Increasing sampling efficiency of lowbush blueberries

Increasing sampling efficiency of lowbush blueberries V. O. Nams

Can. J. Plant Sci. Downloaded from pubs.aic.ca by DALHOUSIE UNIVER on 03/11/13 For personal use only.

Department of Biology, Nova Scotia Agricuttural Cottege, Truro, Nova Scotia, Canada B2N 5E3. Received 1 August 1993, accepted 31 January 1994' Nams, V. O. lgg4.Increasing sampling efficiency of lowbush blueberries. Can. J. Plant Sci. 74: 573-576. Yield estimates of lowbush blueberries (Vaccinium aigusiifolium Ait.) are highly variable. Two techniques for estimating cover, and five quadrat sizes for estimating stem density wereiested in an attempt to minimize this variability. The most efficient technique for estimating cover is pacing foi100 footsteps and counting the number of footsteps falling within blueberry patches. The most efhcient quadrat size for estimiting stem densiiy within blue6erry patches is 0.025-m2. Variability attributable to observers accounted for 23% of the variance in-stem density estimates, Uut tfris decreased to 4Vo with a short training session at the start of each field and sampling day.

Key words: Blueberries, Vaccinium angustifulium, sampling, quadrat, cover estimates, stem densities Nams, V. O. 1994. Amdlioration de l'6fficacit6 de l'dchantillonnage dans les cultures de bleuet nain. Can. J. Plant Sci. 74: 573-576. La pr(diction du rendement du bleuet nain (Vaccinium angustifolium) est sujette ir de larges variations. Afin de r6tt6cir cette variabilit6, nous avons 6valu6 deux techniques d'estimation du couvert v€g€tal et cinq tailles de quadrat pour estimer la densit6 de peuplement. La technique la plus efficace d'estimation du couvert est de marcher 100 pas en comptant le nombre de pas tomLani dans une plaque de bleuets. La taille de quadrat la plus efficace pour l'6valuation de la densit6 de peuplement dani les plaques de bleuetJ eside 0,025 m2. La variabilit6 due aux observateurs comptait porn 23% de la variance des estirnations de densii( de peuplement, mais, moyennant une petite seance de formation au d6but de chaque bleuetidre et de chaquejournde d'6chantillonnage, cefte proportion devrait pouvoft tomber a 4%.

Mots cl6s: Bleuet, Vaccinium angustifulium, lchantillonnage, quadrat, estimations du couvert, densit6 de peuplement

lft. x lft. quadrats (0.09 m2; used before metrification (Trevett 1959; Trevett and Durgin 1972),but no reports state reasons for this choice of size.

The lowbush blueberry (Vaccinium angusffilium Ait.) is a native perennial shrub that is grown commercially in northeastern North America. Commercial fields are managed areas

came from the

of naturally occurring plants, leading to high variability in

Variation among and within observers may affect estimates of stem density. Such variation arises when there is difficulty deciding objectively what constitutes a stem. Blueberry stems emerge from the ground, grow parallel along the surface and then branch or turn upwards. These ground-hugging stems may also send adventitious roots into the ground at several points. Authors generally describe their sampling techniques without defining a stem or without citing a previous definition. This can increase varition within and among observers doing the sampling. My objective in this study was to test various sampling techniques in an attempt to decrease the variability of two of the components of yield estimates: cover and stem density. I tested two different transect techniques for estimating cover. I also measured the efficiency of different quadrat sizes used for estimating stem density and measured the difference in mean stem-density estimates when stems are counted in the field versus in the laboratory. Finally, I measured variation zrmong people counting stems and report on how to minimize observer variation.

yield estimates resulting from a patchy blueberry cover within fields (Trevett 1959) and genetic variability among and within

fields (Trevett 1962; Hepler and Yarborough 1991). Variability in yield estimates of lowbush blueberries may be partitioned into three sources: l) differences in ground surface covered, 2) stem density per unit area ofblueberry patch, and 3) yield per blueberry stem. Of the variation in blueberry yield, most is due to differences in cover (Trevett 1959; Hepler and Yarborough 1991),

which is directly related to field age. Plants in young fields occur in small, dense, sparsely scattered clumps. These clumps increase in size with time, and may require 10-20 yr to completely fill in a young field. Yield estimates using long, narrow quadrats can reduce variation in cover (Krebs 1989), but optimum quadrat lengths for blueberries are very long and slow to sample (Trevett 1959). Nearly half of the variation in yield per area of blueberry patch is accounted for by differences in stem density (Trevett

1959). These differences are due to variability among blueberry clones (Hepler and Yarborough 1991), variability

within clumps and variability introduced by

sampling

techniques.

