Use of Microsite Sampling to Reduce Inventory Sample Size
Item Type text; Article
Authors Larson, L. L.; Larson, P. A.
Citation Larson, L. L., & Larson, P. A. (1987). Use of microsite
sampling to reduce inventory sample size.Journal of Range
Management,40(4), 378-379.
DOI 10.2307/3898743
Publisher Society for Range Management
Journal Journal of Range Management
Rights Copyright © Society for Range Management.
Download date 09/08/2022 18:53:44
Item License http://rightsstatements.org/vocab/InC/1.0/
Version Final published version
Link to Item http://hdl.handle.net/10150/645258
Use of Microsite Sampling to Reduce Inventory Sample Size
L.L. LARSON AND P.A. LARSON
Abstract
The objective of this study was to determine if a stratification of
microsites within range communities could be used to effectively
reduce sampling variation and hence sample size. Two grassland
communities were stratified by microrelief patterns. Random
sampling designs were applied to each community as well as micro-
sites within the community. Stratification of the community, based
on local dniluge patterns, reduced standard errors significantly.
The pooled microsite data sets were not significantly different from
simple random sample
data sets for the communities. Sample size
reductions
of 50 and 60%
were observed using the microsite srm-
pling technique.
Key
Words: inventory, sample size
Vegetation inventories provide information that is used to for-
mulate land management decisions (McQuisten and Gebhardt
1983). The trend in range management and mined land reclama-
tion has been to utilize objective sampling designs that yield valid
statistical inferences about the area being studied. Much of this
trend is the direct result of our increased reliance upon analytical
results for regulation, enforcement, and litigation (McQuisten and
Gebhardt 1983). Unfortunately, this trend has also raised inven-
tory costs.
A number of sampling procedures and improvements have been
described in the literature during the past 5 years (Shute and West
1982, Ahmed et al. 1983, Butler and McDonald 1983, Strauss and
Neal 1983, Taha et al. 1983, and Caranda and Jameson 1986).
However, it is important to be aware that many reductions in
inventory costs depend on the ability of the sampler to identify
sources of sampling variation and to develop study designs that
will minimize the number of samples needed to obtain valid statis-
tical inferences. Partitioning of internal sources of community
variation should be a primary objective when sample size is a
limiting factor. The objective of this study was to determine if a
stratification of microsites within range communities could be used
to effectively reduce sampling variation and hence inventory sam-
ple size.
Methods
Bluegrass
(Poa sandbergiilstipa
comata)and needlegrass
(Stipa
comata/Stipa cohmbiana)
communities near Williams Lake and
Merritt, British Columbia, were selected for this study. Each com-
munity was mapped to form microsites based on drainage patterns
within the community. The partitioning process has the potential
of identifying 4 possible patterns of microsite relief within a given
community: (1) The nose-the driest areas of the community usu-
ally having off-site drainage, (2) Side slopes-areas with straight
contours and gradual off-site drainage, (3) Foot slopes-the gentle
lower parts of the side slope, and (4) Hollows-areas of drainage
accumulation (Cook and Doornkamp 1974).
Each community was sampled using Daubenmire quadrats (20
X 5&m). The placement of the individual quadrats was determined
by using a random sequence of compass bearings and distances
within the communities. Each observation was recorded by sample
number and microsite position. Foliar cover and herbage yield
data were collected by species. Green herbage weights were
adjusted using correction factors obtained from oven-dry samples
(65” C). Sample adequacy for vegetation cover and wet herbage
Authors are assistant professor, Department of Rangeland Resources, Oregon
State University, and ecological consultant, LaGrande Oregon.
Manuscript accepted 9 March 1987.
378
yield were monitored in the field using Stein’s two-stage sampling
procedure (Steel and Torrie 1980). Sample size requirements for
the cover estimates were larger than those observed for wet herbage
yield. Consequently, the sample size requirements for the cover
estimates were used as the minimal sample size requirements for
subsequent comparisons.
Two sets of data were provided by this procedure: (1) A simple
random sample for the entire community, and (2) A stratified data
set based on drainage patterns within the community. Community
estimates from the microsite data were obtained by aggregating the
stratified data sets. Weighted mean and standard error calculations
for stratified samples (Snedeor and Cochran 1978) were used to
obtain these estimates. A planimeter was used to obtain area
estimates (weighing factor) for the microsites.
Results and Discussion
The stratification process recognized 3 microsites in the blue-
grass community (nose, slope, and hollow) and 2 sites in the
needlegrass community (nose and slope). The number of micro-
sites in each community is a function of the drainage pattern of the
land area and the community boundary. The results of an analysis
of variance of these data sets are provided in Table 1. In both
communities microsite differences were shown to be significant
sources of community variation. Mean differences between the
strata ranged from 19-60% cover in the bluegrass community and
1744% cover in the needlegrass community. Similar differences
were observed in wet and dry herbage yield (bluegrass: 11.8-56
g/.lm*, 5.4-24.5 g/.lmr; needlegrass: 13.8-20.8 g/.lmr, 5.7-11.7
g/. lm*). From a sampling perspective, stratification reduced the
standard errors associated with the population estimates by 46%
for cover, 43%for dry herbage yield, and 44% for wet herbage yield
in the bluegrass community. Results were similar in needlegrass
community: standard errors were reduced 42% for cover, 32% for
dry herbage yield, and 23% for wet herbage yield. The standard
error reductions indicate that the accumulative influence of micro-
site drainage patterns contributes a significant proportion of the
variability found in the vegetation cover and herbage yield esti-
mates for the community.
