Climate Change and Managed Ecosystems - Chapter 17 pdf

Challenges in Forest Ecosystems under Climate Change: A Look at the Temperate and Boreal Forests of North America P.Y.. We also concluded that the quantification of carbon-related cos

Challenges in Forest

Ecosystems under

Climate Change:

A Look at the Temperate and Boreal Forests of North America

P.Y Bernier and M.J Apps

CONTENTS

17.1 Introduction 334

17.2 A Short Review of Recent Advances 335

17.2.1 Carbon Budgets and Disturbances 335

17.2.2 Stand- and Tree-Level Processes 336

17.2.3 Landscape-Level Responses 338

17.3 Gaps in Knowledge 339

17.3.1 Propagating Error in Models 339

17.3.2 Interaction between Climate and Disturbance Regimes 341

17.3.3 Impact of Climate Change on Net Forest Growth and Carbon Stocks 342

17.3.4 Carbon Dynamics of Peatlands 343

17.3.5 Verification of Satellite-Based Estimates 344

17.4 Summary and Conclusions 346

Acknowledgments 348

References 348

and their contribution to an altered carbon cycle has been on both sides of the source equation Since the start of the industrial revolution in 1850, it is believedthat conversion of forested areas to areas with lower carbon densities (fields, pastures,and urbanized areas) has contributed to about 36% of the total anthropogenic emis-sions to date (155 Pg C released from deforestation vs 275 Pg C released fromfossil fuel (see Table 5 in Reference 3) Recent estimates still show that nearly 25%

sink-of total annual anthropogenic emissions can be attributed to deforestation (see Table

7 in Reference 3) However, it is also estimated that terrestrial ecosystems have anet yearly uptake of nearly half of total fossil fuel emissions (2.9 ±1.1 Pg C vs 6.3

± 0.4 Pg C; Table 7 in Reference 3), making terrestrial ecosystems a significantcomponent of the global atmospheric CO2 regulation system

In a companion paper,4 we reviewed the mechanisms and uncertainties associatedwith the terrestrial sink, and have concluded that its future is far from certain Aclose watch on how this future unfolds is clearly essential, given the importance ofthis sink We also concluded that the quantification of carbon-related costs andbenefits from sustainable forest management, and of the impact of climate change

on forest carbon sinks and sources requires robust tools for tracking changes incarbon pools, models for predicting their fate, and methods for verifying the esti-mates over large scales This conclusion is, of course, not new, and has prompted asignificant scientific response over the past decade or so, a period during which therehave been dramatic improvements in our ability to quantify the carbon pools offorests and in the scientific understanding of the interaction between climate andforests In particular, advances have been made in the understanding of landscape-and stand-level processes through the implementation of large-scale manipulativeexperiments and of biophysical monitoring programs, as well as the development

of forest-oriented remote sensing All of these advances help strengthen sustainablecarbon management in our forests They also improve our capacity to predict thefate of the large carbon pools of the boreal forest under an uncertain climatic future.Future developments in these tools, refinements of models through the inclusion ofuncertainties, and better integration between scales in modeling and monitoringefforts will continue to contribute to these ends

This chapter is intended to briefly take stock of the recent advances in forestcarbon science, and to identify knowledge gaps, uncertainties, and underlying chal-lenges related to the interaction between forest carbon and climate Since gaps should

be identified with respect to a desired outcome, we define below a set of questionsthat will frame the following review of knowledge and the identification of gaps.These questions are especially policy relevant within the context of the KyotoProtocol and of particular importance to Canada and other forest-rich countrieswhere concerns are emerging with respect to the long-term impact of global changes

on the forest resource These questions are the following: Are current stocks ofcarbon in forests increasing or decreasing? Can we manage the forests to enhancesinks or reduce sources? Will the mechanisms responsible for the present biotic sink

be enhanced, saturate, or reverse sign over time? This chapter focuses particularly

on the temperate and boreal forests of North America Management responses thatpertain to mitigation options are treated in Apps et al.4

