Articles
SAMPLING BEE POPULATIONS IN LOWBUSH BLUEBERRY IN MAINE
Article number
446_14
Pages
101 – 108
Language
Abstract
Between 1992 and 1995, five sampling methods were evaluated for the suitability of sampling bee communities in the lowbush blueberry agroecosystem.
The five sampling methods were: 1m2 visual plot sampling, malaise traps, emergence traps, sweepnet, and linear transects.
Bias inherent in four of these methods was estimated by comparing (chisquare analysis) the relative frequency of different bee taxa censused.
The four sampling methods did sample the bee taxa with differing efficiency.
Malaise traps sampled relatively fewer honey bees in the bee community, than the other methods, but sampled the andrenid bees and other small bees such as megachilids with a high degree of efficiency.
Optimal within-field sample sizes for each method were determined for one level of precision (0.2) over a range of bee densities.
Based solely upon the number of samples necessary to estimate the mean bee abundance observed in 1994, the visual plot method was the most efficient method followed by the malaise trap, the sweepnet, and the emergence trap.
However, when labor and material costs are considered, the most efficient method for sampling the within field bee community is the visual plot count method followed closely by the sweepnet.
In addition, data collected in two fields suggested that fields not dominated by honey bees are best sampled using a stratified random sampling plan compared to a completely random sampling plan.
Allocation of sampling effort, irrespective of sampling methodology, is optimal when 38% of the samples are taken within a 10 m wide strip around the perimeter of the field and 62% of the samples are taken within the interior of the field.
This allocation requires only 82% of the sampling effort when compared to a completely randomized design.
A regional sampling plan was developed for Washington, Co., Maine from bee transect survey data collected in 1995. It was found that if all the bees in the bee community (honey bees included) were to be sampled for obtaining an estimate of regional bee abundance then the desired strategy is to take only one transect within a field and to sample as many fields as the desired precision requires (eg. 91 fields for a precision of 0.2). However, if an estimate of the regional native bee abundance is desired, the optimal strategy is to take four transects within each field and then sample the required number of fields in the region for a given level of precision (eg. 67 fields for a precision of 0.2).
The five sampling methods were: 1m2 visual plot sampling, malaise traps, emergence traps, sweepnet, and linear transects.
Bias inherent in four of these methods was estimated by comparing (chisquare analysis) the relative frequency of different bee taxa censused.
The four sampling methods did sample the bee taxa with differing efficiency.
Malaise traps sampled relatively fewer honey bees in the bee community, than the other methods, but sampled the andrenid bees and other small bees such as megachilids with a high degree of efficiency.
Optimal within-field sample sizes for each method were determined for one level of precision (0.2) over a range of bee densities.
Based solely upon the number of samples necessary to estimate the mean bee abundance observed in 1994, the visual plot method was the most efficient method followed by the malaise trap, the sweepnet, and the emergence trap.
However, when labor and material costs are considered, the most efficient method for sampling the within field bee community is the visual plot count method followed closely by the sweepnet.
In addition, data collected in two fields suggested that fields not dominated by honey bees are best sampled using a stratified random sampling plan compared to a completely random sampling plan.
Allocation of sampling effort, irrespective of sampling methodology, is optimal when 38% of the samples are taken within a 10 m wide strip around the perimeter of the field and 62% of the samples are taken within the interior of the field.
This allocation requires only 82% of the sampling effort when compared to a completely randomized design.
A regional sampling plan was developed for Washington, Co., Maine from bee transect survey data collected in 1995. It was found that if all the bees in the bee community (honey bees included) were to be sampled for obtaining an estimate of regional bee abundance then the desired strategy is to take only one transect within a field and to sample as many fields as the desired precision requires (eg. 91 fields for a precision of 0.2). However, if an estimate of the regional native bee abundance is desired, the optimal strategy is to take four transects within each field and then sample the required number of fields in the region for a given level of precision (eg. 67 fields for a precision of 0.2).
Publication
Authors
F. A. Drummond, C. S. Stubbs
Keywords
bees, pollination, lowbush blueberry, optimal, bias, sample size
Online Articles (71)