Multi-species duck harvesting using dynamic programming and multi-criteria decision analysis

摘要

1.Multiple species are often exposed to a common hunting season,but harvest and population objectives may not be fully achieved if harvest potential varies among species and/or species abundances are not correlated through time.Our goal was to develop an approach for setting a common hunting season that would recognize heterogeneity in species productivity and would select annual hunting seasons conditioned on the status of individual species.2.We first used stochastic dynamic programming to generate optimal,state-dependent harvest strategies for 18 candidate regulatory scenarios.We simulated the performance of these strategies,and then used multi-criteria decision analysis to identify preferred regulatory scenarios for duck hunting seasons in the Atlantic Flyway of the U.S.3.Generally,estimates of annual population size were not correlated among species.Mallards had the highest estimated intrinsic rate of growth,green-winged teal,wood ducks,and ring-necked ducks had intermediate values,and goldeneyes were the least productive.Estimated carrying capacity was highest for mallards and lowest for green-winged teal.4.Managers had greatest interest in maximizing season length (33%) and aggregate duck abundance (28%),and less interest in maximizing aggregate harvest (19%) and the number of years between a change in hunting season regulations (19%).Several regulatory scenarios provided acceptable trade-offs among these objectives.5.Synthesis and applications.Separate hunting seasons for various species of game may be untenable,either due to the added cost and regulatory complexity,or because selective harvesting of stocks may be difficult due to problems in species identification.Rather than averaging species-specific productivities,or basing hunting seasons on the least (or most) productive species,we describe an approach in which productivity and annual population status are considered explicitly for each species.By combining stochastic dynamic programming with multi-cri