 I think that's good enthusiasm. And I hope it carries on for the next two days. The next speaker is Holly Michael, who's a faculty member in the Department of Geological Sciences and Civil and Environmental Engineering at the University of Delaware. Holly received her PhD from MIT. She has been working on numerous areas in global hydrology and water resources and coastal hydrogeology, groundwater surface water interactions, salinization of groundwater, as well as she's worked on the evaluation of the sustainability of arsenic-safe groundwater in Bangladesh. Holly. Great to advance. These two back? OK. I don't know where the point is. I have the pointer. Thanks. Thanks, Venkat. And thanks for inviting me. I'm really looking forward to these two days and learning a lot about remote sensing and how I can potentially incorporate it more into my work. So this is a workshop on remote sensing. And so you might be surprised that I'm not going to talk about that at all. I'm actually going to talk about in-situ data. So Venkat asked me to talk about in-situ data. And my experience is working internationally around groundwater management. And you might also expect me to do a sort of nice overview like Matt just did. But actually, I'm going to do a bit of a deep dive and give some three examples of some of the work that we've done and identify what I see as some of the challenges with in-situ data, particularly in trans-boundary problems. So first I want to mention that in these projects I had lots of collaborators. And in particular, a former PhD student, Mahfuz Khan, did a lot of the work that I'm going to show you. And he's now faculty at the University of Dhaka in Bangladesh. So as we all know, the recent advances in spatial data sets are enabling new regional and global insights. So just a few examples here. One is a map of permeability over the surface of the earth. A lot of these new data sets are derived both from in-situ data, from remote sensing data, and also from models. This one is global patterns of groundwater, table depth, and some work from Matt Rodel, looking at estimates of evapotranspiration, looking at global river discharge, again using a combination of different types of data and modeling. This is an example of a new global coastal database that is used to assess vulnerability to sea level rise. And so they compiled lots of different types of data for world coastlines that are made available. And this one, looking at the vulnerability of groundwater systems to climate change by Petra Dahl, also recharge estimations and estimations of changes and recharge due to climate change leading to development of vulnerability indices. So we have new data sets, things like bulk aquifer properties, recharge, discharge, projected changes in all of those, and new insights, so new process understanding, identification of vulnerable areas, and targets for mitigation, et cetera. But when it comes down to practical management at a site scale or a basin scale, we really need to institute data to supplement those larger scale analyses and larger scale data to ground truth coarser data sets, so some of the remote sensing and the modeling estimates. So as I mentioned, I'm going to show you three examples. They're quite simple, but I think they illustrate different problems, different groundwater problems, and also they vary in the amount of data that was available and the quality of that data, so illustrating some of the data needs and challenges. As I mentioned before, OK, so the first example is in the Ganges basin of India and Bangladesh, which Matt mentioned earlier. And it was a study on conjunctive use of groundwater and surface water in that basin. So this is a major transboundary water quantity problem, and our study suffered from data unavailability. Then the second example is arsenic contamination in the Bengal basin. So this is another transboundary problem, transboundary aquifer and a water quality problem. And in this case, we had some data, but it was sparse and some of it was low quality. And then the third example is pumping, large scale pumping of groundwater in the mega city of Dhaka in Bangladesh. And this is sort of an interesting combined quantity and quality problem. And in this case, we had a lot more high quality data that was available. OK, so this is the Ganges basin. It's huge. I think it's about a million square kilometers, and it drains a large part of India and also Nepal and Tibet. Let's see. So here is the Ganges. Right here are all these tributaries, so some distributaries. The Ganges flows through India into Farakha. And this is the border with Bangladesh. There was a big dam at Farakha. And an international treaty that governs its management. And the reason is that the downstream area of Bangladesh suffers from, well, so the issue is that there are surface water diversions. So there are lots of diversions up here, which I'll talk about. But here, the water is also diverted into the Hoogley River, which flows through West Bengal, India, into the Bay of Bengal. And it's important for transportation conduit, et cetera. And so they want to keep water in that river. But when that water is diverted during the dry season, then it doesn't go into neighboring Bangladesh. And it causes water shortage problems. The opposite is true in the wet season when water is not diverted. So a lot of water is coming through at Farakha, and then it can cause widespread flooding. So there's an international treaty there and an international agreement about how the water is managed at that border. And so like I mentioned, the problem here is just seasonality. It's not overall water availability, at least in most of the basin. Though certainly there are groundwater depletion problems, and those are mostly, I think, in the Western part. But this is a plot of the discharge in the Ganges at Farakha, as well as the rainfall here in red. And so clearly, there's a seasonality. In the monsoon season, there is widespread