Willamette National Forest

 

Stand Structure and Vegetation Changes Five Years after Applying Three Thinning Treatments and Control

Preliminary Results

Liane Beggs and Klaus Puettmann

Oregon State University, Corvallis, OR 97333

 

Introduction

As a comprehensive and integrated long-term ecological study, the Young Stand Thinning and Diversity Study (YSTDS) is designed to test the efficacy of thinning, underplanting, and snag creation in accelerating development of late-successional habitat.  It has been proposed that this combination of techniques will lead to earlier increases in stand heterogeneity, thereby enhancing flora and fauna diversity while maintaining timber production goals.  Impacts of these treatments on understory vegetation, chanterelles, small mammals, amphibians, birds, and invertebrates are being studied.

This paper reports preliminary results of understory vegetation response to thinning four to seven years following treatments.  It is hypothesized that thinning will alter microsite conditions, overstory stand structure, and composition, structure and diversity of the understory vegetation.  Previous work lends support to this hypothesis (Alaback and Herman 1988, Carey and Johnson 1995, Bailey et al. 1998, Harrington and Edwards 1999, Thomas et al.1999, Thysell and Carey 2000). However other studies found no change in the understory following thinning (He and Barclay 2000).  None, however, were able to track the dynamic response over longer periods of time.  Examining this relatively early period of vegetation response will provide a foundation for interpretation of later results.

 

Methods

Study Design

The YSTDS consists of a randomized block design comprised of four treatments within four blocks.  Study blocks are designated as:  Cougar Reservoir (CR), Christy Flats (CF), Sidewalk Creek (SC), and Mill Creek (MC).  All blocks are located in the Willamette National Forest on the western slope of the Cascade Range of Oregon and are composed of 40-50 year old planted Douglas-fir stands occurring within the Western Hemlock (Tsuga heterophylla) zone (Franklin and Dyrness 1973).  Mean annual precipitation for the area is 230 cm.  Elevation of sites ranges from 439 to 800 meters above sea level. 

Each block contains four treatments: Control (C), Light Thin (L), Heavy Thin (H), and Light Thin with Gaps (LG) (see Table 1 for site characteristics of each treatment area).  C treatments retained original stand densities of approximately 250 trees per acre (tpa).  L and H treatments were evenly thinned to approximately 100-120 tpa and 50 tpa, respectively.  LG treatments were lightly thinned to 100-120 tpa with additional cutting of randomly dispersed 0.5 acre circular gaps.  Therefore, the LG treatment can be divided into 3 sub-treatments: Gap (LGG), Edge (LGE), and Matrix (LGM).  All treatments were applied between 1994–1996.  Table 2 displays one-year post treatment average overstory characteristics of the four treatments.

 

Table 1:  Site characteristics of each treatment area

Block

Treatment

Total # of Subplots

Area (Ha)

Stand Age

Elev. (m)

Aspect

Slope %

CR

Control

23

30

40

805

E

18.8

CR

Heavy

13

19.4

40

792

E

24

CR

Light

26

37.2

38

610

E

17.1

CR

Light with Gaps

29*

14.6

40

792

ENE

16

MC

Control

25

52.6

42

902

SSEE

21.1

MC

Heavy

23

34.8

42

658

SE

22.9

MC

Light

30

37.2

43

524

S

20

MC

Light with Gaps

29*

19.8

42

439

SSW

8.9

CF

Control

23

30.8

39

878

SE

6.2

CF

Heavy

15

20.2

36

905

SE

0

CF

Light

24

32

39

902

SE

5.3

CF

Light with Gaps

30

38.9

40

905

SSEE

5.3

SC

Control

17

51

37

634

N

11.4

SC

Heavy

13

19

35

652

NW

16

SC

Light

15

22.3

33

646

NNE

21.8

SC

Light with Gaps

30

30.4

39

671

N

14.5

 

Table 2: Overstory stand characteristics one year after thinning for all trees 8 cm dbh (exception: Overstory % Cover was for all trees 5 cm dbh).  (Averages across four treatment areas.)

