Error measures for template-fit geolocation based on light

Template-fit geolocation offers an opportunity to estimate error in each individual day's position results. The estimate requires determining an empirical calibration constant for latitude and another for longitude based on a trial against realistic noise in irradiance field data, which typically display strong point-to-point correlations. Such a calibration is reported here, making the error estimate available for use. The calibration is based on a 458-day set of blue-light irradiance data taken by an archival tag (Lotek LTD750) at a known location on land (48.57°N, 122.94°W). For the data studied, median values of the estimated daily standard deviations were 0.42° in longitude, 0.95° in latitude. The error estimates include the effects of weather and other data noise, latitude, and season, so (1) they vary widely from day to day and with season, and (2) they can be cautiously expected to remain valid in situations other than the particular one tested. A change in the character of the data noise is the only factor expected to have a significant effect on the calibration; this work should be repeated to sample the expected range of data noise types.