Stanford [00:00:13]:
If you study rivers in the United States, you either have or will encounter the sediment data measurements from the US Geological Survey. Today's guest, Molly Wood, is the national sediment specialist for the USGS Water Resource Mission, who has responsibility over those data. The first time I seriously considered this podcast, I scribbled a couple names in my journal thatd be like my top draft picks for these conversations, and Molly was one of the first names I wrote. Her qualifications as the national sediment specialist for the USGS is plenty of reason to include her in this podcast. But the main reason I wanted to include a conversation with Molly wasnt just because of her position at the USGS or even because of her great research in sediment surrogates before she took over that position. The main reason Molly Wood was one of the first people I wanted to talk to is because I have literally never left a conversation with Molly with my beliefs about sediment intact. Every time I talked to her, I learned something new and substantial about sediment. Sediment data or sediment processes, including this conversation. By the time we were done talking, I had updated my assumptions about how to interpret sample replicate variability and continuous sampling data, among other things. As with all these conversations, we try to cover some introductory topics and then delve into some topics that are more advanced. But by the end of this conversation, I was just asking her some stuff. Ive always, always wondered about sediment data. It was a great conversation that, frankly, I wish I had access to 20 years ago when I first started using the USGS data portal. I'm Stanford Gibson, the sediment transport specialist at the course Hydrologic Engineering center. And in this episode of the RSM River Mechanics podcast, a conversation with sediment data specialist Molly Wood. Molly Wood, welcome to the podcast.

Molly [00:01:47]:
Thank you. It's good to be here.

Stanford [00:01:49]:
So I think that if you were to talk to junior high students or even high school students and ask them, what do you want to be when you grow up? I suspect that exactly zero of them would say that I want to be a global leader in river sediment sampling, although probably some of them should. I'm pretty curious. How did you get here? Like, what was the path that you took? How did sediment sampling capture your imagination?

Molly [00:02:14]:
Yeah, I mean, to be honest, I'm sure that I never thought of this as a possible career direction when I was a teenager, but I wanted to be a park ranger outdoors, that kind of thing. But, you know, I think it's. I never expected to be specifically in sediment. It kind of evolved and grew into that arena. But when I started out in my career, I actually started out working for a consulting firm. And we would go to these sites that were old army ammunition manufacturing facilities. And we would go out and we would sample soil and water for explosives contamination and metals contamination. And I just remember being fascinated by thinking about how these explosives would attach to the sediment and then that sediment would move and move off the facility and have downstream effects. So I think that caught my attention in the beginning. And the thought of how sediment moves and interacts with other things. And that was fascinating to me. And then I started working for the government and started out in Florida there and was really exposed to a very different kind of hydrology. And so sediment was not necessarily the focus. It was more water quality and flow. And I was just fascinated by the hydrology and hydraulics of that area and how it required really advanced instrumentation to really measure flow and water quality the way that we wanted. And then those skills kind of transferred out west when I moved to. So I kind of evolved into this arena of using these advanced instruments to measure things that they weren't necessarily intended to measure. So using them for sediment and that really had piqued my interest and got me, I think where I am today is the interest in innovative instruments to measure sediment in ways that we previously didn't know was possible.

Stanford [00:04:12]:
Oh, cool. So, yeah, so you kind of led right into my next question. Cause your biography, you start in Tennessee, but you end up in Florida and you start your USGS career. Florida. And then you move to Idaho. And those are two states with panhandles. I think that's the only geographical similarity I can think of. So what are these two geographic settings? What did you take from these two early in your career that kind of knit together to kind of who you are today?

Molly [00:04:40]:
Yeah, I think I learned how very different, different hydrologic settings can influence sediment movement and the types of sediment that can move. And I think I really understood the importance of understanding the sizes of sediment that can't be moved and that kind of thing. I mean, when I started in Tennessee and consulting the interests there and the drivers, there were karst systems, so caves and how things move underground and through the tunnels almost. And that was really fascinating. And then in Florida, the sediment transport was mostly fine material stuff that didn't need a whole lot of energy to be moved and that kind of thing. And then I move out west and the setting is very different. So, you know, we have gravel movement, we have sand movement, we have fine. So what I learned, I think, in that is just how there are very different needs for sediment. And a lot of them focus around the type of sediment that's moving. And also the ecological questions that we try to answer are very different in all those different settings. The reason why we would want to have sediment information is very different in all of those settings. So that's kind of, those are kind of the things that I've learned. I mean, it's a lot more complex and diverse, I think, out here, out in the west than it was in the, in the other areas that I've worked. Another point that I wanted to make is, you know, sediment isn't always on everybody's radar. You know, people don't think of sediment as like, sediment's the big enemy, but, but it drives so many things. You know, it, it attaches to things. It transport other things. It's so much more complex in the way it moves than a lot of the things that we consider are pollutants. So that, to me has been really fascinating to see how those ways that things move and attach to other things very, changes very dramatically across the different hydrologic settings.

Stanford [00:06:38]:
So why is that so? You may be comparing it to, like, nitrogen or like, solutes. Why is sediment so. So weird? I guess.

Molly [00:06:47]:
Well, why is it not so recognized, maybe. Why is it so. Yeah, so, I mean, somebody, you know, you eat dirt as a kid?

Stanford [00:06:58]:
I certainly did, yes. That was, that happened.

