Remote radiotracking: a study of the methodology

Abstract

Remote radiotracking is a technique used widely in the study of free-ranging wild animals. Location estimates for radiotagged animals are obtained from the intersection of radiobearings (or radioazimuths) to the radiotagged animals from two or more fixed receiving sites. In this study the geometric model and mathematical procedures used to obtain location estimates from remote radiotracking data were described. Sources of errors in the location estimates were identified. There is a critical review of currently accepted methodologies to deal with errors in location estimates. The failure of error polygons to provide a realistic description of errors in location estimates was demonstrated. It was also shown that techniques based on circular statistics are not necessary for remote radiotracking. Realistic constant probability regions for location estimates from intersecting radioazimuths were proposed. These regions were based on the joint probability density functions for independent radioazimuths having normal and double exponential distributions. Two new methods to be used in the analysis of remote radiotracking were proposed: - Error propagation techniques to obtain the variance-covariance matrices around location estimates from two intersecting radioazimuths. - The least squares adjustment procedure to provide best estimates of location (and associated variance covariance matrices) from the intersection of more than two radioazimuths. A method to obtain error ellipses from variance-covariance matrices was also described. The results of accuracy and precision tests on a remote radiotracking system were presented. Unmodified data from the radiotracking system being tested was unsuitable for further analysis. Samples of the errors in radioazimuths had standard deviation between 9° and 36° and included large numbers of outliers. Methods to reduce the proportion of outliers (and hence the standard deviation) were sought. It was shown that censorship using audible characteristics of the incoming signal could provide an effective means to reduce the frequency of outliers in the data. Further improvement in the quality of data was obtained by restricting analysis to data from experienced observers. Data from experienced observers was used to test the influence of a number of factors on radioazimuth precision and accuracy. The factors were, whether the transmitter type was moving or stationary, the existence of a line of sight between receiving and transmitting sites and the alignment technique used during remote radiotracking. Censored radioazimuths from moving transmitters were relatively precise. The value of the standard deviation of the error in azimuths was 1-2°, which is an acceptable level of precision for real radiotracking. Samples of uncensored radioazimuths where there was no line of sight between receiving and transmitting sites included a high frequency of outliers. Censorship removed most of these outliers. Where there was a line of sight censorship of radioazimuths to moving transmitters removed most outliers whereas censorship of radioazimuths to stationary transmitters did not. The low quality of stationary, line of sight transmitters was attributed to multipath fading. The radiotracking system being tested was aligned using a single stationary reference transmitter. It was shown that this alignment technique caused a sizable bias in the radioazimuths which could not be compensated for during real radiotracking. Modifications to the field procedure for real radiotracking were proposed. It was concluded that the location data for radiotagged kakapo collected using the remote radiotracking system on Stewart Island were not sufficiently reliable to provide the basis for a biological investigation of kakapo. Refined field techniques could have provided more reliable data but the levels of error in location estimates would still have precluded meaningful analyses of phenomena such as habitat use, home range, and territorial interactions. The results of this study and a critical appraisal of the results of other studies lead to the conclusion that remote radiotracking is not a suitable tool to investigate the biology of free living wild animals. There are intractable problems inherent in remote radiotracking: systematic variation in the magnitudes of both linear errors and linear resolution; non-systematic biases in location estimates; and the occurrence of spurious radiobearings. These problems undermine the credibility of the technique.