We introduce a new Bayesian model for detecting and estimating changepoints in spatially correlated functional time series to study the impact of the Mt. Pinatubo eruption on June 15th, 1991 on global temperatures. We model the exact quadratic form of the \(\ell ^2\) -norm of the functional cumulative sum (CUSUM) statistic and leverage a spike-and-slab approach for automatically detecting changepoints while also measuring their magnitudes. Our approach, Spatial Quadratic Regression (SQuaRe), improves over existing approaches by enabling simultaneous detection and estimation while also allowing for spatially correlated changepoints. Extensive simulations demonstrate that our approach more precisely estimates change magnitudes than existing methods, while having comparable detection performance to frequentist approaches. We apply our method to temperature profiles in the upper troposphere and stratosphere to track the impact of the Mt. Pinatubo eruption on global temperatures. The spatial changepoint pattern shows an early impact of the eruption on low-latitude regions, and then the impact gradually moved to high latitudes where intrinsic variation becomes high, whereas there is little impact on the southern hemisphere.