Molly [00:07:00]:
So you don't think of sediment itself as something that's bad, really, you know, and you see it in the river. You see it all the time. You know, you need some level of it to maintain just a normal ecosystem, but you get too little of it or you get too much of it. And that could be a problem. And so I guess people are just don't, don't think of it as a bad thing because they see sediment every day and it's not that big of a deal. So, but people recognize things like pollutants like metals and nutrients, nutrients like nitrogen and phosphorus. They know that those are really associated with some immediate impact, but they don't always see the impact that sediment has because I think sediment, too, has maybe a much broader impact. If you look at on a watershed scale, sediment transport at a watershed scale has a huge impact on downstream things. And I think the pollutants that we think of as problems have maybe a much more localized impact in a specific area. And people can see that they know what the immediate impact is. And while sediment has kind of a longer term impact, that's not so easy to blast on the news as a problem.

Stanford [00:08:18]:
So we've introduced you as the national sediment specialist for the USGS Water Resource mission, which apparently is your old title. You're moving into a new position, potentially, but you oversee a lot of sediment data. But sediment data is a term that kind of lumps together lots of different types of measurements that you and the teams that you're associated with collect. So what are some of the types of sediment data that the USGS teams collect?

Molly [00:08:44]:
Yeah, we collect suspended sediment data in rivers, bed load data, bed material data, which is material that's deposited on a riverbed or in a lake bed that's not moving. The vast majority of the data we collect are suspended sediment, I would say, and then bed load, which is the material moving along the bed of a river that's probably the second most common.

Stanford [00:09:07]:
And which of those is harder?

Molly [00:09:10]:
Bed load is definitely harder.

Stanford [00:09:12]:
That would be my guess, because there seems to be a lot less of it.

Molly [00:09:15]:
Yeah, it's harder. And you're actually lowering an instrument down to the river bed. So there are some challenges there with safety. I mean, you get something stuck on the bed and that kind of stuff. That presents some challenges. That's a majority of what we collect. However, we also do a lot of associated data, like turbidity data collection and what's that? So turbidity is a measure of how much, if you shine light into water, essentially, you're measuring how much that light is scattered at an angle based on the amount of material that's in the rivers. So most commonly sediment. So it's a measure of how much sediment is in the water. So we do a lot of that. We also do a lot of geomorphic measurements, things like pebble counts and going out and doing repeat cross sections and measuring the elevations of a river and doing that over time and seeing how things change and move and sediments move and the river changes and that kind of thing. So we do a lot of those kinds of measurements, too. But by far, suspended sediment samples are the most common types of sediment data that we collected.

Stanford [00:10:20]:
And so it seems to me that you described things that are moving in the water and then things that kind of reflect the bed and or shape of the channel. Would that be a good way of thinking about it?

Molly [00:10:34]:
Sure. Yeah.

Stanford [00:10:34]:
All right. And then the things that are moving in the channel, those are suspended load and bed load. And then you're collecting the kind of the mass but also the gradation.

Molly [00:10:47]:
Yes, often, yeah, we most frequently we collect the mass, so we'll collect concentrations and we'll measure loads and things like that. But we typically also collect some information about size, at least, like what portion of the mass that's moving is considered sand. So coarser material and fine material. So the things that are smaller than a certain threshold between fines and sand. So that's typically a very important division that we want to quantify. But often we also collect even more information. So of the material that we consider sand, like how much is in individual divisions or categories of sizes. And that provides a lot of information to modelers and others to determine what's moving and what's being deposited and that kind of thing.

Stanford [00:11:39]:
Yeah. As a modeler, I'm very thankful. So that it's the sand silt boundary that you report most often. And what size is that?

Molly [00:11:46]:
But 62 microns.

Stanford [00:11:47]:
62 microns or 0.062 mm for those who don't think in microns. Yeah. So when I plot those data, the most common data are the suspended segment. And so when I plot those or when we get those, we often plot them against flow. And so we have what we call a flow load relationship. So for a particular flow, we can kind of come up with a relationship of what's the load for a particular flow or what's the suspended sediment for the particular flow. But those data often have a lot of scatter. So that if you plot the suspended sediment data against the flow, sometimes we call it a data cloud instead of a rating curve because it just has an order of magnitude of scatter. So I guess, why is there so much scatter in sediment data?

Molly [00:12:35]:
Great question. So there's a few reasons for scatter. So first of all, sediment itself is highly variable. You know, we have many videos showing how sediment, sediment doesn't move necessarily in a very uniform way down a channel. It's not completely well mixed in a channel. It doesn't just move in one big slug. Often you've got boils and plumes and things like that moving. And we can see that in our data. And so you could be there at one point in time and then come back and measure five minutes later and the conditions could be totally different. And where you're sampling could be totally different, and you could be picking up something very different than you did before.

Stanford [00:13:15]:
The first time I saw that was in a flume, actually, we were in a flume and I was measuring every five minutes. And my answer was zero. Sediment is traveling through this flume because if you sample every five minutes, you miss the pulses and all of the masses and the pulses. I can imagine that's the same in rivers then.

Molly [00:13:34]:
Yeah, absolutely. And one of the videos is really interesting shows with sonar how you can see plumes of sediment moving in the Mississippi river and how and at the same time, how a sampler is moved through that and how, depending on the timing of that movement, you could capture something very different. So that is one big source of uncertainty and variability is that real variability in how sediment moves in rivers?

