Susan Hubbard: Effects of Climate Change on Watershed Dynamics

Thank you. Jeff, can you hear me? Oh yes, thank you for that very nice welcome. I have to say it's been a pleasure to be here today. I visited Jeff at his previous institution, but this is my first time in this whole area, and it's been great fun talking with different groups this morning. I look forward to continued conversation. So today I'm going to talk about some of the work of our group in using geophysics to understand, as Jeff said, the critical zone. This is the part of our Earth's crust that really has a lot of biological activity, from the top of the bedrock, sometimes within bedrock, to the top of the vegetative canopy. As Jeff said, this has been a focus in the United States and in Europe and many other areas, recognizing that this part of our Earth's crust is difficult to look at behavior in one particular compartment. There are a variety of interactions within and between these compartments that are of interest as we think about how our Earth is evolving, and it's important to think about these together. So I'm going to talk about some of the work we're doing using geophysical methods to understand this critical zone. There are many people involved in the work I'm talking about today. I'd like to particularly mention two early career scientists that work with me: Baptiste Dafflon and Haruko Wainwright. But I'm going to set some of this work in the context of two large team-based projects. I want to mention that I'm only talking about the characterization effort of these projects, but you'll see me mention them. I'll talk about an Arctic project led by Stan Wullschleger at Oak Ridge National Laboratory, and then a new project we call the Watershed Function project that I'm leading, which officially starts next week, so it's in a boot-up mode. My objectives are twofold: I do want to tell you a little bit about these two large team-based projects, what's driving them, what questions we're asking, and then I'll zoom in and talk a little bit about the geophysics. The reason I want to tell you a little bit about these is because we're really open and interested in collaboration. If there's something that you find of interest that makes sense to compare with the work you're doing, I'm very interested in talking with you about that. But then, of course, the main effort here is to really describe some of the work at Berkeley Lab using our geophysical approaches to understand these bedrock-through-canopy processes and the organization of ecosystems and watersheds. First, I want to give a little bit of a flavor for what's driving our research in both of these projects from a pretty high level, starting with thinking about ecosystem feedbacks to climate, which is really what's driving the Arctic project I'll talk about. It's the lack of understanding about how ecosystems will serve in the coming century in retaining the carbon that they have. This graph shows different IPCC models predicting what CO2 distribution will look like in the atmosphere, the ocean, and on the land surface up through the end of this century. Something that's pretty obvious is that in the case of the ocean and the atmosphere, CO2 continues to go up, as we all expect. But what's sometimes surprising is that there's so much uncertainty in terms of the way that the land will retain or release carbon through the end of the century. Some models predict that terrestrial ecosystems will actually give off carbon, be a huge carbon source; some models predict that these ecosystems will be a huge carbon sink. There's a lot of uncertainty in driving how terrestrial ecosystems will evolve in this coming century, and that's what drives the Arctic project I'll tell you about. On the other hand, we're not just concerned about how our ecosystems evolve and how they feed back to climate in terms of greenhouse gases. We're also very interested in how climate, weather, land use, and other perturbations to the land surface influence the way a watershed functions. About halfway through this presentation, I'll switch gears to talk about the Watershed Function project, which is really looking down on the system, trying to understand how perturbations that we're increasingly noticing on these terrestrial environments impact the way that watersheds distribute water and also export nutrients, carbon, metals, and other species in the system. Of course, understanding how watershed systems function is extremely important if we're going to understand not only water distribution but also hydropower, contaminants, and a variety of other services that a watershed provides. So that's a high-level look at the drivers for the Arctic and the watershed systems. I want to jump now into talking about the Arctic system in particular. This is a map of the Earth surface; you can see that there's an awful lot of permafrost shown in purple. Permafrost stores a huge amount of organic carbon. Actually, there's more organic carbon found in permafrost systems than anywhere else in the world's soils, and twice as much as is found in the atmosphere. So a huge stock of organic carbon is locked up in this frozen permafrost system. Let's take a look at a cross-section through permafrost just to make sure everybody recognizes the layers of a permafrost system. Permafrost is usually the deeper layer that's frozen for several successive years. On top of the permafrost, between the ground surface and the permafrost, is this little layer called the active layer. This is the part of the ground surface that seasonally thaws and freezes. What happens when it thaws is that the microbes present in that active layer use that organic carbon as an energy source, and in doing so, they respire CO2 and methane, greenhouse gases, to the atmosphere. The concern is that with warming, and the Arctic is warming faster than any other place on Earth, that active layer is going to get deeper and deeper, potentially exposing these microbes to very bioavailable carbon, which will lead to a huge burst of greenhouse gas into the atmosphere. So that's the threat. On the other hand, of course, when it warms, vegetation grows that can take up CO2; it changes the albedo or the energy balance at the Earth. So there's a huge amount of uncertainty in what will happen in this Arctic ecosystem that has so much organic carbon with warming. Will the microbes win out and produce a huge burst of greenhouse gas, or will the vegetation win out and remain a carbon sink through the end of the century? We have a new Department of Energy project, I should say it's not that new, it's several years old now. The Department of Energy tends to fund a lot of team-based projects in the environmental and climate sciences that