 

 Treatment

Overstory Cover (%)

                     DBH (cm)

Conifer DBH (cm)

    BA (m2 / ha)

TPH

Control

79

24.8

25.3

44.6

767

Heavy

29

29.9

31.4

15.1

182

Light

50

29.5

32.5

24.3

302

Light with Gaps

42

29.0

32.0

19.8

247

 

Sampling Design

“First-year” vegetation sampling occurred during the summer of 1995, 1996, or 1997, depending on time of harvest completion.  For ease of communication, this data will be referred to as “1997” data, inferring it is data describing the earliest post-harvest response.  Resampling was completed during the summer of 2001; the data depict vegetation response 4-6 years post harvest.

All vegetation sampling was conducted using 0.1 ha circular permanent plots.  The number of plots was selected to ensure sampling of 7.5% of each treatment area; therefore the total number of plots in each treatment areas varied (see Table 1).  The LG treatment areas were an exception.  Due to the variability and the desire to sufficiently sample each of the three “sub-treatments,” each LG treatment area contained 30 plots, with 10 plots in each sub-treatment.

In each plot, overstory percent cover was measured centrally and at four cardinal directions (see Plot Diagram). Overstory trees were also tallied for DBH and species.  Understory tall shrub and small tree (diameter at breast height (dbh) < 5cm but taller than 10cm) cover was also recorded within each plot along two 14.5m transects using the Transect Line-Intercept method.  Nested within each plot were sixteen, 0.1 m2 subplots placed along the two 14.5 m transects (see Plot Diagram).  Percent cover of all herbs, low shrubs, graminoids and bryophytes, as well as percentage of exposed mineral soil, coarse litter and fine litter were recorded within each subplot.  Taxonomic nomenclature follows Hitchcock and Cronquist (1973).

 

Data Analysis

All data from the 16 subplots were first averaged to provide mean cover values for each plot.  Because cumulative cover estimates were not taken for separate lifeform groups (Tall Shrubs/Small Trees, Low Shrubs, Herbs, Bryophytes, and Exotics), cover estimates for these groups represent the average combined cover for the species in each group.  Data from all plots within a treatment were then averaged to provide mean values for each treatment area.  For the LG treatment areas, data from plots were averaged for each sub-treatment.  Sub-treatment averages were multiplied by the proportion of each sub-treatment area in the plot to provide a weighted treatment average.  Note that due to the sampling design, gap sub-treatment sampling is more representative of the area within a 10.5 m radius of gap center than the entire gap and edge sub-treatment sampling is more representative of the area 10.25 m from the gap perimeter than the entire “edge” area.  Weighted sub-treatment averages were then added together to obtain overall averages for each LG treatment. Using treatment averages, two-way ANOVA and Tukey-Kramer multiple comparisons were used to test for differences between treatments while controlling for blocks.  All statistical analysis was done using SAS version 8.2.  

Multivariate analysis of the understory plant communities was also conducted.  For this analysis, the LG sub-treatments (LGG, LGE, and LGM) were kept separate.  This provided a total of 6 treatments per block and an overall total of 24 treatment areas for each of the two measurement years.  Cover of understory species, mineral soil, fine litter, and coarse litter from each subplot as well as overstory and tall shrub/small tree species covers were first averaged for each plot.  Mean cover of each species and environmental variable in each treatment area was then calculated by averaging the mean values of the plots.      

From the treatment area averages, two data matrices were constructed.  The species matrix consisted of 48 treatment areas (sample units) and 177 species. The environmental matrix was composed of 48 treatment areas and categorical (treatment, block, year) and quantitative (percent cover of overstory, understory, mineral soil, coarse litter, and fine litter) data.  Overstory, understory, coarse litter, and fine litter were used as indicators of stand and habitat conditions for 1997 and 2001.  Because exposed mineral soil was an indicator of post-harvest soil disturbance only the values from the 1997 data set were used.  