Stanford [00:14:03]:
And that's not just variability from year to year or day to day. You're talking about, like, within a minute or within a certain process or bed form or, like intra hour variability.

Molly [00:14:15]:
Absolutely, yeah. And then, you know, along those same lines, what you mentioned is there's variability in season and year. You have a really wet year. You have a lot of runoff, a lot of sediment transport. You could measure something very different than you would the following year. Maybe you have a dry year. So that seasonal variability and year to year variability drives a lot of things, too. And that's really important. So that's a reason why we collect data over long periods of time is to try to understand those drivers and what causes those things to change over different hydrologic conditions. So the other thing that drives a lot of the scatter in the relationships that you probably see where we try to plot sediment with flow is because sediment doesn't always have a great relationship with flow. And that is a big challenge, you know, because we go out and we collect samples. We can't collect samples all of the time.

Stanford [00:15:11]:
Right.

Molly [00:15:11]:
So our most common and traditional way of getting more sediment data is to relate that those sediment samples with the flow that we do measure all of the time and develop those curves. And so we can pick off for any given flow what the sediment concentration and load might be. But there are so many river systems where flow has a strange relationship with sediment. So I think of one in particular that I've worked on in northern Idaho, where a portion of the watershed was dammed, which was storing, was keeping most of the sediment from being transported downstream. So, and most of the flow was coming from the, you know, another portion of the watershed, you know, and some other tributaries would contribute a lot of the sediment. So the relationship between the flow releases and the flow that was coming down the river and the sediment concentration was highly variable because the sources of the sediment would change, the flow patterns would change, which wouldn't be tied to when, you know, a lot of the sediment was moving from these other tributaries, things like that. We see that a lot. And because of that, it's driven us to look towards other technologies that are much more direct measures of sediment that have a better relationship with sediment to try to get away from that, some of those inconsistencies with flow.

Stanford [00:16:34]:
So two of the places where people talk about uncertainty in data or scatter in data are natural variability. The actual process itself is very variable or like measurement precision. Our ability to actually get in there and measure it with the tools we have, I imagine both of those are in play with sediment data. But what would you say is the first order cause? Is it natural variability or instrument precision.

Molly [00:17:00]:
As the first order cause of the variability? You know, that's a good question. My gut feeling is that it's natural variability. If it were a perfect world, we would go out there and collect a huge scoop of the river at any given time and just analyze all that water and all that sediment. But we can't do that. So the techniques that we've come up with to subsample the water, they are a snapshot in specific locations. They're tested to assess whether they are representative of what's passing in the full cross section. But technically, we could never know that for sure without scooping up the entire river. But I feel confident that the methods that we have developed as a government agency are, you know, are consistent. They're nationally practiced. They've been tested. The samplers that we use have been tested. The techniques that we use are all tested and consistent. So I feel like we have maybe mitigated as much as we can with sampling and minimizing the uncertainty due to sampling, but we can't. We can't get rid of it all.

Stanford [00:18:10]:
Yeah.

Molly [00:18:11]:
You know?

Stanford [00:18:12]:
Yeah. And I think that's my intuition as well. I was curious about what you thought as the expert. But my intuition is, yeah, sediment data is we have less precision. There's more like measurement uncertainty, but it's actually the processes, the natural variability that causes most of the scatter in the data. So you mentioned a few sources of that variability. You have others that come to mind. If I'm a kind of new engineer or scientist looking at the sediment data, and I see that for a given flow, there are lots of different concentrations or loads. What are some of the processes that I should be thinking about?

Molly [00:18:50]:
Another thing to consider is what we call hysteresis. So sediment movement as flow is coming up, we call this the rising limb of the hydrograph. Okay. So sediment can be very different when that flow is first coming up than it can be as flow peaks and comes right back down. In a lot of river systems, we see the peak in sediment transport being actually before the peak in flow. Depends on the sources of the sediment. But often that that timing is not, it's not consistent and it's not, it doesn't coincide with the peak and flow. So if a big reason for scatter and sediment is not really representing those changes over a hydrograph, maybe you always collect your data on, you know, the, the peak or at the tail end or something like that, or you've got collected all over the place and you can see at a given flow how different that concentration can be. That's a big source of variability and uncertainty. When you look at a sediment flow.

Stanford [00:19:51]:
Curve, that's a big one that's always so surprising as you think. You go out, you get the peak flow, you're going to get the peak load, but probably the peak loads already passed or sometimes it hasn't happened yet. And the idea that 5000 cfs on the rising limb and 5000 cfs on the falling limb of the same event are going to have radically different sediment, it's not a straight line of flow load even for the same event. That's remarkable.

Molly [00:20:17]:
Yeah, absolutely.

Stanford [00:20:19]:
So when I go in to look at the USGS sediment data, the main product are these suspended sediment measurements. There are generally two columns. One is load and the other is concentration. What's the difference between sediment load and concentration?

Molly [00:20:36]:
So sediment concentration, simple terms, is the mass of the sediment that is in a volume of water. So if you were to collect a sample of water, it would be that mass of the sediment that's in a liter of water or whatever that you collect. So the load is the amount of mass that's passing a specific location at a specific time and it could be over a period of time. So it typically when we calculate load, we take the concentration and we multiply it by the flow or the discharge at a specific location, and that gives us the load and then we can look at that over different units of time, over an hour, over a day, over a year, and that kind of thing. So those are the major differences between load and concentration.