are very much modeling-driven. Both of the projects I'm talking about today fall in that mode, where we are developing new multiscale, multiphysics models and really prioritizing our experiments and observations in the context of that model to parameterize it, test it, validate it, and improve process insights. The Next Generation Ecosystem Experiment in the Arctic is a large team-based project where the real goal is delivering a process-rich ecosystem model whereby the evolution of this Arctic ecosystem in a changing climate can be modeled at the scale of a high-resolution Earth system model grid or pixel. If you can see these figures, on the left we're looking at some of the landscapes we'll be talking about when I start talking about our Arctic project in Barrow, Alaska. It's ice-wedge polygon land. If we talk about integrating processes that are happening on this ecosystem up to the scale of a high-resolution Earth system pixel, that's a pretty big challenge right now. A pixel of an Earth system model might be on the order of 10 km by 10 km, and what we're looking at here on the right is geomorphology at the site that shows two different types of features that influence the way water is distributed and biological processes, including microbial respiration, occur. You see thaw lakes and in the background the brown part you see permafrost or polygon landscape. If we look into this little box, we move in closer in space. What we see when we move in closer are these ice-wedge polygons. There's differing water distribution in and around that that drives vegetation. This is typically the scale where we collect our information to understand how this critical zone functions. If we take a core from beneath one of these polygons, we take it back to the laboratory. What we're looking at here is just a Barrow Arctic soil aggregate that was created by Marco Volto at Berkeley Lab using our Advanced Light Source. We're digging into the center of this aggregate and taking a look at the Earth from a microbe's perspective. Now we're in the interior of a soil aggregate. I think you can imagine that if water is in one part of that pore space and organic carbon and nutrients in another, the microbe may not have all the constituents together at the same place to perform the metabolic function of respiring and delivering greenhouse gases. I really show you these pictures just to emphasize that the processes that are driving microbial respiration and degradation of this organic carbon in the Arctic that leads to greenhouse gases are really happening at these length scales, and yet the challenge is to deliver this information up to a climate model that is parameterized at much larger scales. It's a pretty big challenge, and that's really the focus of this team effort: to try and not only develop models but develop process insights and information about the system that can really help us meet this goal. The objective of the group that I'm involved in is really focused on the characterization aspect, so that's all I'm going to talk about in the next several slides. But really, we're interested in understanding how and where do these fine-scale processes make a difference. How can we characterize the organization of the landscape and tractably provide information about the processes that do control microbial respiration into models? Let's take a look at this very first field site where the Next Generation Ecosystem Experiment team has focused in the last several years. This is in Barrow, Alaska, the northernmost village in the United States. It's very cold, very continuous thick permafrost. Our typical approach at field study sites with environmental geophysics, which is really my background, is to do nested spatial scale investigation. We come in with more remote sensing data and try to look at variability of below-ground and above-ground processes over large swaths, and typically we zoom in and try to collect finer-scale data in different parts of the system that might have different gross signatures. I'm really going to be talking about today data sets that are at these spatial scales. In fact, this is one of our intensive study sites where we have a lot of data that covers this region. It's still quite small, about a kilometer by kilometer, but we also have some very intensive transects where we've collected a lot of geophysical and other point-type measurements to really understand variability of this active layer and microbial respiration and think about how to scale that small-scale information back up into larger scales. Of course, we collect cores, bring them back to the laboratory, and do a lot of analysis of real physical samples as well. We also collect measurements at various times during the season. This is a very dynamic, in fact the most dynamic environment I've worked at from a hydrological perspective. It goes from a very frozen, long, cold, dark winter to snowmelt, which really rapidly distributes water across the land surface, a rush of a lot of water. It dries out pretty quickly through a shortish growing season, and then we go up into a freezing state again. So we tend to try to get out there about four times a year to understand how the system is responding. I should say that we're in early stages in this process, so we are using this kind of dynamic freeze-thaw annual cycle to really try to understand the processes. We are trying to understand those and put those into models and use the models to project forward, as well as some other strategies for thinking about whether this Arctic ecosystem will remain a carbon sink through the coming century. We've collected a lot of data and a lot of different types of data. These show members of our team collecting geophysics by foot, by snowmobile, a variety of point-based measurements. We'll take a look at some of those, but clearly the things that are of interest for understanding how microbes behave are things like active layer thickness, soil moisture, soil temperature, redox poise. We've collected measurements from the ground surface, from above ground, in the water bodies, mostly on the terrestrial environment. It's been a lot of fun having different types of data, and I have to say the data have been really, really beautiful. What I want to do is tell you a little bit about what I would say is our conventional hydrogeophysical approaches, which is using one particular type of data set to estimate a property or a process associated with one compartment of this critical zone. I mentioned this bedrock-through-canopy critical zone, but oftentimes that includes bedrock and deep groundwater, the vadose zone, the soil layer, the land surface, the vegetative canopy. Traditionally in geophysics, and actually probably