RESEARCH ARTICLE

Diverse responses across soil parent materials during

ecological restoration
Scott R. Abella1,2 , Joseph E. Crouse3 , W. Wallace Covington3 , Judith D. Springer3

A major challenge to advancing the science and practice of ecological restoration is working across large landscapes containing
diverse sites that may respond differently to restoration. We conducted a 5-year restoration experiment, replicated across
9 sites spanning 3 soil parent material types within a 9,000-ha Pinus ponderosa forest landscape. We evaluated plant community
response to restoration Pinus thinning, grazing, and aqueous smoke application. We measured vegetation before (2003) and
3 (2006) and 5 (2008) years after treatment. Plant community responses of species richness, cover, and composition were
diverse, ranging from increases, decreases, or no change depending on soil parent material, tree thinning, and presence or
exclusion of grazing. Restoration outcomes were under hierarchical control: soil parent material constrained response to Pinus
thinning, which in turn influenced grazing effects. On limestone-derived soil, responses included no change in species richness
but increased plant cover with Pinus thinning. Both plant richness and cover increased on benmorite soil after thinning,
and cover generally increased more without grazing. On rocky, basalt soil, plant richness increased but cover did not after
any treatment. Diversity of responses to restoration has implications for: (1) setting goals or monitoring indicators tailored
to inherent soil capability; (2) identifying where grazing most affects restoration outcomes; and (3) forecasting responses to
restoration across landscapes. Diverse responses to restoration along physiographic gradients such as soil parent material
warrant consideration when developing restoration across degraded landscapes.
Key words: environmental gradient, grazing, landscape, parent material, Pinus ponderosa forest, soil, tree thinning, under-
story

restoration across environmental gradients (e.g. topography, soil

Implications for Practice
• Vegetation responded differently among soil types to
restoration tree thinning and grazing treatments.
• Conducting restoration projects across environmental gra-
dients will likely enhance diversity within restored land-
scapes.
• Accounting for diversity of responses across environmen-
tal gradients in restoration projects can help practitioners
optimize treatments (e.g. allowing or excluding grazing at
restoration sites), prioritize sites for treatment to enhance
desired landscape characteristics, and identify monitoring
indicators aligned with inherent soil capability.
parent material) can evaluate generality of site-specific restora-
tion outcomes, and provide practitioners with information on
among-site variability for landscape-scale restoration prescrip-
tions (Eviner & Hawkes 2008).
Some of the few examples of restoration across environmen-
tal gradients have illustrated that restoration outcomes, includ-
ing for key functional attributes such as ecosystem C storage
and rare plant habitat, can vary across environmental gradients.
At 15-year-old grassland restoration sites, for example, soil C
was twice as high on clayey soil compared to on sandy soil
(Baer et al. 2010). Abella (2010) found that Pinus plantation
removal to restore Quercus-prairie ecosystems converted a uni-
formly species-poor set of sites into diverse vegetation types

Introduction
Restoration at a single site or small sites can be essential to
understanding ecological processes and developing restoration
techniques. However, conducting restoration across numerous
sites or large areas is crucial to discerning responses of land-
scapes to restoration at the broad scales where degradation often
occurs. As Eviner and Hawkes (2008) noted, ecological theo-
ries are often formulated as generalities (and frequently based
on few sites) that may not accurately represent many of the
sites where restoration practitioners are working. Implementing
Author contributions: SRA conceived and designed the research; SRA, JEC, JDS
collected field data; WWC arranged for treatment implementation; SRA wrote the
paper; WWC, JEC, JDS edited the paper.
1 Natural Resource Conservation LLC, 1400 Colorado Street, Boulder City, NV 89005,
U.S.A.
2 Address correspondence to S. R. Abella, email abellaNRC@gmail.com
3 Ecological Restoration Institute, Northern Arizona University, Flagstaff, AZ 86011,
U.S.A.
© 2014 Society for Ecological Restoration
doi: 10.1111/rec.12160
Supporting information at:
http://onlinelibrary.wiley.com/doi/10.1111/rec.12160/suppinfo

