SEARCH

SEARCH BY CITATION

Keywords:

  • forest climate sensitivity;
  • regional analysis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

Accurate assessments of forest response to current and future climate and human actions are needed at regional scales. Predicting future impacts on forests will require improved analysis of species-level adaptation, resilience, and vulnerability to mortality. Land system models can be enhanced by creating trait-based groupings of species that better represent climate sensitivity, such as risk of hydraulic failure from drought. This emphasizes the need for more coordinated in situ and remote sensing observations to track changes in ecosystem function, and to improve model inputs, spatio-temporal diagnosis, and predictions of future conditions, including implications of actions to mitigate climate change.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

Climate change has begun to alter the composition and structure of forests worldwide. Drought- and heat-related physiological stress and mortality are documented in temperate, boreal, and tropical ecosystems, highlighting the extent of impacts on forests globally (Adams et al., 2009; Allen et al., 2010; Peng et al., 2011; Peñuelas et al., 2011). The scientific community has made great strides in identifying possible mechanisms of physiological stress, yet we need to better understand how different species and plant communities will be affected by projected climate change and potential feedbacks to climate. This finer resolution of information is essential to inform land management and policy decisions.

In boreal and temperate regions, management and policy decisions, such as where to thin forests to reduce crown fire risk or what species to plant following disturbances, are made at landscape to regional scales, and often before analyses can be conducted to determine if they are ecologically valid. Providing timely information sometimes requires back-of-the-envelope responses (e.g. orders of magnitude estimate of carbon emissions from a wildfire compared to fossil fuel emissions in California). Concurrently, we are continually refining models to improve accuracy and reduce uncertainty for more nuanced prediction of ecosystem responses to climate change in the coming decades.

In these situations, an approach is to provide society with more useful information from the earth system modeling community. It involves analysis of species-level adaptation, resilience, and vulnerability to mortality due to climate change at landscape to regional scales (Fig. 1). Common questions in regional analysis include: Where are tree species most vulnerable to climate change, and how soon will significant impacts be seen? What species are likely to be impacted the most? How does climate change impact ecosystem processes like carbon and water cycling? How do these changes and the resulting feedbacks on climate alter the rate and magnitude of climate change? What mitigation actions can be taken to minimize environmental stress and facilitate migration to a more favorable climate? What are viable management strategies?

image

Figure 1. Conceptual framework for regional land surface modeling.

Download figure to PowerPoint

Species traits and tree size matter

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

There is a range of natural variation in the sensitivity of different tree species and developmental stages to drought, heat, and biotic agents that complicates projections of response to future climate. Typically, biophysical-biogeochemistry models within earth system models (ESMs) use generalized plant functional types (PFTs) to assign physiological and structural parameters for simulation of ecosystem processes, but this doesn't really capture species differences in their sensitivity to climate (Van der Molen et al., 2011). For example, the observed dieback of piñon pine in the Southwestern United States (Breshears et al., 2005) and related changes in ecosystem structure and function would not have been predicted by physiological representation of a single PFT, temperate evergreen needleleaf forests.

The typical PFT classification approach has a limited number of tree classes. For example, the Community Land Model (CLM) has eight tree classes (three boreal, three temperate, and two tropical) for parameterization of physiological and structural traits in global modeling (Oleson et al., 2013). Here, boreal tree classes are evergreen needleaf, deciduous needleaf, and deciduous broadleaf. Temperate classes are evergreen needleaf, evergreen broadleaf, and deciduous broadleaf. Tropical classes are evergreen broadleaf and deciduous broadleaf. Although this is an improvement over more aggregated biome classes, the current approach needs refinement to represent the different suites of traits and adaptive strategies that characterize physiological and demographical responses to climate change for regional modeling.

Physiological and morphological plant traits help explain differences in growth and survival associated with resources, and thus explain species distributions across edaphic, light, water, and nutrient gradients. Traits integrate the ecological and evolutionary history of a species, and represent underlying adaptive mechanisms. Some species are more tolerant of a larger range in climate, while others are more oriented toward competition. Collections of traits at the species level often represent tradeoffs in resource space: the growth rate benefits associated with the ability move water rapidly from roots to leaves, for example, come at the expense of protection against hydraulic failure. This has been explained as a ‘fast-slow’ plant economic spectrum that integrates across plant tissues and explains individual strategies to ecosystem functioning (Reich, 2014). Land system models should be enhanced by creating trait-based groupings of species that better represent climate sensitivity, such as risk of hydraulic failure from drought.

