State/Transition Simulation Models for Ecosystem Management

Video Transcript
Download Video
Right-click and save to download

Detailed Description

This webinar was conducted as a part of the Climate Change Science and Management Webinar Series, put on by the USGS National Climate Change and Wildlife Science Center and the FWS National Conservation Training Center. Webinar Summary: Sustainable management of natural resources under competing demands is challenging, particularly when facing novel and uncertain future climatic conditions. Meeting this challenge requires considering information about the effects of management, disturbance, land use and climate change on ecosystems. State-and-transition simulation models (STSMs) provide a flexible framework for integrating landscape processes and comparing alternative management scenarios, but incorporating climate change is an active area of research. In this presentation, three researchers present work funded by Climate Science Centers across the country to incorporate climate projections into STSMs. - The first case study integrates species distribution modeling with STSMs to project changes in whitebark pine in the Greater Yellowstone Ecosystem. This combination of correlative and stochastic models reproduced historical observations, identified important data gaps, and described potential future declines in whitebark pine. - The second study uses STSMs to address conservation and management of the longleaf pine ecosystem in the southeastern US under climate change and urbanization pressures. Results show that urbanization is likely to be a bigger threat to the future of the ecosystem than climate change. - The third study integrates multiple models to project future rangeland condition and habitat for Greater sage-grouse in eastern Oregon under varying climate and management scenarios. Projections indicate that rangeland condition and habitat are likely to decline due to current stressors, but climate change may have both positive and negative impacts. The three studies highlight the utility of STSMs for natural resource management in disparate ecosystems across the U.S.

Details

Image Dimensions: 480 x 360

Date Taken:

Length: 01:07:33

Location Taken: Reston, VA, US

Transcript

Ashley Isham:  Good afternoon or good morning
from the U.S. Fish and Wildlife Service's

National Conservation Training Center in Shepherdstown,
West Virginia.

My name is Ashley Fortune Isham, and I am
very happy to welcome you all to our webinar

series that's held in partnership with the
USGS National Climate Change and Wildlife

Science Center that's located in Reston, Virginia.

The NCCWSC Climate Change Science and Management
Webinar series highlights their sponsored

science projects related to climate change
impacts and adaptations and aims to increase

awareness and inform participants like you
about the potential and predicted climate

change impacts on the fish and wildlife.

I would like to introduce Dr. Shawn Carter,
senior scientist at the NCCWSC.

Welcome, Shawn.

Dr Shawn Carter:  Thank you.

It's great to be here.

I'm delighted to introduce our next three
speakers.

Megan Creutzburg is a researcher at the Institute
for Natural Resources in Portland, Oregon,

which is affiliated with Portland State University
in Portland and in Oregon State University.

She is interested in landscape scale conservation
and management.

Her current research uses simulation models
to understand the impacts of disturbances,

climate change, and management.

She works in varied ecosystems throughout
the Western US from coastal forests to subalpine

forests and rangelands.

She received her PhD from Utah State University
in 2010.

Jennifer Costanza is a landscape ecologist
with research interests in ecological effects

of global change, land change modeling, and
landscape conservation.

She received her PhD in ecology from University
of North Carolina at Chapel Hill.

She's currently a research assistant professor
in the Department of Forestry and Environmental

Resources at North Carolina State University.

Much of her research involves investigating
future forest dynamics across the US and responds

to changes in climate disturbances, land use,
and land management.

Last, but not least, Brian Miller is a research
scientist at Natural Resource Ecology Lab

at Colorado State University, who's also a
staff member at the Department of Interior's

North Central Climate Science Center.

He's one of our own.

He's working with NCCWSC staff and partners
to develop stimulation and scenario planning

tools that can help resource managers deal
with climate change and is also helping to

manage capacity building projects in the North
Central domain.

Brian earned his degree in ecology at the
University of North Carolina at Chapel Hill

as well.

He's also worked at the Carolina Population
Center.

It's my pleasure to introduce these three
speakers.

I believe we're going to be starting off with
Brian, who I'm going to turn it over to now.

Welcome.

Brian Miller:  Thanks very much, Shawn, for
the introduction, and also thank you to Holly

and Ashley for helping to organize today's
webinar.

Of course thanks to everybody else on the
line listening in.

We're going to break things up today into
three different case studies.

But before we jump into the details there
what I'm going to do is give everybody an

overview of stateandtransition modeling, and
also how climate change is now being brought

into those models.

That's to get everybody on the same page,
and then we'll get into those three different

case studies, the first being a case study
on whitebark pine in the Greater Yellowstone

Ecosystem that I'll be talking about.

Then we'll hear from Jen Costanza about longleaf
pine, and finally from Megan Creutzburg about

rangeland condition and greater sagegrouse.

Each of these case studies brings something
a little bit different to the table in terms

of integrating climate change into stateandtransition
simulations along with other dynamics like

urbanization, and fire, and things like that.

To give you an idea of where we'll be travelling
today, at least virtually, I'll be starting

off here in the Greater Yellowstone Ecosystem,
right at the intersection of Idaho, Montana,

and Wyoming.

Then we'll hear from Jen in the Southeastern
United States, and then Megan back over on

the West Coast in Southeastern Oregon.

What are stateandtransition models?

Initially, and to some extent continuing through
today, they're used as conceptual models,

so ways of organizing our thinking about vegetation
and how it can change through time and space.

This is oftentimes represented in this box
scenario diagram kind of approach where the

boxes represent the states and the arrows
the transitions.

What do we mean when we say statesandtransitions?

One thing I'd point out is this does vary
quite a bit, especially with the term states.

It does depend on your objectives, what you're
trying to model and understand, and also what

data you have available and what level of
detail those data are.

Essentially they're groups of vegetation communities.

They're oftentimes defined by a dominant cover
type, in my case there might be whitebark

pine forest, a structural state, often with
forests that might be open versus closed,

or somewhere inbetween, and then an age.

