Literature Review on Innovative Silviculture and Management Practices Supportive of Conservation Values

March  2009

Mark Swanson
Hiroo Imaki
and
Bruce Lippke

Working Paper 10  Rural Technology Initiative  College of Forest Resources  University of Washington  Box 352100  Seattle, WA  98195-2100  www.ruraltech.org

Olympic Natural Resources Center  (ONRC)  College of Forest Resources  University of Washington  PO Box 1628  Forks, WA  98331  www.onrc.washington.edu

Click here to view the PDF version of Working Paper 10

Acknowledgements

This report was made possible by a grant from the Olympic Natural Resources Center with financial support from the U.S.F.S. Pacific Northwest Experiment Station. An early version of the literature review was funded by the Washington State Department of Natural Resources.  The findings in the report hopefully mirror the results of the many pioneering research studies referenced in the literature review.  Any opinions, findings, conclusions or recommendations in this report are those of the authors and do not necessarily reflect the views of contributing entities.

Abstract

In this literature review we identified over 700 articles following the Washington Forest Landscape Management Project, (WFLMP), which provided one of the early attempts to understand the implications of innovative silviculture and landscape management on conservation values.  We enumerate many of the alternative treatments and important elements starting with the WFLMP while noting the several research pathways that developed from that work.  We do not attempt to highlight the impact of each referenced article but follow the several directions research has taken and rely on the emphasis provided through several publications and conferences that have attempted to provide a comprehensive summary and review of the literature. 

Keywords: Landscape management; biodiversity; habitat suitability; wildlife models; innovative silviculture; spatial forest planning; endangered species protection; harvest scheduling; optimization; meta-heuristic; spatial and temporal landscape patterns; spatial explicit wildlife model; multi-species approach; GIS.

Table of Contents

Background and Introduction

In June 1988, with intensifying public debate about forest management and ecosystem protection, Brian Boyle, Commissioner of Public Lands, established the Commission on Old Growth Alternatives for Washington’s Forest Trust Lands.  The Commission brought together 32 diverse “stakeholder group” representatives.  Their mission was to develop consensus recommendations for the Washington Department of Natural Resources (DNR) on how to balance the goals of providing revenue for education, protecting the biological diversity of the forest environment, and supporting local timber-dependent communities.  Commission recommendations included the establishment of the 264,000-acre Olympic Experimental State Forest (OESF) on the western portion of the Olympic Peninsula as a commercial forest within which there would be a special opportunity to research harvest and regeneration methods to enhance habitat characteristics and commodities production.  To focus and coordinate the long-term research program for the OESF, the Commission recommended creation of the Olympic Natural Resources Center, or ONRC (Commission on Old Growth Alternatives 1989).  These investments in experimental research capacity reflected a broader trend in the field of forest management, in which the desirability of maintaining the inherent complexity and array of functions of natural forest ecosystems is recognized.  Prominent Washington forest scientists offered papers suggesting that modifications to commercial forest management could achieve ecological benefits and accelerate development of desired old forest habitat structures while generating forest products and economic returns (Franklin 1989, Oliver 1992). 

In 1993, the Washington Forest Landscape Management Project (WFLMP), funded by the U.S. Congress and facilitated by the DNR (Carey et al. 1996), created for the first time a broad multi-disciplinary research group with the objective of determining the advantages of simultaneously managing forests for multiple species of threatened, endangered, and sensitive wildlife over large landscapes and across different ownerships including economic considerations. It was hoped that innovative approaches to forest management could integrate species habitat enhancement with harvest of forest resources to maximize environmental benefits and reduce the costs of species conservation on society.  Many ideas about the impacts of changing forest structures on species conservation were discussed and evaluated by the study team. The WFLMP provided the first interdisciplinary opportunity to evaluate the potential costs and benefits of intentionally managing forest stands to restore old forest functionality following harvests and other anthropogenic disturbance. The management treatment strategies developed by the WFLMP scientific team became known as biodiversity pathways, or ‘biopathways’, designed to intentionally produce old forest habitat conditions through active management at a much lower cost than no-action or passive alternatives.  From a literature review perspective, the project considered 150 then-current publications such that projects after 1996 generally had the benefit of the integrative work provided by the project.  The project produced many publications (e.g., Carey, Eliot et al. 1996, Carey, Thysell et al.1996, 1999; Lippke et al. 1996) and began a decade of additional scientific research into the effectiveness of biopathway treatment regimes as compared to no-harvest reserves and commercial or longer than commercial rotations.

In this literature review we identified over 700 articles following the WFLMP with implications for innovative silviculture and landscape management.  We enumerate many of the alternative treatments and important elements starting with the WFLMP while noting the several research pathways that developed from that work.  We do not attempt to highlight the impact of each referenced article, as this approach would be too long for a useable synthesis. Rather, we follow the several directions silvicultural and forest landscape research has taken, and we rely on the emphasis provided through several publications and conferences that have attempted to provide a comprehensive summary and review of the literature.

Biodiversity Management and Protection Pathways

The distinguishing features of “biodiversity pathways” developed in the WFLMP included retention of biological legacies at harvest, plantation and natural regeneration for mixed species forests, pre-commercial thinning to quickly bypass the stage of competitive exclusion in dense young stands, thinning at variable densities in maturing stands, extending final harvest rotation lengths through multiple entry thinning, and creation of snags and downed woody debris.  The range of options examined has been largely oriented toward thinning earlier or later while retaining less or more trees over time with consideration for retention of legacy trees, stumps and debris.  Riparian buffer management incorporated similar thinning treatments with extra attention to stream bank stability, sufficient vegetative and overstory retention for shade, and recruitment of large woody debris (LWD).  Direct measures of forest ecosystem health were largely related to the capacity to support vertebrate diversity, forest floor function for mammals, ecological productivity based on the abundance of arboreal rodents as a critical element of the food chain and production of large vertebrates. These functions/proxies were developed as indices to estimate the ecological impacts of alternative management approaches. 

The biodiversity pathways approach emphasizes changes in forest structure based on growth model simulations, which were analyzed statistically and spatially. The structural composition changes were shown to restore forest stand structure distributions more like those present before European settlement, albeit at a cost to landowners.

The spatial patterns were less revealing since the size of treatment areas was dictated by prior treatments and stand boundaries and regulatory or operational limits on the treatment scale. While biodiversity pathways on the entire landscape produced what were considered to be better spatial distributions for late seral forests, there were few interior species that would benefit. The spotted owl was believed to be more negatively affected by aggregated cuttings than dispersed (Carey et al. 1992, Carey and Peeler 1995). Other studies found no effects from interior forest management alterations and further concluded that no species of concern were “associated” with (benefited from) “competitive exclusion” of dense closed-canopy structures (Hansen et al. 1993). The conclusion that no species benefited from competitive exclusion structure, with most species negatively affected, has provided strong motivation for biopathway thinning in forests with a legacy of commercial management. The dominant forest structure class associated with commercial rotations (or no-action alternatives in previously-harvested forests) following regeneration strategies is competitive exclusion.  This forest condition can potentially last for much more than a century following full stocking regeneration if stem densities are not reduced by thinning treatments or disturbance.  As a result, many studies have emphasized the need to reduce stem densities and avoid or reduce the prevalence of closed-canopy competitive exclusion as important to achievement of biodiversity goals and old forest habitat objectives (Oliver 1992, Poage 2000, Poage and Tappeiner 2002, Acker et al. 1998, Agee 1991, Carey et al. 1999, Hayes et al. 1997, Garman et al. 2003, Tappeiner et al. 1997, Bailey and Tappeiner 1998). 

Functional equivalents of the biopathways approach have been developed for forests in other regions, such as the broadleaf forests of New England.  At a spatial scale of the individual stand, the use of rotated-sigmoid diameter distributions in combination with specific strategies to enhance structural richness (e.g., snag creation, crown release of vigorous trees, etc.) can further contribute to these objectives (Keeton 2006, Kenefic and Nyland 2000).  Lindenmayer and McCarthy (2002) propose similar alterations to intensive forest management in southeastern Australia to conserve fauna dependent on old, hollow trees and intact forest landscapes.  Application of these recommendations in landscapes will validate degrees of effectiveness over the long time scales necessary to conduct research in forested landscapes.        

