DELTA ISB: What we know and don’t know about the Delta food web

The Delta Reform Act of 2009 established the Delta Independent Science Board (or Delta ISB), whose ten members are appointed by the Delta Stewardship Council.  The members appointed to the Delta ISB are nationally or internationally prominent scientists with appropriate expertise to evaluate the broad range of scientific programs in the Delta.

Click here for the Delta Foodweb Conceptual Model.

The Delta Reform Act charges the board with providing oversight of the scientific research, monitoring, and assessment programs in the Delta through periodic reviews at least once every four years.  Since its establishment in 2010, the Delta ISB has produced several reports on various topics, such as restoration, water quality, levee hazards, and findings and recommendations from the Delta ISB reports have helped inform the development and implementation of the Delta Stewardship Council’s Delta Plan.  Many of the recommendations from the Delta ISB are also used to inform the development of science actions to fill critical science gaps through the Science Action Agenda and to better coordinate and communicate science through the Delta Science Plan.

The Delta is one of the most highly invaded estuaries with robust populations of non-native species, some intentionally introduced, and others arriving in ships’ ballast water and other means.  While not all non-native species will become problematic, some of these will, either by modifying habitat, such as submerged aquatic vegetation or by disrupting the food web.  Non-native species that alter food webs make them less able to support native species, either by competing for the food or by changing the food web’s overall composition.  The Delta’s food web has become so altered that the Interagency Ecological Program recognized it as one of the multiple factors contributing to the decline in numbers of multiple Delta fish species, including the Delta smelt.

One of the recommendations in the DISB’s review of non-native species is to build a spatially explicit model of the Delta food web.  At the Delta ISB’s November meeting, a panel discussed what we know and what we don’t know about the Delta’s food web.

PRIMARY PRODUCTION IN THE DELTA

Dr. Jim Cloern, a member of the Delta ISB, has spent decades in the Delta and the San Francisco Bay, studying primary production, algal and zooplankton community dynamics, food web dynamics, and disturbance by introduced species.  He began with a brief presentation on primary production in the Delta.

The Delta is an inland river delta ecosystem formed at the confluence of California’s two largest rivers, the Sacramento River that flows in from the north and the San Joaquin River that flows in from the south.  The map to the left is of the Delta pre-development or about 200 years ago, with the colors showing the distribution of the different wetland habitat types.

Historically, this was a big wetland complex composed of five different wetland habitat types connected by tidal flows and river flows.  The area of the wetland complex in its natural undisturbed state was 2300 square kilometers, which is the size of Chesapeake Bay.

In the Delta as it exists today, the tidal marshes and non-tidal marshes have been leveed and converted into agricultural lands. The dark blue is open water habitat that has expanded relative to the historical Delta. The orange color represents managed floodplains, a habitat type that didn’t exist in the historical Delta.

So the Delta is an extremely transformed ecosystem,” said Dr. Cloern.  “When I first saw these maps, as an ecologist, I wondered, what does this landscape transformation mean in terms of ecosystem processes, like primary production? So we put together a team of people to look at that, and we have an answer to the question of what this landscape scape transformation means in terms of primary production.”

Primary production is at the base of the food web. On the slide, the upper left shows the five different primary producer groups that were considered: phytoplankton, attached microalgae, marsh plants, aquatic plants, and woody riparian plants.   

The approach that we used was really simple,” said Dr. Cloern.  “For each of these plant communities, we scoured the literature and identified characteristic aerial rates of primary productivity, and then we just simply multiplied them by the aerial extent of the habitats where these producers inhabited, comparing the historical and the modern Delta.” 

The outcome from this approach is that total annual net primary production of this historical landscape was on the order of 1300 kilotons of carbon per year,” he continued.  “And if we do that same calculation for the modern landscape after the marshes have basically been removed from the system, it’s less than 100 kilotons of carbon per year.   We’ve known for a long time that one of the consequences of this transformation was lost primary production, but now we have numbers.”

Next, they assessed the consumers of that organic carbon, considering two pathways: direct herbivory and detritus production.  There are quantitative estimates of the massive carbon consumed by herbivores in the historical landscape compared to the modern landscape, and the same thing with detritus production.  This was a marsh system, and the detrital pathway was an important part of moving organic carbon from producers to consumers, and still is important in terms of carbon flow, even though the marshes have largely disappeared, he said.

