Continuous Monitoring for Nutrients: State of the Technology & Science

Download Video
Right-click and save to download

Detailed Description

Making water quality measurements that capture rapid changes due to storms and other events has long been a challenge for accurately measuring the sources, loads and cycling of nutrients in lakes, rivers and streams. However, advances in situ sensor technology and communications over the last 10 years has revolutionized the way water quality monitoring and research can be conducted. In particular, in situ optical sensor measurements for nutrients such as nitrate and orthophosphate are yielding significant insights into the sources and timing of nutrient transport, as well as real-time data for decision support. This talk presents the state of the technology for continuous monitoring of nutrients in rivers and streams and several examples from USGS studies that highlight the opportunities, challenges and importance of making comparable, high quality measurements in our Nation's waterways.


Date Taken:

Length: 00:56:21

Location Taken: US


So, just up front to acknowledge some of the
people who have been really pioneering this

work, and also a list of people who have shared
some information with me for this talk.

It's largely a synthesis of some USGS work,
but certainly there's a much bigger community

of people working on nutrient sensors and
hopefully some of the big ideas at least reflect

their vision on how we can use this new tool.

Also the funding: acknowledge NAWQA and the
Office of Water Quality at USGS because they

are funding a significant part of this work
but there are lots of federal and state cooperators

and partners that are funding water quality
monitoring for nutrients across the country.

So I am not going to spend too much time trying
to convince you that continuous monitoring

is useful.

I think that is pretty well established at
this point; you can learn a lot and also improve

management when you have more data.

In the case of water quality, now we have
instruments where you can take a measurement

all day, every day.

We can make measurements for a wide range
of constituents, or proxy measurements for

a host of things that we cannot measure with
an instrument itself.

We are making measurements on intervals of
seconds to hours rather than the traditional

approaches of collecting discrete samples,
weekly, monthly or seasonally.

We no longer miss events, so these concerns
of our sampling scenarios bias the high flows

or low flows are sort of alleviated when we
collect a lot of data.

And also remote access and control of sensors:
not only the real time component of this,

we're no longer waiting two or three weeks
to get a data point from a lab but we also

have the ability to communicate with instruments
and control how they operate, in some cases,

control where they are on the landscape, and
I'll show you some examples of that.

The state of the technology and the state
of the science are the focus of my talk so

I've broken it up into two parts.

This first part I'm going to give you I guess
my sense and the folks that I work with of

where we believe the state of the nutrient
sensor technology is, and two main points

up front: continuous water quality monitoring
is obviously not new.

We're really focused right now on this nutrient
sensor issues, but we've been making measurements

of water temperature and turbidity and everything
in between for a long time, so the idea behind

this and the instruments are not new, it's
this next generation of tools that is emerging

and is really powerful for water quality monitoring.

The other point I wanted to make up front
is that continuous monitoring, at least for

nitrate, is beyond proof of concept.

You know, the question we hear a lot is: is
it proof of concept or field ready?

We have passed Proof of Concept for some of
these instruments but that doesn't mean that

they are simple, cheap, or easy to operate.

I'll give you a little bit more information
on all three of these.

And note too that I have nitrate here as beyond
"proof of concept."

There are some other technologies for phosphate
and ammonia that maybe haven't gotten over

that hump yet and I'll let you hear a little
bit more.

So I may start with where we've been so that
we have a better sense of where we're going.

We've had ion selective electrodes for nutrients
for a long time, you know, direct potentiometric

measurements between a sensing electrode and
a reference electrode.

These instruments have been available since
at least the 1970s; for many of the people

doing environmental monitoring, there are
a few manufacturers who have built these for

a number of years and they integrate them
into sondes.

They certainly have advantages, the biggest
is that they are inexpensive, on the order

of $400-$500 to maybe 1000 dollars.

They are available for a wide range of constituents:
nitrate, ammonium, potassium.

They're pretty easy to use and have a big
range but there are some real disadvantages.

The resolution accuracy and sensitivity and
precision tend to be pretty low or you have

to invest an awful lot of energy into maintaining
the data quality.

They are subject to ionic interferences, they
can drift, they have fouling problems.

So this is a place we have been.

They are still in use.

From our perspective in the USGS, we don't
use these widely for sure but they are still

a technology that's available for continuous
nutrient monitoring.

Another approach to getting at nutrient concentrations
in situ and continuously and in real time

is wet chemistry.

This is essentially an optical measurement
but reagents are required to develop color

that can be measured optically.

So there are a number of advantages: the resolution,
accuracy, and precision of the instruments

can be really good.

You can make measurements of multiple constituents
doing wet chemistry: nitrate and ammonium,

orthophosphate and silica.

The response time is relatively fast, you
know, you typically need time for chemistry

and color to develop so you may get a measurement
every half hour or every hour but for all

intents and purposes, that resolution is pretty

There are some real disadvantages.

