Regional Ecosystem Dynamics and Climate Feedbacks A. David McGuire Ted Schuur Scott Rupp Helene Genet Eugenie Euskirchen BNZ Annual Symposium 20 February 2015

Regional Ecosystem Dynamics and Climate Feedbacks Questions (1) How will interactive responses of disturbance regimes, ecosystem structure/function, and successional pathways to future climate variability and change influence regional ecosystem dynamics? (2) How will projections of regional ecosystem dynamics affect regional energy and water feedbacks to the climate system? (3) How will projections of regional ecosystem dynamics affect regional CO 2 and CH 4 feedbacks to the climate system?

Tasks (1)Couple model of fire regime (ALFRESCO) with model of ecosystem structure and function (DVM-DOS-TEM), incorporate information from empirical studies, and conduct retrospective analysis of the coupled model framework; (2)Apply the coupled model for future scenarios of climate and analyze changes in ecosystem function/structure at the regional scale; (3)Analyze water and energy feedbacks among the applications; (4)Conduct factorial experiments for future scenarios of climate change and evaluate effects of climate and disturbance on estimates of CO 2 and CH 4

Primary Support for this Research Identifying Indicators of State Change and Predicting Future Vulnerability of Alaska’s Boreal Forest (funded by DoD SERDP) Modeling Objective: Develop models that can forecast landscape change in response to projected changes in climate, fire regime, and fire management. Integrated Ecosystem Model for Alaska and Northwest Canada (funded by USGS and Alaska Landscape Conservation Cooperatives To develop a conceptual framework for integrating important components of an ecosystem model for Alaska and Northwest Canada including: fire, vegetation dynamics/succession, biogeochemistry permafrost dynamics, and hydrology.

Primary Support for this Research (cont.) Alaska Land Carbon Assessment (funded by USGS) Objectives are to assess: 1) the amount of C stored in ecosystems of Alaska 2) the capacity of Alaska ecosystems to sequester C, and 3) evaluation of the effects of the driving forces such as climate and wildfire that control ecosystem C balance. Permafrost Carbon Network (funded by NSF) Address the question “What is the magnitude, timing, and form of the permafrost carbon release to the atmosphere in a warmer world?” through synthesis by linking biological C cycle research in the permafrost region with well-developed networks in the physical sciences focused on the thermal state of permafrost.

Morning Session (Friday, 20 February 2015): Progress and Future Directions of Climate Feedbacks Research 8:30 – 8:50 Climate Change and the Permafrost Carbon Feedback – Ted Schuur 8:50 – 9:10 Projections of Climate and Vegetation Change – Scott Rupp 9:10 – 9:30 Consequences for Carbon Feedbacks – Helene Genet 9:30 – 9:50 Consequences for Water/Energy Feedbacks – Eugenie Euskirchen 9:50 – 10:10 Coffee Break 10:10 – 10:30 Future Directions for Climate Feedbacks Research – Dave McGuire 10:30 – 11:00 Discussion of Future Directions for Climate Feedbacks Research 11:00 – 11:15 Overview of Proposal Development – Roger Ruess 11:15 – 12:00 Theme highlights and directions – Other Theme Leaders 12:00 – 1:30 Lunch

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (1) Substantial progress on the effects of climate, fire, and top down permafrost thaw in uplands on vegetation dynamics and biogeochemistry. Still work to do (such as insects and pathogens). (2) Begun modeling research on the effects of thermokarst on wetland vegetation dynamics and biogeochemistry. (3) Need to begin modeling research on biogeochemical linkages of uplands and wetlands to inland surface waters (lakes and streams) and biogeochemistry of inland surface waters.