This could be a numerical kind of an age or
maybe it's relative to, say, reproductive

status, or something like that.

Transitions are really any kind of process
that has a bearing on that vegetation and

that you're interested in.

Here I've given a couple examples of some
natural and anthropogenic types of transitions,

although clearly there's some overlap there
and those categories are not necessarily mutually

exclusive or so distinct.

This gives you a little flavor of the types
of transitions you can include in these models.

You'll be seeing a variety of these through
the case studies.

Again, I've put this definition of stateandtransition
models up here.

I'm going to leave it there for a moment while
I define one other part of the title of the

slide here.

I've talked about stateandtransition models,
but what do I mean when we insert the word

simulation?

There's a variety of different kinds of simulations,
but really they're any kind of computerbased

representation of the real world.

I think "SimCity" is one that some people
might be familiar with.

It gives you an idea of what we mean.

The benefit of simulations I would say is
twofold.

For one thing, they can really represent complex
interactions that we see in the real world.

Nonlinearities, and thresholds, secondary
effects, or what you might think of as feedbacks,

and also emergent patterns, these are more
complicated looking patterns that emerge from

very simple rules.

For instance, things like bird flocking or
fish schooling behavior.

Not only can we recreate these dynamics, which
is neat and it helps us understand how systems

work, but we can then use those simulations
as a virtual laboratory.

We can ask "what if" questions that we might
not be able to ask in the real world.

"What if we increase the area burned each
year through time?" or something like that.

They're really helpful for digging into managementrelevant
questions.

Together stateandtransition simulation models,
that whole term, essentially we're taking

this box scenario conceptual model and making
it temporally dynamic and in many cases, although

not always, spatially explicit.

As a result we're able to represent biotic
interactions, disturbances, and management

scenarios within this framework.

To give you a more concrete idea of what I
mean I'm going to briefly talk about a simple

example model before we dive into the more
complicated case studies.

Here's a threestate example, an open forest
class, a closed forest class, and finally

maybe there's invasive species that can invade
this forest.

We have some different transitions.

There's succession, fire, invasion, restoration,
giving you a flavor of the states and transitions

you might be interested in.

Then you might imagine a gridded landscape.

Those that are familiar with Raster images
or satellite images, you could think of each

cell on the landscape existing in one of these
three states at any given time.

Then that cell can transition to one of the
other states depending on how you set up your

transitions and whether those are based on
an age, or a probability, or something along

those lines.

To give you a little visual here's an example
image, the three colors representing three

different states.

Then you could imagine this changing over
time.

For instance, the purple being an invasive
species, you might then see that move across

the landscape and have some kind of management
intervention that limits that spread.

That's a basic example, and one that I pulled
from the stateandtransition simulation modeling

software, STSM, that I believe all of us use.

To summarize, stateandtransition simulations
can integrate these different dynamics really

effectively.

They've been shown to be able to do that.

Today I think one of the bigger challenges
that is really being worked on currently is

how to bring climate into this picture.

Through these different case studies we're
going to give you some examples of how that

can be done, so we're really going to hone
in on this connection here.

Now, to bring you back to these more specific
case studies, here's our map again.

I'm going to start off with whitebark pine
in the Greater Yellowstone Ecosystem.

Before I dive in I did want to mention some
collaborators on this project who are listed

here at the bottom of the slide.

Leonardo Frid, Tony Chang, Nate McKelick,
Andy Hansen, and Jeff Morisette have been

helping with this work.

Whitebark pine is a native conifer of the
Western United States, primarily in the subalpine.

It's considered a keystone species because
it provides food for various wildlife, Clark's

nutcracker, red squirrels, and grizzly bears.

It also serves some ecosystem functions, as
stabilizing soil, and moderating snowmelt

and runoff, and also facilitating the establishment
of other species.

But has been effected pretty dramatically
in recent years, primarily by two disturbance

agents, one being a nonnative fungal pathogen,
called white pine blister rust, and the other

being the more infamous mountain pine needle
that will affect trees like lodgepole pine

and ponderosa pine.

We've seen some dramatic declines in whitebark
over the last of couple of decades.

Now the question is, how might climate further
affect whitebark and how might it affect it

indirectly through these different disturbance
agents.

Again, returning to this idea of connecting
climate and state and transition simulations,

the approach that we've taken is to bridge
these two using species distribution models.

Now I'm going to give you just a bit of background
on species distribution models and then get

back to the specific example.

Species distribution models essentially correlate
a set of abiotic variables, so climate variables,

soil, what have you, with species occurrence.

This can be species at presence, or species
presence and absence.

It's a bit of a misnomer because it's not
necessarily telling you where you will find

a species in the future.

It doesn't project species distributions,
but rather it models projected suitable areas

based on those abiotic variables.

In our case, our abiotic variables included
predictor data from the PRISM data set, as

well as some water balance modeling.

These are summarized monthly using means from
1950 to 1980.

Through this collection of predictors, we
then had a large number, 122 predictor covariants.

We needed to narrow that down in order to
reduce colinearity.

We chose among pairs of variables with the
highest level of correlation to reduce that

colinearity.

We selected variables that explained a high
amount of deviance in presence and absence,

and also were ecologically meaningful to whitebark
ecology.

We ended up, through this process, with a
total of eight predictor variables as well

as our set of presences and absences.

We could relate these two data sets, you might
say, through a random forest modeling algorithm.

Then we ran the model for a historical time
period and also projected it into the future.

For the projections, we relied on primarily
two global circulation models as well as two

representative concentration pathways.

I should add that Tony Chang enhanced this
work.

They've explored a variety of other RCPs and
GCMs, but for this work, we just chose to

select the bookends on the most and least
impact on whitebark pine habitat suitability.

The output of these models is really the thing
that I think I'd like you to focus on here.

The output of the models looks something like
this.

It's a grided image here where each cell has
a probability of presence of whitebark pine

based on those abiotic variables.

In this case, red being a higher probability
of presence, blue being the lowest.