Broad patterns of hydraulic safety margins and general patterns of xylem vulnerability to embolism have been identified, but predictors of species ability to survive in specific environments are still not resolved because there are multiple hydraulic solutions (Meinzer & McCulloh, 2013). A critical knowledge gap is the integrated traits that lead to hydraulic failure and carbon starvation under current and future climate conditions. Coordinated observations across a range of species are needed to fill this gap.

Sub-PFT parameterization is more challenging in some tropical forests because of current limited knowledge of physiological attributes. Many tropical forests experience water limitations during a dry season and it is thought that the most pervasive climate effect on tropical lowland forests is higher temperatures, yet there is a range of views on vulnerability (Corlett, 2011) and trait adaptations. Mortality in wet and seasonally dry tropical forests appears to be initiated most often by short, extreme seasonal droughts (Allen et al., 2010). Co-occurring tropical species seem to have different suites of functional and structural traits for tolerating water stress that are linked to wood density (McCulloh et al., 2012). A few attempts have been made to classify traits into species groupings within dry tropical deciduous forests, but there is no universally accepted suite of traits on which to group these forests into sub-PFTs (Chaturvedi et al., 2011). Wang et al. (2014) indicate a twofold increase in carbon cycle sensitivity to tropical temperature variation over the past 50 years, yet this wasn't captured well by terrestrial biosphere models. Improved parameterization is a logical first step. The high biological variability and limited observations in tropical forests pose significant challenges to predicting responses to climate.

Vulnerability to drought-related mortality and associated carbon and water fluxes vary with tree size due to physiological differences and environment (microclimate, nutrients). For example, Spanish inventory data show that for most of the Mediterranean species, mortality rates were highest for small trees (diameter <20 cm) and decreased rapidly then leveled off as size increased (Ruiz-Benito et al., 2013). Larger trees have greater capacity for water acquisition and storage, and nutrient acquisition (Oren & Sheriff, 1995; Stephenson et al., 2014), however, limits of homeostatic adjustment at maximum height, can lead to vulnerability to climate extremes (Zhang et al., 2009; Manusch et al., 2012). Size and age-related mortality are not included in some land system models, and should be considered (Oleson et al., 2013).

Carbon allocation to resource acquisition relative to structural tissue is an important trait that varies with resource availability and developmental stage (Litton et al., 2007). For example, we found that using dynamic carbon allocation with stand age improved Biome-BGC simulations of productivity (Law et al., 2004). More recently, we modified CLM4 with observations from inventories and intensive plots to allow allocation, physiological parameters (foliage carbon to nitrogen ratios and specific leaf area), mortality rate, and biological nitrogen fixation to vary spatially by sub-PFT within ecoregion (Omernik, 1987). We found a 30% and 50% increase in accuracy of simulated productivity and biomass, respectively, compared with predictions using the default configuration (Hudiburg et al., 2013a,b). Plant allocation strategies have implications for mortality. A challenge is having the data needed to effectively parameterize and/or modify models for prediction across regions.

Regional analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

Model development

Research on drought and temperature effects on forests has evolved from intensive measurements at the leaf, tree, and/or ecosystem level to understand mechanisms of response to climate, such as hydraulic limitations to photosynthesis (Meinzer et al., 2009; Ruehr et al., 2012). FACE experiments investigated interactive effects of climate and elevated CO2 (Norby & Zak, 2011), and observations at flux sites show the response of ecosystem carbon and water processes associated with interannual variability in climate among forest age classes and biomes (Law et al., 1999, 2002). Such observations and experiments are critical for model development. Modeling can identify the location and types of observations and experiments that are needed to fill knowledge gaps and input data gaps.

For projections of climate change influences on species distribution in temperate and boreal forests, a hybrid approach uses inventory data on current species distribution and spatial climate over the past 30 years to develop decision rules for species abundance related to future climate, and applies a process-based growth model to predict reductions in photosynthesis associated with those climate variables to map future abundance (Coops et al., 2009). In contrast, more complex land system models include biogeochemistry, vegetation demography, and disturbances from fire and land management (Fig. 1), and can be coupled with climate models to determine land surface feedbacks to climate (Bonan, 2008). The CLM is being developed further to represent vegetation dynamics resulting from ecological-scale interactions (competition, recruitment, migration) in global predictive modeling (CLM(ED); Fisher et al., 2010).

