forest site classification for cultural plant harvest by tribal weavers

forest ecology

Forest Site Classification for Cultural PlantHarvest by Tribal Weavers CanInform ManagementS. Hummel and F.K. Lake

Do qualitative classifications of ecological conditions for harvesting culturally important forest plants correspondto quantitative differences among sites? To address this question, we blended scientific methods (SEK) andtraditional ecological knowledge (TEK) to identify conditions on sites considered good, marginal, or poor forharvesting the leaves of a plant (beargrass; Xerophyllum tenax) used in tribal basket weaving. We relied onvoluntary participation of six expert weavers, a stratified, randomized field sample, discriminant analysis (DA),a standardized color system, and paired t-tests. We accepted each weaver’s classification (good, marginal, orpoor) of forested sites for beargrass harvest and then measured forest and plant attributes on two plots at eachharvest area in each class (n � 72). The DA yielded descriptive but not predictive results. Coarse woody debris(CWD) levels and the number of trees (trees per acre [TPA]) differed significantly between good and poor sitesacross California, Oregon, and Washington, whereas basal area did not. Good sites had less CWD (P � 0.0360)and fewer TPA (P � 0.001) than poor sites. Variations in leaf color decreased as the site class for plant harvestimproved. Results reveal a crosswalk between ecological knowledge derived via SEK and TEK for culturallyimportant plants.

Keywords: forest structure, cultural plants, silviculture, basketry, traditional ecological knowledge (TEK)

S ome plants used by American Indiansare scarce or declining in westernNorth American forests, so informa-

tion about how plant quantity and qualityrelate to site conditions is needed to changethis trend. Our premise is that the best in-formation for sustaining culturally impor-tant plants will come from studies that link

scientific and traditional ecological knowl-edge. Such an approach implies blendingthese types of knowledge to advance under-standing of relations among forest site con-ditions, cultural practices, and land manage-ment. In this article, we focus on oneunderstory plant but contend that a blendedapproach will apply to others. Moreover, a

blended approach could contribute to effi-ciencies in forestry practices, such as manag-ing tree density or using prescribed fire, thatare intended to support multiple objectives.

Traditional ecological knowledge (TEK),refers to a cumulative and evolving body ofknowledge, beliefs, and practices that aretransmitted culturally over generations (e.g.,Berkes and Folke 1998, Berkes et al. 2000,Turner 2001, Menzies and Butler 2006).Other terms are used for such intergenera-tional, cultural knowledge, including indig-enous knowledge and traditional ecologicalknowledge and wisdom (Turner et al. 2000,Corsiglia 2006). We adopt the term TEKfor this article, with the understanding thatit implies knowing a plant within a broadercontext of its landscape, history of place, siteconditions, harvesting practices, and a mul-titude of cultural uses.

Culturally important plants, with asuitable condition for a specific need can bedifficult to locate, access, and harvest (An-derson and Rowney 1999, Deur and Turner

Received November 21, 2013; accepted September 30, 2014; published online November 6, 2014.

Affiliations: S. Hummel ([email protected]), USDA Forest Service, Pacific Northwest Research Station, Portland, OR. F.K. Lake ([email protected]), USDAForest Service.

Acknowledgments: We express our appreciation to each tribal weaver who participated in the study. We thank Breanna Gervais for assistance with fieldwork and forcontributions to Figure 3, Pat Cunningham for assistance with multivariate data analysis, Beth Bambrick for contributions to Figure 4, and Julie Johnson forcontributions to Figure 2 and Table 1. The USDA Forest Service, Pacific Northwest Research Station provided financial support for this research through the CRAGprogram.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Noncommercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Forcommercial reuse, please contact the Journal of Forestry at [email protected].

RESEARCH ARTICLE

30 Journal of Forestry • January 2015

J. For. 113(1):30–39http://dx.doi.org/10.5849/jof.13-082

Copyright © 2015 The Author(s)

2005). Difficulties are compounded whenthe tribal members seeking cultural plantsare elderly or otherwise have reduced mobil-ity because plants suitable for a planned usemay not be at sites they consider good forharvesting. At the same time, their knowl-edge is vital. Who better than experiencedtribal elders and practitioners to teach nov-ice harvesters and youth how to locate po-tential sites, use appropriate harvesting strat-egies, and match plant properties withintended cultural uses? It requires knowingwhat plants to harvest, in which locations,under what environmental conditions, andwith what timing, intensity, and methods.Moreover, there may be traditional or cus-tomary restrictions on who can be involvedor how. Such tribal knowledge and practicesrequire familiarity with a place and its eco-logical processes, plus knowledge of plantproperties and the ability to transform rawnatural materials into an intricate item like abasket.

TEK is not easily shared beyond thecontext of a specific tribal harvesting prac-tice because much of it is tacit, meaning thatit is implied or indicated but not necessar-ily directly expressed (Turner 2001).Nonetheless, if it has relevance for landmanagement, it needs to be generalizedwhen possible and communicated accord-ingly. A role exists for scientific methods,therefore, that can support inferences madeabout the scope of knowledge obtained fromtraditional sources without requiring tribalpractitioners to divulge specific practices.Hence, we used quantitative and qualitativeresearch methods together with TEK to clas-sify and describe site conditions for harvest-ing the leaves of one culturally importantplant. The issue is not solely maintainingplant populations, but sustaining them withneeded quality in places accessible for cul-tural harvests. We use the term “quality” forplant and site properties that are suitable forthe specific purposes of harvesting and weav-ing.