Reseirchers currently use 0.1 m2 quadrats

to

estimate

stem density (Smagula 1979; Ismail and Hanson 1982;Yarborough and Ismail 1985; Jensen 1986). This size presumabrt

,r,

METHODS Three blueberry fields with very different spatial patterns of blueberries were used for these tests. Field I was 12 ha in area with2-4 m2 patches of blueberries spaced 1-2 m apart. Field 2 was 4 ha in area with I m'patches spaced

574

CANADIAN JOURNAL OF PLANT SCIENCE

2-3 m apart. Field 3 was 50 ha in area with a solid mat of blueberries. All three fields were located in Colchester County, Nova Scotia. Each test involved two helds. The f,relds used for a particular test were chosen to be as dissimilar as possible to obtain results in a wide range of sampling situations.


Cover Two techniques for estimating cover by using transects were tested in Fields I and 2. The first technique involved laying out a measuring tape for a specified distance and measuring the lensth of line (to the closest 0. I m) thar fell inside bluJberry pui"tr"r. Two observers were required to do each sampl-e. Each replicate line had a randomly chosen starting point and direction. The five distances tested were 20, 40,60, 80 and 100 m, and all five distances were sampled on the same transect. Four pairs ofobservers each sampled three lines in each field. The second transect technique required only one observer who walked in a straight line in a chosen direction and counted the number of footsteps that fell inside blueberry patches. A footstep was deemed to be inside a patch when more than half of the foot was within the blueberry patch. As in the previous test, each replicate transect line had a randomly chosen starting point and direction. The five distances tested were 25, 50, 75, 100 and 125 paces, which corresponded.to the distances used in the other method (1.25 paces m-'). Seven observers each sampled three lines in each freld. The walking transects were completed very quickly, and it was difficult to accurately time each 25-pace section of the line. Thus, walking speed was first estimated by each observer, and then that estimate was used for timing the actual transects. Each observer's walking speed was apptieO consistently to transects done in both fields. The transect method chosen as an optimum cover estimate was that which maximized effrciency by minimizing variance x time (Wiegert 1962; Cochran 1977, p.97). For example, the longer the transect line the more precise cover estimates are, but the longer they take to do. To test for differences in accuracy (mean cover) of the two transect techniques, a t-test with unequal sample sizes was done. The required sample sizes for desired precisions were estimated by the following equation:

At each sampling location there was placed a 0.2 mz quadrat. If it did not fall completely inside a blueberry patch, it was relocated, so that there was full cover in each quadrat. This ensured that the estimate of stem density did not include effects of cover. The quadrat was partitioned to test five sizes: 0.025,0.05, 0.1, 0.15 and 0.2 m'. The observers counted the number of stems in each quadrat section. The size of quadrat chosen as an optimal stem-density estimate was that which maximized efficiency. However, sampling efficiency is dependent on time and variation, both of which differ when sampling is done over a small versus a large area. Thus the size of the area sampled may affect optimum quadrat size. To take this into account, I also estimated optimum quadrat size when sampling within a small area (several meters across). The sampling time for this small area included the time to count plus the time to walk the several meters to the next site. Variance among the quadrats over the small area was estimated by removing the component of variance resulting from differences across the whole field. Many researchers estimate stem density by first clipping plots and then counting the number of clipped stems (Ismail et al. 1981; Jensen 1986). These clip estimates may differ from those counted in the field because it may be easier to either miss stems or to count them twice. Thus, a comparison was made of counting stems in quadrats in the field versus

first clipping stems and then counting them in the laboratory. Precision was compared by comparing the ratio of variances with an F-test, and accuracy by comparing the means

with a r-test. Observer Effects To estimate the variation of the different component of stem densiry (among quadrats, among people and within people), eight 0.025 m' quadrats were placed within a small area, and seven or eight observers (one was sick one day) counted the number of stems inside each quadrat twice. To minimize the chances of remembering the first quadrat count, the observers counted stems within each of the eight quadrats, then went around and counted again. Table

I.

Percentage cover estimates using pacing Sampling required

for Sampling results

Line

(l)

length (paces)

7o Cover

Mean

SE

(N = 21)

Time ltne ' (min)

Field I

To estimate stem density, five quadrat sizes were tested in Fields I and 3 by eight observers. Each field was divided into eight sections, with each person sampling three randomly located sites within one section. At each site, each person sampled two locations within several meters of each other. This type of sampling allowed separation of small scale variation (between quadrats several metres from each other) from large scale variation (throughout the whole field).