Stanford [00:21:21]:
So one is kind of static, it's like mass per volume and the other is a flux. It's like a mass across an area in a time.

Molly [00:21:29]:
Yeah.

Stanford [00:21:29]:
And you can convert between them if you have the flow.

Molly [00:21:32]:
Yes.

Stanford [00:21:32]:
Okay. So I find that there are kind of load people and they're kind of concentration people. There are people that really like to think in terms of load and will only deal in concentration if they have to or we have to because I'm one of those people. And then there's concentration people that are kind of the opposite. What are the relative advantages of thinking about the world in terms of load or concentration? When might you want to do one or the other?

Molly [00:21:53]:
Yeah, it really depends on the science question you need to answer as to which one is more important to you or has its advantages or disadvantages. So if you were concerned about, you know, a particular aquatic organism that's in the river and the local impact of high sediment on that organism, you might be more concerned with concentration because it is a local impact that's right there. It has a direct impact to whatever is there at that point in time. If you are more concerned about, say, a river reach or a watershed and how sediment is moving in that watershed over a broader area over a longer period of time, or modeling to understand how things move through a system, then load would be the thing that you would be more interested in. So how sediment moves in a system, how it gets deposited or gets washed out of the system over time, those are the things that load is more important to answer than concentration. And you think about folks who manage water quality regulations, who monitor to assess those. Some of them are driven by concentrations, and a lot of those have impacts on the aquatic life that's in a river. Some of them are based on loads. So it's looking maybe more as a broad reach scale system as to how much sediment a particular river system can handle over time. And so those water quality regulations are based more on load. So it really depends on the question that you're trying to answer as to which one would be more important to you.

Stanford [00:23:28]:
All right, so you mentioned that the USGS for many years has been working on. You're not taking perfect samples, but you have strategies and regulations in place to make your samples as good as they can be. And one of the things I heard for many, many years before I had any idea what it meant, was that good sediment samples are isokinetic. And that's just, that's a word. There's syllables that have strung together there. You could actually, I think it's probably greek roots, you know, but I had no idea what it was for a long time. What's an isokinetic sample?

Molly [00:23:58]:
It's a great jeopardy trivia question, right? Yeah. So isokinetic meme that the mixture of water and sediment that's moving in the water and that's moving into your sampler device are moving at the same rate so that you're getting a mixture that is the same as what's in the river. So the procedures that we have in place are that we develop these samplers and we develop these nozzles and we give recommendations on how fast or slow you lower and raise these samplers in the water to ensure that this happens to. Because if you don't sample in a way that allows that mixture to come into the sampler at the same velocity as what's moving in the water, then you can get a sample that's either too high or too low of a sediment concentration than what's actually moving in the water. And so that would be bad. You would be either underestimating or overestimating your sediment concentration if you don't use the right sampler and you don't follow the right practices for sampling. So everything that we have in place is to ensure that samples are collected isokinetically to make sure that that water is moving into the sampler at the same velocity as the river. So, for example, if you lower a sampler too slowly, then that sample container will overfill and you'll get an overestimation of the sediment that's moving in. And the opposite happens if you do it too fast. If you lower the sampler too fast, you will underestimate the amount of sediment that's moving in the river. The pressure in the sampler won't equalize and you won't be able to get the set and the sediment into the sampler. So those are the main considerations that we think about when we're sampling to make sure that we sample isokinetically. Now there are times when you can't do that. You can't sample isokinetically. Maybe the water is moving too slow. Yeah, or the river is too shallow and you can't use the samplers that do sample in this manner. And, you know, typically that's okay, because when we, when you think about isokinetic, we're most concerned with sand movement, things like that. And in lower velocities and shallower streams, sand content is typically not pretty low. Pretty, yeah, it's typically pretty low. So we're not so concerned about that. We can use samplers that don't sample isokinetically and still get a representative sample.

Stanford [00:26:25]:
Right. The USGS is involved in an interagency agreement that all of our samples are isokinetic. And sometimes when we go internationally, those data aren't the same.

Molly [00:26:35]:
Yeah, absolutely. Yeah. I see so often in other countries and even in the US with other agencies collecting data, state and local agencies often will go and just collect a grab sample from near the water surface and that will be their representative sample. Oh, wow. And those of us in the sediment world know that especially for larger material, for coarser material like sands, that the distribution, how well mixed the river is, it's not well mixed. And so by getting just a sample.

Stanford [00:27:08]:
From the surface, you've missed almost everything.

Molly [00:27:10]:
Totally. Yeah. You are not representing what's really in the river. So we see that a lot.

Stanford [00:27:16]:
Yeah.

Molly [00:27:16]:
Yeah. Throw a bucket in. There you go.

Stanford [00:27:21]:
Okay, so, so all the data that the USGS collects you make publicly available after it's been vetted. So how do people get to these data?

Molly [00:27:30]:
The number one way that people get our data is through what's called our national water information system, or nwIs. So every data that we collect and publish is available through that system. So you can search for monitoring sites, you can go see what's available and you can download all of the data that have been collected at that site.

Stanford [00:27:46]:
And then recently you've made it available through our packages and an API as well.

Molly [00:27:52]:
Yeah. People can do that by and those things. Ping our national water information system. Yeah. So the, the original source is always that national water information system. Yeah.

Stanford [00:28:03]:
And these are considered water quality data when people are in there looking for it.

