CoastalupwellingdynamicsinfluencespatialpatternsofhypoxiaandnearshorehypoxiaeventsinthecentralbasinofLakeErie
MarkD.Rowe*1,E.J.Anderson2,S.A.Ruberg2,D.Beletsky1,H.Zhang1,T.H.Johengen1
,S.Moegling3,E.M.Verhamme6, C. A. Stow2
*Presenter,1UniversityofMichigan,CooperativeInstituteforGreatLakesResearch,2NOAAGreatLakesEnvironmentalResearchLaboratory,3ClevelandDivisionofWater,4PurdueUniversity,5USGeologicalSurvey,6LimnoTech
IntroductionLakeEriesustainsanimportantfisheryandisasourceofdrinkingwatertomillionsofpeople,buthypoxiaisdetrimentaltobothoftheseecosystemservices.HypoxiainthecentralbasinofLakeEriehasbeenstudiedsincethe1950’s.Evenso,spatialpatternsofhypoxia,andepisodesofhypoxiainnearshoreareaswheredrinkingwaterplantintakesarelocated,arenotwellcharacterizedowingtolimitedobservations.Hypoxiaisultimatelycausedbyexcessivenutrientloading,stimulatingalgalgrowth,andincreasedbiochemicaloxygendemand.Inaddition,physicalprocesseshaveanimportantinfluenceontheoccurrenceofhypoxia,especiallyprocessesthatinfluencethethicknessofthehypolimnion.
Conclusion
Influence of coastal upwelling on hypoxia in Lake Erie
Hypoxia model and sensor network
Acknowledgements & References
Weevaluatedtheabilityofarelativelysimple,physically-based,dissolvedoxygen(DO)modeltoreproducespatialandtemporalpatternsofhypoxiaobservedinLakeErie.TheDOmodel(Fig.1)usedassignedratesofsedimentandwatercolumnoxygendemand(Muelleretal.,2012;Rucinski etal.,2010)thatweretemperature-dependent,butotherwisespatiallyandtemporallyuniforminthecentralbasin.TheDOinitialconditionwassettosaturationwiththeatmosphereonApril1,atimewhenLakeErieisverticallywell-mixed.TheDOmodelwaslinkedtoNOAA’sLakeErieOperationalForecastingSystemhydrodynamicmodel,anapplicationoftheFiniteVolumeCommunityOceanModel(FVCOM)with0.5-2kmhorizontalresolutionand20verticallayers.Ahindcast simulationwasrunfor2016usingmeteorologyinterpolatedfromstationssurroundingthelake.Inthesummerof2017,themodelwasruninanowcast/forecastmode.Anowcast simulationwasrunevery24hours,initializedfromtheprevious-daynowcast,usinginterpolatedmeteorology,anda14-dayforecastsimulationwasrunusingmeteorologyfromNOAANDFDtoday5andGFSmodeltoday14.ModelDOwascomparedtoobservationsfromreal-timesensornetworksmaintainedbyGLOS,LimnoTech,andMOECCOntarioCanada,inadditiontoanarrayofmooredsensorsthatwedeployedin2017.In2016,anarrayoftemperatureandDOsensorswasinstalledonthebottomofthecentralbasinbyUSEnvironmentalProtectionAgencyandPurdueUniversity(SeaGrant).
EarlieronsetofhypoxiaalongthesouthshoreofLakeErie,comparedtothenorth,wasassociatedwithpredominantsouthwesterlywindsthatcausedathinnerhypolimnion inthesouthduetodownwelling associatedwithEkmantransport.Thisspatialpatternofhypoxiawassimulatedbyamodelthatusedspatially-uniformratesofoxygendemand,indicatingthatspatiotemporalvariationinoxygendemandisnotnecessarytoexplainthispattern.
LowtemperatureanddissolvedoxygeneventsatnearshorewaterintakesinOhiowereassociatedwithupwellingcausedbynortheasterlywindsandassociatedEkmantransport.Sucheventswereforecastbythemodelin2017withgreaterskillthanapersistenceforecast3-12daysinadvance.
ThisresearchisfundedbytheNationalOceanicandAtmosphericAdministrationNationalCentersforCoastalOceanScienceCenterforSponsoredCoastalOceanResearchunderawardNA16NOS4780209toUniversityofMichiganandNOAAGLERL
Muller,B.,L.D.Bryant,A.Matzinger,andA.Wuest.2012.Hypolimneticoxygendepletionineutrophiclakes.Environmentalscience&technology.46(18):9964-9971.Rucinski,D.K.,D.Beletsky,J.V.DePinto,D.J.Schwab,andD.Scavia.2010.Asimple1-dimensional,climatebaseddissolvedoxygenmodelforthecentralbasinofLakeErie.J.GreatLakesRes.36(3):465-476.Hogan,R.J.,andI.B.Mason(2012),Deterministicforecastsofbinaryevents,inForecastVerification:APractitioner’sGuideinAtmosphericScience,2nded.,pp.31–59,JohnWiley&Sons,Ltd.,WestSussex,U.K.