ObjectivesThe main literature on TEK empha-

sizes using it to inform educational pro-grams (e.g., Huntington 2000, Snively andCorsiglia 2001, Kimmerer 2002) and eco-logical restoration activities (e.g., Andersonand Rowney 1999, Shebitz 2005, Senos etal. 2006) or to interpret ecological condi-tions (e.g., Turner et al. 2000). A subset ofthis literature aims to combine TEK and sci-

entific ecological knowledge (SEK). Wecontribute to this small subset by expandingits application in forestry (e.g., Mason et al.2012, Emery et al. 2014).

Our specific objectives in this studywere 2-fold, namely, to develop and applyresearch methods that combine ecologicalknowledge gained via SEK and TEK and todocument the biophysical characteristics ofsites that tribal weavers identified as good,marginal, or poor for harvesting beargrass(Xerophyllum tenax) for tribal basketry.Across our multistate study area, numeroustribes are known to use beargrass in weavingboth twined and coiled baskets (Table 1;Figure 1). It is plausible that classification ofsite conditions for plant harvest could varywith tribal affiliation or weaving style, giventhe variation in how beargrass is used withdesign weaves by different tribes (Table 1).However, some tribes historically wove bas-kets with beargrass obtained in trade (Zobel2002), which means there may be desirableleaf properties that transcend a specificweaving tradition and instead relate directlyto site conditions. Our expectation was thatmeasurable differences exist among forestsites rated by tribal weavers as favorable ornot for harvesting beargrass, independent oftribe and weaving style. To test this hypoth-esis, we relied on the voluntary participationof tribal weavers, a stratified and randomizedfield sample, and a standardized color systemfor plant tissue.

MethodsWe used both quantitative and qualita-

tive methods to combine SEK and TEK forour first objective and then we applied this

blended methodology to learn about bear-grass. For our second objective, we docu-mented the characteristics of sites thattribal weavers classified as good, marginal,or poor for harvesting beargrass by evalu-ating and describing their significant dif-ferences. Using beargrass as an example,the following series of numbered stepsserve as a methodological framework forblending TEK and SEK for other cultur-ally important species.

Step 1. Plant Natural and CulturalHistory Identification

We used existing, written documenta-tion about beargrass to guide the selection ofvariables for field sampling. Particular focuswas given to plant life history and leaf prop-erties desired for weaving baskets. Beargrassis a lily-like plant; it grows in a wide varietyof habitat types and conditions, but in justtwo geographic areas. One area is maritime,from the mountains of northwestern Wash-ington south into western-central Califor-nia, whereas the other is continental, fromCanada south into Wyoming along theRocky Mountains. Its only congeneric rela-tive, eastern turkeybeard (Xerophyllum as-phodeloides), is restricted to the southeasternUnited States where it is classified as threat-ened in portions of its range. Beargrass re-produces both by seed and by sprouting andis adapted to survive disturbances, such asfire and landslides, if the rhizome is unaf-fected. It is found at its highest densitiesunder canopy openings in a variety of foresttypes, but evidence is scarce about the effectsof competition or light on its reproductionand growth. Within the maritime region

Management and Policy Implications

Different types of ecological knowledge can contribute to identifying forest management objectives forculturally important plants. Study results imply that managing tree density can provide understoryconditions to promote the growth of beargrass leaves preferred by tribal weavers. Results also imply thatmanaging down wood levels is important so that coarse woody debris (CWD) (�3 in.) is not a physicalbarrier for tribal harvesters. Across sites in California, Oregon, and Washington, average tree BA per acredid not differ significantly, ranging between 175 and 197 ft2/acre. On sites considered good for beargrassharvest, however, the BA was distributed on fewer trees (average 127) than it was on sites the tribalweavers identified as marginal (average 177) or poor for harvest (average 172). A similar trend existedfor deadwood. Namely, the average amount of total surface wood and litter increased as the siteclassification declined in quality, with good � 14 tons/acre, marginal � 19 tons/acre, and poor � 22tons/acre. Taken together, the forest conditions identified by tribal weavers as good for harvestingbeargrass had, on average, significantly fewer, larger diameter trees and less surface wood, litter, andCWD than did poor sites. The structural elements preferred by tribal weavers for beargrass harvest,therefore, relate directly to those associated with managing fire behavior in similar forest types.

Journal of Forestry • January 2015 31

Figure 1. Examples of ancestral and contemporary baskets (see Table 1). Top row: Coiled baskets. Primary traditional weaving materialsshown: western red cedar (Thuja plicata) and beargrass (Xerophyllum tenax). Left: unknown Klickitat artist, Basket, ca. 1870/1880.Courtesy of Portland Art Museum, Portland, OR. Museum Purchase: Helen Thurston Ayer Fund. Accession number 40.35.21. Middle: Tribalweaver harvesting beargrass, 2012. Courtesy of Frank Lake. Right: Nettie Jackson (Klickitat), Large Cedar Root Berry Basket, 1983.Courtesy of Maryhill Museum of Art, Goldendale, WA. Gift of Mary Dodds Schlick, in memory of William T. (Bud) Schlick, 1925–1992.Accession number 2010.03.001. Bottom row: Twined baskets. Primary traditional weaving materials shown: hazel (Corylus cornuta),bear-grass (Xerophyllum tenax), alder bark-dyed woodwardia fern (Woodwardia fimbriata) and maidenhair fern (Adiantum pedatum).Left: unknown Karuk artist, Woman’s Hat, ca. 1900. Courtesy of Portland Art Museum, Portland, OR. Gift of Miss Mary Forbush Failing.Accession number 18.2.8. Middle: Twined basket, 2012. Courtesy of Frank Lake. Right: unknown Klamath artist, Tobacco Basket, ca. 1950.Courtesy of Portland Art Museum, Portland, OR. The Elizabeth Cole Butler Collection. Accession number 2012.25.22.