flooding. And in the dry season, there are problems of water scarcity. So this is an example of completely dry conditions at Farakha during the dry season. And when that water is not flowing down into Bangladesh and out to the Bay of Bengal, it changes, well, it causes quantity problems and problems of water availability for people, but also major ecosystem problems. So this is the largest mangrove swamp in the world. It's protected. It's the sooner bottom of Bangladesh. And if not enough fresh water is coming into the system, it becomes more salient, and it's really causing speciation to change. It's also home to the endangered Bengal tiger. Anyway, so the World Bank came to us with an interest in the potential for some large-scale, conjunctive use infrastructure for a transboundary water management in the system. And so they recognized these problems, that it was an international issue, and would there be ways to implement major infrastructure that could try to solve it? And they were interested in an idea that was put forward in the 70s is called the Ganges water machine. So just some quick background of this is a really basic representation of how the system sort of works. But there are, of course, major rivers that are perennial rivers. Those are shown in dark blue here. There is a lot of river diversions, so dams, and then canals that are built to distribute surface water for irrigation and other uses throughout the system. And then there are also some ephemeral streams shown here in the lighter color. And there is also some groundwater irrigation. And depending on where you are in the system, there is more or less of that. There are major problems of water logging in some parts of the system, in addition to problems of depletion. So it's quite heterogeneous. So the idea of the Ganges water machine, again, I'm going to be brief just in the interest of time. But I want to give you a flavor of what this was about. And so the idea is to put in wells along the major rivers and to pump during the dry season, essentially, to create a accommodation space that could be potentially filled in the wet season. And then that water that is pumped during the dry season can be distributed out for irrigation into the outlying areas. One of the critical issues here is that in order for this to work, the river has to be dry during the dry season. And so that's obviously a major change to the system. Okay, so in order to take a look at this, we had some data needs. I'm not going to bore you with all the details here, but just to give you an overview, we need model. So we were going to do a modeling analysis to assess the hydrologic effects and also a little bit of an economic analysis. And so we need to populate the model with the boundary conditions and the forcing. So we need things like recharge rates, river widths, distance between rivers, irrigation, return flow. We had to parameterize the model. So hydraulic conductivity in the horizontal and vertical directions, specific yield, conductance of the connectivity between the riverbeds and the canal beds and aquifer. In order to calibrate models, we need in situ data, hydraulic heads, river discharge, river levels. And then also other factors, like energy sources, pump efficiencies, things like that. And so I'm showing here what was available and I've color coded it. So the green is direct data, blue is indirect information that we get from government reports and things like that or estimates from models. The orange is just informed guesses on our part and red is no information. And so there's only one example where we actually had some data and that's river discharge, but it was pretty sparse. There was only a couple of river discharge data locations available. And the rest of it was either taken from aggregated information from the Central Groundwater Board or literature, et cetera. So we were pretty severely data limited. We didn't just consider the Ganges water machine, we also came up with a couple of other scenarios, like, okay, instead of making the whole Ganges go dry, what if there are canals that are diverted and then there's pumping along the canals or what if it's just more typical distributed pumping and recharge, but done in a synchronized way in order to maximize the management of the system. And so to do that, we analyzed those four systems and we set up some really simple models and essentially did a large scale sensitivity analysis that would encompass the variability across the basin and also our major uncertainties in a lot of these parameters. And we looked at things like the potential for reduction in monsoon season river flow, so due to the infiltration along the banks and we looked at the water that would be made available for irrigation, we looked at pumping costs and things like that and these three solutions are the Ganges water machine that pumping along canals and the more distributed pumping. And in this case, it was pretty clear that the Ganges water machine was not the most viable option, even considering all of the uncertainties that we had due to our lack of data. And so I think the Ganges water machine wasn't piloted, at least not to my knowledge. And so this is an example of even a simple data pour analysis being able to provide some insights. But if we were then asked, well, what should be done, that data limitation would really prevent us from giving a good answer. Okay, so moving on to the second example of arsenic contamination in the Bengal basin. So this is the country of Bangladesh, right? This is the Ganges basin that I was talking about before. So this is India here. And there's naturally occurring arsenic in shallow groundwater approximately in this area. So mostly in central Bangladesh, but also extending into West Bengal, India. In that area, there are more than 150 million people drinking groundwater and using it for irrigation, putting about 80 million people at risk for health effects. So there's an example of the