 to be here. So today I'm going to be presenting some work that I've done looking at oxygen dynamics in Chesapeake Bay. This is work that was largely funded through some NSF project that I have as well as a larger NOAA proposal that a lot of people were involved in. Some of them are here. So a quick outline for my talk, I'm just going to give some real basic background and motivation for why I'm looking at this issue. I'm going to talk about a very simplified modeling approach that I've taken and then I'm really trying to understand the role that physical forcing has in the seasonal variation of hypoxic volume in Chesapeake Bay. So I'm going to talk about river discharge, heat flux and temperature, as well as wind direction and magnitude and then at the end sort of talk about sort of inter-annual variability because that's from a management point of view what we're really interested in trying to understand and finish up with some conclusions. So some of the funding that supported this project was through the NOAA IU's office and a proposal that was submitted by Sura and there's quite a large number of people involved. There's three different focus areas. One was on storm surge, one was on the Gulf of Mexico hypoxia and then there was another group looking at the estuarine hypoxia, mainly focused on Chesapeake Bay. So that's the part of this project that I was involved in. We have a number of federal partners, people at NOAA as well as EPA and Army Corps of Engineers and then a whole group of people, many of whom are involved in systems as well. So Carl Friedrich is a project coordinator. Wen Long who's here is really the one of the driving forces behind Chesfroms and so he and Raleigh Hood have done a lot of work developing the model that I'm going to be presenting and then my role was simply to take the models that they've developed and look at some very very simple ways of trying to capture oxygen dynamics. So Chesfroms, which is an implementation of ROMS, which is a regional ocean modeling system, has been incorporated into systems and so this is just a real basic plot. I'm sure you'll be seeing more things like this later but there's a couple of different versions of ROMS that can be used through systems currently. So this is a nice resource. I'm not going to go into how you do that but I'm sure there's people here who are better suited to talk about that but I just want to make people aware that this Chesfroms implementation is available and so I'm going to make the case of linking a lot of other types of models into a system like Chesfroms is going to be a very valuable resource. So just to give some motivation for why we're looking at oxygen dynamics in Chesapeake Bay, this is a map of the mean bottom dissolved oxygen content in the summer of 2005. So these red colors represent regions of very very low dissolved oxygen or the complete absence of dissolved oxygen. So large portions of Chesapeake Bay in the deep part of the channel every summer go hypoxic or anoxic. So obviously that's going to have a large impact on a wide array of biological and ecological processes and from a management point of view it's thought that the nitrogen loading going to the bay is playing a large role in this and there's a study that indicated between 1985 and 1996 about 3.5 billion dollars were spent on nutrient reduction. So from a management point of view this is an issue of significant economic importance and a lot of the way we sort of set the management policy for the bay is based on models. So I think understanding how models predict oxygen dynamics is pretty important and I'm going to make the argument that if you really want to assess the success or failure of these models you have to understand the underlying sort of physical variability because we have lots of physical processes that vary from year to year. So this is just a real simple way in which we think about hypoxia or anoxia in a system like just speak bay. It's thought that we have nitrogen loading so the rivers put in nutrients and they fuel algal growth. These large phytoplankton blooms eventually die and sink to the bottom and when they're in the bottom waters they're going to be respired by bacterial organisms and so we're going to basically draw the oxygen out of the bottom waters. That river discharge is also creating density stratification which is inhibiting vertical mixing and so the sort of conceptual model we have for hypoxia in the bay is that when we have large nutrient inputs we have lots of phytoplankton growth. We also have a lot more stratification with that input of buoyancy. We're going to tend to favor extensive hypoxia. In contrast when we have low river inputs of nutrients we're going to have less phytoplankton growth. We're also going to have perhaps wind mixing and other physical processes that are going to help drive turbulent mixing and so we can have these physical processes in combination with these biological processes reducing the amount of hypoxic water and so I think an important thing that we need to look at is can we sort of try to determine how much of the variability we see from year to year is driven by physical processes versus biological processes. So to try to get at that question I'm using the regional ocean modeling system and again this is a Chesapeake Bay implementation called chess ROMs that went long and Raleigh Hood have put a lot of hard work into developing. This is the model grid here on the right so we have sort of spatial scales of hundreds of kilometers or hundreds of meters for our grid resolution and then the model forcing we have realistic tides and sub-tidal elevation at the ocean boundary so now we're talking about sort of time scales of hours in terms of our forcing. We have realistic surface fluxes and things like surface heating and wind forcing. We have