ALBERT

Description: Marine particulate organic carbon-13 stable isotope ratios (δ13C-POC) provide additional constraints and insights into the cycling of carbon from dissolved pools to marine ecosystems including anthropogenic contributions. For such purposes, a robust spatio-temporal coverage of δ13C-POC observations is essential. In this data product, we collected and merged two large data compilations (Close and Henderson, 2020; St John Glew et al., 2021) into our previous version (Verwega et al., 2021) to provide the largest available marine δ13C-POC data set. Additionally, we have incorporated more meta information including if the samples were acidified before measuring the isotope ratio. The data set consists of 6952 data points covering the global ocean from year 1966 to 2019. We provide the data in the following two formats for best application on specific research purposes: (1) A spreadsheet file including all collected individual data and meta-information; (2) Network Common Data Form (NetCDF) files that only include acidified samples (6633 total data points) interpolated onto a global ocean grid (1°x1° horizontal resolution, 33 vertical levels based on World Ocean Atlas 2009) for each month individually and all months combined, with each file covering the temporal range from year 1966 to 2019.

Description: My work developed a kernel density estimator that well resolves typical structures of probability densities, which was demonstrated on a newly compiled marine data set of organic carbon-13 isotope ratios (δ13CPOC). All work was conducted within the emerging field of marine data science. I identified classical data science, a general understanding of ocean science, communication skills, and confidence as requirements for marine data scientists. In the beginning of my work, the existing δ13CPOC data consisted of about 500 data points in the global ocean. I expanded the existing data set to 4732 data points in a first version, and additionally to 6952 in a second. Both are published at PANGAEA along with meta information such as measurement location, time, and method, and interpolations. I have published a description of the temporal and geographic distribution of the first version at Earth System Science Data. I designed the development of the kernel density estimator algorithm on the existing concept of computing it as a solution of the diffusion equation. My algorithm uses finite differences in space and equidistant time steps with an implicit Euler method, and approximates the optimal smoothing parameter by two pilot steps. Compared to other well-known kernel density estimators, my algorithm produces reliable approximations of multimodal and boundary-close distributions on artificial and real marine data and is robust to noise. My implementation is published as a Python package on Zenodo, its description is submitted to Geoscientific Model Development. I was able to show that my kernel density estimator reliably evalu- ates ocean data and thus lays a foundation for calibrating Earth system models. At the same time, I was able to contribute to the definition and establishment of the field of Marine Data Science.

Description: Probability density functions (PDFs) provide information about the probability of a random variable taking on a specific value. In geoscience, data distributions are often expressed by a parametric estimation of their PDF, such as, for example, a Gaussian distribution. At present there is growing attention towards the analysis of non-parametric estimation of PDFs, where no prior assumptions about the type of PDF are required. A common tool for such non-parametric estimation is a kernel density estimator (KDE). Existing KDEs are valuable but problematic because of the difficulty of objectively specifying optimal bandwidths for the individual kernels. In this study, we designed and developed a new implementation of a diffusion-based KDE as an open source Python tool to make diffusion-based KDE accessible for general use. Our new diffusion-based KDE provides (1) consistency at the boundaries, (2) better resolution of multimodal data, and (3) a family of KDEs with different smoothing intensities. We demonstrate our tool on artificial data with multiple and boundary-close modes and on real marine biogeochemical data, and compare our results against other popular KDE methods. We also provide an example for how our approach can be efficiently utilized for the derivation of plankton size spectra in ecological research. Our estimator is able to detect relevant multiple modes and it resolves modes that are located closely to a boundary of the observed data interval. Furthermore, our approach produces a smooth graph that is robust to noise and outliers. The convergence rate is comparable to that of the Gaussian estimator, but with a generally smaller error. This is most notable for small data sets with up to around 5000 data points. We discuss the general applicability and advantages of such KDEs for data–model comparison in geoscience.