Multivariate community analysis was conducted using PC-ORD v. 4.0 (McCune and Medford 1999).  All data matrices were first examined for outliers.  For the species matrix, several moderately strong outliers both among the sample units and among the species were discovered.  Initial ordinations also displayed a high degree of noise, making patterns difficult to detect.  To correct for this, rare species (species occurring in less than 2 sample units) were dropped from the analysis.  Dropping these species reduced the original 48 x 177 species matrix to 48 x 124, but failed to eliminate the outliers.  In order to reduce the influence of outliers and decrease the high amount of variation in the data (average percent cover values ranged over 4 orders of magnitude), a log transformation was preformed.  To retain all zero values, 0.01 (derived from McCune and Grace 2002) was added to all values prior to transformation.  Following transformation, one sample unit, the LGG treatment from the CR block in 1997, remained a moderately strong outlier.  This sample unit had a relatively high percent cover of Sencecio sylvaticus, an invasive, exotic plant typical of disturbed soil.  Ordinations were run with and without it.  Exclusion of the site lead to a slightly decreased emphasis on the importance of S. sylvaticus in final interpretations and a slight shift of sample units along the ordination axis, but overall interpretations did not change.  Therefore, this site was included in the analysis.

Groups defined by treatment and by year were first tested for differences using multi-response blocked permutation procedure (MRBP; Mielke 1984).  Because it does not require distributional assumptions, this non-parametric technique is well suited for testing for differences among groups in species space.  Euclidean distance measure was used due to the incompatibility of the Sørensen distance with the median alignment of blocks used in the procedure (McCune and Grace 2002).

The test of groups defined by year was segmented into two separate analyses: a test of difference across time of thinned treatments (excluding controls) and a test of difference across time of only control treatments.  Excluding controls permitted shifts of the thinned treatments over time to be revealed that were otherwise concealed by a relatively static control. 

Indicator species analysis (Dufrêne and Legendre 1997) was used to identify differences in species composition among treatment groups and also between years.  This method calculates an indicator value (IV) from relative abundance and relative frequency of each species in each group.  The statistical significance of this value is evaluated by comparison with 1000 randomizations of the data using a Monte Carlo test.  

To illustrate patterns of species distribution, ordination was conducted with non-metric multi-dimensional scaling (NMS; Kruskal1 1964, Mather 1976).  Because assumptions of linear relationships among variables and normality are not necessary with this method, NMS is well suited for community data.  Ordination was run using the Sørensen distance measure on the PC-ORD “slow and thorough” autopilot setting (maximum iterations=400; runs with real data=40; stability criterion=0.00001). 

Using the species matrix, a final 3-dimensional ordination was generated.  While a 2-D solution was preferred for ease of interpretation, the high stress of the 2-D solution (final stress=21.412) may have resulted in misinterpretation and thus rendered it less reliable than the 3-D solution.  A Monte Carlo test of statistical significance was used to test the validity of the final 3-D solution by comparison with 50 randomized runs (p=0.0196, final stress= 14.043, final instability=0.00001, 80 iterations). 

The ordination was rotated to maximize correlations of environmental variables (cover of overstory, understory, and exposed mineral soil) along a single axis, thereby facilitating interpretation.  To illustrate the relation of treatment to the community patterns, treatment types were overlaid on each ordination.  A vector overlay connecting sample units within the same year and block was also used to highlight differences among blocks and time and also to display segregation of treatments within each block. 

 

Results

Overstory

Thinning resulted in significant changes in the overstory conditions in each treatment that persisted four to six years following thinning.  In 2001, overstory cover in the H and LG treatments remained significantly lower than in the control (Figure 1; p = 0.0002).  While the L treatment did not differ from the LG treatment or the control, it did have significantly more cover than the H treatment ( p = 0.0002).  Though statistical tests have not yet been performed, canopy cover in all thinning treatments has increased since initial post-treatment conditions  (Figure 1).       

Figure 1:  Average cover of overstory trees ≥ 5 cm dbh for 1997 and 2001 (averages across four treatment areas).

 

Understory

Tall shrubs/small trees and low shrubs exhibited similar responses following thinning (Figures 2 and 3).  For these groups, all thinning treatments had significantly less cover than the control in 1997, but did not differ from each other (Tall Shrubs:  p = 0.0042; Low Shrubs: p < 0.0001).  However, by 2001, none of the thinning treatments differed significantly from each other or from the control treatment (Tall Shrubs:  p = 0.0639; Low Shrubs: p = 0.5421).

