I work with statistical analysis and modelling of complex marine systems to help improve management and conservation of marine resources. In marine environments, most systems can not be observed directly; data is only available with substantial measurement error. This is for instance the case for stock assessment models and individual animal movement models. In the presence of measurement error, state-space modelling is a valuable framework which my research revolves around.
State-space models separate a time series of interest from the way it is observed. While most of my research is in applied statistics, I am also interested in computational aspects of state-space models. Besides my own research, I enjoy doing statistical consulting in various areas.
Fisheries stock assessment models are an integral part of determining fisheries catch quotas. The aim of the model is to determine the number of fish in the sea through the commercial landings and scientific surveys; a natural application for state-space models.
Currently, my research is focused on extending the models currently used. I am investigating different ways of modelling the observations. Further, I am working on developing models for multiple fish stocks to provide an operational middle way between single stock models and complex ecosystem models.
Examples of assessment model papers I have been part of are:
Like stock assessment models, individual marine animal movement models are perfect applications of state-space models. When animals are tracked under water, measurement error is inevitable.
My research is focused on using the Laplace approximation for maximum likelihood estimation in movement models. Using the Laplace approximation allows more general models than Kalman filters and faster inference than simulation-based methods.
Further, I am interested in developing methods to include location dependent covariates in the movement for maximum likelihood inference on data with measurement errors.
Examples of movement model papers I have been part of are:
Recently, I have been heavily involved in analysing data from fish otoliths. Otoliths are small stones in the ear that grow as the fish gets older. The shape and chemical composition depends on both the species, the stock, and the life history of the fish.
Trajectories of the chemical composition in the otolith — from the centre to the edge — can be used to estimate the age of the fish, and can reveal migration patterns of the fish. Like movement and assessment models, models for chemical composition trajectories must account for substantial measurement errors. A perfect application for state-space models. The otolith shape is used to separate fish stocks to reveal how they mix with each other.
Examples of otolith papers I have been part of are:
While most of my research is in applied statistics, I am also interested in general aspects of state-space models such as estimation procedures, computational issues, estimation problems and model validation. In most state-space models, calculating the exact likelihood of an observation is impossible. It requires integrating over all unobserved variables, which results in both theoretical and practical issues. I am mainly interested in the latter.
Examples of more general papers I have been part of are:
Besides my own research, I do statistical consulting. I can help with planning of experiments and analysis of data — both from experiments and observational studies. Further, I can help with visualization of data and results, and making repeated analysis operational.
Examples of papers where I have been part of analysing experiments are: