Speaker Info

Richard Zimmerman
Professor of Ocean, Earth and Atmospheric Sciences
Old Dominion University

Email: rzimmerm@odu.edu

Biography:

Dr. Zimmerman is Professor of Ocean, Earth & Atmospheric Sciences at Old Dominion University. His research interests include the ecological physiology of marine photosynthesis, metabolic regulation of carbon and nutrient dynamics in marine ecosystems, radiative transfer and remote sensing of optically shallow waters, ecosystem productivity and numerical modeling. Current projects involve the physiological responses of seagrasses to climate warming and ocean acidification, and subsequent impacts on Blue Carbon sequestration in shallow coastal seas.



More info...

Seminar Abstract

Although environmental requirements of submerged aquatic vegetation have been studied for years, reliable metrics for predicting their response to current or future conditions remain elusive. The combined effects of temperature, CO 2 , and light availability controlled by water quality and epiphytes were explored using GrassLight, a bio-optical model that provided a predictive environment for evaluating the interaction of multiple stressors on SAV distribution and density across the submarine landscape of the Chesapeake Bay. Model predictions were validated against in situ measures of spectral diffuse attenuation, SAV density and distribution. The potential for photosynthesis stimulated by ocean acidification to mitigate the effects of high summer temperature, water quality and epiphyte load on SAV populations growing near the southern limit of their distribution were explored. The model accurately reproduced the submarine light environment from measured water quality parameters, and predicted their impacts on SAV distributions throughout the Bay. It also reproduced the negative effects of warm summer temperatures on eelgrass (Zostera marina L.) distribution in the southern Bay, and demonstrated that CO 2 increases projected for the next century should stimulate photosynthesis sufficiently to offset the negative effects of thermal stress, even in the presence of epiphytes. Thus, improved water quality should facilitate the survival of SAV populations in Chesapeake region, even in the face of a warming climate.

Seminar Transcript

>> What I'm going to talk about in the next few minutes is some of the modeling work we've been doing, looking at the impacts of water quality and climate change on SAV. And today I'm going to focus on SAV in the Tidal at Chesapeake Bay. I want to acknowledge my co-author, Victoria Hill, who's been instrumental in all of the work that I'm going to present today, as well as an unindicted co-conspirator here, Chuck Gallegos, who has been really instrumental in helping us move the research model based on the plant physiology into the realm of water quality management areas, stuff we're going to be talking about today. The work has been funded for many years up and down intermittently by NSF, by NOAA Sea Grant, and most recently by EPA through a collaborative modeling program that we've been working with in the last couple years, to try to improve the water quality models in shallow water cribs up bay. Because where we know that the current CHPD model is not working all that well. Okay, so there are two motivations for this work. One of them is the basic science that I've been working on for the past 30 years of my career, which basically involves linking hydrologic optics, that is radiative transfer theory of how light propagates to the ocean with the physiology of the plants to develop fundamental understanding, so physics-based understanding of how climate changes or how environmental forcing impacts, in this case, aquatic photosynthesis of SAV. And more recently, because we've been getting been having increased success with our fundamental models, we're trying to improve our ability to model and manage the water -- the impacts of water quality in shallow water resources in the Bay. And as said earlier, the existing Bay model's not working well, and so we got a little bit of money from EPA to help transition this into that, into more of a management realm. I don't have to tell you about the importance of climate impacts on SAV, everybody in the room pretty much knows that. High light requirements makes them vulnerable to poor water quality and sensitivity to summer heat stress, as JJ has mentioned already. And then when we lose the seagrasses, we lose habitat structure, sediment stability, blue carbon deposits. We get productivity shifts from the plant -- or from the benthos to the plankton that has cascading effects to the higher trophic levels, and you get changes in sediment biogeochemistry and everything else. So as Bill and JJ have said, the seagrasses are the canary in the coal mine, but for the Chesapeake Bay, I would argue that they're more important than that, they're not the canary, they're the keystone organism that keeps everything else in place. But you remove the seagrass, and the system completely changes. And it's not clear to me yet even today that once we've lost seagrasses, that simply improving the water quality is going to bring things back to the way they were. Although, JJ's work with the seeds suggests that recovery may be faster than one had expected. All right, so climate change, it's going to have a number of impacts on SAV. Warming is going to increase summer heat stress. These heat stress events are probably going to become more frequent. And water quality, here's the issue, is it improving or is it not? I got a telephone call a couple of weeks ago from somebody saying there's a new report out that water quality is improving in the mid parts of the Chesapeake Bay and people are going out and wading into the water off their properties and they can see their feet for the first time in 20 or 30 years, [inaudible] are improving. I said, well, that's news to me. But in any event, at least most recently in the past few months, for whatever reasons, in some areas water quality seems to be improving. Which way it's going, up or down, I think is an open question at this point. I don't see any huge evidence that water quality has been making dramatic improvements throughout the Bay. But in addition to those two events, water quality and temperature, which were largely viewed as primarily as insults on seagrasses, there's also this process done as ocean acidification and I actually like to refer to it as ocean carbonation. Because increasing CO2 availability in the environment modifies eelgrass' response to a variety of things. It increases photosynthesis, and therefore improves carbon balance. It has effects on survival and reproduction, plant size, growth, below-ground biomass. And we've been doing long-term experiments on this and, basically, we can take, we've been able to show in the long-term experiments that, that the increased photosynthesis and positive carbon balance that you get from increasing CO2 availability actually leads to long-term survival and growth, and I'll show you some pictures of that. But basically, if we can take seagrasses back to [inaudible], the argument is, it might be able to tolerate those warmer temperatures better than we think. So, we've been combining the biology and the optics and physiology into a model to predict SAV responses to environmental forcing and the model we're now calling GrassLight 2.13, which is publicly available. If you want a copy of the model, we can readily provide it to you, just send me an email and I will point you to the dropbox location where the model exists. There is a user manual and, if you want to get down in the weeds about what goes on in the model, we pumped [inaudible] in 2015. That goes into the model in, the mechanical details in quite a bit on mechanical [inaudible] go into quite a bit of detail. I'm going to skim over that. Today, just to show you basically what we're doing, what we're doing is basically taking Chuck's most recent water quality model, which allows us to predict the spectral attenuation coefficient of water from knowledge of chlorophyll, CDOM, absorption of suspended matter, and generate from that spectral distributions of, of light in the submarine environment. And this is the key factor here, we're now able to not just get to PAR or [inaudible] but we're actually able to predict the irradiance as a function of depth in the water for a variety of water quality conditions and we've been testing this algorithm in the Chesapeake Bay, in the Gulf of Mexico waters and Florida bay waters and, most recently, up in the Arctic. And the model is doing a remarkable job of retrieving estimates of diffuse attenuation from those simple three parameters, you know, the chlorophyll, if you have an idea about turbidity, and you know CDOM, we can give you spectral Kds. That means we can now then integrate that with a vertically, a rigorous vertically structured model of the seagrass canopy and the optical properties, combine that with the physiology to generate what we call a sparse model of the light limited distribution of, in this case eelgrass, as a function of depth in the water column. And so, what we're able to predict is the leaf area index of the, of the, the sustainable leaf area index as a function of that light environment through the water column. And we end up with a relatively simple model that we can often parameterize with a, with a first order or second order polynomial equation that makes it really easy for then, to then combine that with a map of bathymetry and plot the distributions of SAV across the submarine landscape. Keep in mind here that these are -- what we get out of the model are units of absolute biomass, so it's not a percent cover, it's something you can convert to grams of carbon if that's you want to do with it. Okay, so we've been playing with that model in the Goodwin Islands as a study site to test the model's utility and we've, there's a long history of SAV in this area. JJ and Ken Moore have been looking at it I guess since day one, perhaps before that. We know it's vulnerable to thermo- stress in this areas. And because of their work, there's a good time series of water quality measures. There's also a good time series of SAV abundances to compare to model predictions. And then we went out in 2013 and just mowed the lawn and made a very detailed bathymetric model of the area so we'd get high resolution DEMs, about ten meter horizontal DEM and about 10 centimeter in the vertical. So then, we'd take the water quality data from a, that we had measured, as well as some of the water quality monitoring data from the Goodland Islands NERR site and run the model in those conditions. And what we'd find is that typically, as you would expect, seagrass density decreases with depth. The distribution is limited to about a meter and a half. And this is our map of the model predictions for average summer conditions, temperatures 25 degrees C and current atmospheric pCO2 concentrations, about 400 microatmospheres. The blue colors indicate the [inaudible] into deeper water and then the shades of green to yellow indicate the different densities of seagrass, [inaudible] are indexed and we have mapped these four classes to JJ's four classes one through four for comparison. And this is the resulting map that we get driven by the bathymetry. Now, if we compare that to the VIMS map and we ran our, this, we did this one with the 2000, 2011 water quality data, if we compare that to the VIMS map, you see we get pretty good coherence. In fact, when we do an error analysis, what we find out is we get about 90% coherence presence absence. The density coherence is a little bit -- it's not as good, it's about 60 or 70%. Part of the reason we see that the density coherence is not as good is because when we run the model, we calculate the density on a pixel by pixel basis for each pixel based on the bathymetry. And JJ's crew basically groups these into four broad categories. So were seeing a lot of texture here that they're not picking up in their analysis. But this is something that we can be looking for into the future, is seeing how well maybe. But generally, we're getting consistent results, so we're pretty happy with the fact that where we predict the model, where the model predicts seagrasses are growing, that's were we see the seagrasses. And so here's that scenario, again, basically a cool summer temperature 25 degrees, present-day CO2 concentrations. Now, if we heat the water up to 30 degrees, the model predicts that the SAV will die-back. So there it is cool conditions, warm conditions. And you can see the color change is indication of the die- back. All right, so we've got the model working and it's generally consistent again, qualitatively, at least, consistent with what we know historically about the effects of water temperature on these things. And then again, these are average conditions, and what we're showing here is maximum sustainable SAV biomass. It's not trying to simulate any particular day or month, but sort of a optimistic upper end of what the distribution would be. If we add CO2, if we now run the model in a CO2 doubled environment, which is something predicted for the next century, this is the grow-back we would see under the same warm conditions. So that's warm without CO2. That's warm with the CO2 doubling. And that's warm summer temperature with the CO2 quadrupling in the environment, okay. So that suggests that the temperature effects are going to be moderated by the confounding effects of increased CO2 availability in the water column as well. All of this is to say that, although it's clear the temperature is a potential stressor and we need to be worrying about eelgrass, the increasing CO2 availability to date, that we've been working on to date, suggests that we shouldn't abandon efforts to restore, protect, and preserve SAV resources because there is significant likelihood that the increased temperature sensitivity will be offset, at least to some extent, by rising CO2 availability in the water. Okay, and our experimental results support the model predictions. We've done some long-term growth experiments. And this is what happens, we start out with plants at about 50 shoots per tray in these experiments and we grow them over the long hot summer. And what happens over the long hot summer is we lose about, over half of the shoots and they were about half the size they were when they started. We take those same plants and return them to the [inaudible] over that same condition. These are the plants that are growing in a high CO2 environment. Basically the same water supply, same temperature, everything else. The only difference is the CO2 concentration in the water. And we're in the process of just getting this paper written up and submitting it for publication very quick, shortly. Okay, so the model predicts eelgrass performance in the polyhaline regions of the Bay, and we've also used the model in other places. So we're gaining increasing confidence that, at least for eelgrass, the model's doing a pretty good job. Most recently we've been tasked to ask the question, will this work for SAV in some of the fresher parts of the Bay. So as part of the EPA work, we've been asked to apply this to the lower Chester River. This was a site picked by the team in order to try to model the water quality conditions and fluid dynamics in these small tributaries. It would not have been my first choice as an SAV modeling site for obvious reasons. First of all, it's highly turbid. These are sort of average suspended matter loads, or typically higher than 30 milligrams per liter. And it's also got a lot of chlorophyll in the water. So two very highly attenuating substances in the water. I would automatically predict that you're not going to get much depth distribution of the water. And we have gridded bathymetry, 30 meter gridded bathymetry, for this area. So we ran the seagrass model, and I'll show you some of those results. So here is just the Chester River with the three meter DEM, which shows the historic -- [inaudible] three meter boundary for potential distribution of SAV under ideal water quality conditions. And then here is Chester River SAV distribution from the VIMS data set. Again, here's another use of these VIMS maps is to fund exploratory science, as well as doing monitoring. Here you can see it in 2011, there's a little bit of SAV up here and very little bits along the shore. But mostly you see that it's blue. And then if you look at 2013, that SAV has disappeared. So as JJ has shown, it's highly dynamic in these areas. Some years you have reasonably good populations, some years you have reasonably very low populations. We find this is highly variable, it depends on water quality, among other things. And if we run GrassLight over this, we see that GrassLight is giving very similar predictions to where the SAV would be as to, compared to what the VIMS monitoring team is saying again. So again, with the model, we're getting good coherence between where the model predicts the SAV should grow and where we're actually seeing it. Okay, so we were asked most recently to try to do -- apply GrassLight to some DNR restoration sites. And I just want to show you this. This is some of the stuff that we talked about doing for you. We just got it done last week, so here are the results. Okay, so we're looking at three sites here, three DNR restoration sites. We're looking at Back Bay, Magothy River, and then a site in the Upper Chester River. And this table is information that you probably can't read, is basically giving you the water quality conditions for each of those sites. And so what I did was I got the water quality data from Brooke, and I went through and just quickly grabbed the median values for water quality for whatever she gave us. And at the same time, Mike Kent said, well, if you're running it for Brook's areas, why don't you run it for Susquehanna Flats and let's see how the model does. So I got some data from Mike Kent's group as well, and we got data for Susquehanna Flats area, we've got four water quality monitoring sites. There's a monitoring site at Harve de grace right at the edge of town. There is a monitoring site inside the center of the Susquehanna Flats bed, one at the edge of the bed, and one at the DNR common site, which is actually pretty close to the inside of the bed here. I don't know if you can see this very well. The important point to make here is that they are -- each of these sites represent different water quality conditions. And, as you might expect, the water quality conditions inside the center of that huge bed are much, much better in terms of clarity of water than they are at any of the other sites, okay. And you'll see the effects in the model in just a minute. So I ran them for these conditions, and this includes solidity, chlorophyll, turbidity, dissolved organic matter, temperature, pH, and then we kept the biological the plant canopy model conditions all the same. And I need to point out that this is our eelgrass model that we're running here. So I'm not adjusting it for any species composition or anything else. We're using eelgrass vertical biomass distributions and eelgrass leaf optical properties. So it hasn't been tuned for any particular species, although the model could be tuned if those data were available. All right, so, here are the DNR restoration sites, Back Bay up here, Magothy River, several sites around the Magothy River, and then the Upper Chester River. Here's the mouth of the Chester River. This is a site way up here, way up in the upper edge. This plot simply represents the attenuation coefficients for this site. So this is wavelength across this axis, an attenuation coefficient. And you can see that the attenuation coefficients are high, they range from a low about two in the green to about six or seven in the blue. The water's very green, very turbid. And this is the resulting model runs for these sites, showing that basically less than a meter, you should be able to see depth of water. And this is average water depth. You should be able to support seagrasses. But at these three sites, the densities drop off quite dramatically as you get deeper than about a meter in the water. So we take this sparse model resulting from iterative runs of the GrassLight model and propagate that across the 30 meter DEM, the USGS 30 meter DEM site, and here's the results. So here's a potential SAV habitat distribution for the Upper Chester River, all along this main stretch. I think this is -- no, this is not the Corsica, the Corsica is down a little bit farther than that. But anyway, you can see that there is a lot of potential habitat in these shallows, where the SAV might be growing. And this just a blowup of that, showing the GrassLight model predictions for this site dance along the shore, and then as you get into deeper water, the density starts to drop off. We pulled down the VIMS survey data from 2015, and again, we see pretty good coherence between where the model says pretty confident that we're going to get SAV and where the VIMS model says it was for 2015. These are the color varies. The yellow polygons you see here represent the aggregate SAV distributions in this part of the river from 1972 to present, I think. So the model is picking up those same areas pretty well. There's a lot of coherence between distribution. Now, one thing is, up in this skinny little river, a 30 meter DEM is not very precise in its distributions. So what's key here is one key factor in getting the spatial mapping right is having a really good DEM to run the model over. But I think with an improved DEM we could see some improvements. We're showing at least our DEM has a very shallow part of the river right here. And I'm not sure if that's true based on looking at this map here. It looks like it may have eroded out. And this may be an older DEM that may no longer be valid, so keep that in mind. But basically, we're seeing pretty good coherence between where the model says there should be SAV and where they presently are. And this is the restoration site, which is the proposed restoration site, which, Brooke, I think is right about here. So it suggests that there up and downstream, in these flats, there's potential for increased SAV, that's one of your sites where you want to plant. >> We put in seeds last week. >> Okay, good. All right, so here are the results for Back Bay. Same caveats, and now you can see that our 30 meter DEM is not all that good because it doesn't really it doesn't even know that the you so know this Spoil Island here exists at the moment, it's older than that. And what we're seeing is a lot of SAV in the shallow areas. And here is the proposed restoration site for this area. Now, the model is showing that we're getting SAV on both sides of the island and both sides of the shoreline, and that's because we don't have any exposure or fetch limitation on the model, which is something we probably want to add. I suspect growth out here would not be tolerated or really here wouldn't be tolerated under conditions of high wave energy, the high wave energy on open shore. So as you look at it, and when you compare this to the VIMS SAV distributions, focus mostly on the stuff inside this little embayment. And of course, this is dry land across here, but our DEM didn't know that. All right, so here's the comparison then to the DEM's SAV monitor. And again, pretty good coherence. Historically, a lot of SAV in this little cove and along the interior shore where the model says it should be growing, up here, and along the interior edge of this island, again, the model is saying should have high densities of SAV in these areas. Then again this outer area where we're lower predicting, is probably an exposure issue that has very little data, suggesting that there is SAV in these outer banks, this could be an exposure issue that could be added to the model at some point in the future. In fact, we're trying to get on that. Here's the prediction for Magothy. Potential restoration sites are these two small coves right here. Unfortunately our DEM didn't go into those coves. But if you can get us a better bathymetry model for this area, we could re-run it at a higher resolution and provide you with maps of that, if that's what you want to do. The other site is up here in this little cove in Magothy. Again, we're getting a lot of SAV on the outside in which is probably not realistic in a wave-exposed environment. But in the Bay, it covers an extensive area of it. And again, comparing it with the maps of JJ's team, again, we see a lot of coherence, at least with the historic distributions of SAV in the Bay throughout here. And then the colors represent the 2015 distributions in the Bay. So based on the median water quality conditions in these areas, would suggest that there is potential for increased SAV habitat in these sites that we are proposing to revegetate. Okay, so now, in the next couple minutes, I just want to apply this to show you some of the results that we did from Susquehanna Flats. Over here on this plot is the attenuation coefficients for the proposed restoration sites that we just went through. These are the corresponding attenuation coefficients for the Susquehanna Flats area. And you can see, rather than ranging from four to seven on the attenuation coefficients, now our attenuation coefficients are two to four. So the water transparency up in Susquehanna Flats area is about twice what it was down there in the Chester River part of the Bay. Here on the GrassLight vertical distributions that I showed you already, hear are the corresponding vertical distributions for Susquehanna Flats scaled to the same