An updated version of the SZ-plugin: From space to space–time data-driven modeling in QGIS

The geospatial community usually use GIS environments to handle databases and pre-process their information, whereas complex analyses, especially data-driven ones, are performed outside GIS platforms. This interrupts the flow of information and the processing chain in a number of I/O operations that...

Full description

Saved in:
Bibliographic Details
Main Authors: Giacomo Titti, Liwei Hu, Pietro Festi, Letizia Elia, Lisa Borgatti, Luigi Lombardo
Format: Article
Language:English
Published: Elsevier 2025-08-01
Series:International Journal of Applied Earth Observations and Geoinformation
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1569843225003267
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The geospatial community usually use GIS environments to handle databases and pre-process their information, whereas complex analyses, especially data-driven ones, are performed outside GIS platforms. This interrupts the flow of information and the processing chain in a number of I/O operations that inevitably slow down the overall analytical protocols. The first version of the SZ-plugin attempted to mitigate this issue by offering a modeling solution within QGIS. However, the available models in the SZ-plugin essentially boiled down to binary classifiers, whose dimensionality was constrained to address pure spatial problems. In this updated version, we focused on two major aspects: (1) a space–time extension and (2) the inclusion of a regression option in addition to the already existing classification one. These two aspects have been introduced as part of two new models, namely, a Generalized Additive Modeling and a Multi-Layer Perceptron. In short, these would allow users to obtain susceptibility and intensity estimates in space and time. An improved graphical reporting tool has also been implemented. This makes it possible to produce relevant statistical summaries as well as cartographic outputs to be directly integrated into technical reports or scientific documents. The problem of landslide prediction is taken as a reference in Taiwan, where ten years of records are available. The example offers an overview of the new plugin capabilities, to which we added a suite of cross-validation options in space and time, automatically run at the user preference. Despite the specific example framed in the landslide context, the same plugin can be used to perform regressions or classifications for any other phenomenon associated with: digital soil mapping, wildfire and gully erosion modeling, land-use or tree species detection etc.
ISSN:1569-8432