in most measurement fields, we do focus on some measurements in one or two compartments of that field. So what I'm going to do first is just very quickly show you some examples of using geophysics to estimate a property: snow at the land surface, active layer thickness in the active layer, soil moisture, and some permafrost characteristics. But really what I'm most interested in, and I talked today in the journal club about, is kind of new approaches where we can collectively consider at the same time the different sort of information that we get from both our direct measurements of things like soil moisture, soil temperature, geochemistry, and our indirect signatures that we get from geophysics data in one fell swoop. I'm not going to take the time to talk in detail about the geophysical approaches, but I did just generally want to mention that if you're interested in using geophysics and translating that into properties of interest, there are several key steps that we typically take to make that geophysical information useful. First of all, I have a question mark here because I think many of us recognize the value of geophysics: it can provide spatially extensive and sometimes non-invasive, typically minimally invasive, information about the subsurface, so a lot of good information. On the flip side, though, it's not giving us exactly what we want. It's kind of a proxy; it gives us information about electrical connectivity or seismic velocity or electrical resistivity, and what we really want is information about the geochemistry or the lithology or the hydrology, certain properties. So there's a series of steps that we typically take if we are trying to use that geophysical information to inform us about this critical zone. One is collecting the right type of data given your target of interest. This is something where your geophysical colleagues can help you if you have an idea about what you want to measure, and they can match that up with what a geophysical signature will be most sensitive to. A very important part is developing, through theory or laboratory experiments, what we call petrophysical relationships. These are somewhat akin to pedotransfer relationships for soil scientists. These are relationships that link those geophysical observables to your property of interest. This might be, for example, electrical connectivity linking it to soil moisture. And then the last component is where our group tends to spend a lot of time: really how do we integrate these different types of measurements, the point measurements, the non-invasive but spatially extensive geophysical measurements, into a coherent framework so we can actually visualize the estimated properties and associated uncertainties. Again, I'm not going to go through that in any detail, just a quick fly-by of hydrogeophysics and the different parts of it that are important to consider. What I would like to do is just give you an example again of traditional hydrogeophysical approaches, using one method to estimate one property, just so you get a sense about some of the geophysical attributes and also get to see some of the beautiful data that we are getting coming out of the Arctic. The first one is using ground-penetrating radar to estimate snow thickness. Snow thickness, I mean probably everybody here is very familiar with snow and its importance in the Arctic. We're interested in snow thickness and even its density variations because that's the water available for the next growing season. If we can understand snow distribution and how it relates to wind pattern and microtopography, characterize that, put that in a model as initial conditions, we can run that model forward and understand how that water gets distributed on the land surface in the subsequent seasons. Radar is a perfect tool for this. Radar is basically a method that sends out an electromagnetic signal. Under conditions where we tend to use radar, that signal travels as a wave and it reflects off of boundaries where there's a contrast in dielectric constant. The base of the snow and the top of the ground is a perfect example of an interface that'll send a wave back up. So by looking at the travel time of the radar, we can estimate snow thickness. What I'm showing down here on the bottom left is an estimate of radar-obtained snow thickness compared to actual where we've gone around and measured how thick the snow is. What I'm showing over here is actual measurements along this intensive site in two different study blocks, probing that with a snow probe. If we add radar with that, we can see we get much more detailed information. We can see how that snow is distributed as a function of microtopography. If we couple our radar information in a Bayesian algorithm that actually uses the topographic information from the lidar behind it, we can start to get very full coverage of snow, maybe more than you even want, but at least you start to understand how snow is distributed related to topography and can potentially feed into models. So that's an example of using a geophysical method to understand a land surface property. This is one example of using geophysics to understand an active layer property. Actually, I'm going to mention two here. Here we're looking at an electrical resistance tomography profile, about 500 meters long, about 5 meters deep. They're actually the same profile; this one's looking all the way down. So this is a very little active layer at the top, which is kind of blown up on a horizontal scale here. So this is the active layer. Beneath it, in the red, are basically where we see very dense ice or ice wedges. Beneath it, we even see some information about the saline permafrost. We collect these data in two dimensions, in three dimensions, and really understand a lot because that electrical connectivity is telling us something about ice content, moisture content, active layer thickness, and salinity. We have to use point measurements again to interpret that because geophysics is non-unique. But we have done this work where we've extracted from electrical resistance tomography information like moisture content and active layer thickness, and you can see how it compares to real data that we've sampled using direct measurements. Finally, the last kind of conventional example I want to give, as we're looking deeper in this system, is from seismic. This is work that Jonathan Ajo-Franklin and a student, Shand, did, developing a new approach to use surface seismic waves by looking at the full waveform during the inversion. What we're looking at here now is much greater depth. Seismic can easily see much greater depths, and it's a great tool for permafrost environments. We're really looking at four different estimates