Reliable methods for identifying animal deaths in GPS- and satellite-tracking data: Review, testing, and calibration

Assessing mortality factors and survival rates is a cornerstone for the management of threatened and harvestable species. Modern tracking technologies allow unprecedented opportunities to monitor the mortality of individuals for years. However, the methods employed to separate actual deaths from transmitter failures have rarely been validated. Here, we review these previous methods and exploit an intensive satellite-tagging programme on raptorial black kites Milvus migrans to examine through tree models whether any tag-parameter indicators could reliably separate 18 confirmed deaths (recovered corpse) from 32 confirmed cases of radio-failure (tagged individual subsequently observed alive). Classification keys were validated on 21 confirmed deaths or radio-failures from an independent tagging study on two other raptor species. Review of tracking studies on mortality showed that they are exponentially increasing and using a wide variety of indicators (e.g., battery charge, stationary locations, fix frequency/quality, sudden tag disappearance, or common sense), with no preponderance of any. Few studies confirmed enough deaths or radio-failures to validate such methods. For GPS-satellite tags, we found a reliable classification key that correctly classified 100% of the deaths and radio-failures in both the calibration and validation datasets through three simple dichotomous indicators: (a) whether GPS locations were stationary or not; (b) whether background Doppler data confirmed them as stationary or not; and (c) whether GPS-locations' frequency gave indication of transmitter malfunctioning or not (Figure 3). For tags recording Doppler locations alone, we failed to find a reliable classification key and advise future researchers to use Doppler data very cautiously (e.g., by censoring any uncertain terminal events) because misclassification rates are likely to be high. Synthesis and applications. Refining the death surveillance function of satellite and GPS-tags will improve the capability of management practitioners to: (a) uncover under-reported casualties provoked by illegal or conflictual causes (e.g., poisoning, shooting, collision with windmills); and (b) calibrate demographic models which form the basis of management plans (e.g., population viability or sustainable harvest models). Improved exploitation of such remote monitoring will markedly advance the efficiency, planning, and implementation of management projects on imperilled or harvestable species affected by illegal, conflictual, or cryptic sources of mortality.