Great Basin Naturalist

Volume 55 Number 3 Article 5

7-21-1995

Ecology of Celtis reticulata in Idaho

Ann Marie DeBolt Bureau of Land Management, Boise, Idaho

Bruce McCune Oregon State University, Corvallis

Follow this and additional works at: https://scholarsarchive.byu.edu/gbn

Recommended Citation DeBolt, Ann Marie and McCune, Bruce (1995) "Ecology of Celtis reticulata in Idaho," Great Basin Naturalist: Vol. 55 : No. 3 , Article 5. Available at: https://scholarsarchive.byu.edu/gbn/vol55/iss3/5

This Article is brought to you for free and open access by the Western North American Naturalist Publications at BYU ScholarsArchive. It has been accepted for inclusion in Great Basin Naturalist by an authorized editor of BYU ScholarsArchive. For more information, please contact [email protected], [email protected]. Greal Basin Naturalist 55(3), IS) 1995, pp. 237-248

ECOWGY OF CELTIS RETICULATA IN IDAHO

Ann Marie DeBoltl and Bruce McCune2

ABSTRACT.-Tbe small deciduous tree Celtis reticulala (net[eaf hackberry) reaches its northem limit in Idaho, wbere, contrary to most ofits western range, it often occurs as an overstory dominant Two hundred flfty stands of this tree were sampled throughout Idaho. Celfu is slow-growing, averaging 4 m tall at 50 yr, and long-lived {to 300-400 yr). It occurs in a variety of habitats, from riparian to rocky uplands. Trees grow best where topographically sheltered, such as in draws and narrow canyons, and where soils are loamy. Although plants grow more slowly as surface rock cover increases, stands are often associated with ~k, with a mean surface cover of 39% rock. Differences in growth rates were unrelat· ed to parent material and aspect. Most stands are reproducing. in spite of habitat degradation caused by overgrazing. aben plant invasion, and increasing fire frequencies. Stands are typically represented. by one dominant cohort; however, young, even-aged stands are rare and are generaJly found along waterways on stream terraces or at the high-water )jne. Although slow-growmg. C. reticulata shows promise for land managers interested in site enhancement. nus native species is loog.lived, produces fruit used by wildlife, and provides structural diversity in a semiarid landscape {with a maximum height of12 m) in areas that are becoming mcreasingly dominated by exotic plant species.

Key u.>()r'(/,s; Celtis reticuLata, netleaf 1uJckherry, ecology, Idaho, growth. longevity, stand strocture., recruitment, site clwrMterisf;i.c$, livestock gt.azing,. rehabilitation.

Celtis retieulata Torr. (netleaf hackberry, occur as a locally abundant, overstory dominant western hackberry) is a deciduous shrub to (Huschle 1975, Johnson and Simon 1987). small tree in the elm family (Ulmaceae), wide­ Along the Wiley Reach of the middle Snake ly distributed in semiarid regions of the west­ River, it forms narrow, but extensive, gallery ern United Slates (Fig. 1). It occurs in a diver­ forests of nearly monospecific stands (Bowler sity of habitats, including deciduous riparian 1981). On steep shoreline escarpments of the woodlands, mountain shrub, wash scrub, and lower reaches of the Snake River, in the live oak-mixed shrub communities, in rocky "Douglas" hackberry vegetation type described canyons, and as scattered individuals in semi­ by Huschle (1975), it forms a dense, nearly desert grasslands, pinyon-juniper and Joshua closed canopy. On the gentle shoreline slopes, tree woodlands (Glinski 1977, Plummer 1977, alluvial fans, and colluvial cones of the lower Brown 1982, Albee et aI. 1988). Its elevational Snake River, it grows in an open savanna range is from 200 to 2000 m (Elias 1980). (Daubenmire 1970, Huschle 1975). "Open sa­ Populations are often small or highly localized vanna is perhaps the hest way to describe the (Daubenmire 1970, Dooley and Collins 1984), appearance ofa typical Celtis community on an particularly at the northerly latitudes in the upland site in Idaho, where individuals occnr states ofOregon, Washington, and Idaho (Eliot in a random or clumped pattern with exten­ 1938). Despite its broad distribution, little is sive areas ofgrassland between. known about the plant's ecology, presumably Plants produce a small, fleshy drupe in the due to its position as a minor component in WI, favored by a variety ofbirds and mammals many of its habitats, and its fragmented occur­ (Hayward 1948, Larmer 1983, C. A. Johnson rence (Peattie 1953, Larmer 1983). 1990, personal communication). With as many While C. retiJ:;"lata is sparsely distributed as 41 species of birds associated with Celtis in Idabo, near its northern limit (Fig. 2), it communities in Idaho, the tree's importance for appears to exhibit wide ecological tolerances. wildlife cannot be overemphasized (Asherin Howevel; it tends toward the wannest portions and Claar 1976). It provides cover for a variety ofcanyons, especially southerly aspects (TIsdale of big game species, including mule deer and 1986). It is a member of both riparian and bighorn sheep (Asherin and Claar 1976), as upland communities in Idaho, where it can well as much-sought-after shade for domestic

'Bureau or Land M3I\11gelO£ot, 3948 Deve!opmeut AYeOUe, Boi5e, lD 83705. Z~plU1l'(1ent of Botany and PJarlt Pathology, Oreg:ln Sh.le University, ~Iv.llli" 01197331·2002.