of seismic velocity, shear wave velocity, as a function of depth. We see it's very high where there's permafrost. What we found surprisingly is that beneath what we thought was very continuous permafrost, there's actually very low velocity. Paired with what we know now from core data and electrical data, we've interpreted this to be 20-plus meters of unfrozen, and sometimes unsaturated, region beneath what we thought was permafrost. So there's a thin top permafrost layer and a very unsaturated, unfrozen layer. This is a marine area around Barrow, and we think the salinity incursion has depressed the freezing point. This is interesting; it's changed our conceptual model about Barrow, Alaska, and it could have ramifications for other marine permafrost areas too. What's important about it is that if we thaw through this very thin upper permafrost region, we're going to get a very quick mass wasting of this region beneath there that could have ramifications both for carbon transport but clearly for geotechnical. So that was just a drive-by. I just wanted to give you an example of how we use different types of measurements to sample different properties in different compartments of this Arctic ecosystem. What I want to talk about now is this approach that we've been developing over the last several years to try to combine different measurements together to characterize the organization of the landscape and property suites associated with different zones in the landscape. Actually, for the Arctic, this started several years ago when we first started getting these different data sets that I just showed you, and we were interpreting them in terms of different properties or processes. It was really clear from looking at the data that while there are different geophysical signatures and we are interpreting different properties, there was certainly a spatial organization to the data. I could tell that in some parts of the landscape we might always have lower electrical resistivity compared to another part, for example. So the question that we were asking at that time is twofold. First of all, are there unique distributions of geophysical attributes in the subsurface that have a spatially distinct distribution? And if so, or also in addition, how do the land surface properties vary over space? Let me back up for a moment and say that in the Next Generation Ecosystem Experiment, one of our underlying hypotheses is that this ice-wedge polygon landscape, which has very small microtopography, there's not a huge gradient, so the ice-wedge polygon landscape controls the distribution of water across the landscape, and that it would in turn control both the redox state, so the below-ground biological response, and also the vegetation growth, or the above-ground biological response. So here this was a chance where we had all of this data, and rather than taking the geophysics through these different steps of interpreting different properties, folding them into a model, and predicting it forward, the objective here was to understand if we could find unique suites of properties in the land surface and in the subsurface and understand their distribution. What we found out along one of those intensive transects that I showed a few slides ago was that there indeed was zonation. We could just look at the geophysical attributes, and actually in the subsurface we found three unique zones, unique geophysical signatures in three different subsurface zones. If we looked at the lidar, which is giving us information of the microtopography, and we extracted metrics from the lidar like the curvature of these polygons or the size from the center to the lowest part of these polygons, there was also a spatial distribution of the land surface properties, and they coincided with the subsurface. So what it turns out, and again this was one of our hypotheses going in, is that this wasn't surprising. This covariation of surface and subsurface and land surface properties is very much controlled by the geomorphology. Here we're seeing a picture of different types of polygons, and it turns out that these are different type of polygon zones: high-centered, transitional, and low-centered polygons. Here we're looking at an aerial view of two different types of polygons. I should have mentioned before, for those of you that are not familiar with ice-wedge polygon ground, this is land that is formed from this ground freezing and thawing. When it freezes, the ground sometimes cracks and water gets in there. When these ice wedges form, it kind of bunches up the land on top of the polygons to form a certain type of polygon. Eventually, if that polygon thaws, the ground kind of collapses, so you have high-centered and low-centered polygons caused by the health of these ice wedges. So this was exciting to us, and for the first time showed through a bunch of dense data that there was definitely covariance between above- and below-ground properties that was controlled by geomorphology. So we wanted to see if we could take this a step further and move out to this larger site. So this was the transect that we were working on in this last paper. Now we're looking at the entire site. Haruko Wainwright, who I mentioned at the beginning of this presentation, basically took the step of using that DEM from the lidar data to use a data-driven approach to delineate polygons. Basically, this is a watershed approach modified to work in polygon landscape. So now what we're looking at on this map is about 1,500 polygons. You can see different shapes and sizes. This is elevation. These are our two transects. Our question was: if we think that there's a covariance between above- and below-ground properties that vary as a function of polygons, what really controls what we care about in this system most, which is greenhouse gas or CO2 and methane flux? Is it the type of polygon, or is it small geometric or small features associated with these polygons, such as the center of the polygon or the rim of the polygon? So basically what we did is we took all of our dense data that we had along these two transects. Remember, we had a lot of geophysics, a lot of point measurements above and below. We now have this delineated polygon where we can understand what the different types of polygons are, what their shapes and sizes are. The first thing we do is try to understand what controls variability of properties at the site: is it the type of polygon or the polygon feature? So we did a statistical test here, just an ANOVA test, to look at a lot of different properties. We looked at how they varied as a function of polygon type: high-centered polygon, low-centered polygon, flat or transitional polygon, and also as a function of their certain