the observed river discharge for all of the main tributaries and then our boundary condition we're just setting to sort of climatology. So this is stuff that when and Raleigh have spent a lot of time on and they've got some a 15 year time series sort of a hind cast from 1991 to 2005 and so what I've done is I've just taken the sort of hydrodynamic forcing that they've developed and I've added a very very simple way of trying to represent oxygen dynamics. So the oxygen model that I've developed essentially introduces oxygen as an additional model tracer so the model just affects and mixes oxygen and I've prescribed a constant consumption of oxygen you can think of this as a net respiration and it's got a depth dependent vertical distribution so at the surface the consumption of oxygen is zero and it just sort of linearly increases with depth. So this is other than it's depth dependent it's spatially uniform and it's time invariant so I'm really taking a very very crude approach to looking at the oxygen dynamics with the goal of trying to figure out the role of the physical processes. So there's no oxygen consumption outside the eastern portion of the domain. I don't have any net oxygen production. I set the boundaries equal to the saturation based on temperature and salinity so I'm basically sucking oxygen out of the water column and the only way I can get oxygen back into the water column is using a surface flux. So I've got a wind speed dependent piston velocity and that's based on the formulation by Marino and Howard. So all that does is it takes the surface oxygen concentration and the difference between that and its saturated value is simply multiplied by this K which is the piston velocity and that gives you a surface flux and this K is dependent on wind speed. So I'm sucking oxygen out of the water column I'm putting it back in through a surface flux I'm letting the model sort of mix and do what it needs to do to predict oxygen distribution and the key thing is I'm assuming that the biology is constant. So again it's a very gross simplification but I want to really isolate the role of physical processes because I want to understand the dominant physical processes that modulate oxygen. So just to give you some idea of sort of what the cycle of oxygen dynamics looks like in the bay this is some data that was compiled by Rebecca Murphy she's got a nice manuscript that actually just came out in estuaries and coasts and what this is showing is using Chesapeake Bay program data where they go out bimonthly and do CTD surveys and measure a suite of water quality parameters. Rebecca has taken those data and sort of come up with an interpolated volume of how much volume in the bay at any given cruise has an oxygen concentration less than a milligram per liter. So the units are cubic kilometers and then this is sort of the climatology over this about a 25 year record. So the blue circles are just the mean hypoxic volume over this 25 year period and we see that typically as we move from May into the early summer months we start to develop this hypoxic volume it increases in peaks in July and then it sort of wanes in the latter months in the fall. The blue lines are the one is one standard deviation on this time series and then the red line is the maximum and the minimum observed in any cruise and these have been sort of been averaged by bimonthly cruises. So we can see in sort of the peak of the season in July we can have anywhere from say 14 cubic kilometers of hypoxic volume all the way down to say one. So there's a lot of interannual variability as well as a very strong seasonal cycle and so I'm trying to tease out what role of physical processes has not only in the seasonal cycle but also in the interannual variability. So just to give you a quick overview of what are the dominant physical processes and how they vary both seasonally and interannual time scales this is the Susquehanna River discharge this is data that's been averaged monthly going back to 1967 and that's the dominant source of buoyancy and freshwater into the system there are other tributaries but they have a very similar sort of seasonal hydrograph. So again the blues are monthly averages the vertical bars are standard deviations and the red or the maximum and the minimum. So there's a very clear seasonal cycle we have a spring fresh that puts a lot of fresh water into the system that sets up the stratification that is thought to play a role in limiting the mixing and allowing these anoxic waters to develop but there's a lot of interannual variability we can have very very high river discharge and very low river discharges. So typically we think of the river discharge being as one of the most fundamental parameters and contributing to this interannual variability. The tricky thing about the river discharge is it's also bringing in the nutrients and so it has a biological component to it so the approach that I'm using essentially by taking the biology out of the process is I'm parsing the physical effects of the river discharge and I'm decoupling them from their role in delivering nitrogen or phosphorus. The wind speed also has a very nice seasonal cycle we have higher winds in the winter as we go into the spring and into the summer the wind speed decreases so there's less vertical mixing due to wind stress and then as we go back into the fall the wind speeds pick up again and so again this is something that we think contributes to both the seasonal cycle as well as the interannual variability. Temperature obviously in the mid-Atlantic region has a strong seasonal cycle and there is interannual variability and the reason why that might be important is because the amount of oxygen that can be held at saturation is a strong function of temperature so this is just taking this data and plotting what the oxygen