Dr. Cloern then turned to what we know about primary production in the Delta.  “We know something about the magnitude of net primary production in five habitat types, we know something about at least the magnitude of carbon flows to herbivores and detritus production, and we know that the Delta has been transformed from a marsh based wetland system to one that’s aquatic-based,” he said.  “Most of the primary production now is in open water habitats by vascular plants and phytoplankton, whereas those were minor contributors to total ecosystem production, historically.  We’ve also learned about the importance of managed floodplains, especially during wet years.”

Dr. Cloern then showed a list of some of the things we don’t know, acknowledging it’s really a long list, and so these are just some of the things that we don’t know.  He noted that if a quantitative food web model for the Delta is going to be developed, answers to these questions would be needed.

First of all, we know very little about the structure of food webs, so while we can say something that on the order of 20 kilotons of carbon per year are going to herbivores, we don’t know who those herbivores are in terms of their relative importance as secondary producers,” he said.  “So we don’t know the trophic linkages from the plant producers to amphipods, copepods, bivalves, and other mollusks, mysids, or even fish. We know very little about the trophic linkages between the primary producers and the consumers.”

The calculation that we did is static, and it doesn’t consider any kind of transport. So we know very little about the transport of organic carbon, both within the Delta ecosystem and downstream from the Delta ecosystem to the Bay, and the quantity of this exported into the coastal ocean. We know this is only a consideration of net production in the Delta itself; we don’t know anything about net production upstream of the Delta or downstream of the Delta.”

Finally, the big the elephant in the room question is what are the implications for this loss of primary production for California’s goal of protecting and restoring the health of the Delta.”

SECONDARY PRODUCTION IN THE DELTA

Dr. Wim Kimmerer is a Research Professor of Biology at the Romberg Tiburon Center for Environmental Studies of San Francisco State University.  For over 25 years, Dr. Kimmerer and his associates have conducted studies in the San Francisco Estuary on the effects of freshwater and tidal flow on habitat, abundance, the influence of introduced species, the mortality of fish and foodweb organisms, among other things.

Dr. Kimmerer discussed secondary production in the Delta, noting that his presentation will focus on the brackish to fresh habitats for pelagic organisms in the estuary.  There isn’t really a geographic frame of reference because plankton lives in a moving frame of reference.  He also noted that most of what he is presenting is based on long-term monitoring data from IEP and work that he has done with collaborators.

This graph depicts what he calls ‘the hole in the food web.’  “What I mean by ‘hole in the food web’ is that there’s a big dip in the phytoplankton biomass in the low salinity zone and roughly Suisun Bay,” he said.  “This graph shows the calculated net growth rate of phytoplankton. This is from a paper by Jan Thompson of USGS and me, in which we looked at grazing by clams, copepods, and mycorrhizal plankton.  We calculated the net growth rates of phytoplankton by months, which is shown in the box plot.”

Basically, during the spring, there’s a chance of phytoplankton escaping predation, and be able to have a positive growth rate, and we do see still occasional spikes in productivity or spikes in chlorophyll concentration in the springtime, but during the late spring into summer, especially late summer, there’s just no way for phytoplankton in the northern estuary to get ahold of anything. So you just don’t see high productivity or high biomass during that period.”

The slide to the right shows the consequences.  The phytoplankton biomass, as estimated by chlorophyll concentration, is shown in the upper left.  Copepods support the pelagic food web, which in turn supports Delta smelt, longfin smelt, and other pelagic fishes; these organisms really rely on the copepods as their main food source.

The chlorophyll graph has three panels; the left panel shows 1972-1986, the middle panel from 1987 to 1993, and the right panel for 1994 to 2019.  These three time periods were chosen because the left panel is before the potamocorbula clam’s invasion; the middle panel is before the invasions of several new copepods; the right panel is after that invasion.  On all the graphs, the x-axis is the month and the y axis shows salinity.  These charts are drawn from IEP monitoring data.

The upper left panel shows a strong chlorophyll maximum at around a salinity of five, which is typically up in the western Delta in the summertime, but in the spring, it’s usually down in Suisun Bay; that pattern was persistent almost every year from essentially March until November.  Then in 1987, there was a big decline that was coincident with the arrival of potamocorbula.

Since then, many people have done lots of work on that, and it’s all consistent with a cause and effect, and so essentially since then, we haven’t had any chlorophyll to speak of,” said Dr. Kimmerer.