Just keep this image in mind, it'll come back
in about two slides or so.

When we then turn to the state and transition
simulation, if we just look at the core of

it, the most basic portion of this model,
you can think of three states.

There's an open state, which could be a shrubherb
state or maybe an alpine area, where there

is no whitebark yet, and seeds can then disperse
to that location.

But we then need to determine whether or not
a seed could feasibly establish at that location

on the landscape.

Oftentimes you'll do this through probabilities,
but in our case what we did was we brought

in that species distribution model output
to dictate those probabilities through space.

Then we could also allow those probabilities
to change through time with the change in

climate.

In essence, the species distribution model
output serves as a spatial multiplier effect

for whether or not seedlings are able to establish.

I don't want to bore you too much with this
figure, but I did want to mention briefly

that those three boxes and arrows is just
one portion of the broader model that includes

other dynamics, including aging, fire, mountain
pine beetle infection, blister rust, and competition

with other species.

One thing that was important for us to do,
as I said, is to run the model historically,

and then compare the model specifications
to observations.

We pulled some data from the aerial detection
survey about whitebark pine mortality from

mountain pine beetle and that's represented
by the black solid line here.

Then the dotdashed lines are different model
specifications and the grey line there at

the bottom is a no climate change scenario.

You can see that we're not only ending up
at a location where the specification's really

closely matching observations, but we're also
reproducing that temporal signature, as well.

We felt fairly confident, at least for this
way of demonstrating this integration of tools,

to move forward and look at projections, as
well.

I'm going to show two quick results here.

One, if we were to look just at whitebark
pine beetle killed stands through time you

might say, "Oh, well, this is kind of encouraging.

It looks like there's a peak in beetle kill
maybe earlyish in the 21st century that tails

off."

This tails off across the different colors,
which represent different climate change scenarios.

But if we look more carefully at whitebark
on the landscape, we're seeing a fairly consistent

decline through time.

By the time we get to 2100 late zero stage,
whitebark is more mature whitebark, pine stands

are nearly absent from the landscape in all
these climate projections.

I should add this is a bit of a way, again,
of demonstrating the method, but it's a work

in progress.

We have a paper forthcoming describing this
work, but we're continuing to explore other

disturbance agents, and also management scenarios.

Again, our approach was to use species distribution
models to bridge climate science and stateandtransition

simulations.

Not only is this only one way of doing so,
but we see this as being part of a broader

suite of tools that we can use to really make
the simulation modeling relevant to managers.

It's not just modeling for modeling's sake,
but maybe we can bring in scenario planning

and social science to really tune these things
in to meet management needs.

With that, I'm going to go ahead and pass
it over to Jen Costanza and let her talk about

her case study.

Jennie Costanza:  OK.

Thanks, Brian.

I'll be talking about longleaf pine in the
Southeast US.

I'm going to focus on some work we've done
recently to explore some scenarios of urbanization

and climate change and fire management.

Here's our study area, which is at the intersection
of Florida, Alabama, and Georgia.

Again, this work was done with funding from
the Southeast Climate Science Center.

Just a little background before I get into
our modeling project on the longleaf pine

ecosystem.

Longleaf pine woodlands or savannas look like
this, this picture on the left fairly open

canopy with a thick understory of grasses
and herbaceous species and a fairly open shrub

layer, virtually no shrubs present, at least
under historical conditions.

Under presettlement these woodlands and savannahs
would have been burned as frequently as every

one to three years to keep this open structure.

This open structure supports not only a high
diversity of plants and understory, and some

of the highest diversity of understory plants
anywhere in the world.

It also supports a variety of wildlife including
the endangered redcockaded woodpecker.

Longleaf pine would have existed across this
range, the range map you see here.

This is the range of longleaf pine trees.

This longleaf pine ecosystem would have existed
in vast expanses across this range historically.

Now some estimates put the range of the longleaf
pine ecosystem at about three percent of what

it had been.

It's highly fragmented and severely reduced
in range, and the places that haven't been

converted to agriculture and to other human
land uses look like this picture, on the bottom

here.

They're pretty closed in.

The understory may be full of shrubs, the
rich herbaceous species may be gone, and the

canopy may be more closed.

That's mostly due to fire suppression, which
has been fairly predominate over the last

century or more in this ecosystem.

In efforts to try to think about conservation
and restoration of the ecosystem, this report

came out in 2009.

There's been a working group associated with
this ongoing and some efforts ongoing.

This report, this "RangeWide Conservation
Plan for Longleaf Pine" set forth a couple

of major goals.

One was to increase the total area of the
longleaf pine ecosystem by 135 percent.

The other one is to double the area of this
open canopy, firemaintained longleaf state.

This report also set forth two methods to
do that.

One was to focus on maintaining open stands
by keeping fire into these open stands.

The second method was to then, secondarily,
improve stands that had been degraded or restore

stands that had been converted to agriculture
or maybe loblolly pine plantations, which

are big here.

Our research questions for this work really
focused on those goals, and that report, and

looked at, over the next century, whether
those goals for longleaf pine could be met

or would be met under scenarios of changing
climate, urbanization, and then given potential

feasible management scenarios in the future.

Again, we used the same transition simulation
model approach.

We built on the wealth of models that have
been developed under the LANDFIRE program

a few years back.

If you all aren't familiar with them and are
interested in doing this type of modeling

I highly suggest checking these out.

There's one for every ecological system in
the US.

They do focus on preEuropean settlement dynamics.

We use the model for longleaf pine, and then
we changed it to reflect current and then

future conditions.

Here's the basic structure of the model.

Brian did a great job of outlining the overall
setup of these stateandtransition models.

For longleaf pine we have two successional
pathways.

A stand starts out in early succession, and
that represents 0 to 14 years of age.

The normal successional pathway brings that
stand to an open canopy state in midsuccession

for 15 to 74 years, and then to late succession
open after 75 years.

