sensitivity of hypoxia predictions for the northern gulf...

Sensitivity of hypoxia predictions for the northern Gulf of Mexicoto sediment oxygen consumption and model nesting

Katja Fennel1 Jiatang Hu13 Arnaud Laurent1 Martinho Marta-Almeida2 andRobert Hetland2

Received 30 August 2012 revised 10 December 2012 accepted 7 January 2013 published 28 February 2013

[1] Every summer a large area (15000 km2 on average) over the TexasndashLouisiana shelf inthe northern Gulf of Mexico turns hypoxic due to decay of organic matter that is primarilyderived from nutrient inputs from the MississippiAtchafalaya River System Interannualvariability in the size of the hypoxic zone is large The 2008 Action Plan put forth by theMississippi RiverGulf of MexicoWatershed Nutrient Task Force an alliance of multiple stateand federal agencies and tribes calls for a reduction of the size of the hypoxic zone throughnutrient management in the watershed Comprehensive models help build mechanisticunderstanding of the processes underlying hypoxia formation and variability and are thusindispensable tools for devising efficient nutrient reduction strategies and for buildingreasonable expectations as to what responses can be expected for a given nutrient reductionHere we present such a model evaluate its hypoxia simulations against monitoringobservations and assess the sensitivity of the hypoxia simulations to model resolutionvariations in sediment oxygen consumption and choice of physical horizontal boundaryconditions We find that hypoxia simulations on the shelf are very sensitive to theparameterization of sediment oxygen consumption a result of the fact that hypoxic conditionsare restricted to a relatively thin layer above the bottom overmost of the shelfWe show that thestrength of vertical stratification is an important predictor of dissolved oxygen concentration inbottom waters and that modification of physical horizontal boundary conditions can have alarge effect on hypoxia simulations because it can affect stratification strength

Citation Fennel K J Hu A Laurent M Marta-Almeida and R Hetland (2013) Sensitivity of hypoxia predictionsfor the northern Gulf of Mexico to sediment oxygen consumption and model nesting J Geophys Res Oceans 118990ndash1002 doi101002jgrc20077

1 Introduction

[2] The TexasndashLouisiana shelf (TX-LA shelf) experienceslow concentrations of dissolved oxygen (referred to ashypoxia when oxygen concentrations are lt2 mg O2l or 64mmol O2 m

3) every summer The spatial extent of the areaaffected by hypoxia has increased since the 1950s in parallelwith increases in river nitrate loads from the MississippiAtchafalaya River System [Rabalais et al 2002] Since1985 the spatial extent of hypoxic conditions has beenmonitored systematically with at least one shelf-wide

monitoring cruise typically in late July or early August Thelong-term average size of the hypoxic area is 15000 km2however interannual variability in the observed hypoxicextent is large for example due to variations in freshwaterdischarge and nutrient load (drought versus flood years) andatmospheric weather patterns (passage of tropical storms thatlead to enhanced vertical mixing and ventilation of bottomwaters) (see httpwwwgulfhypoxianet)[3] The major driver leading to the formation of hypoxia

on the shelf is decay of nutrient-stimulated phytoplank-ton growth fueled by excessive nutrient inputs fromthe MississippiAtchafalaya River System The MississippiRiver is one of the worldrsquos largest rivers draining about41 of the contiguous US including tile-drained cornfieldsin the Midwest The latter contribute significant amounts offertilizer-derived nitrogen to the river runoff [Goolsby et al2000 David et al 2010] Nitrate load which made up 61of the total nitrogen load from 1980 to 1996 has tripled whencompared to that in the 1970s [Goolsby et al 2000][4] Concerns about increasing nutrient loads prompted the

formation of the Mississippi RiverGulf of Mexico Water-shed Nutrient Task Force in 1996 an alliance of multiplestate and federal agencies and tribal organizations In its2008 Action Plan the Task Force articulated as one of three

All Supporting Information may be found in the online version ofthis article

1Department of Oceanography Dalhousie University Halifax Canada2Department of Oceanography Texas AampM University College

Station Texas USA3School of Environmental Science and Engineering Sun Yat-Sen

University Guangzhou China

Corresponding author K Fennel Department of OceanographyDalhousie University PO Box 15000 Halifax NS B3H 4R2 Canada(katjafenneldalca)

copy 2013 American Geophysical Union All Rights Reserved2169-927513101002jgrc20077

990

JOURNAL OF GEOPHYSICAL RESEARCH OCEANS VOL 118 990ndash1002 doi101002jgrc20077 2013

major goals a reduction of the hypoxic zone area to 5000km2 or less by 2015 through voluntary actions [HypoxiaTask Force 2008]mdasha goal that is unlikely to be met Asillustrated by the Task Force Action Plan scientific under-standing about the mechanisms underlying the generationand maintenance of the hypoxic zone on the TX-LA shelfis influencing decision making at various state and federallevels Decisions should be informed by a comprehensiveunderstanding of the processes influencing the occurrenceof hypoxia to devise effective nutrient reduction strategiesand to build reasonable expectations as to what effects mightbe expected in response to nutrient reductions[5] High-resolution physicalndashbiogeochemical models are

indispensable tools for building a mechanistic understandingof cause-and-effect relationships The objectives of thisstudy are to present such a model evaluate its deterministicsimulations of hypoxic conditions on the TX-LA shelfagainst observations and to assess the sensitivity of themodelrsquos simulated hypoxia to choices that have to be madeduring model development We systematically explore theeffects of three choices all of which could potentially affecthypoxia simulations namely the vertical resolution the re-alism in horizontal boundary conditions and the treatmentof oxygen consumption at the sediment-water interfaceWe use an ecosystem model that has been shown to realisti-cally reproduce many observed features of nutrient andplankton dynamics on the TX-LA shelf [Fennel et al2011] and that has been used successfully before to studythe dynamics of another dissolved gas inorganic carbonon the northeastern North American shelf [Fennel andWilkin 2009 Previdi et al 2009 Fennel 2010] In a recentstudy by Laurent et al [2012] the model was extended toinclude phosphate as additional nutrient however theexperiments discussed here all use the nitrogen-based modelwithout phosphate Here the model was extended to explic-itly describe the dynamics of dissolved oxygen includingairndashsea gas exchange photosynthetic production of oxygenoxygen consumption in the water column due to respirationand nitrification and sediment oxygen consumption Differ-ent treatments of the latter are explored[6] The manuscript is organized as follows in section 2

we first describe the physical model including the differentconfigurations for horizontal boundary conditions (climato-logical observations versus output from operational models)then the biological model component and parameterizations

of oxygen sources and sinks In section 3 we discuss modelresults focusing initially on the areal extent of hypoxicconditions and its sensitivity to the choice of sediment oxy-gen consumption (SOC) parameterization and horizontalboundary treatment (section 31) then we discuss the verti-cal structure of hypoxic conditions (section 32) the role ofvertical stratification (section 33) and finally assess whetherhypoxic conditions are simulated in the right locations(section 34) Our main conclusions (section 4) are that hyp-oxia simulations on the TX-LA shelf are very sensitive to thetreatment of SOC (a direct result of the fact that hypoxia islimited to a relatively thin layer above the bottom) that thestrength of vertical stratification is an important predictorof hypoxia and that modification of physical horizontalboundary conditions can lead to dramatic differences in hyp-oxia simulations because vertical stratification strength canbe affected by such a change

2 Model description

21 Physical model

[7] We used several configurations of the Regional OceanModeling System (ROMS httpmyromsorg [Haidvogelet al 2008]) for the MississippiAtchafalaya outflowregion These configurations differ in terms of vertical reso-lution (20 or 30 vertical layers) and in the treatment ofhorizontal boundary conditions (climatological or more real-istic conditions from operational Gulf of Mexico models)but all use the same horizontal grid (Figure 1) The modelsrsquoterrain-following vertical layers are stretched to result inincreased resolution near the surface and bottom The hori-zontal resolution is highest near the Mississippi Delta withup to 1 km and lowest in the southwestern corner with~20 km The model uses a fourth-order horizontal advectionscheme for tracers a third-order upwind advection schemefor momentum and the turbulence closure scheme for verti-cal mixing by Mellor and Yamada [1982] The simulationperiod is from 1 January 2004 to 31 December 2007[8] In simulations with climatological boundary treatment