geomorphic feature, whether it's in the trough, center, or rim of that polygon. What we found is that the polygon type had more explanatory power, or more power for explaining the variability of the properties that are important for mediating microbial activity and respiration of the site, compared to the polygon features. Now the next thing that says is that this opens the door then, because we can pretty easily characterize polygon type. We've already done a polygon map. We know the statistics of that polygon map. We know which ones are high, transitional, and low. We can take that landscape then and translate that lidar together with other information into these zones based on our information that we have along the transect. So now what we have here, and this is really the final interesting outcome of this particular paper, is that rather than having 1,500 different polygons, which by the way the modelers were interested in moving explicitly and mechanistically into reactive transport models to understand how they function and evolve with warming temperatures, now we basically just have three main zones. So this is the same region again. Ignore the yellow, which is a drainage feature moving into the site. We basically have red regions, green regions, and blue regions, or high, transitional, and low-centered polygon regions. Now we know from our intensive transects that we can distribute across these regions what the distributions of important properties are. So say, for example, in the blue regions, this is the distribution of active layer thickness, this is the distribution of soil moisture, same for the green and the red. The beauty of this now is that it offers potential for representations of these small-scale properties at larger pixel scale. So again, I just kind of plopped this on top, but now you see what a 100-meter by 100-meter pixel looks like. Even though those microbes are really active at much smaller, centimeters sort of length scales, maybe decimeter to a meter, we now are comfortable saying that we can represent the properties that are important for predicting that microbial behavior at much larger length scales. So this is the first time that we have tried this, and we have many other places that we're working at right now where we're testing this zonation concept. But before we leave this, I'd like to kind of check how well this method performed. What we did is we came back into this transect, actually we did it at both, but the next slide I'll show you a cross-section of this transect where we looked at the zone geophysically identified zonation: three key zones, one, two, three. My colleague Neslihan Tas and her team sampled the microbial ecology, so the community that's responsible for the respiration, and then we collected greenhouse gas measurements above ground from chamber-based chambers that are moved along the ground surface. So this is the result, or one very high-level snapshot of the result. Here we have one piece of data again. I'm showing you the ER line again, just one that helped to contribute to this zonation into three regions. These are some examples of characterization of the microbiome over in this high-centered polygon and characterization of the microbiome in the low-centered polygon region. You can't read these, but these are different phyla as a function of trough, center, and rim of the polygon. The important part to communicate is that, well, two things actually: the community is very different in these geophysically identified zones, their metabolic potential or genomic signature is very different, and this is a paper that's in submission right now. But also from the chamber-based measurements, we see that the greenhouse gas signatures are quite different too, as expected based on what microbial communities are there. So a lot more CO2 fluxes over here in this higher and drier region compared to the low-centered polygon where we see a lot more methane flux. So that's one of the two methods I wanted to communicate that I'm excited about from a geophysical front, because it helps us understand the organization of the landscape using kind of one fell swoop of interpreting above- and below-ground data coincident using a more statistical or data-driven cluster-based approach. I think it offers a nice venue for guiding our field characterization for parameterizing models. We need to test it other places, but I'm excited about that. The other thing that I'm excited about is that once we identify these zones, what we've been doing now is moving in and developing intensive monitoring stations. Meaning that now we know that there's these different zones in the landscape, and we want to understand how that critical zone functions and how it responds to these perturbations like freeze-thaw that are quite dramatic and change both above- and below-ground response. So this is one example, and this is led by Baptiste Dafflon at Berkeley Lab, where we have transects of ER, in this case it's about 35 meters. We have a lot of soil probes that are moving into the ground. We have above-ground cameras that are looking directly at this transect all the time to measure things like vegetation growth and water inundation and snow buildup. Right next to this, I'm not showing this, but Margaret Torn at Berkeley Lab has an automatic tram that's collecting a lot of energy measurements and a variety of other measurements that we're comparing as well. So this is a really neat kind of autonomous system. It's running all year long, collecting data every day, and we have just a mountain of data. But I just wanted to show you one slide. This shows ER tomography as a function of season, basically as this is freezing up, and across different types of polygons. This is the land surface image. What I'm showing here is a correlation coefficient between greenness or vegetation, so above-ground property, and average electrical conductivity at about a half meter below ground. So again, that electrical connectivity is kind of a proxy for moisture and temperature and active layer depth, all the things a plant root cares about. What we see is that when we get into the growing season, we see fantastic correlations, up to 0.78. That says a few different things. It says that we can start to think about using plants as an indicator of soil dynamics and variability, which really opens the door because we can see a lot of the plant canopy from remote sensing. But it also really starts to tell us and give us this beautiful window into this world to actually watch this system breathe. So for the first time, we're really watching changes in moisture and temperature and vegetation and freeze-thaw state and salinity and all these things kind of happening as an