17.2 A SHORT REVIEW OF RECENT ADVANCES

The establishment of the Kyoto Protocol in 1997 has provided a strong impetus forimproving our ability to account for past and present changes in forest carbon Sig-nificant advances have been made in this field, leading in Canada to the estimation ofcarbon stocks in the boreal forest (Figure 17.1) and to development of tools to track

or estimate biome-level changes in these carbon stocks.5,6 These developments havethemselves led to the realization that, for the boreal forest, the natural disturbanceregime — that is, the frequency, size, and severity of natural disturbances —- largelycontrols the inter-annual to inter-decadal changes in the carbon balance of the borealforests Disturbances provide fast pathways for direct release to the atmosphere ofcarbon previously stored in the various organic components of the ecosystem, but alsoreset the clock with respect to the carbon uptake capacity of forest ecosystems The

FIGURE 17.1 Estimates of carbon stocks in Canada’s boreal forests showing the importance

of peatlands in the total carbon content of the forest (Estimates are from Apps et al 7 for peat, Apps et al 8 for forest products, and Kurz and Apps 5 for biomass and dead organic matter carbon.)

large amount of decomposing organic debris left behind by disturbances also givesrise to delayed emissions over a range of timescales Disturbance dynamics, and fire

in particular, have therefore been the focus of much recent research

Key progress has been made in the analysis of fire statistics,9 in the construction

of historical fire databases from archives and from field observations,10 and in thequantification of direct fire-related carbon emissions.11 Of particular importance from

a policy perspective is the scale of the inter-annual variability in area burned and inassociated greenhouse gas emissions For example, total area burned in Canada was0.3 × 106 ha in 1978 and 7.5 × 106 ha in 1989.11 This large inter-annual variability

in burned area is caused by nonlinearities in processes driving the generation oflarge wildfires, as most fire activity takes place during a few days with extreme fireweather.12 Small changes in climatic conditions can therefore generate large changes

in the fire regime Direct emissions from fires in Canadian forests were estimated

to be 27 ± 6 Tg C yr–1 for average 5 years, or equivalent to 18% of Canada’s totalanthropogenic emissions,11 but reached 115 Tg C yr–1 in high-fire years Increasedfire frequency can therefore provide strong positive feedback to climate changethrough increased release of CO2 into the atmosphere

Changes in the disturbance regime can also generate change in forest tion at the stand and landscape levels, an indirect climate change effect that could

composi-be more important to species distribution, migration, and extinction than climatechange per se.13 Estimation of the Fire Weather Index based on the Canadian GlobalCirculation Model (CGCM) and a 2 × CO2 scenario suggests an increase in firefrequency by 20% or more in most of the central and western boreal forest, but anabsence of change or a decrease in fire frequency in eastern Canada, where changes

in the precipitation regime and timing of the seasonal warming interact in a differentway.14 Analysis using a 3 × CO2 scenario and the CGCM suggests a 76% averageincrease in area burned across Canada.15

Gains in knowledge have also been made with respect to the relationshipsbetween climate and insect-related disturbances and to interactions among distur-bance types For example, knowledge on the historical climate dependence of moun-

tain pine beetle (Dendroctonus ponderosae Hopkins16) has now been coupled withnew climate interpolation methods and climate scenarios to follow the current spread

of the insect17 and predict its possible future expansion Similar work is under way

for spruce budworm (Choristoneura fumiferana Clem.), and models suggest possible

expansion of outbreaks with a warming climate.18,19 We are also slowly learning how

to tackle the study of interactions among different types of disturbances For ple, Fleming et al.20 have succeeded in quantifying the probability of fire followingepidemic outbreaks of spruce budworm in different parts of the Province of Ontario,Canada The spectral analysis used to obtain this information required detailedtemporal and spatial observations on these disturbances, highlighting the need tomaintain and enhance high-quality disturbance-related data sets

exam-17.2.2 S TAND - AND T REE -L EVEL P ROCESSES

Photosynthesis and respiration are closely coupled to environmental conditions.Long-term changes in the environment must therefore be included in predictions of

future forest carbon stocks In the past, many studies on the interaction betweentrees and their environment were carried out in totally or partially artificial environ-ments, and often only on seedlings or saplings (e.g., review by Wullschleger et al.21).Research emphasis has now moved to observational or manipulative studies at thestand level in natural settings In the boreal forests of Canada and in other forestsaround the world, significant knowledge has been gained in the past 10 years aboutprocesses at the ecosystem level Particular emphasis in this chapter is placed oneddy flux covariance measurements, stand-level manipulative experiments, and keytransect and laboratory studies.