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (1)Substantial progress on the effects of climate, fire, and top down permafrost thaw in uplands on vegetation dynamics and biogeochemistry. Still work to do. – Finish two-way coupling of ALFRESCO with DVM- DOS-TEM and GIPL – Additional research on modeling successional trajectories – Begin research on modeling the dynamics and effects of insect and pathogen disturbances

ALFRESCO GIPL DVM-DOS-TEM Burned area O Horizon Thickness Vegetation Type Soil Thermal Profile Vegetation Carbon Soil Moisture Profile Moss Thickness Vegetation Canopy Elevation, Slope, Aspect. Historical Fire Mineral Soil Texture Air Temperature, Precipitation, Initial Vegetation Snow Water Equivalent Finish two-way coupling of ALFRESCO with DVM-DOS-TEM and GIPL

Additional Research on Modeling Successional Trajectories. (Johnstone et al. 2010)

Relative dominance of spruce in post-fire recruitment is related to fire severity, pre-fire stand age and drainage conditions. The drivers of post-fire recruitment composition Fire severityDrainageStand age Johnstone et al. 2010

Drainage Pre-fire vegetation (PFV) Mesic Subhygric Failure High/moderate low Mixed trajectory Pre-fire stand age Xeric Pre-fire vegetation (PFV) < 50 yr > 50 yr Failure Mixed trajectory high low/mod erate Deciduous trajectory Pre-fire stand age Thermokarst ? < 50 yr > 50 yr Failure Evergreen trajectory PFV = decid. Deciduous trajectory Fire severity PFV = everg. PFV = decid. Deciduous Fire severity PFV = everg. Thermokarst ? Regression Tree Approach to Modeling Dynamics of Post-fire Recruitment and Successional Trajectories : Helene Genet

Build Successional Dynamics into DVM – DOS – TEM (Euskirchen et al. 2009) The dynamic vegetation model simulates carbon and nitrogen dynamic of various plant functional types, competing for light and nitrogen uptake.

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (1)Substantial progress on the effects of climate, fire, and top down permafrost thaw in uplands on vegetation dynamics and biogeochemistry. Still work to do. – Finish two-way coupling of ALFRESCO with DVM- DOS-TEM and GIPL – Additional research on modeling successional trajectories – Begin research on modeling the dynamics and effects of insect and pathogen disturbances

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (2) Begun modeling research on the effects of thermokarst on wetland vegetation dynamics and biogeochemistry. – Development of Alaska Thermokarst Model (ATM) – Development of Peatland DOS-TEM/DVM-DOS-TEM – Coupling of ATM with Peatland DVM-DOS-TEM

Quantifying the Area Susceptible to Thermokarst Function of ice content, landscape position (lowlands), presence of peat (histels), and presence of permafrost (from Helene Genet)

Development of the Alaska Thermokarst Model

Conceptual Model of Landscape Change Associated with Thermokarst Disturbance Permafrost Plateau Thermokarst Lake Young Fen (< ~100 yr old) Young Bog ( < 100 yr old) Fen L Old Fen Old Bog

Thermokarst-Prone Landscape Change Questions What is the current distribution of land cover types in thermokarst-prone landscapes? What are the transition rates among land cover types during the satellite era? What are the controls over transition rates (e.g., climate, permafrost, hydrology)? How is the distribution of land cover types projected to change in the future?

Biogeochemistry Change Questions in Thermokarst-Prone Landscapes How do land cover transitions in thermokarst-prone landscapes (and associated changes in permafrost and hydrology) influence carbon storage and fluxes in land cover types? How do land cover transitions (and associated changes in permafrost and hydrology) influence the loading of carbon into lake and stream networks ? How will climate change (and associated changes in land cover transitions, permafrost, and hydrology) influence carbon dynamics in lakes, stream networks, and wetland complexes?

PERM Region 05B: Discontinuous Boreal Permafrost

Bookkeeping Model: Estimates of Carbon Dynamics of each Land Cover Type (Jen Harden, Jon O’Donnell, Others) Collapse Bogs (CB) Delta Shallow SOC (g C m-2 yr-1) Delta Deep SOC (g C m-2 yr-1) Delta VEGC (g C m-2 yr-1) > Permafrost Plateau Forests (PPF) Treed Bogs (TB)1300 Net CH4 Emissions (g C m-2 yr-1) NEE (+ to atmosphere) (g C m-2 yr-1) DOC Flux (g C m-2 yr-1) Collapse Fens (CF) Thermokarst Lakes (TL)

Bookkeeping Model: Transition Rates (Proportion Per Year) From/To (Proportion) Permafrost Plateau Forest Thermokarst LakeCollapse FenCollapse BogTreed Bog Permafrost Plateau Forest (PPF)NAN NAN Thermokarst Lake (TL)NAN NAN Collapse Fen (CF)NAN NAN Collapse Bog (CB)NAN Treed Bog (TB)0NAN

Bookkeeping Model: Dynamics of Thermokarst Land Cover Types Using Contemporary Transition Rates

Bookkeeping Model: Consequences of Using Contemporary Transition Rates for Cumulative Changes in Carbon Storage

Raised WT Lowered WT Control Development of Peatland DOS-TEM/ DVM-DOS-TEM has relied on data from The Alaska Peatland Experiment (APEX)

Peatland-DOS-TEM Model (Fan et al. 2013) No historical fire In APEX sites No historical fire In APEX sites

Peatland Model (Fan et al. 2013)

from Fan et al (units are g C m -2 yr -1 )

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (2) Begun modeling research on the effects of thermokarst on wetland vegetation dynamics and biogeochemistry. – Development of Alaska Thermokarst Model (ATM) – Development of Peatland DOS-TEM/DVM-DOS-TEM Zhaosheng Fan (now DOE-ANL): Fen Yanjiao Mi: Collapse Scar Bog, Permafrost Plateau, and biogeochemical consequences of thermokast disturbance based on data from chronosequence studies. – Coupling of ATM with Peatland DVM-DOS-TEM This is an activity of the IEM team (science question: How does coupling (internal feedbacks of the system) influence wetland dynamics and biogeochemistry?)

Future Directions of BNZ Regional Climate Feedbacks Research (Dave McGuire Perspective) (3) Need to begin modeling research on biogeochemical linkages of uplands and wetlands to inland surface waters (lakes and streams) and biogeochemistry of inland surface waters. – Mack et al. DOE proposal in review – TEM6 models DOC and DON loading

Addressing Uncertainties in BNZ Research

Afternoon Session (Friday, 20 February 2015): Quantifying Uncertainty and Understand Legacies 1:30 – 1:35Overview of the Afternoon Session – Dave McGuire 1:35 – 1:50Analyzing Uncertainty in Field/Experimental Research: A Hierarchical Bayesian Perspective – Colin Tucker 1:50 – 2:05Role of Model-Data Fusion in Analyzing/Quantifying Uncertainty – Dave McGuire 2:05 – 2:20Characterizing Uncertainty in Applications of Ecological Models – Jeremy Littell 2:20 – 2:40Perspectives on Studying the Role of Legacies in BNZ Research – Michelle Mack 2:40 – 3:00 General Discussion 3:00 – 3:15 Coffee Break

Role of Model-Data Fusion in Analyzing/Quantifying Uncertainty (D. McGuire and H. Genet) Sources of Uncertainty in Models Traditional Parameter Uncertainty Analysis Model-Data Fusion

Sources of Uncertainty in Modeling Conceptual Uncertainty –Compare dynamics of alternative models Formulation Uncertainty (equations) –Compare models with different equations Parameter Uncertainty –Various approaches to analyzing uncertainty Application Uncertainty –Jeremy Littell’s presentation

Parameter Estimation Literature-based estimation Experimentally based estimation Calibration

Traditional Parameter Uncertainty Analysis Definition Relates the variability of model predictions to uncertainty in parameter estimates. Uncertainty analysis is analyzing the effect of the variance of a parameter to the model predictions.