Molly [00:28:07]:
So, yes, if you're looking for individual sediment samples, those are considered water quality data. Now, when we generate what are called time series data. So continuous records of maybe sediment concentration and load based on a relationship with something else like flow, you have to get that by going into the surface water component of ENMIs.

Stanford [00:28:30]:
So what are some of the most common errors that sediment data users make when they download the data? You're very generous. You make these data widely available, but we sometimes go in and take these data and maybe don't use them with all of the context. So what are some of the things that you see people do that you thought, oh, you maybe should not do that?

Molly [00:28:52]:
Yeah, that's a good question. Probably the number one thing is not looking at all of the metadata that comes with the data. So we typically store a lot of information about how the data were collected. So what the data are. So is it suspended sediment? Is it bed load? How was that analyzed? Like, what was the method that the lab used to analyze that data? Also, how was it collected? Yes. So what sampler was used? Was it collected from the full cross section or was it collected just from a single point, like an automated sampler in the river? Those are really important pieces of metadata that someone needs to consider, especially when they're comparing multiple samples. You wouldn't want to necessarily compare a sample from an automated sampler from a single point with samples that were collected from the full cross section without some context. So that is a common thing that I've seen people do.

Stanford [00:29:48]:
So I just need to confess that I'm very guilty of this, that I'll either use R or we have this new tool in raz that does it, or I'll just take the data and put it in Excel spreadsheet. And once it's in there, I strip out everything but the numbers, and then I plot it, and then every point kind of lives on its own, as this is true. And so what sort of codes or metadata should I be looking for when I'm in there dealing with these data?

Molly [00:30:16]:
So the most common ones that we store that are readily available are there's a parameter code which describes what the data are, and then there's information about how the data were collected. We also store the sampler information. There's a metadata for that. And then there's a method code which tells you what methods were followed generally to collect the data. And sometimes you can get laboratory information about how the data were analyzed. Was it analyzed using this technique at the lab versus another technique? And could there be some differences there? So those are kind of the most important things to look for when you're looking at the data and all those codes are in there. Sometimes if you're looking at a big data set, it can be overwhelming, but those are the things.

Stanford [00:31:00]:
So what a good practice be when I'm bringing these data in to maybe color code them differently based on how robust the sampling practices.

Molly [00:31:10]:
So, yes, color coding it is a good idea, and I hesitate to say how robust the sampling practice is because all the sampling practices are robust. But sometimes we collect samples in different ways for different reasons. So the example would be, so maybe we have an automated sampler out there that's collecting samples during a storm event when we can't be there. Okay, so that information is really useful to capture the sediment transport over a specific storm, particularly in maybe a really flashy river that changes very quickly and dramatically. So we collect those data in addition to full cross section samples. And so both of those data sets are useful and they're robust, but we just need to put them in context and compare one to the other. Typically, what we will do with the. When we use an automated samplere is we will develop a relationship between a full cross section sample and a sample that's collected by the automated sampler. And so that we can estimate what the cross section concentration would be based on those point samples by developing that relationship. So both data are very useful, but you just have to compare them in apples to apples in the right way.

Stanford [00:32:25]:
Well, that anticipates my next question because I think the automated sampling data, whenever I go in and see lots of data, I'm excited because there's lots of data, but I'm also hesitant because I think, well, there's no way that a USGS team went out and made all these measurements 24 hours a day for six weeks or whatever. And so I see that there's probably an automated sampler involved, but I don't really know how should I approach automated sampling data versus direct samples? And as far as the uncertainty of those data, or do I need to post process them or has the USgs already post processed them to compare them to direct samples?

Molly [00:33:05]:
Yeah, that's a good question. So we never directly will edit a sample concentration in the database. So the data that you see are the actual concentration data. They're not adjusted data. But we, on our end, if we use those data and we want to relate the point sampler data to the cross section, we will develop a relationship on our own and use those adjusted concentrations in like comparisons and stuff like that. But we never edit the raw data. So one thing that you could do to do that on your own is either you could reach out to the office that's servicing that site and ask for those relationships, or you can look for when a cross section sample was collected at the same, around the same time as an automated sample and develop your own relationship with that set of data to come up with and do your own computations as to what the automated sampler would say about the full cross section concentration. So that's a way to do it. Another thing I wanted to say about autosample data is if you only take samples that are collected over a storm event, a single storm event or two storm events or whatever, and they're all collected over a short, in short succession and things like that. You run into some possible problems there. So I've seen practitioners just pull like, data from one or two storms and think that they fully represented the range of transport conditions.

Stanford [00:34:30]:
Oh, yeah.

Molly [00:34:31]:
So that's a concern. You really need to understand a wide variety of hydrologic conditions and how those drive sediment transport. If you use data from only one or two storm events and use it to do any kind of linear regression with flow or another parameter, then you run into what's called zero correlation.

Stanford [00:34:50]:
Right.

Molly [00:34:51]:
So it's a condition that violates the rules of linear regression. And so that's a concern. So that's another reason why you don't want to just focus on a lot of samples over collected over a very short period of time is because the samples depend on one another.

Stanford [00:35:05]:
You know what I mean?

Molly [00:35:06]:
You're not really getting snapshots over different conditions to represent what's really going on.

Stanford [00:35:11]:
And that probably relates to one of the very first things we talked about is there's these different scales of natural variability. There is within an event scale of natural variability, but there's also a season to season scale and a year to year scale, even a decade to decade scale.

Molly [00:35:27]:
Absolutely.