Flux towers were first installed in forest ecosystems in the early 1990s InCanada, this deployment was done as part of the BOREAS study.22 A decade later,many of the original BOREAS sites are still being monitored as part of the BERMSproject (http://berms.ccrp.ec.gc.ca/), and the network has recently expanded as Flux-net-Canada to cover a variety of forest ecosystems and disturbances(http://www.fluxnet-canada.ca/) Two of the key questions being addressed by Flux-net-Canada are (1) how are disturbances contributing to the carbon budget of Can-ada’s forest, and (2) how sensitive is this budget to the variability in climate Results

to date show that spring temperature and summer drought are key elements of theinter-annual variation in net ecosystem productivity.23–25 Observational bounds arealso being put on post-fire changes in ecosystem carbon stocks (the delayed emis-sions and regrowth).26 These experiments and others are shedding light on themechanisms by which climate change will influence forest ecosystems, and areproviding crucial guidance as well as data for model development

An important body of knowledge is also currently being developed on theecosystem-level impact of elevated CO2 through a handful of in situ stand-level CO2

fumigation experiments located in North America and Europe The most importantfinding to date is that increasing atmospheric CO2 concentrations by 200 ppmenhances growth and net carbon accumulation through increased photosynthesis innearly all systems tested27–32 without apparent saturation of effect after 3 to 6 years

of fumigation Figure 17.2 shows an example of results obtained with loblolly pineafter 3 years of treatment The effect of increased CO2 on growth has been reducedonly where important drought limitations exist.33

The experiments are also helping to determine which other processes within theforest carbon cycle will be affected directly or indirectly by elevated CO2 Examples

of states or processes that appear unaffected by elevated CO2 in the systems testedare bud phenology,34 fine root turnover,35 and leaf area index.28,30 Examples ofprocesses that are affected by elevated CO2 are stem respiration,36 soil respiration,37

and shoot elongation.31 Further manipulations in one particular experiment have alsoshown that additional fumigation with ozone (O3) eliminates the CO2-induced gains

in growth.32 The ozone fumigation also interacts with CO2 to alter the performance

of tent caterpillars.38

Finally, a number of other experiments are providing information on key tainties in the carbon dynamics of the boreal forest For example, soil warmingexperiments in mature coniferous and mixed forests have produced increased storage

uncer-of carbon in biomass and limited loss uncer-of soil organic matter.39,40 Similar conclusionswith respect to the general impact of warming on soil carbon storage have been

reached in studies of climatic transects in Canada.41 Current ongoing research onsoil respiration processes associated with the Fluxnet-Canada network as well as asecond soil warming experiment just starting in northern Manitoba should providefurther insight into these critical components of climate change response.

Over the past 10 years, our ability to interpret and use multispectral signals fromsatellite-borne sensors has increased dramatically Although many problems linked

to the application of remote sensing to operational forestry still need to beaddressed,42 remote sensing technologies are making significant contributions toclimate change-related applications Currently, satellite-based observations are used

to provide regional or global coverage of forest properties such as land cover,43 leafarea index,44–46 and absorbed photosynthetically active radiation.44,47 They are alsoincreasingly used to monitor the progress of dynamic processes such as distur-bances,48,49 phenology,50 stress response,51 and, indirectly, photosynthesis and net

FIGURE 17.2 Cumulative total dry mass (DM) accumulation of loblolly pine stands as a

function of absorbed photosynthetically active radiation over a 3-year fumigation period

in a FACE study The open circles represent the mean of three plots maintained at 200 ppm above ambient CO2 concentration, while the closed circles are the mean values of the three

control plots (From DeLucia, E.H et al., Tree Physiol., 22, 1003–1010, 2002 With

primary productivity.52,53 One of the most notable drivers for advancement in thisarea has been the recent deployment of the MODIS sensors on board the Terra andAqua satellites These sensors now provide near-daily global coverage in spectralbands selected to measure biological and physical processes on land and in oceans.