Figure 1. Process diagram of thedissolved oxygenmodel linked to theLake Erie Operational ForecastingSystem:https://tidesandcurrents.noaa.gov/ofs/leofs/leofs.html
Dominantsouthwesterlywindscausedpersistentdownwelling alongthesouthshore,whichresultedinathinnerhypolimnion andearlierinitiationofhypoxiaalongthesouthshorethaninthenorthinboth2016and2017(Figure4).Occasionalwindsfromthenortheasttemporarilyreversedthispattern,causingupwellingalongthesouthshorethatbroughthypoxicwatertonearshorewaterintakes(Figures5-6).
OC44A-0499
Figure 2. Map of Lake Erie, showingbathymetry and locations of real-time sensors (GLOS, LimnoTech) andsensor moorings installed as part ofthis study.
The dashed line indicates the cross-sectional view from Cleveland toPort Glasgow shown in Figures 4-6.
http://oceanmotion.org/html/background/upwelling-and-downwelling.htm
Figure 3. In oceans and large lakes, the rotation of the Earthdeflects surface current to the right of the wind stress (northernhemisphere), in a process known as Ekman transport. With windoriented parallel to the coast, surface current directed toward(away) from the coast results in downwelling (upwelling), whichresults in deepening (shoaling) of the thermocline. A deepthermocline (thin hypolimnion) causes earlier onset and moresevere hypoxia, for a given rate of oxygen consumption.
Figure 4. Cross-sectional views of the LakeErie model in mid July of 2016 and 2017show upwelling on the north shore anddownwelling on the south (a,b arrows),which is consistent with Ekman transportcaused by the dominant southwesterlywind (e). A thinner hypolimnion along thesouth shore led to earlier initiation ofhypoxia than in the north (a,b arrows; c,d).
Figure 5. An episode of northeasterlywind (e) occurred in July 28-30, 2017,which caused upwelling along the southshore (a,b arrows), as indicated by thethermistor string at buoy 45176. Sensorsin water intakes at Mentor and Ashtabulashowed a decrease in temperature anddissolved oxygen (c,d arrows). The modelover-predicted dissolved oxygen in thecenter of the central basin, which mayhave been caused by the shallow-biasedthermocline depth, as indicated by thethermistor string at buoy 45164 (Figs. 4-6).
Figure 6. A second example of upwelling(a,b arrows) occurred in September 1-3,2016, again associated with anortheasterly wind event (e). Temperatureand dissolved oxygen decreased at theMentor and Ashtabula water intakes (c,darrows).
Cleveland,Ohio PortGlasgow,Ontario
Explanation of Figures 4-6Model temperature and DO are shown ona color scale, with observations fromsensors shown by symbols using the samecolor scale. The location of the cross-sectional view is shown in Figure 1. Thewind barb from buoy 45164 is in the lowerleft. Bottom temperature and DO areshown in the lower panels.
Buoy45164
Buoy45176
Figure 7. Association between pH anddissolved oxygen, measured at sixdrinking water intakes in 2017.
Figure 8. Skill scores for the modeland a persistence forecast inpredicting events of DO < 4 mg/L atsix drinking water intakes in 2017.
Weusedarchivednowcast andforecastmodelrunsfrom2017toevaluatemodelskillinpredictingrapidchangesintemperatureanddissolvedoxygenatsixdrinkingwaterintakes(AvonLake,threeintakesatCleveland,Mentor,andAshtabula).LowDOitselfisnotanissueforwatertreatment,buthypoxicwaterisassociatedwithlowpHandreducedsubstances(Mn,Fe)thatcancauseissuesforwatertreatment.DOandpHwerecorrelated,withDO<4mg/LbeingassociatedwithpH<7.5,anactionthresholdforwatertreatment,>60%ofthetime(Figure7).Forskillassessment,wedefinedeventsasDO<4mg/Loradropinhourlytemperatureof>3°Ccomparedtothemaximumtemperatureoverthepreceding24hours.ThePierceskillscoregivestheprobabilityofacorrectprediction,giventhataneventoccurred,minustheprobabilityofafalsepositive,givenanon-event(HoganandMason,2012).Apersistenceforecastwasselectedasabenchmark,representingtheassumptionofnochangefromthemostrecentobservationandthebestavailableinformationintheabsenceofamodel.Thepersistenceforecastis,bydefinition,perfectforthenowcast (day0),butthemodelskillexceededthatofthepersistenceforecastforforecastdays3-12(Figures8-9).
Figure 9. Skill scores for the modeland a persistence forecast inpredicting events of 3 °C temperaturedrop in 24 hours at six drinking waterintakes in 2017.
a b
c d
e
a b
c d
e
a b
c d
e
