How do forage availability and climate control sheep reproductive performance?. An analysis based on artificial neural networks and remotely sensed data

Abstract

Environmental variability affects life history and fitness of both animal and plant species. For herbivores in particular, climate can have strong direct and indirect effects on demography, which tend to exacerbate in arid and semiarid environments with highly seasonal weather. We studied the joint effect of forage conditions, plant phenology, and climate on the reproductive performance of a "model" population: domestic sheep in the Patagonian steppe of Argentina. In this region sheep behave as semi-natural populations and relatively good population records are available. Using linear models and artificial neural networks trained by second order back-propagation methods, we demonstrated that reproductive performance, characterized by the marking rate (number of lambs per ewe), was associated to the timing of growing season start and to the primary production (as estimated from remotely sensed data) at mating. An ANN model including these variables explained 73% of the variability of normalized marking rate, and predicted observed marking rates with an accuracy of 63%. Our results highlight the importance of forage availability as opposed to weather regulating the reproductive performance of sheep at Patagonia, suggesting that bottom-up controls are of dominant importance for these populations. Using artificial neural networks, satellite imagery, and historical productive and climatic records, we disentangled the controls of sheep reproductive performance in a region characterized by weak but consistent relationships between environment and sheep dynamics. Our work aims to the development of quantitative tools for the management and planning of sheep herd structure from ranches to regions, especially for those situations in which classical methods (i.e. linear methods) fail.