Cost is one that always comes up; these are
typically $15-$20,000 instruments.

They have pretty high power requirements.

They have a high potential for fouling.

And they have high maintenance costs, in particular,
they require a reagent and this is just one

example that you see here.

There are three reagent packs on this instrument
that mix within a central chamber and develop

color and give you measurements.

But not only do you need to replace reagents
that degrade over time, you are also generating

waste and so these are some disadvantages
and challenges for the wet chemical sensor


But, it's been around for a while, it's a
viable option, and there are some parameters

that you can't measure without wet chemistry
presently, like phosphate.

And so more recently, a lot of work has focused
on optical sensors, in particular UV sensors

for nitrate; they measure absorption or more
accurately they measure the light that is

transmitted through a sample and measured
by a detector.

They've been available for a while and the
essential design is pretty much the same for

all of these: you have a light source that
emits light in the UV, you have a detector

that detects light at particular wavelengths
that passes through a sample and certain constituents,

like nitrate, that absorb light at particular

And so the reduction in light, or the change
in transmittance through a sample, can be

correlate to the concentration of that material
in solution.

Not everything absorbs light; nitrate does
and that's where there's been a particular

focus on UV nitrate sensors in surface water,
and their resolution, accuracy, and precision

on the instruments tend to be really good.

You can get instruments that cover a large
measurement range related to the design of

the sensors.

They are chemical free, which is obviously
a plus.

They respond very quickly; you can get a measurement
every second from one of these sensors if

you so desire.

And there's a lot of other information: it's
a spectrophotometer, traditionally people

have used spectrophotometers for surface water
monitoring as excellent ways to better understand

the concentration and type of dissolved organic

These instruments are also recording information
about organic matter even though they are

packaged typically as a nitrate sensor.

Disadvantages: they are expensive, that's
probably the biggest one right now and maybe

the greatest barrier to seeing a larger number
of UV-nitrate sensors in the landscape.

Typically $15-$25,000.

As I mentioned, not everything absorbs light
without wet chemistry, and so nitrate and

technically nitrite are what you can measure
with the current generation sensors.

They require pretty high power; they can require
a lot of maintenance and certainly folks investing

in these should be very careful to think through
the operation and maintenance costs for sensors.

And they are subject to a range of optical
interferences; I'll share more about that.

So essentially what has happened is a lab
instrument has been repackaged so that it

can go right in the water.

This is a picture of a bench top spectrophotometer;
what you don't see here is the computer that

is typically used to operate the instrument.

What manufacturers have done over the last
20 years or so was to miniaturize the component,

built rugged housing, worked through how to
more efficiently handle power, a log internally

to control the sensor, built in anti-fouling
systems like the wipers or compressed air,

and can do on-board data processing.

Essentially this fairly large bench top instrument
has been repackaged into something that is

3-4 inches in diameter and a foot and a half
or two feet long.

All of the UV-nitrate sensors in the market
have essentially the same components at this

point, and there are three main features;
there is a light source on one end, the lamp

types differ but it's either xenon or deuterium.

There is a sample path that the water passes
through and the light is passed through that

gap in the water sits, and then there is a
detector on the other end that measures transmitted

light at specific wave lengths.

They all do the same thing, but they all are
different in other ways.

Those differences affect the measurement range
and the accuracy of drift, the tolerance for

interferences, the power, the field maintenance,
and ultimately the cost.

And so here are the ways that they differ:
history in particular is a really interesting


This top picture is from Monterey Bay, a research
institute, where they initially developed

a UV-nitrate sensor called Isis, maybe 20,
25 years ago for coastal applications.

Here's one where it's on a buoy whereas the
other primary manufacturers have come from

the wastewater side.

So the difference between a coastal ocean
and a wastewater treatment facility is pretty

significant and that's been part of the design

As much as the environmental monitoring, obviously
freshwater, is somewhere between those two.

We don't have very low nitrate and clear water
in many cases, often times we're dealing with

particles and nitrate concentrations that
can get quite high.

At the same time, we're not dealing with the
sludge and other issues that are part of the

design for wastewater applications.

As I mentioned, they differ with light sources,
they differ in terms of the measurement path

length, and the types of materials that are

They differ in terms of spectrophotometers
and I'll mention a little more about that


They use different algorithms, they use different
references, and they also have different anti-fouling


I'll give a few quick examples of each.

UV Nitrate Sensor design: there are a few
things that are really important for making

high-quality UV nitrate sensor measurements.

One of them is, you can get a variety of pathlengths.

This picture shows just a few of them.

On the left is a 35 mm pathlength sensor,
this middle picture is a 10 mm pathlength

sensor and this one on the right is a 2 mm
pathlength sensor in this little gap you see

up at the top here That's a really important
design feature.