(henceforth referred to simply as climatological simula-tions) radiation conditions [Flather 1976] are imposed forthe three-dimensional velocities with no mean barotropicflow Temperature and salinity at the boundary are relaxedto a horizontally uniform monthly climatology with a time-scale of 10 days for outgoing and 1 day for incoming

88 W

28 N

29 N

30 N Atchafalaya Bay

MississippiRiver Delta

10 m

20 m

30 m

40 m50 m

100 m200 m300 m

Miss R

Atch R

Sabine R

Texas Louisiana

95 W 89 W90 W91 W92 W93 W94 W

Figure 1 Model domain grid (light gray lines) and selected isobaths (dark gray lines)

FENNEL ET AL SENSITIVITY OF HYPOXIA PREDICTIONS

991

information All climatological simulations were initializedwith an average profile of temperature and salinity (basedon historical hydrographic data and assumed to be horizon-tally uniform) and spun up for at least 1 year This configu-ration was used previously in the studies by Hetland andDiMarco [2008 2012] and Fennel et al [2011][9] In the configurations with more realistic boundary

treatment initial and daily boundary conditions were takenfrom two Gulf of Mexico operational models which are sub-sequently referred to as parent models A nudging regionwith a width of 6 grid cells was implemented along the threeopen boundaries in which model temperature salinity andbaroclinic velocities are relaxed toward those of the parentmodels The nudging time scale was 8 hours at the bound-aries decaying to 0 inside the nudging layer At the bound-aries radiation conditions are used for tracers and baroclinicvelocities and sea surface height and barotropic currentsfrom the parent models are imposed[10] The two parent models are 1) the Gulf of Mexico op-

erational hybrid coordinate ocean model (HYCOM) and 2)the Intra Americas Sea Nowcast Forecast System (IASNFS)HYCOM (httpwwwhycomorg) [Wallcraft et al 2009]has a horizontal resolution of ~4 km and 20 vertical layersand assimilates observations from several sources includingsatellite altimetry satellite and in situ temperatures and ver-tical temperature and salinity profiles from XBTs and ARGObuoys The IASNFS model [Ko et al 2008] has a horizontalresolution of ~6 km 41 vertical levels and assimilates sim-ilar observations as HYCOM More details on the off-linenesting procedure and the parent models can be found inthe study by Marta-Almeida et al (Evaluation of model nest-ing performance on the Texas-Louisina continental shelfsubmitted to Journal of Geophysical Research 2013)[11] The models are forced with 3-hourly winds from the

NCEP North American Regional Reanalysis (NARR) andclimatological surface heat and freshwater fluxes from thestudy by da Silva et al [1994a 1994b] Freshwater inputsfrom the Mississippi and Atchafalaya Rivers use daily mea-surements of transport by the US Army Corps of Engineersat Tabert Landing and Simmesport respectively[12] The climatological model realistically captures the

two distinct modes of circulation over the TX-LA shelfnamely mean offshore flow during upwelling favorablewinds in summer and mean westward (downcoast) flowduring downwelling favorable winds for the rest of the year[Hetland and DiMarco 2008] Model skill in simulatingobserved salinity distributions was quantified for the climato-logical and nested configurations in the study by Hetland andDiMarco [2012] and Marta-Almeida et al (submitted manu-script 2013) respectively

22 Biological model

[13] The biological component of our model uses thenitrogen cycle model described in the study by Fennelet al [2006] but was extended by including dissolved oxygenas a state variable and a parameterization of the airndashsea flux ofoxygen as described below The nitrogen cycle model is arelatively simple representation that includes two species ofdissolved inorganic nitrogen (nitrate NO3 and ammoniumNH4) one functional phytoplankton group Phy chlorophyllas a separate state variableChl to allow for photoacclimation

one functional zooplankton group Zoo and two pools ofdetritus representing large fast-sinking particles LDet andsuspended small particles SDet[14] The main processes described in the model are 1)

temperature light- and nutrient-dependent phytoplanktongrowth with ammonium inhibition of nitrate uptake 2)zooplankton grazing represented by a Holling-type III param-eterization 3) aggregation of phytoplankton and small detritusto fast sinking large detritus 4) photoacclimation (ie avariable ratio between phytoplankton and chlorophyll) 5)linear rates of phytoplankton mortality zooplankton basalmetabolism and detritus remineralization 6) a secondorder zooplankton mortality 7) light-dependent nitrifica-tion (ie oxidation of ammonium to nitrate) and 8)vertical sinking of phytoplankton and detritus The modelis shown schematically in Figure 2 For further details onmodel justification equations and parameters we refer thereader to Fennel et al [2006 2008][15] Here we only report the new model equation describ-

ing the biochemical dynamics of oxygen Ox as follows

Ox

tfrac14 mmax f Ieth THORN LNO3R02NO3 thorn LNH4RO2NH4eth THORNPhy 2nNH4 RO2NH4 lZoothorn rSDSDet thorn rLDLDeteth THORN

(1)

where mmax is the maximum growth rate of phytoplanktonf (I) is a nondimensional light-limitation term LNO3 and LNH4correspond to nutrient-limitation due to nitrate and ammo-nium respectively RO2NO3 = 13816 mol O2mol NO3 andRO2 NH4 = 10616 mol O2mol NH4 are stoichiometric ratioscorresponding to the oxygen produced per mol of nitrateand ammonium assimilated during photosynthetic produc-tion of organic matter n is the nitrification flux l is thezooplankton excretion rate and rSD and rLD are the reminer-alization rates of small and large detritus respectively Thefirst term on the right-hand side of equation (1) correspondsto the production of oxygen during photosynthesis Whilephytoplankton growth is limited by the total available dis-solved inorganic nitrogen (ie the sum of nitrate and ammo-nium) the corresponding oxygen gain differs depending onwhich nitrogen species supports photosynthesis hence thedistinction between LNO3 and LNH4 in the first right-hand sideterm of equation (1) The second term represents the

Figure 2 Schematic of the biological model


992

consumption of oxygen during oxidation of ammonium to ni-trate (2 mol of O2 are consumed per mol of ammonium oxi-dized) and the last term corresponds to oxygen sinks dueto respiration by zooplankton and heterotrophic bacteria thatdegrade detritus The hats in equation (1) mark terms thathave been modified compared to Fennel et al [2006] to ac-count for low oxygen concentrations Specifically

r frac14 r maxOx OxthkoxthornOxOxth

0

(2)

where kox= 3 mmol O2 m3 is an oxygen half-saturationconcentration and Oxth = 6 mmol O2 m

3 an oxygen thresh-old below which no aerobic respiration or nitrificationoccurs The same formulation is used for remineralizationof large and small detritus and for nitrification[16] In addition to the biochemical sources and sinks of

oxygen there is gas exchange across the airndashsea interfacewhich directly affects the oxygen concentration in the toplayer of the model and is parameterized as

F frac14 vkO2Δz

Oxsat Oxeth THORN (3)

where F (in units of mmol O2 m3) is the flux of oxygen into

the top layer vkO2 is the gas exchange coefficient for oxy-gen Δz is the thickness of the respective grid box and Oxsatis the saturation concentration of oxygen The gas exchangecoefficient is parameterized followingWanninkhof [1992] as

vkO2 frac14 031u210

ffiffiffiffiffiffiffiffiffi660

ScOx

r (4)

