Assessment of temporal homogeneity of long-term rainfall time-series datasets by applying classical homogeneity tests

• Published in: Environment, development and sustainability development and sustainability, 2024-07, Vol.26 (7), p.16757-16801
• Main Authors: Kabbilawsh, P., Kumar, D. Sathish, Chithra, N. R.
• Format: Article
• Language: English
• Subjects: Calendars; Confidence intervals; Datasets; Earth and Environmental Science; Ecology; Economic Geology; Economic Growth; Environment; Environmental Economics; Environmental Management; Homogeneity; Hydrologic data; Inhomogeneity; Long term; Null hypothesis; Rainfall; Sustainable Development; Tests; Time series
• ISSN: 1573-2975; 1387-585X; 1573-2975
• DOI: 10.1007/s10668-023-03310-0

The word "inhomogeneity" in a time-series analysis is defined as those changes that occur in the time-series datasets due to non-climatic factors. Rainfall datasets without the presence of inhomogeneity are needed for conducting accurate hydro-climatological studies. The homogeneity of long-term rainfall (exceeding 100 years) datasets encompassing Kerala's entire state has not been checked. This research article bridges the gap by conducting homogeneity tests on every calendar month, monthly, seasonal, and annual scale from 1901 to 2013. The classical absolute homogeneity tests, such as the Standard normal homogeneity test (SNHT), Pettitt test, Buishand range (BR) test, and Von Neumann ratio (VNR) test at a 95% confidence level were applied for this purpose. Based on the results obtained, the rainfall datasets are further classified as "useful", "doubtful", and "suspect". The time-series datasets were generated by arranging the monthly rainfall data at each calendar month, seasonal and annual scale. The results indicated that more than 80% of the time-series datasets were homogeneous and accepted the null hypothesis; SNHT (88.50%), Pettit's test (88.12%), BR test (87.35%) and VNR test (82.15%). The research article presents two major findings. Firstly, it identifies cases where a data series deemed homogeneous by one test is considered non-homogeneous by other tests, emphasizing the importance of using multiple tests for homogeneity analysis. Secondly, the study suggests that researchers and practitioners should carefully consider the homogeneity test type and temporal scale/arrangement of the rainfall data while selecting and analyzing it for various applications.