saturation is and we can say in cold water we can hold say maybe 12 milligrams per liter of oxygen and during the summer months at saturation maybe we're less than eight milligrams per liter so when the water is warm it can hold less oxygen and so there's less inventory to draw down to get to zero and so again this temperature we think has a role in the seasonal cycle and maybe perhaps also in the interannual variability. So I just want to convince you or show you that this model that I'm using that has essentially biological processes that are constant can represent at least the sort of seasonal cycle of hypoxia in the bay reasonably well. So these are Chesapeake Bay program observing stations going from the upper bay sort of moving southward to the lower bay and I'm showing a three-year simulation of just the bottom dissolved oxygen concentration so the blue circles are individual Chesapeake Bay program data where they they measured the bottom oxygen concentration and the red line is what the model is predicting so the model with no biological variability is getting a nice seasonal cycle of the oxygen dynamics in the bay it's even capturing some of these events like this one in August of 2003 that was a big wind mixing event so the model actually shows remarkable skill I think given that there's no biological variability at all and I think it highlights the role that physical processes play in the sort of seasonal cycle of oxygen dynamics in the bay. This is just a showing you the model is actually getting some of the spatial variability as well what's contoured in the background is the bottom dissolved oxygen concentration that the model is predicted and on the left we're looking at a period in July and then the circles are again the Chesapeake Bay program's data from these cruises so the model largely is getting this large anoxic zone in the bottom of the deep part of the channel it's getting roughly the sort of spatial extent correct and if we go from July to a couple weeks later when the next cruise in August occurred there was a big erosion of this low oxygen water in the lower part of the bay which the model is largely capturing so it's not perfect there's a lot of areas you could highlight where there's a mismatch between the model and the observations but in a very sort of qualitative sense the model seems to be capturing sort of large-scale variability that's seen in the observations so from a management point of view one of the metrics that we're really interested in is what's the total hypoxic volume any given year and so the sort of theme of this meeting is time scales and space scales I'm now looking at sort of the integration of both of those things and trying to come up with sort of spatially and temporally integrated hypoxic volume from any one season and so the model can give us some estimate of that and what I'm doing here is I'm showing the hypoxic volume that the model predicted for this three-year period in black in the red is the observed or interpolated hypoxic volume from data and so we can see that from year to year we can have you know 50% differences in the integrated amount of hypoxic volume in space and time and the model is getting some of that variability so 2004 has roughly 50% less integrated hypoxic volume than 2005 and the model is getting some of that variability it's not doing very well in the early part of 2003 so the model gets some of the inter-annual variability but it's not getting all of it and I'm going to return to that later in the talk but this is purely due to physical variability because the biology has been assumed to be constant so clearly from a year to a year point of view physical processes play a role in controlling the water quality in the system. Okay so now I just want to go through some sort of sensitivities that I've looked at sensitivity runs I've done looking at the importance of these various physical forcing so I'm going to look at river discharge I'm going to look at heat flux and temperature and I'm going to look at wind speed and direction in terms of the role they're playing in the seasonal cycle so just to remind you here's that climatology of Susquehanna River discharge we have the big spring freshet that decreases as we move into the summer months and that sets up the stratification because of the sort of long residence time in Chesapeake Bay the stratification has a lag and it peaks in the summer and so it's thought that the seasonal hydrograph that sets up the stratification is really important to the seasonal cycle of hypoxia in the Bay so what I'm going to do is I'm going to just take the model and I'm going to run it with the full physical forcing where everything's allowed to vary and then I'm going to do another model run where I just hold the river discharge from all the tributaries constant so again this is hypoxic volume on the y-axis and this is time and all of my plots are going to focus on the 2004 year the blue line is the base case run and so this shows that there's no hypoxic volume is in the winter and as we move into the spring and early summer we develop this hypoxic volume that gets mixed away and so the blues the base case where everything all the parameters vary in time and the red case the red line is if you assume that the river discharge was just held at its constant mean value and you see that there are some changes but in general the time variability the river discharge isn't having a pronounced influence on this seasonal cycle so when you remove the sort of biological component where the river discharge is delivering nutrients you don't see a very strong influence of the river discharge which I found rather surprising the next thing I did is I just took the base case and I took all the tributaries and I for the red line I doubled their river discharge for the black