Figure 2:  Average species cover of tall shrubs and small trees for 1997 and 2001 (averages across four treatment areas).

 

Figure 3:  Average species cover of low shrubs for 1997 and 2001 (averages across four treatment areas).

 

For bryophytes, only the H treatment had significantly less cover than the control in 1997 (p = 0.0171; Figure 4).  Thinning treatments did not differ from each other in bryophyte cover in 1997.  By 2001, none of the treatments differed significantly (p = 0.7495).   

Figure 4:  Average cover of bryophytes for 1997 and 2001 (averages across four treatment areas).

 

Herbs responded somewhat differently than the other lifeform groups (Figure 5).  In 1997, thinning treatments did not differed statistically in herb cover from each other or from the control (p = 0.0763).  However, by 2001, the H treatment had significantly more herb cover than the control (p = 0.0412).  Neither the L nor the LGG treatments differed from the control in 2001. 

Figure 5:  Average species cover of herbs for 1997 and 2001 (averages across four treatment areas). 

 

Though cover of exotic species was low in all treatments, the LGG treatment did have significantly higher cover of exotics than the other thinning treatments and the control in 1997 (p = 0.0321; Figure 6).  By 2001, none of the treatments differed from each other or the control.    

Figure 6: Average species cover of exotics for 1997 and 2001 (averages across four treatment areas).

 

In the multivariate analysis, understory plant communities showed significant differentiation among treatment types (MRBP; 1997 and 2001 combined:  A=0.073, p<0.001; Only 1997:  A=0.105, p=0.021; Only 2001:  A=0.062, p=0.001), as well as a weak divergence over time in the thinned treatments (A= 0.123, p=0.055).  No difference between controls of 1997 and 2001 was present (A=0.081, p=0.1590). 

Ordination of the sample units in species space clearly illustrated the separation of treatments and years.  Of the total variation, 84 .4% was accounted for by the three axis in the ordination.  A 114° rotation was used to maximize alignment of environmental variables along Axis 2.  Accounting for 29.3% of the variation, Axis 2 (Figure 7) was strongly correlated with treatment and stand conditions, showing a strong positive relationship to overstory and understory tall shrub/small tree cover of 1997 and 2001 (r = +0.626 and r = +0.627, respectively) and a strong negative relationship to harvesting disturbance (i.e., exposed mineral soil in 1997, r = -0.584).  Overlays with the main matrix revealed that plants, such as bryophytes, Gaultheria shallon, and Chimaphila umbellata, (r > +0.6 for all) fell on the positive end of the second axis while plants, such as Epilobium watsonii, Cirsium spp., and Senecio sylvaticus, fell toward the negative end (r < -0.5 for all).   Axis 3, which was successful in capturing 36.2% of the variation, represented differences in species composition among treatment blocks.  The CR block was heavily dominated by Oxalis oregana, Montia sibirica, and Anaphalis margaritacea (r > +0.5 for all), while the CF block contained high abundances of Linnaea borealis, Rubus ursinus, Smilacina stellata, and Rubus nivalis (r < -0.5 for all), causing the two blocks to strongly differentiate along opposite ends of the axis.  The MC and SC blocks were similar in abundance of Anemone deltoidea, Trifolium latifolium, and Whipplea modesta, resulting in similar scores and an intermediate positioning for the third axis. 

Figure 7:  Ordination of treatment areas  in species space.  Vectors connect treatments within same block, same year.  'CR' = Cougar Reservoir Block; 'CF' = Christy Flats Block'; 'MC' = Mill Creek Block; 'SC' = Sidewalk Creek Block.  

 

Axis 1, which accounted for 21% of the total variation, illustrated compositional shifts of the communities over time (Figure 8).  Graminoid and bryophyte species as well as species such as W. modesta, Hieracium albiflorum, Berberis nervosa, R.  ursinus, Trientalis latifolia, and Satureja douglasii strongly correlated to the negative range of the first axis (r < -0.4 for all), while species such as S.  sylvaticus, Cirsium spp., Stachys mexicana, Rosa gymnocarpa, and Tiarella trifoliata var. unifoliata fell within the positive range of axis 1 (r > +0.3 for all).  