Stanford [00:35:28]:
All right, so in these podcasts, some of which will run after yours, there's been some expression of various expressions of trust of data. And I think that one of the things I'm interested in is with your experience with collecting and analyzing sediment data, do you think that it makes you more or less confident in the result than, say, a downstream user like me?

Molly [00:35:55]:
I would say I feel more confident in the data because I know that we have developed samplers that have been tested, that have been shown to be representative and sample isokinetically. I know we're using nationally consistent methods. We're representing as much of the river as we feel like is necessary to represent what's really going on. So I would say I feel more confident than probably others who are just using the data and seeing all that scatter. I feel confident in the data that we collect.

Stanford [00:36:29]:
Okay, so maybe changing gears a little bit, you're obviously one of the national specialists here in the United States, but you've also consulted with several agencies in Asia and South America. But we've overlapped on a couple of those projects, actually. Where and why have you gotten involved in some of these international programs?

Molly [00:36:46]:
Yeah, I've always been really interested in international efforts. I mean, I think it started for me in college. I left college for a year and went out of the country and experienced South America and lived in Australia for a little while. So that sort of piqued my interest. But in my career, you know, my career has taken me to the Middle east, you know, in Asia and South America and Central America and all over. And I really enjoy the different perspectives on water and the different challenges that people face with water security and just having a handle on. It's amazing to me how few people across the world know what kind of water they have, how much and the quality of their water.

Stanford [00:37:29]:
Yeah.

Molly [00:37:30]:
So, yeah, so that's kind of. I've really found that to be one of the most rewarding parts of my career is to be involved in international work.

Stanford [00:37:38]:
So what are some of the challenges of setting up a sediment data program in other settings? And if you kind of look at the international settings that you've worked in, are there some conditions that have. That the successful programs have had in common?

Molly [00:37:52]:
Yeah, so the most successful programs have had plenty of funding to support them, infrastructure in place to support that, and also a willingness of a community to take on something and own it. You know, the most successful programs have been where actually a lot of the ideas came from the community or the group that you're working with. And they own it. They have pride in it, they own it and you know that they're going to take it forward. So those are some of the big things. So I find that, you know, safety and security is a big challenge internationally. I mean, you can't just put out a bunch of, you know, expensive equipment and expect it to stay there. So that's a big challenge for people. And then just, you know, safety in getting out on rivers and on bridges and things like that is a very different world compared to the US, given all the safety regulations we have. Another thing that I find is really challenging to convey is the importance of monitoring in extreme events. We always say in the sediment world that 90% of the sediment load happens in 10% of the time the events. I find that a lot of international entities don't want to be out there for safety reasons. They don't want to be out there when the majority of things are, are moving. And so that really limits them as to what they can collect and what they can say about sediment moving in rivers. Another big thing I see is that the people who interpret and try to make decisions based on data are usually not the people that are out there collecting the data. So when they need to make a decision or a change in something, they don't have that firsthand knowledge of what was happening at the site and what was causing this particular anomaly or this data to happen. And there's that disconnect. I find a lot that I hope will improve in some agencies, but that's a problem.

Stanford [00:39:54]:
Sometimes I feel like that's a problem for us too, because sometimes we go and just download USGS data and then tell a story about it. And there probably should be a lot more integration with the people who actually collected it.

Molly [00:40:07]:
Yeah, absolutely. I think of one particular example, a story in Chile. The government in Chile, they were really trying to assess what their water resources were, the sediment loads in particular, in advance of a lot of hydropower development in the region in southern Chile and Patagonia. The people who were developing the ratings, the discharge ratings up in Santiago had no sense of what the rivers looked like or what was happening. And so they were making decisions about how to change these ratings that were not based on, didn't make hydraulic sense, based on the conditions in the rivers where they were collecting the data. And so that disconnect really caused them to make decisions that were, probably gave them lower quality of data because they didn't have that really understanding of what might be possible hydraulically at a site.

Stanford [00:40:55]:
So you mentioned the 90 ten rule that 90% of the sediment moves in 10% of the time. When I teach my sediment transport class, I tell them that if you take one thing away from this class, it should be sediment transport is nonlinear. All right. I think it's the most interesting and important thing to understand. How does USGs and your agencies and your teams, how do they deal with that here? Do we try to get out and measure sediment at those high events? Is that part of your culture?

Molly [00:41:22]:
It is, yeah, definitely. Yeah. We teach classes specifically, too, in where we focus on how do you, what are the considerations for getting out there in high flow and high sediment transport conditions and collecting a sample safely? Because we recognize the importance of that. But yes, it is sort of ingrained in us to build that into our sediment monitoring strategies and programs is to make sure that most of our samples are collected in high flow, high transport events.

Stanford [00:41:51]:
So, Molly, what is a surrogate, another word that got banny around a little bit before I understood what it was?

Molly [00:42:01]:
Yep. So it's a great, you know, the. The whole conversation we had before about the inconsistencies between flow and sediment is a really great lead into the concept of a surrogate. So a surrogate is a substitute for something else. So you think about a surrogate mother or something like that.

Stanford [00:42:18]:
The first thing that comes to mind and kind of a weird association with.