[17] Here u10 is the wind speed 10 m above the sea sur-face and ScOx is the Schmidt number which we calculateas in the study by Wanninkhof [1992] Oxsat is based onthe study by Garcia and Gordon [1992][18] In combination with the freshwater discharge

described in section 21 the model receives inorganic andorganic nutrients specifically nitrate ammonium andparticulate nitrogen which are based on monthly nutrientflux estimates from the US Geological Survey [Aulenbachet al 2007] Particulate organic nitrogen fluxes are deter-mined as the difference between total Kjeldahl nitrogenand ammonium[19] Three different parameterizations of sediment oxygen

consumption (SOC) have been implemented and are com-pared in this study

A The option for ldquoinstantaneous remineralizationrdquo (re-ferred to as IR) relates the flux of organic matter intothe sediment to a corresponding efflux of ammoniumIn IR all organic matter that reaches the sediment is in-stantaneously remineralized accounting for the loss offixed nitrogen due to denitrification Key assumptionsunderlying this parameterization are that denitrificationand SOC are linearly related according to the

relationship presented by Seitzinger and Giblin [1996]that oxygen is consumed in the sediments only by nitri-fication and aerobic remineralization and that all or-ganic matter reaching the sediments is remineralizedimmediately (bottom water concentrations are updatedaccordingly in the same time step) The detailed deriva-tion is given in the study by Fennel et al [2006] Herethe parameterization was extended to include thecorresponding oxygen sink Because this parameteriza-tion accounts for denitrification (ie the production ofbiologically unavailable N2) it represents a sink of bio-available nitrogen in the system An important limitationof this parameterization is that it neglects temporaldelays in SOC which occur in nature and would resultin smaller SOC at the height of blooms and largerSOC after bloom events in late summer and fall and fur-ther downstream from nutrient sources

B We also implemented the SOC option of Hetland andDiMarco [2008] (henceforth referred to as HampD) whichwas parameterized based on sediment flux observationsby Rowe et al [2002] HampD parameterizes oxygenuptake by the sediments dependent on bottom wateroxygen concentration and temperature as

FHampDSOC T Oxeth THORN frac14 6 mmol O2 m2 d1

2T=10

C

1 exp Ox

30 mmol O2 m3frac12

(5)

SOC is assumed to increase linearly with increasing bot-tom water oxygen for concentrations below ~50 mmolO2 m

-3 and to saturate above ~100 mmol O2 m3 Tem-

perature dependence follows the Q10-rule with SOCdoubling for each temperature increase of 10C Herethe parameterization was extended to include a flux ofammonium into the bottom water that is proportionalto the SOC flux

FNH4 frac14 rNH4SOCFSOC 6eth THORN

with rNH4 SOC = 0036 mol NH4 per mol O2 The deri-vation of rNH4 SOC = is given in Appendix 5 With thisparameterization organic matter essentially leaves thesystem when sinking out of the water column but a frac-tion of the nitrogen that sank out is returned to the watercolumn as ammonium (at the rate dictated by the param-eterization) This approach is more realistic than the IRparameterization in the sense that it predicts largersediment oxygen consumption when water temperaturesare highest and microbial decomposition of sedimentaryorganic matter is accelerated (ie in summer) IRpredicts the highest rate of oxygen consumption when theorganic sinking flux is largest (ie in spring) Howeveran important limitation of the HampD parameterization is thatit does not account for any direct effect of changes in nutri-ent load on SOC essentially disconnecting the two Thesame limitation applies to the next parameterization

C The third SOC option is based on the linear relationshipbetween SOC and bottom water oxygen suggested byMurrell and Lehrter [2011] (henceforth referred to asMampL) and is based on their sediment flux


993

measurements The MampL relationship was modifiedslightly here by forcing the linear fit to be zero for zerobottom water oxygen and by including temperature de-pendence following the Q10-rule resulting in thefollowing

FMampLO2 T Oxeth THORN frac14 00235 mfrac12 2T=10

COx mmol O2 m3

(7)

[20] The SOC flux observations from Murrell and Lehrter[2011] are smaller than those observed by Rowe et al[2002] leading to generally smaller SOC fluxes with theMampL option compared to HampD (Figure 3) Ammoniumfluxes are determined according to equation (6) as for HampD[21] The three SOC parameterizations result in an oxygen

drawdown in the bottom-most grid cells (simulating the ef-fect of oxygen diffusing into the sediments) and are separatefrom and in addition to oxygen drawdown in the water thatresults from remineralization of detritus nitrification etc[22] The biological variables NH4 Phy Chl Zoo Sdet

and LDet were initialized with small constant values whileNO3 and Ox were initialized with horizontally homogenousmean winter profiles based on available in situ data All bio-logical simulations were spun up for 1 year At the openboundaries NO3 and Ox were prescribed using the NODC

World Ocean Atlas All other biological state variables atthe boundary were set to small positive values[23] We present results from 4 years of simulation from 1

January 2004 to 31 December 2007 An overview of all thesimulations discussed here is given in Table 1

3 Results and discussion

31 Spatial extent of simulated hypoxic conditions

[24] The spatial extent of hypoxic conditions has beenused historically as an important metric for the TX-LA shelffor example in monitoring interannual variability of hyp-oxia [Rabalais et al 2002] as dependent variable in statis-tical hypoxia models [Greene et al 2009 Forrest et al2011 Feng et al 2012] and as target for nutrient reductionstrategies put forth by the Hypoxia Task Force [2008] Sim-ulated hypoxic areal extent (temporal means and ranges) forthe end of July (coincident with the hypoxia monitoringcruises of Rabalais et al [2002]) are given in Table 2 andtime series from selected simulations are shown in compari-son to the observed spatial extent in Figures 4 and 5 andFigure S1 in the Online Supplement[25] Of the three simulations with different treatments of

sediment oxygen consumption (SOC) but the same climato-logical boundary conditions and 20 vertical layers simulationB20clim (with HampD SOC parameterization) simulates theobserved hypoxic extent best and agrees with the observedextent in all 4 years (Figure 4) Simulation A20clim (withIR) simulates hypoxic extent equally well in 2004 but under-estimates the observed extent in the other 3 years Almost nohypoxic conditions are produced in simulation C20clim (withMampL SOC parameterization)[26] No systematic differences in simulated hypoxic extent

result from changes in vertical resolution (ie an increase invertical resolution from 20 to 30 layers) in the climatologicalruns This is illustrated for two different SOC parameteriza-tions (IR and HampD) in Figure S1 (see Online Supplement)[27] Changes in the boundary conditions however lead to

one systematic difference namely a doubling in hypoxicextent for simulation B30IAS (with IASNFS boundary con-ditions Figure 5) The hypoxic extent simulated in B30HYC(with HYCOM boundaries) is very similar to that in theclimatological simulation B30clim[28] It should be noted that uncertainties in model para-

meters initial and boundary conditions and forcing (whichcan only be known approximately) can lead to amplifieduncertainties in hypoxia simulations (Mattern P et alUncertainty in hypoxia predictions for the TX-LA shelfsubmitted to Journal of Geophysical Research 2013) (seeFigure S2 for an example of uncertainty in hypoxic extent

Figure 3 Parameterizations of sediment oxygen consump-tion (SOC) of Hetland and DiMarco [2008] (black line) andMurrell and Lehrter [2011] (red line) at 25C and observa-tions of Rowe et al [2002] (black dots) and Murrell andLehrter [2011] (red dots)

Table 1 Overview of Model Simulations

Run SOC Treatment Vertical Resolution Horizontal Boundary Treatment

A20clim instantaneous remineralization 20 layers climatologicalB20clim HampD SOC parameterization 20 layers climatologicalC20clim MampL SOC parameterization 20 layers climatologicalA30clim instantaneous remineralization 30 layers climatologicalB30clim HampD SOC parameterization 30 layers climatologicalA30HYC instantaneous remineralization 30 layers HYCOMB30HYC HampD SOC parameterization 30 layers HYCOMA30IAS instantaneous remineralization 30 layers IASNFSB30IAS HampD SOC parameterization 30 layers IASNFS


994

Table 2 Observed and Simulated Hypoxic Area at the End of July (in 103 km2)a

2004 2005 2006 2007

Observed 110 101 94 153Min Mean Max Min Mean Max Min Mean Max Min Mean Max

A20clim 105 129 143 16 28 39 19 33 59 57 70 93B20clim 108 182 246 24 56 118 43 80 143 78 113 177C20clim 04 15 28 01 02 04 00 02 05 02 06 12A30clim 130 152 163 22 34 50 13 26 48 40 50 74B30clim 108 201 275 27 73 147 50 86 144 77 107 168A30HYC 29 63 97 01 06 11 04 13 26 16 22 33B30HYC 144 213 247 39 80 135 112 158 215 113 147 186A30IAS 60 110 152 06 16 31 05 19 38 24 35 51B30IAS 295 373 403 159 230 331 266 329 410 257 307 357

aMinimum mean and maximum extent of hypoxic area during the July monitoring cruises are reported for the simulations The area was estimated bylinearly interpolating the observed oxygen concentrations onto the model grid with the help of Matlabrsquos grid data function and then calculating the area withoxygen concentrations below the hypoxic threshold