Measuring the Biological and Economic Success of Pathways

Development of quantifiable and statistically robust measures of ecosystem health with which to assess the comparative success of alternative management treatments has been a challenge for researchers. Early studies were constrained by limited empirical stand data and with limited statistical analysis enriched by expert opinion. Gehringer (2006) developed a non-parametric statistical target of old forest conditions by analyzing actual old and previously not harvested forest inventory data and determining the attributes that most reliably discriminate those stands from all others.  Following this analysis Gehringer was able to create management targets to test for development of old forest characteristics within a confidence interval representative of an old forest range of variability for Western Washington.  Gehringer found that trees per acre, diameter and height were the dominant metrics of interest that differentiated old forest characteristics. When simulated treatments and growth for forest stands are modeled forward in time then the point at which stand characteristics resemble those of the target older forests can be identified.  Two important performance metrics for assessing comparative old forest treatment options emerge: which treatments create desired conditions soonest and which treatment regimes most successfully extend the duration of time in the old forest target.  Other variables thought to be important to habitat such as coarse woody debris (CWD) were characterized by such large variance that they did not provide a discriminating function for old forests. However, CWD may be a discriminator of better habitat for some critical species (Bunnell et al. 1999), and can be used as a fine filter discriminator to identify selected species habitats within the broader class of old forests (Zobrist and Gehringer 2004).  Other metrics include an index of old growth (Acker et al. 1998), which has been used to quantify the rate of structural development in unmanaged Douglas-fir stands (Larson et al. 2008).  Stand-level diameter distributions are also useful for quantifying differences between structurally simple young stands and complex, multi-cohort late-successional stands (Zenner 2005), especially under chronic disturbance regimes (D’Amato et al. 2008).  

Simulated treatment activities such as thinning harvests can be used to forecast potential log yields and economic returns.  Old forest structure targets and simulated economic returns provide a means for identifying least-cost pathways for achievement of biological forest targets.  Similar methodologies for statistical discrimination between forest conditions have been explored by Lundquist (Lundquist & Beatty 1998, Lundquist & Lidner 2000).

et al. (2005a) used the Gehringer procedure to determine the most economic pathways to reach old forest structure targets.  The method was to inform forest management options within riparian zones in the PNW Douglas-fir zone as well as to aid development of legacy structures in the U.S. Southeast, a project supported by the National Commission on Science for Sustainable Forestry, NCSSF (Zobrist et al. 2005b).  A similar approach was used to investigate management options for old forests and spotted owl habitat on the DNR Olympic Experimental State Forest (Lippke, et al. 2007).  The performance metrics for a zone-specific target of old forests with their discriminating structural differences were used in all these studies to identify which treatments most successfully reach the structural target objective at the lowest economic costs.

Habitat Suitability Models

Other forest structure research efforts have targeted the development of various habitat suitability measures for individual species, which are logically more sensitive to detailed segmentation of forest structure classes. The U.S. Fish and Wildlife Service developed a habitat evaluation procedure (HEP) (USDI 1980) based on characterizing the forest in a number of discrete conditions in order to provide the inputs for habitat suitability models.  These methods were automated for a study in Western Washington that used a stand-level simulation model to conduct habitat suitability tests, demonstrating repeatability in evaluation methods (Ceder et al. 2002a, Marzluff et al. 2002). 

Johnson and O’Neil (2001) developed a wildlife-habitat model that covers entire vertebrate communities in Washington and Oregon. This habitat model recognizes three factors: wildlife habitats = wildlife cover type(s) plus structural condition(s) plus habitat element(s).  Wildlife cover types refer to vegetation classifications that were determined based on similarity of wildlife use.  Structural conditions describe the forest structure at the stand level. There are a total of 26 structural conditions defined by tree size, percent canopy cover, and the number of canopy layers. Finer scale habitat features are habitat elements such as downed logs and snags. The resulting Johnson and O’Neil wildlife-habitat matrix associates species types with 26 habitat structural conditions further defined by three habitat qualities: closely-associated, generally-associated, and present. Although habitat elements are considered important components in habitat models, elements are too variable and species-specific to generalize in the Johnson and O’Neil’s wildlife-habitat model although potentially important at the site or forest type level.

Habitat suitability measures are useful tools for analysis but currently lack sufficient data to be as statistically robust as the Gehringer approach.  As more data is collected on species-specific habitats, statistical links between habitat suitability and forest structure class will logically improve.  However, another problem with classification systems is that habitat definitions tend to be limited to “in or out” rather than probabilistic gradients and consequently can exhibit knife-edge properties relative to management simulations obscuring performance at the margin (University of Washington College of Forest Resources 2007). 

Models integrating a non-binary perspective on habitat utilization are more likely to offer insights into landscape and stand influence on vertebrates (Marzluff et al. 2004, Folliard et al. 2000).  The former focus in landscape ecology on the patch-mosaic model of landscape structure is being enriched by an understanding that many phenomena occur as gradients in landscapes (McGarigal and Cushman 2005).  This is reflected by the use of logistic regression models parameterized at multiple scales to assess the importance of landscape context (including the abundance of various structural classes of forest) in determining organism abundance or viability (e.g., White et al. 2005, Daw and DeStefano 2001).  These models can then be used to create maps of probability gradients, assess actual landscapes for areas of high-quality habitat, or plan restoration treatments that will create appropriate combinations of landscape elements.

Variable Densities and Variable Retention

Research on the ecological impacts of biopathways silvicultural interventions is still in formative stages.  Individual tree growth in variable density units tends to vary as a function of proximity to gaps and whether the tree is in the thinned matrix of the unit (Roberts and Harrington 2008). While variable densities were considered important in the early biodiversity pathway work (Carey et al. 1996), natural variation as well as variation induced from equipment impacts made it difficult to establish useful measures of density within otherwise uniform stand structures.    Thinning clearly favors the development of shade-tolerant trees in the understory of stands in the competitive exclusion/biomass accumulation stages (Chan et al. 2006).  Underplanting may have to be employed to accelerate this process, especially when seed sources for shade-tolerant conifers are distant (Beach and Halpern 2001).  A meta-analysis of thinning impacts on understory vegetation (Wilson and Puettmann 2007) found that variable-density thinning increased spatial variability in understory communities, but effects on richness varied from case study to case study, often as a function of past site management.  Concerns regarding potential for wind damage related to larger openings associated with variable-density thinning were allayed by Roberts et al. (2007), who found no difference between unthinned controls and treatment plots.  Certain taxa may benefit from variable-density thinning, such as litter invertebrates (Schowalter et al. 2003).   Many of these results await confirmation over longer time scales and application in other forest types, but the results of current research favor the concept that spatially variable thinning is an effective, ecologically-based method for enhancing complexity and diversity in young forest stands. 

Disturbance in forest ecosystems, by definition, kill or remove trees and other organisms (Pickett and White 1985).  Most natural disturbance processes, however, tend to leave a variety of living or dead structures as biological legacies (Franklin et al. 2000).  These legacies promote ecosystem recovery (e.g., Keeton 2000) and the persistence of a wide variety of organisms at multiple spatial scales.  It has been proposed that retaining structures at harvest can capture some of the functional benefits of biological legacies within the context of the managed forest environment (Franklin et al. 1997, Mitchell and Beese 2002).  Variable retention has been shown to be effective for the conservation of understory plant communities (Nelson and Halpern 2005), water quality (MacDonald et al. 2003), lichen communities (Esseen et al. 1996), and other values.   Ectomycorrhizal fungi have been found to be more abundant in the vicinity of retained individual trees (Cline et al. 2005), with proximity to patches is relatively more important than the size of the patch itself (Jones et al. 2008).  A meta-analysis of research on green tree retention (variable retention) demonstrated the utility of the approach for maintaining organisms from several important functional groups, such as dispersal-limited lichens, birds, and ectomycorrhizal fungi (Rosenvald and Lohmus 2008). 

