Adaptive harvest management for the Svalbard population of pink-footed geese: cooperator report

Abstract

This document describes progress to date on the development of a harvest‐management strategy for maintaining pink‐footed goose abundance near their target level by providing for sustainable harvests in Norway and Denmark. Many goose populations in western Europe have increased dramatically in recent decades. The Svalbard population of pink‐footed geese (Anser brachyrhynchus) is a good example, increasing from about 10 thousand individuals in the early 1960’s to roughly 80 thousand today. Although these geese are a highly valued resource, the growing numbers of geese are causing agricultural conflicts in wintering and staging areas. The African‐Eurasian Waterbird Agreement (AEWA; http://www.unep‐aewa.org/) calls for means to manage populations which cause conflicts with certain human economic activities. We compiled relevant demographic and weather data and specified an annual‐cycle model for pink-footed geese that reconciles the different dates of monitoring activities and the timing of harvest-management decisions. We then developed dynamic models for survival and reproductive processes and parameterized them using available data. By combining varying hypotheses about survival and reproduction, we developed a suite of nine models that represent a wide range of possibilities concerning the extent to which demographic rates are density dependent or independent, and the extent to which spring temperatures are important. These nine models varied significantly in their predictions of the harvest required to stabilize current population size, ranging from a low of about 500 to a high of about 17 thousand. For comparison, the harvest in Norway and Denmark was about 11 thousand in 2011 and the population increased from 70 to 80 thousand. We relied on the passive form of adaptive management in formulating a harvest strategy. In passive adaptive management, alternative population models and their associated weights of evidence are explicitly considered in the development of an optimal harvest strategy. Unlike active adaptive management, however, there is no explicit consideration of how harvest management actions could reduce uncertainty as to the most appropriate model of population dynamics. In optimizing a harvest strategy, we assumed equal probabilities for all nine models and assumed relatively course control over harvest. We used a management objective that seeks to maximize sustainable harvest, but avoids harvest decisions that are expected to result in a subsequent population size different than the population goal of 60 thousand. Optimal harvest strategies were calculated using stochastic dynamic programming, and Monte Carlo simulations were used to investigate expected strategy performance. The optimal passive adaptive‐management strategy is expected to maintain mean population size near 60 thousand, regardless of the most appropriate model. However, mean harvest rates and harvests varied substantially depending on the most appropriate model of population dynamics. With an average number of days above freezing in May in Svalbard, optimal harvest rates (i.e., the proportion of the population to be harvested in autumn) increase rapidly once there are more than about 50 thousand birds in the population. Generally, optimal harvests were on the order of 10 – 20 thousand for population sizes > 60 thousand, and 0 – 5 thousand for population sizes < 60 thousand. For the observations of young of 15.4 thousand and adults of 54.6 thousand in autumn 2010, and 10 days above freezing in May 2011 (a relatively warm spring compared to the average of about 7), the optimal harvest rate in autumn of 2011 would have been 0.16, or a harvest of about 14 thousand. Based on the optimal strategy, hunting‐season closures would be required as the number of adults in the autumn population falls below about 52 thousand, regardless of the number of young in the population. As the number of adults and young decrease, the number of warm days in May required to keep the hunting season open increases. We also investigated the ability of the optimal strategy to stabilize the population at around 60 thousand birds, assuming varying values of the maximum harvest rate that could be implemented. Harvest strategies that contained a maximum harvest rate of 0.16 (equivalent to a harvest of about 17 thousand) were effective at stabilizing the population at 60 thousand within 4‐5 years, regardless of climate scenario. Harvest strategies with a maximum harvest rate of 0.12 (harvest ≈ 13 thousand) were also able to stabilize the population near 60 thousand, although it took more time. Harvest strategies with a maximum harvest rate of 0.08 (harvest ≈ 8 thousand) were unsuccessful at stabilizing the population at 60 thousand. Continued monitoring of the pink‐footed goose population on an annual basis is critical to an informed harvest management strategy. At a minimum, the ground census in November should be continued to determine population size and proportion of young. Continued estimates of harvest from Norway and Denmark are also necessary to help judge the credibility of the alternative population models. However, an adaptive management process that relies on periodic updating of model weights will depend on acquiring either estimates of the realized harvest rate of adults or the age composition of the harvest. We also recommend that a census conducted during spring migration be operationalized, and that estimates of survival based on mark‐recapture data be updated. Finally, the International Working Group has expressed a desire to adopt a three‐year cycle of decision making related to the regulation of pink‐footed goose harvests. The idea is that once a target harvest level is adopted, it would remain in place for three years, after which time population status would be assessed and a potentially new management action chosen. We have developed a preliminary framework to implement a three‐year cycle using stochastic dynamic programming, and we hope to have it fully operational later this year . We note, however, that application of this 3‐year framework will still require annual resource monitoring and assessments to facilitate learning, and to allow managers the opportunity to respond to any unforeseen change in resource conditions.