interconnected critical zone system. That's been super fun. We have been kind of expanding this. I mentioned this during the journal club meeting this morning, using UAVs. It sounds like that'll be a big direction of development here as well. I think it's a fantastic opportunity to be able to do some of this intensive monitoring and then occasionally move across the landscape with UAV. So now I want to switch gears. That was just a little bit of a description of the Next Generation Ecosystem Experiment Arctic project that's interested in predicting ecosystem feedbacks to climate. Again, I didn't talk about any of the prediction or the work going on in the other teams; I really focused on the work our group is doing in the geophysics. But it really is looking up. And here we're looking at a new project called the Watershed Function Scientific Focus Area, which is the Department of Energy's term for a large team-based project. Here we're asking how does this watershed function, and how will it function with increasing perturbations, and what will it deliver downgradient as far as water, nutrients, contaminants at the watershed scale. This is our East River site. I'll talk about in just a moment. But I do want to say that this project officially starts on October 1st, so in a few days. But we've had the last three years working on another related project that we actually call Genomes to Watershed, although we work from the genome to the floodplain, as a kind of a springboard for what I'm going to describe in this new Watershed Function project. Another team-based project with a lot of great people involved. Underlying both our Genomes to Watershed and Watershed Function are, I think, what many of us would recognize as three really big challenges. One is that there are extreme complex interactions between plants, organic matter, dissolved constituents, migrating fluids. These happen within this structured critical zone, so they're not happening in a well-mixed fluid; they're happening in groundwater, in the vadose zone, in soils, at the land surface, in the vegetative canopy, in the terrestrial system, in the aquatic system, in the interface between. So these interactions are not only complex, but they vary within and across these compartments. This drives a lot of our work, a lot of the work in the whole entire environmental community. What I'm going to talk about is some of the work that we've done to try to understand genome through floodplain, tackling these questions, and then really where we're going in this new watershed question is to try to understand not only how these complex processes occur and where it's important to represent them, but how can we think about what the aggregated signature is of all of these complex processes across a watershed. That's really our new push. So I want to step back a few years and say just a maybe two or three slides on this project we call Genomes to Watershed. We really started looking at the very small scale, the genome, because as you saw on the last slide, it can potentially give information about what is the metabolic potential of the microbiome to perform functions that we care about, like cycling carbon and cycling iron and cycling sulfur. Actually, the Department of Energy has a really big effort in quantifying the microbiome and understanding metagenomics. We have seen through our work that that sort of information can be helpful at this site. This is at Rifle, Colorado. This is a floodplain in the Colorado River where we started our work. It's really a nice natural laboratory. It also was snowmelt-dominated, so we get a seasonal pulse of oxygenated water into the subsurface, so we can really look at how this floodplain responds to this changing moisture and redox conditions. Our real goal in these first few years was to develop and explore the value of a genome-enabled reactive transport watershed simulator up to the floodplain scale. The idea is that if we found that to be useful, we would move up to the watershed scale and try to consider how we move that to the watershed scale. We've published as a group a couple hundred papers basically at this site. Here I'm just giving a very few highlights just to kind of show the sort of things that we've learned because it informs the structure of this new project called the Watershed Function SFA. On the left-hand side, I think big contributions from this project have really been to quantify the subsurface microbial diversity and the metabolic potential, basically through the work of some of our microbial ecology team leads, Harry Beller and Jill Banfield. There's probably been more work done to characterize the environmental microbiome at the Rifle site than probably any other place on the planet. Basically, Jill and her group have discovered a whole new tree of life, basically using the metagenomic information. They've discovered and just started to scratch the surface on recognizing that there are key metabolic handoffs that happen, so one single organism doesn't complete a full carbon or sulfur or iron cycle; the community works as an ensemble. Also, through using metatranscriptomics, actually showing that even through this whole diversity of the microbial ecology, there's a small fraction that actually does the work. In this case, this is Harry Beller's work showing that chemoautotrophy was a key reaction pathway that we needed to include in the model. Also done a lot of work in understanding and using different types of approaches to quantify hot spots and hot moments in this floodplain system. This is an example of using geophysical data to map subsurface hot spots. Here they're naturally reduced zones or biogeochemical hot spots that influence reaction fronts moving through them. I think one of the more exciting aspects was the work with Tetsu Tokunaga and Jiamin Wan, who did a lot of work in deep vadose zone, deep meaning more than a half a meter. So for those of you that are familiar with climate models in thinking about carbon in the climate models, a lot of that really only considers about the first half a meter or so. This is working down through the deeper vadose zone, a couple meters down and into the groundwater. What Tetsu and his group found is that actually the vadose zone played an important part in the carbon cycle and actually contributed up to about 15% of the respiration from the deep vadose zone, so something that climate models currently don't consider. Finally, probably the highlight of our first three years was actually developing this genome-enabled reactive transport watershed simulator using some of this data to parameterize and test it, and actually using that model to predict carbon export to the Colorado River. What we found is that the use of that genome-enabled watershed reactive transport simulator improved prediction of carbon export to the river by 200% compared to conventional reactive transport watershed simulator. So we'll come back to that because that makes a big difference when we really want to move up in scale in terms of both length scale and complexity. So we are moving up in scale. I'll show you the pictures again in a moment. We're moving up towards the headwaters here in a headwaters catchment in the Upper Colorado River Basin, up from the Rifle floodplain. We've chosen to work in mountainous watersheds because while they're extremely important for a variety of different services, they're also extremely vulnerable to climate change. 