Survival and growth of retained trees is a major concern when applying variable retention harvest strategies. Survival is often influenced by tree wounding during the harvest or proximity to skid trails (Thorpe et al. 2008), and wind firmness is a major consideration when selecting trees for retention.  Healthy, uninjured trees (even in older age classes), however, typically respond well to the release provided by a variable retention harvest event (Latham and Tappeiner 2002, Bebber et al. 2004).  Experience in wet sclerophyll forests of Tasmania demonstrates that variable retention may aid in regeneration of species-rich forest communities by reducing dispersal distances (Tabor et al. 2007), and that while windthrow may damage retained trees, old-growth individuals tend to be relatively windfirm (Neyland 2004).  Additionally, the retention of undisturbed forest aggregates in Tasmanian forest has been shown to favor amphibians (Lauck et al. 2008) and beetles (Baker et al. 2007). 

Experimental On-the-Ground Treatments

Most early work on innovative silviculture to promote conservation values relied on simulations as the historical data on past treatments was generally lacking. This prompted a series of experimental studies, collectively known as the Capitol Forest Study, to apply different treatments on the ground in sample plots to estimate cost and environmental benefit over time (Curtis et al. 2004).  Located on public land in Washington State, this is an exemplary experimental effort aimed at clarifying the impacts of various silvicultural treatments on the structure and composition of Douglas-fir stands in the competitive exclusion stage.  These studies, unlike many from the ecological literature, provide useful operational cost information; however, the impacts on ecological measures are only now beginning to have enough maturation time to be indicative of the future.  Future monitoring will help to determine comparative environmental effectiveness. 

Silvicultural research to assess the efficacy of different timber harvest regimes in maintaining habitat values or ecosystem processes have shown the value of various modifications of traditional timber harvesting.  The DEMO study (Aubry et al. 2004) has produced a number of studies on stand-level response to various forms of variable retention, some of which have been discussed in the section on variable retention.    

Silvicultural experimentation has not been limited to the creation of later-successional structural conditions, however.  An excellent example may be found in the oak forests of the Puget Sound region in western Washington, where partial and full crown release prescriptions have been shown to be effective in restoring vigor to overgrown Oregon white oak (Devine and Harrington 2006).   

At Fort Lewis in Western Washington, Churchill (2005) showed that on dry-sites Douglas-fir can grow in the understory and maintain its release potential for at least 20 years under moderate overstory stocking levels (30-55% full stocking). His results indicate that by combining elements of shelterwood, group selection, and single tree selection systems, multi-cohort, structurally complex stands can be created and maintained in a shifting mosaic of patches while also producing significant wood volume over time. Historic stand conditions cannot be duplicated rapidly following either stand replacement fires or high volume timber harvests.

Dry pine forests present a further set of restoration challenges, since these forest types have experienced exploitative timber extraction, fire suppression, and agricultural conversion. Everett et al. (2007) provide a unique analysis of history based on stand reconstructions over 150 years.  They concluded that in pre-settlement forests landscapes varied greatly with many structural stand types reflecting different stages of recovery from multiple disturbances of varying intensity and frequency but had a well-represented tree-understory with very few dead and down fuels. Following euro settlement, decreased fire effects allowed existing understory to continue development, which increased stand density and the proportion of shade tolerant species.  Under fire suppression, landscape diversity of forest structure has been lost with a decline in early succession stages and an increase in old forest structure (Everett et al. 2008).  Stands are currently transitioning from high post fire suppression tree densities to less dense stands as insect and pathogens thin stands from above and below.  This stand thinning process has increased amounts of dead and down wood, previously maintained at low levels by frequent surface fires.  Although the dry forest landscape has always operated under a mixed fire regime, the proportion of landscape subject to high severity fires has increased over the last several decades reflecting unsustainable conditions (see also Hessburg et al. 2005). Both forest structure and disturbance regimes are evolving. Restored landscape and forest structures will need to be sustainable under disturbance regimes that differ significantly from euro settlement conditions.  Restoration objectives that include thinning and prescribed fire to emulate the inherent disturbance regimes of the area will also need to reflect post settlement socio-economic expectations. 

Research in the forests of the Southeast has shown that it is possible to restore old-growth longleaf and loblolly stands through modified group selection and single-tree selection, the application of prescribed fire, and the introduction of spatial heterogeneity (Bragg et al. 2008, Brockway et al. 2002).  Experimental silviculture in the Ouachita Mountains of Arkansas has focused on restoring shortleaf pine stands formerly maintained by fire (Guldin 2004).  Research in ponderosa pine ecosystems in the Black Hills (Shepperd and Battaglia 2002), the interior Pacific Northwest (Youngblood et al. 2006), and the Southwest (Allen et al. 2002) indicate that restoration of coarse woody debris, retention of large-diameter pines, reintroduction of fire, and creation of spatial complexity (gaps, clusters of trees, and randomness at various spatial scales) are important elements that diverge from traditional tenets of silvicultural practice.  Arno and Fiedler (2005) examine long research histories from a number of western forest types where carefully applied silvicultural treatments and the resumption of historic fire regimes lead to a host of ecological, economic, and social benefits.        

Temperate systems from around the world offer similar experiences.  For example, silvicultural trials at Warra, Tasmania, in eucalyptus-dominated stands show benefits to creation of ecological complexity within stands and restoration of elements of natural disturbance regimes (Neyland 2004). 

Stand Variability Over Greater Spatial Scales

Lindenmayer and Franklin (2002) focused on enhancing biodiversity with the perspective that the scale requirements can not be met by partitioning off some lands as reserves while other lands are used for intensive timber production. They characterized a checklist for achievement of forest biodiversity conservation across a landscape based upon a matrix of conditions needed. A number of cases studies were summarized to highlight the contribution of a variety of landscape characteristics.

Important areas: The Lindenmayer and Franklin checklist begins by identifying important areas that need to be protected. Important areas include aquatic systems such as stream networks and wetlands, wildlife corridors, specialized habitats such as cave and thermal protection, biological hotspots, and remnants of late successional and disturbance refugia such as forest areas with no prior history of harvest.   Culturally and socially important areas are identified for protection as well. These areas are mid-spatial scale and are to be given special consideration before making a forest management plan.

Aquatic ecosystems: Protecting aquatic ecosystems are given special attention as uniquely important to biodiversity in a forested landscape (Naiman et al. 1993, Naiman et al. 2000, Brinson and Verhoeven 1999, Calhoun 1999). Sixty percent of 480 wildlife species were observed in riparian forest in Washington State (Raedeke 1988). Important roles of riparian forests are 1) light and water temperature control, 2) organic matter inputs, 3) bank protection, and 4) source of large woody debris.

Selection of the appropriate widths and lengths for streamside corridors should be based on the objectives of the buffer and the spatial pattern of relevant influences (Lindenmayer and Franklin 2002).  Recruitment of large woody debris is a good example. Relatively narrow buffers (e.g. 10 meters or less) have generally not been considered adequate to address the biological and physical interactions between riparian forests and streams.   The Forest Ecosystem Management Team (FEMAT 1993) created one standard that uses a site-potential tree height to define protection areas (the more productive the site; the wider the buffer). Two tree heights for fish-bearing streams and one tree height for other small streams were used as standards for federal lands in the northwestern United States.  However, these standards were arbitrary, based upon expert opinion, and not derived from effectiveness studies.  Effectiveness studies (Cross 2002, Ice 2000, NCASI 2000) tend to show the vast majority of the LWD, shade and particulate matter available to streams are concentrated in the first ten meters. Microclimatic impacts, however, may not be attenuated with narrower buffers (Brosofske et al. 1997).  The length of protection along the stream is also an issue since low order streams are more common and have been difficult to identify (University of Washington College of Forest Resources 2007, DP7).  A final issue is related to management on migratory floodplains, since buffers established adjacent to the current channel position will be insufficient to accommodate fluvial shifts in the channel (Naiman et al. 2000). 

Lakes, ponds, and other wetlands are considered important elements within aquatic ecosystem that are afforded special protection when landscape management plans are developed.  The Northwest Forest Plan (USDA Forest Service and USDI Bureau of Land Management 1994a, b) assigned two site-potential tree heights to protect forests around lakes and natural ponds and one site potential tree height for other wetlands.

et al. 2004).  In particular, low gradient pools for salmon rearing and the LWD important to their formation were considered of high importance.  Only certain elevation profiles, streams banks and stream configurations likely to contribute LWD are critical in the formation of these pools. Greater protection from urban influences was thought to be more important than needed for rural areas.  While these and other elements were considered of high priority and could potentially support more effective criteria for management, there does not appear to be a direct body of research quantifying such relative priorities as a part of management planning.  