Then there's an alternative succession pathway
that brings the stand to similar closed late

and midstates.

I'll add in the probabilistic transitions
under fire.

There are several kinds of fire that can occur
in this system and under this model.

A replacement fire would bring any of these
states back to early succession.

A mixed fire would bring a closed stand to
an open condition, and surface fires act to

maintain that open condition.

There are probabilities associated with these.

They're annual probabilities.

There are also time steps inbetween them.

There are constraints where these fires may
or may not be experienced in any given year.

We, again, updated this model to reflect current
conditions by gathering data on recent wildfires

in the region where we were working.

You can see here, here's the updated wildfire
probabilities.

I mentioned earlier that an open canopy state
under presettlement may have experienced fire

every one to three years.

Here we're seeing, according to the recent
wildfire data, a 0.02 annual probability,

which translates to a fire every 50 years.

Severe fire suppression, as I mentioned, is
reflected in the data that we were able to

gather on this.

The second update we did was to adjust all
these probabilities for a climate change,

so all the fire probabilities.

Here we're focusing on climate change's effect
on fire.

There have been a lot of studies recently
done suggesting that the range of longleaf

pine might not be affected, at least in our
study area, by climate change.

It actually might do better under climate
change because it's resistant to drought,

the species itself.

The effect of climate on fire was what we
were interested in examining.

We did that by gathering historic climate
records, and then using that same wildfire

record for the area.

We related recent area burned per month to
some climate variables.

They're both temperature and precipitation
related variables.

They also looked at different lag times.

How hot is it and how hot has it been in the
recent past, and then how wet is it and how

wet has it been in the recent past.

We did that on a monthly basis and got a time
series regression relationship.

That was based on recent data, and then we
used that same relationship to project, then,

how much area would be burned under future
climate.

We did that using downscaled climate projections
done by Katherine Hayhoe for a number of GCMs

under two emission scenarios, A1FI, which
is the high emission scenario, and the B1,

which is the lower emissions scenario.

Then we translated that into what we call
a fire multiplier, which is the ratio of recent

fire, or area burned, to future area burned.

You see that relationship for the A1FI here.

This is the distribution.

Since we had multiple GCMs we did a probabilistic
distribution into the future.

The middle line here is the mean, and then
you see the min and the max for every year.

We were able to sample across that distribution
to get a series of multipliers.

We did that 50 times so we could run our simulation
model 50 Monte Carlo simulations.

You see, I do want to point out that even
this high emission scenario, the projection

is close to one, meaning no change in fire
and slight increase, maybe, over time in fire

under that scenario.

The third thing we did is to add urbanization
to this stateandtransition model.

We had the possibility of any of these state
classes transitioning to an urban class.

We did that based on some recent projections
of urbanization we did for the Southeast.

Here you see the orange is the current urban
area, urban footprint, and the red is the

projected future urban footprint.

This is based on the cellular autonomy model
called Sleuth.

Brian mentioned the spatial multiplier.

We also incorporated this time series of future
projections as spatial multipliers in our

modeling.

For our study area it worked out to be about
a 0.0018 annual probability of urbanization.

Once we had the urban and the climate change
added we needed to look at some management

scenarios.

Based on that rangewide conservation plan
for longleaf pine we identified some of the

most likely management actions in the ecosystem.

The focus in that report was on maintaining
open longleaf.

That was one of the management actions we
added to the stateandtransition model.

That just acted to keep mid and late open
state classes in their current states.

We also added an option to improve closed
stands to open.

This is like the mixed wildfire, but it also
takes into account a series of management

type burns and thinning actions done.

That's the state of the art in terms of the
latest thinking on how to transition closed

longleaf to open longleaf.

Then there was the option to restore longleaf
from planted pine or agriculture.

That, again, represented a series of management
actions including cutting planted pines, and

then planting longleaf, et cetera.

This is our modified stateandtransition simulation
model.

We ran this for 14 scenarios.

We wanted to, again, look at the influence
of urbanization, climate, and management.

At the top, the top two boxes here represent
scenarios that incorporated climate change

only, but no management, and with and without
urbanization.

Then we added four management regimes, and
we used the A1FI climate scenario in those.

Again, we have without urban and with the
urbanization.

Our management scenarios, again, incorporated
those three management actions that I showed

on the previous slide.

Two of our scenarios incorporated maintenance
burning only, so maintaining open longleaf

only at a level similar to current levels
in the landscape based on some management

records that we had for the landscape.

Then doubling the management level.

Still maintaining open longleaf, but doubling
that amount that gets managed every year.

Then a portfolio of management actions, which
included all three of those actions that I

discussed previously, but with a level that
reflects current levels, and then twice current

levels.

That was our 14 scenarios.

We mapped our whole ecoregion that we were
working in.

We mapped the longleaf, agriculture, planted
pine, and urban areas for the initial landscape

and we labeled all the longleaf with one of
those five state classes.

Then we ran that simulation model on an annual
time step through 2100 for all 14 scenarios.

Here's a look at the kinds of results we got.

This is those five longleaf state classes
over time.

The blues are the open canopy and the reds
are the closed canopy.

You see, if you look at this top line and
the first two scenarios here, these first

six scenarios include climate without management.

You see that essentially you get a buildup
of these red classes, these closed canopy

classes.

That's because even with climate change fire
suppression is still so high in this landscape

that you get a buildup of these classes.

You also see that there was no substantial
difference in the results among the different

climate scenarios.

Without climate, with the B1 and the A1FI,
you still get this buildup of the closed canopy.

Any scenario with management, any of the other
scenarios, did better than the ones without

management in terms of the amount of open
canopy that resulted towards the end of the

simulation period.

The scenarios that included a portfolio of
management, that included the restoration

of nonlongleaf to longleaf, increased the
total area of longleaf.

Here, briefly, is a different way to look
at the results, relating them directly to

that rangewide plan for conservation.

This is plotting the end results, so the result
in 2100.

