This file was created by scanning the printed publication.
Errors identified by the software have been corrected;
however, some errors may remain.
Assessing the Impacts of Timber Harvest on
a Northern Arizona Rare Plant, Clematis
hirsutissima yare arizonica, Through Canopy
Manipulation and Matrix Demographic
Analysis
Edward Bennett Smith 1
Abstract; - In a study of the demographics of a rare plant, the Arizona
Leather 'Flower (Clematis hirsutissima var. arizonica), we assessed the
effects of canopy cover on Clematis survival and reproduction. Varying
amounts of ponderosa pine canopy cover or shading were removed
(branches, poles, and saplings) from some plots, while artificial shading was
added to other plots in the Coconino National Forest. Results indicate that
experimental removal of canopy had detrimental effects on seed production,
while shade addition in previously low-shaded areas had a positive effect
on seed production. These changes may be stochastic, but matrix analysis
shows that all plots had eigenvalues below 1.0 (0.129-0.931), indicating that
they are not growing, and are in long-term decline. These findings are
important, because there are only 1500 extant plants of this southernmost
subspecies, and more than half the population exists in areas of current or
planned timber sales. Factors besides tree canopy that may be affecting
Clematis' survival include climate, fire, insects, forest floor depth, introduced
ungulates, and other ecosystem variables and their interactions. The
U.S.D.A. Forest Service should continue to protect this plant's habitat and
monitor its demography, and if possible, expand studies into other landscape
processes that may have affected the reproductive biology of this plant.
INTRODUCTION
its existence on lands under its control. The author and Dr. Joyce
Maschinski continued a long-term demography study on plots
of Arizona leatherflower within the Lake Maty timber sale area.
Demographic studies quantify the change of individually
mapped plants from state to fate, so that comparisons can be
made of the eigenvalues, fitnesses or growth rates of different
samples over time intervals. Demographic analysis using
matrices is very useful because the source of a plant's decline
can be pinpointed very accurately. Plant populations are divided
by age-, size- or stage class, and the relative contributions from
these classes, or elasticity analysis, can be quantified, and the
class or classes that are diminishing or not contributing to the
plants overall fitness can be identified. If the particular class
does not contribute to the overall fitness of the plant, for instance
the seedling stage, then experiments can be designed to increase
the survival of plants in this stage. Experimental results from
The process of listing plants as threatened or endangered takes
up to ten years, compared to animals, which take 2-5 years
(Phillips, pers. comm.). The scientific and political processing
necessary for listing therefore may take too long to protect a
plant before numbers of individuals drop below minimum viable
populations. The Arizona leatherflower (Clematis hirsutissima
var. arizonica, fig. 1) is a U.S. Fish & Wildlife Setvice categOly
two (C2) plant, for which there is insufficient data to decide
between listing or not listing. In the interim, the U.S. Forest
Service manages this plant as 'sensitive' so as not to jeopardize
1 Master's degree candidate at Northern Arizona University.
Flagstaff, Arizona. USA.
348
Figure 1. -
Clematis hirsutissima var. arizonica.
these manipulations could be quantified with the matrix model,
and management decisions could be based on the analysis of
these data. Examples of this type of model analysis have been
done for several species, including the effects of prescribed fire
on a tropical savanna grass, testing spatial and temporal variation
of a perennial bunchgrass, predicting growth of temperate
deciduous forest stands, and for conserving an endangered
animal species.
juveniles have neither wide leaves nor reproductive structures.
Canopy cover for the plots was detennined with a spherical
densitometer and a photometer, and plots were divided into
subgroups of 'high', 'medium' or' low' canopy, with the 'high'
canopy intercepting the greatest amount of solar radiation. For
each subgroup, half the plots were randomly designated as
controls, and half as experimental, in which the canopy or degree
of shading was manipulated. In early spring 1992, the plots with
'high' and 'medium' canopy levels were altered by removing
branches or small trees surrounding the plots (reduced 5-25%),
and during the summer of 1992, plots of 'low' canopy cover
were altered -by the additi~n of 85-95% shade structures (fig. 2
and fig. 3). Control plots provide baseline data and comparisons
for the experimental plots.