Road networks: Because road networks can heavily influence sediments and organic materials supply to wetlands, road location and construction in order to minimize sediment is another important forest management consideration (Lindenmayer and Franklin 2002).   While there exists a substantial body of literature on sediment management it is generally beyond the scope of this review. 

Skid roads in thinning units may contribute to spatial variability in stand structure, as well as sites for establishment of shade-tolerant trees (Nyland 2002).  For example, bats may utilize roads and trails for travel between feeding (Hayes and Loeb 2007).  Additionally, demographic processes such as post-harvest tree mortality (a source process for snag/downed wood) may be influenced by the proximity of skid trails (Thorpe et al. 2008).    
Wildlife corridors: Wildlife corridors have been assumed by some scientists to be an important element for landscape management (Noss 1987); however, the research evidence has not always supported this assumption.   For example, the hypothesis that corridor structure facilitates northern spotted owl dispersal was rejected (Thomas et al. 1990, Murphy and Noon 1992). Conversely, however, Verboom and Huitema (1997) reported that population densities of some bats were positively correlated to the existence of wildlife corridors.  Instead of supporting animal movement, migration, and dispersal, there is some evidence that corridor-like linear strips of forests serve as residential habitat instead of migration or dispersal habitat (Bennett 1998).   For some species, riparian forest buffers may work as wildlife corridors. The difficulty in proving the usefulness of wildlife corridors originates partly in the difficulty of defining wildlife corridors in a landscape. Therefore, the effectiveness of wildlife corridors in forest planning may need to be considered on a species-by-species basis with better-developed spatial arrangements.  Bailey (2007) notes that organisms of intermediate dispersal ability benefit most from this type of connectivity, and calls for further empirical work into the utility of corridors for the conservation of biological diversity. 

Geologic features: Specialized habitats such as cliffs, caves, thermal habitats, meadows, and vernal pools were identified as needing special attention.   Calving sites for ungulates, high-quality spawning habitat for fish, foraging sites with rare but essential food resources, overwintering habitat areas were also to be considered for special attention (Lindenmayer and Franklin 2002).

Harvesting and natural disturbances: Spatial and temporal arrangement of harvest units were identified as important in landscape-level forest management.   The size of harvest units, levels of structural complexity retained within units, and time interval between rotations and temporal arrangement of harvest units and their management prescriptions were all identified as important considerations (Lindenmayer and Franklin 2002) and are covered in considerable depth under the spatial management sections of this review.  

While young stands do not develop features associated with old forest species such as the spotted owl, heavy thinnings at 50 and 80 years were considered more effective than not thinning (Andrews and Perkins 2005).  In effect long rotations without thinnings are less effective both economically and for producing old forest-like conditions. Multiple treatment approaches may also help to mimic the variability of natural disturbance regimes.  Lindenmayer and Franklin (2002) summarize stand-level management methods by including biodiversity pathways, dispersed retention, aggregated retention, variable retention harvest systems, variable-density thinning, and snag creation. 

Longer rotations have been linked to greater biological diversity (Lindenmayer and Franklin 2002, Oliver 1992, Carey et al. 1999). However, extended rotations result in losses of economic return that, without compensation, will limit acceptability.  A range of rotation periods instead of a single rotation period has been proposed. Multiple rotation ages better mimic the variability in frequency of natural disturbance regimes (Seymour and Hunter 1999), and short rotations on certain sites may enhance certain structures that benefit biodiversity (e.g., aspen stands, Arno and Fiedler 2005).

A useful frame of reference that has been used by researchers is natural disturbance history. Disturbance regimes have been developed to imitate landscape patterns caused by natural processes. Natural disturbance regimes are divided into two classes, 1) intense episodic disturbance regimes and 2) chronic disturbance regimes. Intense episodic disturbance regimes are defined by catastrophic stand-replacing disturbances such as forest fire and floods. However, the scale of size and time of episodic disturbance regimes are usually much larger than can be implemented on the ground. Therefore, only a few cases have employed intense episodic disturbance regimes (Cissel et al. 1998, 1999).  Events associated with many chronic disturbance regimes occur at smaller spatial scales than those associated with episodic disturbance regimes, and are common in landscapes with frequent wind disturbances or insect/disease patches (Deal and Tappeiner 2002).  Gap based timber harvesting or group selection may mimic the spatial effects of many chronic disturbance regimes (Lindenmayer and Franklin 2002).  Silviculture based on chronic disturbance regimes is intended to create spatially complex landscape structures through the use of relatively small harvesting units. The current management plan for federal lands throughout the Sierra Nevada provides one example of a project emulating a chronic disturbance regime (USDA Forest Service 2001). 

In the New England landscape, agricultural clearing and subsequent abandonment of farmlands and imposition of patch clearcutting timber harvest regimes have resulted in substantial, measurable changes to forest landscape patch size distribution, stand composition, and structures such as woody debris (Seymour et al. 2002, Howard et al. 2005).  Seymour et al. (2002) emphasize the shift from intermediate-scale clearcutting to a within-stand, gap-based cutting system to regenerate the species composition and structure of historic stands in addition to restoring the large, chronically disturbed landscape patches of historic New England.  Such approaches are being investigated for other parts of the world, such as the true-fir/European beech forests of southeastern Europe (e.g., Nagel and Svoboda 2008) and the southern beech forests of southern Argentina and Chile (e.g., Martínez-Pastur et al. 2000). 

The many reports on thinning cited earlier are largely focused on achieving outcomes similar to the impact of disturbances (Bailey and Tappeiner 1998, Carey 1998, Tappeiner et al. 1997).  Bailey and Tappeiner (1998) noted substantial structural differences by different thinning treatments contributing to diversity and complexity but that there were greater differences in response across sites than across stand types suggesting stand locations are very resilient in restoring prior growth and structural conditions after disturbances.

The dispersed management model that spatially distributes harvest units across a landscape has been applied by many government agencies. The dispersed management model is appealing since it creates heterogeneity across the landscape. However, there are disadvantages as well, including negative impacts on biodiversity (Franklin and Forman, 1987), average patch size and interior habitat reduction, edge habitat increase, and increase in wind susceptibility (Sinton et al. 2000).  Many of these issues are related to lack of congruence with the spatial characteristics of the natural disturbance regime.  Lindenmayer and Franklin (2002) recommend diversifying landscape-level approaches by imitating natural disturbance regimes as much as possible in harvest planning.
Internal structural complexity can be enhanced through live tree retention and partial cutting. The retention harvesting approaches and partial cutting can result in increased capacity to sustain biodiversity, increased structural complexity, and reduced edge impacts (Lindenmayer and Franklin 2002). Many different partial cutting and patch cut regimes have been proposed (Curtis et al. 2004). Beese and Bryant (1999) reported species richness and abundance greater just 3 years after a shelterwood harvest, but with different levels of canopy retention potentially having “drastic” effects on breeding birds.  A comparison of clearcuts, mature forest, and shelterwoods in northern New Hampshire showed that shelterwood units had higher bird species richness and diversity than either clearcuts or mature forest (King and Degraaf 2000). Understory re-seeding was reported to improve plant diversity over natural regeneration (Adams and Zuo 1998), and the underplanting of conifers in managed riparian zones may also accelerate the development of structurally complex riparian forest (Beach and Halpern 2001).  Since this was believed to be a response to broader access to seed species variation, it may be become more important with climate change as it would likely speed-up the prevalence of more resilient species.

Harvest unit size and shape are considered important in landscape-level forest management.  The size of interior habitat required for target species and edge/interior ratio can provide one index (Hof and Joyce 1992, 1993; Bevers and Hof 1999; Ducheyne et al. 2006; Wei and Hoganson 2006).  Bayne and Hobson (1997) did not find fragmentation contributed to predation and that forest patch harvesting may not be a serious problem. In other studies (1998) they reported the impact of fragmentation to be largely species specific with some doing better and others worse with the impact of total habitat the more dominant predictor.  They noted many species did better outside of isolated patches of dense older forests.  The conclusion that a variety of stand types is needed to meet the needs of all species is frequently noted (Chambers and McComb 1999, Oliver 1992, Daw and DeStefano 2001).  Spatial connections between different habitat types are also noted in the literature; for example, Mladenoff et al. (1993) noted that connectivity between riparian forest and upland old growth has been disrupted in managed vs. unmanaged forest landscapes by timber harvest regimes that spatially differ from the natural disturbance regime.  The early work on biodiversity pathways concluded that while sequencing different treatments had some impact on creating diversity in structures across the landscape, regulations and guidelines limiting the size of harvests or other treatments were more important in determining ultimate outcomes (Carey et al. 1996).    