This is the change in total area of longleaf
along the Xaxis and the change in area of

open longleaf along the Yaxis for all 14 of
those scenarios.

Each scenario is represented by a dot or a
triangle.

The triangles are with urbanization and the
dot is without urbanization.

Here you see the scenario named along the
right, corresponding to the colors.

The first thing to notice is that all of the
climate change scenarios, again, don't show

any significant or substantial difference
from one another.

They're stacked up here.

They do worse, they have a decrease in the
area of open longleaf, and they don't do any

better in terms of the total area of longleaf.

All of the management scenarios did better.

Then I put the conservation goal of doubling
the area of open longleaf here on this slide.

You can see that the management portfolio,
when the management effort is doubled, the

scenarios with the management portfolio far
exceed that conservation goal.

One of these other maintenanceonly results
does as well, but none of the scenarios meet

that second conservation goal.

That conservation goal would be out here,
the increasing the total area of longleaf

by 135 percent.

None of them came close to that.

In sum we're seeing that climate scenarios
aren't really affecting the results.

Climate change didn't have that substantial
of an effect on our fire regime here, but

management scenario and urbanization do affect
the results.

No scenario we examined was able to meet both
of those conservation goals.

With that, I'll pass this over to Megan who
is going to bring us back west and talk about

sagegrouse habitat.

Megan Creutzburg:  Thanks, Jen.

I'm Megan Creutzburg.

I'm going to be presenting the third case
study on rangeland condition and sagegrouse

habitat.

This work was funded by the Northwest Climate
Science Center.

First, to give a little context on the rangelands
in Eastern Oregon, for those of you that aren't

familiar with this ecosystem, the region is
dominated by sagebrush steppe.

There are various management challenges that
we're facing in this area.

In the warmer, drier region exotic grasses,
such as cheatgrass, have invaded large areas

and have promoted this frequent fire cycle
causing a shift from native perennial shrub

steppes to annual grasslands.

In the cooler, more moist areas there's a
very different problem where western juniper,

which is actually native to Eastern Oregon,
is expanding into areas historically occupied

by shrub steppe and changing it from shrub
steppe to wetland.

Then, of course we also know that climate
change is already and will cause increasing

changes in temperature and precipitation patterns.

Looming over all of this in the sagebrush
steppe throughout the West is the greater

sagegrouse, which is the sagebrush obligate
species that's been declining for several

decades throughout its range and actually
will be considered for listing under the endangered

species act in just a few months here.

The objectives for this work are to model
vegetation change into the future under varying

scenarios of climate and management, and then
interpret implications for both vegetation

condition and sagegrouse habitat.

To address these questions we integrated three
types of models.

First my colleague Emily Henderson built a
sagegrouse species distribution model, shown

in the lower left, which basically defined
a rule set relating environmental variables

to sagegrouse presence in the region.

We integrated this rule set into the MC2 dynamic
global vegetation model to track vegetation

that is climatically unsuitable for sagegrouse.

The MC2 model projects broadscale changes
in the distribution of plant functional types,

carbon, nutrient pools and fluxes, and also
wildfire.

It can project future conditions under scenarios
of climate change.

That's a mechanistic model.

From MC2, and also integrating this species
distribution model rule set, we incorporated

changes in the distribution of these plant
functional types and wildfire into a set of

stateandtransition simulation models, which
have been introduced in the last two talks.

Again, these model succession, disturbances,
and management at finer scales than the MC2

models.

They're more applicable for management questions.

Incorporating information from MC2 allowed
areas to shift from one vegetation type to

another with climate change, and also allowed
us to incorporate trends in wildfire with

climate change.

Then we ran the stateandtransition models
and summarized outcome in terms or rangeland

condition and sagegrouse habitat.

We ran these models under several climate
and management scenarios, which I will outline

next.

For our climate change scenarios we used three
global circulation models.

We purposely chose three very different models
with projections that spanned the range of

what we expect in the region.

First, the HadGEM model, which is the hottest
and driest model in the group, the NorESM

model, which is hot with wetter conditions,
and then the MRI model is warm, not as hot

as the other two.

All of these three models project higher temperatures
in all seasons, quite a bit higher in some

of the hotter scenarios, and similar or higher
annual precipitation so, in general, hotter

and wetter, but varying in the degree and
also in the seasonality of that change in

precipitation.

For our management scenarios, we ran a baseline
nomanagement scenario, assuming no restoration

activities, and then a current management
scenario, where we worked with various agencies

in the region to determine what treatments
are currently happening, the types of treatments,

and also what rates, for each ownership type,
and apply those into the future.

Then a restoration management scenario, which
was an alternative scenario where we increased

the rates of treatment and, in some cases,
changed the types of treatments to try to

better restore sagegrouse habitat.

I'm going to delve into some results here.

First, we're going to look at potential vegetation
types.

These are those broad vegetation categories.

As you can see from the legend here, we have
just a couple of types of forested vegetation

types, shown in gray moist and dry forests.

Then we're primarily interested in these shrubsteppe
vegetation types in the black lines.

The solid black line is moist shrub steppe,
the dashed line dry shrub steppe, and then

the dotted line is xeric shrub steppe.

The xeric shrub steppe is the vegetation type
that we got out of that species distribution

model that I mentioned, and it tracks areas
that are hot and dry in the summer that, according

to this model, were climatically unsuitable
for sagegrouse.

These two, the moist and dry shrub steppe,
are suitable for sagegrouse, and then, of

course, the forested types are not habitat
for sagegrouse, and then the xeric shrub steppe

is also too hot and dry in the summer for
sagegrouse.

Those are considered unsuitable vegetation
types.

I'm going to show four panels here of the
different climate scenarios.

All these projections for the first few slides
are under no management, just a baseline scenario.

With current climate, there is no trend shown,
because we actually just assume that vegetation

types stay constant into the future.

We're showing out to the end of the century
on the Xaxis and then the area on the Yaxis.

