PRMS Simulation of Depression Storage

Video Transcript
Download Video
Right-click and save to download

Detailed Description

Instructions for simulating depression storage when using the USGS Precipitation Runoff Modeling System (PRMS).

Details

Date Taken:

Length: 00:18:43

Location Taken: Lakewood, CO, US

Transcript

Roland Viger: Hello, my name is Roland Viger.

I'm with the Modeling of Watershed Systems
Project.

I'm going to talk to you a little bit about
simulating water storage in surface depressions

with PRMS.

This is an optional functionality within the
Watershed Model.

Certainly not required and clearly not all
locations have a lot of surface-water storage

that you need to worry about.

We decided we needed this kind of functionality
because we thought we were doing a very good

job of simulating hydrology in general on
a watershed.

But for some reason some of our flows were
showing up too early, but not all of our flows.

We realized, in fact, that our runoff from
the land surfaces was in some cases ending

up in the stream too quickly.

When looking at more spatial data and the
history of the area, we realized there was

a large number of farm ponds and mill ponds
and other kinds of structures on the land

surface that runoff would actually end up
in, and spend at least a little bit of time

in before ending up in the stream network.

That was our motivation.

Hopefully, if you're using this functionality
it's similar to what you're seeing.

Here is an extreme example, where clearly
you might have a ton of surface depressions.

Water getting across all of this into a stream
could take a long time, and really change

the overall hydrology of our watershed.

Our initial development watershed was nowhere
near as heavily pitted as this, but it still

made a big difference.

Terms of the overall importance of surface-water-holding
bodies in watershed...This is a very aggravated

view looking across the whole country.

If you look to the graph at the right in particular,
we see total area in the country associated

with water bodies of a certain size.

What you can see is that, even fairly early
on in the process, towards the left end of

the graph, after a category or two, you've
actually got quite a lot of landscape that's

covered by water bodies.

This just drives home that water bodies are
really everywhere, even if they're not on

the stream network.

They can add up and have an overall effect,
a visible effect on stream-flow response.

This might motivate you to think a little
bit more about whether you need to simulate

these explicitly in your watershed with PRMS.

Surface depression is a term that we sort
of -- I'm not going to say invented -- but

we have adopted.

There are certainly many synonyms.

You can see a few of them listed down in the
middle of the slide.

The idea is that these are depressions in
the surface that will hold up the water, and

do so with enough volume that your stream-flow-calibration
process should be...understand it that they

exist.

A couple of points at the bottom, there are...

Although you can have an HRU that is really
just embodying a lake or is dominated by a

lake -- which is described in other lectures
in this series -- these water bodies tend

to be so small that we really don't care to
represent them as discreet HRUs individually,

because we'd end up with hundreds or even
thousands of extra HRUs.

That's just a bit too much in the views of
the modeler.

Even though their explicit spatial representation
within the model doesn't exist, they're sort

of part of or characteristic of each HRU.

The model will in fact keep track of the acreage
of inundation and the volume in those things.

It will make sure that any precipitation on
to the HRU will also include distribution

of that precipitation into the surface depressions
and will do things like figure out when it's

going to spill over, or whether the water's
going to seep into the ground, and so on.

We have a full suite of physical processes
associated with these surface depressions,

even though they are sub-HRU resolution, which
is an HRU's sort of basic building block for

PRMS.

Here is a much more detailed view of the various
states associated with a surface depression.

These are all simulated for each time step.

You can see they correspond to some of the
things we showed on the last slide.

Don't need to retain this.

There is documentation we will talk to you
about in the end, which will let you look

at this graphic again and read more about
the terms involved.

Surface depressions, coming up with estimates
of parameters to describe all the things that

PRMS needs to know about them it takes a little
bit of work.

It's not an undue amount of work but we wanted
to go over how we have approached coming up

with this formation.

You can see we have really four broad steps
and the boundaries between, say, three and

four, they're a little fuzzy.

We'll go through what we're talking about
and hopefully it'll be clear.

There's a handful of parameters that we feel
relatively confident in setting, based on

spatial analysis in readily available GIS
data sets.

There are a number of other ones which relate
to flux rates and things like that.

Those are really more set heuristically based
on the expertise of the modelers.

This slide, we don't usually describe a method
for coming up with.

It's up to you to look at things like hydraulic
connectivity or how your soil's glacial till

and things like that, the bedrock and so on.