237 238 GREAT BASIN NATURALIST [Volume 55

- INTERMITTENT DISTRIBUTION ... CONTINUOUS DlSTRmUTION

100 km

MOSCOW

•

•

-- --

• • •

Fig. 1. Glohal distribution of Ct:ltis reticulata (revised from lillIe ]976). TWIN. \ FALlS .lOll IlIt livestock along the Snake River (Daubenmire Fig. 2. Idaho distribution ofCeltis n~tictlltJtn. 1970). Due to an apparent tolerance of harsh. water­ stressed growing (..anditions, a strong potential METHODS to resprout followin~ disturbances such as fire and herbivory, and its high wildlife values, Field Methods public land managers are interested in using 'Rvo hundred thirty stands spread over much C. reticulata to rehabilitate disturbed habitats. of ~,e Idaho range of C. "eticulata were sampled However, we must know more of the growth in 1990 and 1991. Approximately 20 stands on rate, longevity, stand structure, and ec'OlogicaJ the west side of the Snake Rive); in adjacent tolerances of the species to properly evaluate Oregon and Washington, were also sampled its potential in site enhancement or rehabilita­ (total N = 250). Stands were selected based on tion projects. within·site homogeneity of apparent history, This study sought to answer the following topography, and parent material, and a mini­ questions: (1) What are the growth rates and mum population size of six individuals (many longevities of C. retie,data, and do they differ more individuals were usually present). With with aspect, parent material, soil texture, per­ these constraints for homogeneity, the sam­ cent surface rock cover, topographic position, pling areas were typically irregularly shaped topographic shelter, and grazing intensity of a and small, usually less than 0.25 ha. Stands stand? (2) How does the size class structure of were chosen to represent a range of sites and C. reticulaia slaods diller with the environ­ disturbance histoloies. mental parameters Jisted above? Is the species Stands were assigned to topographic posi­ reproducing in Idaho, and does recruitment tions (Table I) that included river terrace, high­ differ under different environmental conditions? water line, ilia",', rocky draw, bench, toe slope, (3) Are environmental conditions related to lower slope, broken lower slope, mid-slope, differences in growth form of the plant (i.e., uppcr slope, and talus. The II categories were the formation ofsingle vs. multiple stems)? narrowly defined on the assumption that 1995] CELTIS RETICULATA IN IDAHO 239

TABLE L Definitions of topographic positions in which size of rhyolite (4) and its chemical similarity Celtis reticuWta was sampled. to granite, the two were combined for analy­ River terrace Relatively flat horizontal surface cut or sis. A similar situation existed for oolitic lime­ built by river or stream action stone, an uncommon and geographically High-water line Transition line between flood-tolerant restricted coarse-grained rock that typically and -intolerant plant species Draw Shallow incision in a slope, with <30% occurred as a lens within sandstone-dominated total surface rock cover strata. Therefore, the eight stands on oolitic Rocky draw Shallow incision in a slope, with limestone were combined with sandstone for >30% total surface rock cover analysis. Bench Nearly level surface usually well above active floodplains and terraces Each stand was categorized by "topograph­ Toe slope Gently inclined, basal part ofa slope ic shelter": open (0), intermediate (1), and shel­ continuum that grades to the valley, tered (2). For example, exposed stands grow­ 0 usually <14 slope ing within a valley were classified as "interme­ Lower slope Lower 1/3 ofa hillside (above the toe 0 diate," while stands growing in a side canyon of slope, when present); ifsteep (> 14 ) and toe slope absent, the basal part of the same valley were classified as "sheltered." the slope that meets the valley floor "Open" stands were those with unobstructed Broken lower Similar to lower slope but with exposure to solar radiation. They were typically slope extensive surface cover oflarge boulders and outcrops not associated with a major, incised drainage; Mid-slope Middle 1/3 ofa hillside, relative to the rather, they faced broad, expansive valleys. surrounding landscape To evaluate recruitment and growth of C. Upper slope Upper 1/3 ofa hillside, relative to the relieulata under different livestock grazing surrounding landscape Talus slope Coarse, angular rock fragments derived pressures, we scored grazing intensity within from and lying at the base ofcliffs or a stand as none to moderate (1) or extreme (2). rock slopes; slopes typically >25° Stands scored as extreme were recognized by (1) heavy browsing of trees, with a hedged or "pasture-tree" growth form; (2) elimination of vegetation under trees by trampling; (3) tree combining them at a later time, ifneeded, would roots exposed by soil compaction and erosion; be possible. Based on field observation and and (4) dominance of alien plant species. reconnaissance, the number of stands sampled Thirty-six of the 250 stands were classified as within each topographic position was approxi­ extreme, mately proportionate to how frequently those The overall density ofCeltis stands was cat­ topographic positions were occupied by tlle egorized as (1) widely scattered [mature indi­ species. Stand-level data recorded, in addition viduals more than 10 crown widths apart]; (2) to topographic position, included elevation; scattered [mature individuals separated by latitude; longitude; aspect; slope; percent sur­ gaps of4--10 individual crown widths]; (3) sub­ face rock cover; surface soil texture; parent continuous [breaks in the total canopy exist but material; topographic shelter, grazing intensi­ mature individuals average no more than 3 ty, total stand density; density within four crown widths apart]; or (4) continuous [little structural classes, including seedling, juvenile, open space in the canopy; crowns form a con­ mature, and decadent individuals; number of tinuous matrix with occasional gaps]. Inter­ cohort modes; and associated dominant plant mediate sites were recognized with a mid­ species (explained below). point value (e.g., 3.5 for staods approaching a Surface soil textures were evaluated by moist­ closed canopy). ening in the field according to the Soil Conser­ To evaluate the composition of C. reticulata vation Service "Guide for Textural Classificaton" stands, densities in four structural classes were (Brady 1974). When soils were unreachable due also estimated in a similar fashion. The four to surface rock, the surface rock matrix was structural classes were defined as follows: (1) classified instead. For example, stands on talus seedling [individual of the year and < 2 yr slopes had soils categorized as "talus." old]; (2) juvenile [individual >2 yr old and Six categories ofparent material were iden­ < 1.5 m tall]; (3) mature [>1.5 m tall]; and (4) tified initially, including granite, sandstone, decadent [> 1.5 m tall and experiencing signif­ basalt, river alluvium, rhyolite, and oolitic lime­ icant dieback, i,e" at least one major dead stone, However, because of the small sample branch present]. 240 GREAT BASIN NATURALIST [Volume 55