groundwater use, which is domestic wells and irrigation wells, which are primarily shallow. And the health effects are issues like skin lesions and also internal and external cancers, problems with brain development in children, et cetera. So this is a major problem. But one interesting aspect is that if you plot arsenic concentration with well depth, there's a pretty big break with depth. So at about 150 meters depth, the concentrations become fairly low. This green line is the World Health Organization standard of 10 micrograms per liter. So you can see in the shallow aquifer, the concentrations are orders of magnitude higher than the standard. It's a big problem, but deeper, it's much better. The concentrations are much lower. And so if we look at the mitigation options that are available and that people are using, deep wells are a major component. So at this point, about 30% and I think growing pretty rapidly. About 42% of the population is still exposed. And so our approach was to look at the sustainability of that deep groundwater resource if it is pumped. So is arsenic just going to migrate to those deeper depths? And if so, how long will it take to get there? And again, this is a big system. And so in order to get at what's happening with the deep system, we really had to model all of it because it's possible that the recharge areas and the discharge areas for deep groundwater are very far apart. So this is the system, this is the Ganges, again coming in, this is approximately Farakha. This is the Brahmaputra River, draining the eastern part of the Himalayas. They come together in Bangladesh and flow into the Bay of Bengal here. This was our model boundary. So we encompassed all the permeable sediments relevant to the groundwater flow system in the basin. And it was just a mod flow model. This is what it looked like. It was about 600 by 600 kilometers to about three kilometers depth. So we developed this model in order to test scenarios for sustainable pumping of that groundwater and the data needs. In this case, we needed topography. We needed pumping data, hydraulic conductivity, of course. And then in order to estimate parameters, we need hydraulic head, groundwater, age, driller logs, river discharge, et cetera. So one of the challenges that we found was in collecting this data. So I already mentioned that in the study we did in India in the Ganges basin, we had almost no data. And that's, I'll talk about that in a minute, but there was a pretty big contrast in the data collection and the data availability between the two countries. So in Bangladesh, there are a lot of both governmental and non-governmental institutions that are dealing with water data. There's the Bangladesh Water Development Board that had water levels at the time. They had, I think several hundred now, there are 1,000 wells that are monitored nationwide, but they're mostly shallow. They've done lots of pumping tests. They have water quality data, rainfall, river stage, and discharge. There's the Bangladesh Meteorological Department, which has a lot of meteorological data, but some data are missing. The Department of Public Health Engineering has the largest database of deep borehole logs, but there are some quality problems with it. Coordinates are sometimes wrong. The data is sometimes poor quality. And it's not well organized. So despite a project that was funded by the Japanese Aid Agency maybe 15 years ago, which my former PhD student, Mafuze, was actually involved in as an undergrad. And so he took lots and lots of paper records of driller logs, digitized them, put them into a huge database, and made that available to DPHE. But unfortunately, that's still not available publicly. We have it because Mafuze was working on a project, but it's not publicly available. And then there's the Water Resources Planning and Organization Agency, which is supposed to be a hub for the water-related data, but doesn't function so well in doing that. Non-governmental institutions, which have projects that are mostly government funded, there's an Institute for Water Modeling, which collects data from lots of different places, and does water modeling studies. And so they have a lot of data, but it's difficult to get it from them. And there's CGIS, the Center for Environmental and Geographic Information Services, which collects data from different government agencies as well and compiles it. A lot of it is GIS coverages. And data is sometimes difficult to get, although you can buy it, you can pay for it. In India, there's the Central Groundwater Board, which Matt mentioned, and they have lots of data. So as far as I know, there's a pretty well-organized groundwater monitoring network and it's high-quality data. The Board produces really nice reports, but they're primarily aggregated information. And then there's also the Geological Survey of India, which collects data. But the problem is just getting it. So as Matt mentioned, it's available in country, but not internationally. So that presented a major problem. Navigation of all of this involved a lot of collaboration. We had really good collaborators in both Bangladesh and India. We had great students and spent a lot of time there just meeting with people, meeting with these different agencies and trying to understand where data was and how it could be, how we could get it. So going back to these data needs, just some idea of what we got. So for topography, there's the, we use the SRTM data, there's data available. For domestic and irrigation pumping, we used data on population density in irrigated areas. We actually, in that case, were able to get data for West Bengal, though it was a little bit lower resolution than we had, which was, I think, or district-wise in Bangladesh. And same thing for irrigated area. For parameterizing the model, we estimated the hydraulic conductivity using three different data types. And those were in situ hydraulic head data, groundwater age, and driller logs. We