The second graph down on the left is the eurytemora carolleeae, a copepod, and the pattern is similar to that for chlorophyll, where it was very abundant at a salinity in the low salinity zone, at essentially a salinity of about 2-5 ppt until 1987 when it declined precipitously.

Since then, it has been abundant in the winter, late spring, and fall, and essentially absent from the estuary the rest of the time,” said Dr. Kimmerer.  “It’s not really absent, but its abundance is in very low numbers, and that’s how it’s gone on ever since that 1987 event.

The graphs on the bottom left are Sinocalanus doerrii, a freshwater species of copepod that follows a similar pattern except that it maintains a freshwater population that dwindled after 1993.  In the upper right-hand corner is Pseudodiaptomus forbesi, a copepod native to Asia that was introduced in 1989. Since it wasn’t present before 1989, it isn’t shown in the first panel.  In the second panel, it was quite abundant in low salinity regions.  After 1993, its abundance has shrunk to the point where it’s really now abundant in freshwater, with individuals of that species leaking into saltwater or brackish water.

The last two sets of graphs are related. Acartiella sinensis is a copepod native to estuaries in Southeastern China, Thailand, and Sri Lanka that arrived in 1993.  It’s quite abundant in 2-8 ppt salinities in the late summer and into the fall.  And the next graph shows the Limnoithona tetraspina, a copepod from China, that is highly abundant during the summer, and in pretty much all salinity ranges from about 1-10 ppt.  That was rather perplexing, so for over the last decade and a half, Dr. Kimmerer’s lab has been pulling the strings of this pattern and trying to figure it out.

I think we have it figured out,” he said.  “Going back to the beginning, we have the loss of chlorophyll, which is the base of the food web.  Eurytemora carolleeae is no longer present in the system in the summertime, so it essentially has a temporal refuge from clam predation. I should also mention that in addition to removing the phytoplankton, the clams also removed the larval stages of these copepods to a great extent.”

Pseudodiaptomus forbesi in the upper right-hand corner has a spatial refuge. It’s most abundant in freshwater, and its larvae and its young sort of leak into the low salinity zone by dispersion. But in 1994, that was greatly reduced. And the reason is that Acartiella sinensis is a predatory copepod that eats the nauplii larvae of these other copepods. So that curtailed some of that flux of copepods into the low salinity zone.  As a result, there was a whole lot less available for fish to feed on since 1993. ”

Dr. Kimmerer said that Acartiella sinensis can survive because it’s a predator that feeds mainly on the nauplii larvae of copepods and the Limnoithona tetraspina is its food supply.  Limnoithona tetraspina is very abundant, very small, not preyed upon much upon by fish. So why is it present?

We’ve actually done some work on mortality of these guys. And it turns out that Pseudodiaptomus forbesi has a ridiculously high mortality rate of its larval stages in the low salinity zone in salinity above about 1 ppt.   Limnoithona tetraspina also has a high mortality rate in its larval stages, but as adults and as juveniles, they are essentially invulnerable to predation.  They’re hardly preyed upon by fish, their mortality rates are very low, and so basically, they have a life stage refuge from predation.   Acartiella sinensis has a food supply refuge.   Pseudodiaptomus forbesi has a spatial refuge, and Eurytemora carolleeae has a temporal refuge.  So these are the reasons why we have anything at all in the low salinity zone on which, without these refuges, we wouldn’t have anything to supply food for fishes.”

What are the big unknowns right now?  “We actually don’t quite understand what supports high zooplankton production,” said Dr. Kimmerer.  “We thought we did until we started actually looking at the foods that the zooplankton eat and what supports their high growth rate. It turns out that we have lots of data now on observations of high growth rates and phytoplankton biomass, and yes, they do grow faster when phytoplankton biomass is high – but it hardly ever is in most of the estuary.  We also know that the feed on cyanobacteria, which is supposed to be junk food for copepods, and on microzooplankton, which is pretty well known, but we don’t really have a good handle on which foods actually promote high zooplankton production.”

We’ve been doing some work on longfin smelt with some are our colleagues in various agencies and other organizations.  My lab has been looking at their food resource. The problem is the geographic limits of sampling; essentially, longfin smelt live throughout the estuary and mainly in brackish water, but they spend a lot of time down in the lower bays where the zooplankton monitoring doesn’t go.  From an ecosystem perspective, we’re not actually doing a very good job of monitoring longfin smelt and their food supply where they actually live. And even the longfin smelt larval survey, which is intended to look specifically for this fish misses a lot of the larvae during high flow periods.”