Within each stand at least three individu­ radial growth. Nine stands were dropped, for als, chosen to represent the modal size in the a final sample size of241. SPSS (1988) was used stand, were measured and aged. Modal size for all analyses. was defined as typical size ofindividuals in the A heat load index was generated to account dominant (most abundant) cohort. Measure­ for diflerences in heat load from northeast- to ments recorded for each tree included height, southwest-facing slopes (Whittaker 1960, Muir age, diameter at core height (typically 20 em and Lotan 1985). For each stand, index values above ground level), number of live and dead were calculated with the following equation, stems, and percent rock cover below the can­ where e = aspect in radians east of north: opy as centered over the main trunk. When heat load = (L - cos(8 - 110/4))/2. index values two or three modal sizes were present, all ranged from 0 (NE slopes) to 1 (SW slopes). modes were sampled for a minimum total of To compare C. retieulala growth rate and either six or nine individuals. When stands stand structure differences under various en­ were all-aged with no apparent modal tree vironmental conditions, we developed 50-yr size, at least six individuals of the dominant site indices as measures of growth potential canopy cohort were sampled. The number of (i.e., site quality), as outlined in Husch et a!' modes present, from 1 to 4, with 4 equivalent (1972). Site index is based on average heigbts to an all-aged stand, was recorded as a stand­ of dominant trees at a specified index age level variable. Most height measurements (usually 50 or 100 yr) and is the most widely were obtained with an 8-m, extendable level used method of evaluating site quality for tree rod. For taller trees, height was determined growth (Husch et aI. 1972, Daubenmire 1976). with a clinometer, Site index curves are constructed to allow for Increment cores were taken at the same estimation of site index for stands older or height the diameter was measured (20 em). younger than the index age, as index age Cores were transported in plastic straws, glued stands are seldom encountered (Rusch et al. onto slotted boards, sanded, and annual growth 1972). rings were counted under a dissecting micro­ The commonly used relationship of tree scope. When cores did not reach the tree's height to age formed the basis for one index, center (i,e., because of rot), the number of and the relationship of tree diameter to age missing years was extrapolated by first sub­ formed the basis for the second (DeBolt 1992). tracting the length of the core from the tree's The best linear fit was achieved when log radius. This remainder was multiplied by the (height, m) and log (diameter, em) were number of rings counted in the core's inner regressed on the log of tree age (R2 = .25, R2 centimeter, which was then added to the num­ = .54, respectively; N = 939). The resulting = X ber of rings counted for an estimate of the equations were log (height) 0.428 log (age) - 0.135 and log (diameter) = 0.764 X log total age. When cores were off-center, the miss­ (age) - 0.165. Using these two equations, we ing radius was estimated by overlaying a clear obtained the expected (mean) height and transparency with a series of circles of known diameter at 50 yr, then back-transformed to radii over the core, matching the ring pattern improve interpretability, yielding an expected in the core with a circle, and multiplying its size at 50 yr of3.9 m tall and 13.6 em in diam­ radius by the number of rings in the centime­ eter. ter nearest the core's center. This amount was For each tree in the data set, the site index added to the number of counted years to esti­ was calculated by first finding its residual from mate tree age. Small-diameter noncoreable the regression line, then shifting this residual individuals were cut down and a cross section to the 50-yr point on the line, which yields an was removed, sanded, and the rings counted estimated height and diameter at 50 yr. Thus, as above. tbe equations to calculate site index (SI) for Analytical Methods each tree were:

Stands were not included in the analysis if Log (height SI) = 0.591 + the sample size within a particular topograph­ (LOGheight - «0.428 X LOGage) - 0.135)) ic position or parent material was too small, or if the majority ofcores from a stand were illeg­ Log (diameter SI) = 1.134 + ible after sanding due to contortions in the (LOGdiam - ((0.764 X LOGage) - 0.165)) 1995] CELTIS RETICULATA IN IDAHO 241

To analyze structural class differences under between environmental variables and stand differing environmental conditions, the vari­ structure (TYPE) and the number of modes able TYPE, representing types of stand struc­ were analyzed by contingency tables and ture, was created. Based on the density of ANOVA. juvenile, mature, and decadent size classes in a stand, the five TYPEs were defined as fol­ RESULTS lows: (1) young (juvenile); (2) mature, nonre­ Growth producing, nondecadent; (3) mature, repro­ ducing, nondecadent; (4) mature, reproducing, Log-log regressions best represented the decadent; and (5) mature, nonreproducing, statistical relationship between height and age decadent (Table 2). (Fig. 3) and diameter and age of C. relieulata Based on field observations, mortality of C. individuals. An initial impression that regres­ reticulata seedlings during year one is extreme­ sion lines do not fit the scatter of points at ly high. Because most seedlings were year­ log(age) < 1.2 can be reconciled by recognizing lings, seedlings were not used to deflne TYPE. that the dense central elliptical clouds ofpoints Stands were classified as reproducing when have controlled the regression results. In both the juvenile density class was 1 or greater (i.e., cases the least-squares fit resulted in a good fit > 5 individuals). to the dense cloud ofpoints representing mid­ Celtis relieulata growth rate, expressed by dle-aged trees, but resulted in almost entirely site indices, was analyzed as the dependent vari­ negative residuals for trees younger than able in one-way analyses ofvariance (ANOVA) 10-25 yr. Because these younger trees were against the environmental parameters topo­ from a small number of sites, many of which graphic position, parent material, soil texture, showed battering by floods, distributions of grazing intensity, and topographic shelter. residuals were judged to be acceptable. Relationships between site indices and ordered Celtis reticulata diameter and height were categorical independent variables were ana­ tightly related in a log-log regression (R2 = lyzed by linear regression. vVith few excep­ .75). Mean height and diameter of dominant tions, height site index was a more sensitive and codominant C. reticulata, regardless of predictor of growth differences than diameter age, were 5 m and 18 em, respectively. Wbile site index. Celtis reticulata growth rates and diameter is a better predictor of age than relationships with topographic position and height (R2 = .53 and .25, respectively), height other environmental parameters were also is more responsive to site characteristics than analyzed with analysis of covariance, to com­ is diameter, both in the literature and in this bine categorical and continuous factors. study. Thus, height was the preferred basis for Included in the model was the categorical vari­ the site index. able topographic position, with soil texture, Fifty-year-old C. relieulala trees in Idaho topographic shelter, grazing intensity, and par­ averaged 3.9 m tall and 13.6 em in diameter. ent material as four covariates. Relationships Using height, we constructed site index cunres