didn't have river discharge, so we did not use that. And so just some examples. So these are 147 wells that we found from in Bangladesh Water Development Board data. There are more wells than that, but we tried to quality control. So that brought the number down quite a bit. And the problem is that they're only in Bangladesh and not in India here. But we use them in any case. The data are reasonable. They rise, the water tables generally rise to the ground surface during the monsoon, and then they fall during the irrigation dry season. But one of the water quality or data quality problems was that the elevations were very different from the topography, right? So it was clear that the elevation surveys were probably not right. And so we didn't want to trust using the actual elevations to calibrate the model, but instead dealt with that by using the transient fall during the irrigation system to calibrate these parameters. The problem is that these wells are only shallow and the deep aquifer, which we were most interested in, is actually insensitive to that data. We were able to find only four groundwater age data in the deep aquifer, but they ended up being very useful. So the good thing about them is that because they were deep groundwater ages, that the deep aquifer parameters were highly sensitive to them. But some of the problems are that, one, there's only four of them, but also there was quite a bit of uncertainty about the actual ages because of a lack of additional geochemical data that would have helped us constrain those ages a little bit better. But still it was a pretty good calibration target. And the third one was driller log data. We had 147 driller logs. We had them in both Bangladesh and India. These were mostly from the literature. And from those, we were able to estimate the vertical and horizontal hydraulic conductivity for various locations in the basin. And so we looked at spatial trends in hydraulic conductivity. But again, there was a quality control problem because these are collected by different drillers using different drilling methods, and they have different ways to interpret the facies. But in any case, they were the best that we had and so we use them. But even with that data, we really could only reliably calibrate two parameters for that whole huge system. So even doing something simple like zoning it into physiographic regions didn't really produce something that we felt we could believe it was better than just assuming the whole system was homogeneous. And so as a first approximation, we said, well, let's just assume that it's homogeneous and isotropic that due to all this heterogeneity in the system, we have strong anisotropy and then go from there with sensitivity analysis to understand how wrong we might be. Okay, so we looked at two management alternatives. One is putting both irrigation and domestic pumping deep into the low arsenic aquifer and the other is splitting it. So keeping the irrigation wells where they are shallow and only putting domestic pumps deep into the system. And what we found was that there is a really big difference between those two management alternatives and this is for our base case scenario, but even if we really varied the parameters over our range of uncertainty or variability, we still convinced ourselves that we had a robust result. So with the split pumping scheme, 90% of the area was sustainable compared to only 14% with all the pumps being deep in the system. So one thing that came out of this, so despite the fact that we had sparse data and we had to really make a simple model because of that, I think there was useful and I think hopefully robust information that came out of it which is to regulate the indiscriminate extraction of deep groundwater. So really just using that for the much smaller amount of domestic pumping that's necessary as opposed to irrigation. Okay, so on to the third example. Currently in Bangladesh that regulation is not occurring but I think it's possible to discourage irrigation with deep groundwater in rural areas. On the plus side, tens of thousands of deep wells have been installed all over the basin which has really reduced arsenic exposures but there are still other problems that need to be dealt with. One of them is that there's a huge population in Dhaka that is extracting massive amounts of groundwater and that's causing big problems and it's not just Dhaka of course, this is Dhaka, this is a map of mega depletion cases that around the world and the size of the dot here equals the size of the mega depletion and Dhaka is actually only a small one compared to many others in the world. But we wanted to know, okay, what are the effects of this pumping and is this water use in Dhaka making this arsenic problem worse? So the idea here is that, so this is just a schematic of the Dhaka system. Here's Dhaka city, they have these very high capacity and deep pumping wells within the city that's created a cone of depression. The blue here is the deep groundwater level and the red is the shallow groundwater level. There is no arsenic within Dhaka because it happens to be this uplifted Pleistocene block and the Pleistocene aquifer is the low arsenic aquifer. But just outside of the city, there's widespread high arsenic and the shallow aquifers. So there's shallow pumping out there. There is no access to the city water supply outside the city, right? So there are pumps here and there are deep wells that have been installed for arsenic mitigation. So the water table in the city has declined more than 80 meters in some areas. It's declining two to five meters per year and they keep having to put the pumps deeper over time. So this is a major water quantity problem. But our concern was the associated water quality problem not in the city, but outside the city where people aren't having access to the water that's being pumped. And the issue is that it's creating these massive downward gradients that could drive downward