Restoration Ecology 1
Restoration across soil parent materials
differentiating along soil drainage gradients. Moreover, four
state-listed rare plant species became established on restora-
tion sites, and not all the species would have appeared had
restoration not occurred across the full soil gradient. In both of
these examples, conducting restoration across landscape gradi-
ents allowed identification of a range of restoration outcomes,
which displayed spatial fidelity to particular environmental set-
tings within landscapes.
Some of the most common restoration treatments conducted
in many of Earth’s biomes include decreasing woody plant
encroachment, reestablishing fire regimes, and manipulating
grazing (Vandvik et al. 2005; Brudvig & Asbjornsen 2009;
Busse et al. 2009). We used Pinus ponderosa (ponderosa pine)
forests of northern Arizona, United States, as a model sys-
tem to assess how plant communities respond to restoration
tree thinning, smoke addition, and grazing exclusion across
different soil parent materials. Suppressing formerly frequent
fire from dry-site conifer forests (such as those dominated by
P. ponderosa) has resulted in over 23 million ha requiring
restoration in the western United States (Schoennagel & Nel-
son 2011). Restoration seeks to ameliorate hazardous fuels and
re-balance biomass distribution among trees and other ecosys-
tem components such as understory plants (Covington et al.
2001).
Owing to negative relationships between P. ponderosa den-
sity and understory plant abundance, restoration thinning of
Pinus would be expected to increase plant species richness and
cover (Laughlin et al. 2006; Bakker & Moore 2007; Dodson
et al. 2007). Whether changes are consistent among soil types
is uncertain, as is potential differentiation of plant species com-
position among soil types after thinning. As an example of
this uncertainty, nutrient-rich, productive soil could be antic-
ipated to display the greatest increase in understory plant
cover because of a favorable environment for plant growth
(Welch & Klemmedson 1975). But for the same reason, these
soils could instead display minimal response to restoration if
their high productivity has maintained understory cover despite
P. ponderosa encroachment. As a germination cue associated
with restoring natural fire regimes, smoke would be antici-
pated to increase plant richness and cover (Rokich et al. 2002;
Flematti et al. 2013). Previous research in P. ponderosa forest
found that aqueous smoke triggered 21–115% greater emer-
gence from soil seed bank samples in a greenhouse (Abella
et al. 2007). Grazing also can influence restoration outcomes,
and a global synthesis found that grazing effects reversed across
soils: grazing increased plant diversity on nutrient-rich soil but
decreased diversity on nutrient-poor soil (Proulx & Mazumder
1998). Grazing is common in P. ponderosa forests, and any rela-
tionship between grazing and soil type could be mediated via the
Pinus overstory that strengthens (or weakens) grazing effects
(Bakker & Moore 2007).
Our experiment examined interactions among these influ-
ences in a practical restoration setting, specifically evaluating
the question: do plant communities respond differently to com-
binations of restoration tree thinning, smoke addition, and graz-
ing among sites that differ in soil parent material?
2 Restoration Ecology
Methods
Study Area
We conducted this experiment within a 9,000-ha area of the
Northern Arizona University Centennial Forest, 10 km south-
west of the city of Flagstaff in northern Arizona, United
States (Supporting Information, Fig. S1). The Flagstaff Air-
port weather station has reported the following climate means:
precipitation: 53 cm/year (about half as snow), July daily high
temperature: 28∘ C, and January daily low temperature: −9∘ C
(2,137 m elevation; 1950–2012 records; Western Regional Cli-
mate Center, Reno, NV, U.S.A.). Most soils, including those
we studied, have basalt, benmorite, or limestone parent mate-
rial and are classified as Typic, Lithic, or Mollic Eutroboralfs
(Miller et al. 1995). Forests are predominately pure Pinus pon-
derosa, with some Quercus gambelii (Gambel oak). Fire-history
studies adjacent to and within the study area recorded mean fire
intervals of 2–4 years from the 1600s to the 1880s, indicating
that forests frequently burned before Euro-American settlement
in the 1880s (Van Horne & Fulé 2006). Anthropogenic exclu-
sion of fire has since occurred at our study sites, typical for dry
conifer forests of western North America (Falk et al. 2011). Dur-
ing the fire-free period, densities of P. ponderosa trees increased
by orders of magnitude, often from less than 100 trees/ha in the
1880s (Abella & Denton 2009) to greater than 1000/ha today
(Fulé et al. 1997).
Livestock—Bos primigenius (cattle) and Ovis aries
(sheep)—were introduced with Euro-American settlement
and were kept on the landscape at greater densities than in
contemporary forests (Loeser et al. 2007). Current livestock
density, including at study sites during our study period,
approximates 0.03–0.20 animal unit months/ha (Bakker &
Moore 2007; U.S. Forest Service, Coconino National Forest,
Flagstaff, AZ, U.S.A.). Other ungulates include Odocoileus
hemionus (mule deer), which is indigenous; and Cervus ela-
phus (Rocky Mountain elk), native to elsewhere in the United
States, but introduced to Arizona in 1913, and now numbering
about 35,000 animals in the state (Arizona Game and Fish
Department, Phoenix, AZ, U.S.A.).
Study Sites and Soil Types
We randomly selected three mapping units of each of three
soil parent material types (limestone, benmorite, and basalt)
mapped by soil scientists under the U.S. Forest Service’s ter-
restrial ecosystem survey with a 16-ha minimum mapping unit
(Miller et al. 1995; Fig. S1). Each of the three soil parent
materials exhibited a unique combination of characteristics.
Some individual properties could be similar between two soil
types, but differences in one or more properties (including sub-
surface properties such as depth to clay layer) distinguished
sites (Abella & Denton 2009; Table 1; Fig. 1). For example,
clay concentration in surface soil varied sharply by a factor
of 1.5 (16 to 24%), and subsoil organic C concentration dou-
bled from limestone to basalt. In summary, limestone supported
sandy soil, few rocks, high plant species richness and cover,
and mixtures of perennial grasses and forbs. Benmorite had

