Today's multitude of remote sensors and environmental models provides an invaluable data source to understand and predict spatiotemporal phenomena. These data are inherently multidimensional and most often come in a discrete computer representation as arrays. To efficiently integrate array information from space, time, and various thematic dimensions, there is still a strong demand for easy-to-use data management tools and analysis methods.

In this course we focus on managing large geoscientific data with the array-based data management and analytics system SciDB and on statistical methods for extracting multidimensional information from satellite imagery time series. Specifically, this course will:

1. Introduce the array model and discuss how it is implemented in open source data management and analytical software (SciDB, R, rasdaman, GNU Octave, Python), how arrays facilitate representing spatiotemporal phenomena, and where typical challenges arise.


2. Introduce array-based data analysis using examples in dimension reduction and spatiotemporal change detection.

3. Provide hands-on experiments with SciDB and R to conduct complex geoscientific data analysis. The topics are:
(i) A novel change detection method that extends statistical testing for structural changes in time series to account for the spatial autocorrelation.
(ii) Exploring spatiotemporal and spectral patterns in satellite image time series using principal component analysis / empirical orthogonal function analysis.