age. Science,253(5018):390–395, 1991. ! pages 11C. Folland, J. Renwick, M. Salinger, and A. Mullan. Relative influences of theinterdecadal pacific oscillation and enso on the south pacific convergence zone.Geophysical Research Letters, 29(13):21–1, 2002. ! pages 38D. K. Hubbard. Reefs as dynamic systems, pages 43–67. Life and death of coralreefs. 1997. ! pages 6IMaRS-USF. Millennium coral reef mapping project., 2005. Unvalidated maps.These maps are unendorsed by IRD, but were further interpreted by UNEPWorld Conservation Monitoring Centre. Cambridge (UK): UNEP WorldConservation Monitoring Centre. ! pages 30IMaRS-USF and IRD. Millennium coral reef mapping project., 2005. Validatedmaps. Cambridge (UK): UNEP World Conservation Monitoring Centre. !pages 30A. Juillet-Leclerc, S. Thiria, P. Naveau, T. Delcroix, N. Le Bec, D. Blamart, andT. Correge. Spcz migration and enso events during the 20th century as revealedby climate proxies from a fiji coral. Geophysical research letters, 33(17), 2006.! pages 38H.-Y. Kao and J.-Y. Yu. Contrasting eastern-pacific and central-pacific types ofenso. Journal of Climate, 22(3):615–632, FEB 2009. PT: J; UT:ISI:000263908000010. ! pages 2, 3H. Kim, P. Webster, and J. Curry. Impact of shifting patterns of pacific oceanwarming on north atlantic tropical cyclones. Science, 325(5936):77–80, 2009.! pages 11, 26, 27, 34S. T. Kim and J.-Y. Yu. The two types of enso in cmip5 models. GeophysicalResearch Letters, 39(11):L11704, 2012. ! pages 27S. A. Klein, B. J. Soden, and N.-C. Lau. Remote sea surface temperaturevariations during enso: Evidence for a tropical atmospheric bridge. Journal ofClimate, 12(4):917–932, 1999. ! pages 7, 2744V. Krishnamurthy and B. P. Kirtman. Variability of the indian ocean: Relation tomonsoon and enso. Quarterly Journal of the Royal Meteorological Society, 129(590):1623–1646, 2003. ! pages 38J.-S. Kug and I.-S. Kang. Interactive feedback between enso and the indian ocean.Journal of climate, 19(9):1784–1801, 2006. ! pages 38M. Latif and A. Gro¨tzner. The equatorial atlantic oscillation and its response toenso. Climate Dynamics, 16(2):213–218, 2000. ! pages 27T. Lee and M. J. McPhaden. Increasing intensity of el nino in thecentral-equatorial pacific. Geophysical Research Letters, 37:L14603, JUL 242010. PT: J; UT: ISI:000280326400005. ! pages 2, 10, 26G. Li, B. Ren, C. Yang, and J. Zheng. Indices of el nin˜o and el nin˜o modoki: Animproved el nin˜o modoki index. Advances in Atmospheric Sciences, 27(5):1210–1220, 2010. ! pages 8, 9, 10B. Linsley, L. Ren, R. Dunbar, and S. Howe. Clipperton atoll coral stable isotopedata, 2000a. IGBP PAGES/World Data Center-A for Paleoclimatology DataContribution Series # 2000-048. NOAA/NGDC Paleoclimatology Program,Boulder CO, USA. ! pages 31B. Linsley, G. Wellington, and D. Schrag. Rarotonga sr/ca and sst reconstructiondata, 2000b. IGBP PAGES/World Data Center for Paleoclimatology DataContribution Series #2000-065. NOAA/NGDC Paleoclimatology Program,Boulder CO, USA. ! pages 31Z. Liu and M. Alexander. Atmospheric bridge, oceanic tunnel, and global climaticteleconnections. Reviews of Geophysics, 45(2), 2007. ! pages 27J. Lough and M. van Oppen. Introduction: Coral BleachingPatterns, Processes,Causes and Consequences, pages 1–5. Coral BleachingPatterns, Processes,Causes and Consequences. Springer, 2009. ! pages 5, 6J. J. McCarthy, O. F. Canziani, N. A. Leary, and D. J. Dokken. Climate change2001: Impacts, adaptation, and vulnerability. 2007. ! pages 6S. McGregor, A. Timmermann, M. England, O. Elison Timm, and A. Wittenberg.Inferred changes in el nin˜o–southern oscillation variance over the past sixcenturies. Climate of the Past, 9(5):2269–2284, 2013. ! pages 29M. J. McPhaden. Evolution of the 2002/03 el nin˜o. Bulletin of the AmericanMeteorological Society, 85(5):677–695, 2004. ! pages 145C. J. Mock. PALEOCLIMATE MODELING — Paleo-ENSO, pages 1934–1941.Encyclopedia of Quaternary Science. Elsevier, Oxford, 2007. ISBN978-0-44-452747-9. ! pages 2, 3, 4G. Muller-Parker and C. F. D’Elia. Interactions between corals and theirsymbiotic algae, pages 96–113. Life and death of coral reefs. Chapman & Hall,1997. ! pages 6I. Nurhati, K. Cobb, C. Charles, and R. Dunbar. Line islands coral sst and d18oswreconstructions, 2010. IGBP PAGES/World Data Center for PaleoclimatologyData Contribution Series # 2010-012. NOAA/NCDC PaleoclimatologyProgram, Boulder CO, USA. ! pages 31J. K. Oliver, R. Berkelmans, and C. Eakin. Coral bleaching in space and time,pages 21–40. Coral BleachingPatterns, Processes, Causes and Consequences.Springer, 2009. ! pages 6, 7M. Pfeiffer. Corals as archives of past climate, volume 1 of The Biology of CoralReefs, chapter 9, pages 241–243. Oxford Scholarship Online Monographs,2009. ! pages 3, 4R. Preisendorfer and C. Mobley. Principal component analysis in meteorologyand oceanography, volume 425. Elsevier New York, 1988. ! pages 10T. Qu, Y. Du, G. Meyers, A. Ishida, and D. Wang. Connecting the tropical pacificwith indian ocean through south china sea. Geophysical research letters, 32(24), 2005. ! pages 38, 39T. Quinn and D. Sampson. Amedee coral elemental ratios and isotope data, 2003.IGBP PAGES/World Data Center for Paleoclimatology Data ContributionSeries # 2003-073 NOAA/NCDC Paleoclimatology Program, Boulder CO,USA. ! pages 31E. M. Rasmusson and T. H. Carpenter. Variations in tropical sea surfacetemperature and surface wind fields associated with the southern oscillation/elnio. Monthly Weather Review, 110(5):354–384, 1982. ! pages 1, 2R. Reynolds, T. Smith, C. Liu, D. Chelton, K. Casey, and M. Schlax. Dailyhigh-resolution-blended analyses for sea surface temperature. Journal ofClimate, 20(22):5473–5496, 2007. ! pages 8, 12B. Santer, J. Boyle, J. Hnilo, K. Taylor, T. Wigley, D. Nychka, D. Gaffen, andD. Parker. Statistical significance of trends and trend differences in46layer-average atmospheric temperature time series. Journal of GeophysicalResearch, 105(D6):7337–7356, 2000. ! pages 13C. R. C. Sheppard, S. K. Davy, and G. M. Pilling. Coral reefs in the modernworld, pages 223–254. The biology of coral reefs. 2009a. ! pages 6C. R. C. Sheppard, S. K. Davy, and G. M. Pilling. The main reef builders andspace occupiers, pages 33–65. The biology of coral reefs. 2009b. ! pages 6T. M. Smith, R. W. Reynolds, T. C. Peterson, and J. Lawrimore. Improvements tonoaa’s historical merged land-ocean surface temperature analysis (1880-2006).Journal of Climate, 21(10):2283–2296, 2008. ! pages 29, 30M. Spalding, C. Ravilious, and E. P. Green. World atlas of coral reefs. Univ ofCalifornia Press, 2001. ! pages 30C. Stephans, T. Quinn, F. Taylor, and T. Correge. New caledonia porites coralpaleothermometry assessment data, 2005. IGBP PAGES/World Data Center forPaleoclimatology Data Contribution Series # 2005-031 NOAA/NCDCPaleoclimatology Program, Boulder CO, USA. ! pages 31K. E. Trenberth and D. P. Stepaniak. Indices of el nino evolution. Journal ofClimate, 14(8):1697–1701, 2001. ! pages 2, 3UNEP-WCMC, W. Centre, WRI, and TNC. Global distribution of warm-watercoral reefs, compiled from multiple sources, and including imars-usf and ird(2005), imars-usf (2005) and spalding et al. (2001), 2010. Cambridge (UK):UNEP World Conservation Monitoring Centre. ! pages 30J. E. N. Veron. Corals in space and time: the biogeography and evolution of theScleractinia. Cornell Univ Pr, 1995. ! pages 7C. Wang and P. C. Fiedler. Enso variability and the eastern tropical pacific: Areview. Progress in Oceanography, 69(2-4):239–266, 6 2006. ! pages 1, 4, 10R. Wood. The Changing Fate of Coral Reefs: Lessons from the Deep Past, pages3–27. Geological approaches to coral reef ecology. Springer, 2007. ! pages 6,7S. Yeh, J. Kug, B. Dewitte, M. Kwon, B. Kirtman, and F. Jin. El nin˜o in achanging climate. Nature, 461(7263):511–514, 2009. ISSN 0028-0836. !pages 1, 2, 347S. Yeh, B. Kirtman, J. Kug, W. Park, and M. Latif. Natural variability of thecentral pacific el nin˜o event on multi-centennial timescales. GeophysicalResearch Letters, 38(2):L02704, 2011. ! pages 10J.-Y. Yu, Y. Zou, S. T. Kim, and T. Lee. The changing impact of el nin˜o on uswinter temperatures. Geophysical Research Letters, 39(15):L15702, 2012. !pages 1, 8, 1048


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items