MATERIALS AND METHODS
Forty 2.3 m2 (square meter) plots were established in the
summer of 1991, encompassing 401 of the approximately 1500
extant plants in Arizona. Plants were identified on an x-y axis
grid within each plot, and measurements taken on each plant for
the number of stems, number of flowers, number of flowers
eaten, and number of flowers that set seed. Plants were identified
as seedling (1), juvenile (2), or reproductive development stage,
based upon leaf motphology and presence of flowers. Seedlings
and young ramets have characteristically wide, flattened
!cotyledons, although it is vety difficult to differentiate between
them. Reproductive plants have flowers or seec:iheads, and
Matrices
TIle transition matrices were constructed from the raw data
collected over the two field seasons. TIle columns represent the
'state' or condition of the plants in 1991, while the rows
represent the 'fate' or condition of the plants in 1992. The entries
349
Control
Experimental
within the matrix are proportions of plants that went from one
state to the corresponding fate. The stages for the plants were
delineated as follows:
High shade (n =12)
1991
1992
x =89.3%
1991
89.3%
Cover
1992
Stage 1
Seedlings or new ramets
Stage 2
Juveniles, nonreproductive
94.5%
65 -75%
Shade Removal
Stage 3
Reproductive, <9 stems
1991
Stage 4
Reproductive, 9-25 sterns
Stage 5
Reproductive, >25 sterns
Medium shade (n =12)
1991
1992
x =55.6%
55.6%
1992
The transition matrices were analyzed using Eigenfmder in
the MacMathTM package. The eigenvalues (A) give the growth
rates for the group of plots as indicated. This is a complex
distillation of how each stage class contributes to the overall
fitness of the group of plants (Caswell, 1989). Unfortunately, it
is not possible to compare the eigenvalues from only one
transition matrix to another, rather two or more years are
necessmy for comparisons of spatial, temporal or experimental
variability (Moloney, 1988). Sensitivity analysis is used to
identify the most important stage(s) changes that are contributing
to a plant's growth or decline (A).
56.8%
35 -25%
Shade Removal
Cover
Low shade (n =16)
'.
1991
1992
1991
1992
x = 9.2%
9.2%
7.13%
85 - 95%
Cover
Shade Removal
Figure 2. - Experimental design for canopy manipulation
(Maschinski, 1990).
Figure 3. -
Shade addition was accomplished with portable structures which were covered with pine boughs or wood lath.
350
RESULTS
Low Control
93.1 %
J.. - 0.931
There were 401 plants sampled from the 40 plots, with 283
adults (70.57%), 84 juveniles (20.95%), and 34 seedlings
(8.47%). As is true for many rare plants, this demographic
distnbution indicates very low mtes of regeneration Maschinski
found that the number of flowers that set seed in low canopy
cover experimental plots increased from 162 in 1991 to 385 in
1992 (q'= 7.48, <pO.05), indicating that the addition of shade
significantly improved the potential for sexual reproduction
Also, there was a high canopy removal treatment effect on the
number of flowers that set seed (fig. 4). Plots that had high
canopy cover reduced through thinning had significantly lower
numbers of flowers that set seed (-100 in 1991 decreased to 44
in 1992), compared to controls (37 in 1991 increased to 79 in
1992; F=2.76. p<O.I).
The eigenvalues presented in figures 5 and 6 indicate that all
sampled plots are in general decline (J... < 1.0), which is further
evidenced by the lack of plants in the earlier stages, and general
lack of larger, reproductive plants. An eigenvalue of 1.0 indicates
a group of plants that is neither declining nor increasing, whereas
an eigenvalue greater than 1.0 would indicate growth.
2
2
1--=-1-.------ ...------+-0-.9-31-0-1
3
3
Q,.?LL _.....Jl_1.:.1.~
2
~~.t!.. __ ...