75 100

54 46 44 44

125

45

25

48 47 44 44 44

25

Stem Density

50

Field

4.9 3.5 2.7 2.1 2.2

1.0 1.1

1.5 1.7

2.0

2

50 75 100

125

5.6 4.8

1.0

J.l

1.5 1.7 2.O

3.2 3.2

1.1

Var.

x time

a

precision 0.1

N

490 94 290 48 220 27 170 t7 200 18 1ffi 150 660 l1l 510 65 440 4'l 500 47

of

Time (h) 1.5

0.90 0.68

0.50 0.62

)4 2.1 1.6

t.4 1.6

NAMS

Table

2.

for Sampling results

Line length (m)


Field 2

Table

20 4A 60 80 100 20 40 60 80 100

7o Covet

Mean

SE

8.0 4.5 4.3 3.6 3. r 7.7 6.5 4.2 4.0 4.3

35 39

40 38 3'1

28 29

29 27 28

(N

=

area (m2) Mean

Large scale variation 0.025 Field

3

0.050 0.100 0.150 0.200

Field

I

tlme

l1.3

8670 3500 3870

203

39

64 58

l5

150

40

l4

2730

30

12

286

r4.r

5690 4960 2480

46 40 20

t6.4

25ffi

18.6

3390

75 88

t4.3

t7.4 20.4 23.4

9.6

0.050 0.100 0.150 0.200

Field

I

0.025 0.050 0.100 0.150 0.200

:

quao

(min)

1.6

1110

90

11 10

87 86

2.2 3.2 4.2 5.2

X

2l 27

103)

493 s99 62'7

766 916

t.2

4t2

1.7

428 622

I 130

96

0.9

r99

150

93 82 78

1.5

313

n5

78

2.5 3.5 4.5

810

t23

0.7

780 780 780

103

1.2

100 93

2.0

226 265 428

2.8

511

790

96

3.t

703

2.6 3.4

4)

'734

993

520 667

a

precision 0.1

of

Time

tlme

(x

89 77 76 72 75

I 110 l 100

85

for

780 780 780 790

I110

t7

202

12)

Var.

110

1

(h)

Sampling required

Time

110

810

Time

N

densities using different sized

130

1100

3

11.9

I 150

1

0.025 0.050 0.100 0.150 0.200

Small scale variation 0.025 Field 3

SE

of

precision 0.1

(min)

Sampling results (N

stem density (m -)

The evaluation of estimation procedures must take into

a

Var.

x

575

Cover

Time line-l

3. Efliciency of estimatrng stem

ouadrat

12)

SAMPLING EFFICIENCY OF BLUEBERRIES

RESULTS AND DISCUSSION

Percentage cover estimates using measured lines Sampling required

Field I

-

N

(h)

94 2.5 82 3.0 59 3.2 55 3.9 53 4.6 204 4.2 153 4.4 149 6.4 t34 ',7.5 146 l0.l 66 62 48 44 44 195 138 129 111 118

1.0 1.6 2.1

2.6 3.4 2.3 2.7

4.4 5.3 7.2

account both efficiency (precision and time) and accuracy. The most efficient method for both fields tested was the pacing method with 100 paces (Thbles I and 2). For example, to achieve a precision of 0.1 (i.e. , the confidence interval is l0% of the mean) in Field 1, the pacing estimate would require 17 samples and ake 30 min, whereas the tape-measured lines would require 30 samples and take 12 h- more than double the number of samples and 20 times the time. Given these results, pacing transects are the best alternative to more time - and labour-intensive methods of estimating cover, such as the line transect technique, or the pin-frame technique used by Smith and Hilton (1971)' Therb was no significant difference in the accuracy ofcover estimates of blueberries in Field 1 between the pacing and line transect techniques (t-test for differences in mean cover) (P > 0.1) (Tables I and 2). However, the two techniques

gave significantly different mean estimates for Field 2 ip < O.Ot) (Tables I and 2). Field 2 was pruned by the producer after the line technique was tested but before the pacing technique was tested (Field 1 was not pruned). This iuggeits that accuracy of cover estimates are affected by pruning, and the estimates should probably be carried out before pruning.