A meta-analysis on the impact of patch sizes (Bender and Contreras 1998) concluded that generalist species were sensitive to total habitat, not configuration; interior species experienced somewhat greater loss than others from fragmentation while edge species experience somewhat less loss; and migrating species less loss than resident species.  McGarigal and McComb (1995) found that bird abundance for several common forest birds in western Oregon increased in relatively fragmented landscapes.   Fragmentation would appear to be most important to just a few species for management planning.   Boutin and Hebert (2002) believe that landscape structure (fragmentation) has been overemphasized for some landscapes and that threshold effects that can be better detected by spatial modeling comparisons to natural disturbance regimes will become more important.  For some organisms and processes, fragmentation is a serious concern, but it must be carefully disentangled from co-occurring processes such as simple habitat loss in order to design effective conservation strategies (Fahrig 2002, Lindenmayer and Fishcher 2006).  

Consideration of disturbance processes operative at landscape scales is also critical to the success of forest management.  The enhanced contiguity of forest stands vulnerable to fire, disease, and insect infestation in the inland Northwest is a result of prolonged fire suppression in a system once characterized by short fire return intervals (Hessburg et al. 2000, Hessburg et al. 2005).  Recommended silvicultural treatments would focus on restoring the spatial interspersion of patch types, reducing density in pine and Douglas-fir stands, and conserving late-successional, complex forest at points in the landscape where the biophysical environment would have historically favored them (Camp et al. 1997).  

Many of the advances in conceptual and empirical understanding discussed here reflect a trend towards recognizing that management imperatives are derived from considerations at multiple spatial scales.  Hummel and Barbour (2005) employ the term “landscape silviculture” to demonstrate coordination of silvicultural activities across multiple spatial scales to achieve objectives.  This is an important trend that diverts focus from standard operational scales of focus such as the production area and harvest units.   

Effectiveness measures

While there would appear to be some support in the literature for the concepts of biodiversity enhancement as described above, the depth of research studies that include effectiveness is limited, long-term conclusions potentially premature, and many of the concepts would be difficult to implement technically as well as economically. Such approaches are natural outgrowths of efforts to preserve or restore naturally found conditions rather than to characterize effectiveness measures for integrated management. One difficulty with characterizing the effectiveness across such large scales is how to measure the benefits. Statistical measures across large scales reveal the substantial variation associated with habitat and old forest functionality.  While this validates conceptually the idea that substantial variation in forest conditions across the landscape has been the historic precedent, it is still difficult to determine what treatment to apply where for what result. Spatial analysis has emerged over the last two decades as a means of quantifying more directly the impact of any given set of treatment strategies. A substantial body of the literature has attempted to assimilate the latest landscape-level forest management studies that focus on the effects of spatial and temporal forest stand management arrangements on ecological as well as economic outcomes.

Spatial and Temporal Forest Stand Management Modeling

Recent spatial forest planning studies that provide methods to optimize both economic and ecological outcomes through managing forests are summarized in this section.  After “new forestry” was proposed by Franklin (1989), various forest management approaches that consider landscape-level parameters into account have evolved under the general topic area of spatial modeling. Early approaches often dealt with treatments at the stand or multi-stand level. Difficulties in developing landscape-level forest management methodologies originated from the need to consider both large scale and diverse parameters. Parameters that need to be considered in landscape-level management include stand shape, size, location, juxtaposition, timing of harvesting and thinning, stand level treatments, and other economic and ecological constraints.  It is difficult to set up landscape size silvicultural experiments and set aside the area for a period of time long enough to collect sufficient data. Therefore, simulation studies became an alternative way to explore landscape-level forest management. 

Multi-Stand-Level Approaches

Franklin and Forman (1987) and Li et al. (1993) explored the possibility of implementing landscape objectives such as patch size, edge length, and patch configurations into a forest management plan. They examined only clear-cutting patterns in a static landscape. To make their simplistic approaches more applicable in real situations, Baskent and Jordan (1996) and Baskent (1999) developed the management design model LANDMAN and a landscape pattern analysis model, PATREC. LANDMAN was developed to explore forest landscape design under different management strategies. In their approach, general landscape management strategies such as scatter harvesting, cluster harvesting, edge progressive harvesting, and nuclei progressive harvesting are pre-defined, while spatial objective variables and intervention strategies can be developed.  Li et al. (2000) integrated their landscape model, LEEMATH, with a habitat suitability model to aid in management decisions. However, their approaches did not have the capability to optimize both economic and ecological goals through a mathematical optimization process.  Aside from the forest landscape management approaches, landscape considerations in harvest scheduling have been developed in operational research studies. Spatial components within a harvest scheduling process started in 1990 and optimization studies for timber revenue or volume with wildlife habitat objectives have also increased (Bettinger and Chung 2004). Since the goal of forest landscape management and optimization studies with wildlife habitat constraints have common objectives, both ideas can be integrated into spatial forest planning.

 Spatial Forest Planning 

Spatial forest planning accommodates spatial requirements as well as multiple, often conflicting, management objectives over a landscape (Baskent and Keles 2005). It uses a mathematical optimization approach such as linear programming, simulation, or heuristics to achieve conflicting goals in an optimal manner.  Spatial requirements often relate to size, shape, juxtaposition, and distribution of management units, minimum and maximum harvest block size limits, adjacency restrictions, connectivity and proximity, and interior and edge habitat availabilities. Spatial forest planning focuses on forest management activities and the specific tools used to develop, implement, and evaluate spatial forest plans and alternative policies (Bettinger and Sessions 2003a). In spatially managed forests, each stand is treated as an individual component and its relative location to other stands is considered. Spatial forest-planning problems are often solved through optimization calculations in order to achieve all management goals in an efficient manner. Outputs from optimization can be graphically presented using geographic information systems (GIS), and utilized in planning processes as well as communication with interest groups.

Bettinger and Session (2003b) pointed out several reasons to adopt spatial forest management approaches over other traditional non-spatial approaches. First, some public agencies charged with managing resources for the public good operate under a wide variety of regulations that include landscape considerations such as limiting the size of clearcuts and maintaining core area requirements for wildlife (Oregon Secretary of State 2002). Second, voluntary programs such as the Sustainable Forestry Initiative program (Wallinger 2003) and habitat conservation plans (HCP) may require landscape-level forest management for indigenous flora and fauna. Third, in order to maintain and develop environmental resources to meet various objectives including economic ones, an efficient use of the landscape is necessary (Nalle et al. 2002). For example, when adjacency restrictions which require the arrangement of harvests over space and time are required, spatial forest planning can evaluate the problem mathematically before activities are implemented on the ground. Fourth, forest landscape planning efforts require a broad perspective looking across ownership boundaries, and spatial forest planning can help evaluate potential activities that are likely compatible across multiple ownerships. Fifth, the ability to graphically display the results of alternative forest plans on maps using GIS may greatly help an organization’s forest planning. Sixth, when biodiversity is a stated objective, forest management ideally emulates the natural disturbance regime with respect to size and frequency of disturbance patches.  This tends to perpetuate the biophysical environment in which native forest fauna and flora evolved (North and Keeton 2008).  Spatial planning is necessary to ensure that the size and frequency distributions associated with silvicultural activities do not deviate excessively from the natural disturbance regime (e.g., Seymour et al. 2002).  Finally, spatial forest planning may help build trust among interest groups and organizations (Bettinger and Session 2003b). This is because a systematic and organized planning process that recognizes and accounts for spatial concerns and displays the results graphically may ensure that the resulting forest plans can withstand rigorous evaluation (Bettinger and Session 2003b).

