Logo image
Uncovering spatial process heterogeneity from graph-based deep spatial regression
Journal article   Peer reviewed

Uncovering spatial process heterogeneity from graph-based deep spatial regression

Di Zhu, Sheng Wang and Peng Luo
ISPRS journal of photogrammetry and remote sensing, Vol.232, pp.509-523
02/01/2026
DOI: 10.1016/j.isprsjprs.2025.12.008
url
https://doi.org/10.1016/j.isprsjprs.2025.12.008View
Published (Version of record) Open Access

Abstract

One can never specify the true statistical form of a complex spatial process. Model misspecification, as linearity and additivity, will lead to fundamentally flawed interpretations in the estimated coefficients, particularly in empirical geographic studies involving a large number of observations and complex generation processes. Motivated to learn representative patterns of spatial process heterogeneity, we propose a deep explainable spatial regression (XSR) framework based on graph convolutional neural networks (GCN), which bypasses the conventional parametric statistical assumptions in spatial regression modeling and generate deep spatially varying coefficients that depict the heterogeneity structure of spatial processes. introduce an analytical framework to (1) perform deep spatial regression modeling in multivariate crosssectional scenarios, (2) reconstruct spatial heterogeneity patterns from the learned deep coefficients, and explain the effectiveness of heterogeneity through a simple diagnostic test. Experiments on Greater Boston house prices modeling demonstrate better fitting performance over spatial regression baselines. The spatial patterns of deep local coefficients consistently exhibit stronger explanatory power than those derived geographically weighted regression, indicating a better representation of the true spatial process heterogeneity uncovered by graph-based deep spatial regression.
Geography, Physical Geology Geosciences, Multidisciplinary Imaging Science & Photographic Technology Physical Geography Physical Sciences Remote Sensing Science & Technology Technology

Details

Metrics

Logo image