State
Fate ..-1
0.04444
0.04444
Hioh Control
J..- 0.662
State
1
2
Fate
1 0. 0366Z
_9.~
0.09091 1-'0.01818
0.12713
0.03636
0.65454
2
3
2
3
0.71111
_1
0.16667
0.65483
~!l.~J.l:!!1J.J~.!!!i!!l.!!!!..t.:: 1 ..':..Q:~~?__.
3
l ___.._. __ . _..._._._._..__.... ___ ._ _.__.._. __ _
2
3
0.08333
3
~~~.!!l_(_. ___ .. __ ... __ ..~_:' ..
State
93.8%
Low EXDerimental
J..- 0.646
State
1
2
Fate
1 0.04167
~~ __ I -__--" -.---~ _ _ _l
Fate
1
1
... _Jl_~2~
2
3
______ _______ _ __ . _. ____
0.06667
0.09333
0.17333
0.01333
0.53333
t!i9!!....Experimental
88.4%
~~ _.. _--._'-'
State
1
2
3
Fate
1
3
2
3
0.02326
0.11628
0.09302
0.02326
0.h2791
Figure 5. - These transition matrices were constructed from data
for the 1881 (state) and 1882 (fate) data. The numbers within
each cell represent the proportion 0 fplants from each
specified canopy class that went from one state to another.
Each matrix represents plants from 6 - 8 plots, and does not
include any recreltment, either sexual or aseuxal. In these
3 x 3 matrices, stages 1 and 2 represent the seedling and
Juvenile stages, respectively, as in the 5 x 5 matrices, but
stages 3, 4, and 5 have betm lumped into one reproductie
stage, 3 (as explained in text). the A. values are the dominant
eigenvalues, or overall growth rates for the group of plots.
The overall percentage of living plants remaining in 1992 is
given in the upper right had corner. For ease of comparison,
and to keep as many cells as possible filled with values,
these 3 x 3 matrices are preferable to the 5 x 5 matrices.
Number of Flowers That Set Seed
Low Canopy Experimental (LE) &
High Canopy Experimental (HE)
Medium Control
Fate
60
en
•
84.4%
5
1
5
LE Set Seed
Medium EXDerimental
SiatB-
50
Gi
en
4
2 0.04444 0.04444
3
0.04444 0.57778 0.08889
4
0.02222
0.02222
'C
GI
GI
Ie - 0.578
&~
70
1
Ie = 0.499
3
88.0%
-"4 -~-5
Fate
1
-~_
2 0.06667 O~0~333~R1J_n ________:.~
0.17333 0.49333 0.013~4 _. _ _ _
_____.2. ___
40
--11---- -----,.9:.Q?§.!E-e.---
~
GI
0
==
30
u::
20
Hiah Experimental
State
1
Fate
1
10
Year
83.7%
4
5
2 0.02326 0.11628 0.02326
3
0.09302 0.48837 0.02326
4
0.02326 0.02326
5
0.02326
0
1991
Ie - 0.494
1992
Figure 6. - Transition matrices for the data presented in a 6 stage
format, which is how the data were collected. The A. values
are the dominant eigenvalues, or overall growth rates for
the group of plots. The overall percentage of living plants
remaining in 1982 is given in the upper right had corner.
Note that many cells are empty below the diagonal,
indicating a severe lack of plants in these transition stages.
Figure 4. - Plots that had high canopy cover reduced through
thinning had significantly lower numbers of flowers that set
seed compared to controls, and the addition of shade
Significantly improved seed set in low canopy cover plots.
351
Depth of Forest Floor
DISCUSSION
Logging and fire exclusion have drastically changed the
character of the canopy of these forests. The litter layer has also
presumably changed, the depth of which is a function of leaf
deposition rates, fire frequency, and decomposition rates. Natural
fire frequency disruption through suppression could have slowed
the litter decomposition and nutrient mineralization processes,
limiting the amount of nutrients available to growing and mature
plants, and making it more difficult for seedling establishment.