Jan04 Jul04 Jan05 Jul05 Jan06 Jul06 Jan07 Jul07 Jan080

5

10

15

20

25

30

observedA20climB20climC20clim

Hyp

oxic

are

a (1

03 km

2 )

Figure 4 (Middle panel) Time series of simulated hypoxic extent for the three different treatments of sed-iment oxygen consumption A20clim (blue line) B20clim (green line) and C20clim (red line) and observedextent in late July (black dots) See Table 1 for acronyms Also shown are the observed bottom oxygen con-centrations (colored dots) and the inferred hypoxic area (red area) for the years 2004 (top left) 2005 (bottomleft) 2006 (top right) and 2007 (bottom right) The observed bottom oxygen concentrations (colored dots)are in units of mmol O 2 m

3 and correspond to the colour scale shown in the top left panel

observedB30climB30HYCB30IAS

0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )

Figure 5 Time series of simulated hypoxic extent with the HampD SOC parameterization for the three dif-ferent boundary treatments B30clim (blue line) B30HYC (magenta line) and B30IAS (green line) andobserved extent in late July (black dots)


995

resulting from an assumed 20 error in freshwaterdischarge) Of course the estimate of the observed hypoxicextent although represented without error bars is based onnecessarily incomplete station observations and containserrors as well (we estimated the hypoxic extent here bylinearly interpolating the observed bottom oxygen concentra-tions onto the model grid with Matlabrsquos grid data functionthen summing the area of all grid cells that are hypoxicaccording to the interpolated observations) for examplethe 2005 extrapolation on the inner shelf side of the hypoxicarea at the mouth of Atchafalaya Bay may be an overesti-mate Another important limitation of using the spatialextent of hypoxia as metric for comparisons between modeland observations is that a model may well simulate the righthypoxic extent but in the wrong location hence for the wrongreasons[29] Also obvious in Figure 4 is that large temporal

changes in simulated hypoxic extent can occur over shorttime scales (a few days) and that there are systematic differ-ences in simulated hypoxic extent between the differentformulations on longer (monthly) time scales Evaluationof these features would require highly resolved time seriesmeasurements to address variability on short time scales aswell as more than one shelf-wide cruise per year to addressthe monthly evolution of spatial hypoxic extent[30] The results discussed so far raise three questions

which will be considered next[31] 1 Why are simulated hypoxia so sensitive to the

treatment of sediment oxygen consumption[32] 2 Why is hypoxic extent roughly doubled when the

IASFNS boundary conditions are used[33] 3 Are hypoxic conditions simulated in the right

locations

32 Importance of sediment oxygen consumption forthe generation of hypoxic conditions

[34] On the TX-LA shelf hypoxic conditions are oftenrestricted to below a secondary pycnocline well below themain pycnocline an observation that was first made byWiseman et al [1997] Our model simulations show thesame behavior For example the median thickness of thehypoxic layer in simulation B20clim is 26 m (Figure 6)while the median thickness of the layer below the main pyc-nocline is much larger with 84 m during hypoxic eventsWe defined the main pycnocline as the depth where densityfirst exceeds the density at the surface plus 01 st -units andonly included days and horizontal grid cells where hypoxicconditions occurred The median thickness of the hypoxiclayer differs between simulations (Table 3) but is typicallyless than half of the median thickness of the layer belowthe main pycnocline We defined the bottom boundary layeras the layer in which density does not exceed the density atthe bottom plus 05 st -units In B20clim this bottom bound-ary layer has a median thickness of 27 m during hypoxicevents (Figure 6) and coincides with the hypoxic layerThe hypoxic layer and bottom boundary concur remarkablywell in all simulations (Table 3)[35] The fact that hypoxic conditions are typically re-

stricted to within 2ndash3 m above the sediment suggests thatSOC is an important contributing factor to hypoxic bottomwaters Kemp et al [1992] suggested based on a compila-tion of measured water column and sediment respiration

rates from a number of coastal environments that the impor-tance of sediment respiration is inversely related to waterdepth Water column respiration refers here to the consump-tion of oxygen due to microbial respiration measured in asample of water while SOC or sediment respiration refersto the consumption of oxygen by processes in the sedimentie the flux of oxygen into the sediments Kemp et al[1992] derived an empirical relationship for the SOC frac-tion of total respiration Based on this relationship they pre-dicted that water column oxygen sinks should exceed those

0 5 10 15 20

Thickness of layer below the bottom pycnocline (m)

Pro

babi

lity

dist

ribut

ion

mode = 12 mmedian = 27 m25minustile = 14 m75minustile = 46 m

0 5 10 15 20

Thickness of hypoxic layer above bottom (m)P

roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01

Thickness of layer below the main pycnocline (m)

Pro

babi

lity

dist

ribut

ion


Figure 6 Probability distributions of simulated thick-nesses of layer below the main pycnocline (top) layer belowthe oxycline of 63 mmol O2 m

3 (middle) and bottom pyc-nocline (bottom) for simulation B20clim Includes days andhorizontal grid cells where hypoxic conditions occurredMain pycnocline depth was defined as depth at which den-sity first exceeds that at the surface plus 01 st -units Bottompycnocline was defined as height from bottom at whichdensity first exceeds that at the bottom plus 05 st -unitsMode and median are shown by solid green and red linesrespectively and 25th and 75th percentiles are shown bydashed red lines


996

of the sediment for systems deeper than 5 m By relatingtheir observed water column respiration rate in the layer be-low the main pycnocline to sediment oxygen consumptionMurrell and Lehrter [2011] suggested that sediments onthe TX-LA shelf account for 22 of total oxygen consump-tionmdashan estimate that is consistent with the empirical rela-tionship of Kemp et al [1992] We would argue that thethickness of the bottom boundary layer (ie the layer thatactually turns hypoxic) should be applied instead of thethickness of the layer below the main pycnocline a pointthat has already been made by Hetland and DiMarco[2008] When applying the water column respiration ratesof Murrell and Lehrter [2011] to a 2 m thick layer of waterabove the bottom the SOC fraction of total respiration risesto 46 When applying the water respiration rates ofMurrell and Lehrter [2011] and the sediment respirationrates of Rowe et al [2002] the SOC fraction rises to 60Both of these estimates are also consistent with the empiricalrelationship of Kemp et al [1992][36] It should be noted that a main pycnocline is a prereq-

uisite for the existence of a secondary pycnocline as dis-cussed by Hetland and DiMarco [2008] Overall stratifica-tion of the water column thus continues to be a usefulindex for hypoxia and will be discussed next

33 Importance of Stratification for the Generation ofHypoxic Conditions

[37] Vertical stratification of the water column is generallythought to be a prerequisite for the occurrence of hypoxia incoastal systems as stratification is a barrier to vertical oxygensupply from the surface We analyzed the correlationbetween bottom water oxygen concentrations and stratifica-tion in our simulations As a measure of stratification wecalculated the potential energy anomaly (f J mndash3 ) definedby Simpson [1981] as

f frac14 1

h

Z 0

hr reth THORNgzdz with r frac14 1

h

Z 0

hrdz (8)

where h is the depth of the water column r is density g isthe gravitational acceleration and z is the vertical coordinateThe value of f is equivalent to the amount of work requiredper unit volume to vertically homogenize the water columnthrough mixing [Simpson 1981] f has been widely used as

a measure for stratification strength (see the study byBurchardand Hofmeister [2008] and references therein)[38] Time series of stratification index f and bottom oxy-

gen concentration are shown in Figure 7 for a representativestation on the shelf for each of the four simulated years ofB20clim As expected f and bottom oxygen are anticorre-lated meaning that stronger stratification corresponds tolower bottom oxygen The correlations are strong with abso-lute values of r larger than 070 in all years except for 2006when absolute correlation is 045 Absolute correlation coef-ficients are even higher for May to August ie the period