TABLE 2. Categorization ofthe Celtis reti~ulata stand structure variable TYPE. TYPE represents the five types ofstand structure that were recognized from the density classification. Within each stand, the three size classes of trees (juve­ nile. mature, decadent) were assigned to a density class based on the following dofinitions. Mid-point values were used as needed. Juvenile: (1) widely scattered-5 or fewer juveniles present; (2) scattered- >5 juveniles present in H nonag­ gregated distribution averaging> 10 canopies apart; (3) subcontinuous-breaks in the total canopy exist hut juveniles average >3 and < 10 canopies apart. MaturelDecadent: (1) widely scattered-mature individuals> 10 crown widths apart; (2) scattered-mature individuals separated by gaps of >4 and <10 individual crown widths; (3) subcontinuous­ breaks in the total canopy exist but mature individuals average < 3 crown widths apart; (4) continuous-mature trees form a continuous matrix with only occasional gaps.

Density class TYPE Description Juvenile Mature Decadent 1 Young > 1 :5: .5 <2 2 Nonreproducing, nondecadent :5: .5 >.5 ,;2 3 Reproducing, nondecaJent , 1 >1 <2 4 Reproducing, decadent , 1 >1 >2 5 Nonreproducing:, decadent <1 >.5 >2 242 GREAT BASIN NATURALIST [Volume 55

15 32 stands (13%) were located on the coolest 0 8 sites between 0.00 and 0.20, or between 350 Site index and 98 0 east ofnorth. The mean heat load index 7 Height at 50 years was 0.69. No stands were found between 349 0 6 0 10 and 9 east ofnorth. In spite of C. reticulata's affinity for souther­ 5 ly exposures, heat load was not a good predictor 4 ofhackberry growth characteristics. More often than not, stands have an affinity for southerly 5 exposures, but because oftopographic shelter­ , ing, growing conditions are often not as harsh or water stressed as they first appear. Of 241 1 Celtis stands, 168 (70%) had at least an inter­ 0 mediate topographic shelter. In a stepwise regression analysis from a 0 100 200 300 400 pool of six independent variables (soil texture, AGE, years rock, grazing intensity, shelter, heat load, and Fig, 3. Nontransformed log~log regression of Celtis slope), shelter was the most important predictor reticulata height (m) on age and site index curves for the of site index (R2 ~ .13, P < .001, F ~ 35.5). Idaho stands. Site index values were largest when shelter was greatest, with well-sheltered stands differ­ ing from intermediate and open exposures for Idaho Celtis stands to allow site classifica­ (Table 4). However, variability in growth rates tion for a stand at any age (Fig. 3). Site quality within a given class of shelter is large, as of an area can be assessed by determining shown by the low R2. average height and age of dominant trees and Presence of C. reticulata is correlated with locating the position of these coordinates on surface rock or rock outcrops. Ofthe 241 stands the site index graph. The stand's site index is sampled, 96 (40%) had a surface rock cover of then read from the closest curve. 50% or more (Fig. 4B). Twenty percent of the Site quality, as expressed hy the height­ stands were extremely rocky, vvith rock cover­ based site index, differed among the eleven ing 75-98% of the ground surface. Average topographic positions identified (P ~ .0001, F rock cover was 39%, ~ 4.4) (Table 3). However, variation within A weak, inverse relationship between per­ topographic positions was large, so that at the cent surface rock cover and site index was .05 significance level, only draws differed found (R2 ~ -.28, P ~ .0001). As rock cover in­ from any other specific topographic position. creased, site index tended to decrease slightly. Growth was faster in draws than on talus Rock was a statistically significant variable in a slopes, upper slopes, mid-slopes, and stream stepwise multiple regression as well, follovving terraces. topographic shelter in order of entry. Although site index means did not differ Including rock in the model increased the R2 statistically hetween most topographic positions, value from .13 to .20 (F ~ 28.9, P < .001). On a relatively predictable biological ranking of sites classified as draws, where topographic topographic positions was expressed, with a shelter is maximized, surface rock cover is less general trend offaster growth where sheltered important. and mesic to slower growth on more xeric and Neither parent material nor grazing inten­ exposed sites. For example, site index values sity was a statistically significant predictor of were smallest on talus slopes, followed by upper site index (P ~ .43 and .14, respectively). How­ slopes, mid-slopes, and stream terraces (Table 3). ever, site index values differed with soil tex­ Celtis reticulata occurred infrequently on ture (P ~ .023, F ~ 2.07). As with topographic north- and east-facing slopes (Fig. 4A). Twenty­ position (Table 3), means were ranked by five percent (60) of stands were found on SW Fisher's LSD procedure in an intuitively pre­ slopes, with a heat load between 0.95 and dictable order. Growth rates were higher on 1.00, the hottest values of the heat load index; finer-textured soils (clay or loam) than on 58% (140) were between 0.74 and 1.00. Only coarse-textured soils (sand). At alpha ~ .05, the 1995J CELTIS RETICU/ATA IN IDAHO 243