Under each of our climate change scenarios,
here's the hottest and driest scenario, and

the warm scenario, and the hot and wet scenario.

You can see that vegetation types are projected
to change quite a bit in the next century.

There's quite a bit of variability, which
you would expect because the climate models

were quite different in their projections,
but there are also some commonalities.

One commonality is that moist shrub steppe,
in the solid black line, is increasing in

all of those scenarios and dry shrub steppe
is declining.

This is largely due to the increase in precipitation
that we're seeing in the climate change projections.

We also see that the forested types, in grey,
are remaining relatively constant over time.

Then there's a fair amount of variability
in the xeric shrub steppe, in the dotted lines.

Each climate model is pretty different in
terms of its summer conditions, how hot and

dry those summer conditions get, so quite
a bit of variability there.

Next, I'm showing vegetation composition.

This is the more detailed vegetation composition
that we get from the stateandtransition models

across all of our vegetation types.

We have in green the native shrub steppe,
which is of course the most desirable state,

and then in red exotic shrub steppe.

There's two different flavors of that.

The solid red line is areas that dominated
by exotic shrub steppe or exotic grasses.

Then the dashed red line is semidegraded areas
where there's still some remnant perennial

native population, but also some invasion,
so it's at risk for transitioning to exoticdominated.

In the yellow there's seeded shrub steppe,
which is pretty minor in the landscape, and

then in blue we see juniper woodlands.

The woodlands in the blue solid line, and
then threshold woodlands, which are areas

which are scattered juniper that are at risk
of transitioning to full woodland, in the

dashed blue lines, and then forest in the
grey.

Under current climate we can see that, first
of all, the initial landscape starts out with

a lot of semidegraded shrub steppe.

That's really a state at risk of transitioning
to exoticdominated.

That's what we see where that solid red line
increases.

Native shrub steppe in green declines, and
then juniper woodlands continue to expand,

which is what we've been seeing across a lot
of the landscape in recent decades.

Again, this exotic shrub steppe and woodland
forest are unsuitable for sagegrouse.

Under our three climatechange scenarios, under
the hottest, driest scenario, here's what

we see with the warm scenario and the hotandwet
scenario.

In the first few decades of the simulations,
things look fairly similar across all the

climate scenarios, and then later in the century,
they really diverge.

Actually, under the climatechange scenarios,
we see less exotic grass invasion, and this

is because, related to the last slide I showed,
the areas are transitioning from dry shrub

steppe to moist shrub steppe, which is more
resilient to exotic grasses.

These areas are actually able to recover and
resist invasion better.

We see, also, better outcomes in terms of
native shrub steppe, where it's maintaining

or increasing over time, depending on the
climate scenario.

Then, interestingly, in terms of juniper woodlands,
in the blue, under the climatechange projections,

since we're seeing more precipitation and
a shift from dry to moist shrub steppe, we

would expect to see more juniper woodlands.

There's more juniperwoodland potential, but
we're also seeing an increase in wildfire.

Fire is keeping those woodlands at bay, so
they actually are at lower levels, across

these climatechange scenarios, compared to
the current climate scenario.

Speaking of fire, this is just showing, real
quick, under current climate on the left bar,

what we're seeing early, mid, and late century,
in terms of fire, a little bit of an increase,

due to exotic grasses, but not much.

Then, under the three climatechange scenarios,
to the right of that, we see a two to fourfold

increase in fire, so quite a bit more fire
at the end of the century under climate change.

Again, that's related to the projections we
saw with the juniper, where we weren't seeing

as much juniper under climate change because
of all the fire.

Now I'll look a little bit at management scenarios.

This is showing, again, what you've seen with
exotic grass, but with the initial conditions

on the left, and then endofcentury conditions
on the right, shown for each of the climate

scenarios, in the groups of bars, and then
management scenarios are the different shades

of gray and black.

For exotic grass, in terms of latecentury
projections, there's a lot of variability

among the climatechange scenarios, but those
three bars are pretty similar.

[laughs] Management is not having much of
an effect, and it's really difficult to restore

some of these areas.

Conversely, with juniper, we actually see
that management is fairly effective.

There's not a whole lot of variation among
the climate scenarios, but you do see that

the difference between that black bar of no
management and that lightgray restoration

management is pretty substantial, so management
is really better able to keep juniper at bay.

One more thing I'll show you.

Bringing it all together in sagegrouse habitat,
the legend is at the bottom there, and I don't

know how easy it will be for you guys to read
that.

I apologize for the small font.

Basically, what I'm showing here is sagegrouse
habitat over time.

The black lines are general habitats that
include some kind of marginal habitat states.

Then, in gray is highquality habitat.

Then the difference types of line.

No management is in the solid line, current
management is in dashed, and then restoration

management is in dotted.

Under current climate, you do see declines
in sagegrouse habitat, but we also see that

management can have some effect in buffering
that and making the decline not quite as sharp.

Then, under our three climatechange scenarios,
little bit of a different picture to what

we were seeing with some of the other, previous
slides.

In recovery from these undesirable state classes,
we see a little bit better outcomes, still

declines in some cases, and under this really
wet NorESM model, actual recovery to pretty

high levels over time so some variation, again,
among climatechange scenarios, and also some

potential for management to impact those outcomes.

Really briefly, a few conclusions.

In our study, climate change is likely to
cause expansion of moisture of steppe and

increases in wildfire.

The current landscape composition presents
a high risk for expansion of exotic grasses

and juniper, but actually, climate change
is projected to reduce exotic grass and juniper

relative to current climate, although they
did both increase.

Management was effective in controlling juniper
but not exotic grass.

Sagegrouse habitat declined in the short term
but then did rebound late in the century under

some of the climatechange scenarios.

We did publish this work in a special issue
of "AIMS Environmental Science" that is all

devoted to stateandtransition simulation modeling.

If you'd like to get in touch about these
results, I'm happy to discuss them.