As you can imagine, you're trying to get light
from one end of this instrument to another

through an environmental sample and so things
like very high turbidities make it incredibly

difficult to get light through the sample,
in which case you'd want a very short distance

that the light needs to pass through.

This also affects the range of nitrate concentrations
that you can measure.

It affects the active dissolved organic matter
on the sensor measurement.

So it's a design feature and it's selectable
by manufacturers but it's a really important

component to making good measurements.

Another is as I mentioned they use different
spectrophotometers and this is pretty important

because lots of things absorb light in the
UV range.

This is a figure from Mbari again.

What you see in the blue here is the absorbance
at a number of wavelengths.

Nitrate typically 220 nm is used as a wavelength
to measure nitrate absorbance but you can

see bromide is absorbing light, nitrite absorbs
light in this range, and we're not even showing

in this image dissolved organic matter which
can certainly absorb a large percentage of

the light at 250, 260, 280 nm.

So there are some instruments that simply
take an approach where you can measure absorbance

due to nitrate, and then something else.

I think the standard method for this for a
lab measurement says look at 280 nm as the

comparison for everything other than nitrate.

There's an instrument that looks at 350 nm
and then there are instruments that look at

every nanometer, up to 256 of them, all of
those wavelengths to calculate the nitrate


There are differences in the design that affect
how a sensor can handle interferences from

a variety of other constituents.

It's an important consideration, you have
to think through if you're in a coastal setting

you should really be measuring absorbance
so you can account for bromide.

If you are in an environment with very high
dissolved organic matter, you would want an

instrument that can account for that.

So it's this design criteria that's part of
the development of the spectrophotometer and

related to the history of the sensor application,
but it's a very important criteria.

And the other is the algorithm, so ultimately
what the instrument is doing is it's taking

a bunch of data on transmitted light at different
wavelengths, subtracting out some interferences,

and it gives you the output of a nitrate concentration

The algorithm is obviously critical because
that's what's converting absorbance or transmittance

into nitrate concentrations.

And these vary by manufacturer.

Some have a single algorithm that you can
consider sort of a global algorithm, some

recommend developing more local-scale calibrations
where you would collect a bunch of discrete

samples and get the sensor measurements and
compare those too and build your own algorithm.

It's really important to getting good data
from these sensors.

So a quick example: here's the same sensor,
in the same solution, with a different algorithm.

So this is nitrate concentration on the y-axis,
we started with a standard of 1 mg/L and we

varied the dissolved organic carbon concentration
in that sample.

And you can see the river calibration does
fine, if you were to use a drinking water

calibration it would really over estimate
nitrate concentration.

If you used an influence calibration it's
also underestimating now.

So this is again one instrument, just changing
which algorithm you want to process the data.

It's a very important consideration for ultimately
getting high-quality data out of these sensors.

This is probably the biggest one.

A dirty optical sensor cannot give you good

It's based on light passing through a window,
through a sample, and through a second window

to a detector.

Whether you're talking about, in the upper
right here, this isn't a nitrate sensor but

it's a fluorimeter, whether you're talking
about really fine algal material or you're

talking about small organisms or you're talking
about big organisms, you need to keep these

sensors clean in order to make good measurements.

There are a range of antifouling approaches.

Wipers are probably the most popular, some
manufacturers use compressed air to clear

out the flow path, others recommend the use
of copper or other biosides.

There are lots of options but its critical
for an optical sensor to be clean to get good


So back to wet chemistry: in particular, we
can't make measurements of orthophosphate

or ammonium or silica by just interaction
of light and constituents alone.

So we do need reagents and this is probably
the most widely accepted or emerging approach

in doing that.

The instruments are currently available for
orthophosphate and a number of other constituents,

I think the orthophosphate sensor is probably
the most stable at this point and has been

the focus of more recent work, certainly by
the USGS although we're only operating something

like 5 of them across the country.

Part of the challenge is figuring out how
to maintain those sensors for continuous monitoring

in the long term.

Short term deployments are typically successful
but anything where you're trying to run a

sensor for months at a time gets trickier.

The detection limits are really low and the
ranges are good.

We can get samples every 30 minutes or so
and you can typically collect something like

1000 samples from any of these instruments
before you ultimately have to change the reagents.

So a lot of people, maybe you're thinking
this yourself, but a lot of people often ask:

what do I do about total nitrogen or total
phosphate if that's what I'm concerned about?

Nitrate is fine but I also have to account
for particular nitrogen and ammonium and dissolved

organic nitrogen?

And unfortunately there's no instrument that
can really do that.

You may read spec sheets that say they can.

In some cases you could probably put unicorns
and fairy dust on the spec sheets too.

But the reality is the standard lab methods
measure total nitrogen or total phosphate

require chemistry, typically alkaline persulfate
oxidation of that sample or really high temperature,

meaning you have to combust the sample, and
no in situ sensors at this point can do that.