Stem Densities The most efficient quadrat size for estimating blueberry stem density in fields with both open cover (Field 1) and closed cover (Field 3), and in both large and small areas was 0.025 m2 (Table 3). Precision increased slightly for larger

quadrats, but sampling time increased much more' The

optimum size

of O.0r5m2 is

smaller than the 0.1 m2

quadrat traditionally used by blueberry researchers. The iraditional, larger quadrats required about25% more sampling time to achieve the same precision as the 0.025 m' quadrats. The smaller quadrats were also easier on worker morale, since the larger the quadrat, the more difficulty there is in counting stems when there are many stems. Accuracy did not differ signihcantly among quadrat sizes. Mean stem density did decrease slightly with larger quadrats (Table 3), but the change was small and not statistically significant. Stem densities counted by counting clipped stems in the laboratory can be directly compared with field counts of live there was no significant difference in accuracy stems -stem density was similar (P > 0.5)] or precision [mean were similar (P > 0.5)] (Table 4). [variances

Observer Effects To see if variation among people counting stems could be

minimized, a brief training session was run. At the start of each day and f,reld, all people sampling spent l0 min together counting stems in one quadrat to ensure that everyone classified and counted stems in the same way. The ratio of the variance among people with versus without a training ses-

sion was compared with an F-test.

Variation attributable to the observer in stem-density estimates accounted for 23% of the total variance. However, a brief training session significantly decreased this variation to 4%

and

(P < 0.025) (Table 4). Both variation among people each person decreased; that is'

in the estimates of

observers counted blueberry stems more consistently and more alike after training. Even if observers train together to classiff stems in the same way at the start of a project,

576 Table

CANADIAN JOURNAL OF PLANT SCIENCE

4. Variance

components for estimates of stem density when counting blueberry stems in 0.025 m2 quadrats (estimates are per quadrat rather than per square metre); effects of training session, and comparisons of counting in field versus clipped stems in laioraiory

Field 4: Joint training

Field I Source of variation

Variance component

Within people

Variance component

649 t76 776

6.43

Among people Among quadrats

19.5

88.8

Mean


a

df

2U 27 967

3.98 3.45 184.3

Variance component

164 967

1.00 5.87 167.4

5I

31.3

43.3

44.6

+2.0

+2.4

!1

95% confidence interval

Field 4: counting clipped stems in lab.

immediately before sampling

brief training session (10 min) at the srart of each day and

field can significantly standardize estimates even further. Because frelds vary in the amount of dead leaves, sparseness.of cover, proportion of dead stems, etc., tr;ining together for each field ensures that everyone treats thes; differences equally.

ACKNOWLEDGMENTS f thank Sharon Andrew, Beth Bishop, Karen Henley, Nancy Kolstee, Michelle Langille, Scott Murphy and Lisa Wilkinson

for help in field work, and Magi Nams, Chris Jordan and Leonard Eaton for editorial assistance. Cochran, W. G. 1977. Sampling techniques. John Wiley & Sons, New York, NY. 428 pp. Hepler, P. R. and Yarborough, D. E. 199f . Natural variability in yield of lowbush blueberries. HortScience 26: 245-246. Ismail, A. A. and Hanson, E. J. ftn2. Interaction of method and date of pruning on growth and productivity of the lowbush blueberry. Can. J. Plant Sci. 62: 677-682. fsyail, A. A., Smagula, J. M. and yarborough, D. E. l9gl. Influence of-pruning methods, fertilizer and terbaii,t on the growth and yield of lowbush blueberry. Can. J. plant Sci. 6t: O1_lt. Jensen, K. I. M. 1986. Response of lowbush blueberry to weed control with atrazine and hexazinone. Honscience 2l: ll43-ll4.

a

Krebs, C. J. f989. Ecological methodology. Harper & Row, New

York, NY. 654 pp. Smagula, J. 1979. Effect of long-term N or NPK fertilization on Iowbush blueberry yield and plant stand. Pages 80-83 jn J. N. Moore, ed. Proc. of the 4th North American Blueberry Workers Conf., University of Arkansas, Fayetteville, AR. Smith, D. W. and Hilton, R. J. 1971. The comparative effects of pruning by burning or clipping on lowbush blueberries in north-

J. Appl. Ecol. 8: 781-789. Trevett, M. F. 1959. Growth studies of the lowbush blueberry 1946-1957. Maine Agricultural Experiment Station, University of eastern Ontario.

Maine at Orono, ME. Maine Agr. Expt. Sta. Bull. 581. 58 pp.

Trevett, M. F. f962. Nutrition and growth of the lowbush blueberry. Maine Agricultural Experiment Station, University of Maine at Orono, ME. Maine Agr. Expt. Sta. Bull. 605. l5l pp. Trevett, M. F. and hrgin, R. E. 1972. A progress report on pruning lowbush blueberries with Paraquat, Dinitro herbicides and mowing. U. Maine, Res. Life Sci. 19: l-4. Wiegert, R. G. 1962. The selection of an optimum quadrat size for sampling the standing crop ofgrasses and forbs. Ecology 43:

r25-r29. Yarborough, D. E. and Ismail, A. A. 1985. Hexazinone on weeds and on lowbush blueberry growth and yield. HortScience 20:

406-Q7.