A Critical Review of the Error Polygon Method

Abstract

The Error Polygon Method (EPM) of Heezen and Tester (1967) is used most frequently to quantify telemetry error. Complete scientific reporting for this method should include the confidence arcs with associated confidence levels, a measure of distance from the receiver to the estimated location, and a measure of the angle of intersection. The EPM assumption ofa normal distribution of bearing errors was rejected with a large data base (n = 940) collected on 13 transmitter locations that resulted in 388 estimated locations. Empirical data of actual error demonstrated the EPM's ability to delineate 90% error polygons that contained the actual location >90% of the time, although both the area and longest diagonal ofthe error polygons were 6.3 and 7.4 times, respectively, larger than actual error. The EPM did not give an accurate measure of location error. Int. Conf. Bear Res. and Manage. 8:251-256 The scientific method demands the reporting of accu? racy and precision be complete so future research can be built on a solid base. Wildlife researchers commonly use radiotelemetry in the study of elusive or far-ranging species, such as bears. If telemetry is used to estimate an animal's location through triangulation, complete report? ing ofthe error associated with such an estimated location is necessary. There are 5 primary sources of telemetry error: system error, topographic error, reading error, movement error, and map error. System error is the result of inaccuracies inherent in the receiving system under standard field conditions. For example, once the receiving system is in the field, the directional sensitivity of its antennas can change through twisting by the wind, cable wear, or bent elements. Topographic errors arise from the radio signals being absorbed, deflected, or reflected by landscape components (Lee etal. 1985). Topographic errors are the most significant source of telemetry error, especially in mountainous terrain. Reading errors result from misreading the compass, the rosette bearing, or from incorrectly recording a bearing and can differ between observ? ers. Reading errors are usually considered negligible or included in the angle error. Movement errors result when bearings on a moving animal are not taken simultane? ously from different receiving stations. The time lag between bearings allows a radio-collared animal to move some distance, thereby giving single bearings on differ? ent actual locations. This type of error depends on the time lag and the activity of the animal between conseeu? tive bearings (MacDonald and Amlaner 1980). Mech (1983) mentioned map error, which normally is not included in discussions of telemetry error. Plotting locations of receiving stations on maps will vary in accuracy and precision depending on the scale and accu? racy ofthe maps used. An estimated location read from hand-plotted line bearings, as done in some field work, is influenced by the width of the marked lines and the placement of the compass rosette. These sources of error affect angle error and location error. Springer (1979), MacDonald and Amlaner (1980), and Lee et al. (1985) described methods of data collection to minimize overall error. All locations should be on an x-y grid system (e.g., the Universal Transverse Mercator [UTM] system). Computer triangulation should be used to prevent increasing the effect of map error. The most common method used to quantify radiotel? emetry error is the Error Polygon Method (EPM). Heezen and Tester (1967) developed the concept of error poly? gons to allow improved locations for fixed receiving towers. They did not originally apply it to error assoei? ated with estimated locations, but to delineate a study area within which their estimates would have a known confi? dence. Springer (1979) discussed error polygons in depth. Currently, the EPM is used to report error in telemetry studies that often use the estimated location as a location without error in the subsequent analyses. The standard reporting of EPM includes a measure of bias and precision of bearings. Empirical data will show this is incomplete reporting. I collected data on the actual error of several hundred estimated locations. I was then able to compare actual error to the predicted error ofthe EPM. My purpose was to investigate: 1) if an EPM 90% error polygon encompasses the actual location 90% ofthe time when applied to a comparable data set; and 2) if the 90% error polygon is effective in reflecting telemetry location error. I appreciate the hours J.D. Hole, M.A. Horner, and E.O. Jones devoted to data collection and the support from my wife and son. Special thanks go to R.A. Powell and D.E. Seaman for their support and discussions of