As a complement to direct field measurements, remote sensing offers two attractiveadvantages: it can cover large areas, and it can do so repeatedly These two strengths

of remote sensing products can be combined to provide retrospective analysis ofvegetation dynamics One important outcome of such research has been to showsignificant changes in timing and extent of greening at continental scales in recentdecades, suggesting an observable response to recent climatic trends.54,55 Remotesensing outputs from the MODIS sensor now feed into processes for the production

of near-real-time estimates of photosynthesis and net primary productivity.56 Thechallenge in this field now lies in the validation of such products (see below)

17.3 GAPS IN KNOWLEDGE

We posed at the outset three questions that relate to the current state of our forests,their long-term response, and the management options for adaptation or mitigationactions Currently, key areas exist in which progress is needed to better guide forestcarbon science and policy and address these questions Much of the list of researchtopics proposed by Woodwell et al.57 in a comprehensive review on biotic feedbacks

to the carbon cycle remains pertinent today, although much knowledge has beengained since then Here, in light of current policy needs, we suggest where the mostsignificant gains can be made to improve our ability to assess and monitor climatechange impacts on the forest resource, and to evaluate the impact of our actions onthese changes

A fundamental problem that is seldom addressed in large-scale biological pursuits

is the absence of estimation and propagation of errors in models At the experimentallevel, we are required by peer review to quantify the uncertainty around estimates

so that experimental outcomes can be declared statistically significant or not ever, the same requirement is usually absent from higher-level studies Estimatingand propagating errors through integrative procedures is complex but not alwaysimpossible.58 Such an exercise provides a number of significant benefits

How-As a first advantage, quantification of uncertainty in the estimates of lower-orderprocesses permits the intercomparison of these processes and the identification ofthose on which resources should be spent A second advantage is that the estimationand propagation of uncertainty in models should make it possible to identify pro-cesses that can be left out because their contribution is masked by the overall modelerror This is particularly important in scaling-up exercises where temptation is great

to include detailed processes — at high data and computational costs — becausethey are known or thought to be important at finer scales At coarse scales, gainsmade by propagating fine-scale elements such as canopy structure or the use of short

(e.g., hourly) time steps, are likely lost in the noise due to fine-scale variability orerrors in model inputs (e.g., maps of leaf area index or interpolated rainfall).The third benefit of error propagation is that such analyses carry forward to thedecision maker a significant quantity of information that is lost if only mean valuesare reported Whereas a mean difference may elicit a particular response fromdecision makers, the same difference with a confidence interval that includes 0 maywell generate a totally different one Figure 17.3 provides an example of a simulationresult on tree growth following thinning, in which the 95% confidence intervalreaches the “no-effect” level far before its mean.59 Such an outcome can be used toestimate when the effect ceases to be significant It can also be used in a riskmanagement sense to determine the probability of being wrong if a decision is madebased on the presumed presence of a difference Decision makers may tolerate risksfar above the 1 and the 5% confidence levels generally used by the scientificcommunity, and thus would truly benefit from the presentation of uncertainties.Methodologies to quantify and propagate uncertainties in models exist but areoften overlooked because of the effort (and expense) involved in their implementa-tion Variance of soil carbon within and among plots, variance of the errors in thecomparison of estimated to measured tree volume increments in permanent sampleplots, and interval of confidence in modeled allometric equations are all examples

of uncertainties that are within the reach of the analyst Uncertainties such as futureclimate conditions or disturbance regimes that are more difficult to quantify can betreated through the use of scenario modeling In the final analysis, trading off the

FIGURE 17.3 Predicted evolution of a treatment effect (commercial thinning) on

merchant-able volume as a difference from a control, with the 95% confidence interval (From Raulier,

F et al., Can J For Res., 33, 509–520, 2003 With permission.)

possibility to propagate errors for increased model complexity may well be thewrong choice We therefore recommend pursuing the inclusion of model uncertain-ties in all analytical efforts related to climate change and forest carbon.

