The general theme of our work is to gain a quantitative understanding of plant ecosystem and community dynamics across scales from the individual to the globe. This is achieved by a balanced combination of field research, novel statistical methods, and numerical models.

Assimilation of imaging spectroscopy data to improve the representation of vegetation dynamics in ecosystem models

The ability to seamlessly integrate information on forest function across a continuum of scales, from field to satellite observations, greatly enhances our ability to understand how terrestrial vegetation-atmosphere interactions change over time and in response to anthropogenic and natural disturbances. This project focuses on the use of field and high-spectral resolution remote sensing observations (i.e. imaging spectroscopy, IS), within an efficient model-data assimilation framework, to improve the characterization of vegetation dynamics in terrestrial ecosystem models. This effort comes at a crucial time because the experimental, remote sensing, and modeling communities have entered into an increasingly data-rich era; however the tools needed to make use of the numerous but disparate data for model improvements are currently lacking. For example, remote sensing can provide detailed spatial and temporal information on a number of important biophysical and biochemical properties of ecosystems. State-of-the-art dynamic vegetation ecosystem models, such as Ecosystem Demography (ED2.2) model (Medvigy et al., 2009), a physiologically-based forest community model, can potentially use this information to improve model representation of vegetation dynamics. ED2 is especially relevant to these efforts because it contains a sophisticated structure for scaling ecological processes across a range of spatial scales: from tree- level physiology to stand demography to landscape heterogeneity to regional carbon, water, and energy fluxes, which allows for the direct use of remotely sensed data at the appropriate spatial scale. The project leverages extensive field and imaging spectroscopy (IS) data that have been collected by Co-PI's Shawn Serbin and Phil Townsend within the upper Midwest, US, directly within an ecosystem modeling framework. We are working to utilize a radiative transfer modeling (RTM) module being developed by Serbin and Dietze for use with the ED2 model and Predictive Ecosystem Analyzer (PEcAn, LeBauer et al., 2013) workflow system (www.pecanproject.org) to enable efficient assimilation of spectral reflectance observations from imaging spectroscopy data (and eventually any optical remote sensing observations, such as Landsat and MODIS/VIIRS). Through this open-source workflow system we will facilitate direct assimilation of spectral observations rather than derived products. This will improve the models parameterization of canopy optical properties and the surface energy balance. Through state-variable data assimilation we will fuse AVIRIS, flux towers, forest inventories, and model projections to reconcile estimates of vegetation composition and carbon pools and fluxes. The resulting data product will be analyzed to better understand the drivers of spatial and temporal variability in the carbon cycle and the sources of uncertainty in these estimates. This project would be an important step toward the operational capacity to assimilate reflectance observations, uniformly, within sophisticated ecosystem models with the goal to accurately constraining model projections of carbon pools and fluxes of terrestrial ecosystems.

Brown Dog

Press Release Researchers in Professor Dietze's Ecological Forecasting lab at Boston University will be using the Brown Dog tools in two case studies focused on the responses of forests to climate change. In particular, both case studies aim to improve the representation of forests and other vegetation in Earth System Models, which is currently the second largest source of uncertainty in climate change projections behind cloud physics. The first case study will focus on extracting vegetation data from historical land survey information collected by the US General Land Office in 1800's under the Homestead Act. Machine learning will be used to interpret hand-written survey notes and convert them to maps of pre-settlement forest composition and biomass. This data will form an important baseline from which to calibrate earth system models and judge the impacts of human land use and climate change. The second case study will focus on curating and synthesizing contemporary data on the forest carbon cycle collected by individual researchers around the world. These data are 'small-data' when viewed individually, but collectively their information have the transformative potential of 'big data' to improve and calibrate Earth System Models. This case study will build upon an ongoing bioinformatics project in the Ecological Forecasting lab, the Predictive Ecosystem Analyzer http://pecanproject.org, which has developed tools for making forest models more accessible and their analysis more automated. Within the Department of Earth and Environment, Professor Dietze and his Ecological Forecasting lab specialize in assimilating data into forest models in order to improve predictions of how climate change, forest management, and natural disturbance will affect forest biodiversity and the carbon cycle.

