Absolute homogeneity assessment of precipitation time series in an arid region of Pakistan

Abstract

Homogeneity evaluations are usually performed on the total annual precipitation data, which often fails to detect non-homogeneity in seasonal precipitation. Furthermore, it is required to assess homogeneity using multiple methods as the performance of homogeneity testing methods depend on the distribution of the data. This is particularly important for the arid region where distributions of seasonal and annual rainfall are often non-normal. The homogeneity of annual and monthly precipitation datasets of 14 meteorological stations located in the arid region of Pakistan was assessed in this study using the Pettitt's test, the standard normal homogeneity test (SNHT), the cumulative deviation test, the von Neumann's ratio test, the Bayesian test, the Worsley's likelihood ratio test, and Student's t-test at a 95% confidence level. The rainfall series were categorized into three classes, namely "useful", "doubtful" and "suspect" based on the results of different homogeneity tests. Results suggest that rainfall time series for most of the months in all the stations are useful. The rainfall time series are found doubtful for the month of June at two stations, for April at one station, and suspect for November at only one station. On the other hand, the annual series were found useful at 12 stations and suspect at two stations. Comparison of different homogeneity tests revealed that SNHT and Worsley's tests are the most sensitive, and cumulative deviation test is the least sensitive to changes in monthly precipitation data. In the case of annual series, the von Neumann's test was found most sensitive compared to other tests.