migration of arsenic and contaminate these deep wells. There's also an associated problem with low groundwater levels here, which are in some cases making the hand pumps inoperable. Okay, the information that we needed here was all the information we had in the basin scale model, but additional more high density model parameterization of hydraulic conductivity and specific yield. And then again, we needed data for parameter estimation, the heads, the drill logs, the river information. In this case, we had a lot more direct data. We have pretty reliable hydraulic head measurements. Some of that is from D-WASA, which is the DACA Water Management Agency. And we also had a lot of our own wells and loggers. So this is a much smaller area than we were talking about before and it's actually feasible to go in and make our own more reliable measurements. We had pretty high quality driller logs from those high capacity wells that have been drilled and some of our own data on the river levels. Just to note, even though this is a massive problem in DACA, they could potentially run out of water in the next decade, the data collection is sporadic and it's really driven by sort of one-off government projects. And there is not a standard practice of modeling by the DACA Water Agency or other agencies, which I think presents a problem for figuring out how to manage this issue. Okay, so how did we address this? We used a nested model, so we took that big model and then we nested a higher resolution model within it. And so that's the river system that's in that small area. This is, the colors in this map indicate arsenic concentrations. I'm not showing that arsenic data, but this is aggregated here. And so within this purple area, which is DACA city center, it's green, it's arsenic safe and that's again, because it's Pleistocene, but just outside the city center, there is high arsenic. So in some cases, greater than 50% of the area has high arsenic. And then the black pluses here are locations of our groundwater level data. And we had both shallow data and deep data, which was important. Here are the locations of the lithologs that we had and that we used in the modeling. So again, pretty densely spaced. Going back to what the system looks like, it's pretty complex hydrologically and also geochemically and geologically. So these river systems are braided. They're huge, this is kilometers wide. And you have lots and lots of other surface water bodies on the surface, distributaries and tributaries and ponds and things like that. And so you have a very complex surface which has created a very complex subsurface. And of course we ignored that in the large scale analyses, one because it would be pretty hard to get that into such a large model, but also because we didn't have the information. But we had the information in this case, and I'll just illustrate to you actually first how heterogeneous this system is. This is that database that I mentioned that was digitized and compiled by Muffoo's. All of that data is shown here. And so the colors are, the greens are clays and silts and then the warmer colors are sands and gravels. And so the system you can see is really quite heterogeneous. This is just looking at it all from the side. But of course we're just looking at the small area and we used our localized data to create models of heterogeneity using geostatistics. So this is an example of one of our stratigraphic models and we generated lots of these that were statistically equivalent. And so essentially what came out of it, so we looked at, for example, probability of contamination in 200 years. And these areas in pink are the areas that would be contaminated within that timeframe for an equivalent homogeneous model, equivalent to all of these, with equivalent permeability of the heterogeneous models. But if we look overall at those 60 geologic models and what comes out on average in terms of probability, the area where there is a high probability of contamination is much larger. And the uncertainty is really great throughout the whole study area. So the first message here is that incorporating this geologic data is important, especially in this area where we have the strong pumping that's associated preferential flow and transport of arsenic. And the contamination is likely in some areas, but it's really uncertain. It's also difficult to predict. So one of the things we wanted to do with this model was try and identify in-situ data that could be collected that would be indicative of whether or not a certain location is vulnerable to arsenic migration. And it turns out it didn't work. The typical indicators like water levels and sediments and distance from the city were really not predictive. And that's because the system is so heterogeneous and three-dimensional. So our main conclusion here is that monitoring is really critical in order to protect the health of the people drinking the water. Okay, so just to summarize, with the first example, I just wanted to show that the data can be hard to get. But also, even in the absence of data, you can get some informative results, but they're limited to certain questions. Analysis with sparse, sometimes low-quality data can still provide useful high-level management guidance. But the more data you have, the better your understanding, right? And I think I'm sure I'm preaching to the choir here. Everybody understands these ideas, but I just wanted to illustrate it, some of these challenges through this example. And yeah, so we may have better understanding, even if that understanding means that it's uncertain, but that's still important information. Okay, so these examples are simple models, right? But our groundwater and hydrological models are rapidly becoming capable of really large-scale analysis, like Matt pointed out. So this example is one from Reed Maxwell and Laura Condon. I don't know, maybe Laura will talk about it later, but this is a continental-scale integrated hydrologic