line I reduced their discharge in half and the green line represented 20% reduction to 20% of their values so we're dealing with an order of magnitude change in the annual river discharge and again these are the predicted hypoxic volumes the blue is the base case these are sort of the integrated hypoxic volume days for that entire year and there's a couple interesting things first of all if we take the case the black case where we reduce the river discharge by 50% and we look at the base case and the doubling of the river discharge we're actually seeing hypoxic volume is going down slightly however the very lowest river discharge is actually slightly lower so there's not a linear dependent so it's a little bit more complicated than that but overall a two an order of magnitude change in river discharge is only leading to about a less than 10% change in integrated hypoxic volume so the sort of conventional wisdom that the river discharge is one of these fundamental physical parameters does not seem to be supported by this model it seems to suggest that it's very insensitive to river discharge again when you take out the fact that the rivers are delivering the nutrients here's our temperature climatology so again I wanted to look at the role that temperature has in the seasonal cycle as well as inter-annual variability and so I did a couple of similar sensitivity runs where the blue again is the base case where I where the all the real all the forcing is realistic the red line is if I just assume the water temperature in the bay was constant 25 degrees Celsius year-round so again we we have this seasonal cycle in the early spring there's slightly more hypoxic volume slightly less later in the season but it's not showing a really strong difference when you have constant water temperature at 25 degrees however if you assume that the water temperature was five degrees year-round we would have a much more muted amount of hypoxic volume and so clearly the fact the water temperature is warm in the summers and cold in the winters is playing some role in this seasonal cycle however this kind of analysis is pretty unrealistic we're not ever going to have a case where the Chesapeake Bay doesn't have any seasonal variability and we're not going to have five-degree water temperatures year-round so to try to more realistically simulate the kinds of temperature changes we do observe I took the air temperature and I increased the monthly air temperature by one standard deviation and I decreased it by one standard deviation so what that's doing is affecting the sensible heat flux into the bay and it's having an effect on the mean annual water temperature so this is the modeled water temperature for a case where the air temperature is increased by a standard deviation and the blue is where it's decreased by a standard deviation so that leads to a sort of an annual difference of about one and a half degrees Celsius when you average the two over the course of the year and this is just taking some data 1998 and 1992 which are sort of a example of a cold year and a warm year so these have very similar sort of changes in temperature so this is a much more realistic representation of sort of the inter-annual variability that we might see in temperature so when you do that we've got the warm case in red in the cold case in blue they're very similar but there are some sort of noticeable differences now we have measurable changes in the total integrated amount of hypoxic volume just based on water temperature so from year to year warm years are going to tend to have higher amounts of hypoxic volume and cool years are going to have less hypoxic volume and so this increase in surface heating is resulting in roughly 20% increase in integrated hypoxic volume so this is a stronger influence than a two order of magnitude or one order magnitude change in river discharge so again I was somewhat surprised by this result next thing I want to do is I want to look at the role of wind forcing so again we've got higher wind speeds in the winter lower wind speeds in the summer increasing into the fall we also have changes in wind direction over Chesapeake Bay the winds tend to blow sort of out of the north northwest during the winter we transition to the summer months we switch to that sort of Bermuda high setup where we get sort of south-southwesterly winds and so we've got changes both in speed and in direction and I want to sort of try to systematically look at those as well so the first thing I did is the blue again is this base case this is the time series of hypoxic volume so the red line represents if the winds blew like they did in July year round so there's a 30-day cycle of wind variability but I just repeated that for every single month of the model so there's daily variability but there's no seasonal variability and I assume that each month had the winds from July if you do that you see that the hypoxic conditions start a lot earlier they're more persistent and they last a lot longer so clearly if we had weak winds year round we would have a lot more hypoxic volume and so the winds are playing a role in the seasonal cycle the black line is if I just took the January time series of winds and I just repeated that for every month that I ran the model and the black line basically shows no hypoxic volume so if the winds blew like they did in January year round we would not develop hypoxic volume according to this model again that's not a very realistic simulation so I wanted to do some more systematic and realistic changes to the forcing so I just took the summer months so these are histogram showing the mean wind speed as a function of month so we had this decrease in the summer and I just changed the summer wind speeds by plus or minus 15% I just wanted to see what a 15% change in wind speed would do during