|Home > HOW to monitor? > Setting sample size > Decreasing variability|
|Helpful Hints on How to Decrease Variability|
In general, anything that decreases the variability of your counts is a blessing. With decreased variability you will increase your power to detect trends and fewer samples will be needed to detect a trend of a given magnitude.
Basic Rules of Thumb for decreasing your sampling variability (after you have chosen which species to monitor):
The variance of a population count has three components: the variability of the animal’s populations from year to year, the imprecision in the counting technique itself, and the variability of population trends from site-to-site (if more than one site is being included).
Animal populations are inherently variable and counts are inherently imperfect.
Species Choice. If you have the choice to measure a number of different populations then one of the considerations may be the degree to which the populations fluctuate from year-to-year. For example in the amphibian world terrestrial salamander populations change little from year to year compared to frog populations. Terrestrial salamanders are long-lived, have few young each year, and live in protected environments. Frogs are short-lived, lay many eggs each year, and live in a very unpredictable environment. Consequently, the differences in population variability is great.
Counting Technique. Counting techniques vary in their precision and accuracy. Indices based on territory mapping techniques will likely have a higher correlation with the real number of birds breeding in the area that do 3 minute point counts.
Choosing When to Count. All species have times of the year when they are less detectable and counts more variable. Counts should usually be taken when detectability is highest and/or most consistent. Including data from less than optimal time periods will increase the variation in the results. Weather conditions usually have a strong effect on detectability. Restricting when data are collected to certain weather conditions can also decrease the variability in the counts.
Covariates. Any factor that might influence the detectability of the animals being counted (e.g., temperature, wind, cloud cover, habitat) that is easily measured along with the count should be taken. Any influence that these factors may have on the counts can then be accounted for in the statistical analyses.
Observers. Observers are often one of the greatest sources of counting error within a monitoring program and consequently one of the greatest sources of controllable variability. Training is the most important factor. The goal of training is to insure that all observers are counting in as close the same way as possible. In particular observers should be calibrated so that they all have equal skills in identifying the animals and that they are estimating the numbers seen or heard in the same way.
While written and computer tests can be used to measure observer skill, it most important to test them in the field. A group of observers, including at least one known qualified observer (warning, experts should not be assumed to be infallible) should all be in one spot or plot and count the animals present at the same time without any communications to one another. A comparison of the results will quickly and more accurately reveal identification problems. Observers can then be retested until they meet a certain criteria for the quality of their results.
Appropriate Statistical Tests. While not something that controls the variability of the counts per se, the statistical test you use is a factor in your ability to deal with variability. The degree to which your data fit the statistical test, model, and sampling distribution used in analyzing the results will determine how much additional variability will be squeezed out of the results, effectively increasing your power to detect trends.
home | START HERE | worksheets | counting techniques | CV tools | site guide
U.S. Department of the Interior
U.S. Geological Survey
Patuxent Wildlife Research Center
Laurel, MD USA 20708-4038
Contact Sam Droege, email email@example.com
USGS Privacy Statement