Last time we talked about how establishing trends are important for monitored natural attenuation and how things like the Mann-Kendall Analysis can help you do this. We'd like to continue down this road by emphasizing that environmental data can sometimes make all this challenging. I think everybody who's ever tried to plot a few years of quarterly or semiannual monitoring data would recognize that these data can look pretty messy at times. >> Yeah, and refer to this is variability of the data. And we really see it here in this example plot of concentration versus time data. >> So high monitoring and variability, it can be problematic for several reasons. It can create misleading patterns, maybe apparently increasing trends that are not there, maybe it looks like there's some seasonality, apparent changes in attenuation rates. So it makes it a lot harder to verify that your concentrations are decreasing and what that rate really is. >> Yeah, it adds a lot of uncertainty things, and this has implications for sort of the amount of monitoring you have to do, the amount of data that you have to collect. And one of the big problems with this is that the sources of that variability aren't really always well understood. One thing that we've got available to us though is this series of sort of interesting data that was collected in a series of studies sponsored by SERDP and ESTCP that looked at the relative contributions of various sources of variability. And so on this next slide, what we'll sort of go through, at least a few of those, and take a look at whether they're big deals or not, something we need to concern ourselves with. Well first is sort of looking at source attenuation. So that's this idea that you've actually got a trend, you've actually got things sort of decreasing over time in terms of your contaminant concentration. >> That's really what we're trying to find here. >> That's our overall goal. We've got well and sampling dynamics. Things that are going on then while you're actually sampling or maybe even things that are going wrong in the well between sampling events. >> And then there's this lab analysis, and so here's a picture of in the lab and maybe we'll start with that one first and then go to some of the well stuff, so- >> Yeah, well I think we should start with that one first because just to clarify in terms of source attenuation, really we're talking about variability. We're talking about short-term variability, and so that's sort of that event to event, things that are maybe out of your control. And source attenuation is really describing a trend. So in this case, in terms of variability, we're not really interested in source attenuation, we're interested in things like the other two. >> So number one is really the signal and these other things are the noise. >> Yeah. >> And we're trying to distinguish between those. >> Yeah, exactly, so lab analysis. >> So there were some studies done on a lot of lab data from Hill Air Force Base and some other sites as part of some of these projects. It was done by Dr. Tom McHugh, and it found some interesting things. They would switch labs but the big picture is that the variability in the labs is not too bad. It causes less than 10% of the overall variability. There are some interesting things that they discovered that when you change labs sometimes that baseline concentration can change a little bit. But overall, the labs aren't a significant part of the variability problem, don't worry about the labs. >> So I can't blame all my data variability problems on the labs. >> Only in extreme circumstances. >> Okay, well if that's not the case then you look at something like the well or the sampling methods that you're using. And so you might imagine that if you've got a better choice of sampling, maybe that's going to give you less variability in your data, more confidence in that result. It turns out same data studies not exactly a major source in terms of variability. >> Okay, can we talk more about that? >> Yeah, and then we're going to look at that then here in the next couple slides because it really ends up being sort of an interesting story, something that maybe isn't all that well appreciated. And the way to think about it is that a monitoring well can be pretty dynamic situation. There's a lot of things that might be happening between sampling events, so before youre out there, as well as while you're actually collecting those samples. And so we'll go through a few of those here as we talk. So one of the things, in terms of what's going on between sampling events, is this idea of ambient flow. And so what's ambient flow, Chuck? >> Well it's this idea you have this flow in the groundwater. And so if you have this well that you could have this flow sort of horizontally go through the well and out the same side. And so this well is sort of capturing this stuff. So that's just this natural groundwater flow. >> Yeah, and that's because you really got less hydrostatic resistance when you put that well into the formation, right? >> The well will actually concentrate some flow at least the streamlines will converge in there and go out. So that's the first thing that sort of going on. >> Yeah, and so as a result of that process, you've got flushing of that well that happens, a well from the formation does make its way through that well. And so this is sort of semi-continuous flushing process that goes on. But as Chuck mentioned, you do have some influences from the permeability. So you will have a greater contribution of zones of higher permeability