Table 3 Median (Med) Mode (Mod) 25th Percentile (25-pct) and 75th Percentile (75-pct) of Simulated Thickness of the Layer Belowthe Main Pycnocline Thickness of the Hypoxic Layer and Thickness of the Bottom Boundary Layera

Thickness of Main Pycnocline Thickness of Hypoxic Layer Thickness of Bottom Layer

25-pct med mod 75-pct 25-pct med mod 75-pct 25-pct med mod 75-pctA20clim 51 75 59 140 18 30 20 48 12 21 10 46B20clim 56 84 59 137 16 26 20 41 14 27 12 46C20clim 49 58 59 71 07 12 08 18 07 11 08 16A30clim 40 66 38 125 18 34 09 60 10 19 09 46B30clim 48 80 38 136 18 32 19 51 14 28 10 52A30HYC 36 50 38 71 11 19 10 32 08 13 07 22B30HYC 53 83 60 149 22 39 24 58 16 31 09 54A30IAS 39 56 39 89 12 22 12 35 09 16 09 28B30IAS 65 119 61 192 28 43 35 63 22 40 23 64

aThe upper limit of the latter is defined as the depth where density equals the density at the bottom plus 05 st units All numbers are in units of m Thecalculations only include days and horizontal grid cells where hypoxia was encountered Histograms of the underlying distributions are shown for B20climin Figure 6

Apr 04 Jul 04 Oct 04

corr(yr) = minus073corr(MayminusAug) = minus079









0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )

Figure 7 Simulated time series of potential energy anom-aly (f blue line) and bottom oxygen concentration (greenline) for a representative station on the TX-LA shelf (stationis indicated by a magenta dot in Figure 7) for the years2004ndash2007 for simulation B20clim Time series was low-pass filtered with a filter window of 1 week to remove timescales shorter than 1 week The correlation coefficient be-tween both variables over the whole year and for the monthsMay to August is given as well


997

during which hypoxia forms except for 2007 where theabsolute value of the annual correlation coefficient is 078while the May to August value is 073 This strongcorrelation between stratification index and bottom oxygenconcentration is not an anomaly as can be seen in the map

in Figure 8 and the histogram in Figure 9 The map in Fig-ure 8 illustrates that the two variables are strongly correlatedover most of the shelf except for the shallow nearshore areawest of Atchafalaya Bay Most of the monitoring stationsthat were found to be hypoxic at least once between 2004and 2007 (shown as black dots in Figure 8) have absolutecorrelation coefficients of 050 and higher Absolute valuesof mode and median of the correlation coefficients are 060and 063 respectively (Figure 9) Strong correlations arefound for all simulations (Table 4)[39] The question arises whether the differences in areal

extent of hypoxic conditions simulated by B30HYC andB30IAS result from differences in strength of stratificationIt should be noted that the biological model configurationis identical in both simulations (ie the same biological ini-tial and boundary conditions and parameterizations areused) The only difference between B30HYC and B30IASis in their physical boundary conditions In Figure 10 timeseries of f and bottom oxygen are shown for a shelf stationclose to the western boundary of the domain for both simu-lations The stratification index is significantly higher forB30IAS while bottom oxygen concentrations are signifi-cantly lower and at hypoxic levels for extended periods ev-ery summer The stratification index in B30HYC is muchlower and bottom oxygen concentrations are higher reach-ing hypoxic levels only occasionally[40] Thus the significant difference in simulated hypoxic

conditions between B30HYC and B30IAS (a doubling ofhypoxic extent in the latter) must be due to differences inthe physical boundary conditions with the latter model pro-ducing stronger vertical stratification because of its bound-ary conditions It is noteworthy that this difference was notdiagnosed in the salinity skill assessment of Marta-Almeidaet al (submitted manuscript 2013) who compared the skillof both models in simulating observed salinity profiles andfound that skill scores were practically indistinguishable

Figure 8 Spatial distribution of temporal correlations be-tween stratification index f and bottom oxygen concentra-tion over the whole simulation period for B20clim Alsoshown are stations (black dots) that had hypoxic bottomwaters during at least one of the July monitoring cruisesbetween 2004 and 2007 The magenta dot indicates the sitefor which time series are shown in Figure 6

minus08 minus06 minus04 minus02 0 020

1

2

3

4

5

Temporal correlation between φ and bottom oxygen concentration

Pro

babi

lity

dist

ribut

ion

mode = minus066median = minus06325minustile = minus06975minustile = minus053

Figure 9 Probability distribution of simulated temporalcorrelation coefficients between f and bottom oxygen forB20clim Mode and median are indicated by green and redlines 25th and 75th percentiles by dashed red lines

Table 4 Median Mode 25th Percentile (25-pct) and 75thPercentile (75-pct) of Temporal Correlation Between StratificationIndex f and Bottom Oxygen Concentration in Percenta

Temporal Correlation

25-pct Median Mode 75-pctA20clim -71 -65 -63 -61B20clim -69 -63 -66 -53C20clim -68 -62 -62 -51A30clim -73 -66 -69 -59B30clim -71 -66 -70 -58A30HYC -75 -70 -73 -61B30HYC -73 -67 -72 -55A30IAS -74 -70 -74 -62B30IAS -76 -70 -76 -58

aHistogram of the underlying distribution is shown for B20clim in Figure 9

Jan 04 Jan 05 Jan 06 Jan 07

B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)

Figure 10 Time series of stratification index (top panel)and bottom oxygen concentration (bottom panel) at a shelfstation close to the western boundary of the model domainfor simulations B30HYC (blue lines) and B30IAS (redlines) Oxygen concentrations below the hypoxic thresholdof 63 mmol O2 m

3 are highlighted in magenta


998

One possible explanation is that model skill is similar forboth models east of 93W where most of the observed pro-files are located while stratification (and thus simulated hyp-oxic conditions) differ most strongly west of 93W

34 Spatial Comparison of Simulated HypoxicConditions

[41] The simulated hypoxic area during the hypoxiamonitoring cruise in late July of 2004 and the correspondinghypoxia observations are shown in Figure 11 for all simula-tions When comparing the different treatments of the sedi-ment oxygen sink for climatological boundary conditions(top row in Figure 11) it is obvious that treatment A(IR) generates hypoxia closer to the river freshwater sources(ie near the Mississippi Delta and Atchafalaya Bay) whiletreatment B (HampD) generates hypoxic conditions also atsome of the deeper shelf stations Treatment C (MampL) simu-lates almost no hypoxia[42] To assess the effects of the different vertical resolu-

tions and different boundary conditions one can compareall the A treatments and all the B treatments In both casesA and B the increase in vertical resolution does not leadto qualitative changes although the shape of the hypoxicarea differs slightly between simulations Slight changesare to be expected because of the turbulent nature of the circu-lation on the shelf and the resulting uncertainty in distributionfeatures for small perturbations to the model (Mattern Pet al submitted manuscript 2013)[43] In the A treatment switching from climatological to

more realistic physical boundary conditions (ie HYCOMor IASNFS) decreases the simulated hypoxic area while inthe B treatment hypoxic area increases for the more realisticboundary conditions The decrease in treatment A is a directconsequence of the fact that more organic matter is trans-ported off the shelf in the simulations with realistic horizon-tal boundary conditions and a more dynamic circulation (seeFigures S3 and S4 in Online Supplement) In simulation