FISH FEEDING ECOLOGY OF THE DELTA

Dr. Ted Sommer is the Lead Scientist for the Department of Water Resources.  He has spent nearly 30 years researching ecology, native fishes, and food webs in the Delta.

For better or worse, the endpoint of a lot of our management activities in the estuary are fish,” Dr. Sommer began.

The Department’s work on fish feeding ecology began around the 60s or so before the State Water Project came online.  Many traditional diet studies looked at the diets of fish, so we know a moderate amount, historically, about what species like striped bass and salmon ate.  Increasingly over the years, a lot of the emphasis shifted to delta smelt.

The other tool that is being used increasingly is experimental enclosures.  It’s a classic ecology tool that places enclosures in different habitats and looks at the food web responses.  A lot of this information was applied to bioenergetic modeling to understand when there’s a food web change, what’s the response?  Delta smelt and Chinook salmon are both examples where we used a modeling approach to integrate that data.

Isotopic studies have also been informative.  One of the better examples is looking at how the isotope signatures for fish feeding change when they’re near submerged aquatic vegetation, such as large beds of invasive vegetation change versus the open water pelagic habitat. More recently, researchers at the USGS and UC Davis have done quite a bit of work looking at marsh habitats and how the isotopic signatures compare to other regions of the estuary.  New tools such as DNA, eDNA, and metabarcoding are starting to be applied to feeding ecology.

One of the areas that we’ve really learned a lot and enough to make actual management changes is in the Yolo Bypass, the San Francisco estuary’s primary floodplain. It’s the former flood basin of the upper Sacramento River, which was partially leveed to maintain it as a flood basin. In the process, a remarkably productive seasonal floodplain was created.

It is a unique system to study because it acts in parallel to the adjacent Sacramento River, so it provides an opportunity to look at foodweb responses and other environmental responses in the seasonal habitat in parallel to sampling in the adjacent Sacramento River.

We’ve had a couple of decades of using different approaches to look at this off-channel habitat,” said Dr. Sommer.  “This has allowed us to do comparative studies looking at diets, bioenergetics, and cage studies, and we found that these off-channel habitats are much more productive. In fact, this is really one of the major rearing habitats for some of the key native species, like downstream migrating juvenile Chinook salmon.”

Insect production is very high during flood events, but that isn’t the only food source; zooplankton can also be a major food source. During lower flow conditions, the floodplain is still connected to the estuary at its base through tidal flows, and fairly high levels of productivity can accumulate there.  They have also determined that even with modest pulse flows through that system; they can get subsidies to the downstream estuary.  This has led to some major management changes to try and improve the functioning of the system.

One of the problems with the system is that during drier periods, there isn’t inflow from the Sacramento River, so connectivity is poor, fish don’t have access to the habitats from the Sacramento River up at the top of the Yolo Bypass, and there aren’t any downstream subsidies.  So the Department of Water Resources, the Bureau of Reclamation, and other partners are working on a large project to build a notch at the top of the floodplain to provide better connectivity to the Sacramento River to hopefully increase food response and rearing habitat.

They have also learned that it’s not all about the high flow periods; the tidal portion of the Yolo Bypass is very productive.  They have been doing active large scale experiments during summertime with managed flow pulses to move food web production downstream.

Here’s an example where we have studied a habitat, intensively looked really in detail at the food web, and managed to provide some lessons that could be used for management,” Dr. Sommer said.

So what do we need to know?  “We mentioned a lot of the progress in pelagic habitats, but there are other sorts of food sources out there,” he said.  “There are benthic food sources; we’ve seen a lot of the fish are eating amphipods. So epibenthic fauna, we don’t sample well, and we don’t have a good handle on their contribution to the food web.  Other off-channel areas, tidal wetlands submerged aquatic vegetation – historically, those weren’t really major inputs to the system. And so we need a better handle on how these areas contribute now.”

We need greater use of experimental approaches.   I mentioned the use of enclosure studies to kind of do direct comparisons and manipulations; we need to do more active manipulation so that we can learn using more of a classic ecological strategy.”