model that has produced these really amazing results. And the problem is that the availability of in-situ subsurface data to constrain simulations like this, and in the U.S., it's much better. There's a lot more data than there is elsewhere. That availability of data is lagging behind. There are some areas in hydrology that I think are more advanced than we are in subsurface hydrology. So meteorological data is much more available. There's global databases of river discharge, but like Matt pointed out, only eight countries are contributing to a global database of groundwater. And so that's a huge problem. The remote sensing and surface-based data are providing exciting new insights. They're enabling these global-scale analyses. Certainly the examples that I showed would have benefited from that information. If anybody wants to collaborate, let me know. But I still think that the remote sensing and the models can't substitute for in-situ data, particularly for subsurface processes that can be difficult to image in high resolution. And another issue is that while data may be dense and available in some areas, there are always problems of collection and organization that are global. So this is an example from that fan at all paper that I showed earlier where they showed the all of the shallow groundwater measurements that they used in their analysis. And so in the U.S., it's very dense and in other nations, but in many areas, it's quite sparse. And so I'm not meaning to pick on Bangladesh and India. I've talked about those examples because that's where I've been working, but it's also the case in Delaware, for example. So we're the second smallest state. Everybody knows everybody, you imagine, that we're really coordinated in our environmental data and people are actually very environmentally oriented. We're very worried about sea level rise and things like that, but it's not the case. So we have the Delaware Geological Survey that has a really nice monitoring network and a database, but they only have their data. And then there's the Department of Natural Resources and Environmental Control that has lots of different departments and they're collecting all kinds of different data, including really valuable water quality data, but that's not coordinated. It's not put in any database. You have to know somebody who knows which report you should look in order to get that data. And so it's a major problem, even in Delaware, even just looking at salinity, for example. So I'm saying all this to highlight that there's, which I think all of you know that there's a need for a pie in the sky, global database of fully quality control, densely spaced, in situ, hydrological, geological, geochemical data that's publicly available in real time all over the globe, right? Yeah, that's the pie in the sky, but there are lots of challenges, of course, the cost, we would have to have international data sharing agreements. I think that's a major issue. Coordination, technology transfer, infrastructure. Political will, of course, is a big one. Expertise, sensor development. So if we want to extend these databases to geochemistry, having really good geochemical sensors is important. And then of course, uniformity and quality control. Collecting high quality data, of course, is really expensive. So there are trade-offs. On one hand, if you have a global database, you want it to all be uniform. I can just use that data in all the same way and my models are to inform my remote sensing. But on the other hand, I think there is value in lower quality data as long as its shortcomings are identified. Okay, and so just lastly, solving the major water challenges of the future, especially transboundary conflicts, are going to require large data sets that are both in situ and remotely sensed. And I think transparency in the data collection and quality and access to that are critical to coming up with policies that make sense and are agreeable to all sides in these conflicts. Thank you. Thank you, Holly. Thank you very much for that wonderful talk. Questions for Holly. Mark, Persaud, New Mexico Tech. Can, I know you feel that your models have a fair amount of uncertainty, but do you think, for example, with your DACA model, that you can forecast potentially the decade when the water concerns, either contamination or loss of water levels will become very critical for the city? Yeah, I think for the water levels for sure, we did actually a really good job in both a homogeneous and heterogeneous models of getting the cone of depression, right? So I didn't show any of that, but there's concern about that drawdown cone expanding into the outer areas. And so, yeah, I think we can do, as long as we know the pumping rate into the future, we can do a pretty good job of predicting that. The more difficult thing is predicting the arsenic transport, not only because of these uncertainties in preferential flow and how these sediments are connected, but also because of the geochemistry, which is really complicated. And I didn't go there at all, but how arsenic sorbs, how it transforms, as it moves through different sediments, is still a big question. So in terms of just water transport, maybe we could do that, but in terms of the actual arsenic transport, I think it's a harder question. Ed Bigley, Northeastern University. In all the examples that you showed, you had the data that you needed for parameterization and kind of listed the sources. For discharge, I think all of them said none. You had water level in one example. I'm just curious, if you had river discharge for each of those examples, what's the value gain from having that discharge in the modeling work that you did? Yes, a good question. So we had some, there were a couple points in India that we used, but we didn't have much high quality discharge data. If we had it, it may have changed the way we modeled. So that's something I didn't talk about, is that the data that's available actually changes the