Discover coincident data across satellites in dynamic Arctic environments: exploring scalable strategies for spatio-temporal joins on Apache Sedona

The increasing diversity of Remote Sensing (RS) platforms and products has brought unprecedented opportunities for monitoring the rapidly changing Arctic region as air and sea surface temperatures are rising. Coincident data discovery, as an important building block for the integrative use of data across multiple platforms, is of great importance in such dynamic environments. However, the unprecedented volume and variety of increasingly available RS data have become computationally challenging for geoscience researchers using traditional platforms. Although distributed computing frameworks have been extended to support geospatial data types, few consider both the spatial and temporal perspectives needed to find coincident observations. To fill this gap, we developed and compared three computing strategies based on the open-source Apache Sedona for discovering coincident RS data and evaluated these strategies through three real-world scenarios using metadata from three widely used sensing platforms: Sentinel-2, Landsat-8, and ICESat-2, which cover both multispectral imagery and LiDAR-based height observations. The experimental results showed that the strategy with a temporal slicing index significantly improved computing performance by orders of magnitude. The study also provides an in-depth performance analysis that combines detailed physical execution plans of the query strategies with experimental metrics on runtime variability, communication cost, and memory-related behaviour.