SIParCS 2022 - Mozhgan A. Farahani
pyTIER: Development of a Python Knowledge-based weather station interpolation algorithm
Converting point-based observations from irregularly placed weather stations to values on a uniform high-resolution grid is challenging, especially in areas of complex terrain, and many different methods have been developed to create these data products. Gridded precipitation and temperature products are used widely across the Earth Sciences as input data to process models, model verification datasets, and analysis datasets for features of interest (e.g., changes in temperature over time). The Topographically InformEd Regression (TIER) method, written in MATLAB®, implements a well-known, yet proprietary, knowledge-based interpolation system in an open-source repository. Providing an open-source version makes the concepts of this popular method more widely available and allows users to easily experiment with aspects of the interpolation routine to better understand the impacts of different options such as algorithmic decisions on the final interpolation product and associated uncertainties. This project develops a python version of the TIER weather station interpolation algorithm (pyTIER). We also applied Information Theory to further improve our methodological understanding of these types of algorithms. This updated TIER repository will enable users to better contribute to the TIER codebase by using a more popular, and free programming language and make TIER more efficient for large spatiotemporal domain processing. Furthermore, this project enables future developers to easily add parallelization for use on clusters and other high performance computing environments.
Mentors: Andrew Newman, Seth McGinnis, Rachel McCrary
Slides and poster