A30clim more organic matter is remineralized on the shelfleading to more oxygen consumption and thus a larger hyp-oxic region Hypoxia simulations with the B treatment aremore sensitive to stratification on the shelf and the increasein hypoxic area for the more realistic boundary conditionsis due to an increase in stratification over the outer shelfThe most pronounced difference however is the increasein hypoxic area in the B treatment when switching from cli-matological or HYCOM boundaries to IASNFS In the pre-vious section we have shown that this increase is due to anincrease in stratification west of 93W This is reflected inFigure 11 by the westward expansion of the hypoxic area(shown in red) for B30IAS compared to B30HYC[44] The same qualitative patterns with regard to SOC treat-

ments vertical resolution and boundary conditions hold forthe other 3 years (see Figures S5ndashS7 in the Online Supplement)[45] Quantitative comparisons of simulated and observed

hypoxic conditions for all 4 years and all simulations arepresented in Figures 12 and 13 Figure 12 shows the fractionof the total number of sampling stations for which presenceor absence of hypoxia was correctly simulated (ie the frac-tion of stations with positivendashpositive and negativendashnegativehits to total number of stations) The bars show this measureusing model output for the date half-way through theapproximately week-long cruise (same as in Figure 11)However the number of hypoxic stations can change withina few days (see eg Figure 4) To take this variability intoaccount the fraction of stations simulated correctly was alsocalculated using model output from the first and last day ofthe cruise the range is given by the error bars in Figure 12To help gauge the metric of ldquofraction of stations predictedcorrectlyrdquo we also show how the two extreme predictionsthat either ldquoall stations are hypoxicrdquo or ldquono station is hyp-oxicrdquo would fare This is meant to be analogous to compar-ing for example the simulated time series of a variable liketemperature or chlorophyll concentration against a climatol-ogy of the same variable to test whether the model is able to

Figure 11 Simulated hypoxic area (red) and sampling stations (open and filled circles) from hypoxia mon-itoring cruise for late July of 2004 Filled circles indicate that hypoxic conditions were encountered duringsampling The simulated area is shown for the half-way date of the cruise which typically lasts 5 days


999

make better predictions than the climatology Here we areinterested whether the model is able to produce better resultsthan would be obtained by predicting that either all stationsor no station are hypoxic For the case that ldquoall stations arehypoxicrdquo the metric is simply the fraction of hypoxicstations out of the total number of stations sampledConversely for the case of ldquono station is hypoxicrdquo it is thefraction of stations that are not hypoxic

[46] We would like to note that not all stations are sampledeach year often cross-shelf transects are terminated whenoxygenated bottom waters are encountered The fraction ofstations that are hypoxic in a given year is thus a biasedmeasure We would also like to point out that the fractionof stations simulated correctly by pure chance can be calcu-lated as pn 100 where p is the probability of a stationbeing hypoxic or not and n is the number of stations Thisnumber is very small well below 1 for 10 stations or morewhile the total number of stations during one monitoringcruise is on the order of 100[47] Figure 12 shows that no model outcome is ever worse

than the worst of the extreme cases ldquoall hypoxicrdquo and ldquonohypoxiardquo However no model is consistently better thanthe better of the two extreme cases In 2004 the climatolog-ical simulations with SOC treatment A simulate about 65of the stations correctly more than the better of the extremecases ldquoall hypoxicrdquo In 2005 all models are on par with thebetter of the extreme case (ldquono hypoxiardquo) at about 63 ex-cept for B30HYC and B30IAS In 2006 simulationB30HYC is notably better than all others and the extremecases In 2007 only the B treatments (except B30HYC) dobetter than the better extreme case of ldquoall hypoxicrdquo[48] Figure 13 shows the normalized RMSE of bottom wa-

ter oxygen between the simulations and observations In theclimatological simulations the error is consistently higher forthe A treatment but this difference disappears in the nestedsimulations where errors tend to be smaller and the A and Btreatments often have similar errors Another systematic differ-ence is that among the nested simulations those withHYCOM boundary conditions tend to have slightly largererrors than those with IASNFS boundaries This is despitethe fact that B30IAS simulates an unrealistically large spatialextent of hypoxia (twice the size simulated by B30HYC)

2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct

A20climB20climC20climA30climB30climA30HYCB30HYCA30IASB30IASall hypoxicnone hypoxic

Figure 12 Percentage of sampling stations predicted correctly as hypoxic or not hypoxic during thehypoxia monitoring cruises from 2004 to 2007 The colored bars use model output corresponding to thehalf-way date of the cruise The error bars give the range when using model output for the first or lastday of the cruise and can be considered as an estimate of uncertainty Also given are the fractions of sta-tions that are hypoxic and not hypoxic as they represent the fraction of stations that would be correctlypredicted for the extreme predictions of ldquoall stations are hypoxicrdquo and ldquono station are hypoxicrdquorespectively

2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE

A20climB20climC20climA30climB30climA30HYCB30HYCA30IASB30IAS

Figure 13 Root mean square error (RMSE) of dissolvedoxygen between simulations and measurements from hyp-oxia monitoring cruises from 2004 to 2007 The colored barsuse model output corresponding to the half-way date of thecruise The error bars give the range when using model out-put for the first or last day of the cruise and can be consid-ered as an estimate of uncertainty


1000

[49] According to these metrics none of the SOC orboundary treatments consistently outperforms others It isnoteworthy that the treatment with IASNFS boundaries per-forms as good as or better than other models according toboth metrics despite its propensity to overpredict hypoxicconditions west of 93W This shows that reliance on alimited set of skill metrics can be problematic

4 Conclusions

[50] Hypoxia simulations for the TX-LA shelf are verysensitive to the size of the sediment oxygen sink Model-simulated hypoxic extent for the three parameterizations thatwere tested here (instantaneous remineralization a parame-terization by Hetland and DiMarco [2008] and one byMurrell and Lehrter [2011]) varied significantly Of thesethree parameterizations the one by Hetland and DiMarco[2008] performed best in simulating the observed hypoxicextent while instantaneous remineralization slightly under-estimated the observed extent and the parameterization ofMurrell and Lehrter [2011] led to almost no hypoxia Wewould like to emphasize that the parameterizations ofHetland and DiMarco [2008] and of Murrell and Lehrter[2011] were used here to study model sensitivity to SOCmodel resolution and physical forcing Neither of theseparameterizations accounts for the effects of changing nutri-ent loads on SOC thus neither would be appropriate forscenario simulations with varying nutrient loads[51] On the TX-LA shelf hypoxic conditions are restricted

to a relatively thin layer above the sediment unlike for exam-ple Chesapeake Bay where hypoxic conditions occur in thebottom waters of a relatively deep channel with restricted cir-culation and extend tens of meters above the bottom up to themain pycnocline [Pierson et al 2009] In our simulationshypoxic conditions are typically constrained to a 2ndash3 m thickbottom boundary layer which we defined operationally as thelayer where density does not exceed that at the bottom bymorethan 05 st -units A direct consequence is the sensitivity ofhypoxia simulations to the rate of sediment oxygen consump-tion because the latter is of similar magnitude as water-columnrespiration when integrated over the bottom boundary layeronly It follows that an adequate resolution of near bottomstratification and sediment oxygen consumption are crucialfor properly simulating hypoxia on the TX-LA shelf[52] On the TX-LA shelf stratification strength (expressed

by the potential energy anomaly) and bottom water oxygenconcentrations are highly correlated in time over most ofthe shelf (the exception is the shallow near-shore area westof Atchafalaya Bay where correlation coefficients tend tobe smaller than 050) Mode and median of absolute corre-lation coefficients are always larger than 060 regardlessof which sediment oxygen parameterization is usedStratification strength is thus an important determinant ofhypoxic conditions[53] Replacing the physical climatological boundary con-

ditions with more realistic conditions from HYCOM andIASNFS both data-assimilative operational forecast modelsfor the Gulf of Mexico resulted in large differences in sim-ulated hypoxic extent for IASNFS boundaries but not forHYCOM With IASNFS boundaries the hypoxic extentroughly doubles which is due to increased stratification west

of 93W These differences in stratification were not pickedup by the salinity skill assessment of Marta-Almeida et al(submitted manuscript 2013) nor were the differences inbottom oxygen concentrations picked up by our skill metrics(percent of sampling stations simulated correctly as hypoxicor not RMSE of bottom oxygen concentration) Accordingto the latter two skill metrics the configuration with IASNFSboundaries performs as good or sometimes better thanthe other configurations It is important to recognize thatconclusions about one model formulation performing betterthan others do depend on the metric used Our results hereshow that the statistical measures RMSE and ldquofractionof stations predicted correctlyrdquo are not appropriate forthe given model and available observation data setMetrics like the spatial extent of hypoxic conditions aremore useful One reason for this is that differences inhypoxia simulations appear west of 93W where almostno sampling occurred Thus we would like to cautionagainst relying on a single or a few closely related skillmetrics