TABLE 3. Site index values ofCeltis reticulata (5 ;::: standard deviation) for each topographic position. Mean site index (51) values have been b'ansformed back into the original scale of measurement to aid interpretation. Topographic posi~ tions with no overlap of similarity groupmg letters are different from each other at the .05 significance level (Fisher's LSD). Mean 51: Mean $1: Topographic transformed back- Similarity position (s) transformed N grouping D...w 0.74 (0.16) 5.5 30 A High.water line 0.65 (0.15) 4.4 37 AB Toe slope 0.61 (0.14) 4.1 17 AB Rocky d.-aw 0.58 (0.17) 3.8 15 AB Lower slope 0.57 (0.19) 3.7 40 AB Bench 0.56 (0.24) 3.6 16 AB Broken lower slope 0.55 (0.12) 3.5 20 AB Stream terrace 0.51 (0.31) 3,2 13 BC Mid.slope 0.50 (0.18) 3.2 28 BC Upper slope 0.48 (0.20) 3.0 12 BC Talus slope 0.47 (0.13) 2.9 13 BC

only pairs thai differed from each other were were absent from most individuals. The num­ talus and loam. ber of dead stems at mid-slope (x = 0.6) was Interactions between soil texture and topo­ greater than all other topographic positions ex­ graphic position were highly significant (Chi­ cept upper slopes (P = .0001, F = 6.5). Stands square, P = .001). When the analysis of site at high-water line, rocky draw, stream terrace, index and soil texture was restricted to just draw, and broken lower slope topographic upland siles, the effect was slightly more pro­ positions averaged only 0.1 dead stems per nounced (P = .014, F = 2.49). individual. Growth form did not differ with the number ofsize modes within a stand. Growth Form Individuals on sandstone were more com­ "Shrubbiness" was quantified by counting monly multi-stemmed than those on the three the number of live and dead main stems or other parent materials, for both living and trunks of each individual. Regression analysis dead stems (P < .001, F =8.5; P < .001, F = of stem number with the variables grazing 14.7, respectively). intensity, topographic shelter, soil texture, heat Longevity load, slope, average height, average diameter, and percent surface rock cover produced sev­ The mean age of individuals sampled dur­ eral statistically significant, albeit weak, rela­ ing our study was 66 yr, with a range of 1-374 tionships. Live and dead stem density per yr (Fig. 5). Old age and large size are not tight­ individual decreased as topographic shelter ly related. For example, it is common to find increased (R2 = .20 and .30, respectively). trees 10 m tall but less than 75 yr old. Average height decreased slightly as the num­ Diameter was often a better predictor of age ber oflive stems increased (R2 = .20). In gen­ than was height (R2 = .54 and .26, respective­ eral, on sheltered sites C. reticulata has a sin­ ly, after log-log transformation). gle stem (treelike) rather than multi-stem The oldest C. retioulata recorded in this study (shrublike) growth form. (about 374 yr) grew on an exposed talus slope Differences in plant growth form were found approximately 300 m above the Salmon River; among topograpbic positions and among par­ it was 4.6 m tall and 48 cm in diameter at 20 ent materials. Individuals growing at mid-slope em above ground level. Percent surface rock were generally shrubbier, with a greater num­ cover ofthe site was 90%, with the small stand ber of live stems (x = 2.5), than individuals of scattered trees restricted to talus margins growing at high-water line (x = 1.4), in draws where pockets of soil were exposed. Other (x = 1.6), and in rocky draws (x = 1.5) members of the stand ranged in age from 191 (ANOVA, P = .003, F = 2.71). Dead stems yr (3.35 m tall, 28 em diam) to 320 yr (5.48 m were far less numerous than live stems and tall, 46.5 em diam). 244 GREAT BASIN NATURALIST [Volume 55

70 60

60 A 50 B

50 40 m m 0 I 0 ~ 40 ~ m ...m 30 ...0 0 ~ w~ w 20 '"~ 20 '"~ "z "z 10 10 1~8~~~~~~m~m~1 0 II 0

- 10 :-:-----:':-----:'':---:-'-:--:'"c---:'::---c'::--.J -10 '---'---'-----'-----'---'-----'---' -0.2 0.0 0.2 0.4 0.6 0-6 1.0 1.2 -20 0 20 40 60 80 100 120 HEATLOAD PERcENT SURfACE ROCK

Fig. 4. Frequency djstributiolls of the number ofCeltis reticulata stands by (A) heat load and (8) percent surface rock cover.

Stand Structure line. Rock}' draws were never a<;signed an over~ Of the 241 Celtis stands, 178 (74%) were re­ all density <2.5, where 3 = subcontinuous. In producing and only 4 (1.7%) ofthese were deca­ lact, 75% of rocky draws had closed or nearly dent. Fifty-seven stands (23.5%) were classified dosed canopies (overall density = 3.5 or 4). as nunreproducing, 6 (2.5%) of them decadent. Juveniles were often present on the margins of The remaining 6 stands (2.5%) were recently rocky draws. established (juvenile dominated), with no The few decadent stands were found higher mature individuals present. on the slope, on steeper slopes, and in less­ Structure of C. reticulata stands, in terms of sheltered positions than nandecadent stands. their relative densities of juvenile and mature Nonreproducing, decadent stands were more size classes, was unrelated to soil texture (Chi­ steeply sloping than young and nondeC'ddent, square, P = .31). Structural type was weakly nonreproducing stands (ANOVA, P = .003, F related to topographic position of the stand = 4.03) (Table 5). Of the 10 decadent stands, (Chi.square, P = .08). Of the II topographic 50% were at mid-slope and 20% were on talus. positions, rocky draws had the highest juve­ While none ofthe decadent stands were ex­ nile density, or recruitment. In general, juve­ tremely overgrazed. their distance from water nile densities increased as the percent of sur­ may have confounded this result. Overgrazed face rock cover increased. Density of C. retie­ stands were typically found on fairly gentle ulata juveniles was highest when rock cover terrain (x = 14·, S.D. =8.6) and in close prox­ was 50% or greater. imity to a water source, where livestock tend Rocky draws consistently had the densest to concentrate, while decadent stands were on canopies, followed by draws and high-water steeper slopes (Table 5) and at higher slope

TABU!: 4. Mean site index (51) values for Celtis reticufau, for three levels of topographic shelter, in both transformed and back-transformed scales. Topographic sbelters with no overlap ofsimilarity grouping letters are different from e'dch other at the .05 significance level (Fisher's LSD).

