Long-term demographic fluctuations in an orchid species driven by weather: implications for conservation planning.
Abstract
Management decisions are increasingly based on matrix models intended to predict the long-term fate of endangered species. However, certain elements of these models, such as life-state transition probabilities (vital rates), are difficult to parameterize and their values may vary depending on external conditions such as weather. Details of how weather might influence population performance are rare, yet necessary to assess the effects of global climate change on a species' distribution. Based on a 26-year data set of a population of Himantoglossum hircinum in a nature reserve in Germany, variations of life-history traits and vital rates were studied. Matrix analysis was used to identify the most important life-state transitions for population growth. Multiple linear regression was used to quantify the response of population traits and vital rates to changing weather conditions. Population size increased exponentially and density effects could not be observed. Flowering plants and large plants had the highest and second highest reproductive value, respectively. The population's finite rate of increase fluctuated strongly among years; life-history traits varied strongly and were interlinked, thereby violating the assumptions of matrix modelling in a population viability analysis. Some vital rates and the population growth rate showed a trend over the total period. A certain and sometimes large amount of that variability could be attributed to variability of weather conditions, with warmer winter conditions favouring population performance. Prediction of population size was fairly accurate within a time frame of 10 years, but size class structure was not. Synthesis and applications. Matrix modelling proved to be unreliable for predicting long-term population dynamics, despite the long-term data set used for matrix construction. This can be explained by weather-dependent variability of vital rates driving population dynamics. A minimum study period of 4 years is necessary to produce relevant information for model development. Our study emphasizes the need for critical evaluation of management decisions based only on single short-term studies and for studies covering longer time intervals than 2-3 years.