Table 1. Examples of tribes (California, Oregon, and Washington) using beargrass in the design weave of twined and coiled basketry.

Design weave (beargrass)

Twined Coiled

Full-twistoverlay

Half-twistoverlay Overlay

Wrappedtwine

Falseembroidery Imbricated

California Hat CreekModocPit RiverShastaWintun

HupaKarukPit RiverTolowaWiyotYurok

Yana

Oregon KalapuyaUmatilla

SiletzTakelmaTututniUmatilla

Siletz ClatsopUmatilla

TillamookUmatilla

Washington Klickitat ChehalisQuinaultSkokomish

ChinookNisquallyPuyallupQuileuteSkokomish

ChinookMakahQuileute

ChehalisKlickitatNisquallyPuyallupYakama

Example of basket (and date)of each design weave(beargrass)

CAS 0090-0015*(early 1900s)†

CAS 0473–0023*(ca. 1977)†

2013.10.44*(ca. 1840)‡

NA 26*(ca. 1830–1840)�

Turnbaugh andTurnbaugh,1986§

2005.001*(ca. 2004)�

* Museum accession number.† California Academy of Sciences, San Francisco, CA: Anthropology collection database. Available online at research.calacademy.org/anthro/collections; last accessed May 15, 2014.‡ Portland Art Museum, Portland OR. 2014. Native American art collection. Available online at www.portlandartmuseum.us/mwebcgi/mweb.exe?request�home; last accessed May 15, 2014.§ Peabody-Turnbaugh, S., and W.A. Turnbaugh. 1986. Indian baskets. Schiffer Publishing, Atglen, PA. 264 p.� Hallie Ford Museum of Art, Salem, OR: Native American collection. Available online at willamette.edu/arts/hfma/collections/native_american/index.html; last accessed May 15, 2014.


http://research.calacademy.org/anthro/collections

http://www.portlandartmuseum.us/mwebcgi/mweb.exe?request=home

http://willamette.edu/arts/hfma/collections/native_american/index.html

in particular, many American Indian tribesharvest beargrass for cultural purposes, theprimary use being basketry with a twined orcoiled foundation (Table 1). Beargrass leavesare durable yet flexible, so they can be wo-ven, braided, or wrapped tightly. Evidencesuggests that tribal basket weavers favor lon-ger, thinner, more pliable leaves with lesspigmentation and a snowy white color at thebase. Leaves with these properties are oftenassociated with postfire conditions (Hum-mel et al. 2012).

Step 2. Preliminary Site IdentificationSome areas of interest to land managers

and tribal harvesters where beargrass grew wereidentified before field reconnaissance. Thebiophysical characteristics of the sites weredescribed from existing soil maps, elevation,fire or forest management history, and generallocation near roads or other features. Road ac-cessibility to sites with high densities of bear-grass was desired by many tribal gatherers. Weused a rule set that included distance from aroad, forest vegetation community type, pub-lic ownership, fire history, and personalknowledge of the area to identify potentialsites. We aimed to include a geographic rangeon the Pacific Coast from northwestern Cali-fornia to central Washington.

Step 3. Participant Recruitment andFinal Site Selection

Expert weavers were identified by referral,with our main sources being contemporaryweavers, basketweaver associations, schools,and community organizations. Six weaversfrom three states and four tribes agreed to par-ticipate in the study. They are not identifiedmore specifically to uphold confidentiality.We consulted with tribal cultural resource spe-cialists and federal policies to confirm thatknowledge being shared by tribal memberswas not subject to regulation. Moreover, weagreed that research sites would remain confi-dential if requested by tribal participants (USCTitle 25 § 30561). It is important that scien-tists learn in advance about existing policies,laws, and regulations that affect any culturalplants to be studied, including protection oftribal TEK and confidentiality of sites.

We selected our final study sites in Cal-ifornia, Oregon, and Washington (Figure 2)based on information from the tribal partic-ipants and associated land use history,including, for example, ancestral or cededterritories, past or present forest or fire man-agement history or related activities in coop-eration with tribes, land use designation, site

accessibility, and physical stamina or mobil-ity of participants. Trees species were pre-dominantly coniferous, including Douglas-fir (Pseudotsuga menziesii), grand fir (Abiesgrandis), western hemlock (Tsuga hetero-phylla), subalpine fir (Abies lasiocarpa), pon-derosa pine (Pinus ponderosa), and westernlarch (Larix occidentalis).

Step 4. Field SamplingInput from both written and oral

sources helped us identify variables of poten-tial importance for classifying sites by har-

vest quality. For our blended research ap-proach, we elected to accept each weaver’squalitative classification of a site as good,marginal, or poor for harvesting beargrassand then used field methods adapted fromecology and forestry to sample the site andplant variables. Our conceptual model inthis study can, therefore, be described asTEK � SEK � latent variables � error.This approach acknowledges the existenceof traditional knowledge that may not bemeasurable but, nonetheless, quantifies somesite attributes that may be important in the

Figure 2. Study area (California, Oregon, and Washington) for field sampling good,marginal, and poor sites for harvesting beargrass for basketry.


identification of forest management objectivesand planning management activities.