Figure 8: Ordination of treatment areas in species space.  Vectors connect 1997 sites to their corresponding 2001 sites.  'CR' = Cougar Reservoir Block; 'CF' = Christy Flats Block; 'MC' = Mill Creek Block; 'SC' = Sidewalk Creek Block.

 

Indicator species analysis revealed a distinction of species among treatments similar to that noted along axis 2 (Table 3).  Between years, a separation of species comparable to axis 1 was also produced by analysis of indicator species (Table 4).  

 

Table 3Species representative of understory plant communities within each treatment 

(* p < 0.05;  ** p < 0.01)

TREATMENT

INDICATOR SPECIES

CONTROL

Aster canescens *

 

Berberis nervosa *

 

Blechnum spicant *

 

Boykinia elata *

 

Chimaphila umbellata *

 

Goodyera oblongifolia *

 

Rubus nivalis *

 

Trillium ovatum *

 

Bryophytes **

 

Cornus canadensis **

 

Asarum caudatum **

LIGHT

Acer circinatum *

 

Adiantum pedatum *

LIGHT GAPS -- EDGE

None Identified

LIGHT GAPS -- GAP

Cirsium vulgare *

 

Conyza canadensis *

 

Epilobium angustifolium *

 

Epilobium paniculatum *

 

Epilobium watsonii *

LIGHT GAPS -- MATRIX

Smilacina stellata **

 

Linnaea borealis *

 

 

Table 4: Species representative of understory plant communities within each year  

(* p < 0.05;  ** p < 0.01)

YEAR

INDICATOR SPECIES

1997

Aster canescens **

 

Cirsium spp. *

 

Rosa gymnocarpa *

 

Rubus parviflorus *

 

Senecio sylvaticus *

 

Stachys mexicana *

 

Tiarella trifoliata var. unifoliata *

 

Vaccinium parvifolium

 

Viola spp. *

2001

Achlys triphylla **

 

Asarum caudatum **

 

Asplenium viride *

 

Berberis nervosa *

 

Bryophytes *

 

Campanula scouleri *

 

Cirsium vulgare *

 

Epilobium angustifolium *

 

Fragaria virginiana *

 

Gnaphalium microcephalum *

 

Graminoid spp. *

 

Hieracium albiflorum *

 

Hypericum perforatum *

 

Lactuca muralis *

 

Linnaea borealis *

 

Polystichum munitum **

 

Rhododendron macrophyllum **

 

Rubus ursinus *

 

Satureja douglasii **

 

Stachys spp. *

 

Symphoricarpos mollis *

 

Trientalis latifolia *

 

Violoa oregana *

 

Viola sempervirens *

 

Whipplea modesta *

 

 

Discussion

Thinning of overstory trees in young Douglas-fir stands produced distinctive changes in the understory plant communities among treatments.  Immediately following treatment, all thinned treatments had significantly less cover of tall shrubs/small trees and low shrubs than the control, presumably due to damage in the harvesting operation.  However, the rapid recovery of these lifeform groups indicates that this initial decline was likely a result of harvest disturbance rather than a response to changes in resource conditions. 

While it is likely that bryophytes also suffered from harvest disturbance, much of their initial decline may have been due to alterations in microclimate.  Thinning probably resulted in decreased soil moisture due to increased solar radiation reaching the forest floor.  Because several bryophyte species generally depend on moist environments, this drop in moisture may have resulted in desiccation and initial decline of cover.  However, as the canopy closes, bryophytes seem to be recovering to levels similar to those found in the control.

As a group, herbs seemed to respond favorably to thinning.  Unlike other lifeform groups, they did not appear to suffer significantly from harvest disturbance nor did they exhibit initial declines due to altered microclimate.  The increase in cover 4-6 years following thinning indicates a gradual positive response likely due to increased resource availability. 

