Affiliations: [a] Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
| [b]
Department of Mathematics, Dalian Maritime University, Dalian, Liaoning, China
| [c] State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan, Hubei, China
| [d]
Navigation College, Dalian Maritime University, Dalian, Liaoning, China
Correspondence:
[*]
Corresponding author. Jun Xia. E-mail: xiaj@igsnrr.ac.cn.
Abstract: Detecting temporal and spatial trends of annual and seasonal land surface temperature (LST) can contribute to study the effect of climate change and climate variability on temperature behaviors both in time and space. The temporal and spatial gridded dataset of monthly mean LST series produced by Berkeley Earth were chosen for the analysis of LST trends. The dynamic programming (DP) based segmentation algorithm is a fast and efficient time series segmentation algorithm which can identify multiple change points in a given time series. Multiple change points of annual and seasonal LST average anomaly time series during the period 1880–2013 (reference to the 20th century average) were identified by the DP based time series segmentation algorithm. Schwarz’s Bayesian information criterion (BIC) was applied to automatically determine the optimal segmentation order. BIC selected one change point for annual and seasonal average LST except for the autumn season’s. Regardless of the number of change points, at the first segment, trends are always increasing and at the last, they are sharply increasing except for the winter season. Moreover, all the change points locate around El Nin˜o years, La Nin˜a years, and phase transition years of the Pacific decadal oscillation (PDO). Based on optimal time series segmentation results selected by BIC, the spatial distributions of linear trends (slope estimates) of annual and seasonal LST anomalies corresponding to different homogeneous periods are showed. Comparing to adjacent segment, the recent warming trends not only appear in Greenland and the surrounding area, but also dominate most parts of land surface, and significant warming trends appear in high latitude regions of the Northern Hemisphere.
Keywords: Time series segmentation, temporal-spatial variability, trend analysis, climate change