So here are two examples though, where on
the left is nitrite plus nitrate measured

in situ or actually from lab data versus the
total nitrogen from lab data on the Mississippi

River at St. Francesville, you can see that's
a pretty good correlation so you may be able

to develop some sort of surrogate approach,
whether you use the nitrate sensor data or

other data as a proxy for total nitrogen.

In some place surrogates are trickier; on
the right you see again lab nitrate plus nitrate

data, versus total nitrogen for the Potomac

The Potomac has a greater influence of organic
nitrogen and also particulate nitrogen at

certain times.

Nitrate as a proxy for total nitrogen is obviously
not as strong of a relationship but there

are other options for developing proxies.

I want to highlight the work that Andy Ziegler
and others have been doing until Kansas at

the USGS and point you towards the National
Real Time Water Quality site and the surrogate


Here's one where the computed concentrations
of loads from total phosphorous are based

largely on turbidity measurements.

So what you see here is, this is for the Little
Arkansas River, this year to yesterday and

you can see discharge in blue but then you
can also see this black line with the red

dots which is computed instantaneous total
phosphorous concentrations.

The nice thing about this approach is that
you can also apply uncertainty to the measurement

because it is a model, it is a surrogate,
and so the uncertainty becomes a really important

component of that prediction.

And you can do this for a wide range of constituents,
not just total nitrogen and total phosphorous

and e.coli and a variety of other parameters.

It becomes a bit of a black box and it's site

In some cases there are a number of parameters
that are a part of the model, in some cases

there's only one.

But anyway, this is a viable approach at this
point for things that we can't measure.

To do any of this, whether it's UV-nitrate
or surrogate approaches or wet chemical sensors,

it's critical that we have guidelines and

So we have to continue to characterize instruments.

We have guidelines for a range of instruments
and we have to continue to work with manufacturers.

In some cases, particularly if they come from
wastewater or oceanography, fresh water monitoring

is not entirely sure of the specifications.

What sort of turbidities do we deal with?

What sorts of nitrate concentrations do we
deal with?

For a lot of these folks, nitrate concentrations
in parts of the Midwest, like in the flooding

period following a drought n 2012, if we had
20 or 30 or almost 40 mg/L at sites and for

a lot of folks, that was a big surprise.

We have to continue to work with manufacturers
to get the right instruments for freshwater


Part of this, we need training, we need time
in the field to develop methods, and then

if you're interested there is a recent guidance
document that the USGS put out as part of

a series of techniques and methods approaches
for optical nitrate.

So that's sort of a general State of the Technology
from my perspective.

A lot of these instruments are field ready
but they still do require a lot of attention

and we still need to put more effort towards
common protocols and standards.

But when you do that, there are a lot of really
interesting questions you can ask.

So I want to get into some of the state of
the science, it's just going to be a few examples.

And two points about this: one, it is a revolution,
and I've heard this from some people who have

been particularly interested in nutrients
and how this kinda changes the game.

What this changes though is the density of

Now we have really time-dense data and we
can deliver that data in near real-time.

It's not that we couldn't measure nitrate
before or orthophosphate before so it isn't

that we're making new parameter measurements;
we're just making a lot more of them, and

so that time density is a really important
part of the application of these tools.

But having that now opens up all kinds of
new opportunities for water quality monitoring,

for load assessment, for source identification,
for event detection, getting a better sense

of what's happening within streams and rivers,
and also ultimately I think real time decisions

support is may be one of the best applications
of these tools.

So I'm going to go through a couple of examples,
a couple are ours within our group in California

and a few are examples from other USGS folks
across the country.

So here's the one that really opened our eyes

We piggybacked on a study a few years ago
in the San Joaquin River (in California) has

some issues with very low dissolved oxygen
in some of the lower reaches and that affects

fish migration and is related to nutrient
loading and algal productivity in the river.

We set up some instruments for a short time
and we also sat out on a dock for two days

and collected samples every two hours.

This is 2005.

It's in the summer, there was no rain, it's
very dry in the Central Valley of California

in the summer other than this irrigation runoff.

What you see is nitrate concentration from
discrete samples.

We sort of stared at this and we thought we
were just going to fit a line through here

and look like it was going up but ultimately
this is where our eyes were really opened

to the application.

When you look at the continuous nitrate data,
there actually is a lot of structure here.

It's not quite what we expected.

We anticipate given the high algal productivity
in the system that nitrate would decrease

throughout the day and then rebound at night
and there's a little bit of evidence for that

but then there is this trend with the spikes,
sometimes they are at night and sometimes

they are during the day.

Ultimately the story here is that there may
be some in stream biological activity involved

in generating this signal but there is also
lots of other factors, like irrigation runoff

is probably the biggest one that would cause
variability in the signal at these sorts of

time scales.

