Pattern hunting in climate: a new method for finding trends in gridded climate data

Trends are very important in climate research and are ubiquitous in the climate system. Trends are usually estimated using simple linear regression. Given the complexity of the system, trends are expected to have various features such as global and local characters. It is therefore important to develop methods that permit a systematic decomposition of climate data into different trend patterns and remaining no‐trend patterns. Empirical orthogonal functions and closely related methods, widely used in atmospheric science, are unable in general to capture trends because they are not devised for that purpose. The present paper presents a novel method capable of systematically capturing trend patterns from gridded data. The method is based on an eigenanalysis of the covariance/correlation matrix obtained using correlations between time positions of the sorted data, and trends are associated with the leading nondegenerate eigenvalues. Application to simple low‐dimensional time series models and reanalyses data are presented and discussed.