SHAVER: Web-Optimized Lossy Compression for High-Resolution Unstructured GIS Datasets

Abstract

As Geographic Information Systems (GISs) increasingly migrate to the web, efficiently delivering high-resolution datasets for interactive visualization pre-sents a substantial challenge. In this thesis, we propose a shaver, a novel compression system tailored for Web-Based GIS platforms, combining unstructured grid simplification with state-of-the-art lossy floating-point compression. Building on an angle bounded simplification method to preserve grid quality, we compress scalar fields using the zfp compressor and implement client-side decompression for web deployment. Evaluations on oceanographic simulation data from the Oceanbox WebGIS demonstrate compression ratios up to 12x and a 6x reduction in end-to-end response time, with minimal impact on visual fidelity. We conclude that our compression system significantly improves the efficiency and responsiveness of high-resolution WebGIS visualizations, enabling interactive use from remote locations and alleviating server resources and bandwidth in the face of a global user scale-up. Offering high tunability, shaver can also be applied to other GISs with similar unstructured archives. After extensive fine-tuning and more seamless system integration, Oceanbox is ready to deploy the compressor in their WebGIS, currently serving major Norwegian aquaculture customers.