And it isn't trivial, right, I mean we're
talking 1.7 to 2.5 mg/L of nitrate almost

over very short periods of time.

You can imagine the challenges if you're only
collecting one data point a week or a month

in trying to describe these loads accurately.

A few years ago this really opened our eyes
to the variability in a freshwater system.

Since that point, the Survey, USGS has made
really significant investments into UV nitrate

monitoring and it certainly is not due to
that study, it's due to in a variety of locations

recognizing the potential of higher frequency

Much of this really emerged out of Iowa and
out of the Midwest where in 2008 USGS Iowa

Water Science Center had the first two continuous
nitrate sites and grew steadily to 2011 but

it was still only something like 15 sites
and I think Iowa made up the bulk of those


In the last 3 years, it's grown to, we have
probably close to 80 sites that are publically

available at the point and a few more that
will be publically available soon.

But, we're close to 100 sites and you can
see the distribution on a map here where much

of it is focused on the Mississippi River
Basin, a big cluster down the East Coast and

again in Florida and the network in California
in the Bay Delta and a few other sites.

It is currently operating in 24 states and
most of this is funded by cooperators which

is both really good in that there's obviously
a need for this kind of data and so cooperators

support this partnership and delivering that
data to them.

It's also a bit of a challenge, we don't have
as much stability in that network as we probably

would like.

Ultimately if funding changes some of these
sites will go away.

So it is a consideration as we grow national
networks and as we build on more capability

how we do that and how we support it.

So here's an example from the Mississippi
River; this is work that we've been doing

with the NAWQA program for a few years now
and we have a sensor in Baton Rouge; now there

are a number of sensors on the Mississippi
River Main Stem and also in small watersheds

and larger tributaries to the Mississippi

This one in particular shows some interesting

Part of the challenge to us from NAWQA was
to show that we can make these types of measurements

in big rivers and also show its usefulness.

So I'll point you towards this upper right
figure which shows field sensor nitrate versus

the lab discrete nitrate and I think that
this will convince you that we can get really

good sensor data and it correlates very well
with lab data.

And so if you look at this time series now,
this is back to November 2011 when we started

the data collection you can see discharge
in 2011 there were a few moderate peak flows

then we had very low flow during this drought
period in the late summer and fall of 2012.

2013, a pretty big pulse of water come out
of the system and flooding and then this year

it is buried a bit, there are a few pulses.

A couple things to note: obvioulsy I just
showed you this upper panel, so the discrete

data show this general pattern.

There area also some interesting things for
such a big river.

Timing of the nitrate pulses: typically many
people that work in the basin know occurring

in May and June and into July so we see the
concentration peak, very low nitrate concentrations

in September, back up to pretty high concentrations,
over 3 mg/L high.

So there certainly is variability, not always
timed with discharge, and within a week it

can change by 10% easily within a few weeks
you could have a pretty dramatic swing in

this case.

This is a pretty big example, over 3 to 0.5
mg/L over a period of a month.

Well and so obviously the goal here is to
get a better handle on loads which for a long

time the USGS has estimated loads using the
LOADEST model and other regression-based techniques

and those work very well, they are based on
a limited number of samples collected every

year but they rely on longer-term records
to relate discharge to concentrations.

But the charge from NAWQA was: can we do better?

Can we improve our load estimates with continuous

And so what you see here is the monthly nitrate
load calculated from the sensors.

And a couple things I want to point out: as
I mentioned monthly, the uncertainty on a

monthly load is very small when you have a
lot of data.

This is typically 1-2%.

Lowering the uncertainty is certainly a good
thing when you are making pretty expensive

management decisions based on load information.

It also helped to reduce the uncertainty of
that input into the Gulf Hypoxia model, that's

another plus.

The other thing I wanted to point out though
is if you look at this time period relative

to these gray shaded areas which are the 10th
and 90th percentiles of the monthly loads

from 1968-2010 and also the average which
is in red, 2011 and 2012 we are below the

average monthly load and in fact In the lowest
10% for the end of 2012 and then we transition

to this period where the loads were very high,
the highest 10% for June.

So we had the sensors up for only a short
period but we captured this very abrupt transition

I think some folks on the weather channel
dubbed this weather whiplash or something,

so you know, these transitions from one extreme
to the next was captured with this part of

the record and that sort of highlights something
models have difficulty with.

A regression based model is really a model
focused on average conditions and does very

well with that.

But it's hard to model extremes when we hardly
see them.

So if you look at the comparison of the sensor
model and the sensor estimates versus the

model estimates, here's some data for three
different models: LOADEST, the composite methods,

and WRTDS.

These are all regression-based models, don't
worry about details of these models if you're

not familiar with them but just note that
in general, as a percentage, this is a little

bit deceiving, so the percent difference each
month from the sensor load to the model estimate.

