Anomalous cluster detection in spatiotemporal meteorological fields

Abstract

Finding anomalous regions in spatiotemporal climate data is an important problem with a need for greater accuracy. The collective and contextual nature of anomalies (e.g., heat waves) coupled with the real‐valued, seasonal, multimodal, highly correlated, and gridded nature of climate variable observations poses a multitude of challenges. Existing anomaly detection methods have limitations in the specific setting of real‐valued areal spatiotemporal data. In this paper, we develop a method for extreme event detection in meteorological datasets that follows from well known distribution‐based anomaly detection approaches. The method models spatial and temporal correlations explicitly through a piecewise parametric assumption and generalizes the Mahalanobis distance across distributions of different dimensionalities. The result is an effective method to mine contiguous spatiotemporal anomalous regions from meteorological fields which improves upon the current standard approach in climatology. The proposed method has been evaluated on a real global surface temperature dataset and validated using historical records of extreme events.

Piece‐wise constant cluster modelling of dynamics of upwelling patterns

“…• Generative model clustering: The work in Ramachandra et al (2019) models spatial clusters with Gaussian density functions, so that patterns that are highly different from the Gaussian are considered anomalous. The authors in Sambe and Suga (2022) put together temperature and salinity profiles while applying the Gaussian mixture model to cluster about 200 ocean depth layers; to discover that the resulting clusters form spatially contiguous regions.…”

“…Among applications, one should point to image data analysis including such areas as Medical imaging, 3D imaging, Oceanography, Industrial automation, Remote sensing, Mobile phones, Security, Face related applications, and so on (Wazarkar & Keshavamurthy, 2018). Other application areas are Earth Sciences (Donatelli et al, 2022; Martino et al, 2018; Ramachandra et al, 2019; Tonini et al, 2022; Yu et al, 2020), Health Informatics (Kamenetsky et al, 2022; Kulldorff & Nagarwalla, 1995; Mattera, 2022), as well as other societal phenomena (Mozdzen et al, 2022).…”

“…In the relevant literature, one may distinguish between two streams of work: (1) methods for clustering at oceanographic phenomena and (2) computer‐oriented studies of upwelling systems. Stream (1) includes works on Trajectory clustering: Specifically, a method from Liu et al (2019) proceeds by using a similarity measure between trajectories, whereas the algorithm from Zhang et al (2018) first partitions a trajectory into a set of line segments by characteristic points at which the object makes a sharp turn, and then groups similar line segments together into a cluster. Generative model clustering: The work in Ramachandra et al (2019) models spatial clusters with Gaussian density functions, so that patterns that are highly different from the Gaussian are considered anomalous. The authors in Sambe and Suga (2022) put together temperature and salinity profiles while applying the Gaussian mixture model to cluster about 200 ocean depth layers; to discover that the resulting clusters form spatially contiguous regions. Hotspot discovery: The work in Martino et al (2018) applies a version of fuzzy method extended by using hyperspheres as cluster centers into a mix of spatial and temporal features.…”

On the nature and types of anomalies: a review of deviations in data

“…Deviant spatio-temporal patterns also figure largely in the video surveillance of streets, parks and train stations. In non-crowded scenes anomalies may be an individual person or vehicle that demonstrates abnormal walking, running, crawling, driving or stopping behavior [ 68 , 293 ]. In crowded scenes individual people cannot easily be distinguished, so the focus is on even more aggregated and abstract motion patterns that capture multiple subjects simultaneously [ 237 ].…”

Dynamic graph embedding for outlier detection on multiple meteorological time series

“…Outlier detection, which identifies instances that are distant from most other observations, is an important field of computer engineering and data mining. Outlier detection in meteorological time series is a necessary research issue because learning the patterns of abnormal climatic events can help reduce losses due to meteorological disasters [ 4 , 5 ]. The concealed information obtained from meteorological data can be detected to analyze climatic changes.…”