1. The Downscaled Climate Projection Has Arrived – NOW WHAT?
What does an 8°C (14.4 °F) rise in annual average temperature mean to a resource manager? THAT DEPENDS….
GCMs tell us to expect warming of in the range of 8 degrees C ( annual air temperature by the end of the century, when compared to the last decades of the 20th century. That sounds like a large change, but what does it mean to the biotic resources and ecosystem functions that resource managers care about?
To address most management-relevant questions, downscaled climate projections are just the beginning, and translating climate projections into an understanding of ecosystem change requires consideration of complex processes, and multiple feedbacks, both positive and negative

2. Habitat & Species Vulnerability
Ecosystem Response Models – linking climate projections to the needs of resource managers
Habitat & Species Vulnerability
Disturbance
Permafrost
Hydrology
Vegetation
Climate
Conservation Action
Thinking in terms of a climate vulnerability analysis for species and habitats, the challenge is to assess exposure in an ecologically meaningful way. Changes in temperature and precipitation may directly affect species at the individual or population level, but more often influence indirectly through ecological processes, such as disturbance, and changes in vegetation/hydrology.

3. Conceptual Model
Ecosystem response models can take many forms, from simple to complex. At the simple end, is a non-quantitative conceptual model linking climate to habitat to species response. This box-and-arrow diagram outlines pathways by which changes in the coastal zone might affect birds. Blue boxes are physical processes, green boxes represent habitat change, and white boxes represent anticipated response of species or species groups. The model is a useful way to organize our thinking, and one of its utilities is to highlight areas of uncertainty (Click)
One such area of uncertainty occurs in this part of the model. We are unsure of the future availability of delta mud flat habitat, because there are opposing influences. Sea level rise and increasing storminess may result in diminished habitat availability, while increased sediment load due to thawing permafrost may result in higher rates of deposition and greater habitat availability. To resolve this issue, we would need a numerical model that quantifies these processes and predicts the net effect of opposing trends.

4. Soil Thermal Properties
Downscaled GCM Data
ALFRESCO
GIPL-1
Soil Thermal Properties
DOS-TEM
Vegetation type
Moss & Organics
The project I would like to discuss for a few minutes is on the complex end of the scale. It is a multi-year, multi-discipline effort supported by Alaska LCCs and the Alaska Climate Science Center. The objective is to link several “specialist” models to that they integrate important components of an ecosystem model for Alaska, including: fire, vegetation succession, permafrost dynamics, and hydrology. This figure illustrates the concept: downscaled climate models drive the system and 3 sub-models pass information back and forth to represent the mutual influences. ALFRESCO simulates fire and vegetation succession processes, DOS-TEM simulates organic layer thickness, carbon and nitrogen fluxes, and competitive interactions among plant functional types. GIPL-1 simulates energy flux between land and atmosphere and the resultant effects on active layer thickness and permafrost distribution. Each specialist does its own job well, but to improve our predictions of ecosystem response to warming, we need them to work together.
Burned area
Soil moisture
Fire Severity
Veg. canopy

5. Permafrost
I will not delve into the details of each model, but a little background will help you understand the need for coupling the three component models. The photos on the left show a typical boreal forest scene and a soil exposure. You can clearly see the vegetated surface, and a layer of brown peat and organic soils, and then the gray mineral soil. The surface layer that thaws annually is the “active layer” and the permanently frozen soil beneath is permafrost. It isn’t completely clear from this photo where that boundary lies, but there is ground ice visible within the permafrost at the left side of the figure, so we know that permafrost is very near the surface. Permafrost prevents downward water percolation – it acts like bedrock, except it is bedrock that melts away when temperature warms.
Air temperature is a critical driver of permafrost distribution around the globe. So, if the atmosphere warms, we can expect the ecological foundation to deform and fundamentally alter the ecosystems that sit at the surface. The ecosystem, however, also feeds back to the permafrost. Vegetation and snow cover act as insulation. Thick moss cover, for instance, keeps the permafrost from warming as much in the summer, conversely, thick snow cover prevents permafrost from cooling as much in winter. Disturbances such as fire that remove the vegetative cover also change the soil temperature and permafrost distributions. The take-away message is that there are feedbacks among climate, permafrost, vegetation, and disturbance that all must be taken into account simultaneously if we are to project future habitat conditions.
Ground Ice

6. Permafrost Model Alone
To illustrate ouput from just the GIPL. The map shows modeled mean annual ground temperature for the beginning of the century. Blue colors indicate stable permafrost, red colors indicate a mean annual soil temperature above 0 C, thus melting of permafrost. Keep an eye on the regions highlighted in the ovals. (click) At the beginning of the century, the west coast is losing permafrost, interior is mostly stable with a few areas of loss, and northern Alaska is stable. (click)
Here is a projection of permafrost extent at mid-century Western Alaska will be too warm for permafrost to persist, esatern is about half and half, and northern Alaska is still stable. (click)
By end-of-century, however, the North Slope is the only portion of the state cold enough to retain permafrost. So, this model alone is informative – it predicts that there is a certain degree of resistance in arctic Alaska, and deep cold permafrost will not be disappearing there in the next few decades.
Mean Annual Soil Temperature – 1 m Depth

