Abstract
Population monitoring programmes plays a central role in biodiversity conservation. They provide estimates of population sizes and trends, both of which are crucial for identifying populations at risk of extinction, understanding the causes of their decline, and subsequently testing and validating ways to halt this phenomenon. However, the effectiveness of monitoring programmes is controversial as they suffer from recurrent methodological problems that may prevent them from providing precise and unbiased estimates of population sizes and trends. This seems to be particularly the case for plant population monitoring, especially as accounting for observation errors, one of the main sources of error in population monitoring, seems to be rare. Furthermore, reducing the uncertainty of estimates and ensuring that they are unbiased reduces the risk of making decision errors, which may result in the implementation of ineffective or even harmful conservation measures. The overall aim of my PhD was to improve plant population monitoring methods so that they can provide unbiased and more accurate estimates of population sizes and trends. To this end, I addressed the two main methodological problems of population monitoring: sampling designs and observer error, and specifically imperfect detection. My work was conducted along three main axes.The first axis consisted of a synthesis of the literature on plant population monitoring methods. I structured this synthesis according to the two main challenges of population monitoring, i.e. sampling and accounting for observation errors, and the two scales at which monitoring can be carried out, i.e. at the level of the population as a whole or at the level of the individuals that constitute it. Thanks to this synthesis work, I was able to identify priorities for methodological research to improve monitoring, and to highlight the differences between gaps in methodological development and the lack of implementation of existing methods.In the second axis, I used computer simulations to compare the precision of population size estimates obtained with three different sampling methods for populations whose individuals were more or less spatially aggregated. Spatial aggregation is a common feature of plant populations, and this makes population size estimates less accurate compared to spatially homogeneous populations. The simulations allowed me to highlight substantial differences in precision between the sampling methods, and thus to propose a method to improve the precision of estimates by adapting the sampling design to the level of aggregation of the studied population.The goal of the third axis was to verify that the detection of individuals is imperfect when counting unmarked individuals, which had not yet been formally established. To this end, I conducted an experiment in which a large number of observers counted individuals of plants of different species, in different habitats and with three counting methods. The results showed that the detection of individuals was indeed imperfect, and allowed to describe the effect of the main ecological and observational variables on the detection probability of individuals. In addition, it allowed to give recommendations to reduce the variability of counts by adapting the observation method to the study conditions.