Spatial forest planning offers a very quantitative framework to implement forest management plans seamlessly from long-term and wide scale strategic plans to ground level operational plans. However, stand-level forest planning has also played a significant role in the implementation of diverse management goals into forest planning (Li et al. 1993; Marzluff et al. 2002).  A stand-level approach is especially well suited for strategic level planning (Murray and Church 1995, Bettinger et al. 2004b). However, stand-level forest planning requires the additional step to translate the strategy into operational planning (Van Raffe 2000, Bettinger et al. 2004b). In operational planning, the execution of ground-level detailed operational activities is planned on a yearly or seasonal basis (Baskent and Keles 2005). Additionally, stand-level trade-off analysis may not guarantee whether a proposed plan can achieve all management goals in an optimal fashion.  Therefore spatial forest planning that includes both spatial pattern considerations and harvest schedule optimization in one place may be the better approach for forest management planning.  The difficulty with spatial models has largely been with the complexity of spatial criteria and computational difficulty associated with problem size. It may not be computationally practical to consider complex spatial criteria in large problems.

Management Objectives

Objective functions modeled in spatial forest planning are diverse. Spatial requirements such as harvest unit or habitat patch size, shape and distribution (Cox and Sullivan 1994; Baskent and Jordan 1995, 2002; Başkent 1997, Gustafson et al. 2006; Hurme et al. 2007, Hof and Raphael 1997; Holzkamper et al. 2006; Kurttila 2001; Rempel and Kaufmann 2003; Kurttila et al. 2002; Authaud and Rose 1996; Bettinger et al. 2003, Bettinger et al. 1997), adjacency restrictions (Jones et al. 1991; Weintraub et al. 1994; Yoshimoto and Brodie 1994; Murray and Church 1995, 1996; Snyder and Revelle 1997; Hoganson and Borges 2000; McDill and Braze 2000, 2001; Nalle et al. 2005), connectivity and proximity (Nelson and Finn 1991; Sessions 1992; Hof and Joyce 1993; Church et al. 1998; Williams 1998; Lu and Eriksson 2000, Weintraub et al. 2000; Richards and Gunn 2000, 2003), interior and edge habitat (Hof and Joyce 1992, 1993; Bevers and Hof 1999; Ducheyne et al. 2006; Wei and Hoganson 2006), habitat attributes (Rohweder et al. 2000), habitat effectiveness (HEI, Bettinger et al. 1999a), and wildlife populations (Moore et al. 2000; Spring et al. 2001; Calkin et al. 2002; Juutinen et al. 2004; Nalle et al. 2004; Polasky et al. 2005; Loehle et al. 2006) were taken into account with economic goals.

Wildlife Habitat as a Management Objective

Spatial forest planning studies that chose wildlife habitat as an objective function were more common in our references. Bettinger et al. (1997) developed a tabu search algorithm to find efficient harvest schedule subject to even-flow of timber volume harvested, adjacency constraints, and spatial wildlife habitat quality (distribution and amount of foraging and thermal habitat for elk) goals. Bettinger and Boston (1999a) developed a method to integrate a Habitat Effectiveness Index (HEI) for Roosevelt elk into the objective function in an optimization process. They set a commodity production goal with the HEI objective and adopted a tabu search algorithm to find efficient relationships between those two conflicting objectives.  Boston and Bettinger (2001) set similar management goals with red-cocked woodpeckers and used linear programming and heuristic techniques to solve their problems. The hybrid two-stage approach was also used in Boston and Bettinger (2002) to solve four problems with spatial constraints such as maximum opening size and maximum average opening size. Bettinger and Boston (1999b) also integrated different management objectives such as even-flow and adjacency considerations into an optimization process. The goal in their study was to achieve the highest, and most even, flow of timber volume over a certain time period with an adjacency harvest restriction rule.

Bettinger et al. (1996) set five management goals, such as stream temperature for fish habitat, habitat effectiveness index (HEI) for elk, habitat corridors, old seral stage stands, and an even flow of timber, and tested three riparian management strategies and two forest road construction scenarios.  The objectives of Bettinger et al. (2003a) was to develop a process where the amount of habitat for northern spotted owl could be maintained within a certain radius of an owl nest location, while using thinning and group selection harvests to assist in the development of mid- to late-successional forest conditions. They reported that when nesting, roosting, and foraging (NRF) habitat levels were constrained to a minimum level of 40 %, net present value declined by almost 24%, while average NRF value increased 11% over a 100-year time horizon. When NRF habitat levels were constrained to a minimum level of 80%, NPV declined almost 70%, while average NRF increased 29%. Although Kurttila et al. (2002) did not include an economic evaluation of different management decisions; they evaluated spatial forest patterns for flying squirrel and moose that show conflicting habitat demands. They used habitat suitability indices (HSI) to evaluate the amount of habitat availability for both species. They calculated the proportion of a specified type of stand boundary to the total boundary length, spatial autocorrelation of HSI and a weighted mean of the stand-level HSI to evaluate the distribution of different habitat types. They evaluated results from an optimization process in terms of available habitat areas, spatial patterns and habitat connectivity. Hurme et al. (2007) set timber production and available suitable habitat for the Siberian flying squirrel as management objectives. They developed an empirical site-specific habitat model and integrated it into their optimization framework. They created five alternative forest plans under different combination of objective levels. They concluded that the formation of flying squirrel habitat in the landscape was enhanced, and did not always incur severe reductions in harvestable timber volume.

Wildlife Populations as a Management Objective

There were few studies that modeled wildlife populations or communities in their optimization processes. Hof et al. (1994) combined wildlife growth and dispersal as a dynamic process in their mixed integer linear programming approach. Hof and Raphael (1997) used a method in Hof et al. (1994) and developed an optimization procedure that can analyze habitat layouts for northern spotted owl. Calkin et al. (2002) integrated a wildlife model into their optimization process. Simulated annealing was used to solve for harvest schedules that maximized the net present value of timber harvest subject to a target value for likelihood of species over a 100-year planning period. Nalle et al. (2004) developed a method that used two heuristic algorithms and spatially explicit wildlife population models for the common porcupine and the great horned owl to evaluate the tradeoff between ecological and economic outputs with various harvest schedules. Polaksy et al. (2005) developed an optimal land allocation algorithm for working forests, agricultural lands, and reserve sites.  They created a spatially explicit model for analyzing the consequences of alternative land use patterns on the persistence of various species and on market-oriented economic returns. Through an optimization process, they found that the degree of conflict between conservation and economic returns appears much less using their joint biological and economic modeling approach than using a reserve-site selection approach.  Lindenmayer and Possingham (1996) used a population viability model to predict persistence of the Leadbetear’s possum in Australian mountain ash forests under a number of different timber harvest scenarios. 

Optimization Techniques

Spatial forest planning relies on optimization techniques to solve problems that often include conflicting ecological and economic management goals. Among many optimization approaches, meta-heuristic such as simulated annealing, tabu-search, and genetic algorithms have been used for wildlife studies. This is because heuristics are an effective approach when the size of a problem is large, such as landscape-level forest management with wildlife objectives. Meta-heuristics provide alternative approaches for spatial forest planning studies. Meta-heuristics cannot guarantee optimality in the solutions (such as reaching a global optimum), however, and often integer programming cannot be applied to wildlife problems for this reason.  However, they can deal with non-linear, complex, and large size problems. Therefore, meta-heuristics are becoming more mainstream in recent spatial forest planning studies (Bettinger and Chung 2004). Bettinger et al. (2002) tested 8 different heuristic techniques (random search, simulated annealing, great deluge, threshold accepting, tabu search with 1-opt moves, tabu search with 1-opt and 2-opt moves, genetic algorithms and hybrid tabu search / genetic algorithm search) and reported that simulated annealing, threshold accepting, great deluge, tabu search with 1-opt and 2-opt moves and tabu search / genetic algorithm performed better than other methods in their testing environment. Boston and Bettinger (1999) conducted a similar study to compare Monte Carlo integer programming, simulated annealing, and tabu search and reported that simulated annealing found the highest solution value for three of the four planning problems. Studies that used heuristics were listed.  Simulated annealing was adopted by Baskent and Jordan (2002), Calkin et al. (2002), Chen and Von Gadow (2002), Nalle et al. (2004). Tabu search and its variations were used by Bettinger et al. (1997, 2007), Boston and Bettinger (2002), Caro et al. (2003), Nalle et al. (2004). Threshold accepting was used in Bettinger et al. (2003a).

