A spatio‐temporal approach to estimate patterns of climate change

Abstract

We introduce a method for decomposition of trend, cycle, and seasonal components in spatio‐temporal models and apply it to investigate the existence of climate changes in temperature series. The method incorporates critical features in the analysis of climatic problems—the importance of spatial heterogeneity, information from a large number of weather stations, and the presence of missing data. The spatial component is based on continuous projections of spatial covariance functions, allowing the modeling of complex patterns of dependence observed in climatic data. We apply this method to study climate changes in the northeast region of Brazil, characterized by a great wealth of climates and large amplitudes of temperatures. The results show the presence of a tendency for temperature increases, indicating changes in the climatic patterns in this region.

Tornado Occurrences in the United States: A Spatio-Temporal Point Process Approach

“…This structure is important as it allows estimating the impact of non-permanent shocks on temporal patterns, being especially useful in modeling climate processes as it allows estimating the aggregate effects of phenomena such as droughts, cyclical changes in ocean temperature and other periodic patterns. A detailed discussion of this decomposition applied to climate change patterns can be found in Laurini (2019).…”

“…Our model follows the approach introduced by Laurini (2019) which proposes a method for decomposition of trend, cycle and seasonal components in spatio-temporal models, where the spatial component is based on a continuous projections of spatial covariance functions. Indeed, our proposed model can be seen as an extension of Laurini (2019) approach to spatial point process, where the dynamics in point process are captured by persistent term and mean-reverting components, plus the spatial term, which is time-varying by the autoregressive group structure. The persistent term is modeled as a first order random walk for a latent component whereas the cyclic component is based on a second-order latent autoregressive structure.…”

Spatial Heterogeneity in Climate Change: Evidence from Brazilian Biomes

“…Our work contributes to the analysis of the spatial heterogeneity of climate change effects by proposing a spatio-temporal decomposition of the permanent and transient components of changes in climate patterns, allowing and testing for the existence of spatially localized components for these effects. As stated in [24] (and references therein), one way to verify the existence of changes in climate patterns is through the estimation of trends and periodic components. In this work, we generalize this structure to estimate models with permanent (trend) and transitory (seasonal and cycle) components for each region (biome), comparing it with a model of unique global components, related to the hypothesis of spatial homogeneity in warming patterns linked to climate change.…”

“…In this work, we generalize this structure to estimate models with permanent (trend) and transitory (seasonal and cycle) components for each region (biome), comparing it with a model of unique global components, related to the hypothesis of spatial homogeneity in warming patterns linked to climate change. In particular, our analysis is focused on testing whether the trend component, which captures non-reversible patterns of climate change [24,25], is global or local. We analyze the hypothesis of a single global trend, implying spatial homogeneity of climate change, versus a hypothesis of multiple trends and spatial heterogeneity in climate change effects, through a Bayesian test using the Bayes Factor, and also through information criteria (DIC and WAIC).…”

Quantifying deviations from separability in space-time functional processes

“…Spatio-temporal data is ubiquitous in many branches of modern statistics, such as medicine (Worsley et al, 1996;Lindquist, 2008;Skup, 2010), urban pollution (Krall et al, 2015), climate research (Laurini, 2019;Chattopadhyay et al, 2020), spectrograms derived from audio signals or geolocalized data (Rabiner and Schafer, 1978;Bel et al, 2011) and perhaps most prominently in geostatistics (Mitchell et al, 2006;Gneiting et al, 2006). For a review of spatio-temporal models with an extensive list of fields of applications see Kyriakidis and Journel (1999) and the references therein.…”