depth. I realize you can't read it, but this is three meters here. So the model's predicting SAV within the bed at least to go down to three meters. So the black line represents the run with the water quality from the inside of the bed. This is the DNR common site, which is very close to the inside of the bed. Then we get to Harve de grace, the edge of the bed. And then since Mike's team had provided epiphytes data, we ran the model with epiphytes, which we can include in the runs. And we see the results here. And so quickly, you've sort of gotten a sense with these pictures now, I'll just run through them very quickly. But here's the model predictions for the distribution of SAV if the water quality throughout the region were equivalent to that at the water quality monitoring station inside the bed in the center of the bed. If we run it with the common site, which is not quite as clear as the inside of the bed site, and the yellow dot is an indication of where those sites are, you see you get a bit of a retraction in seagrass density. It sort of starts to shallow as the water gets more turbid. We go to the Harve de grace site, you get more retraction. And if we go to the edge of the bed site, it looks like this, okay. Now, if we add epiphytes to the model simulation, now the distribution of SAV looks like that. Now I'm going to show you the comparison to the latest SAV map from 2015 from the VIM survey shows that this patch is remarkably coherent to the distribution that you see here in the darkest area, including the presence of a small channel between the main bed and these small patches over here. Here's that little channel. The presence of SAV over here on this bank. And the retreat of SAV out of this little bay here, which was much more heavily vegetated in the clear water conditions, and which, according to the VIMS maps, indicates that there used to be extensive SAV distributions up into some of these areas. So it looks like we're starting to get pretty good coherence with it. The model seems to be responsive to the median water quality conditions that we are providing. And, again, this is sort of an optimistic steady state model run, having not done a lot to it in terms of trying to tune the model to these different environments. So quickly, just back to the Lower Chester River, here was the original model run distribution with the median water quality data for the Chester River. So we were playing this game with it saying, okay, what if something were to happen and one could improve the water quality in the Chester River to be equal to Sandy Point, which is across the bay near Anapolis, okay. And so if we drop the suspended matter load from 1230 to 10 milligrams per liter, and the chlorophyll we cut it in half, that's the potential SAV habitat distribution, okay. So here's a way you could use the model and say, if you really want to target water quality, you can now use the model and you could start to get estimates up, not just aerial coverage, but also how much grass you're going to get in terms of density from the model calculations. Okay, so at this point I think, you know, we're becoming increasingly confident that GrassLight is a pretty good tool for predicting SAV distributions in the Chesapeake Bay and in other places. And it's a good predictor of the light-limited distributions. We can use the median values of four simple parameters, chlorophyll, suspended matter, temperature, and pH. And there's good evidence, playing around with our model, that in the center of that bed at Susquehanna Flats that those plants may actually be CO2 limited during certain times of the day because the pH is going up above nine, and we can actually simulate that in the model, I kind of showed you that today. So the water quality requirements aren't massive here, they're fairly simple. And basically we can use standard water quality monitoring data. We're using eelgrass morphology and optical properties. We did not tune this, we did not make a ruppia model, we did not make a [inaudible] area model, but we could do that if the data were available. For these runs we assume that the light environment was equivalent to the summer solstice, so it was the longest day of the year where we ran these models. Could also run it for the equinox and see how that would affect the differences. And we were using a 30 meter DEM for all these runs. What we know is the pH may limit the density in some areas. Epiphytes are probably really important in these areas, that could be included in the model. And ultimately the predictions are only as good as the underlying bathymetry in which we layer them. Because the GrassLight model gives you that polynomial distribution with depth that we then populate across the space. So all that texture comes from the DEM. So knowing the bathymetry is really important in terms of making model predictions. Just a few caveats at this point. Important things that GrassLight is not considering at the moment, are the optical properties and canopy architecture of freshwater SAV, wave exposure and fetch, and other things that might be limiting in the distribution of the plants besides light. This includes sediment characteristics such as sand versus mud, organic content, sulfide content, water column anoxia, and other things. But those are we view those as incremental additions that we can put onto the model to help improve the reality, and make the model more realistic for certain management applications. But I think right now as a light- limited model, this seems to be working pretty good. And in any event, we'll try to incorporate some of these other things. And so I'd just like to close, paraphrasing from George Fox, although we think the model's working pretty well, we know it's wrong, but that's okay, we also think it's probably useful. So thank you. [ Applause ]

Seminar Discussion

Coming Soon