the summer months so again this is another plot of hypoxic volume with time blue is the base case when you increase the wind speed by 15% in the summer months which is the red line you see a very big reduction in the total amount of hypoxic volume the black line is when you reduce the wind speed we see this big increase in hypoxic volume so this is roughly a factor of three difference in the amount of hypoxic volume we might observe simply based on subtle changes in the wind speed wind direction is another thing that I've spent a lot of time looking at in the summer months we typically have winds that blow sort of out of the south southwest so I wanted to examine the role of changes in wind direction so I rotated the winds during the summer months positive 90 degrees which makes the wind blow mainly out of the west negative 90 degrees where they're rotated to blow mainly out of the east and then I rotated them 180 degrees so they blow mainly out of the north and again I just want to look at how so the wind speed in all these cases going to be the same it's just going to be the wind direction is going to be rotated during the summer months and again we have pretty large differences between say the black line which is when the winds are blowing mainly out of the west to the base case and then the red line is when they're blowing out of the north so just based on wind direction even though the wind speed is the same we can see very large differences in hypoxic volume again we're talking sort of factor of two differences based on differences in summer wind direction so wind speed and direction appear to be the most important physical parameters that both drive the seasonal cycle and they might be important to the sort of interannual variability as well so I just want to finish up and look at the interannual variations hypoxic volume because again from a management point of view we want to know why one year the water quality is bad and has a large hypoxic volume versus another year when say the hypoxic volume is much smaller and again we want to be able to try to manage this this resource and we've been doing that largely looking at nitrogen loading but again all those influences are obscured by this physical variability so I ran the model with constant biological forcing for this 15 year period so the model has been sampled the same way the observations were sampled and compared to these Chesapeake Bay program observations in green so you know we get the seasonal cycle from year to year it's repeatable in all 15 years some years like 2004 and 2005 the model does very well some model some years like 2003 it doesn't do so well and I've already sort of hinted at this but it doesn't there's a lot of years where it does pretty badly so this is zooming in on 1995 and 1994 1995 the models doing very well I would say but in 1994 it's significantly under predicting the observed hypoxic volume so can the model with just physical variability capture what's really going on in in reality and it's not doing all that great this is the the monthly or bimonthly averages with the maximum and the minimum as a seasonal cycle and also showing the inter-annual variability this is the observations on the left and then the model on the right so the model with no biologic variability does show significant inter-annual variability but it doesn't have enough variability when you compare the model to the observations and there's some subtle differences like it's under predicting the early summer and it's over predicting the late summer on average so does the physical forcing explain the observed inter-annual variability on hypoxic volume well this is a plot of the the observed versus the model for this 15-year time series where I'm just averaging the hypoxic volume over the entire season and it gets a little bit of the variability but it's not doing a very good job it actually does not explain the inter-annual variability in a statistically significant way so this is very preliminary but I wanted to take a very simple approach and say are the residuals correlated with any obvious other process at the models leaving out so I just took the residuals and I did a sort of lag correlation with the previous nitrogen loading from Susquehanna River and I see that no matter which definition of hypoxic volume you use you end up with these statistically significant correlations with nitrogen loading so then I just came up with a very simple load dependent respiration rate so I scale my respiration rate now based on the integrated nitrogen loading so this is a time series over that 15-year period of what the respiration rate would be based on the integrated nitrogen loading and I run the model again just to sort of see how that does and sort of cut to the chase when you actually include this sort of long this low dependence or the amount of nitrogen that's going into the bay and you make your respiration rate proportional to that you actually do get sort of a better statistical representation of this inter-annual variability so this is very preliminary and it's and it just finished doing this but the sort of take-home message is to get the inter-annual variability you not only have to have the physics but you also have to have some sort of biological component and that's going to need to have accurate nitrogen loading so you're probably gonna have to have a watershed model that can deliver nitrogen to the bay you're gonna have to have a more sophisticated biological model that's gonna process that nitrogen and that's gonna have to be coupled with the hydrogen amic model so again I think in terms of what systems does well we can start pairing up these models we can start putting nitrogen in we can link that to a hydrogen amic model and then we can link other biologic models to try to get at this process and I