R EGIMES

Global circulation models (GCMs) predict climate anomalies in response to assumed

CO2 emission scenarios Their predictions therefore incorporate both the ties in the model representation of the climate system and the uncertainties related

uncertain-to the CO2 emission scenarios Regional climate scenarios are then down-scaledfrom GCM outputs with some additional errors injected in these finer estimates.Larger errors are still likely to be incurred, however, when these regional climatescenarios are used to forecast future disturbance regimes Relationships betweenenvironmental variables and drivers of disturbances are often highly nonlinear, such

as those captured in the different components of the Canadian Forest Fire WeatherIndex,60 or in the proposed climate control on mountain pine beetle populations.17

Nonlinear relationships with disturbances amplify errors present in climate estimates.Finally, additional biotic interactions such as those between host and parasite play

a large role in the downregulation of epidemics and are even more difficult toincorporate in models, even empirically, simply because of the paucity of observa-tions that could be used to fit simple statistical models

As mentioned above, we now know that forests are in perpetual adjustment toshifts in the disturbance regimes, and that it is these regimes that control to a largeextent the age-class distribution of the forest and hence the carbon storage within

In view of the changing environment, predicting with some certainty how replacing disturbances will change under new climatic conditions is a high priority.The benefits of improving the representation of climate–disturbance interactions, oreven including their uncertainty, would be substantial for all the reasons cited abovefor the benefits related to the propagation of uncertainties in models

stand-Historical studies provide key data and insights needed for the development

of predictive models Fire histories can be mapped and dated to an extent andaccuracy that decrease with the remoteness of the site and the time since fire.Nevertheless, considerable progress has been made in the mapping of past fires.61

Current databases are also capturing recent and current fires using remote sensing62

and other methods, thus adding to an ever-increasing database of large fires in theboreal forest.10 For insects, tree ring-based methods are being developed that can

be used to reconstruct past insect outbreaks and their effects on forest growth.61,63,64

Statistical techniques are also helping to unravel the relationship among differenttypes of disturbances.20 All these methods are slowly helping to build the types

of databases that can be used to tackle the climate–disturbance relationships andestimate the first-order uncertainties in our models In light of the importance ofthis topic, we recommend pursuing the application of these methodologies to awider variety of landscapes, forest types, and environments in order to provide abroader data domain for models

17.3.3 I MPACT OF C LIMATE C HANGE ON N ET F OREST G ROWTH

AND C ARBON S TOCKS

Across vast landscapes, disturbances are dynamically counterbalanced by the netgrowth of forests This net growth is made up of tree-level and stand-level processesthat control gross carbon uptake, respiration losses, shedding of tree parts, mortality

of individual trees, net accumulation of organic material on the forest floor andwithin the mineral floor, and loss of dissolved organic and inorganic carbon throughleaching Our ability to predict how the net growth and the attending carbon seques-tration in forest stands will behave in the future depends on knowledge of how theseindividual processes interact with environmental variables related to climate oratmospheric properties such as its CO2 and ozone concentrations Significantprogress has been made recently with flux tower data, FACE studies, and other suchexperiments, but significant gaps and uncertainties exist to this day in this area.Three examples of processes in which large uncertainties exist are (1) long-termresponses to environmental changes (climate, CO2, and other atmospheric constitu-ents such as ozone), (2) autotrophic respiration losses, and (3) within-tree carbonallocation

We have just begun to study stand-level responses to increased CO2 andozone.30,32 Long-term effects in interaction with other limiting factors will remainuncertain and hypothetical for some years to come, but bounds of uncertainty must

be quantified Climate change will also cause significant decoupling between speciesdistributions and optimum growth ranges with unknown consequences to forestgrowth, in addition to changing the vulnerability of forests to specific natural dis-turbances.13 Clearly, long-term assessment of growth impact and of vulnerability todisturbances still requires considerable work

Significant research is carried out on photosynthesis and its acclimation totemperature and increased CO2 About 50% of the carbon captured by photosynthesis

is respired back to the atmosphere by the plant when the photosynthates are usedfor growth and maintenance of vegetation structures Yet, in spite of this large loss,relatively little is known about the mechanisms that control this autotrophic respi-ration, or the relationship between it and the gross productivity of trees As anexample, in the boreal forests of Canada, there has been only a handful of studiesdedicated to the quantification of total respiratory losses from trees.65–68 As a result,present estimates of percent losses to autotrophic respiration for black spruce, aspen,

and jack pine (Picea mariana, Populus tremuloides, and Pinus banksiana) vary

between 35 and 54% of gross photosynthetic uptake In many models, a fixed value

in the growing environment of the trees The largest gap in this area is certainly theallocation to belowground processes that involve, in particular, fine roots, root