Computational Methods
Computational Methods for Quality & Utilization of Models & Data (Speaker: M. Flechsig)
Utilisation of computer models and digital data, and establishing mechanisms assuring their quality are essential for research at PIK. Concerning this matter, the group aims in close cooperation with the other Research Domains and PIK-IT at the development of concepts and the implementation of computational methods.
The following interrelated focal points are addressed
by the group. They are of significant and continuous relevance to sustain PIK’s
research process.
- Integrated data utilisation aims at integrated and flexible utilisation of available, yet considerably heterogeneous data. General and extensible data frameworks are designed and developed that provide standardised and integrated access to data resources and data transformation and processing functionality.
- Data quality analysis is approached by systematic quality investigations. The base idea is to generally systemise quality measures, adapt them to specific disciplines and provide a set of strongly linked statistical and visualisation methods.
- Analysis of model sensitivity and uncertainty to fluctuations in a model’s input space is a key technique for model evaluation and quality assurance. Methods to generate efficient deterministic and probabilistic experimentation plans and to analyse multi-run experiments are identified and implemented.
- Sustainability and assurance of model quality by software engineering addresses source code quality maintenance, testing and debugging in a transdisciplinary research environment by modern software engineering methods like code analysis, refactoring and automatic documentation for large legacy model codes.
- A cross-cutting focus is the adaptation and development of visualisation methods including exploration methods for large-volume and multivariate datasets and comparative approaches for scenario / experiment analyses.
The overall approach is to bridge the gap between
domain experts and/or decision makers and available information technologies.
Problem analysis is carried out in close co-operation with domain experts to
systematically identify requirements and match them with IT methods and
software solutions. State-of-the-art technologies, concepts and methods are
integrated / provided / adapted / enhanced for research processes. New concepts
and methods are developed together with external research partners to meet
PIK-specific requirements. The group builds on its expertise in a number of
common base technologies, including simulation, data bases / warehouses,
compute and storage grid, parallelisation, human-computer interaction, visual
and statistical data analysis, software engineering and code analysis.