7. Charring Boreal Forest Forecast
But we know that permafrost distribution will be influenced by fire, and the permafrost model results I just showed you do not take changing fire regime into account. This graph shows the expected area burned within a test area (Yukon River basin), historically (black) and forecasts driven by two different GCMS (red and blue) under the IPCC A1B emissions scenario. This model predicts an increase in fire for the next couple of decades, but then decline again as the dominant vegetation and fuel loads change.
Comparison of the distribution of the relative frequency of annually burned area percentage in bins of <0.5%, %, 1.0–1.5%, %, and >2.0%, over AKYRB region from the historical database ( ) and ALFRESCO simulations for the projected period (2007 – 2099). Note that there are two ALFRESCO fire projections driven by GCM outputs of CCCMA-CGCM3.1 (CCCMA) and MPI ECHAM5 models (ECHAM5) for the A1B emissions scenario.
Driven by GCM outputs of CCCMA-CGCM3.1 (CCCMA) and MPI ECHAM5 models (ECHAM5) , A1B emissions scenario.

8. Pilot Year : Completed in January 2012
Identified and obtained all variables needed to drive and couple the individual models
How to modify outputs from one model so that they are meaningful to another model
How to downscale driving variables
How to compare similar output variables between models to understand differences and/or uncertainties in these outputs
The modeling team has completed the pilot phase of this project. They have built and tested the platform, completing such tasks as assembling the variables needed to run the model, putting variables in a form so that one model can talk to the other, and downscaling to a common scale.

9. Phase II Objectives: 4-year Plan
1) Model coupling
To develop and apply a fully synchronous coupled model over the Western Arctic integrating important components of an ecosystem model for Alaska including: fire, vegetation succession and permafrost dynamics
2) Tundra fire and treeline dynamics
To evaluate, test and incorporate tundra fire and treeline and shrub dynamics into the Alaska IEM
3) Thermokarst dynamics
To develop a conceptual model of thermokarst dynamics at a landscape scale and to evaluate, test and incorporate thermokarst dynamics into the Alaska IEM
4) Wetland dynamics
To design wetland field dynamics program to support the Alaska IEM and develop, evaluate, test and incorporate wetland dynamics into the Alaska IEM
The modeling team has an ambitious 4-year plan of implementation. In the current year they will run the model in a fully-coupled dynamic mode. At each annual time step, the models will output all the variables describing ecosystem state, exchange information, and reiterate that for as many simulated years as we ask it to. The team will also be working to improve the way in which the model simulates tundra fire and treeline dynamnics. Some of the more difficult issues – representing thermokarst dynamics (the slumping or sinking of the surface when permafrost thaws and ground ice melts) and attendant changes in wetland dynamics – are getting initial attention now but are not expected to be implemented until the 3rd or 4th years.

10. Domain and Scale
The domain of the model will be expanded to include portions of Canada adjacent to Alaska. It will include nearly the entire area of the NW Interior Forest LCC, shown in brown. Data availability (no PRISM climatology) prevents us from including the northeastern edge of that LCC.
Model structure is spatially explicit – input and output are based on grids that are on the order of 1 km in scale. This is a compromise level of resolution – it is not as fine as many land managers would like – for instance, it blurs the heterogeneity contained in typical land cover maps based on 30-m pixels derived from LANDSAT imagery. On the other hand, it stretches and strains the resolution of GCMs and many remote-sensing products that output data at the scale of 10s to 100s of kilometers.

11. Other Stakeholder Groups
Links to Management
Other Stakeholder Groups
Alaska Integrated Ecosystem Model
Impact Models
modeoutput x
Hypothetical
Model
model output y
Conservation & Resource
Management
Decisions
canopy cover
thermokarst
Habitat
Change
Models
species composition
probability of fire
vegetation cover
Fire
Management
Models
probability of fire
This is a busy slide that is simply meant to convey that the IEM, on the left, is not a one-stop shop to answer all our questions. We anticipate the development of impact models that tier off the IEM to address specific questions of interest to managers and other stakeholders.
An important aspect not really represented in this diagram is the role of model development and testing in identifying areas of greatest uncertainty, which feeds back onto the design of research and monitoring activities. This relates back to the data sparse environment in which we live, and the desire to target scarce monitoring dollars where they can do the most good.
surface hydrology
vegetation cover
Animal Performance
Models
biomass productivity
surface hydrology

12. Collaboration Across Disciplines
Terrestrial Ecology
Ecosystem Modeling
Fire Science
Programming
Hydrology
I’d like to close with some observations about the sociology of this project. The Pis are both with the U of Alask: Dave McGuire from the Coop Research Unit, and Scott Rupp from Scenarios Network for Alaska Planning. The team includes faculty, post-docs, USGS staff, and university staff representing a wide range of expertise and disciplines. It has been a bit of a revelation for me to see that academics from different departments and labs need this kind of project to motivate collaboration across their own organizations.
Geophysics
Remote Sensing
Data Management

13. Collaboration Across Organizations
Academic departments
Alaska Climate Science Center
Arctic LCC
Western Alaska LCC
In terms of financial support and oversight, the Alaska Climate Science Center is now taking a central role, but with substantial contributions from the Arctic and Western Alaska LCCs. NWIF LCC has no project funding to contribute, yet, but participates in team meetings providing guidance and conveying the needs of the resource management community to the modeling team.
In closing – the problem of forecasting future condition under changing climate scenarios is enormously complex, but we can make progress if we bring multiple disciplines together, fund the work in a collaborative fashion, and take the approach of patient investors who are in this for the long haul.