In summary, bringing together all three of
these talks, we think that stateandtransition

simulation models are a useful and flexible
tool for conceptualizing vegetation dynamics

and projecting future conditions.

They've been used quite widely to inform management
of ecosystems, and I think they will continue

to be used quite a bit.

They're useful for a wide variety of ecosystems
because they're so flexible.

As we showed in these studies, although they
haven't been historically used this way, they

can be used to assess the impacts of climate
change in various different ways, and they

also allow us to incorporate data from multiple
models and other sources.

With that, thank you for your attention, and
we'd be happy to take any questions.

Ashley:  Excellent.

Thank you very much, Megan, Brian, and Jen.

We will open up the conference for questions.

We had one come in from John Wilson, and it
says, "Does the increasedprecipitation, hotter

scenario take into account ET rates?"

Megan:  I believe that question is for me.

John Wilson:  Sorry.

Can you hear me now?

Megan:  Yes.

John:  OK.

It was about the sagegrouse one.

ET rates might be expected to increase and
counter the increased precips.

I was just wondering if those scenarios took
that potential into account.

Megan:  The MC2 model does take evapotranspiration
into account.

It is incorporated in there, but we also have
to set thresholds in the model.

It's a little complicated.

Our threshold is actually based solely on
precipitation.

Actually, increases in evapotranspiration
are taken into account in the model but then

not in our threshold for distinguishing the
vegetation types, and that's actually something

that we probably could improve with a finertuned
rule set.

John:  OK.

As a quick follow, did you consider doing
a decreasedprecip scenario?

Megan:  We looked at, what was it, 40 scenarios,
I think, so a really large set of them.

Almost all of them show increases in precipitation.

The Hadley model was one of the driest.

That was just a little bit more moist overall,
but it was drier in the summers.

There were just a few that were drier overall.

Since we had to pick three, we wanted to pick
high performing models and also capture this

cloud.

The Hadley model was the closest to having
decreased precipitation, but even it had increased

precipitation.

Really, the projections are for more precipitation
overall.

John:  Thanks.

Ashley:  We have a question from Leila.

It says, "Is there any spatial output from
any of these studies?"

Jen:  This is Jen.

We initialized our landscape with spatial
data but we did not do spatial projections

in this case.

Megan:  The same with me.

This is Megan from the Sage grouse study.

I believe Brian has some spatial output.

Brian:  Yeah.

We did run them all spatially, and this included
things like the sizes of the fires, and also

distances for seed dispersal, and things like
that.

Spatial dynamics were definitely a part of
the model in that sense.

We haven't looked too carefully at the spatial
outputs, because if we run it multiple different

times, because of the randomization, and the
model, you will get some fairly different

spatial results.

That being said, we could probably come out
with some general conclusions regarding where

whitebark might, or might not be found on
the landscape.

Through time, then we can see this in the
species distribution model work.

You would expect that suitable conditions
are moving up slope as temperatures increase,

and things like that.

Those high elevations become more suitable
for whitebark.

In terms of looking at, say, early seral stage
whitebark, versus late seral stage whitebark,

and how that's distributed on the landscape,
we haven't looked at that as carefully yet,

but yeah, it's certainly possible.

Ashley:  Thank you.

Jen:  Thank you.

Ashley:  Then along with more of the models,
it says if you have no experience doing this,

how difficult is it, and is there a good reference
for getting started?

Like the transition models 101.

Megan:  Good question.

This is Megan.

In my experience, this is one of the more
user friendly models to use.

We all use this STSM platform.

The company who developed that product, the
ApexRMS does do trainings a couple of times

a year, I think.

There are various resources available on their
website, as well.

I would say, there's certainly a learning
curve, but it's not too bad.

Jen and Brian might have some other thoughts.

Brian:  I was going to add that I actually
copy and paste the Web links to the ApexRMS

website into the chat box for you to be able
to click on.

Simply, that the nice thing about that platform
is it's a graphical user interface.

You don't need to learn a coding language
necessarily in order to run it.

I think that's what makes the learning curve
a little bit shallow, at least initially.

But it does have a lot of flexibility and
things that you can do with it.

It's easy to get your feet wet.

It allows for making more complex models as
you get more comfortable.

Jen:  I was going to add this is Jen that
I'll agree with Megan and Brian that it's

pretty user friendly and that it has been
used quite a bit in working group meetings

with stakeholders and that kind of thing.

It is easy enough to use that you can update
some of these models on the fly depending

on the complexity and whether you're running
them spatially.

That may not be as possible to do on the fly.

But you can use this to communicate and work
with folks to update models and run them real

time even in some cases.

Megan:  I know Jen mentioned during her presentation,
the LANDFIRE models are a great place to start,

if you want to start modeling some ecological
systems.

You just need a template of historical conditions.

It's a fairly simple model.

You can modify it from there.

There is a starting place.

Ashley:  Excellent.

Thank you.

We have a few more questions.

One is for Brian.

It asks, "Can you explain a little bit more
about the depth of the model, that complex

diagram that you had showed?

Where does STM integrate with process modeling?"

Brian:  Sure.

Let me go ahead and see if I can jump back
to that slide that I believe Gregor is mentioning

there to give a little bit of a visual of
what I'm talking about.

This is one of the vegetation types.

This is the whitebark pine vegetation type.

We also have, what we call, strata.

Or, what you might call potential vegetation
type for spruce fir forest and also lodgepole

forest with their own sets of states and transitions.

I guess to keep it as brief as I can to give
you a little bit more meat, this area in the

upper left is essentially that three box and
arrow diagram I showed before with seedling

establishment.

These green arrows are representing agebased,
sort of deterministic transitions through

time.

You might think of a succession to mid and
lateclosed forest and midopened and lateopened

whitebark forest.

Then over here, we have sort of, you may think
of it as a sort of blister rust module.

Infection of various state classes by that
nonnative fungal pathogen, white pine blister

rust and also sort of the mountain pine beetle
kind of a module.