Mean 51: MC

, >0 TABLE 5. Average slopes of decadent, nondecadent, and young stands of Celtis reticulata, with the variable TYPE we in its original five-category format. TYPE represents the five types of Celtis reticulata stand structure that were recognized. 8C TYPE N Mean slope S.D. w "w (degrees) 60 " "0 Young Guvenile) 6 12 3.4 •"w ;,> " Nonreproducing, nonclecaclent 51 20 9.6 >0 Reproducing, nonclecadent 174 22 10.3 ~~~~D.. Reproducing, 0 0 • decadent 4 27 7.7 Nonreproducing, ->0 decadent 6 33 5.0 -100 Hm '00 ,co ,co TREE AGE (year$)

Fig. 5. Frequency distribution of the number of Celtis 0 1'etkulata trees by tree age. (x ~ 12 ). All had at least 15% surface rock cover, but most had 75% or greater rock cover (x = 65%). Five of the six stands were on allu­ positions. Less intensively grazed stands aver­ vium, including stream terraces, high-water 0 aged 23 (S.D. = 10.1). lines, and henches. All five had sandy soil. The Grazing level was related to stand structure sixth stand was atypical, occurring near a mid­ (TYPE; Chi-square, P = .0002). A larger per­ slope, sparsely vegetated band of sandstone centage of heavily grazed stands (53%) were with intermediate shelter. All individuals were nonreproducing than were stands with light or shrubby, decadent, and old (18-33 yr) relative moderate grazing intensity (18.5%). Even to the average height of0.7 m (expected age = though sample sizes were very different (light 8 yr). Soils were sandy loam in texture. or moderate = 205, extreme ~ 36), the pattern While young stands were only on sites with confirms field observations of low recruitment intermediate topographic shelter, reproducing under extreme grazing pressure. However, it and nonreproducing stands differed little in is perhaps even more noteworthy that recruit­ the degree of shelter they received (Chi­ ment on heavily grazed sites is as high as it is, square, P = .06). Thirty-three percent ofrepro­ given how few, if any, other shrub species are ducing stands were sheltered, compared to 25% present on such sites. ofnonreproducing stands. Among the four parent materials, 37% of C. The amount of surface rock differed weakly across stand structure (TYPE; ANOVA, P = reticulata stands growing on sandstone were .038, F = 2.58). Differences were greater when nonreproducing, as compared to 21%, 22%, the variable TYPE was restructured to three and 21% of stands growing on granite, basalt, categories (mature reproducing, mature non­ and river alluvium, respectively (Chi-square, reproducing, young), eliminating decadence P = Table 6). A greater numher of sand­ .014; as a factor (ANOVA, P = .015, F = 4.26). stone-associated stands were nonreproducing Under the three-level categorization, young than expected (14 and 9, respectively), wbile stands were rockier than mature, nonrepro­ fewer were reproducing than expected (22 and ducing stands (x = 32%) but did not differ 27, respectively). Expected and observed val­ from those that were reproducing. ues for the three other parent materials were more similar. Number ofModes Newly established C. reticulata stands are Celtis reticulata stands typically appeared to apparently rare, as few were observed during be unimodal (73%), with one dominant cohort. the study in spite of efforts to locate them. Stands with two modes were far less common Only six young «33 yr) stands were sampled. (11%), but a slightly greater number were all­ These were typically on rocky sites with inter­ aged (16%). Since only two stands had three mediate topographic shelter and gentle slopes modes, they were dropped from analyses; 246 GREAT BASIN NATURALIST [Volume 55

TABLE 6. Cross tabulation of the number of Celtis reticulata stands by stand structure and parent material. The hypothesis ofindependence ofstand structure and parent material is rejected with P == .014. Mature Parent Observed! material expected Nonreprod. Reprod. Young Row total Row %