Vulnerability of forest biodiversity to climate change: Individual risks and regional responses

Individual trees bear the burden of responding to global climate change but predictions of associated changes to forest biodiversity are usually based on aggregate species level responses. However, species responses are not synonymous with individual responses to drought and changes in growing season length. To address this discontinuity, we are collecting individual-scale data in forests spanning a continental gradient of temperature and precipitation in eastern North America. We measure annual fecundity, germination, recruitment, growth, and mortality along with climate drivers in a series of new and established > 1ha forest plots. This collection of forest demographic and ecological data is being assimilated into Ecosystem Demography 2.2 (ED2) model to produce improved process-based predictions of regional vulnerability of forest biodiversity to the combined risks of increasing drought and growing season length.

In the Dietze Lab, postdoc Brady Hardiman is managing 2 of the 13 project sites. These are located in the White Mountain National Forest, New Hampshire at the Bartlett Experimental Forest (BEF) and in northern Wisconsin at the University of Notre Dame Environmental Research Center (UNDERC). Plots at BEF (4x 1ha each; 2 high elevation and 2 low elevation) and UNDERC (2x 1ha each w/ 0.5ha buffer) were installed in 2012. Each plot was surveyed, and gridded to 10m and all stems were mapped. In 2013, artificial gaps will be created in plots at UNDEC. Each 1ha plot contains nested sapling (6x 100m2) and seedling recruitment plots (25x 2m2), as well as a grid of seed rain traps (25). Ancillary measurements include annual hemispherical photography, soil cores, ground based portable canopy LiDAR, and continuous measurements of soil moisture & temperature, air temperature & relative humidity, and below-canopy PAR. Each of the 13 sites is associated with, or near, an Ameriflux eddy covariance tower and falls within a core or relocatable NEON site.

This work is supported by the National Science Foundation under award no. #1318164. Collaborators include Jim Clark, Alan Gelfand, Andy Finley, Sean McMahon, Jackie Mohan, and Maria Uriarte.

Building forest management into Earth system modeling: Scaling from stand to continent

A wide array of human-induced disturbances can alter the structure and function of forests, including climate change. Changes in climate are anticipated to impact forest ecosystems and the services they provide. Under changing climate, forest management is necessary to sustain biodiversity, biogeochemistry, and services of forest ecosystems. Unfortunately credible incorporation of forest management practices, for climate change mitigation and adaptation, into continental scale Earth system models has been hampered mainly due to scaling assumptions. These assumptions stem from stand-level studies and are not rigorously evaluated at regional to continental scales. The primary objectives of this research are to determine how the variations in forest management, climate, and disturbance regime impact the structure and function of forest ecosystems, and to quantify the relative importance of these drivers of ecosystem structure and function at stand to continental scales across North American forests. The study areas are the Pacific Northwest (PNW) and Southeastern United States (SE) which are two major U.S. forested regions that have significant, and very different, management practices.

In Dr. Dietze's Ecological Forecasting lab, postdoc Afshin Pourmokhtarian is in charge of model simulations and incorporation of different forest management practices and disturbance regimes for two forest domains using the Ecosystem Demography model (ED2; Medvigy et al., 2009). Afshin works closely with two other postdocs (U. Florida and U. Alabama) under the mentorship of senior scientists of the project in ManDiForE (MANagement and DIsturbance in FORest Ecology) which is an interdisciplinary team of forest and landscape ecologists, remote sensing scientists, biostatisticians, and ecosystem modelers. Their main goal is to improve understanding of how management and disturbance influence forest complexity, ecosystem services, and climate feedbacks. Within each region, remote sensing land-cover information will be analyzed to select 400 candidate 10x10 km areas with different management type (total of 800 runs), plus ensemble runs that modify model disturbance assumptions.

