Comparison of long-term and short-term trends of annual rainfall in India: a case study

Abstract

ABSTRACT India experiences rainfall from southwest and northeast monsoon, with large spatial and temporal variability reported in recent years. Rainfall trend could be essential for disaster preparedness or long-term planning of agriculture and economic advancement. Long-term and short-term trends observed in rainfall across different subdivisions of India could be different. The result of trend (or any other) analysis could vary when different tests are employed and hence it is advisable to employ more than one test for any statistical check. Therefore, four tests were employed for detection of trend including Mann-Kendall test, Spearman’s Rank Order Correlation test, Wald-Wolfowitz Run test on data, and Wald-Wolfowitz Run test on successive difference of data; and the slopes were estimated by Sen’s slope as well as linear regression analysis. For detection of possible change point, the four tests included Pettitt’s test, Von Neumann Ratio test, Buishand’s Range test, and the graphical test of Cumulative Departures from Mean. Cases of conflicting results in the four tests were addressed with categorical inferences: useful (one or less test rejects null hypothesis); doubtful (two tests reject null hypothesis); suspect (three or more test rejects null hypothesis) – based on approach described in literature. For seven subdivisions in India, trend was statistically significant at 5% level, out of which for five subdivisions, change point was also identified. Decreasing trend was observed for: Nagaland, Manipur, Mizoram, and Tripura; East Uttar Pradesh; East Madhya Pradesh; and Chhattisgarh, whereas increasing trends were identified for Konkan and Goa; Coastal Karnataka; and Telangana. Change point was detected for four other subdivisions as well. The mean was observed to be unstable for nineteen subdivisions, which included all nine subdivisions with identified change points. The short-term trend was found to be at variance with the long term trend in several subdivisions and significant trend was observed for Saurashtra, Kutch and Diu in the last quarter period (1980–2016). Such findings highlight the necessity of short-term trend analysis for indications of possible climate change effects on recent rainfall records.