For each variable, we evaluated differ-ent sampling methods by considering theirsuitability and availability to land managers.This was an important step, because a vari-ety of methods exist and all choices involvedtradeoffs. For example, percent canopycover can potentially affect beargrass growthand resultant leaf quality, but this variablecan be estimated ocularly (with high bias,greater error, and lower price) or measuredby using a spherical densitometer (moderatebias, minor error, and moderate price) orlight spectrum instrument (low bias, highprecision, and high price). Another variable,deadwood, can affect site access and it too,can be sampled with different methods thatinvolve tradeoffs in repeatability, bias, preci-sion, unit of measure, time, and cost to doc-ument. We selected particular instrumentsor methods that matched limitations im-posed by time and budget to standardizemethods across the range of conditions weexpected to encounter.

Each of the six weavers was driven todifferent locations and asked to offer a pre-liminary assessment of sites as good, mar-ginal, or poor for harvesting beargrass. Aftera range of sampling sites was seen, each par-ticipant was asked to personally check thequality of leaves on some individual plants ata site to make their final determination.Across each of the three site types selected(�6–25 acres), we marked out six patches(100–500 ft2) that represented conditionsin which the weaver would harvest beargrass.We then threw a die to select two of the sixpatches at random and established two sam-ple plots per patch. This resulted in a totalnumber of 12 sample plots per weaver (3 sitetypes � 2 patches � 2 plots). Thus, for thesix weavers we sampled 72 plots.

The sample plots were circular, with aradius of 45 ft. On each plot, the followingsite attributes were measured or estimated:diameter of each tree, height of tallest tree,deadwood in tons per acre by size class (us-ing the Maxwell and Ward 1980 visualphoto series), trees per acre (TPA), percentslope at plot center (clinometer), and per-cent canopy cover at four cardinal directions(spherical concave densitometer). In addi-tion, several plant attributes were measured.To do so, we first divided each plot into fourquadrats by extending two measuring tapesin cardinal directions. By beginning at duenorth and walking toward plot center, wesampled the first five beargrass plants en-

countered along the tape and extending 18in. on either side. If five beargrass plantswere not encountered on the first transect,we continued by beginning at due eastand walking toward plot center until fiveplants were sampled. On each sampled

plant, the code that most closely corre-sponded to the color at the midpoint of lon-ger leaves of a central whorl was assignedusing the Munsell plant tissue color guide(Munsell 2011). We tallied the number ofemergent whorls of new leaf growth because

Figure 3. Decision key created from field visits with tribal weavers to study sites (California,Oregon, and Washington).


they contain the best basketry materialand thus are targeted for harvest by tribalweavers.

In addition to plot measurements, a de-cision key was created from informationshared by the weavers during field visits (Fig-ure 3). The key illustrates the considerationsthese women and men gave to certain siteand plant conditions.

Step 5. Preliminary Data SummariesWe summarized preliminary results

from the field study, which we then pre-sented during an informal panel at the 2012Northwest Native American BasketweaversAssociation (NNABA) meeting in Auburn,Washington. The panel included everytribal study participant and USDA ForestService research staff. The lead agency scien-tist, who described the purpose and objec-tives of the study, was the only nontriballyaffiliated panel member. Panel observers in-cluded interested weavers attending theNNABA meeting. Our research goals for thepanel were to determine whether the weav-ers felt we had measured the right variablesand had not omitted something important.In qualitative research, this step is referred toas “member checking” (Janesick 2000). Ouroutreach goal was to honor the contributionof study participants by sharing some pre-liminary results. We learned several thingsfrom the weaver panel, namely, (1) there wasinterest and discussion about down wood(e.g., amount, size, and distribution) as a sitedescriptor, (2) no key variables were identi-fied as having been omitted from the fieldstudy, (3) it was easier for weavers to de-scribe and identify “good” and “poor” sitecharacteristics than “marginal,” and (4) par-ticipants appreciated that we brought infor-mation to them.

The NNABA weaver panel reinforcedthe value of the blended approach we used inthe study because traditional and scientificknowledge each made a contribution. Re-searchers learned that the weavers who par-ticipated in the study had classified sitesbased on personal integration of many fac-tors, including what they saw in the field andwhat they had previously observed, beentold, or experienced. They did not distillsite and plant qualities into individualvariables. Plant qualities, when consideredgood enough to harvest, were evaluated inrelation to the site conditions and proximityto better plants. Nonetheless, tribal partici-pants were interested in how the researchersidentified and measured variables because

they felt it could help them describe to nov-ice weavers what they were seeing whenidentifying the characteristics of a good site.In turn, the researchers learned from theweavers why some seemingly importantvariables might indeed affect their harvestcriteria. For example, the amount and size ofdown wood could affect ease of physical siteaccess, harvester mobility, and speed of leafcollection. Hence, down wood may be botha direct factor affecting potential fire behav-ior and an indirect factor affecting culturaluse harvesting efficiency. Importantly, it is avariable that can be quantified and man-aged. We chose to investigate both the totalamount of down wood and the subset ofdown wood �3 in., which we distinguish ascoarse woody debris (CWD) in this article.Without the interaction of researchers andweavers, less would have been learned aboutthe potential importance of down wood bysize class and arrangement on site classifica-tion for cultural harvest than we learned incombination. The NNABA panel also con-firmed the merits of multivariate analysis toevaluate the qualitative classes because theweavers themselves relied on a multivariateassessment of each site.