when you're talking about this sort of ambient flow. >> But looking at the bottom of that graphic there, the original idea was that you did have this flow going from left to right sort of through the well, that's not the case in a lot of conditions. There is actually stuff going on in the well itself. >> Yeah, and so there's also this idea that ambient mixing can also happen within the well. And so when we say ambient mixing we're sort of talking about this process where you see vertical mixing of groundwater within the water column within the well. And so that can happen for a lot different reasons a few of them are listed here then on the slide. But it's basically these gradients of things like temperature, pressure or salinity. They might cause density differences such as that water isn't stable in there. It wants to sort of mix around. >> So a lot of new knowledge about this, it's maybe the elevator effect. Contaminants can come in here, go down and exit down here, or the reverse. >> Yeah, so that would cause redistribution of those containments within there. It would mean that you may have a closer to a flow-weighted average that's actually going on if you sampled within that well. And so there's a lot of interest in this, there's particularly interest in how common this particular phenomenon might be. And so one way to maybe take a look at ambient mixing is to sort of look at the effects of temperature. So the temperature obviously changes seasonally and so you can take a look at this little graph here. So you imagine you're sort of at the depth on the y-axis in a well and you've got temperature on the x-axis. And so as you go deeper into that well, there's some particular seasons where you might see the temperature actually increasing with depth. >> Okay, cold water is denser than warm water, right? >> Yeah, and so in this case you've got colder water near the surface. It's more dense than warmer water below and so what would this cause in terms of within that well. We actually get a case then where you want to have mixing occur, right? That density gradient can't be really sustainable, so you'd see water then within that water column mixing. And so that might be when you've got colder water near the surface, that might happen sort of late winter, early springtime periods. >> Early on in my career, I did a lot of work with lakes and there's this idea that in the spring there's this lake turnover and that cold water sort of plunges down. And so similar concept here, right? >> Yeah, pretty analogous. So there's an opposite sort of situation that you'd see. What's that, Chuck? >> Okay, so think it's late summer. It's early fall, that shallow groundwater nearest to the surface has heated up, so it's less dense, so it stays there. And what you'll have is this stratification in the well. And some of those stuff we've sampled in this case. You can see tremendous orders of magnitude change from just a couple of feet, even less than a foot. 10 micrograms per liter, 10,000 micrograms per liter here. So it can be pretty stark. >> Yeah, so it's not promoting that sort of mixing, not promoting that redistribution, so you might be able to get actually depth data that would be very different at various points within that well. And just a couple other points on here, we've sort of done this in terms of collecting field data as well as modelling, but this sort of phenomenon, it's really driven by the surface temperature. So maybe about ten meters is where you'd see that sort of effect happen. >> That's right. >> So then let's move to what's going on during sampling events. So we use well purging in a lot of cases, low flow purging is sort of the standard within our field for collecting samples. The idea here is you're trying to get sampled water that's more representative of the formation water, right? >> Right, and then I think the graphics and some of this stuff are coming from this ESTCP project by, is it Sandy Britt and then Jamie Hayden? Is that right? >> Yeah, yeah. >> And they did some computer modeling and they did some sampling. So a lot of these insights on what's happening in the well was coming from that particular ESTCP program. >> Yeah, but one of the things that happens then when you're purging a well is you're still pushing towards that sort of flow-weighted average concentration. You've got high flow zones moving into that well as you're purging it. So you're going to obtain a flow-weighted average if you sufficiently purge that well. In this case though, one thing to remember then is that concentration result, if there is in-well mixing, vertical mixing, ambient mixing going on in there, the purge sample and a no-purge sample would probably be roughly the same answer, right. So one way to sort of go about determining whether this was important is to take a look at sort of what the field data would show us. And you're going to describe a little bit about this, right? This is showing some different sampling methods that were tested out, right Chuck? >> Pretty cool study, this is an ESTCP Project again down there. And the idea is that it does it matter which sampling method you use? And as our conventional method, which is doing this low flow sampling to parameter stability, is this still the king or how much variability affects these different things. So we have these five different sampling methods, the first again is the standard, the thing we've been doing for a long time, standard low flow. You're going to purge that well, pump out that water til your dissolved oxygen and your