Skip to main content

Learning from long time series of harvest and population data - Swedish insights for European goose management

Cite this dataset

Liljebäck, Niklas (2021). Learning from long time series of harvest and population data - Swedish insights for European goose management [Dataset]. Dryad.


Goose management in Europe is faced by multiple challenges, as some species are declining and in need of conservation actions, while other populations have become very abundant, resulting in calls for increased harvest.

Sweden has long-term series of harvest data and counts of breeding and autumn-staging geese. We used national data (indices) for greylag goose, bean goose, and Canada goose to study shifts in temporal trends and correlative patterns, and to infer possible causal links between harvest and population trends. Our study provides an opportunity to guide management given the data collected within the present monitoring, as well as to suggest improvements for future data collection.

The populations of greylag and Canada geese increased in Sweden 1979–2018, but this long-term trend included a recent decrease in the latter species. Bean goose breeding index decreased, whilst staging numbers and harvest varied with no clear long-term trend. For Canada goose, our analysis suggests that harvest may affect population growth negatively. For bean goose and greylag goose we could not detect any effect of harvest on autumn counts the following year.

We find that the present data and analysis of coherence may suffice as basis for decisions for the current management situation in Sweden with its rather unspecific goals for greylag (very abundant ) and Canada goose (invasive species) populations. However, for management of bean geese, with international concerns of over harvest, data lack crucial information. For future management challenges, with more explicit goals, for all goose species we advocate information that is more precise. Data such as hunting effort, age-structure of goose populations, and mark-recapture data to estimate survival and population size, is needed to feed predictive population models guiding future Swedish and European goose management.


We based this study on data from three independent long-term monitoring programs in Sweden, providing annual data of: 1) breeding season abundance (1998-2017), 2) autumn staging counts (1978-2017) and 3) national harvest estimates (1978-2017). These datasets are here used as indices for breeding and autumn staging population development and changes in harvest levels respectively, and represent the available nation-wide monitoring of goose populations in Sweden.

Usage notes

Temporal trends (finite growth rates) based on breeding season counts along the fixed routes, were analysed using TRIM (Trends & Indices for Monitoring data, v.3.53, Pannekoek and van Strien, taking into account that not all routes were done every year. The statistical model in TRIM builds on Poisson log-linear regression, estimating site and time (year) effects on species abundance (counts) as well as an overall linear trend (log-scale). The basic TRIM model is: expected count = year + site, where both year and site are fixed effects. Effects are estimated using maximum likelihood and generalized estimating equations, the latter to handle potential overdispersion and serial (auto) correlation. For autumn counts and harvest estimates, the finite rate of increase was calculated according to Caughley & Sinclair (1994).

To identify possible changes in trends, we performed a breakpoint analysis using the package strucchange for all three species and all three indices (Kleiber et al. 2002). Breakpoint analysis is based on piecewise linear models and it identifies the time of significant shifts in trends (Zeileis et al. 2002). Coherence in timing of changes (breaks) between times series as well as differences in slopes of regression lines before and after breakpoints were used as basis to discuss plausible mechanisms and causality behind shifts in trends.

To analyse the effect of harvest on the population growth rate, we need to relate harvest level to population size. By assuming that our indices provide relative numbers over time we calculated an annual (1977–2018) ‘relative harvest rate’ by dividing the harvest estimate (hunting year t) by the autumn count (year t). In a second step, relative harvest rate (year t) was related to the exponential rate of increase (Caughley and Sinclair 1994, Steidl et al. 1997) based on the autumn count data from year t to year t+1, by using linear regression. If harvest affect the population growth rate, we expect a negative relationship between the relative harvest rate (year t) and the exponential rate of increase (year t+1).  All tests were performed using R 3.3.3. (R core team 2013).