This project is supported by the National Science Foundation (NSF) under award #1241894 and will collaborate with National Ecological Observatory Network (NEON). PI of this project is Christina Staudhammer and collaborators include Binford, M., Desai, A., Dietze, M.C., Starr, G., and Stoy, P.C.

PalEON: A Paleo-Ecological Observatory Network to assess terrestrial ecosystem model


PalEON is an interdisciplinary research group of paleoecologists, statisticians, and ecosystem modelers working together to study how climate variations shape forest dynamics across a range of timescales.  Specific goals include developing a coherent inferential framework with rigorous estimates of uncertainty for paleoecological data, applying these techniques to reconstruct variations in forested ecosystems for the last 2000 years from the Great Lakes to New England, and then assimilating these datasets into a suite of regional-scale ecosystem models to infer presettlement biogeochemical cycles.  PalEON has recently received funding from NSF-Macrosystems to begin a two-year effort towards these goals, with an emphasis on initial development of methods and datasets, community-building, and interdisciplinary training in paleoecology, statistical ecology, and ecosystem modeling.

Prof. Dietze is a co-PI on the PalEON project and responsible for coordinating the different ecosystem modeling teams involved in the project. The long-term goal is to be able to make inferences about presettlement biogeochemical cycles and understand how these legacies affect current and future ecosystem dynamics. Research questions focus on validating ecosystem models at centennial time-scales, making inference about pre-settlement ecosystem dynamics and biogeochemical cycles, and exploring the sensitivity of models to historical vegetation. The PalEON project kicks off May 2011 so stay tuned for future results.

Integrating paleoecological analysis and ecological modeling to elucidate the responses of tundra fire regimes to climate change

Recent climate warming has resulted in profound environmental changes in the Arctic, including shrub-cover expansion, permafrost thawing, and sea-ice shrinkage. These changes foreshadow more dramatic impacts that will occur if the warming trend continues. Among the major challenges in anticipating these impacts are surprises in system components that have remained relatively stable in the observational record (typically past few decades in arctic regions). Tundra burning is one such change, with available evidence suggesting that ongoing climate and vegetation change could significantly increase tundra burning. Tundra burning is emerging as a key process in the rapidly changing Arctic, and knowledge of tundra fire-regime responses to climate change is essential for projecting Earth system dynamics, developing ecosystem management strategies, and preparing arctic residents for future change. The short duration of observational fire records, paucity of fire-history studies, and possibility of novel future climate and vegetation greatly hinder our ability to evaluate how tundra fire regimes may respond to future climate and vegetation change.

Paleoecological analysis and ecological modeling circumvent these limitations and offer the only ways to acquire such information. This project, collaborative with the Hu lab (U. Illinois), the Higuera lab (U. Idaho), and Paul Duffy (U. Alaska), takes advantage of the complementary properties of paleoecological and modeling approaches to (1) quantify historic climate-vegetation-fire relationships in the tundra ecosystems of the North American Arctic, (2) conduct multi-proxy analyses of lake sediments to reconstruct tundra fire regimes during periods of the late Glacial and Holocene with novel combinations of climate and vegetation, (3) re- parameterize ALFRESCO, a landscape ecosystem model initially developed to study the response of subarctic vegetation to changes in climate and fire regimes, for predicting tundra fire regimes under the suite of IPCC climate scenarios for the 21st century, (4) modify ED, a state-of-the-art physiologically- based model for tundra ecosystem studies, and (5) couple ED with ALFRESCO to simulate carbon dynamics related to 21st-century shifts in tundra fire regimes. Each of these elements is at the forefront of ongoing research in the respective areas, and together they promise to substantially advance our knowledge of climate-vegetation-fire interactions of tundra ecosystems for the past, present, and future.

Biofuels and ecosystem services

As part of the Energy Biosciences Institute the lab is working to construct and parameterize models of biofuel agroecosystems at the regional scale. This project focuses on assessing crop suitability in different regions an forecasting the effects of different crops on various ecosystem services, such as carbon, water, and nutrient cycling, under different crop and land-use scenarios.

