Chopin: An Open Source R-language Tool to Support Spatial Analysis on Parallelizable Infrastructure

TL;DR

The paper introduces 'chopin', an open-source R package that simplifies parallel geospatial data analysis, enabling scalable computation across various hardware setups without requiring domain-specific high-performance computing knowledge.

Contribution

It presents a flexible, easy-to-use R package that supports parallel processing for geocomputation, reducing technical barriers and improving scalability for large spatial datasets.

Findings

Significant reduction in execution time for environmental exposure assessments

Supports interoperability with multiple R spatial packages

Effective on diverse computing hardware from laptops to HPC clusters

Abstract

An increasing volume of studies utilize geocomputation methods in large spatial data. There is a bottleneck in scalable computation for general scientific use as the existing solutions require high-performance computing domain knowledge and are tailored for specific use cases. This study presents an R package `chopin` to reduce the technical burden for parallelization in geocomputation. Supporting popular spatial analysis packages in R, `chopin` leverages parallel computing by partitioning data that are involved in a computation task. The partitioning is implemented at regular grids, data hierarchies, and multiple file inputs with flexible input types for interoperability between different packages and efficiency. This approach makes the geospatial covariate calculation to the scale of the available processing power in a wide range of computing assets from laptop computers to high-performance computing infrastructure. Testing use cases in environmental exposure assessment demonstrated that the package reduced the execution time by order of processing units used. The work is expected to provide broader research communities using geospatial data with an efficient tool to process large scale data.