And this can be high in some months, you know,
over 100% for this particular approach in

the summer; those high percentages are related
to times when the mass loads are very small

so not all that meaningful to look at as a
percent; it's more meaningful to look at as

a mass and when you do that's pretty small.

But there are these periods, particularly
in the springs, where the models in this case

underestimated the loads in May and June and
July of 2012 and again underestimated the

loads in May and June and July of 2012.

Again, these are extreme periods and that's
really hard for a regression-based model to


But the good news is for much of this record,
the models do very very well.

And so I think the take-home point is when
we think about the historical record we think

about what we know based on the models so
far, the data is good and we've learned a

lot and that's been really valuable.

Moving forward we have some new technology
which will help us do it even better.

In particular, having sensor data like this
that for example in Baton Rouge will help

us to refine models and sites that don't have
instruments by understanding the factors that

give us these offsets in the model output.

So looking back, there certainly is a lot
of valuable information; looking forward we'll

be able to do even better.

So real time management, I think that's this
is one of the biggest selling points for the

instruments given that you can make measurements
in real time and deliver that data.

So there's really interesting examples that
comes out of the Iowa Water Science Center;

Jesse Garrett provided this to me.

And so in 2013 there was a unprecedented nitrate
levels, and how that was a challenge for drinking

water, and in particular the Des Moines water
works has a $4 million nitrate removal system

that costs an awful lot to operate every day.

So having two sensors, one on the Des Moine
River and one on the Raccoon River, allows

them to better manage how those two sources
of water are blended to deliver water that's

below the regulatory limit.

That can be a real challenge sometimes, I'm

In particular, that really high concentration
period; here's again a graph from Jesse that

shows nitrate from the Raccoon River and the
cutoff here is 10 mg/L and you can see that

the Raccoon River certainly is high in nutrient
concentrations every year but in particular

if you look at this 2013 this is a period
where you're hitting 20, 30 mg/L and so operating

the treatment systems based on this data,
or blending water if that's an option, certainly

benefits from having that in real time.

There are some interesting things too, beyond
just looking at loads and looking at concentrations.

You can see some real patterns, and here's
an example from Jennifer Graham in Kansas

where there's a study on Indian Creek, this
is a watershed that has wastewater discharge

and the treatment plant has been upgraded
and there's a study to find out how changes

in wastewater loading have affected nutrient
dynamics in the watershed.

First thing I want to show you for anyone
else who is thinking "wow, he only showed

me the Mississippi," sensor versus discrete
data, here's another example with a different

instrument than what's used in the Mississippi;
for a number of sites showing the sensor and

laboratory data and again the fit is very

There's typically a little bit of an offset
and that can be corrected or not.

One of the things that Jennifer and her colleagues
saw was that at some sites there's this really

clear diurnal variability and you can see
that here in the site in red at "college"

and it's pretty significant; on a given day
you can vary from 4-8 mg/L; obviously that's

a challenge if you're collecting one sample
every so often to characterize the nitrate


One sample would also entirely miss this pattern
and so you lose that information about what's

happening in the rivers and in the streams.

What's interesting here is that you do have
some sites that show a lot of diurnal variability

and in fact it was highest immediately downstream
of the wastewater treatment facility, lowest

at the upstream sties.

So there's some loading pattern here that
appears to be related to wastewater discharge

that affects aquatic productivity but it differs
as you move downstream or upstream above the


A lot of information here.

There are some really interesting things you
can do with that kind of data.

One of them is quantify the nutrient retention
in rivers; this is some work from Matt Cohen

and Jim Heffernan in Florida who really pioneered
this approach, there have been approaches

in the past using dissolved oxygen to look
at metabolism and how nutrients are taken

up in streams.

What Matt and Jim are doing is using nitrate
sensors to do that instead, and actually some

really interesting work with phosphate sensors
too which we're not going to highlight here

but essentially what you can do is look at
the daily drawdown and recovery and by calculating

in between those peaks, you can get at how
much nitrogen is taken up.

If you're looking at the size of those night-time
peaks relative to say an upstream source,

that then tells you how much is removed and
that's typically related to denitrification.

So if you want more details on how this is
done and whether it can be done in certain

types of systems, follow up with me or look
at this paper and others that have come since

but it's a really interesting approach and
once you really start to do that, you can

start to ask some important questions.

So here's some data from the Potomac River
at Little Falls; this Potomac River is a pretty

big river; this is over just a couple of days
and you can see the nitrate concentration

here do vary quite a bit and we're talking
10-20% today.

Discharge is not changing; it's the blue line;
specific conductance is not changing, it's

the black line, and so this can be attributed
to in-stream productivity.

What we're doing with this is part of the
NAWQA integrated watershed studies program

is trying to figure out what this tells us
about aquatic retention.

