Evaluating the Efficiency of Various Styles of Distributed Geoprocessing Chains for Visualising 3D Context Aware Wildfire Scenes from Streaming Data

Abstract

Big data refers to the ever-increasing volumes of data being generated continuously by a large variety of sensors, such as smartphones and satellites. In this chapter, we explore solutions for challenges related to the velocity characteristic of big geospatial data. The research was inspired by the Advanced Fire Information System (AFIS), which provides near real-time fire detection from earth observation satellites. Users in Southern and East Africa, South America and Europe are automatically notified of active fire detections, with the hope that timeous information may help to mitigate the impact of wildfires. This chapter evaluates the efficiency of various styles of geoprocessing chains for generating enriched notifications containing 3D fire visualisations from an intermittent stream of active fire detection data generated by remote sensing satellites. The visualisation should be ready for viewing as soon as the user investigates the notification; this implies the requirement for rapid geoprocessing, since there may be hundreds of fire detections disseminated to hundreds of parties at any satellite overpass. Geoprocessing chains were implemented in Python using open-source libraries and frameworks. This study investigated efficiencies across three dimensions: (1) software libraries, (2) tightly-coupled/serial versus loosely-coupled/distributed geoprocessing chain implementations, and (3) standardised geoprocessing web service (Web Processing Service) implementations versus bespoke software solutions. Results show that bespoke software, using specific geoprocessing libraries, implemented on a loosely-coupled messaging architecture significantly outperforms other combinations.