Quantifying fitness in terms of survival and reproduction has been an obstacle in many evo-eco studies. As survival is binary (survived/died) and reproduction ordinal (number of offspring), tracking changes in individual fitness can be especially challenging, particularly in long lived, slow reproducing species. Thus far, researchers have been limited to using rough proxies of fitness, such as body condition and blood cell count. Here, we coalesce veterinary physiological data and time to event models to construct informed, precise continuous measures of survival (probability of survival) and reproduction (probability of reproduction) in African buffalo (Syncerus caffer). We measured common indicators of animal health (blood chemistry, acute phase proteins, hematology, body condition, age) every two months for two years in 75 African buffalo contained in a 900-hectare enclosure in Kruger National Park (KNP), South Africa. Animal health, mortality and reproduction data was used to construct Cox proportional hazard models; models were then used to calculate survival and reproduction probabilities for each animal at each capture period.
Results/Conclusions
Backwards model selection revealed that combining traditional proxies of fitness (body condition and age) with blood chemistry parameters yielded models with the lowest AIC. Models with the lowest AIC were selected as the most suitable models to construct survival and reproduction probabilities. To test how well model outputs estimated fitness, we first used our models to calculate survival and reproduction probabilities for a herd of 200 KNP free ranging African buffalo captured every six months for four years, a study which collected identical measures of animal health, mortality, and reproduction and then conducted a sensitivity and specificity analysis for predicted probabilities as classifiers of mortality or reproduction. Importantly, we found our model outputs to be more sensitive and specific than solely using traditional proxies of fitness. Therefore, survival and reproduction probabilities derived from time to event models offer a precise way to track change in individual fitness over time. Such methodology offers unparalleled utility in estimating the longitudinal effects of environmental variables on individual and population fitness. Future work involves using our metric of survival and reproduction to analyze how individual fitness changes with parasite burden (tolerance), an important issue within the field of disease ecology.