Appendix A Sediment Oxygen ConsumptionParameterization

[54] Here we derive the value of rNH4 SOC which relatesthe SOC flux to its corresponding efflux of ammonium Wemade the following assumptions[55] 1 Organic matter is remineralized via two pathways

aerobic oxidation and coupled nitrification-denitrification[56] 2 Net oxygen consumption only occurs during aero-

bic carbon oxidation and nitrification[57] 3 Denitrification and sediment oxygen consumption

are linearly related according to the relationship of Seitzingerand Giblin [1996]

FDNF [mmol N m-2 d-1] = 0105 [mol N mol O2] SOC[mmol O2 m-2 d-1] (A1)(see also Fennel et al [2009] for alarger more recent compilation of observations validatingthis relationship)Let x2 [01] be the fraction of carbon oxi-dation that occurs through denitrification Then assumingthe stoichiometries as in Fennel et al [2006] for 1 mol oforganic carbon to be remineralized 106(1 x) + 1696x=106 + 636x mol O 2 are consumed The value of x can bedetermined using relationship A1 as 848x = 0105(106 + 636) hence x= 014 In other words 14 of organicmatter is denitrified and 86 is oxidized aerobically Thecorresponding oxygen consumption is 1149 mol O2

(086106 mol in aerobic remineralization and 0141696mol in coupled nitrification denitrification) Thecorresponding production of ammonium is 413 mol NH4

(08616 mol in aerobic remineralization and014848 + 01416 =963 mol in coupled nitrifica-tionndashdenitrification) Hence for 1 mol O2 consumed 4131149 = 0036 mol NH4 are produced

[58] Acknowledgments We would like to thank three anonymousreviewers for their constructive comments on an earlier version of this man-uscript This work was supported by NOAA CSCOR grantsNA06N0S4780198 and NA09N0S4780208 and the US IOOS CoastalOcean Modeling Testbed NOAA NGOMEX publication no 173 KFwas also supported by NSERC and CFI


1001

ReferencesAulenbach B H Buxton W Battaglin and R Coupe (2007) Streamflowand Nutrient Fluxes of the Mississippi-Atchafalaya River Basin andSubbasins for the Period of Record Through 2005 Open-File Report2007-1080 US Geological Survey httptoxics usgs govpubsof-2007-1080report_site_map html

Burchard H and R Hofmeister (2008) A dynamic equation for the poten-tial energy anomaly for analysing mixing and stratification in estuariesand coastal seas Estuarine Coastal Shelf Sci 77 679ndash687

da Silva A C Young-Molling and S Levitus (1994a) Atlas of SurfaceMarine Data 1994 Vol 3 Anomalies of Fluxes of Heat and MomentumNOAA Atlas NESDIS 8

da Silva A C Young-Molling and S Levitus (1994b) Atlas of SurfaceMarine Data 1994 vol 4 Anomalies of Fresh Water Fluxes NOAAAtlas NESDIS 9

David M L Drinkwater and G McIsaac (2010) Sources of nitrate yieldsin the Mississippi River Basin J Environ Qual 39 1657ndash1667

Feng Y S F DiMarco and G A Jackson (2012) Relative role of windforcing and riverine nutrient input on the extent of hypoxia in the North-ern Gulf of Mexico Geophys Res Lett 39 L09601

Fennel K J Wilkin J Levin J Moisan J OrsquoReilly and D Haidvogel (2006)Nitrogen cycling in the Middle Atlantic Bight Results from a three-dimensional model and implications for the North Atlantic nitrogen budgetGlobal Biogeochem Cycles 20 GB3007 doi1010292005GB002456

Fennel K J Wilkin M Previdi and R Najjar (2008) Denitrificationeffects on air-sea CO2 flux in the coastal ocean Simulations for the north-west North Atlantic Geophys Res Lett 35 L24608 doi1010292008GL036147

Fennel K et al (2009) Modeling denitrification in aquatic sedimentsBiogeochemistry 93 159ndash178

Fennel K and J Wilkin (2009) Quantifying biological carbon export forthe northwest North Atlantic continental shelves Geophys Res Lett36 L18605 doi1010292009GL039818

Fennel K (2010) The role of continental shelves in nitrogen and carboncycling Northwestern North Atlantic case study Ocean Sci 6 539ndash548

Fennel K R Hetland Y Feng and S DiMarco (2011) A coupledphysical-biological model of the Northern Gulf of Mexico shelf Modeldescription validation and analysis of phytoplankton variabilityBiogeosci 8 1881ndash1899

Flather R (1976) A tidal model of the northwest European continentalshelf Mem Soc R Sci Liege 10 141ndash164

Forrest D R R D Hetland and S F DiMarco (2011) Multivariablestatistical regression models of the areal extent of hypoxia over theTexas-Louisiana Shelf Environ Res Lett 6 045002

Garcia H and L Gordon (1992) Solubility of oxygen at different temper-ature and salinity Limnol Oceanogr 37 1307ndash1312

Goolsby D W Battaglin B Aulenbach and R Hooper (2000) Nitrogenflux and sources in the Mississippi River Basin Sci Total Environ 24875ndash86

Greene R J Lehrter and J Hagy (2009) Multiple regression models forhindcasting and forecasting midsummer hypoxia in the Gulf of MexicoEcol Appl 19 1161ndash1175

Haidvogel D et al (2008) Ocean forecasting in terrain-following coordi-nates Formulation and skill assessment of the Regional Ocean ModelingSystem J Comput Phys 227 3595ndash3624

Hetland R and S DiMarco (2008) How does the character of oxygendemand control the structure of hypoxia on the Texas-Louisiana continen-tal shelf J Mar Syst 70 49ndash62

Hetland R and S DiMarco (2012) Skill assessment of a hydrodynamicmodel of the circulation over the Texas-Louisiana continental shelfOcean Modell 4344 64ndash76

Hypoxia Task Force (2008) Mississippi RiverGulf of Mexico WatershedNutrient Task Force Gulf Hypoxia Action Plan 2008 for ReducingMitigating and Controlling Hypoxia in the Northern Gulf of Mexico andImproving Water Quality in the Mississippi River Basin Washington DC

Kemp W M P A Sampou J Garber J Turtle and WR Boynton(1992) Seasonal depletion of oxygen from bottom waters of ChesapeakeBay roles of benthic and planktonic respiration and physical exchangeprocesses Mar Ecol Prog Ser 85 137ndash152

Ko D S P J Martin C D Rowley and R H Preller (2008) A real-time coastal ocean prediction experiment for MREA04 J Mar Syst69 17ndash28

Laurent A K Fennel J Hu and R Hetland (2012) Simulating the effectsof phosphorus limitation in the Mississippi and Atchafalaya Riverplumes Biogeosci 9 4707ndash4723 doi105194bg-9-4707-2012

Mellor G and T Yamada (1982) Development of a turbulence closuremodel for geophysical fluid problems Rev Geophys 20 851ndash875

Murrell MC and Lehrter J (2011) Sediment and lower water columnoxygen consumption in the seasonally hypoxic region of the Louisianacontinental shelf Estuaries Coasts 34 912ndash924

Previdi M K Fennel J Wilkin and D B Haidvogel (2009) InterannualVariability in Atmospheric CO2 Uptake on the Northeast US ContinentalShelf J Geophys Res 114 G04003 doi1010292008JG000881

Pierson J J M R Roman D G Kimmel W C Boicourt and X Zhang(2009) Quantifying changes in the vertical distribution of mesozooplank-ton in response to hypoxic bottom waters J Exp Mar Biol Ecol Suppl381 S74ndashS79

Rabalais N R Turner and W Wiseman Jr (2002) Gulf of MexicoHypoxia aka ldquoThe Dead Zonerdquo Annu Rev Ecol Syst 33 235ndash263

Rowe G T M E C Kaeki J WMorse G S Boland and E G E Briones(2002) Sediment communitymetabolism associated with continental shelfhypoxia northern Gulf of Mexico Estuaries 25(6) 1097ndash1106

Seitzinger S and A Giblin (1996) Estimating denitrification in NorthAtlantic continental shelf sediments Biogeochemistry 35 235ndash260