silt loam surface soil, clay subsoil usually within 30 cm of
the surface, intermediate rock content, and lower plant rich-
ness and cover than limestone. Basalt contained abundant rocks,
clayey soil, and dominance by Muhlenbergia montana (moun-
tain muhly), forbs, and small-statured graminoids such as Carex
geophila (White Mountain sedge) and Sporobolus interruptus
(sand dropseed). Linear distance between mapping units aver-
aged 8 km (2–16 km range), and map units of different parent
material were interspersed except for clustering of benmorite
sites corresponding with geological patterns (Fig. S1). Mean
elevation varied by less than 35 m among soil types, forest man-
agement history of selective harvest and grazing was similar,
and precipitation modeled by PRISM was nearly identical, sug-
gesting that the main difference among map units was soil parent
material (Table 1; Abella & Denton 2009).
Within each mapping unit, we randomly selected a site (using
a randomly generated geographic coordinate) that met the fol-
lowing criteria: within 1 km of a forest road for accessibility to
implement treatments, contained P. ponderosa trees greater than
50 years to represent typical forests undergoing thinning (Fulé
et al. 2012), and no visual evidence of fire in the post-settlement
period since 1880. Pinus ponderosa forests exhibit a spatial
structure of groups of trees (usually <0.1 ha in size) alternat-
ing with canopy openings of 0.01–0.1 ha (Larson & Churchill
2012). At each site, we selected three treed and one open area
closest to the randomly generated point. Within each area, we
established a 20 m × 25 m (0.05 ha) experimental plot in which
treatments were implemented.
Experimental Treatments
We designed a balanced, four-factor experiment including soil
type (three levels: limestone, benmorite, or basalt), patch type
(nested within soil type and with four levels: unthinned control,
thinned, thinned + smoke, or open), grazing (nested within soil
and patch type and with two levels: excluded or not), and time
(three levels: 2003 pre-treatment and 2006 and 2008 represent-
ing 3 and 5 years post-treatment). Patch type was designed to
represent a range of realistic ecological structures maintained
or created during restoration projects, including practices such
as leaving some unthinned patches for wildlife habitat diversity,
heterogeneity of patches with burning, and remnant openings
that remain untreated. The experiment was originally intended
to include a burn treatment, but owing to dry conditions and fire
restrictions, burning was not implemented. Instead, we applied
smoke as a potential influence of fire (Lloyd et al. 2000).
Tree thinning was performed by hand using chainsaws (no
mechanized equipment drove across sites) on two randomly
selected treed plots (the other plot remained an unthinned con-
trol) at each site in September 2003 (Fig. 1). Trees (all P. pon-
derosa) were thinned from an average density of 1,362 trees/ha
(SD = 824) to either 60 or 80 trees/ha (3 or 4 trees/plot). The
slight variation (1 tree) in post-thinning density among plots
resulted from tree sizes (all trees >40 cm diameter at 1.4 m
height were retained) or from avian nests in a tree’s crown that
resulted in retention. A post-thinning tree density as similar
as possible among soil types was targeted to isolate effects of
Restoration Ecology
Restoration across soil parent materials
parent material, and this target density was within a range of
circa 40–100 P. ponderosa trees/ha reconstructed to have occu-
pied these sites in 1880 before fire exclusion (Abella & Denton
2009). To reduce edge effects, trees were also thinned in a 5-m
buffer surrounding each plot. Logs and slash were transported
off-site.
A grazing exclosure, 3.16 m × 3.16 m (10 m2 ) and 2 m tall,
consisting of four metal fence posts and 1-mm thick wire with
5 cm × 10 cm openings, was installed in the center of half of
each plot. An area also 3.16 m × 3.16 m (10 m2 ), but without
an exclosure, was delineated in the center of the other half of
each plot.
To a randomly selected thinned plot at each site, we applied
undiluted aqueous smoke (Regen Direct, Forest Flavors, Glas-
gow, KY, U.S.A.) to entire plots by backpack spraying on
23 June after thinning and exclosure installation. The appli-
cation rate followed manufacturer recommendations and was
100 mL/m2 . Our smoke treatment used the same brand of smoke
and application rate that increased density of emerging seedlings
by 10× in a Western Australia Banksia woodland (Lloyd et al.
2000). We timed the treatment within the May to early Septem-
ber period when most fires historically occurred in the study area
(Fulé et al. 1997).
Data Collection
Pre-treatment data were collected on each plot in September
2003 within a 9-m2 area inside each exclosure and within the
paired 9-m2 area outside the exclosure. We divided each area
into nine, 1 m × 1 m subplots. We visually categorized aerial
cover of each plant species rooted in each subplot as 0.1%,
0.25%, 0.5%, and 1% up to 1% cover, at 1% intervals to 10%
cover, and at 5% intervals above 10% cover. The same data
were collected in 2006 and 2008, 3 and 5 years after treatment.
Nomenclature follows Natural Resources Conservation Service
(2013). Species richness and cover in the nine 1-m2 subplots
were averaged per m2 for each plot, and cumulative species/9 m2
also was tabulated for each plot.
Data Analysis
We analyzed the response variables of species richness/m2 ,
species richness/9 m2 , and total plant cover using a balanced,
nested mixed model analysis of variance (ANOVA) includ-
ing the four factors of soil type (tested over site within soil
type), patch type (tested over site within soil by patch), graz-
ing (tested over site within soil by patch by grazing), year
(tested over site within soil by patch by year), and all inter-
actions (SAS Institute 2009). The minimum adequate model
was determined by Aikaike Information Criterion (AICC), cor-
rected for small sample size. Residuals for each model were
normal and homoscedastic. Full models including the four-way
interaction among all factors were identified by AICC as min-
imum adequate models. To limit the numerous multiple com-
parisons that could be generated by a four-way interaction (3
soil types × 4 patch types × 2 grazing treatments × 3 years = 72
combinations plus interactions therein) and to identify which
3
Restoration across soil parent materials

Table 1. Characteristics of soil parent material types on which responses to restoration treatments were evaluated in Pinus ponderosa forests, northern Arizona,
United States.