We could really use more integration of this information into fish lifecycle models,” said Dr. Sommer.  “And finally, we need to understand better the potential management actions that we can use to improve fish response. Those include habitat restoration, but also some of the flow actions I mentioned. And whether or not there are ways that we could manage invasive species, for example, aquatic plants.”

DISCUSSION PERIOD

Dr. Jay Lund noted there’s a lot of agriculture that goes on in this area. Does any of that have drainage channels? “Each of these islands has a network of drainage channels, and maybe there’s some detrital material that comes off the agricultural landscape as it gets drained and pumped into the Delta. Does any of that have any significance? There’s so little natural ecosystem input. I’m wondering if that’s an important aspect to consider.”

I’m going to give a partial answer to that question,” said Dr. Cloern.  “I showed you in the calculation that detritus is the largest source of organic carbon to consumers. But that calculation did not include any sources of detritus that originated with agricultural production.  But I do know that these managed floodplains where rice is grown are important in terms of fish production.”

We’ve looked at the managed flooding idea of directly providing habitat for fish,” said Dr. Sommer.  “But the other thing that we’ve been testing in recent years is flow pulses to try and move some of the productivity.  The Cache Slough complex and the  lower Yolo Bypass is essentially a tidal slough network, but it’s also used for agricultural drainage.   These are working landscapes that are also used for different purposes. We found that chlorophyll and zooplankton levels are quite high.  But one of the issues in these areas is that there are also heavy diversions. A lot of that productivity during peak diversion season is going the wrong way, and that flows are going upstream away from the Delta. And so a lot of the purpose of the flow pulse is to see if we can provide more natural flow patterns, and perhaps subsidize downstream habitats.” 

We worked with Ted’s group up in the toe drain area of Yolo Bypass in the summertime and looked at growth rates of the copepods, the most abundant ones up there,” said Dr. Kimmerer.  “We found that most of the time, they’re quite a bit higher than we usually see in the rest of the estuary.  It was related to chlorophyll concentration, and it seemed to be related to the flow pulses, but the abundance was inversely related. So basically, the flow pulses were diluting the populations even as they were growing faster.  There must be some sweet spot in there that we haven’t figured out yet where the flow is high enough to bring fresh nutrients and fresh phytoplankton in but not so high that they just dilute the copepods out there. But we should bear in mind that the Toe Drain where this is going on is tiny – you can throw a rock across it. So so it’s a great living laboratory or field laboratory, but it’s really more of a demonstration for what you might be able to accomplish with a larger system.” 

Question: Is there anybody or any activity going on to develop a foodweb for the Delta, either pelagic or benthic or system-wide, as a quantitative or even qualitative foodweb model or anything like that?

A fluid model, you mean? A bunch of us have kicked around ideas about that,” said Dr. Kimmerer.  “My answer is that I’d like to at least know a bit more about the mechanisms involved, and so I’ve been focusing on focusing on those. I’ve been involved in a lot of modeling projects, so I think it would be a timely thing to do. What I have seen in the modeling work is that you often have modeling going on, independent of any new research. So I think that if you really want to develop a decent model, you need something like a 10-year program where you have a modeling team and various research teams that interact and feed information back and forth. Because if you don’t do that, then they stand alone.  For example, our delta smelt individual base model that I worked on with Kenny Rose. We didn’t have any research component; it was all done on our computers.

Question: So what do you think a role of the ISB could be to move this forward?

What it would take is a person to lead such an effort, and some funds, at least, maybe some seed money to get something off the ground that could then be could be supported over a reasonably long time,” said Dr. Wimmerer.  “It takes a long time and a lot of work to develop good models, and it takes also takes a lot of information that we may or may not have at our fingertips.

It’s particularly challenging when food webs are changing at the speed of light,” said Dr. Cloern.  “You showed shifts in food web structure from one decade to the next. So one question is, which food webs are we going to build models for? And what parts of the system during what eras?

The last time point in there was 1993, so actually, copepod or the zooplankton food web has been steady,” said Dr. Kimmerer.  “There hasn’t been a new introduction that I know of, and people are looking.”

I’m more of a fish guy,” said Dr. Sommer.  “At the same time, I recognize that we might focus maybe too much at times, just on fish responses, and not spend enough time understanding the basics of the food web behind it. We have plenty of examples showing that the food web is critically important. And we’re underpowered at many times. There’s a lot of good information out there. Based on the monitoring, I think, a more focused approach, looking at food web interactions would be helpful.”