other parameters stabilize, then you're going to pull that sample. >> Okay. >> Then we said, let's go back in time, originally we said we did the sample well, pump out three casing volume. So we said, put away your dissolved oxygen meter, put away your pH meter. Just take out, in the case of 2A, just 3 liters from that well and then sample. And the other one's a sort of a standard consistent volume each time, 18 liters per purge. And the idea is that if you do the same thing every time, maybe you'll decrease that variability. Then we looked at two no purge methods. One is this SNAP sampler which is this device that sort of has a VOA vial on the surface that clamps down. You can pull the sample once it's installed in there, and then a really neat device, HydraSleeve where we pulled up, that will open up and then collect the sample. So the question was, we have these sort of five different types of sampling, are some of them better for variability and some of them for worse? And so went out to two sites, did a lot of sampling, one after another. Hey, let's compare the variability. >> And okay, so a lot of that is shown. And you collect a lot of data I know from there. But some of this is shown really well in terms of this sort of almost conceptual diagram of plotting the cost associated with each of those methods versus the variability associated with each of those methods. So what you saw in terms of the amount of variability of the data that you collected with each of those individual types of sampling, right? And so that's shown on here. What are the actual findings from here then? >> Okay, cost is on the y-axis in this and variability's on the x-axis, and we've got the five different methods. The costliest is our standard low flow standard where you sort of keep purging until you reach that stabilization. And then this active no-purge is the least expensive. But it also was the highest variability. There appeared to be a couple of sites where it seemed to did not match the other one. So that may have been the way that this particular device opens up and collected the samples. So we did notice that type of difference in these things, but then this sort of puts these different things in there. Another idea is that this purging, that parameters stability, didn't really reduce the variability at all. >> It's sort of the same as these other methods that. >> Just looking at those green boxes there most of them are really in terms of the variability that x-axis lining up pretty close together so not maybe a lot of changes. >> Yeah, so what was interesting is sort of different than some of thoughts going into the study. But if we wanted to summarize some of the results that are coming into this thing, we can go to the next slide and look at some of this. The key thing was that sampling method appeared to have little effect on this concentration or this variability except for this no purge sampling at some sites. >> Okay, and there's really no benefit then in this case to monitoring, to using purge perimeters to sort of monitor where you at in that process. Doing that versus fix volume, you didn't really see a lot of difference. >> Yeah, so you select your sample method based on cost, ease of implementation, sample volume requirements. And that's maybe how you make your decision. >> So not necessarily all that impact on data quality itself. It's these other factors that might be more important. >> That's right. >> Okay, so we touched on these three things originally and [COUGH] we basically said that first one doesn't necessarily matter, source attenuation, lab analysis, maybe a minor source. Well and sampling dynamics, we just spent a few slides going over those, not really a major source as well. That leaves us with one thing we didn't mention before and that's this idea of signal variability, and so Chuck, what's signal variability? >> Well, it's just part of the natural change of these things. That the contaminants are leaving these sources and they're moving down gradient in these plumes. And you might think of the subsurface as being this reactor with billions and billions of tiny baffles. So there's just not that much of mixing in there, so you can have these high concentration packets that enters a well then maybe a lower concentration. So in some ways it's just the nature of the beast, and you have to live with this and you have to account for it with things like statistics and things like that. >> Okay, so it's a major source but maybe not one that we have all that much control over in terms of short term variability. >> So number one is sort of the signal. You're trying to measure this, but number four says there is that variability in the signal. And you just have to sort of overcome it by being very judicious about how you analyze your data. But the actual sampling method, maybe it's not that important. >> Okay, so let's wrap up and go through some of the key points from this lecture. During monitoring, purging is typically used to obtain a flow-weighted average concentration. >> Okay, and then between monitoring events, ambient flow and mixing of water can occur in that monitoring well and cause the water to be really representative of the formation, and that's without purging. >> And based on the great field study that we went through, there's really little difference that we're seeing based on the sampling method itself. >> And our ability to reduce that short-term variability in that signal, that's pretty limited right now. We just have to account for it or deal with it in different ways.