Limestone Benmorite Basalt

a
Elevation (m) 2190 ± 21 2225 ± 40 2214 ± 11
Rock cover (%) 1±1 4±2 5±1
0–15 cm soil
Gravel (%) 28 ± 2 34 ± 3 38 ± 14
Sand (%) 46 ± 11 28 ± 1 30 ± 1
Clay (%) 16 ± 4 18 ± 3 24 ± 3
pH (1:2 CaCl2 ) 5.9 ± 0.1 6.0 ± 0.2 5.9 ± 0.0
Organic C (%) 1.5 ± 0.7 1.5 ± 0.1 1.6 ± 0.2
Total N (%) 0.08 ± 0.03 0.09 ± 0.01 0.10 ± 0.01
15–50 cm soil
Gravel (%) 45 ± 6 52 ± 3 46 ± 11
Sand (%) 40 ± 5 25 ± 3 29 ± 3
Clay (%) 31 ± 8 30 ± 6 33 ± 5
pH (1:2 CaCl2 ) 6.0 ± 0.1 5.9 ± 0.1 5.9 ± 0.0
Organic C (%) 0.7 ± 0.1 0.8 ± 0.2 1.4 ± 0.2
Total N (%) 0.06 ± 0.01 0.06 ± 0.01 0.09 ± 0.01
Dominant species Poa pratensis Festuca arizonica Carex geophila
Erigeron formosissimus Elymus elymoides Muhlenbergia montana
Dense tree canopyb
Species/m2 6±2 2±1 3±1
Species/9 m2 13 ± 3 5±3 9±2
Plant cover (%) 8±6 1±1 3±1
Open tree canopy
Species/m2 12 ± 5 6±4 6±2
Species/9 m2 23 ± 7 14 ± 7 14 ± 4
Plant cover (%) 15 ± 8 8±7 12 ± 11
Values are mean ± 1 SD, calculated from three replicate sites per soil type.
a Environmental variables, including soil properties, were obtained from measurements at the sites in 2003 described in Abella and Denton (2009). Gravel is material greater than

2 mm in diameter. Soil percentages are by weight.

b Vegetation variables for both tree canopy types are pre-treatment in 2003 and were obtained by data collection in this study.

treatment combinations changed significantly through time, we
used Bonferroni-adjusted pairwise tests to compare pre- (2003)
and post-treatment (2006 and 2008) least-squares means within
treatment combinations (patch type and grazing within soil
type).
To evaluate species composition using the same partially
nested model as for the univariate analyses, we used permuta-
tional multivariate analysis of variance (PERMANOVA; Ander-
son 2004). We constructed a matrix of relative species cover
(cover of speciesi /sum of all species for each plot), converted
to a matrix of Sørensen distances on which we performed PER-
MANOVA (999 permutations and post-hoc contrasts for signif-
icant model terms). To discern patterns in individual species,
we calculated difference in cover between pre-treatment and the
two post-treatment years within treatment combinations.
Results
Species Richness and Cover
According to AICC, full models including four-way interac-
tions of soil parent material, patch, grazing, and year were
significant for all three univariate response variables of plant
species/m2 , species/9 m2 , and cover (Table 2). Overall plant
community responses to treatment were established by 3 years
(2006) after treatment and were maintained for the duration of
the 5-year experiment. Based on ANOVA and pairwise compar-
isons, responses to tree thinning and grazing differed across soil
parent materials and among measures (richness vs. cover) of the
plant community (Table 3; Fig. 2).
Thinning alone within patch type significantly increased
species richness (both scales) on benmorite and basalt, but not
on limestone soil (Fig. 2). In unthinned control plots under tree
canopy, grazing did not affect richness or cover of the plant
community on any soil type. However, grazing interacted with
thinning to increase species richness on limestone and basalt.
On benmorite, both grazed and ungrazed plots exhibited signifi-
cantly higher richness post-treatment on thinned plots. Aqueous
smoke had little influence, except possibly for a subtle increase
in richness (compared to thinning alone) on limestone.
Patterns for plant cover differed from those of richness.
While richness did not change with thinning on limestone,
cover significantly increased (Fig. 2). Cover increased most
without grazing on limestone, although grazed, open-canopy
plots 5 years post-treatment exhibited significantly greater cover
than before treatment. Thinning sharply increased cover on
benmorite, and this increase was most pronounced without
grazing. Unlike species richness that significantly increased on
basalt with thinning (most strongly on thinned, grazed plots),
cover did not significantly change on basalt with any treatment.