The reason why that's really important is
because we rely on models like the SPARROW

model, spatial regression models, that have
a term for aquatic decay; those models incorporate

what happens in the river in their load and
concentration models.

This is a way we can really start to tease
that out and refine that term and in a place

like the Potomac where that really helps us
is as we move back into the watershed and

try to quantify loading from groundwater.

Now we're accounting for that in-stream productivity
that otherwise is just incorporated into a

discrete measurement at a single point.

Both the loading and whatever happens once
it's loaded into the river.

We can start to calculate that and work our
back towards better estimates of groundwater


You can also start to think about what drives
that variability and here's an example from

Liberty Island in California, in the Bay Delta,
where over just a couple of days you can see

it's a tidal system and you can see discharge
changing here.

You see nitrate being drawn down, and these
are pretty low concentrations already, but

you can see it gets drawn down and recovers
and gets drawn down again and recovers and

what you see if you couple that with a fluorimeter
for something like phycocyanin, an indicator

of pigments from blue green algae, you can
see that that goes up at the same time.

While we may also be interested in drives
nitrate variability, this tells us something

about how nutrients support algal productivity
in a system.

When you start to couple these types of measurements,
you get a whole new picture of what's happening.

So that's diurnal variability, but then we
also deal with sites that have tidally driven


This is one from New York, Hog Island, you
can see that there's a clear signal of nitrate

that's related these tidal dynamics.

You see some signals of riverine impact versus
coastal waters coming in and there's a lot

of information here that's difficult to tease
out with an occasional discrete sample.

And to highlight another application from
the folks in New York, from Dick Cartwright,

USGS in the New York Water Science Center,
is that they also use this to compare the

sensor data to ecological conditions so you
can see that for Hog Island they've highlighted

daily mean nitrate in comparison to fair and
poor ecological conditions.

And again, you're talking about tens of thousands
of measurements a year versus what may have

traditionally been 12 or 18, and we get a
much clearer picture of when exceedances occur.

This is something you just can't get with
traditional approaches.

These don't have to be fixed sites and so
moving boat surveys are a possibility.

Here's some data from Amanda Booth and Eduardo
Patino in Florida with USGS where they put

the sensors on a boat and they pump water
up into the boat and they go down the Caloosahatchee

and they go down into coastal environments
and they are able to map the distribution

of nitrate concentrations, sort of this LaGrangian
approach where you make various water be captured

by spatial mapping.

Another really promising approach, really
interesting approach, doesn't require people

on a boat, it requires someone driving an
underwater vehicle.

And so this is from Ryan Jackson with the
USGS in Illinois, and he's using the ecomapper

and right now the ecomapper holds a typical
water-quality sonde, a YSI multi-parameter

sonde, in the nose here and you're able to
do profiles with depth and this is a really

nice visual of being able to go up and down
within a water column and make a variety of


Ryan has done this in a variety of locations
including Milwaukee River Estuary.

If you just had a fixed site of water moving
into the estuary, the picture of what's happening

as different rivers all converge and mix really
wouldn't be clear.

Through these sorts of approaches where he
is using a profile, so this is really, this

is obviously a really important direction
and nutrient sensors will surely be integrated

into these AEVs.

They have been in coastal environments.

So I want to catch up with some thoughts on
the future and on the near future and what

should we be expecting in the next 1-5 years?

And it's a really important question, I think,
as Yogi Berra said: if you don't know where

you're going, you might not get there.

Here are a few thoughts.

We are going to continue in the very near
future developing protocols and guidelines;

the USGS and other organizations realize how
essential this is to getting good data, particularly

for real-time networks.

So there are lots of good examples.

The Monitoring Council has NEMI and the Sensor
Workgroup and everyone is doing important

work to build protocols and guidelines.

The USGS really values this; it's critical
to how we run our sites across the country.

One of the things we're working on is a USGS
Techniques and Methods report on in situ fluorometers,

it's going to focus mainly on chloraphyll,
algal pigments but also FDOM, the fluorescent

organic matter, I think more people are making
the measurement because it's now part of some

traditionally used sensor platforms like YSI
and Exo.

I don't think everyone knows exactly the challenges
of correcting for interferences, and how to

calibrate those sensors, and what sorts of
variability we see in the field.

This is just a quick visual example of what
happens with a fluorometer; this is a lot

like a spectrophotometer than nitrate sensors
in that there is a light source that's emitted

into a sample that the fluorometers are different
in that rather than measuring how much light

is transmitted by a particular wavelength,
it's measuring how much light is absorbed

and reemitted a longer wavelengths as fluorescence.

But you can see just visually that excited
light in quinine sulfate fluorescing which

is pretty blue, when you add color to it,
it's much harder to see, and when you add

turbidity it's virtually impossible to see,
and your eye is much like the detector of

these instruments and so these sorts of challenges
need to be addressed.