60 to 90% of waters come from mountainous watersheds, and they're being impacted in terms of both their water discharge and resulting biogeochemical cycles in ways that are not very well understood at all. So we are in the Upper Colorado River Basin. This is actually an extremely important basin for the US. It provides water for one in every 10 Americans. It provides a lot of hydropower along the Colorado River corridor. It also has a lot of redox-sensitive contaminants that sorb or release as a function of redox state, so the amount of water makes a difference along that contaminant. It's been called an essential water tower for the world, meaning that its behavior has severe ramifications for the downgradient population, the whole Southwest of the United States. We're looking at the site again. This doesn't start for a few more days, but we're looking at an aerial view downgradient. It's about 300 square kilometers. It's a paired catchment with one part being a pristine catchment and another being metals-impacted, so there's some mining that's gone on and there's some contaminants. We're taking a phased approach, focusing on the pristine catchment early on. There are extreme gradients here. It's about an average elevation of 3,000 meters, but there's about 1,500 meters in elevation difference. At the site, we're asking how mountainous watersheds retain and release water, nutrients, carbon, and metals, and in particular how perturbations to this system or other mountainous watershed systems, such as floods and droughts and early snowmelt, really change the downgradient water availability and the associated biogeochemical cycles. We're asking at time scales of episodic through decadal, so we're asking kind of shorter time scale questions compared to what climate modelers tend to think about, because these are the time scales that water resource managers care about. These systems are already changing, and these are the time scales actually where we think we can develop and test these models. I'm going to go pretty quickly through the next few slides here. I put up this slide. Let me say that that little Rifle watershed that I mentioned is about this size. It's not located there, but in terms of scale and complexity, you can see that we're moving to a much larger scale and we are also moving to much larger complexity. So what I've done here is pulled out a few different pictures showing the different subsystems of this site. You can imagine that if you're trying to understand how this system functions and responds to perturbation, what happens here in this metals-impacted catchment might be quite different than what happens at a floodplain, what happens in a mountainous hillslope or in the hollow. So how do we capture that? What's our conceptual model for how we're going to predict this watershed and its response to perturbation? So rather than genome-enabled reactive transport watershed simulator, which we found to be effective but which we recognize is much too complicated and intractable when we move to the watershed scale, we've developed a concept called a systems-of-systems perspective and a scale-adaptive approach. Let me unpack that a little bit. The systems-of-systems perspective says that we envision this watershed as a collection of these subsystems, and each of these subsystems might have different responses to a perturbation, different interactions between the vegetation, the hydrology, the biogeochemistry. So each of these subsystems might deliver something preferentially or differently compared to neighboring regions to the downgradient riverine system in terms of concentration-discharge. So let's just take a thought experiment here and say if we look at an upgradient alpine system and a lower montane system, and we're looking at how these systems respond to an early snowmelt. The alpine subsystem, the vegetation may not be ready to use that water and may just pass it to the downgradient riverine system, and with it maybe leach nitrogen from the system. Down at the montane system, the vegetation may be quite well ready to receive that water, may grow more vegetation, may have more evapotranspiration, may use the nutrients, so may release very little to the downgradient system. What we're interested in is studying these individual subsystems, but ultimately our big milestone in this project is to create an approach to estimate and then test the aggregated signature of concentration-discharge from these individual subsystems. The way that we do that, in terms of a modeling perspective, again a lot of our projects are very much modeling-driven, is that rather than genome-enabled, we're taking a scale-adaptive approach. Meaning that we're creating a bedrock-through-canopy model that has a depth of mesh refinement that can zoom in to capture and simulate processes where and when we think they're important, which means where and when we think it contributes to the larger system behavior. That's the concept we're testing in the next few years. So I'll go through this pretty quickly, but we're doing this by asking a bunch of different science questions. One: how do these different subsystems respond to perturbations like droughts, floods, early snowmelt? Two: how does connectivity between these subsystems change with those perturbations? Three: how do these interactions in these different subsystems aggregate to yield accumulative output of water discharge, carbon, phosphorus, metal export? And then as we go into out years, we'll start comparing the pristine and the impacted watershed and also look at what information is critical for improving the more operational forecasting we have. Again, a really great team, maybe more discipline-oriented sub-teams. Each of these teams has hypotheses that they're asking about hydrology, ecohydrology, organo-mineral dynamics, and so forth. Many of these people from Berkeley Lab: Tetsu Tokunaga, Heidi Steltzer from Fort Lewis, Jill Banfield from UC Berkeley, Reed Maxwell from Colorado School of