4 Restoration Ecology
 Restoration across soil parent materials

Figure 1. Examples of sites before treatment (2003, top) and 3 years after tree thinning and construction of grazing exclosures (2006, bottom) in Pinus
ponderosa forests, northern Arizona, United States. Stars mark the same location in before–after paired photos. Note high plant cover on limestone compared
to coverage by rocks of the soil surface on basalt. Photos by S. R. Abella.

Table 2. Statistical results for treatment influences on plant community characteristics of Pinus ponderosa forests, northern Arizona, United States.

Species/m2 Species/9 m2 Cover Composition

Effect dfa Fb pb F p F p F p

Soil (S) 2,6 2.9 0.131 2.5 0.164 1.9 0.227 1.9 0.100
Patch (P) 3,18 6.3 0.004 5.6 0.007 4.0 0.023 1.9 0.002
Grazing (G) 1,24 0.5 0.491 0.4 0.543 5.8 0.024 1.3 0.206
Year (Y) 2,48 71.2 <0.001 81.5 <0.001 47.5 <0.001 7.5 0.001
S×P 6,18 0.9 0.515 1.3 0.308 1.3 0.327 1.0 0.421
S×G 2,24 1.1 0.343 0.4 0.680 0.7 0.491 1.3 0.164
P×G 3,24 0.4 0.786 0.2 0.892 2.0 0.142 1.1 0.261
S×Y 4,48 6.7 <0.001 7.1 <0.001 3.7 0.011 1.7 0.002
P×Y 6,48 20.0 <0.001 24.8 <0.001 6.0 <0.001 1.5 0.004
G×Y 2,48 1.5 0.231 3.0 0.058 3.8 0.029 1.7 0.023
S×P×G 6,24 1.1 0.379 0.8 0.602 1.1 0.411 1.0 0.400
S×P×Y 12,48 2.0 0.043 3.1 0.003 1.6 0.120 1.1 0.170
S×G×Y 4,48 1.8 0.142 3.6 0.012 2.1 0.099 0.8 0.769
P×G×Y 6,48 1.6 0.156 1.4 0.223 1.4 0.244 0.7 0.986
S×P×G×Y 12,48 1.6 0.133 1.6 0.138 1.1 0.382 1.0 0.490

Akaike Information Criterionc

Full model 588 792 970 —

Reduced model 634 858 1048 —
a Numerator, denominator degrees of freedom.
b F-statisticsand p values for the univariate richness and cover variables in analysis of variance, and Pseudo-F and permuted p value for multivariate species composition in
permutational multivariate analysis of variance. p Values in bold are less than 0.05. Italicized p values indicate minimum adequate models based on Akaike Information Criterion.
c Preferred models according to AICC are those with the smallest AICC values, which in our case are full four-way interaction models rather than reduced models with three or fewer

interacting factors.

Species Composition
We recorded 145 species, of which 130 (90%) were native,
across all plots during the experiment. Growth form distribu-
tion consisted of: 110 forb species (76%), 25 grass (17%), 4
sedge (3%), 4 shrub (3%), and 2 tree species (1%). Most were
capable of perennial life spans (104 species, 72%), with the
remaining 41 species (28%) annuals or biennials. The most
common species in 2008, occupying at least 50% of the 36
plots, included: Elymus elymoides (bottlebrush squirreltail;
100% of plots), Carex geophila (White Mountain sedge; 89%),

Restoration Ecology 5
Restoration across soil parent materials

Table 3. Summary of plant community responses to tree thinning, grazing, and thinning + grazing across soil parent material types of Pinus ponderosa forests,
northern Arizona, United States.

Species Richness/m2 Species Richness/9 m2 Plant Cover

Thin Graze Thin + graze Thin Graze Thin + graze Thin Graze Thin + graze

Limestone 0 0 0 0 0 0 ↑ 0 0
Benmorite ↑ 0 0 ↑ 0 ↑↓ ↑ 0 ↑↓
Basalt ↑ 0 ↑↑ ↑ 0 ↑↑ 0 0 0
Zeros signify absence of a statistical response (p > 0.05). Arrows note significant responses and their direction (increase or decrease), including for the thinning + grazing interaction
where the first arrow signifies thinning and the second arrow grazing.

Figure 2. Plant community responses (down columns) relative to before treatment (2003) among soil parent material types (across rows) and patch types
(x-axis labels), with and without grazing 3 (2006) and 5 (2008) years post-treatment in Pinus ponderosa forests of northern Arizona, United States. Asterisks
note differences from pre-treatment significant at p < 0.05. Pre-treatment values for vegetation variables under tree canopy or in openings are given by parent
material in Table 1.