Data AnalysisWe used the measured variables in dis-

criminant function analysis (DA) (SAS In-stitute, Inc. 2000), either directly (e.g.,down wood by size class) or calculated (e.g.,basal area [BA]), to predict class member-ship by using continuous or binary indepen-dent variables. In ecology, DA is also useddescriptively (Williams 1983). A multivari-ate analysis of variance, such as Wilks’ � thatwe selected, tests the statistical significance

of mean differences of the groups. We con-firmed that our data were not highly corre-lated (�0.7) and assumed multivariate nor-mality. Our original interest in the potentialrole of leaf color in site classification was in-creased by weaver comments during mem-ber checking at the NNABA panel. In thesystem of color theory we used, hue is acombination alphanumeric scale, whereaschroma and value are numeric (Munsell2011). The Munsell system is qualitative, inthat it uses perceptive color space instead ofa quantitative measure of visible light(Viscarra Rossel et al. 2005). All of ourrecorded hues were in the same alphabeticclass of “green-yellow” (GY), so we usedonly the numeric component of the huescale in the DA.

We identified the combination of site,tree, size classes of deadwood, and leaf vari-ables that maximized the number of cor-rectly classified good sites by using the resub-stitution method in DA; the list of finalvariables is shown in Table 2. The site fea-tures of longitude and latitude were not inthe final model, although elevation was in-cluded. From the final complete set, we ex-cluded the color “triplet” of hue, chroma,and value to evaluate the effect on classifica-tion error and on distance between classes.Resubstitution uses the fitted discriminantfunction to reclassify the sample data thatwere used to fit the discriminant function inthe first place. The effect is that the model is“validated” with in-sample observationsonly.

The results of the DA were not statisti-cally significant, yet identified categories ofimportant structural and ecological vari-

Table 2. Variables from discriminant analysis of good, marginal, and poor sites(California, Oregon, and Washington) for harvesting beargrass for basketry.

Variable Good Marginal Poor

SiteElevation (ft) 4,152 (340) 4,097 (426) 4,123 (410)

TreeBA (ft2/acre) 197 (175) 181 (148) 175 (144)Total trees (TPA) 127 (57) 177 (105) 172 (67)

DeadwoodResidue (in.) 0.20 (0.10) 0.30 (0.20) 0.20 (0.10)Duff/litter (in.) 1.14 (0.90) 1.08 (0.70) 1.20 (0.70)Size class 1 (�1 in. tons/ac) 0.95 (0.66) 1.30 (0.67) 1.40 (1.04)Size class 3 (1–3 in. tons/ac) 1.40 (0.55) 2.10 (1.09) 1.90 (0.96)Size class 9 (3–9 in. tons/ac) 5.30 (2.00) 9.80 (6.40) 10.70 (06.5)

Leaf colorHue 5.08 (0.22) 5.18 (0.51) 5.48 (0.94)Chroma 4.95 (0.52) 5.03 (0.70) 4.91 (0.74)Value 4.14 (0.27) 4.04 (0.32) 4.10 (0.38)

Data represent means (SD).


ables, including tree (BA and TPA), downwood (total surface), and leaf color. We usedpaired t-tests to evaluate differences inmeans between the individual tree and downwood variables and harvest site classification(Husch et al. 2003). We also added a test ofmeans for CWD because of the discussionabout it at the NNABA panel.

ResultsAcross all weaving styles and forest

types represented in the study, there wasconsistency in the leaf color that tribal par-ticipants associated with good harvestingsites for beargrass. When leaf color wasmainly within the middle hue (5GY) andhad few instances of different hues, the site

was generally judged to be good (Figure 4).In contrast, several instances outside of themiddle hue downgraded weaver judgmentof the site. This color consistency held trueeven though leaves were harvested from dif-ferent whorl layers on individual plants ac-cording to different tribal weaving styles.

The average BA (ft2/acre) was similaracross all site classifications, but it was dis-tributed on fewer average TPA on the goodsites (Figure 5). That is, sites classified byweavers as marginal or poor had, on average,more TPA than sites classified as good. Asimilar trend existed for deadwood. Goodsites had fewer TPA (P � 0.001) and lessCWD (P � 0.036) than poor sites. The av-erage amount of total dead surface wood intons per acre increased as the site classifica-tion declined in quality (good � 14, SD 5;marginal � 19, SD 10; and poor � 22, SD11). The average CWD (�3 in.) amountswere good (12 tons/acre), marginal (17 tons/acre), and poor (20 tons/acre). The 3–9 in.size class of deadwood included in the model(Table 2) and important to weavers (Figure6) followed the same trend.

Evidence was lacking for statisticallysignificant mean differences in our threeclasses (Wilks’ � � 0.475, P � 0. 55). Theclassification summary from the full modelreturned the highest number of correctlygrouped sites. The highest classification ratewas for good (83.3%), followed by marginaland poor (66.7%) (Table 3). The distancefrom good to marginal was equal to the dis-tance from marginal to poor (2.2). The colortriplet changed these distances and alteredthe relationship among sites. Without color,marginal sites were judged to be closer topoor sites than they were to good sites. Colorwas the least variable on good sites, becom-ing more variable as quality was judged todecline (Figure 4). Despite this trend, everyclass had instances in which it was misclassi-fied as either of the other two (Table 3).