Simpson J H (1981) The shelf-sea fronts implications of their existenceand behaviour Phil Trans R Soc Lond A 302 531ndash546

Wallcraft A J E J Metzger and S N Carroll (2009) SoftwareDesign Description for the HYbrid Coordinate Ocean Model (HYCOM)Version 22 Tech Rep Naval Research Laboratory Stennis SpaceCenter MS

Wanninkhof R (1992) Relationship between wind speed and gasexchange J Geophys Res 97 7373ndash7382

Wiseman W N Rabalais R Turner S Dinnel and A MacNaughton(1997) Seasonal and interannual variability within the Louisiana coastalcurrent stratification and hypoxia J Mar Syst 12 237ndash248


1002

major goals a reduction of the hypoxic zone area to 5000km2 or less by 2015 through voluntary actions [HypoxiaTask Force 2008]mdasha goal that is unlikely to be met Asillustrated by the Task Force Action Plan scientific under-standing about the mechanisms underlying the generationand maintenance of the hypoxic zone on the TX-LA shelfis influencing decision making at various state and federallevels Decisions should be informed by a comprehensiveunderstanding of the processes influencing the occurrenceof hypoxia to devise effective nutrient reduction strategiesand to build reasonable expectations as to what effects mightbe expected in response to nutrient reductions[5] High-resolution physicalndashbiogeochemical models are

indispensable tools for building a mechanistic understandingof cause-and-effect relationships The objectives of thisstudy are to present such a model evaluate its deterministicsimulations of hypoxic conditions on the TX-LA shelfagainst observations and to assess the sensitivity of themodelrsquos simulated hypoxia to choices that have to be madeduring model development We systematically explore theeffects of three choices all of which could potentially affecthypoxia simulations namely the vertical resolution the re-alism in horizontal boundary conditions and the treatmentof oxygen consumption at the sediment-water interfaceWe use an ecosystem model that has been shown to realisti-cally reproduce many observed features of nutrient andplankton dynamics on the TX-LA shelf [Fennel et al2011] and that has been used successfully before to studythe dynamics of another dissolved gas inorganic carbonon the northeastern North American shelf [Fennel andWilkin 2009 Previdi et al 2009 Fennel 2010] In a recentstudy by Laurent et al [2012] the model was extended toinclude phosphate as additional nutrient however theexperiments discussed here all use the nitrogen-based modelwithout phosphate Here the model was extended to explic-itly describe the dynamics of dissolved oxygen includingairndashsea gas exchange photosynthetic production of oxygenoxygen consumption in the water column due to respirationand nitrification and sediment oxygen consumption Differ-ent treatments of the latter are explored[6] The manuscript is organized as follows in section 2

we first describe the physical model including the differentconfigurations for horizontal boundary conditions (climato-logical observations versus output from operational models)then the biological model component and parameterizations

of oxygen sources and sinks In section 3 we discuss modelresults focusing initially on the areal extent of hypoxicconditions and its sensitivity to the choice of sediment oxy-gen consumption (SOC) parameterization and horizontalboundary treatment (section 31) then we discuss the verti-cal structure of hypoxic conditions (section 32) the role ofvertical stratification (section 33) and finally assess whetherhypoxic conditions are simulated in the right locations(section 34) Our main conclusions (section 4) are that hyp-oxia simulations on the TX-LA shelf are very sensitive to thetreatment of SOC (a direct result of the fact that hypoxia islimited to a relatively thin layer above the bottom) that thestrength of vertical stratification is an important predictorof hypoxia and that modification of physical horizontalboundary conditions can lead to dramatic differences in hyp-oxia simulations because vertical stratification strength canbe affected by such a change

2 Model description

21 Physical model

[7] We used several configurations of the Regional OceanModeling System (ROMS httpmyromsorg [Haidvogelet al 2008]) for the MississippiAtchafalaya outflowregion These configurations differ in terms of vertical reso-lution (20 or 30 vertical layers) and in the treatment ofhorizontal boundary conditions (climatological or more real-istic conditions from operational Gulf of Mexico models)but all use the same horizontal grid (Figure 1) The modelsrsquoterrain-following vertical layers are stretched to result inincreased resolution near the surface and bottom The hori-zontal resolution is highest near the Mississippi Delta withup to 1 km and lowest in the southwestern corner with~20 km The model uses a fourth-order horizontal advectionscheme for tracers a third-order upwind advection schemefor momentum and the turbulence closure scheme for verti-cal mixing by Mellor and Yamada [1982] The simulationperiod is from 1 January 2004 to 31 December 2007[8] In simulations with climatological boundary treatment

(henceforth referred to simply as climatological simula-tions) radiation conditions [Flather 1976] are imposed forthe three-dimensional velocities with no mean barotropicflow Temperature and salinity at the boundary are relaxedto a horizontally uniform monthly climatology with a time-scale of 10 days for outgoing and 1 day for incoming

88 W

28 N

29 N

30 N Atchafalaya Bay

MississippiRiver Delta

10 m

20 m

30 m

40 m50 m

100 m200 m300 m

Miss R

Atch R

Sabine R

Texas Louisiana

95 W 89 W90 W91 W92 W93 W94 W

Figure 1 Model domain grid (light gray lines) and selected isobaths (dark gray lines)


991






22 Biological model





Ox


(1)




992



0

(2)




F frac14 vkO2Δz




vkO2 frac14 031u210


ScOx

r (4)








2T=10

C

1 exp Ox

30 mmol O2 m3frac12

(5)








993



COx mmol O2 m3

(7)


















994


2004 2005 2006 2007





5

10

15

20

25

30


Hyp

oxic

are

a (1

03 km

2 )




0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )



995






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002






22 Biological model





Ox


(1)




992



0

(2)




F frac14 vkO2Δz




vkO2 frac14 031u210


ScOx

r (4)








2T=10

C

1 exp Ox

30 mmol O2 m3frac12

(5)








993



COx mmol O2 m3

(7)


















994


2004 2005 2006 2007





5

10

15

20

25

30


Hyp

oxic

are

a (1

03 km

2 )




0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )



995






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002



0

(2)




F frac14 vkO2Δz




vkO2 frac14 031u210


ScOx

r (4)








2T=10

C

1 exp Ox

30 mmol O2 m3frac12

(5)








993



COx mmol O2 m3

(7)


















994


2004 2005 2006 2007





5

10

15

20

25

30


Hyp

oxic

are

a (1

03 km

2 )




0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )



995






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002



COx mmol O2 m3

(7)


















994


2004 2005 2006 2007





5

10

15

20

25

30


Hyp

oxic

are

a (1

03 km

2 )




0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )



995






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002


2004 2005 2006 2007





5

10

15

20

25

30


Hyp

oxic

are

a (1

03 km

2 )




0

10

20

30

40

50


Hyp

oxic

are

a (1

03 km

2 )



995






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002






locations





0 5 10 15 20


Pro

babi

lity

dist

ribut

ion


0 5 10 15 20


roba

bilit

y di

strib

utio

n


0 5 10 15 20

0

005

01

015

02

025

0

005

01

015

02

025

0

005

01


Pro

babi

lity

dist

ribut

ion





996





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002





f frac14 1

h

Z 0


h

Z 0

hrdz (8)


















0

1000

2000

0

1000

2000

0

1000

2000

0

1000

2000



0

100

200

0

100

200

0

100

200

0

100

200

φ (J

mminus

3 )

botto

m o

xyge

n (m

mol

O2

mminus

3 )



997







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002







1

2

3

4

5


Pro

babi

lity

dist

ribut

ion








B30HYCB30IAS


0

2000

4000

6000

8000

0

50

100

150

200

250

φ (J

mminus3

)bo

ttom

oxy

gen

(mm

ol O

2 m

minus3)




998











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002











999





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002





2004 2005 2006 20070

10

20

30

40

50

60

70

80

s

tatio

ns c

orre

ct



2004 2005 2006 20070

02

04

06

08

1

12

Nor

mal

ized

O2

RM

SE




1000


4 Conclusions
















1001





































1002


4 Conclusions
















1001





































1002





































1002

sensitivity of hypoxia predictions for the northern gulf...

Documents