Polygonum douglasii (Douglas’ knotweed; 69%), Poa fend-
leriana (mutton grass; 69%), Packera multilobata (lobe-leaf
groundsel; 67%), Erigeron divergens (spreading fleabane;
67%), Taraxacum officinale (common dandelion; a non-native,
58%), Lotus wrightii (Wright’s deervetch; 56%), and Vicia
americana (American purple vetch; 50%).
Analyzing multivariate species composition in PER-
MANOVA indicated no four- or three-way interactions, but two-
way interactions occurred between year and each experimental
factor (Table 2). The soil by year interaction indicated that rel-
ative change through time in species composition across treat-
ment combinations varied among soil types. This interaction
resulted from differential accrual of new species through time,
combined with differential changes in plant cover among
soils. Limestone was species-rich before treatment (Table 1)
and accrued fewer new species than other soils, affecting

6 Restoration Ecology

compositional changes. Moreover, basalt exhibited minimal
increase in plant cover with treatment, especially for peren-
nial graminoids, also resulting in a compositional trajectory
differing from other soils. Species composition between
post-treatment years did not differ, indicating that post-
treatment composition was established by 3 years after treat-
ment and little additional change occurred at 5 years.
Consistent with multivariate results, cover of most species
was either maintained or increased following thinning. More-
over, magnitude of the pre-and post-treatment cover increase
for native perennial graminoids was larger overall without
grazing, but specific responses varied among soils (Table S1).
For example, the pre- and post- (5 years) treatment increase
in cover on thinned plots for Carex geophila was 1.3× greater
without grazing on limestone and 13× greater without graz-
ing on benmorite. Basalt exhibited the reverse: C. geophila
cover was 2.6× higher on grazed compared to ungrazed
plots. Generally, the interactions of soil, patch, and grazing
with pre- and post-treatment years in PERMANOVA were
driven by a multivariate cumulative response across numerous
species, rather than strong responses in a few individual species
(Table S1).
Discussion
Interactions among type of soil parent material, patch type cre-
ated through restoration (e.g. via Pinus overstory thinning), and
grazing resulted in patterned responses to restoration across
the landscape. First, experimental factors had a hierarchical
influence on the plant community: soil parent material affected
response to tree thinning, which in turn constrained effects of
grazing exclusion. Second, plant community characteristics dif-
fered among soil parent materials before treatment, and sub-
sequent post-treatment trajectories also differed among soils.
Third, components of the plant community—species richness,
cover, or species composition—responded differently to exper-
imental factors. Fourth, within the 5-year study period, vegeta-
tion responses to treatment were largely established by 3 years
post-treatment (2006) and changed little more at 5 years (2008)
post-treatment. Annual precipitation during the post-treatment
period was slightly above average the first two post-treatment
years (115% of the 53 cm long-term mean in 2004 and 117% in
2005) and slightly below average for the last three (76% in 2006,
85% in 2007, and 92% in 2008; Flagstaff Airport Station, West-
ern Regional Climate Center, Reno, NV, U.S.A.). Fifth, relative
importance of treatments varied along a continuum of strong
effects of tree thinning, intermediate importance of grazing, and
weak influence of smoke application.
Across all soils, tree thinning induced some change in at least
one component of the plant community. However, magnitude
of the change, and which component (species richness, cover,
or composition) was most affected, varied among soils. Despite
voluminous literature on Pinus ponderosa thinning (e.g. Laugh-
lin et al. 2006; Dodson et al. 2007; Busse et al. 2009), this find-
ing of variation in plant community responses to tree thinning
across soil types was previously little articulated. This under-
scores value of a landscape environmental gradient perspective
Restoration Ecology
Restoration across soil parent materials
in restoration research and monitoring to place site-specific find-
ings in a broader spatial context to identify ranges of potential
responses to restoration.
Differential response to tree thinning across soils and among
plant community components likely relates to inherent soil
productivity. Limestone had the greatest plant richness and
cover before treatment, and the soil supported increased plant
cover but not richness after treatment. Based on a hypothe-
sis that high productivity constrains species richness through
competitive exclusion (Schultz et al. 2011), high plant cover on
limestone, especially of competitive perennial grasses, might
have limited further accumulation of species. At the other
extreme, basalt displayed a sharp increase in richness but weak
response in cover. Rocky substrates reducing area suitable for
plant growth, coupled with clay subsoils limiting root devel-
opment, could have constrained plant cover increases, while
allowing accumulation and coexistence of new species after tree
thinning.
Grazing exclusion only had significant effects on thinned
plots. This finding concurs with Bakker and Moore (2007),
who reported that effects of 63 years of grazing exclusion
between 1941 and 2004 hinged upon dynamics of the P. pon-
derosa overstory. Increased tree density during the 63 years
overwhelmed grazing effects. Our results further suggest that
interactive effects of overstory and grazing depend on soil type.
Contingency of grazing effects on soil agrees with results of
a global synthesis (Proulx & Mazumder 1998). We found,
however, the opposite pattern from Proulx and Mazumder
(1998). Rather than grazing increasing richness most on the
most productive soil, we found that grazing increased rich-
ness most after tree thinning on one of our least productive
soils (basalt). This finding was especially perplexing given
that plant cover displayed the weakest response to tree thin-
ning on basalt, so richness response to grazing might not have
been mediated extensively through plant cover as it could have
been elsewhere. Increased richness on grazed plots was largely
driven by forb colonization, and it is possible that mecha-
nisms such as seed dispersal by herbivores, or soil disturbance
on the rocky soils, enhanced plant richness on thinned basalt
plots.
Although aqueous smoke stimulated germination from soil
seed bank samples (assayed in a greenhouse) collected from the
study area in previous research (Abella et al. 2007), aqueous
smoke had minimal effect in this field experiment. Moreover,
our type of aqueous smoke and application rate was identical
to that used in an Australian field experiment that found a 10×
increase in plant emergence (Lloyd et al. 2000) and was similar
to other studies that also found enhanced plant establishment
(e.g. Rokich et al. 2002; Burne et al. 2003). Reasons for lack
of smoke stimulation in our experiment remain unclear. Cor-
responding with predominant timing of historical fires, our 23
June smoke application was followed by a 0.4 cm rain event
the next day and no rain until 1.3 cm fell 23 days later. This