Biofuel Ecophysiological Traits and Yield Database (BETY-db)

We are on the verge of launching a publicly-accessible electronic clearinghouse on information about biofuel productivity and ecosystem impacts. You can visit the beta version of the site HERE

PECAn: The Predictive Ecosystem Carbon Analyzer


In support of our EBI work David LeBauer (post-doc) is leading development on a system to automate model parameterization, data assimilation, and model analysis. PECAn is not a model but scientific workflow which encapsulates ecosystem models and connects and automates a number of analytical tools designed to manage the flows of information in and out of ecosystem models. It is designed to make ecosystem models more accessible to the research community and in particularly focuses on making models easily updatable and enabling more direct feedbacks between field research and modeling. The current beta version of PECAn focuses on using plant trait information from databases like BETY-db to constrain models, to propagate uncertainies through models, and to attribute model uncertainty to particular parameters. By identifying and quantifying the key parameters that limit model performance, PECAn allows further experimental work to be optimized to gather the greatest amount of new information for a given amount of effort.

Woody biofuels: hybrid poplar and beyond

The lab is closely involved with the woody biofuel trials being conducted at the EBI Energy Farm. Working collaboratively with Gary Kling, Sarah Davis, and Evan DeLucia we are investigating the yields of scores of different hybrid poplar and hybrid willow clones. In addition, we are working very intensively on the Novel Woody Feedstocks experiment, where 21 different trees and shrubs are being investigated for their potential use in coppice forestry. In this project we are measuring not only overall growth, but also leaf traits and photosynthesis, instantaneous and isotopic water use efficiency, biomass partitioning, nonstructural carbohydrates, phenology, and soil carbon and soil respiration.

At a regional scale, Dan Wang (postdoc) has lead efforts to model the potential yields and ecosystem services of woody biofuels using the ED model. Focusing our initial efforts on hybrid poplar, Dan has successfully calibrated and validated ED using literature trait data and yield estimates from a number of sites and clones. She has also produced maps of potential yields and optimal rotation period across the lower 48 and compared poplar against Miscanthus, a particularly productive perennial grass biofuel crop. This analysis demonstrates that there are many areas where poplar yields are economically viable and a few key regions, such as New England and the upper Great Lakes, where poplar yields are predicted to be higher than Miscanthus. Dan is currently working to expand this analysis to additional woody biofuel crops

Perennial grasses and native prairie

Xiaohui Feng (a.k.a. Sunny) is taking a closer look at the dynamics of mixed-species polycultures using a combination of field work and modeling. Thus far her field work has shown intriging relationships between community composition and photosynthetic rates, both looking across species and looking at the seasonal variability within individual species. Sunny is next looking to represent prairie dynamics within the ED model to begin to tease out these interesting relationships. She will also use ED to project prairie dynamics through time and at a regional scale.

In addition to her work on woody biofuels, Dan Wang (post-doc) has been studying the yield and ecophysiology of two perenial grasses, switchgrass and Miscanthus. Dan conducted an intensive ecophysiological field study of switchgrass and Miscanthus that investigated the effects of canopy depth, nitrogen fertilization, and within season development on photosynthetic properties and leaf economic traits. This study suggests that while traits vary significantly with depth, the main response to nitrogen was a change in leaf area rather than leaf traits. Dan also used this data to model seasonal GPP, from which she found that GPP was more sensitive to seasonal variation in leaf traits than to their vertical distribution, which has important implications for how regional models represent grass canopies. In another study, Dan has conducted literature meta-analysis of switchgrass which assessed the impacts of climate and management on yield. This analysis suggests that switchgrass polycultures containing legumes can have yields comparable to switchgrass monocultures.

Edaphic variation in community and ecosystem dynamics

It is well known that vegetation patterns often vary predictably across the landscape. However, it is not always clear what processes drive this variation, and how they affect ecological dynamics at both larger and smaller scales.