Mines. The rest of the folks are from Berkeley Lab: Deb Agarwal, Carl Steefel, Kim Williams, Haruko Wainwright, Owen Brody, Peter Nico, and Tetsu Tokunaga. So a great team, and we're working together at this site to address these scientific questions using again this very iterative model development, experiment observations, iterative approach. We have a series of satellite sites where we're asking very specific questions. We have intensive sites where all the teams are working together to address these bigger questions. Although we don't start for another week, we've been collecting data over the last year or so to start to understand the climate, our weather variability, the biogeochemistry variability, the metagenomic information. We have a series of climate stations. We've done a big snow experiment, ASO recently. We've started to look at exports and how they relate to discharge. I mention this because we really are developing kind of a community observatory. That's something the Department of Energy is very interested in doing. I just have to say that we have many collaborators already working at the site, and we certainly welcome other collaborations. We've started to do in our Environmental Geophysics Group the same sort of work that I mentioned in the Arctic project. Just very briefly, we're now starting to look at watershed zonation, not just understanding the shallower soil and the vegetation, but also how bedrock and topography and aspect play a role. We are zooming into watershed zones to do our detailed instrumentation here, with a real focus on different transects that also very much look at root functioning. We are coupling this into models, both the larger models that are looking at the hillslope transect up to the watershed, but also our hydrogeophysical models where we can start to take this streaming data from the field and use it in a coupled way. So for example, this is a surface-subsurface model that jointly inverts hydrological, thermal, and direct geophysical information to estimate things like soil temperature and saturation that again is important for the microbial functioning. So I think as we're all moving into this realm of a lot of autonomous and ubiquitous data sets, we have a real possibility to test the value of detailed information for larger system functioning, and this is headed in that direction as well. So thank you very much. What I wanted to try and tell you about our two team projects, one focused on ecosystem feedbacks to climate, one on perturbations to watersheds, they both really have at their heart development of models and scale-based information. These are team projects. This is our Watershed Function team. As part of these efforts, our geophysical sub-team is trying to develop new ways to think about watershed ecosystem zonation and function. Thank you.

In these two projects, we are not specifically, and not in the watershed project. However, as I mentioned, we are interested in this being a community observatory, and there are conversations going on about improving our understanding of mountainous orography and atmospheric physics in mountainous regions, including some of the atmospheric chemistry as well. We've been having some conversations about trying to improve our understanding of that climate and then being able to couple it into the watershed model. So it is a direction, but it's not explicit yet.

That's a great question. We talked about that a little bit this morning too in the journal club. I went through those pretty quickly, but we all have a grand challenge question. For example, both of those projects had one. In this watershed one, it was: how do mountainous watersheds retain and release water, and what does it mean for downgradient discharge and biogeochemistry? Underneath that, we have a series of questions that start to unravel that. All the teams, well, I shouldn't say all, most of those questions require all the teams to work together to address it. So we are basically driven by science questions. Then we are all working together at a series of intensive sites to develop process understanding. But as I mentioned earlier, a lot of our work is really moving towards developing and building these models. So really all of our work is done to try to understand and inform and test these models so that we can predict forward. So I guess the short answer is that we have common questions, we have teams that are working together to address those, and we can't do it without that. We're working at common field sites, and we all recognize the value of our work for informing the model. All those things help keep us as a glue. I'm not saying that it is easy all the time, and certainly there are parts of the science discoveries that are made that kind of go off as an offshoot from the whole team direction. But all these things really help keep us together. And then regular meetings, retreats, telecons, and regular exchanges. As I was mentioning this morning, we have a really great set of early career scientists and postdocs that really form a glue across some of these big projects too. Thank you.

In our scope, there's not. We don't have other approaches that we're taking that are more top-down. We of course are doing the watershed zonation, so we look at the organization and start to think about what and start to test what, when, and where we might need that information. But this is where I think it'd be great, and there are investigators that are interested in working with us to kind of co-test these approaches. I think there's a lot to be gained from that.

Yeah, well, that would be great if we can. So this is work that's starting up. We haven't done that. We have done some work in this area where we are using the climate output and predicting forward parts of this system behavior. There's a lot of things that are still pretty unknown about how this system will behave, well, most systems will behave. Part of this work will be developing ways to improve the way change in climate will lead to temperature and changes in moisture and how this responds in the watershed. For example, there's a lot of work going on in creating dynamic vegetation models that will be coupling on top of this system to understand, and that's a big part of the water balance. If you're not getting the vegetation right, you could be missing a big part. So we're at a pretty early stage. We have done some pretty simple feeding of some climate output into this model to understand what it looks like. Reed Maxwell's certainly done a lot in this area as well. But really doing a full treatment that considers the variability in both subsurface and vegetation and what this means, that's the heart of this project. So we haven't done that yet.