Hof and Joyce (1992) conducted one of the original studies that introduced an optimization technique to maximize wildlife habitat availability and stand allocations. They used non-linear integer programming to solve optimal stand allocation for the amount of edge and interior habitat as well as connectivity. In order to improve several shortcomings such as the non-convex nature of spatial models and limitations in the solving ability of their method, Hof and Joyce (1993) proposed a mixed-integer linear programming approach for wildlife and harvest schedule problems and showed that their method could solve more complex constraints such as accessibility and larger size problems. However, both integer and mixed-integer programming have critical limitations for solving real forest management problems such as the large problem size and difficulty in formulating problems. Mixed-integer programming was used in Hof and Joyce (1992, 1993), Hof and Bevers (1994), Bevers and Hof (1999) and Boston and Bettinger (1999).  Boston and Bettinger (2001) combined linear programming and a tabu search / genetic algorithm technique and claimed that their two-stage approach was superior to the one-stage approach that only used a tabu search / genetic algorithm technique. Limitations of integer and mixed-integer programming to wildlife habitat optimization were detailed in Baskent and Keles (2005).

Recent spatial forest management studies that focused on wildlife habitat suggested that finding optimal relationships between conflicting ecological and economic management objectives create more cost efficient management solutions compared with traditional approaches (Juutinen et al. 2004, Nalle et al. 2004). Other research also suggested that a large fraction of conservation objectives could be achieved at little cost to economics through spatial forest planning (Nalle et al. 2004, Polasky et al. 2005, Hurme et al. 2007).

Spatial forest planning can include and test complex forest stand treatments such as biodiversity pathways. Imaki et al. (2007) adopted a biodiversity pathway in their study and constructed production possibility frontiers (PPF) to examine trade-offs between habitat conservation levels and opportunity costs for an interior and edge bird species. This study showed that the trade-off was not linear and spatial dependencies from stand arrangements and stand treatment allocations influenced the shape of the PPF. Ecological traits of species such as dispersal distance, home range size, and habitat preference also significantly changed the optimal forest management plans.

Complexity and Limitations

Spatial forest planning concepts and techniques have been developed for two decades; however, there are still practical limitations to the application of spatial forest planning to real forest management problems such as including more complex and diverse management objectives and personnel who can develop complex optimization processes and GIS analysis.  Spatial forest management can provide the capability to test and implement conceptual landscape management approaches on the ground.  Too much complexity, however, can work against developing a common understanding of what was learned and how to implement a practical program on the ground given many more specific details than could be handled in an optimization program.  “The devil is in the details”, to quote an old adage. Furthermore, spatial forest management methods do not eliminate stand-level considerations as the impacts of treatments at the stand-level generate the metrics that can be used and assessed spatially. 

Salient Messages

What are the salient messages from these many studies?

The process starts with understanding the impact of treatments at the stand level although managing a few stands will not be sufficient to meet most objectives since there are many different desired conditions across a large area (i.e. no one treatment fits all).

Since overly dense stands (stem exclusion structure) are in surplus relative to natural history yet provide the least habitat, a range of thinning treatments in dense stands will likely be necessary to reach a range of habitat objectives and they might best emulate disturbance history even though that does not by itself guarantee functionality.  Thinnings are essential for restoration of an understory and greater biological complexity from managed forests. 
If the objective is largely focused on old forest species for which the supply has undoubtedly been reduced the most, dispersed treatments may be more important than protecting interiors. Moving mosaics of dispersed treatments generally allow better results across conflicting objectives such as economic impact vs. old forest habitat. 

One can target and manage for future conditions, however natural old forest conditions were very variable so that any target needs to include variation, and protecting certain critical legacies may be the best way to provide for some specific future conditions.  However, in the long term, legacies may not be protectable from all disturbances and some of the more mature stands may need to be managed for legacy restoration. 

There are many tradeoffs in meeting different ecological and economic objectives and learning how to be effective is important.  Many modeling efforts, from simulation to spatial optimization, have been examined but have not greatly simplified the complexity inherent in meeting a range of future conditions across a broad landscape as the objective.  Situations are different and local objectives will most likely determine best local treatments although that does not guarantee fulfilling all important criteria across a broad landscape.  There are tradeoffs both within ecological objectives and between ecological and economic objectives that can be displayed.  Establishing hard criteria that weight these objectives to produce best or optimum results is rarely practiced except by optimization programs where any one objective can be optimized while the others are characterized as constraints. Achieving multiple objectives can be approximated by iteratively relaxing or tightening constraints.

Authors that have attempted to summarize management opportunities to maintain biodiversity tend to consider one of two approaches, the coarse filter approach or the fine filter approach (Loehle ,Wegley et al. 2005).  The coarse filter approach assumes that maintaining a range of forest patterns and successional stages that are similar to the historic natural landscapes is sufficient to maintain a degree of biodiversity. The fine filter approach tries to meet the functional needs of specific species even if based on empirical comparisons of structural conditions as a proxy for function.  Some features may be good for a broad range of species while others appear to be necessary only for a few species. 

What are the more effective treatments characterized in these studies? 

A 2006 conference on managing for wildlife habitat in west side production forests resulted in many papers (PNW-GTR-695 March 2007) endorsing and customizing the biodiversity pathway concepts originating a decade earlier (Carey et al. 1995). Some of the papers focused on structural features as targets (large trees, dead wood and snags, structural diversity including floristic (understory) diversity).  Diversity was considered necessary if not sufficient for fungi, lichens, mosses, terrestrial and aquatic invertebrates and the process supportive of these organisms that are important to wildlife (Hagar 2007). 

A key theme running through much of the recent literature is that maintenance of structural and compositional richness in each developmental stage is critical to support most species and processes in forested landscapes (Lindenmayer and Franklin 2002).  Traditional silvicultural methods, which are relatively agricultural in nature and not tightly modeled on natural disturbance regimes, do not tend to function in this way (Puettman et al. 2009).  These methods may, however, be modified to produce desired values (O’Hara 2001).  Primary elements of silvicultural systems that recognize processes in natural ecosystems include: extended rotations (Curtis 1997, Lindenmayer et al. 2006), variable retention at harvest (Franklin et al. 2007), and intermediate operations to enhance complexity during a rotation (essentially the biopathways approach).  Much research remains to be done to assess the effectiveness of combinations of these principle elements, and modeling efforts should continue to receive the initial focus while field trials are conducted at longer time scales.       

What constitutes the relevant range of treatments for a modeling exercise? 

 A range of biodiversity pathways and commercial alternatives that focus on the production of a range of structural features found to be important to habitat and economic returns are needed as the primary input to the modeling exercise. While one should not focus on a single or even just a few species, there is ample evidence that the habitat in shortest supply is best described as being characterized by old forest structures. The structures in greatest surplus are overly dense stem exclusion structures. Studies focused on commercially managed forests have generally found adequate levels of most species except those preferring old forest structures. In contrast the lack of management on federal lands and perhaps some state lands including no clearcuts or thinnings is resulting in no open areas except those that may result from natural disturbance. Disturbance patterns have however been substantially reduced by human interventions such as fire suppressions and patterns of land management. 

The most successful economic prescription on industry lands relies heavily on pre and post regeneration vegetation control.  The rapid growth of young conifer does not eliminate a short period with open structures but very quickly produces very dense albeit young stands, a less than desirable structure for most habitats (University of Washington College of Forest Resources 2007). 

As the purpose of a modeling exercise is to explore ranges of innovative treatments while pushing the boundaries of current practices to better understand ecological/economic tradeoffs, a wide range of treatments needs to be considered. Similarly a wide range of ecological metrics will need to be analyzed. Metrics of greatest interest will logically focus on older forest structures but should not do so exclusively.

What metrics for ecological sufficiency are highlighted in the literature?