A very important question, obviously, is,
"Where do we get a map of these surface depressions?"

That's our starting point.

What we wanted to just drag you through was
a few slides sort of emphasizing how different

estimations of where the water bodies are,
and the quantity of those things are, depending

on which data source you use.

On this slide we've got some red speckles
in the inset, that you can see maybe a little

more clearly.

Those are sinks derived from a DEM.

Many of those are actually spurious artifacts
associated with re-projecting a DEM.

Then when you analyze it in their projected
coordinate system, you have a lot of local

spurious pits.

You've probably heard of filling the pits
in your DEM, filling the sinks, as almost

just a standard issue DEM massaging process.

That tends to get rid of these by design.

Even if you didn't do that, we find that the
resolution of most DEMs...Obviously this might

change with the advent of LIDAR, things like
that, or use of that in your modeling application.

We don't generally view the DEM as the best
place to get estimates of where water bodies

or surface depressions lie.

In the yellow we're seeing enhanced thematic
mapper data, which tends to give you a sense

of where water is resting on the surface and
that's looking a lot more realistic.

Obviously we still get the main stream path,
which is different than what we're after,

so we would probably throw that out if we
use that.

Here's yet another view.

We in the lower 48 or in the Unites States
are very fortunate that we have something

called the National Hydrography Dataset, which
is a vector-based representation of the nation's

water bodies and streams.

What we're showing here is, we've taken an
NHD high resolution, which is a very local-scale

data product, and extracted just what they've
designated as water bodies as opposed to rivers

or streams.

We're showing them on the map here.

For instance, at the bottom right you see
a very large water body, The Great Swamp.

You could argue whether that's a water body
or wetland or what.

We're going to zoom in on the upper area for
our sample watershed.

You can see a lot more resolution.

We can also see how we are comparing that
with, say, what we got off of the enhanced

thematic mapper in the green.

We tried to highlight where the two datasets
were similar in the purple.

You can definitely see there's a lot of red
stuff and there's plenty of green stuff.

The degree of agreement between those datasets
is kind of low.

It's pretty important to choose the right
data source for how you're getting your surface

water bodies.

It can vary quite a lot.

This is just another example of where we might
find surface depressions.

We'll move on from there.

One of the most basic parameters we use to
describe the presence of water bodies is just

how much of the HRU area is covered by these
things.

Obviously, the preceding slides, you could
imagine pretty directly how this parameter

would react to this choice of data depiction.

Once you go from this sort of pixelated view
that I've been showing you, we come up with

averages per each of our hydrologic response
units.

Really we're just showing here, with similar
color schemes across all three examples, the

HRU values for this parameter, depending on
which data source you use.

On the left we've got the ETM stuff.

In the middle we've got the NHD stuff.

And on the right we've got DEM stuff.

You really do get different values and it
pays to think about which one you would like

to use for your application.

A lot of the following parameters are really
based on how much of the runoff generated

in the HRU drains to the stream versus draining
to the surface depressions.

We tend to need to find the contributing areas
to the surface depressions within an HRU.

We have two major depictions of how runoff
gets routed.

It's surface runoff to depression storage
and surface runoff to depression storage impervious.

The first parameter is about the pervious
land surface and obviously the second one

is about the impervious stuff.

In both those cases we're using this watershed
analysis associated with each of these depressions

within the HRU.

Then we're going to come up with that number.

The next couple of slides, we're going to
illustrate to you how to find any of the contributing

areas.

Takes a little bit of filtering.

It's definitely a judgment call there.

By and large, a lot of the ways in which we
get contributing areas includes what we'll

call on-stream water bodies.

We tend to want to exclude those, because
all of the HRU will eventually leave through

the associated stream path.

If you maintain surface depressions that are
actually on-stream, you're going to find that

a hundred percent of your surface runoff,
no matter where you start from, ends up going

into a surface depression of one kind or another.

We think that's not quite accurate, not what
we want to say.

We're going to filter on that.

For example, on the far left we see -- if
we take our water bodies straight ahead and

we say, "Where's the contributing area?" -- you
can see there's almost no white in that map.

Almost every square foot of this whole basin
ends up running into some sort of water body.

What we did in the next, figure B, is we said,
"All right.

Well, you need to actually not be touching
the stream for your water body to be considered

for this watershed analysis."

We see essentially a drastic reduction.

We feel this is legitimate in that, if you've
got a water body that's sitting on a stream

and runoff hits you, you're probably a free-flowing
water body.