While a thick litter layer has the 'mulch effect' of moderating
moisture loss and s~owing soil temperature change, it also has
a significant effect on the amount of precipitation available to
plant roots by interception and subsequent evaporation
The low eigenvalues, and paucity of plants in the earlier
stages of growth indicate that the plant is having trouble
reproducing, sexually and asexually. It has been previously noted
that presumably mammalian herbivory accounts for the loss of
significant numbers of flowers and seeds within some plots. It
has also been noted that many insects frequent these plants, and
that up to 90% of once-viable seeds are parasitized by some
unknown weevil or beetle (Maschinski, 1989).
The plant's decline may be in response to the periodic drought
of 1988-89, but one would expect a quick rebound to such a
transient change in precipitation. With the record-breaking
rainfall we have had over the last two years, observations from
another field season should test this hypothesis.
Insects
Since there is an abundance of insects that utilizes these
plants, and the number of seeds parasitized by insects may be
significant, the suppression of fire may have a positive impact
on any populations of insects that ovelWinter in the soil,
allowing them to maintain imbalanced, epidemic numbers.
Perhaps frequent fire in the past maintained insect populations
at stable levels that allowed. higher levels of Clematis
reproduction than is observed today. Further study is warranted
in this area.
Elk tracks were noticed around some of the
manunalian-herbivore impacted plots. The management of this
introduced species remains controversial, as some believe elk
wmbers are out of control. While there was a natural population
of Merriam's elk, many believe its range was closer to the
Mogollon Rim, and did not frequent the areas now inhabited by
the larger herds of Rocky Mountain elk. This increased grazing
pressure from large herds of elk (and cattle) may result in elk
having to consume less desirable plants like Clematis, which is
in the Ranunculaceae family. This family contains several
poisonous genera (Cimcifuga or bugbane, Aconitum or
monkshood, Delphinium, Anemone and Columbine), which are
nonnally avoided by ungulates.
to
Fire
. Clematis grows under a ponderosa pine canopy, which is
fire-tolerant. Perhaps Clematis is fire-tolerant or even
fire-dependent, relying on periodic removal of the chaff,
composed of old stems, to allow a higher rate of photosynthesis,
and to eliminate competition from grasses, forbs and pine trees.
A similar demographic study on a tropical Andropogon grass
in Venezuela showed dramatically different eigenvalues for
unburned (0.2762) versus burned plots (1.2524). Most of this
discrepancy was shown to be due to the growth, survival and
reproduction in the two smallest size classes, which were shown
by elasticity analysis to be the two most important classes to
population growth (Silva et alia, 1991). Analysis of fire scars
in northern Arizona has shown that some forests experienced
bum frequencies of 2-15 years (Covington and Moore, 1992),
indicating that fire may have played a critical role in maintaining
this ecosystem. Fire played a role in the thinning of trees, so
that there were fewer but larger trees per acre. This could have
affected water relations in the soil, as fewer young, fast-growing
trees may have reduced competition for available soil water.
With more trees in the older age classes, the forest would have
had interlocking canopies providing more shade with fewer
trees. The character of ponderosa canopy changes with age, to
a more patchy, heterogeneous spatial arrangement, allowing light
to pass through while maintaining leaves at lower temperatures,
thereby reducing evapotranspiration There probably exists an
ideal canopy closure level that maximizes plant fitness by
maintaining high photosynthetic rates, while minimizing water
loss through evapotranspiration and evaporation from the soil
and litter. This idea of there existing an ideal overstory canopy
composition that is neither too dense nor too sparse is supported
by preliminaty analysis of 1993 field data (Maschinski, pers.
comm.).