I think lower cost, easier to use, more efficient
are going to be really important themes over

the next two or three years.

I think we are going to see more plug and
play integration for sensors, realize that

you take a variety of instruments from various
manufacturers and that you now have to figure

out how to log them and transmit them and
talk to them and that's a real challenge,

so I think we're going to see an easier to
use system of sensors in the next few years.

We're going to be doing more with our phones
and tablets, this is certainly an option for

us now, but it's not widely used and it certainly
needs some more development.

This is a direction we are clearly headed.

I think automating the QA/QC and collecting
additional metadata is going to help us work

with these very large datasets and that's
one of the ways we're going to be way more

efficient with how we manage instruments in
the field and how we manage data.

Also it's going to be really critical to getting
lower cost and easier to use sensors is partnerships.

And I wanted to highlight one of them which
is really exciting: the nutrient sensor challenge.

This is a collaboration mainly coming form
the White House Office of Science and Technology

Policy and EPA but in collaboration with the
Alliance for Coastal Technologies, funded

by NOAA, USDA, USGS, and lots of other partners.

There's a challenge to accelerate the development
and commercial availability of affordable,

reliable, and accurate in situ nutrient sensors.

There are incentives for testing and verification,
publicity, recognition, market access.

It's a really exciting effort and there's
a real vision here for how we can make these

instruments not only more accessible by lowering
up front costs, but also thinking through

applications and this is a really strong partnership
with manufacturers.

We had a meeting last week in DC and this
ended in a really positive direction.

And so there are ways you can be involved:
there's a website and you can email, but I'd

encourage anyone who is interested in nutrient
sensor technology, it initially focuses on

nitrate because we're a little bit closer
to refining that technology but it also certainly

focuses on phosphate and other parameters
that we're interested in.

We're going to certainly expand applications,
you know, what I showed you is surface water

but we are starting to make more continuous
nutrient measurements in groundwater.

It's an interesting question whether or not
having that higher density temporal data actually

tells you something about groundwater and
I think early indications are that there are

places where you can learn a lot by sampling
more frequently.

And as you see, this is probably one of the
most promising applications where you get

out of the stream and you get on the edge
of a field.

The Great Lakes Priority Watershed, the USGS
is a partner with a variety of other agencies

for some of that work and this is a place
where I think we are going to see the application

of these sensors to identify and inform best
management practices on the lakes.

It's just a really promising application that's
fraught with challenges for sure but once

we get it right, it's going to be really good

And so I also wanted to highlight another
challenge and it is related to Edge of Field,

and that's Tulane is offering a $1 million
prize for innovations that reduce nutrient

loading and Gulf Hypoxia.

Check out this website if that's something
that you might be interested in.

So next generation sensors: this is, next
generation is certainly within the next 2-5

year timeframe, and one of the things I'm
most excited about is LED technology fingerprinting;

that's allowing us to get into regions for
fluorescence in particular, that traditionally

have been really difficult.

If you look at this plot on the right, excitation
emission plots, think of it as a fingerprint

for organic matter, it's bench top measurement,
and what you see in the fingerprint is this

part down here that's really bright red.

It's a very strong fluorescence peak and it's
related to what we think of as protein rich


This is a wastewater sample through work with
Steve Corsi of the Wisconsin Water Science

Center and the technology now is such that
the LEDs are pretty stable down in this region.

Ultimately what this means is that we may
have instruments that are very good at detecting

wastewater because those peaks are typically,
and so the Low UV fluorometers are instruments

that are really promising.

They are instruments that are good at identifying
algal classes, sediment size, ammonia, and

even some instruments now that are certainly
in testing phase and are being potentially

used to identify DNA and microorganisms.

Some of these are further off than others,
but the next generation of tools is really


Finally, how do we build national networks?

How do we do this consistently, develop national
networks with real time data that helps us

meet monitoring needs, whether that's drinking
water, TMDL, edge of field loads, that also

helps us accelerate the pace of discovery,
have some long-term stability, funding is

always a challenge and we need to figure out
how to do more of this without budgets increasing

and so we have to figure out how to do this
more efficiently, from data collection to

decision support.

There are lots of questions on how to build
these sorts of networks regionally and nationally.

I think ultimately the charge I would have
for the group is to think about where time

dense data will change what you know or what
you do about water quality.

Ultimately we don't need to do this everywhere,
but there are places where this will revolutionize

what we know or how we manage water quality.

We need to identify those places, we need
to get sensors in at those places, and we

need to be really clear about how this changes
what we know or what we do.

So, thanks, that's all I wanted to present.

I'm happy to hear any questions or ideas from

Also feel free to follow up with me or I'd
be happy to pass you to those folks in the

different USGS offices whose data I showed
here; there's a lot of expertise within the

survey and other agencies and academia and
I'm happy to point you in certain directions.