Abstract

This paper presents a novel clustering framework for spatial interval-valued time series (Spatial ITS). To this aim, we introduce an adaptive distance metric that explicitly incorporates both the inherent autocorrelation structure and the deterministic trend parameters of the interval time series. Furthermore, the clustering mechanism is enhanced by integrating the spatial dependence structure of the units, thereby facilitating a comprehensive spatio-temporal modeling approach. To demonstrate the empirical validity of the proposed framework, we provide applications to real-world georeferenced interval-valued time series, such as CO2 emissions across Italian provinces and suicide rates across European regions.