Molly [00:42:21]:
Sediment measurements, you know, and so really what it means is that we have identified something that we can measure more easily than sediment, but it's a substitute for sediment. It can be related to sediment. We can get more information about sediment because of this more easily measured parameter. And so because we, over the years, have understood that inconsistency in the relationship between flow and sediment, we started looking at other technologies that might give us a much more direct measure of sediment than flow. Okay. Flow or discharge. So the things that we have looked at to be surrogates for sediment, most commonly are things like turbidity. That's a common one. We've also developed a lot of techniques associated with what's called acoustic backscatter, which is kind of like sonar.

Stanford [00:43:12]:
All right.

Molly [00:43:12]:
You send sound out into the water. That sound bounces off of things in the water like sediment, and it gets reflected to the instrument.

Stanford [00:43:19]:
This is remarkable to me because sediment is not very big. Most of it is not very big, but it reflects sound in water.

Molly [00:43:26]:
It does, yeah. You pick the right frequencies of sound and it reflects it. Absolutely.

Stanford [00:43:31]:
And it's different frequencies for different sizes.

Molly [00:43:33]:
Yeah. So different frequencies are more sensitive to different sizes.

Stanford [00:43:38]:
Okay. Oh, wow.

Molly [00:43:39]:
So you can use multiple frequencies and get different responses for the types of sizes of sediment that are, that are moving. So that sound reflex, it gets returned to the instrument, and the instrument measures the strength of that reflection. So with those technologies, we can put them in a river. They can be measuring all of the time. We go out to the site periodically and we collect physical samples, and we relate the results, the concentrations and loads of those samples, to the readings at the time we were there. And we develop what's called a surrogate relation, a rating.

Stanford [00:44:12]:
Okay.

Molly [00:44:13]:
And so then, just like the flow curves that you were talking about, we can use those ratings to pick off for any reading of turbidity or backscatter. We can get an estimate of what the sediment concentration and load is at that point in time based on that rating. So at a simple level, that's what it is. And it really, this concept really allows us to get sediment data more available to people, ultimately at a less cost. Once we develop that relationship, we don't have to sample as much. And so hopefully, over time, that results in a less cost to give more sediment data.

Stanford [00:44:52]:
So you're still going out there and taking some physical sediment samples, but instead of relating them to flow, you're relating them to something else that you can measure continuously. That's a better relationship.

Molly [00:45:02]:
Exactly.

Stanford [00:45:02]:
And then you measure that continuously and you have this relationship that connects it to load or concentration.

Molly [00:45:09]:
Yep. Absolutely.

Stanford [00:45:11]:
So federal agencies aren't the only ones that collect turbidity data. A lot of times there'll be like a water treatment plant downstream that has to collect turbidity data. So when a modeler is looking at their system and they don't have USGS data, a lot of times they'll say, yeah, but the water treatment plant has turbidity data. Can I use turbidity to parameterize my sediment model? What would you say to that?

Molly [00:45:37]:
So I would say that it really depends on the sediment that you're concerned with. Turbidity is very sensitive to very fine sediment, fine material and it will pick that up, you know, great. It is not very sensitive to sands, coarser materials. So, you know, we have these great plots of where we collected profiles in various rivers. Particularly the one that comes to mind is in the Missouri river near St. Louis. We collected these profiles of sediment samples and turbidity data at all those points. And then we've used other instruments to collect profiles of sediment through water column. And you see almost no variation with turbidity, with depth. And in our samples you can clearly see how sediment concentration, because of increased sand transport near the bottom, right. You can really see that in the samples, but you don't see any of that picked up in the turbidity. So if your concern is fines and that's really all that's moving in your system, and turbidity might be a nice way to, you know, represent sediment, but turbidity is affected by a lot of things. Turbidity can be affected by the color of the water and other things in the water. It's not on its own, it's not a direct measure, but it can be a surrogate or allow you to estimate roughly the level of sediment, a fine sediment that might be moving in a river.

Stanford [00:47:04]:
So what, what would you say? You've kind of been in this game a while. What would you say is the most important innovation in the last 20 years that's changed the way you collect sediment data and change the kind of sediment data that you're making available. But maybe someone downstream wouldn't really be aware of what are the kind of mature innovations in the data community that maybe the user community has been slow to adopt?

Molly [00:47:28]:
I would say back to our previous discussion about surrogates. So the concept of surrogates is people don't understand them a lot of times and don't always know when we've developed those relations and have those data available. So I think that's probably a lot of those technologies like turbidity and back scatter are fairly mature. And so that's an innovation that's happened in the last 20 years that I think people are probably underutilizing as a resource to increase the amount of sediment data that could be available.

Stanford [00:47:59]:
So if I was a user on a system that has some backscatter data, would those be available on the website?

Molly [00:48:06]:
So if someone has developed that relationship between the backscatter and the sediment? Yeah. They would be available on our website.

Stanford [00:48:12]:
And they would be available as loads or concentrations. Yes. What about looking forward? If you were to kind of look forward in the next 20 years? What do you think are the next 20 years for sediment date? What are the big things that are promising?