Microsite sampling reduced the overall sample size requirements
for the 2 communities by 50% for the bluegrass community and
60% for the needlegrass community (Table 2). The sample size
numbers reported in Table 2 represent the largest sample size
required to meet the stated confidence level for all 3 community
attributes. The number is a composite of the random samples
collected from the respective microsites. The weighted mean esti-
mates for the bluegrass cover, dry herbage yield, wet herbage yield
and the needlegrass cover and wet herbage yield estimates fall
within 1 standard error of the simple random sample estimates for
communities. The weighted estimate for dry herbage yield in the
needlgrass community exceeded
I
standard error but is within 1.7
standard errors of the simple random estimate of the mean. Conse-
quently, the confidence limits for both communities indicate that
there is no significant difference between population estimates
obtained using simple random sampling or microsite sampling.
Conclusion
Rangeland inventories are normally conducted on limited
budgets, which underscores the need for maximizing the informa-
tion obtained through the inventory process. In addition inventory
funding determines to a large extent inventory sample size (Steel
JOURNAL OF RANGE MANAGEMENT 40(4), July 1987
T8bk 1. An8lysk of v8rhnce of stmt8 d8t8 sets from tbe bluegmss 8nd needle communitka.
Attribute Source
Cover between
within
total
Herbage yield between
(dry) within
total
Herbage yield between
(wet) within
total
lSigniiicant at
.Ol level.
DF
2
26
28
2
26
28
2
26
28
Bluegrass
Needlegrass
MS
F SE
DF
MS F
SE
3178 318 1 4094 52’
100 1.80
320 3.32
;; 78 1.66
226 2.84
697 29*
I
379 32.
24 .90 26 .67
71 1.57 27
::
.98
2116 30*
I
510
19’
69 1.54 26 27 .98
215 2.72 27 45 1.27
Table 2. Compukon of the dmpk raodom and random microsite date sets
for the bluepass
811d
needkgmss communitks.
Bluegrass Needlegrass
Simple Random Simple Random
Random Microsite Random Microsite
Attribute
Estimate
Sample Sample Sample Sample
Cover Sample size 29 14 28 I1
Mean (%)
38.
I
38.4 26.8 24.0
Standard Error 3.32 2.85 2.84 1.72
CI (.05) 31.4-44.7 32.1-44.6 21.1-32.5 20.1-27.9
Herbage Sample size 29 14 28 11
yield
Mean (g/. lm) 13.7 14.0 8.8 7.3
(dry) Standard error 1.57 1.48 .98 .69
CI(.O5) 10.5-16.8 10.7-17.2 6.8-10.7 5.7-8.8
Herbage Sample size 29 14 28 II
yield
Mean (g/. lm) 27.6 29.4 17.9 15.7
(wet) Standard error 2.72 .I8 1.27 .98
CI (.05) 22.1-33.0 25.4-33.3 15.3-20.4 13.6-17.8
and Torrie
1980).
Consequently, land managers are often faced
with a decision to either revise their objectives, due to the necessity
of reducing the number of samples collected, or delay the inventory
until adequate funds are available.
Microsite sampling within a plant community is an alternative
when sample size requirements exceed budgetary constraints.
When properly applied, stratified sampling results in smaller var-
iances than simple random sampling designs (Cochran 1977). A
sample size savings of
50
and 60% were observed in the tested
communities without sacrificing randomization. However, the
area occupied by the microsite must be measured accurately to
avoid sources of error in the calculation of weighted means and
standard errors.
Literature Cited
Ahmed, J., C. Bonh8m, 8nd W. Laycock.
1983.
Comparison of techniques
for adjusting biomass estimated by double sampling.
J. Range Manage.
36~217-221.
Butler, S., 8nd L. McDon8ld. 1983.
Unbiased systematic sampling plans
for the line intercept method.
J. Range Manage.
36463468.
C8r8nde, V., 8nd D.
J8meson.
1986.
Combination of weight estimates with
clipped sample data.
J. Range Manage. 39428-431.
Cochrrm, W.G. 1977.
Sampling techniques.
John Wiley, New York.
Cooke, R.U., end J.C. Doomkunp. 1974.
Geomorphology in environmen-
tal management. Clarendon Press, Oxford.
McQukten, R., 8nd K. Cebb8rdt. 1983.
Analytical reliability in
the de-ci-
sion making process-the numbers game. J. Range Manage. 36: 126-128.
Shute, D., 8nd N. West. 1982. Two
basic methodological choices in wild-
land vegetation inventories; their consequences and implications.
J.
Appl. Ecol. 19~249-262.
Snedecor, G.W., 8nd W.G. Co&an. 1978.
Statistical methods, 6th ed.
Iowa State Univ. Press, Ames.
Steel, R., end J. Torrie. 1980.
Principles and procedures of statistics, 2nd
ed. McGraw-Hill, New York.
Strauss, D., 8nd D. Ne8l.
1983. Biases in the step point method on bunch-
grass ranges. J. Range Manage. 36~623-626.
Taha,
F., H. Fkser, 8nd R. Rks. 1983.
A modified IO&point frame for
vegetation inventory. J. Range Manage. 36~124-125.
JOURNAL OF RANGE MANAGEMENT 40(4),
July
1987
379