Pre-treatment conditions, simulated by the control, harbored species more sensitive to disturbance and indicative of shaded environments while species typical of high light environments and disturbed soil dominated the LGG treatments.  Opening of the canopy and thereby increasing resource availability likely permitted these more opportunistic species to quickly colonize and displace the less competitive species (Grime 1979).  Epilobium spp. and other similar early-successional species are capable of rapid seed dispersal, allowing them to quickly occupy a disturbed site with abundant exposed mineral soil and exclude other species possessing slower reproductive mechanisms (Meier et al. 1995).  However, the strong tendency for species composition to remain relatively stable among blocks (Figure 7, axis 3) illustrates the important influence pre-harvest conditions hold on post-treatment development (Hughes and Fahey 1991).    

Though inclusion of early-successional species may increase species diversity, the high degree of soil disturbance and complete opening of canopy within the gaps did permit limited introduction of problematic invasive and exotic species.  Within the LGG treatments, invasive and exotic species flourished as evident by the location of the LGG sites in the lowest negative range of axis 1.  It is possible that the relatively young understory plant communities in the pre-thinned forest were not robust enough to defend against these types of invasions following significant disturbance (Sakai et al. 2001).  In this sense, artificial gap formation may be counterproductive to the overall goal of accelerating late-successional species composition.  However, in this study, exotics were not dominant in any of the thinning treatments and none differed from the control after 4-6 years.  This may indicate a lack of seed source or that exotic invasion may be an immediate effect of thinning, but may not have long-term impacts on plant communities. 

Other thinning treatments (H, L, LGE, and LGM) did not show a strong differentiation.  Indicator species analysis revealed only a few species demonstrating strong associations with these treatments.  Among these, L. borealis, a plant capable of capitalizing on heterogeneous resource availability (Antos and Zobel 1984, Spies 1991) did strongly associate with the LGM treatment, signifying a variety of habitat conditions within this treatment.  However, within the species ordination (Figure 7) the treatments lie intermediately scattered along the second axis.  It may be that more subtle distinctions are occurring within these treatments that are obscured by the drastic distinctions between the C and the LGG treatments.  To investigate this possibility, future analysis may need to either exclude the C and LGG treatments in order to magnify any potential differences or examine within treatment variation more closely at the macroplot level.

Within all thinning treatments, conspicuous changes in plant communities occurred between the initial post-harvest conditions and conditions 4-6 years later.  Even within this brief post-harvest time, the decrease in treatment effects of the four years indicates treatments may be homogenizing rather than diversifying.  This may be, in part, to an overall shift in all treatments from pioneer species to a shrub-dominated understory.  As the overstory begins to close in and effects of soil disturbance begin to diminish, early-successional, shade-intolerant species are slowly being displaced by a shrub community of Berberis nervosa, R. ursinus, and W. modesta, along with more shade-tolerant herbs.  Combinations of these species along with “weedier” species in the 2001 sites suggest the understory plant communities are undergoing a transition from an extreme early-successional community structure to a more shrub dominated young–seral composition.   

 

Conclusions

Thinning successfully created a range of overstory densities resulting in a variation of understory light conditions.  Preliminary analysis of the understory vegetation response to the thinning treatments indicates that there is subtle differentiation of plant communities among all the thinning treatments, with the gap treatment and control being the most dissimilar.  Increased similarity of these communities over time indicates that a transition from an early-successional community to a more shrub-dominated community may be occurring.  Future analysis of the understory response will explore questions such as: what are the strongest patterns, what is driving the changes, and more in-depth examination of how the communities are changing over time.      

 

Acknowledgments

This study is made possible through cooperation from Oregon State University, University of Oregon, Cascade Center for Ecosystem Management, Willamette National Forest, and the USDA Forest Service Pacific Northwest Research Station.  We would also like to thank John Cissel, Maureen Duane, Steve Garman, James Mayo, Brenda McComb, John Tappeiner, Gabriel Tucker and numerous others for providing data and background information for the study.

 

Literature Cited

 

Alaback, P.B., and F.R. Herman.  1988.  Long-term response of understory vegetation to stand density in Picea-Tsuga forests.  Can. J. For. Res.  18: 1522-1530.

 

Antol, J. A., and C. B. Halpern.  1997.  Root system differences among species: implications for early successional changes in forests of western Oregon.  Am. Midl. Nat. 138: 97-108.