That said, though, you have to congratulate the IEP for having started zooplankton monitoring back in 1972,” said Dr. Kimmerer.  “Even the Chesapeake Bay hasn’t doesn’t have a record that long.”

Dr. Steve Brandt said he thought it would be a great organizing topic.  “There are so many facets of this; if it were well organized, these facets could actually coalesce into a much greater understanding of the system and how it’s working, which is vitally important,” he said.  “Also how it might be modified or order kind of restored in such a way that you could you could move the system towards food webs, food web processes, and trophic processes that actually have good management outcomes.”

In the paper that we’ve submitted recently, we did a third scenario of primary production in the Delta,” said Dr. Cloern.  “The third scenario is if we meet California’s goals for restoring tidal marsh habitats, non-tidal marsh habitats, and floodplain habitats, we have assessments of what attaining those habitat restoration goals would mean in terms of carbon fixation by photosynthesis, but nothing about what that would mean for production at higher trophic levels. So that’s an example of if we had an appropriate food web model, we could transform habitat, acreage increases to potential fish production increases.”

If there’s a good strong leadership on this, there’s no telling how far we could actually go towards doing something that would have positive outcomes for many, many years to come,” added Dr. Larsen.

I come from the perspective of a landscape ecologist and an ecological modeler,” said Dr. Sommer.  “It seems to me that this is exactly the time to start thinking about a model that works more at a larger scale in terms of dimensions, where because there’s all this data, you could start on conceptualizing the model, start with the conceptual model, help to identify where the gaps are, and then start exploring more details on the smaller studies or, or calls for different kinds of data. But if you sit down now and see what you have, that will help identify the gaps. And a sensitivity analysis would be really important to figure out where the potential changes could be.”

We’ve been down this particular road before years ago,” said Dr. Kimmerer.  “It was at the behest of the ISB that we needed conceptual models for all the various components of the system. And so people started working on those, and they kind of came dribbling out over time.  I don’t think they really had much of an impact because it wasn’t actually a concerted effort; it was an effort to produce conceptual models, period. And a conceptual model is mainly useful to those who build it because it gets them to organize their own thoughts, ideas, and predictions and start thinking about what do we need to know … that’s why I suggested that you need a team that has some longevity or some persistence over time with a strong leader.

Dr. Robert Naiman noted that it was over 40 years ago that they started working in Canada on some very large watersheds, and asking the question, what was the biophysical basis for fish production coming out of these big watersheds?  “I probably still have the original conceptual models that we put together, which fit on a little kind of 11 by 18 piece of paper, with maybe half a dozen boxes, and some arrows and so on.  But we also had ongoing research to inform those models and fill in the gaps, and we would sit down at the end of the year and revise our conceptual models and put in actually some numbers here and there.   It was quite surprising to me that at the end of probably six to seven, eight years out, that we had a lot of numbers, and we had a conceptual model that was just greatly expanded with numerous boxes and arrows and flexes and lots of stuff going on, then we could actually begin to see how those big watersheds worked to actually produce Atlantic salmon.”

In the process, we began to discover some of the finer details that allowed us to pinpoint what was actually driving the Atlantic salmon production in many of those rivers,” he continued.  “It was the marriage of the modeling combined with a with the ongoing on the ground research and as well as what we knew already about the system, and the revisiting it from year to year that really allowed us within a decade to come up with some really strong answers that to this day and are used in Quebec and in eastern Canada to try and manage those fish populations. These things, when you first start them, they seem almost impossible. But through time, and especially these days where we have so much in the way of existing information and great people out there doing the work, if there was some way to kind of organize and synergize all these efforts, then we could really see something positive come out of this probably within a decade, if not sooner.”

The funding for us largely, in the beginning, was private and through endowments,” he continued.  “We had a cooperative program between the Woods Hole Oceanographic Institution, the government of Quebec, and a number of universities in eastern Canada. The endowment really came out of Johnson and Johnson, Seward Johnson, when he was the CEO of Johnson and Johnson, I believe he loved Atlantic salmon, and he was willing to foot the bill for this in the beginning, and then later on, it was the National Science Foundation and other kinds of traditional funding sources came in to help pick up the tab once we had proof of concept.”

FOR MORE INFORMATION ...