DiscussionThis study suggests how land managers

could combine management objectives forforest overstory and understory plants whendetermining target levels for key stand struc-tural elements, namely, by identifying theforest conditions associated with desired un-derstory plant properties and then incorpo-rating this information into silvicultural, fu-els, and fire management activities. In themixed-conifer forest types included in thisstudy, for example, a body of literature existsabout relations among size class distribu-

Figure 4. Distribution of beargrass leaf color on sites (California, Oregon, and Washington)classified by tribal weavers as good, marginal, or poor for harvesting beargrass forbasketry.


tions of trees and down wood and fire behav-ior and severity (e.g., Hummel and Agee2003). In particular, fire modifies foreststructure along a gradient, thinning smallerto larger trees and thin- to thicker-barkedspecies as intensity increases. This findingimplies that managing beargrass sites forfewer, larger trees and lower levels of down

wood in all size classes could moderate fu-ture fire severity and contribute to good con-ditions for tribal harvesting of leaves forweaving.

We used the example of beargrass, butother culturally important plants lend them-selves to a similar approach. Our study dem-onstrates a crosswalk between ecological

knowledge derived empirically via scientificmethods and via TEK because clear differ-ences between good and poor harvestingsites were identified by both. This result im-plies that general guidelines can be devel-oped to aid in managing specific sites forculturally important plants as required byfederal law. The blended approach we devel-oped and applied demonstrates that SEKcan be advanced by combining qualitativeand quantitative methods and that TEK canbe generalized using scientific methods. Thisinvolves paying attention to the units andmethods of measure for different variables ofinterest used by tribal practitioners and howthey are similar to or different from what isunderstood, described, and used by scien-tists or land managers.

The importance of beargrass leaf colorfor both tribal and commercial harvestershas been reported elsewhere (Hummel et al.2012). Evidence suggests that overstory con-ditions are related to leaf color, with a densecanopy casting more shade and thus creatingconditions for darker leaf pigmentation(Schlosser and Blatner 1997). To ourknowledge, this study is the first to investi-gate leaf color by using a standardized sys-tem and, in so doing, document its consis-tency across several forest types and tribalweaving techniques. In this respect, our re-sults are informative beyond identifying theleaf color desired for weaving: they can berelated to the levels of CWD, BA, and TPAon sites considered good for harvesting bear-grass. The implication is that some optimalrange of BA/TPA exists that is suitable forbeargrass to develop leaves of the desiredcolor for weaving. Although our results pro-vide information about conditions in thisrange, we cannot yet identify the entirerange itself, which would require data frommore years, sites, and tribal weavers.

Despite a consistency in color prefer-ence across all sites and weavers, we were

Figure 5. Average BA and TPA on sites (California, Oregon, and Washington) classified bytribal weavers as good, marginal, or poor for harvesting beargrass for basketry. *Signif-icant difference (� � 0.05) between good and poor sites.

Figure 6. Dead surface wood (tons/acre) on sites (California, Oregon, and Washington)classified by tribal weavers as good, marginal, or poor for harvesting beargrass forbasketry. *Significant difference (� � 0.05) between good and poor sites.

Table 3. Classification rates fromdiscriminant analysis of good, marginal,and poor sites (California, Oregon, andWashington) for harvesting beargrass forbasketry.

Good Marginal Poor Total

Good 10 (83.3%) 1 (8.3%) 1 (8.3%) 12Marginal 2 (16.7%) 8 (66.7%) 2 (16.7%) 12Poor 2 (16.7%) 2 (16.7%) 8 (66.7%) 12Total 14 11 11 36

Error rate 0.17 0.33 0.33 0.28


unable to identify any combination of vari-ables that explained statistically significantmean differences among good, marginal,and poor sites as classified by the weavers.Although it was possible to correctly identifyclass membership more than 66% of thetime for our sample, it was not possible topredict it. The main reasons for this out-come probably relate to the size, timing, andscale of our sample. It is possible, but lesslikely, that we excluded important variables,given our blended approach and the mem-ber checking that we undertook with tribalparticipants. We consider each of these as-pects in turn in the following paragraphs.

Size of SampleOur original intent was to have a larger

number of expert weavers from more tribesand thus the ability to sample more sites andplots within each harvest suitability class.We found it difficult to enroll participants,however, due to negative views of or experi-ence with the USDA Forest Service, con-cerns about disclosing proprietary culturalresource information, and agency restric-tions on travel. Our final set of participantsmay not have represented all relevant groupswithin the population of tribal weavers.Given the variability we encountered, the fi-nal sample size limited our ability to use DAto identify significant differences amongclasses or to create predictive models. Wewere, however, able to identify statisticallysignificant differences in TPA and CWD be-tween good and poor sites because we in-cluded SEK methods with TEK. The study,therefore, demonstrates how to advancecommunication and knowledge among dif-ferent ways of describing the forest struc-tural conditions desired for different man-agement objectives. Any future research likethis that blends qualitative and quantitativemethods to study culturally importantplants will benefit by ensuring that the sam-ple of tribal participants covers all relevantsocial variables and that the number of sitesand plots sampled covers the expected vari-ation in key sites and plant attributes.