Any water that comes into you will probably
go out of you unfettered.

It's just a matter of distance or travel time
as to when it leaves, and not about impoundment.

Growing further from that idea, we said, "All
right.

Well, if you're a water body...We're going
to exclude at least the portion of your area

that is within a certain distance buffer,
in this case 300 meters from a stream."

"If your water body is half in what you might
call the riparian zone and half out, we're

going to exclude the half that's in, and then
find the contributing areas."

What we see is a relative reduction from B
to C in terms of the overall drainage area.

The far right one is the most extreme example
of what we just said.

Your water body footprint for each individual
water body must be fully beyond 300 meters

from the stream.

You can see how these differences could play
through and seriously impact the values you

would assign for SRO to depression storage,
impervious 

and not impervious.

As you noted, those SRO-to-depression-storage
parameters, they care about whether the surface

is impervious or pervious.

Talking about imperviousness a little bit,
we tend to get grids from sources like the

Nation Land Cover dataset, for various years.

It's telling us a floating-point number from
zero to one, about how impervious a given

pixel or cell is.

We want to...The parameters in PRMS require
that you define, "Is it impervious or not?"

In other words it's a binary differentiation.

We need to turn this floating-point number
into a binary designation.

What we did was, we looked at the impervious
maps.

This is an example in our sample watershed
in the Apalachicola-Chattahoochee-Flint.

We've got a little zoom in area at the north
end.

Then what we did is, we applied a minimum
floating-point-impervious value.

That way, it would allow to maintain an area
being called, "impervious."

You can see there's a pretty tremendous reduction
there.

The threshold we applied, it was actually
very, very low, was five percent or less,

if I remember correctly.

Ultimately, this is a judgement call for you
as the modeler making an application.

You need to pick that threshold and you can
play around and see how you get fallout in

the red areas -- the impervious areas -- depending
on what sort of threshold you set.

We wanted to get rid of areas that were really
not very restrictive, as far as infiltration.

We fiddled with our threshold until we saw
things like the road network emerge.

We figured roads are nice and impervious,
and as long as we're honoring those, we're

happy with what that looks like.

Another way to think about this is we actually
-- based on land cover -- said, "What's your

average, mean, min and max type of imperviousness?"

You can see for, for instance, the developed
categories, they range from over 90 percent

down to 8 percent or so.

As I mentioned earlier, a threshold of five
percent was a little bit lower, but we felt

like it still maintained images, if you will,
of certain features like roads and landscape.

You can see the non-developed areas, like
forests and so on, have a much, much lower

imperviousness.

We felt like we weren't artificially describing
things like forests as impervious with the

threshold we chose.

This is how we made our binary image an effective
impervious as opposed to what might be considered

just any sort of imperviousness.

Another view on this, if we did not do this
effective imperviousness filter, you can see

the before and after in terms of what the
average HRU value would be.

The effective imperviousness is generally
lower, which you would expect.

It's an important filter and it will impact
your model.

Generally speaking, PRMS is very sensitive
to imperviousness of any kind.

What we're showing here is a little synoptic
view of the change in performance of PRMS

for the sample watershed, the Apalachicola-Chattahoochee-Flint.

We've been showing you images of, so far,
where we have no surface depression on the

y-axis and surface depression on the x-axis.

We are seeing a change that's important and
actually shows up in the results.

The usual finishing slide, we have our MoWS
home page and this is where you should go

to for information.

If you have questions, you can follow those
links and email us through this web-form.

We encourage you to use this instead of trying
to email us directly, so we can decide who

the best person to answer your question is
and make sure we're all seeing the same questions

from you.

In terms of documentation, there's actually
a fair bit of it by now.

Some of the early work was done in collaboration
and almost unilaterally by an early user of

ours, Kevin Vining from North Dakota.

There's a couple of pubs that really describe
how he put things together and the motivation

for it.

We took that approach and began modernizing
it with the 2010 paper you see there.

That gives you probably a lot more background
and rationalization as far as the techniques

that you've seen in this particular presentation.

You should always be familiar with the PRMS
manual, which is there.

There have been some updates in terms of the
parameter names, which are reflected in this

presentation, but might not be in earlier
versions.

For instance, the 2010 paper might use slightly
old parameter names and things like that.

You should keep your eyes open for that 2015
publication.

It has been approved and is being released
through our website right now.

That's all we have.

Thanks very much.