Moisture and temperature impacts on microbial activity have direct consequences for forest growth through stoichiometric coupling of carbon and nutrient cycles. Nutrient dynamics have been shown to have a first-order influence on climate system feedbacks (Thornton et al., 2009; Zaehle et al., 2010), with forest systems playing a dominant role. The role of phosphorus as a limiting nutrient, especially in tropical forest systems, is being explored in new model developments (Wang et al., 2007), and seems to help explain observed patterns of forest biomass accumulation (Yang et al., 2014).

Model inputs

A range of datasets is required to support model development and applications. Although fine-scale predictive modeling is already under development for temperate, tropical, and arctic ecosystems, sub-PFT parameterization of trait sensitivity to climate will only be feasible where sufficient observations are available. Parameterization across regions with sub-PFT maps would benefit from coordinated physiological and morphological trait measurements to find suites of traits that represent species or groups of species sensitivity to climate extremes.

Other model inputs are downscaled climate projections, remote sensing-based landscape characterization, and soil characteristics. Climate projections at finer spatio-temporal scales should account for orographic effects of mountain ranges and elevated land masses on climate (Abatzoglou & Brown, 2012), and capture climate extremes (e.g. temperature, vapor pressure deficits). The remote sensing land cover input should further resolve forest types or species densities. Gridded soils data are required to parameterize soil hydraulic and thermal properties, and soil carbon and biomass are needed for initial and boundary conditions. There is much room for improvement in the density of soils data.

Model evaluation

Long-term in situ and remote sensing observations are needed to detect and understand impacts of climate change on forests, and evaluate model performance. In tropical forests, the density of in situ observations should increase significantly, and repeated measurements (e.g. biomass, mortality) should be more frequent in forests with rapid growth and turnover rates. Some tropical forests can reach high biomass levels that challenge satellite detection. An approach for calibrating LiDAR measurements to field observations has been proposed to improve accuracy of airborne and space-based LiDAR estimates of tropical forest carbon (Asner & Mascaro, 2014), and Fourier transforms of remote sensing LiDAR vegetation density measurements have been tested for estimating tropical forest biomass (Treuhaft et al., 2010).

Even in temperate forests, a reliance on calibrated and validated remote sensing observations will be critical for spatio-temporal diagnosis of model performance. Approaches have been developed to quantify chlorophyll fluorescence (Frankenberg et al., 2011) and changes in photosynthetic light-use efficiency from space (Hilker et al., 2011), and they could be combined. Satellite remote sensing of photosynthesis and phenology can be calibrated by near-surface digital observations from towers, aircraft, or lightweight unmanned aerial vehicles (Anderson & Gaston, 2013; Richardson et al., 2013).

Evaluating outcomes of potential human actions

In regions affected by land management policies, I suggest a model framework that incorporates region-specific scenarios of proposed or implemented decisions and policies for prediction of how they affect forest processes (e.g. carbon storage, net emissions) and ecosystem services (e.g. water availability) over the next several decades. Demography, physiology, and land management are being integrated in models such as ORCHIDEE (Delbart et al., 2010) and CLM (Fisher et al., 2010; Hudiburg et al., 2013b), but few have used these models in fine-scale regional simulations to evaluate specific land management scenarios and how the scenarios might affect forest ecosystem function and atmospheric emissions (Fig. 1).

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

Regular assessment of trends in species mortality and ecosystem function, such as a regional annual report card, and fine-scale predictive modeling both argue for integrated observation systems to track ecosystem changes over time (Ciais et al., 2014). A grand challenge is fine-scale predictive modeling (e.g. 1 km2 grid) over regions using remote sensing and in situ observations to determine trends in forcing factors, species-level responses, and feedbacks into the future. Modeling can provide feedback to inform decisions on new observations, experimentation, and data synthesis critical for filling knowledge gaps. One important gap is how plant traits integrate to lead to species-level ability to survive in a given environment, urging coordinated physiological and structural observations across a range of species to inform models. Ultimately, the observations and modeling will provide information needed to produce high quality and timely regional assessments of vulnerability and resilience of forests, and outcomes of potential human actions to mitigate climate change.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References

Expression of this opinion benefited from conversations with Drs. Peter Thornton, Mathew Williams, Chris Still, Phil Mote and Rick Meinzer. The work was supported by the USDA NIFA program and the Office of Science (BER), US Department of Energy (DOE grant no. DE-FG02-07ER64361). There is no interest or relationship, financial, or otherwise, that might be perceived as influencing the author's objectivity.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Species traits and tree size matter
  5. Regional analysis
  6. Conclusions
  7. Acknowledgements
  8. References