Functional geographically weighted regression with conformal uncertainty for seismic data

This work presents a functional geographically weighted regression model for seismic intensity measures, capturing spatially varying relationships and multiple sources of non-stationarity, including site and event specific effects. To validate predictions, a functional conformal inference framework is integrated, providing finite-sample, distribution free predictive bands. Applied to seismic curves, the model captures complex spatial patterns and delivers reliable uncertainty quantification critical for seismic hazard assessment.