Timing of SampleOur sampling occurred during just one

summer, which meant that plant growthand leaf properties did not cover the entireamplitude of available conditions. For ex-ample, although it has been documentedthat leaf properties associated with burnedsites are desirable for weaving, we did notexplicitly address the role of fire history in

our analysis of site condition. We observedthat weavers’ identification of a good sitechanged more than did that of a poor site,based on what was available in our samplingyear. The weavers wanted to see all the po-tential sites before deciding which belongedin which class. Among the weavers fromnorthwestern California there was a markedpreference to harvest beargrass from sitesburned within the past 2 years, whereas thiswas less of a factor for tribal weavers farthernorth. Areas or sites considered “good” or“marginal” one year could be better or worsein other years, depending on interveningfire, management, or other factors such asdrought and summer temperature that affectthe population of beargrass.

Scale of SampleIn the process of conducting the study,

it became apparent that tribal gatherers usedtwo scales of assessing the site before actualharvesting. The first condition was physical,meaning ease of access to the forest sitewhere beargrass grew. The second was thecondition or quality of beargrass at the site.Our attempt to cross-walk SEK samplingmethods with TEK logic and practices in-volved uniting different assessment tech-niques. Sometimes in field conversations withstudy participants, TEK information was ob-tained that SEK sampling alone would nothave revealed. For example, weavers comingfrom geographically distinct basketry tradi-tions (e.g., from Northwestern Californiatwined to Columbia Plateau coiled) (Table1) expressed different preferences for the leafsize (e.g., leaf blade base width) and forwhich leaves (e.g., internal compared to ex-ternal) were harvested and used from thewhorls. In addition, we observed that leaveswith higher moisture content were valued byweavers during harvesting as determined bythe “squeaky” sound accompanying extrac-tion. Although not quantified in this study,moisture content can influence color.

It was also true that our SEK approachprovided a new way for the weavers to thinkabout how they assessed a site. Sometimes itwas evident that they went to sites becausethey were shown those sites by someone elseand so had never thought about what madeit good other than that recommendationand/or they had never tried to deconstructwhat they observed at a site; they just knew itwas good enough (or not). The importanceof down wood in different size classes was afactor not explicitly considered yet clearlyinfluential, particularly with respect to how

much time was spent harvesting comparedwith how much usable material was ob-tained. Some of the classification errormight relate to differences in microsites be-tween sampled plots and the site in whichthe beargrass was located.

In addition to the aforementioned sam-pling issues, there are ongoing social and en-vironmental issues related to the growth andharvest of cultural resources that providecontext for this and any subsequent fieldstudies that blend SEK and TEK. The bestresearch on specific, culturally importantplants will be a product of cooperation be-tween scientists and tribal practitioners. Inthis study we learned that compensating theparticipants for their time and travel re-quired different types of agreements or con-tracts. For example, suitable cultural plantsare not always located on tribal lands sothere is a need for tribal members to accessand harvest botanical resources elsewhere.Going somewhere to harvest is not strictlyan issue of rights and access, however, but isalso negotiated through complex interper-sonal and intertribal relationships. For ex-ample, if personnel on a national forest con-duct a project to improve or enhancebasketry materials, many tribal weavers willstill seek “permission” or make “payment”to the recognized tribal elder or senior basketweaver on whose traditional territory theproject has taken place. Further, weaversmay be of mixed tribal and ethnic ancestry,with enrollment affiliation with one tribebut cultural and family ties with other tribalareas. Respect and reciprocity to follow tra-ditional customs may or may not be compat-ible with modern forest management and ju-risdiction. One direct impact is that weaversmay have to travel from their home or reser-vation to their ancestral lands or to an areawhere access and suitable basketry plant ma-terials occur. This means that those living atgreater distance from their ancestral territorywill incur greater expenses and need to takemore time to participate in forest manage-ment planning, consultations, coordination,and collaboration and to harvest valuedplant resources. In addition to these andother social issues, a key environmental issueis the presence of other culturally importantplants at the same site. Importantly, TEK israrely single species focused but is biophysi-cally and metaphysically related. Eventhough beargrass was our target species forresearch and harvesting, tribal participantsoften evaluated, harvested, or tended toother plants in the vicinity. The totality of


the gathering experience includes the de-sired plants and the surrounding landscape,which underscores the appropriateness ofusing multivariate methods. Evidence fromthis study supports our premise that the bestinformation for sustaining culturally impor-tant plants will come from studies that linkscientific and traditional ecological knowl-edge. Scientists working with tribal practi-tioners and land managers working to in-crease access to suitable plant materials needto be mindful of factors that perpetuate tra-ditional practices, including those involvedin harvesting.

Endnote1. US Code (USC), Title 25, Chapter 32A, Sec-

tion 3056: Prohibition on disclosure.

Literature CitedANDERSON, K., AND D. ROWNEY. 1999. The edi-

ble plant Dichelostemma capitatum: Its vegeta-tive reproduction response to different indige-nous harvesting regimes in California. Restor.Ecol. 7(3):231–240.

BERKES, F., AND C. FOLKE. 1998. Linking socialand ecological systems for resilience and sus-tainability. P. 1–25 in Linking social and eco-logical systems-management practices and socialmechanisms for building resilience, Berkes, F.,and C. Folke (eds.). Cambridge Univ. Press,Cambridge, UK.

BERKES, F., J. COLDING, AND C. FOLKE. 2000.Rediscovery of traditional ecological knowl-edge as adaptive management. Ecol. Appl.10(5):1251–1262.

BUTLER, C. 2006. Historicizing indigenousknowledge-practical and political issues. P.107–126 in Traditional ecological knowledgeand natural resource management, Menzies,C.R. (ed.). Univ. of Nebraska Press, Lin-coln, NE.