You might think of, down here, that here showing
mountain pine beetle infection and spread,

and also mortality events, both from rust
and from pine beetle.

Down here at the bottom, we have a placeholder
for areas that have burned from fire.

What you can't see on the same edge, I'm going
to show you these arrows going off into space

here are transitioning from dead forest to
a shrub herb state class or shrub herb potential

vegetation type.

That can then subsequently sort of transition
back to whitebark possibly or transition to

those other forest types, lodgepole pine or
sprucefir forest.

Ashley:  Thank you.

Megan:  Actually, I think there's a question
that may have just been sent to me from Robert

Schantz.

It says why is the moist shrubsteppe increasing
during the hot dry climate scenario change

from conifer and shrub?

As I mentioned, even the hot dry climate scenario,
it's dryer in the summer but it's actually

more moist overall.

There's a couple of things going into those
projections.

One is just the change in precipitation where
moisture is increasing although it's not increasing

a whole lot.

The other component that I didn't have time
to get into is that we assume, when we make

these state-and-transition transition models
climate sensitive that we're basically drawing

arrows between these ecological systems and
allowing transition from one to the other

and we're assuming that you can only get a
climaterelated shift following wildfire, otherwise

they'd be jumping back and forth yeartoyear.

There wouldn't be any sort of inertia of the
existing vegetation.

Because there's so much more fire, not only
are we getting more precipitation causing

that shift to the moist shrubsteppe, but also
these frequent fires are allowing opportunities

for the vegetation type to shift.

It's sort of twofold.

Hopefully, that makes sense.

Ashley:  Thank you.

We do have another sage-grouse question, and
it's from David Wood.

In accounting for sagegrouse habitat following
wildfire or other disturbances, is there a

lag effect until the sagebrush is tall enough
to meet seasonal habitat requirements?

Megan:  Good question.

One thing that we probably overestimate sagebrush
habitat postfire and that's I think one of

the things we mentioned in the paper.

If we get really big fires where we're not
getting sagebrush seeding back in, our model

is not capturing that well.

Actually, we determined which state classes
were suitable for sagegrouse based on basically

overlaying our maps of current state class
distribution with known locations of sagegrouse.

Some of those locations were lex.

We think we're actually probably overestimating
the habitat value of some of those more open

areas and that, as you point out, some shrubs
are going to have to come in and reach a certain

height and sort of a successional stage to
really provide quality habitat, at least for

some life stages.

We really sort of paint a broad brush of that's
one thing about sagegrouse is they have these

different requirements in different seasons.

We're kind of glossing over all of that.

It's a very broad picture view and we don't
really get to that level of detail.

Some spatial modeling would probably help
with that.

Some of it is sort of our rule set in how
general it is and some of it is that we're

not modeling these fire events spatially and
accounting for that seeding of sagebrush back

in.

Some of our habitat projections are a little
bit optimistic in that way.

Jen:  Brian forwarded me one from Wendy Ledbetter
asking was our model able to take into account

that with increased temperatures, natural
disturbances including wind events, hurricanes

and tornadoes could impact forest stands with
increased field loading and in some instances

hazardous fuels ?
Thinking about interactions among fire and

other disturbances.

Then, the impact of fire effects and safety
and frequency for safety as well.

Unfortunately, we didn't take into account
from these interactions and that's a good

point, but with increasing, especially hurricanes,
there might be even more potential for more

wildfires and more severe effects of the wildfires
in the future.

That was something that we weren't able to
take into account.

Another thing along those lines is changes
in some of the ability to conduct some of

these management actions like prescribed burning
as urbanization takes place.

Maybe with the increases in the wild land
to urban interface, we might have more safety

issues and constraints on burning.

That's another thing that we didn't take into
account, which would be great to account for.

Future work on that.

Brian:  I just had one more question come
in to me from Robert Schantz.

His question was, it seems like whitebark
pine should be increasing over time, all things

being equal.

Due to climateinduced reductions in subalpine
fir, Engelmann spruce, current distribution

might be confounding the climate envelope
model.

I guess I'll just add, I focused on whitebark
pine in terms of bringing in the species distribution

model with whitebark.

We also did the same thing for both lodgepole
forest and spruce fir forest.

That was work done by Nathan Butelic at Penn
State to sort of also have the ability of

spruce or lodgepole establishment be dictated
by their own sort of climate envelope models.

That was one piece of it.

I think that sort of the interesting thing
of what you're pointing out here is that you've

kind of have complex intersection of a variety
of dynamics.

Although we may expect spruce fir areas to
decline due to climate, but there are also

some opportunities that are presented.

For one thing, it's a shade tolerant species
as you may realize.

Over time it can replace whitebark stands.

Also, as whitebark stands are being impacted
by pine beetle and blister rust, that'll present

some additional opportunities for subalpine
fir and Engelmann spruce to sort of establish

their spread across the landscape.

I hope that answers your question.

Ashley:  All right, and we have time for
one more question.

It's from Nancy Green, and it is for Megan,
and it says, does your paper acknowledge the

possibility that local populations of sagegrouse
could be extricated before conditions or rebound

in the future, that the habitat may rebound
but the species might no longer be there to

use it.

Excuse me.

Megan:  That's a good point.

We make it very clear that we're not doing
any sort of sagegrouse population modeling.

It's all vegetation modeling.

I don't know that we mentioned that specifically.

That is true, and we definitely don't take
that into account in any of the demographic

stuff.

Yeah, hopefully sagegrouse won't be in that
bad of shape that quickly, but that is a good

point.

Ashley:  Thank you very much, Brian, Jen,
Megan for the excellent presentation and I

think that was a wonderful question afterwards
so thank you very much.

Brian:  Thanks again.

Megan:  Yeah, thanks a lot.

Ashley:  Shawn or Holly, did you have any
closing remarks?

Shawn:  Nope, thank you very much for the
presentations.

Really enjoyed the question and answer session
as well, so great job guys.

Ashley:  Thank you to the participants.