• Non-parametric statistical tests of the goodness of a sample treatment replicating old forest samples (i.e. time in the desired old forest structure described earlier  (Gehringer 2006).
• Meeting defined habitat thresholds such as structural conditions specific to owl nest sites or structural definitions for nesting-roosting-foraging habitat (e.g. WAC definitions of nesting roosting foraging habitat).
• The Old-Growth Habitat Index (OGHI) developed to define old-growth forests. The OGHI integrates five stand structure parameters and successional status into a single measure by the Old-Growth Definition Committee in 2004 (Washington Department of Natural Resources 2005). Those five parameters are:
  
  
  1. Large trees (number of trees per hectare > 100 cm dbh)
  2. Large snags (number of standing dead trees per hectare > 50 cm dbh and > 15 m tall)
  3. Volume of down woody debris (cubic meters per hectare)
  4. Tree size diversity
  5. Stand age (years)

Each of the five elements making up the OGHI is scored on a scale of 0 to 100. There are three forms of the OGHI for western Washington. The standard OGHI is simply the average of the five scores. A modified OGHI excludes stand age from the calculation and emphasizes stand structures. A weighted OGHI uses the same 4 elements as a modified OGHI but weights each parameter to emphasize the density of large trees in a stand. The density of large snags was highly variable and the least informative in the weighted OGHI. Thresholds in the indices to define old growth forest were subjectively selected based on the experience of the ecologists. An index score of 60 (the standard OGHI) was generally corresponded to stands whose dominant overstory Douglas-firs were > 200 years old (Washington Department of Natural Resources 2005).

Assessment of stands in the relatively drier forests in the east Cascades requires the integration of structural differences in these stands from more mesic forests (Franklin et al. 2007, Everett et al. 2007). These include fewer large trees per acre, fewer standing dead trees, lower volumes of downed wood, a patchier distribution of regenerating trees, and simpler canopy structure in general.

• A lower threshold such as the amount of a late seral forest structure class (Oliver 1981, Franklin et al. 2002).  This metric is often based on a coarse filter approach to forest diversity and hence somewhat easier to measure testing for presence of all of about 5 broadly defined structure classes.
• Finer filter objectives such as downed wood, snags, canopy closure while not good discriminators of old forests in general, they may still be important for some habitat such as woodpeckers and owls.  Finer filter objectives can provide better or more critical habitat within the broader class of old forest structures.  These structural elements tend to indicate ecological complexity and function during all developmental phases (including naturally disturbed stands in early succession).  The importance of variable densities imposes statistical problems since characterization at the stand level requires breaking a stand into sub-parts in order to measure the variation.  There is substantial natural and managed variation in stand metrics caused by natural conditions (rocks, disease patches, and wind disturbance) as well as management impacts (thinning trails, thinning densities). 
• Habitat Suitability Indices that depend on the segmentation of stand structures into a large array of forest conditions that can be linked to use by different species.  Early efforts may evolve over time as more data is collected and the link between suitability and structure class is developed by more thorough statistical testing.  Gradations from one condition to another tend to produce knife edge measures of suitability rather than a more continuous transition.   
• Economic “forest value” for different treatments i.e. the discounted present market value net of discounted costs. Soil expectation value (SEV) (after harvest removal) is the most representative value for perpetually sustained forest management.  For dry forests with a stand-replacing fire risk, the economics of avoided suppression costs may play an increasingly important role in the decision to thin or not (e.g., Snider et al. 2006).
• Fire risk/severity measures (e.g., Hummel and Barbour 2007, Hessburg et al. 2005).  This is especially relevant for dry forest types where fire suppression has resulted in greater contiguity of fuels and the potential for stand-replacing fire.       
• Landscape patch-size distribution characteristic of historic landscapes (e.g., Seymour et al. 2002).  Other landscape-scale metrics may also be useful. 

The National Commission on Science for Sustainable Forestry (NCSSF) sponsored many studies with a focus on sustainability of biodiversity among critical attributes. They recognized the importance of characterizing biodiversity locally. They published a guidebook aimed at helping interested parties determine how to identify and maintain biodiversity within the sustainability context (NCSSF 2007).  They suggest using condition indicators, pressure indicators and policy response indicators. Condition indicators include metrics like the percent of area in a mature condition by forest type, large tree density, and mature forest reserves. Pressure indicators include negative counter parts like the percentage of acres not yet mature, as well as positive indicators like the percentage managed for some degree of forest retention.  NCSSF policy indicators incorporated process motivation such as a written conservation policy and the application of incentives recognizing the non-market aspects of sustaining biodiversity.  Collectively this process integrates many of the indicators provided in the literature.   

What range of silvicultural treatment alternatives are considered?

Variables such as initial density, rotation length, and internal variation in treatments may be modified to fit landscape-specific conditions in research at the OESF and elsewhere.

1. Commercial rotations (e.g. 45 years or optimal economic rotation for a medium site class in the Pacific Northwest, with optional thinning for dense stands). Commercial rotations provide a baseline for economic comparisons in order to determine the opportunity cost (net revenue or present value loss) to provide alternatives with increased conservation values.
2. No action (no harvest within planning interval after stocking, although not necessarily replicating natural development).
3. No action areas restricted to important legacy "biodiversity hotspots".
4. Long rotations (e.g., 60-100 years with medium density and high density alternatives) with longer rotations or periodic restoration thinnings when old forest retention objectives are critical.
5. Short rotation intensive (35 years), with pre-commercial thinning [PCT] if needed.
6. Biopathway with multiple thinnings and a long rotation clearcut i.e. only green tree retention with no retention of the largest trees in the overstory (approximate trees per acre each treatment = 300, 150, 60, green tree).  Thinning operations may be omitted for lower density stands.
7. Biopathway on extended rotation but no retention at harvest.
8. Fast biopathway (accelerated thinning schedule) with no retention at harvest.
9. Biopathway with retention of 15 to 30 large overstory trees for RMZ (PCT, single thin low density, 2 thin higher density, overstory retention/harvest).
10. Group selection of variable-size compartments with retention of CWD.  This system may be tested for functional similarity to the windthrow regime common to many forest types (e.g., western Olympic lowlands or southern beech). 
    
    Finer filter refinements
    
    
11. Manage #5 and/or #6 for narrower range in canopy closure (more thinning entries throughout rotation).
12. Manage #5 and /or #6 for recruiting more logs/snags.
13. Manage #8 for recruiting more CWD.

Summary of Trends and Challenges In Silvicultural and Forest Management Research

Innovative silviculture modeling over range of treatments noted in the literature applied across landscapes based on the ecological sufficiency metrics noted in the literature provide instructive alternatives serving a range of objectives from economic to ecological restoration.   Gaps in ecological metrics will be noted. The modeling generally starts by simulations of various treatments at the stand level and progresses to simulated spatial applications across example landscapes.  The outputs provide a quantitative assessment of comparative impacts across treatment alternatives and ecological sufficiency metrics.  The lack of retrospective studies on the impacts of earlier treatments limits the degree of validity testing between simulations and outcomes.  More long term monitoring of treatments will be needed but will be slow in coming. 

Although there is much that remains to be done in terms of empirical research, this review clearly documents encouraging advances in knowledge and practice in temperate forests in the Pacific Northwest and around the world.  Improvements in the spatial modeling (and concurrent advances in remote sensing and computational power) will allow managers to best approach complex, multiple-output resource management scenarios.  Silvicultural methodology is beginning to draw inputs from multiple spatial scales, instead of the single forest stand alone, and the heterogeneity of any forested landscape is increasingly addressed in planning.  Elements of natural disturbance regimes are emulated, from single-tree gap creation to the incorporation of biological legacies into large harvest units and inclusion of mesoscale reserves in managed landscapes.  Traditional silvicultural systems are increasingly modified for site-specific circumstances, much like the originators of those systems intended.  And the rigor and sophistication with which ecological functionality is quantified continues to expand in concert with a deeper understanding of both managed and natural forest ecosystems.  

The OESF and numerous similar management situations in temperate forests around the world represent key opportunities to perform research at stand and landscape scales at a time when natural resource managers around the world are recognizing that traditional models of silviculture may not recognize the full range of disturbances, developmental pathways, and landscape patterns represented in functional forest ecosystems.  Evolving silvicultural systems still recognize the importance of producing wood and other forest products, but have expanded their recognition of serving biological diversity and complexity objectives as well.

How to properly assess the economic value of achieving conservation objectives relative to the measureable cost of production or protection, while not the focus of this review, was generally not very evident but is receiving increased attention in the growing literature on ecosystem services.    

 Continue to Reference List