CONCLUSIONS
It is interesting to note that the effects of the experimental
canopy manipulation were statistically significant within one
season. The number of flowers that set seed increased in low
canopy plots where shade was added, and the number of
'seedlings' decreased in high canopy plots that had canopy
removed. Since this variation may be due to other factors, such
as El Nifto weather patterns, this demographic study should be
continued. Because the eigenvalues are so low, and comparisons
among treatments require more years' data, the intensity of this
352
Covington, W. W., Moore, M.M. 1992 Postsettlement changes
in natural fire regimes:Implications for restoration of
old-growth ponderosa pine forests. in Old growth forest in
the Southwest and Rocky Mountain regions. USDA Forest
Service Technical Report RM-213. pp 81-99.
Crouse, D.T., Crowder, L.B., Caswell, H. 1987. A stage-based
population model for loggerhead sea turtles and implications
for conservation Ecology 68(5) 1412-23.
Galeano-Popp, R. 1988. Clematis hirsutissima Pursh var.
arizonica (Heller) Erickson: Preliminary study and sUlVey of
the Lake Maty timber sale area Report prepared for the
Coconino National Forest, Contract No. 40-8167-8-276.
Manders, P. T. 1987. A transition matrix model of the population
dynamics of the Clanwilliam Cedar (Widdringtonia
cedatbergensis) in natural stands subject to fIre. Forest
Ecology and Management. 20: 171-86.
Maschinski, I 1993 Report of 1992 research on Clematis
hirsUtissima Pursh var. arizonica for Coconino National
Forest cost-share studies.
Maschinski, I, Phillips, B.G. 1991. Impacts of experimental
manipulation of overstory on Arizona Leatherflower
(Clematis hirsutissima var. arizonica) demography. Report to
Coconino, Kaibab, and Prescott National Forests, U.S. Forest
Service.
McDougall, W.B. 1973. Seed plants of Northern Arizona.
Museum of Northern Arizona, Flagstaff. p. 181.
Moloney, K.A. 1988. Fine-scale spatial and temporal variation
in the demography of a perennial bunchgrass. Ecology
study should be maintained, and possibly expanded to include
some of the Clematis populations in Walnut Canyon, Rio de
Flag drainage, and Volunteer Canyon.
Logging activity has changed the Southwest landscape to such
a degree that sensitive areas, such as Clematis habitat, should
be protected until enough infonnation has been gathered to
delineate a clear picture of all the long term effects of
management activities.
Priorities for Further Study
1.
2.
3.
4.
5.
Maintain or expand present demography research.
Begin prescribed fire or siIIlulated fire studies along
with ungulate-exclusionary fencing.
Capture and identify insects that impact the plant.
Emphasize those insects that forage on seeds quantify samples (% seed heIbivory) from different
canopy, experimental, and fire regime plots.
Develop stem map from stUmps and current stems.
With tree coring, timber sale history and regression
from current canopy and site index, develop fire
and stand density history to determine forest
structure in Clematis habitat for the last 100+ years.
Measure predawn and noon water potential values
under different canopy levels and treatments to
elucidate water stress relations.
69(5): 1588-98.
ACKNOWLEDGEMENTS
Padien, OJ. 1992. Plant spatial pattern and nutrient distribution
in pinyon-juniper woodlands along an elevational gradient in
northern New Mexico. Int. J. Plant Sci. 153: 425-33.
Palmer, M.E. 1987. A critical look at rare plant monitoring in
the United States. Biological Conservation 39:113-27. ~
Silva, IF., Raventos, I, Caswell, H., Trevisan, M.C., 1991.
Population responses to fIre in a tropical savanna grass,
Andropogon semibeIbis: a matrix model approach. Journal of
Ecology 79:345-56.
Sokal, RR., Rohlf, FJ. 1968 Biometrics. W.H. Freeman, San
Francisco.
Solomon, D.S., Hosmer, R.A. 1986 A two stage matrix model
for predicting growth of forest stands in the Northeast. Can.
I For Res. 16:521-28.
This research was funded by the Arboretum at
Flagstaffffransition Zone Horticultural Institute and the
Coconino National Forest's Rare Plant Program.
REFERENCES
Caswell, H. 1989. Matrix Population Models. Sinauer
Associates, Inc., Sunderland, MA, USA. 45-53, 118-55,
161-227.
353