Keywords: long-term ecosystem research, time-series analysis, data integration
Motivation:Holistic time-series studies of arctic marine ecosystems are rare. This is not surprising since polar regions are largely only accessible by means of expensive modern infrastructure and instrumentation. In 1999, the Alfred Wegener Institute, established the LTER (Long-Term Ecological Research) observatory HAUSGARTEN, crossing the Fram Strait between Greenland and Svalbard to detect and track the impact of large-scale environmental changes on the marine ecosystem in the transition zone between the northern North Atlantic and the central Arctic Ocean. Multidisciplinary investigations cover major parts of the open-ocean ecosystem. They are carried out at more than 20 permanent sampling sites in water depths ranging between 250 and 5500 m (Soltwedel et al., 2016). From the outset, repeated sampling in the water column and at the deep seafloor during regular expeditions is complemented by continuous year-round sampling and sensing. Measured parameters cover a very broad trans-compartmental spectrum. It ranges from remote sensing data for sea-ice coverage and primary production, variations in the hydrography, SST variations, transport processes of organic matter to the deep sea and the effects of those changing processes on marine communities.
Aim:The vast data set from this ecological time-series comprises a temporal component, generated over twenty years, and a spatial component for each year investigated. By now, we have not yet sufficiently succeeded in statistically combining the large number of investigated parameters in a way it would allow to derive direct or indirect dependencies between those parameters. Based on the design of the monitoring study the project will be divided into three main aspects, each of them aiming to find the best possible statistical method to assess if respective gradients have an effect only on individual parameters or directly on a whole set of parameters.
Objectives:(1) Statistical analysis of the time-series component using Dynamic Factor Models (Dickhaus and Sirotko-Sibirskaya, 2019) in combination with statistical change point detectionto identify time periods of biological or climate regime shifts, (2) Application of Copula Models to investigate special dependencies and structures (Neumann et al., 2019 and the references therein), (3) Amalgamation of key findings of the two previous tasks (Kröncke et al., 2019) and application on high-resolution remote sensing spatial data to interpolate parameters sampled with lower spatial resolution and frequency.
Competences:The candidate should have a solid background in mathematics with programming skills and statistics with emphasis on time-series analysis and pattern detection methods. A true motivation to work with different disciplines of climate and ecological sciences is desired.
Dickhaus, T., Sirotko-Sibirskaya, N. (2019). Simultaneous statistical inference in dynamic factor models: Chi-square approximation and model-based bootstrap. Computational Statistics and Data Analysis 129: 30-46.
Kröncke et al. (2019). Comparison of biological and ecological long-term trends related to northern hemisphere climate in different marine ecosystems.Nature Conservation 34: 311-341. https://doi.org/10.3897/natureconservation.34.30209
Neumann, A, Bodnar, T, Pfeifer, D, Dickhaus, T. (2019). Multivariate multiple test procedures based on nonparametric copula estimation. Biometrical Journal; 61: 40– 61. https://doi.org/10.1002/bimj.201700205
Soltwedel, T., et al. (2016). Natural variability or anthropogenically-induced variation? Insights from 15 years of multidisciplinary observations at the arctic marine LTER site HAUSGARTEN. Ecological Indicators 65 (Supplement C): 89-102.