Error Propagation Amplification Compensation Error amplified : σ z > σ x Error compensated: σ z < σ x

Multiple parameter uncertainty analyses: 100 independent sets of parameters value using Monte Carlo iterations applied to 43 parameters varying within an “acceptable” range, assuming a uniform probability density function for each parameter. SWHi = Threshold snow water equivalent at which forage intake goes to zero for each ungulate category ; GWP = Water content per kg protein ; FGB = Forage intake per kg of body mass ; BMi = initial body mass for each ungulate category; PSN(24) = Snow depth in unburned forest. From Turner et al. (Ecological Applications).

Model-Data Fusion Model and data integration, also called model–data fusion or model–data synthesis, is defined as combining models and observations by varying some properties of the model, to give the optimal combination of both (Raupach et al., 2005). Model–data fusion encompasses both calibration and data assimilation. Model–data fusion can be characterized as both an inverse problem, analyzing a system from observations, and as statistical estimation. Calibration: Parameter estimation to produce desired outputs for a given input. Data Assimilation: observations are used to refine estimates of the evolving model state. Model–data fusion brings together four components: – external forcing, – a model that relates model parameters, state and external forcing to observations, – observations, and – an optimization technique.

Example: Uncertainty in the fate of soil organic carbon: A comparison of three conceptually different decomposition models in a larch plantation (He et al. JGR-B in press) Compared three structurally different soil carbon (C) decomposition models (one driven by Q10 and two microbial models of different complexity) The models were calibrated and validated using four years of measurements of heterotrophic soil CO 2 efflux from trenched plots in a Dahurian larch (Larix gmelinii Rupr.) plantation. Parameters in each model estimated using a Bayesian data assimilation framework

Model Validation

Differences in Model Dynamics and Uncertainty Simulated 100 years responses of SOC stock for the three models. Top panel (a-c) is trenched plot simulation; bottom two panels (d-i) are model simulations under 4.8 °C progressive increasing soil temperature and litterfall. The deep blue and red lines (for 3-pool Q10 model) represent ensemble mean from the 100 independent optimization runs for each model, the light colored lines are the results from each ensemble member.

Afternoon Session (Friday, 20 February 2015): How should BNZ quantify uncertainties and understand legacies? 3:15 – 3:30Organization of and charge to working groups and discussion points 3:30 – 4:30Working groups convene 4:30 – 5:30 Working Group Reports 5:30 – 7:00 Mixer, Dinner, Poster Session

Charge to Working Groups Data and Experimental Uncertainty Group: How do we quantify uncertainty across the extended site network? Model Uncertainty Group: What aspects of uncertainty should/can be addressed? What approaches should be used to addressed understand/quantify parameter-based uncertainty? Which data from the extended site network can be used to address uncertainty? Legacy Group:

Morning Session (Saturday, 21 February 2015): BNZ Management and Outreach 8:30 – 9:30BNZ Site Management – Discussion re gearing up the ESN – Jamie Hollingsworth 9:30 – 9:50Progress and Directions of BNZ Information Management – Jason Downing 9:50 – 10:15Progress and Directions of BNZ Education and Outreach – Elena Sparrow 10:15 – 10:30Coffee Break 10:30 – 10:45USDA Forest Service Perspective on BNZ LTER – Paul Anderson 10:45 – 11:00LTER Citizen Science – Christa Mulder and Katie Spellman 11:00 – 11:15 ITOC – Mary Beth Leigh 11:15 – 11:45Discussion of the Future Directions of BNZ Outreach – Marie Thoms and Alison York 11:45 – 12:00 Organization of and Charge to Afternoon Working Groups – Roger Ruess 12:00 – 1:30 Lunch

Afternoon Session (Saturday, 21 February 2015): Next Steps in Developing the BNZ Renewal Proposal 1:30 – 3:00 Theme Working Groups Convene 3:00 – 3:15 Coffee Break 3:15 – 4:30 Reports from Theme Working Groups 4:30 – 5:00 General Discussion on Next Steps 5:00 Adjourn