think that's sort of ultimately where this is all going is to have much more sophisticated approaches where we link various models together so my conclusions are that a relatively simple model with no biologic variability at least reasonably accounts for the seasonal cycle of hypoxia in the bay wind speed in direction appear to be the two most important physical variables the results are largely insensitive to river discharge when you isolate the role of nutrient delivery air temperature seems to be playing a role as well at both seasonal and inter-annual time scales but the key thing is this 15-year simulation with constant respiration rate is not able to capture the inter-annual variability so clearly from a management point of view that's what we want to know and so you're gonna have to some way count for the fact that the nitrogen loading is driving the biology and so preliminary attempts to include the effects of nitrogen loading through this low-dependent respiration are showing promise for at least getting some of this long-term variability. Thank you very much. Any questions? Malcolm I just had a comment not really a question it's really surprising to me that when you change the river discharge and you don't see much oxygen hypoxia or water change I think it's because as you I think you already mentioned that you assume that the the river loading or your respiration rate is decoupled from the discharge and that actually shows up when you do this 15-year run and you say that it cannot explain the internal variability and that's one part I think you helped me to understand that process. My comment is really when you swatch the summer wind with the January wind you see a huge difference between the hypoxia volume. There are two things going on there one is when you do the swapping you swap to the surface LSE into exchange rate why because your exchange rate is a function of the wind speed and also the saturation oxygen saturation concentration which is also function of temperature. The other factor is when you switch the wind it changes the stratification so it's hard to know whether it's because of the change of the surface flux or it's because of the change of the stratification inside of the water gala. So my suggestion is what if you keep the flux the surface LSE fluxes the constant to be constant but you change the wind speed and change the stratification regime and and that might tell which one is more important in changing the oxygen. Yeah I agree with that and one of the ways I haven't done the run that you're suggesting but one of the things I've done to try to get at that is you can you can look at budgets for how does oxygen get into that lower layer and the along channel residual circulation in the bay is very sensitive to wind speed and direction and so what's happening I when you do these sort of budgets as you realize that the strong northerly winds are enhancing the along estuary exchange and they're bringing in a lot of oxygen from the lower portion and bringing that up the bay and so that's a component of it so that's decoupled even from the turbulent mixing it's not direct wind driven mixing but it's these long channel and as well as lateral circulations that are driven by winds so it's not simply a stratification effect in terms of the way the wind drives the circulation and you can get at that when you start looking at these budgets you know for the purposes of this talk I didn't get into all the details but there's some very clear patterns about what the wind is doing in terms of altering circulation and how that introduces oxygen into the lower waters so but yes your suggestion about changing the surface flux making that constant is a very good suggestion in terms of teasing out the different thank you I'll stay tuned question oh sorry I didn't see you go ahead thanks for a great presentation I have two questions one very simple straightforward the other maybe a little more difficult but the first one is how did you relate your nitrogen input to the rest a respiration rate you just made some simplifying assumptions there yeah so I did a really basic I took all the nitrogen and I converted it to carbon via the redfield ratio and assumed that that had to be respired over some time scale over some volume and made my respiration rate match that in an integrated sense which ended up with the models way too sensitive so I had to damp out that sensitivity so it's sensitive to the nitrogen loading but if I assume that every gram of nitrogen that entered was respired and that led to an oxygen demand it seemed to be way too sensitive so I had to sort of damp that out a little bit the interesting part about that is of course the nitrogen actually cycles through the system many many times before being washed off shore so it's interesting that it should have been under sensitive not oversensitive to it interesting so the other question is what would be needed to apply the ROMs model globally to coastal systems awesome really big computers I mean I think when you start looking at oxygen and you're looking at processes like vertical mixing you have to have the resolution to capture things like picnic lines and interactions with the symmetry and so I mean at a global scale yeah having aren't they one in the same having spent 10 years at Rutgers where ROMs lives right I've always gotten that same response but it seems like it certainly would be a very could be very useful for many different reasons to be able to move ROMs really into the coastal system globally oh I agree 100% I just think in terms of the physical mechanisms that lead to the mixing the scales that we could resolve with a global model right now just would be it would be very hard to get at the at least in a system like this maybe on a larger system it would work out better great thanks Malcolm