CORSIGLIA, J. 2006. Traditional wisdom aspracticed and transmitted in NW BC, Can-ada. P. 221–235 in Traditional ecologicalknowledge and natural resource management,Menzies, C.R. (ed.). Univ. of NebraskaPress, Lincoln, NE.

DEUR, D., AND N.J. TURNER. 2005. Keeping itliving: Traditions of plant use and cultivation onthe Northwest coast of North America. Univ. ofWashington Press, Seattle, WA. 404 p.

EMERY, M.R., A. WROBEL, M.H. HANSEN, M.DOCKRY, W.K. MOSER, K.J. STARK, AND J.H.GILBERT. 2014. Using traditional ecologicalknowledge as a basis for targeted forest invento-ries: Paper birch (Betula papyrifera) in the USGreat Lakes region. J. For. 112(2):207–214.

HUMMEL, S., AND J.K. AGEE. 2003. Westernspruce budworm defoliation effects on foreststructure and potential fire behavior. Northw.Sci. 77(2):159–169.

HUMMEL, S., S. FOLTZ-JORDON, AND S. PO-LASKY. 2012. Natural and cultural history ofbeargrass (Xerophyllum tenax). USDA For.Serv., Gen. Tech. Rep. PNW-GTR-864, Pa-cific Northwest Research Station, Portland,OR. 80 p.

HUNTINGTON, H. 2000. Using traditional eco-logical knowledge in science: Methods andapplication. Ecol. Appl. 10(5):1270 –1274.

HUSCH, B., T.W. BEERS, AND J.A. KERSHAW JR.2003. Forest mensuration, 4th ed. John Wiley& Sons, Inc., Hoboken, NJ. 456 p.

JANESICK, V.J. 2000. The choreography of quali-tative research design: Minuets, improvisa-tions and crystallization. P. 393 in Handbookof qualitative research, Denzin, N.K., and Y.S.Lincoln (eds.). Sage Publications, ThousandOaks, CA.

KIMMERER, R.W. 2002. Weaving traditionalecological knowledge into a biological edu-cation: A call to action. BioScience 52(5):432– 438.

MASON, L., G. WHITE, G. MORISHIMA, E. AL-VARADO, L. ANDREW, F. CLARK, M. DURGLO, ET

AL. 2012. Listening and learning from traditionalknowledge and Western science: A dialogue oncontemporary challenges of forest health andwildfire. J. For. 110(4):187–193.

MAXWELL, W.G., AND F.R. WARD. 1980. Photoseries for quantifying natural forest residues incommon vegetation types of the Pacific North-west. USDA For. Serv., Gen. Tech. Rep.PNW-GTR-105, Pacific Northwest ResearchStation, Portland, OR. 229 p.

MENZIES, C.R., AND C. BUTLER. 2006. Introduc-tion. P. 1–17 in Traditional ecological knowl-edge and natural resource management, Men-

zies, C.R. (ed.). Univ. of Nebraska Press,Lincoln, NE.

MUNSELL, A.H. 2011. Munsell color charts forplant tissues. Munsell Color and X-rite, GrandRapids, MI. 20 p.

SAS INSTITUTE, INC. 2000. The SAS system: SASOnlineDoc, version 8.2, HTML [CD-ROM].SAS Institute, Inc., Cary, NC.

SCHLOSSER, W.E., AND K.A. BLATNER. 1997. Spe-cial forest products: An eastside perspective.USDA For. Serv., Gen. Tech. Rep. PNW-GTR-380, Pacific Northwest Research Sta-tion, Portland, OR. 27 p.

SENOS, R., F.K. LAKE, N. TURNER, AND D. MAR-TINEZ. 2006. Traditional ecological knowledgeand restoration practice. P. 393–426 in Restor-ing the Pacific Northwest: The art and science ofecological restoration in Cascadia, Apostol, D.,and M. Sinclair (eds.). Island Press, Washing-ton, DC.

SHEBITZ, D. 2005. Weaving traditional ecologi-cal knowledge into the restoration of basketryplants. J. Ecol. Anthropol. 9(1):51–68.

SNIVELY, G., AND J. CORSIGLIA. 2001. Discover-ing indigenous science: Implications for sci-ence education. Sci. Ed. 85(1):6–34.

TURNER, N., M.B. IGNACE, AND R. IGNACE.2000. Traditional ecological knowledge andwisdom of Aboriginal peoples in British Co-lumbia. Ecol. Soc. Am. 10(5):1275–1287.

TURNER, N.J. 2001. “Keeping it living”: Applica-tions and relevance of traditional plant man-agement in British Columbia to sustainableharvesting of non-timber forest products. P.66–77 in Forest communities in the third mil-lennium: Linking research, business, and policytoward a sustainable non-timber forest productsector, Davidson-Hunt, I., L.C. Duchesne, andJ.C. Zasada (eds.). USDA For. Serv., Gen.Tech. Rep. NC-217, North Central ResearchStation, St. Paul, MN.

VISCARRA ROSSEL, R.A., B. MINASNY, P.ROUDIER, AND A.B. MCBRATNEY. 2005.Colour space models for soil science. Geo-derma 133(3):320–337.

WILLIAMS, B.K. 1983. Some observations on theuse of discriminant analysis in ecology. Ecology64(5):1283–1291.

ZOBEL, D.B. 2002. Ecosystem use by indigenouspeople in an Oregon coastal landscape.Northw. Sci. 76(4):304–314.


forest site classification for cultural plant harvest by tribal weavers

Documents