Bailey, J.D., and J. C. Tappeiner.  1998.  Effects of thinning on structural development in 40- to 100- year-old Douglas-fir stands in western Oregon.  For. Ecol. Mgmt.  108: 99-113.   

Carey, A. B., and M. L. Johnson. 1995. Small mammals in managed, naturally young, and old-growth forests. Ecol. Appl. 5:336-352.  

Dufrêne, M., and P. Legendre.  1997.  Species assemblages and indicator species:  the need for a flexible asymmetrical approach.  Ecol. Mono.  67: 345-366.

Franklin, J.F., and C.T. Dyrness.  1973.  The natural vegetation of Washington and Oregon.  USDA For. Ser. Gen. Tech. Rep.  PNW-GTR-8.  Pac. Northwest Res. Stn., Portland, OR.

Grime, J.P.  1979.  Plant Strategies and Vegetation Processes.  J. Wiley and Sons, New York.  222 pp.

 

Harrington, T.B., and M.B. Edwards.  1999.  Understory vegetation, resource availability, and litterfall responses to pine thinning and woody vegetation control in longleaf pine plantations.  Can. J. For. Res.  29: 1055-1064. 

 

He, F., and H.J. Barclay.  2000.  Long-term response of understory plant species to thinning and fertilization in a Douglas-fir plantation on southern Vancouver Island, British Columbia.  Can. J. For. Res.  30: 566-572. 

 

Hitchcock, C.L., and A. Cronquist.  1973.  Flora of the Pacific Northwest.  University of Washington Press, Washington, USA.

 

Hughes, J.W., and T.J. Fahey.  1991.  Colonization dynamics of herbs and shrubs in a disturbed northern hardwood forest.  J. Ecol.  79: 605-616.

 

Kruskal, J.B.  1964.  Nonmetric multidimensional scaling: a numerical method.  Psychometrica.  29: 115-129.

 

Mather, P.M.  1976.  Computational methods of multivariate analysis in physical geography.  J. Wiley and Sons, London.  532 pp.

 

McCune, B. and M.J. Mefford.  1999.  PC-ORD.  Multivariate Analysis of Ecological Data.  Version 4.0  MjM Software, Gleneden Beach, Oregon, USA. 

 

McCune, B., and J.B. Grace.  2002.  Analysis of Ecological Communities.  MjM Software, Gleneden Beach, Oregon, USA.  300 pp. 

 

Meier, A.J., S.P. Bratton, and D.C. Duffy.  1995.  Possible ecological mechanisms for loss of vernal-herb diversity in logged eastern deciduous forests.  Ecol. Appl.  5: 935-946.

 

Mielke, P.W., Jr.  1979.  On the asymptotic non-normality of null distributions of MRPP statistics.  Communications in Statistics A5: 1409-1424.

 

Sakai, A.K., F.W. Allendorf, J.S. Holt, D.M. Lodge, J. Molofsky, K.A. With, S. Baughman, R.J. Cabin, J.E. Cohen, and N.C. Ellstrand.  2001.  The population biology of invasive species.  Annu. Rev. Ecol. Syst.  32: 305-332.

 

Spies, T.A.  1991.  Plant species diversity and occurrence in young, mature, and old-growth Douglas-fir stands in western Oregon and Washington.  In: Wildlife and Vegetation of Unmanaged Douglas-fir Forests.  USDA For. Ser. Gen. Tech. Rep.  PNW-GTR-285, Pac. Northwest Res. Stn., Portland, OR. pp. 111-121.

 

Thomas, S.C., C.B. Halpern, D.A. Falk, D.A. Liguori, and K.A. Austin.  1999.  Plant diversity in managed forests: understory responses to thinning and fertilization.  Ecol. Appl.  9: 864-879.

 

Thysell D.R., and A.B. Carey.  2000.  Effects of forest management on understory and overstory vegetation: a retrospective study.  USDA For. Ser. Gen. Tech. Rep.  PNW-GTR-488, Pac. Northwest Res. Stn., Portland, OR.  ii+41 pp.