In one project, we are using the Ecosystem Demography model (ED v2.1) to investigate what factors actually drive variation in dynamics along an environmental gradients. It's not uncommon for a single physical gradient (e.g. elevation) to be closely correlated with numerous other factors (e.g. temperature, precipitiation, soil chemistry and texture, solar radiation, etc.) Models allow us to disentangle these correlations by varying each factor independently. The focus area for this work is central NH, where the ED model is being parameterized using eddy covariance and vegetation data from the Bartlett Experimental Forest and then validated using data from the Hubbard Brook LTER's long-term experimental watersheds.

In another project, we are looking at the spatial heterogeneity in regeneration dynamics in the southeast US. We are measuring the growth, mortality, and resprouting of understory saplings in plots scattered across the landscape stratified by key environmental variables such as topographic soil moisture, elevation, and soil type.

Tree demography, carbon reserves, and climate change thresholds across eastern temperate forests

This project is aimed at understanding regional patterns of recruitment and mortality in the temperate forests of eastern and central U.S., the processes that drive these patterns, and the role these processes play in both forest responses to gradual climate change and the potential for rapid ecosystem shifts. The project grows out of a recent project focused on patterns of tree mortality across the eastern US, which looked at drivers of mortality across scales from regional scale climate to meso-scale air pollutants to landscape scale heterogeneity to stand scale biotic interactions. Our current aim is to relate the observed patterns in demography across these various scales to the variations in non-structural carbohydrate reserves (sugars and starches) that we believe underly the patterns in how trees respond to stress. To assess this we began installing plots across the east temperate region in summer 2009 and hope to study at least a dozen sites in total.

Plot locations for carbon reserves cross-site study
At each site, censuses of tree demography are combined with measurements of plant carbon reserves to better understand the physiological drivers of community dynamics. Field data, literature meta-analysis, and process-based models will be combined to improve predictions of how forests will respond to climate change and to predict the vulnerability of temperate forests to rapid ecosystem change.

Model-data intercomparisons and synthesis

The lab is actively involved in a number of intercomparison projects aimed at evaluating how state-of-the-art models compare with data and with each other. These projects include:

Within the NACP intercomparision Prof. Dietze is leading up two analysis teams. The first, in collaboration with Rodrigo Vargas (UC Berkeley) and Andrew Richardson (Harvard) looks at the spectral characteristics of model error. The second takes a focused look at model/data mismatch for the eastern temperate forest region.

Regeneration Dynamics in Large Forest Gaps

At the finest scale we are studying the dynamics of large treefall gaps through a combination of field experiments and modeling. The Experimental Gaps project consists of a set of 18 replicated gaps of two sizes (40m and 20m diameter) split between the Duke Forest in the piedmont of central North Carolina and the Coweeta LTER in Southern Appalachians of western North Carolina. The study site was established in 2000 and gaps were created in 2002 by mechanically pulling trees till they uprooted or snapped. The project is large collaborative effort among many researchers, with our research focusing on the demography of advanced regeneration, damaged trees, and sprouts, as well as the heterogeneity in understory light conditions.

Data collected from the Experimental Gaps project, as well as from other research in the southeast US, are being used in the development and parameterization of the SLIP (Scalable Landscape Inference and Prediction) forest gap model. Prof. Dietze is one of the primary developers on the SLIP project, which defines the cutting edge in individual-based forest gap modeling. SLIP implements a number of novel spatial algorithms and a fully Bayesian parameterization, which allow for efficient simulation and error propagation. The model currently represents every individual tree, sapling, seedling, and seed on an spatially-explicit, topographically heterogenous landscape as well as the 3D light environment and 2D seed dispersal and seed-bank. The lab is currently using SLIP to assess the role of determinism vs. stochasticity vs. sensitivity to initial conditions in forest gap dynamics. In the future we hope to use SLIP to elucidate the key drivers of forest dynamics, the effects of climate change of future forests. SLIP also serves as a test bed for understanding better what are the key factors that need to be included in scaling up predictions from the landscape to regional and global scales Pilot work has also begun in applying SLIP to other ecosystems, such as the Great Dividing Range in Australia.