Granite obs. 11.0 40.0 0.0 51 21 expo 12.2 37.6 1.3 Sandstone obs. 14.0 22.0 1.0 37 16 expo 8.8 27.2 0.9 Basalt obs. 20.0 71.0 0.0 91 38 expo 21.5 66.3 2.3 Alluvium obs. 12.0 45.0 5.0 62 25 expo 14.5 44.9 1.5 Column total obs. 57.0 178.0 6.0 241 100 Column % 24% 74% 3% thus, the sample size for this portion of the development (Schnell et al. 1977). On the results is based on 239 stands. Growth form or eastern Great Plains, C. tenuifolia (dwarf hack­ number of stems of the individuals was unre­ berry) is a gnarled, shrublike tree when grow­ lated to number ofmodes. ing on its typical rocky, shallow, calcareous Although of marginal statistical significance, substrate, but in the bottom of ravines it may all-aged stands were more common on shel­ reach heights of 8-10 m (Stephens 1973). In tered sites (Chi-square, P = .07). For example, addition to the influence of an ameliorated 33% of stands in draws, which typically have environment, sheltered stands may be less at least an intermediate topographic shelter, prone to repeated disturbances such as fire, to were all-aged. The percent of all-aged stands which a vegetative sprouter such as C. reticu­ at other topographic positions ranged from 6% lata will often respond with a shrubbier growth to 16%. fonn. Livestock grazing intensity may restrict enby In Oklahoma, Celtis occurs almost exclu­ of new cohorts within a C. reticulata stand as sively on loamy bottomland soils (Dooley and shown by the strong tendency for overgrazed Collins 1984), and in west Texas it is best de­ stands to be unimodal (92%; Chi-square, P = veloped on alluvium (Van Auken et al. 1979). .0008). In contrast, 70% of light to moderately In the canyon grasslands ofIdaho, Tisdale (1986) grazed stands had only one mode, 11% were recognized two types ofCeltis-dominated vege­ bimodal, and 18% were all-aged. tation on soils of two principal origins. The C. Size structure of Celtis stands did not differ reticulata-Agropyron spicatum habitat type with topographic position, parent material, occurs on lower valley slopes with rocky (50­ soil texture, slope, percent surface rock, or 60%), weakly developed loam soils derived heat load (all P > .2). from residual and collm1al materials. The sec­ ond vegetation type, unclassified because of DISCUSSION heavy livestock disturbance and alien plant In our study, trees were typically tallest and dominance of the understory, occurs on allu­ least shrubby when located in draws, on sites vial terraces with deep sandy soils. with surface or subsurface moisture, and in areas Soil texture appears to have a greater influ­ where they received maximum topographic ence on C. reticulata growth on upland sites shelter. Similar observations of C. reticulata than on sites associated with a perennial water have been recorded by others (Eliot 1938, Van source. W'hile C. reticulata grows on a range Dersal 1938, Peattie 1953), and for different of soil textures in Idaho, we found the tallest species of Celtis as well. For example, Hitch­ trees on loams, possibly because of their cock and Cronquist (1964) noted that Celtis greater water-holding capacity and nutrient reticulata is taller in moist areas in the Pacific content. However, 80% of the stands occurred Northwest. In Oklahoma, C. laevigata (sugar­ on soils with some sand component, and 30% berry) is typically a small tree in open areas, were on sand or coarse sand. The presence of but in lowland forests it reaches its maximum good drainage may be an important limiting 1995] CELTIS RETICULATA IN IDAHO 247 factor for C. retieulata, as finer-textured soils and 13.6 cm in diameter in Idaho, with a mean of the uplands were nearly always skeletal. tree height and diameter, regardless of age, of The increased percolation of sandy or skeletal 5 m and 18 cm, respectively. Unlike some shrub soils provides greater moisture availability for and tree species in the Intermountain West, deep-rooted shrub and tree species. populations are generally maintaining them­ In Idaho, C. retieulata is most prevalent on selves by vegetative sprouting or seedling rocky sites with southeast to westerly aspects, recruitment, despite historic and prevailing although heat load was not an important pre­ large-scale habitat alterations resulting from dictor ofgrowth. The presence ofrock, particu­ overgrazing, exotic plant invasion, and chang­ larly hedrock, may in fact be critical for hack­ ing fire frequencies (Tisdale 1986, Whisenant berry's existence on certain sites. It may also 1990). Hackberry's general resiliency and abil­ partially explain the fragmented distribution ity to resprout following disturbance or injury of the species in Idaho. Other rock-associated likely playa role in this, as does its positive asso­ species have been observed in semiarid regions ciation with rock Recruitment, as expressed as well. In the shrub-steppe region of eastern by the density of juvenile individuals, in­ Montana, Rumble (1987) found that scoria creased as surface rock cover increased. How­ rock outcrops provide a unique habitat for ever, under extreme grazing pressure, recruit­ several relatively mesic species. Rhus trilobata ment was significantly lowered and stands (skunkbush sumac), Prunus virginiana (choke­ were nearly all unimodal. All-aged stands cherry), Ribes spp. (currant), and Juniperus were absent from severely grazed sites. Even spp. Guniper) were found only in association though rock favors Celtis recruitment, its with rock outcrops in that ecosystem. He con­ growth is favored on less-rocky sites, such as cluded that their occurrence is prohably related draws. to protection from wind, snowdrift accumula­ The most likely explanation for relatively tion, shading, and mulch effects of rocks. slow C. reticulata growth on stream terraces, Oppenheimer (1964) and Potter and Green in spite of the assumed availability of ground­ (1964) suggested that the association of mesic water, is the extreme annual fluctuation of the species with rocky substrates is due to tempo­ water level and battering by flood debris. These rary water reservoirs that rock fissures pro­ sites are located below the high-water line. vide. In Arizona, Johnsen (1962) reported that Above the high-water line the mean site index Juniperus monosperma (one-seed juniper) is is larger and mechanical stresses are less largely limited to rock outcrops, where he extreme. While newly established C. retieulata recorded 2-2.5 times as much available mois­ stands were uncommon, they typically occurred ture. The theory of extra moisture availability on these riparian sites, where establishment in rock fissures could also hold true for the conditions occur more frequently than in the deeply rooted C. reticulata, helping explain its uplands. frequent presence on southerly aspects. Although individuals are often slow-grow­ Other plausible explanations for the infre­ ing, the variation in site conditions that the quency of C. retieulata