Predicting population fluctuations with artificial neural networks
Lindström, J., Kokko, H., Ranta, E. & Lindén, H. 1998: Predicting population fluctuations with artificial neural networks. - Wildl. Biol. 4: 47-53.
Successful predictions of population fluctuations are valuable in game management, as population estimates are instrumental in increasing the time available for management decisions. However, finding a population model which produces predictions accurate enough to be used for management purposes is often precluded due to scarcity and noisiness of population data. Using two long-term population data sets, 1964-1984 data on Finnish grouse (Tetrao urogallus, T. tetrix and Bonasa bonasia) and 1914-1950 data on coloured fox Vulpes fulva from Canada, we demonstrate the use and power of an artificial neural network in predicting population fluctuations. The performance of an artificial neural network model is compared to two benchmark forecasts: time series mean and the previous data value. Unfortunate as it is, in practise management decisions often have to be made with limited data. Therefore, a notable advantage of neural network modelling is the forecast accuracy even in cases when the time series available are short and noisy, and the processes underlying population fluctuations are not fully understood.
Key words: artificial neural network, forecasting, population dynamics, wildlife management
Jan Lindström*, Hanna Kokko & Esa Ranta, Department of Ecology and Systematics, Division of Population Biology, P.O. Box 17 (Arkadiankatu 7), FIN–00014 University of Helsinki, Finland
Harto Lindén, Finnish Game and Fisheries Research Institute, P.O. Box 202, FIN–00151 Helsinki, Finland
*Present address: University of Cambridge, Department of Zoology, Downing Street, Cambridge CB2 3EJ, UK, e-mail: J.LINDSTROM@zoo.cam.ac.uk
Received 24 March 1997, accepted 12 November 1997
Associate Editor: Nigel G. Yoccoz