is typical for the study area where rain is sparse in May and
June until monsoon rains begin sometime in July or August.
Smoke chemicals can be retained both in soil and seeds for vary-
ing duration (Rokich et al. 2002), but it is possible that climate
7
Restoration across soil parent materials
or other factors were not optimal for emergence. While we
found minimal effect of smoke, other delivery methods and new
smoke technologies warrant testing for plant responses in west-
ern United States frequent-fire forests (Flematti et al. 2013).
Results suggest several practical considerations for imple-
menting landscape-scale restoration projects. For example, fuel
reduction via reduced tree density is needed across soil types,
and given equal other factors and limited resources for treat-
ments, targeting specific soil types for treatment could max-
imize desired responses. Prioritizing treatment on soil types
with greatest capacity for increasing species richness could
enhance biodiversity benefits of restoration. Likewise, prioritiz-
ing soils of highest productivity could maximize forage value of
restoration sites. Decisions on grazing management at restora-
tion sites could be assisted by the differential effects that graz-
ing had across soils. Findings further inform restoration goals
or monitoring indicators, such as refining expectations based
on identifying inherent soil capability. The diversity of restora-
tion outcomes would not have been detected without imple-
menting treatments in a landscape context spanning multiple
soil types.
Acknowledgments
We thank the Ecological Restoration Institute (ERI) and U.S.
Forest Service for funding; D. Normandin, M. C. Tuten, and
L. Brandy (ERI) for installing grazing exclosures; Coconino
Rural Environment Corps (Flagstaff, AZ) for implementing tree
thinning; S. Crace (“Charcoal Sam”) and Forest Flavors, Inc.,
for donating Regen Direct aqueous smoke; J. J. Smith and C.
Miller (NAU) and K. Pajkos (Arizona State Lands Department)
of the Centennial Forest for facilitating access to field sites; C.
H. Sieg (Rocky Mountain Research Station) and M. M. Moore
and P. Z. Fulé (Northern Arizona University), for suggestions on
study design; K. Christie, R. Hastings, T. DeKoker, and many
students and staff of the ERI for help with vegetation sampling;
C. H. Vanier (University of Nevada Las Vegas [UNLV]) for
providing statistical advice and performing statistical analyses;
S. Altman (UNLV) for creating tables and figures; and V. Eviner
and four anonymous reviewers for helpful comments on the
manuscript.
LITERATURE CITED
Abella SR (2010) Thinning pine plantations to reestablish Oak Open-
ings species in northwestern Ohio. Environmental Management 46:
391–403
Abella SR, Denton CW (2009) Spatial variation in reference conditions: histor-
ical tree density and pattern on a Pinus ponderosa landscape. Canadian
Journal of Forest Research 39:2391–2403
Abella SR, Springer JD, Covington WW (2007) Seed banks of an Arizona Pinus
ponderosa landscape: responses to environmental gradients and fire cues.
Canadian Journal of Forest Research 37:552–567
Anderson MJ (2004) DISTLM v.5: a FORTRAN computer program to calculate
a distance-based multivariate analysis for a linear model. Department of
Statistics, University of Auckland, Auckland, New Zealand
Baer SG, Meyer CK, Bach EM, Klopf RP, Six J (2010) Contrasting ecosystem
recovery on two soil textures: implications for carbon mitigation and
grassland conservation. Ecosphere 1, art5
8 Restoration Ecology
Bakker JD, Moore MM (2007) Controls on vegetation structure in southwestern
ponderosa pine forests, 1941 and 2004. Ecology 88:2305–2319
Brudvig LA, Asbjornsen H (2009) The removal of woody encroachment restores
biophysical gradients in Midwestern oak savannas. Journal of Applied
Ecology 46:231–240
Burne HM, Yates CJ, Ladd PG (2003) Comparative population structure and
reproductive biology of the critically endangered shrub Grevillea althofer-
orum and two closely related more common congeners. Biological Con-
servation 114:53–65
Busse MD, Cochran PH, Hopkins WE, Johnson WH, Riegel GM, Fiddler GO,
Ratcliff AW, Shestak CJ (2009) Developing resilient ponderosa pine forests
with mechanical thinning and prescribed fire in central Oregon’s pumice
region. Canadian Journal of Forest Research 39:1171–1185
Covington WW, Fulé PZ, Hart SC, Weaver RP (2001) Modeling ecological
restoration effects on ponderosa pine forest structure. Restoration Ecology
9:421–431
Dodson EK, Metlen KL, Fiedler CE (2007) Common and uncommon under-
story species differentially respond to restoration treatments in ponderosa
pine/Douglas-fir forests, Montana. Restoration Ecology 15:696–708
Eviner VT, Hawkes CV (2008) Embracing variability in the application of
plant-soil interactions to the restoration of communities and ecosystems.
Restoration Ecology 16:713–729
Falk DA, Heyerdahl EK, Brown PM, Farris C, Fulé PZ, McKenzie D, Swetnam
TW, Taylor AH, Van Horne ML (2011) Multi-scale controls of historical
forest-fire regimes: new insights from fire-scar networks. Frontiers in
Ecology and the Environment 8:446–454
Flematti GR, Waters MT, Scaffidi A, Merritt DJ, Ghisalberti EL, Dixon
KW, Smith SM (2013) Karrakin and cyanohydrin smoke signals provide
clues to new endogenous plant signaling compounds. Molecular Plant 6:
29–37
Fulé PZ, Crouse JE, Roccaforte JP, Kalies EL (2012) Do thinning and/or
burning treatments in western USA ponderosa or Jeffrey pine-dominated
forests help restore natural fire behavior? Forest Ecology and Management
269:68–81
Fulé PZ, Covington WW, Moore MM (1997) Determining reference conditions
for ecosystem management of southwestern ponderosa pine forests. Eco-
logical Applications 7:895–908
Larson AJ, Churchill D (2012) Tree spatial patterns in frequent-fire forests of
western North America, including mechanisms of pattern formation and
implications for designing fuel reduction and restoration treatments. Forest
Ecology and Management 267:74–92
Laughlin DC, Moore MM, Bakker JD, Casey CA, Springer JD, Fulé PZ,
Covington WW (2006) Assessing targets for the restoration of herba-
ceous vegetation in ponderosa pine forests. Restoration Ecology 14:
548–560
Lloyd MV, Dixon KW, Sivasithamparam K (2000) Comparative effects of
different smoke treatments on germination of Australian native plants.
Austral Ecology 25:610–615
Loeser MRR, Sisk TD, Crews TE (2007) Impact of grazing intensity during
drought in an Arizona grassland. Conservation Biology 21:87–97
Miller G, Ambos N, Boness P, Reyher D, Robertson G, Scalzone K, Steinke
R, Subirge T (1995) Terrestrial ecosystems survey of the Coconino
National Forest. U.S. Forest Service Southwestern Region, Flagstaff,
Arizona
Natural Resources Conservation Service. (2013) The PLANTS database
http://plants.usda.gov (accessed 10 October 2013)
Proulx M, Mazumder A (1998) Reversal of grazing impact on plant species rich-
ness in nutrient-poor vs. nutrient-rich ecosystems. Ecology 79:2581–2592
Rokich DP, Dixon KW, Sivasithamparam K, Meney KA (2002) Smoke, mulch,
and seed broadcasting effects on woodland restoration in western Australia.
Restoration Ecology 10:185–194
SAS Institute. 2009. SAS/STAT 9.2 user’s guide. SAS Institute, Cary, North
Carolina
Schoennagel T, Nelson CR (2011) Restoration relevance of recent National Fire
Plan treatments in forests of the western United States. Frontiers in Ecology
and the Environment 9:271–277

