Discovery of anomalies and teleconnection patterns in meteorological climatological data

Abstract

Climate change is an ever-increasing problem in the modern world, and the analysis of data sets to discover patterns and anomalous behaviour, with regards to the climate, is more crucial than ever. However, analysing such data sets can prove overwhelming, as these data sets tend to be too large to manually inspect. As a result, a need for techniques to efficiently scour and manipulate such extensive data, often referred to as data mining techniques has arisen. The work that is carried out within this research consists of using different data mining algorithms to extract anomalies and teleconnections from a data set of monthly global air temperatures, covering a period of 72 years (1948-2019). Anomaly detection is a significant step in data mining that aims to identify data points that deviate from the remainder of the data, often called anomalies. The purpose of anomaly detection in climate data is to identify any spatial (across space), temporal (across time) or spatial-temporal (across both space and time) anomalies within the data set. These anomalies are important to understand and forecast the nature of Earth’s ecosystem model. The anomalies are detected using three different algorithms; K-Nearest Neighbour (KNN), K-Means Clustering, and Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Results obtained from anomaly detection show that the KNN algorithm is the best performing one out of the three. Teleconnections are recurring and persistent patterns in climate anomalies, which connect two distant regions. Teleconnections are important because they reflect large-scale changes in the atmosphere and influence temperature, rainfall, and storms over extensive areas. As a result, teleconnections are often the culprit responsible for anomalous weather patterns occurring concurrently over widespread distances. The teleconnections are detected using three different association mining techniques; Apriori, FP-Growth, and Generalized Sequential Pattern (GSP), over the spatial-temporal anomalies identified before. Results obtained from teleconnection detection show that the Apriori and FP-Growth algorithm output very similar results, while the GSP algorithm failed to output any results due to hardware limitations. The extracted anomalies and teleconnections, obtained from the previously mentioned algorithms, are presented using interactive graphs and heat maps.