on northerly aspects species appears to tolerate and its other posi­ and sites with less surface rock cover include tive attributes (i.e., wildlife food, cover, land­ its sensitivity to late spring frosts (personal scape structure, potential large size, tolerance observation) and poor competitive ability with of southerly aspects), are favorable qualities fast-growing species. In Idaho, C. retieulata is for those seeking rehabilitation species. The the last shrub to break dormancy and expand species' persistence in heavily degraded its leaves in the spring. This strategy, in com­ ecosystems may speak to its suitability for bination with the tendency to grow on warmer rehabilitation projects as well. slopes, generally prevents frost damage from occurring. The greater effective soil moisture ACKNOWLEDGMENTS and dense vegetative cover of north slopes probably create an environment too competi­ This study was funded in part by the Boise tive for this slow-growing species. District Office ofthe Bureau of Land Manage­ In summary, Celtis reticulata can generally ment, with additional support provided by Idaho be described as slow-growing and small­ Power Company. Nancy Shaw, Ed Tisdale, statured. Fifty-year-old trees averaged 4 m tall and Steve Monsen provided insight and 248 GREAT BASIN NATURALIST [Volume 55 encouragement during the earliest phase of HUSCHLE, G. 1975. Analysjs of the vegetation along the the research. Roger Rosentreter assisted in the middJe and lower Snake River. Unpublished master's tbesis, University of Idalm. Moscow. 271 pp. field and provided helpful suggestions and JOHNSEN. T N. 1962. One-seed juniper invasion of north· encouragement throughout the study's dura­ ern Arizona grasslands. Ecological Monographs 32; tion. We thank Patricia Muir, Boone Kauffinan, 187-207. and Kermit Cromack for their valuable com­ JOHNSON, C. G., JR., AND S. A. SIMON. 1987. Plant associa­ tions of the Wallowa-Snake Province, Wallowa­ ments on an early version of the manuscript. Whitman National Forest. USDA Pacific Northwest Thanks are also due to Stanley D. Smith, Shere! Region Publication R6-ECOL-TP-255B-86. 272 pp. Goodrich, and an anonymous reviewer for UNNl::R, R. M. 1983. Trees of the Great Basin. University their constructive review of the manuscript. of evada Press, Reno. 215 pp. U1TU::, E. L, In. 1976. Atla~ of UnittXi States trees. Vol· ume 3--Minor western hardwoods. U.S. Department LITERATURE CITED of Agriculture Miscellaneous Publication No. 1314. Washington, DC. 215 pp. ALBEE, H. J.. L. M. SCUULfZ, AND S. Gooumen. 1988. Atlas MUIR, P. S., AND 1. E. lm'AN. 1985. Disturbance history of the vascular plants of Utah. Utah Museum of and serotiny of Pil'WS contorla in western Montana. Natural History, Occasional Publication No.7. Salt E<'Ology 66, 1658-1688. Loke City, UT. 670 pp. OPPF.NHF:I~RR, H. R. 1964. Adaptation to drought; xero· AsHERIN, D. A., AND J. J. Q...v.R. 1976. Inventory of riparian phylism. Pages 105-135 in Plant-water relation..;;hips habitats IUld associated wildlife along the Columbia in arid and semi-arid conditions. 111e Hebrew Uni­ and Snake rivers. Volume 3A. College of Forcstry, versity, Rehovot. brael. Wildlife, and Range Sciences, University of Idaho, PBATTlF., D. C. 1953. A natural histOly of weslcrn trees. Moscow. 556 pp. Houghton Millin, New York. 751 pp. BoWLEI\, P. A, 1981. Natuml history studies and an evalua­ PWMMF.R, A. P. 1977. Revegetation of disturbed inter· tion fol' eligibility of the Wiley Reach of the Snake mountain area sites. Pages 302-337 in J. C. Thames, River for National Natural Landmark designation. editor, Reclamation and use ofdisturbed lands ofthe Unpublishtld report, National Park Service, Seattle, Southwest. University ofArizona Press, TUCSOn. WA. 86 pp. POWER, L. D., .... NI) D. L. GREEN. 1964. Eoology of pon· BRADY, N. C. 1974. The nature and property of soils. derosa pine in western North Dakota. Ecology 45: Macmillan Publi~hing Co.> Inc., New York. 639 pp. 10-23. BROWN, D. E. 1982. Great Basin montane scrubland. RUMBLE, M. A. 1.987. Avian use of scoria rock outcrops. Pages 83-84 in D. E. Brown, editor, Biotic commu~ Grcat Hasin Naturalist 47; 625-630. nities of the American Southwest-United States SCHNI!:LL, G. D., P. C. RISSER, AND J. F. HELSEL. 1977. and Mexico. Desert plants. Volume 4. Factor analysis of tree distribution patterns in Okla­ n"UIlENMUlll:, R. 1970. Steppe vegetation of Washington. homa. Ecology 58: 1345-1355. Agricullllre Experiment Station Technical Bulletin SPSS. 1988. SPSS/PC+ V2.0 base manual. SPSS Inc., No. 62. Pullman, WA. 131 pp. Chicago. --d~uctivily. 1976. The use of vegetation in assessing the pro­ STEPHENS, II. A. 1973. Woody plants of the North Central of forest lands. Botanical l\eview 42: Plains. University Press of Kansas, lawrence. 530 Pl'. 115--143. TISDALE, E. W. 1986. Canyon grasslands and associated DEBoLl~ A. M. 1992. The ecology ofCeltis reticulata Torr. shrubhmds of we~1-C,,'t:Jltral Idaho and adjacent areas. (nctleuf hackberry) in Idaho. UnpubHshed master's BulJetin No. 40. lobrestry, Wildlife, :UJd Hange Exper­ thesis, Oregon State University, Corvallis. 161 pp. iment St-.ltion, University of ldaho, Moscow. 42 pp. ))OOL':;Y, K. L., ....ND S. L. CoLLINS. 1984. Ordination and VAN AUK-EN, W W, A. L. FORD, AND A. STEIN. 1979. A com~ classification of western oak forests in Oklahoma. parison ofsome woody upland and riparian plant com­ American Journal of Botany 71: 1221-1227. munities of the southern Edwa,·ds Plateau. South· ELIAS, T S. 1980. The complete trees of North America. westem Naturalist 24: 165--180. Times Mirror Magazines, me., New York. 948 pp. VAN DIl:RSAL, W. R. 1938. Native woody plants of the EU01: W A. 193B. Forest trees of the Pacific Coast C. P. United Stutes, their erosion control and wildlife val­ Putnam's Sons, New York. 565 pp. ues. U.S. Department of Agriculture Miscellaneous CUNSKI, R. C. 1977. Regeneration and distribution of syca­ Publication No. 303. Washington, DC. 362 pp. more and cottonwood trees along Sonoita Creek, WHISENANT, S. G. 1990. Changing fire frequencies on Santa Cruz County, Arizona. Pages 116-123 in R. R. Idaho's Snake River Plaim: ecological and manage· J~hnson and A. Dale. technical coordinators, Impor­ ment implications. Pages 4-10 in E. D. McArthur, et tance, preservntion, and management of riparian al., compilers, Cheatgrass invasion, shrub die-off, habitat-proceedings. USDA General Technical and other aspects of shrub biology and manage­ Report RM-43. r'Ort Collins, CO. ment-proceedings. U.S. Department ofAgriculture HAYWARD, C. L. 1948. Biotic communities of the Wasatch General Technical Report INT-276. Ogden, UT. chaparral, Utah. Ecological Monographs 18: 473--506. WHl"TTAKRK, R. H. 1960. Vegetation of the Siskiyou HITCHCOCK, C. L., AND A. CRONQUI~'T. 1964. Vascular Mountains, Oregon and CalifoOlia. Ecological Mooo­ plants of tho Pacific Northwest. Volume 2. Univer· gl'Ophs 30: 279--338. sity ofW