strategy for modeling and the potential outcomes for it. And so if we were able to use discharge data, we can do better at getting, say, recharge and hydraulic conductivity together. So there are sort of paired parameters that really benefit from not having only hydraulic heads to calibrate to, but also fluxes. So that's one of the examples. That's one reason it would be really great to have that data. And so I think both discharge and level data are really valuable. Thanks, Holly. That was a wonderful talk and very clear as to the points you wanted to make. I don't think you're doing this and I want to throw it out because I would like to have people react to it. Many times when I listen to hydrologists and or non-academics interested in water, especially politicians, the focus is on how changes in supply or resource availability are likely to have an impact without really thinking about adaptation on the use side or preventive side. And I wonder if you'd like to also compliment your excellent presentation on the analysis with some thoughts on what needs to change that could help in these cases. That's an excellent question. So the analyses that I showed, we were looking at if the current scenario continues essentially, right? And so what we certainly could have incorporated was management scenarios in which behavior changes. And so changing pumping rates and illustrating how that could then affect those outcomes I think is really important. And one thing that I think, so one thing that we've been trying to do in some other projects and that I think is really important moving forward in hydrological modeling is incorporating decision-making and incorporating effects of policy. So using coupled models that are not just coupled hydrological processes but that also couple in the feedbacks between what happens in the natural system and the way people perceive issues and the way that they therefore behave and the way that policy follows the changes in the natural system. So the examples that you put up that were in the news of all the problems globally that people are paying attention to, how is that potentially making changes that could feed back into the natural system? So I think that's a really important part of it. Burke Minsley with the USGS, thanks Holly. I was wondering if you looked at sort of a data worth analysis of looking at and supporting these decisions of which data sets are most important and which ones would you go forward in trying to collect or reduce uncertainties in your model to inform these decisions if you've done that or if you have an idea of which data types would be most useful in doing that? Yeah, that's a great question. For these analyses, I don't think we did that explicitly but by calibrating you get a good sense of what's useful data and what's not, right? And so like I mentioned that shallow groundwater data wasn't helping us so much but just those foreground water age dates were huge. And so in our case, we're looking at a big system with really long travel times constraining the ages would be something I'd wanna do more of but I think it depends a lot on the question but I do think that it's definitely important as we're thinking about the various questions that need to be asked what data is the most relevant and what is the most bang for the buck in terms of monitoring? Mike Cosh, USTA, Agriculture Research Service. Comment on in situ data collection. We're all academics here mostly and we all have this perspective of more data, better data. There's a large group of people that would say that they're not thinking about collecting data for altruistic purposes, they're looking at it as a tactical purpose. And I'm referring to even the sponsor of this workshop that's looking at a variety of purposes that could be on academics. There are economic purposes, there are economic advantages to be gained from just natural resource monitoring and for us remote sensing is a term that means more data, better data everywhere and for some people remote sensing is essentially spying. We're not there, why aren't we there? Well, maybe they're a different company, maybe they're a different country, maybe they're a different state and we're trying to negotiate or navigate contracts and where's there gonna be data issues in the future or water issues in the future? So that perspective can be, we forget about it because it's not really in our realm, we're worried about trying to help people and make sure they don't get poisoned. But other people may be like, well, I'm looking where to install my next factory, I don't wanna put it in a town that's gonna have water issues. So that town doesn't really want people to know they have water issues. So sometimes it's nice to pull back and go, okay, wait a minute, how can this go wrong? I get asked this question when I go on a farm, how is this gonna screw me over? How is what you're collecting gonna harm me in some way, either through regulation or change in policy? Yeah, that's a really great point. And we have the same interactions with farmers because we're looking at salinity and coastal zones on farms and there's a worry that, if it's known I have a water quality problem, how can the value of my property goes down for home owners and things like that? Yeah, I think that's a problem and that's one of the reasons that countries don't wanna share their data, right? There are reasons to keep data private. But I think that, we have to somehow overcome it if we want this pie in the sky to happen. And one way to do that is communication. So communicating the potential benefits that could potentially outweigh these issues. So if there is a water quantity problem somewhere and maybe having the data available to people who can help manage the system would actually make things much better as opposed to making them worse. So part of it maybe is changing perceptions and to communicate more the potential benefits of data. Thank you, Paulie. Thank you very much.