Schultz NL, Morgan JW, Lunt ID (2011) Effects of grazing exclusion on
plant species richness and phytomass accumulation vary across a regional
productivity gradient. Journal of Vegetation Science 22:130–142
Van Horne ML, Fulé PZ (2006) Comparing methods of reconstructing fire history
using fire scars in a southwestern United States ponderosa pine forest.
Canadian Journal of Forest Research 36:855–867
Vandvik V, Heegaard E, Måren IE, Aarrestad PA (2005) Managing hetero-
geneity: the importance of grazing and environmental variation on
post-fire succession in heathlands. Journal of Applied Ecology 42:
139–149
Welch TG, Klemmedson JO (1975) Influence of the biotic factor and par-
ent material on distribution of nitrogen and carbon in ponderosa pine
ecosystems. Pages 159–178. In: Bernier B, Winget CH (eds) Forest soils
and forest land management. Les Presses de l’Universite’ Laval, Quebec,
Canada
Coordinating Editor: Valerie Eviner
Restoration Ecology
Restoration across soil parent materials
Supporting Information
The following information may be found in the online version of this article:
Figure S1. Location of experimental sites within a Pinus ponderosa forest landscape
in northern Arizona, United States. The nine mapping units we studied ranged in size
from 85 to 4,610 ha (median = 210 ha). The center of the City of Flagstaff is: 35∘ 12′ N,
111∘ 38′ W.
Table S1. Ten species displaying the greatest difference in raw cover from pre- (2003)
to 5 years post-treatment (2008) inside (ungrazed) and outside (grazed) of exclosures
in Pinus ponderosa forests, northern Arizona, United States. Values are post-treatment
minus pre-treatment raw cover and averaged across tree canopy and smoke treatments
within a soil parent material type. Gray shading marks soil parent materials where a
species had greater cover (relative to change from pre-treatment) in ungrazed compared
to grazed areas. Asterisks signify non-native species, and dashes note a species was
absent.
Received: 6 December, 2013; First decision: 7 April, 2014; Revised: 13 October,
2014; Accepted: 14 October, 2014
9