The Swamp Stomp
Volume 16, Issue 17
A few months ago, I was observing one of the Swamp School’s wetland delineation training courses in the field. I was interested in some of the techniques that the students were using to collect data. I watched several of them taking measurements and then entering those values into the data form. Those values were then used calculate the prevalence index and dominance test and to evaluate hydrology and soil indicators. All these measurements were then used to determine if the assessed area was considered a wetland. The Army Corp of engineers would review the area at a later time and determine if these measurements were done correctly and if the assessment was accurate.
The first student I observed was reviewing the vegetation section of the data form. She was determining the absolute percent cover of three tree species that were recorded in the data form. Here observations were: River Birch – 34%, Red Maple 12%, and Black Oak – 6%. Her wetland partner had the same tree species identified but his absolute percent cover values were quite different: River Birch – 54%, Red Maple – 33% and Black Oak – 18%. Why were their values so different? Who was closer to the true values of absolute percent cover? This is course led me to wonder that if both students were assessing the same area would their wetland conclusions be different?
In statistics, the variability that I observed is called “measurement error” and is present in all wetland delineations, as well as any process where measurements are taken by individuals. Measurement error has two components: accuracy and precision. Accuracy is the difference between the average measured values and the true value. Precision is how close all the measured values are to each other. The graph below visually demonstrates the difference between accuracy and precision. Source: (http://kaffee.50webs.com/)
Going back to our example of the percent cover, let’s assume that the true values were River Birch – 43%, Red Maple – 21% and Black Oak – 15%. If we took the average of the percent covers for each person it would be Birch – 44%, Red Maple – 22% and Black Oak – 12%. So it would seem that the accuracy of all the assessors (average values) was good. But the precision in the measurements (variability between assessors) was not very good. If we had asked each person to measure the same percent covers multiple times (unknowingly of course) we could have also measured the variability within the assessors.
Why does this error in measurement exist? The answer, in general terms, is that each operator has slightly different methods for calculating percent cover. In order to correct measurement error, the wetland delineation team would have to improve the process of how the percent covers were estimated. For instance, they could have a written procedure that explains exactly how the process should work, including pictures that demonstrate different percent covers.
There are of course other measurements taken during a wetland delineation that have potential for measurement error. Examples include measuring the soil depth and determining the color percentage for the soil section. Or determination if hydrology indicators are present at the site, such as surface soil cracks or moss trim lines.
Error exists in all wetland delineation processes where measurements are taken. You will not be able to eliminate all the error but you will need to take steps to ensure that the error is minimized. Error in your measurements could lead you to making incorrect conclusions regarding the decision about a site being a wetland. Sometimes these errors can cost your company thousands of dollars.
There are statistical methods available such as control charts, analysis of variance, and attribute assessments to quantify the amount of measurement error that exists in your processes. These techniques can be useful in assessing measurement error in any data collection process. Understanding the concepts of measurement error, the tools to measure it, and being able to improve your measurement processes will provide you with meaningful data which you can use to make fact-based decisions.