Molly [00:48:25]:
Yeah, I think we are going to rely much more on non contact and remote sensing methods. So non contact means that you don't have something physically in the water that could be damaged. It could be a camera that's sitting above the water. It could be a drone that's flying over a river reach and things like that. Remote sensing techniques have really advanced and that kind of stuff. So I think we will see a lot more of that. I don't think that we are going to be getting away from physical sample collection anytime soon because really that still is our closest measure of the truth and we need that to validate all these other methods that we're using. But to get an estimate of sediment concentration or load by not even being in the river at a given point in time just to get some information, I think that is really going to advance the other area that I think that we are already actively advancing and I think is going to get more operational. We use a lot of instrumentation for measuring flow all of the time at stream gauges across the nation. So we can leverage some of that same instrumentation to get an estimate of sediment because it has that underlying concept of acoustic backscatter. So it's still bouncing off of particles in the water. So we are developing methods to get to where every time we collect a flow measurement, we can get a picture and an estimate of the sediment concentration and load at the same time. Getting there would be so valuable because it would just expand the amount of sediment data we have at all of our monitoring stations across the US. And to make that available would be, to me, transforming, like it would be transformational, like what that would do.

Stanford [00:50:08]:
I guess you have to be a special kind of person to see that as a beautiful, beautiful future. But I'm that kind of person. Right where the idea is if you go to a sediment gauge on the data portal and you're kind of playing data roulette, you're like, how much data are they going to have? And is it going to be in the right time period? But the idea that every water gauge could eventually have pretty consistent sediment data, that is a transformative future.

Molly [00:50:35]:
Absolutely. Yep. Two other things that I'll mention that I think are in our future is one is my agency is working towards more nationally available predictive models for sediment using a lot of different tools. I mean, broad, regional based statistical tools down to, you know, individual models that have been developed in a watershed. So having that available nationally will be really powerful. And I also want to get to a place in the future where we share data more effectively among all of the data collectors out there. And we make different kinds of data more available. So right now, the way we serve data is typically at a single point. Can we share information that's collected along an entire river, reach more effectively a bathymetric survey of a river? And can we share data among agencies much more effectively and make it available to everyone? So that's a future I want to get to. And I think we'll get there. Just a lot of decisions to be made along the way.

Stanford [00:51:39]:
I know you've been involved in that. And those are the two kinds of data that as a modeler I'll use is I go to the G's website and I get all the flux data I can find. But then I have to go out and hunt for any sort of bathymetric data that was available. Because it is kind of a different, different conceptual space model. But that also is a lovely future.

Molly [00:52:00]:
Yeah, absolutely.

Stanford [00:52:02]:
Okay, so I've been trying to ask this question in most of the podcasts. And so let's wrap up with this. If you had to start your career over, let's say that you're in your mid twenties and you're just getting into rivers, except this version of you, the one that has all the experience and has spent years doing this, you get to give the novice version of you that's just starting two pieces of advice about rivers or sediment data. Just at the very beginning, what wisdom would you pass on that would give that novice version of you the biggest shortcut to becoming the expert version of you?

Molly [00:52:37]:
Wow, that's a big question. Number one, I would just say, be flexible and see where your career takes you. I never would have expected that I would be focused on sediment, you know, along the way. Yeah. But this is where, you know, my interest took me. And it's really fun and exciting. So I would just say, don't expect that you need to have everything figured out right away. Just see where life takes you and follow, you know, different interests and see what interests you the most.

Stanford [00:53:07]:
Great.

Molly [00:53:07]:
Yeah. The other thing I would say is when you're evaluated with different decisions in life, go with things that give you that gut level of excitement. Because if you're doing things that excite you, it's going to show up in everything that you do. You're going to do great at it. You're going to, you know, it's going to be contagious. People will get excited about things because you're excited. So go for those things that give you that gut level of excitement.

Stanford [00:53:32]:
Molly, anything else that you think that beginning sediment modelers or scientists should know?

Molly [00:53:38]:
The only thing I would say is that working with other people, other experts in my field and other agencies has been really fun. You know, really expand your horizons and reach out to those kinds of people because you get completely different perspectives on whatever it is you're working on. You know, those kinds of partnerships and relationships are, I think, foundational to growing yourself in your career and your skills.

Stanford [00:54:05]:
Yeah. It's not an accident that I wanted you on the first season of the podcast because I've had that experience working with you. Just the perspective that other agencies bring with the different questions they have to ask is just irreplaceable. All right. Well, Molly Wood, thank you for being on the podcast.

Molly [00:54:20]:
Yeah, thanks for having me.

Stanford [00:54:24]:
Like Molly, I work in a number of international settings, and the more I do, the more I appreciate the great work the USGS does and the uncommon resource that our nation's sediment data program is. I really appreciate Molly taking the time to help us make the most of it. We've uploaded links and bonus content, including video shorts of podcast clips and cuts at the podcast website linked in the episode notes and Molly provided some really cool sediment videos for these this week, and we've edited them into videos of this conversation. You'll want to check that out, subscribe, or check back in a couple weeks for the next episode, where we'll talk to senior core sediment modeler and Einstein award winner Doctor Ron Copeland. I talked to Ron about the early days of river restoration, numerical modeling, and Mississippi river morphology. This podcast was funded by the Regional Sediment Management program of the US Army Corps of Engineers and is part of the tech transfer mission, which we like to call RSMU. Doctor Katie Boucher leads the RSM program, and Tate McAlpin is the RSM inland lead who oversaw this project. This season's music is by Mike Loretto. This project was also supported by Core's Flood and coastal storm damage reduction R and D program and the Hydro hydrology, hydraulics, and coastal science and Engineering Technology program. H Agency set this is an informal conversation and the views expressed by the hosts and guests do not necessarily reflect official positions or policies of the USAC or USGs. I'm Stanford Gibson, the sediment specialist at HEC. Thanks for tuning in to the RSM River Mechanics podcast.