ABSTRACT This study analysed long-term rainfall data (1851–2006) over seven climatic zones of India at seasonal and annual scales based on three techniques: (i) linear regression, (ii) multifractal detrended fluctuation analysis (MFDFA) and (iii) Bayesian algorithm. The linear regression technique was used for trend analysis of short-term (30 years) and long-term (156 years) rainfall data. The MFDFA revealed small- and large-scale fluctuations, whereas the Bayesian algorithm helped in quantifying the uncertainty in break-point detection from the rainfall time series. Major break points years identified through Bayesian algorithm were 1888, 1904 and 1976. The MFDFA technique identified that high fluctuation years were between 1871–1890, 1891–1910 and 1951–1970. Linear regression-based analysis revealed 1881–1910 and 1971–2006 as break-point periods in the North Mountainous Indian region. A similar analysis was carried out for India as a whole, as well as its seven climatic zones.

中文翻译：

---

基于分段的印度降雨时间序列（1851-2006）趋势分析和结构中断方法

摘要 本研究基于以下三种技术分析了印度七个气候区的长期降雨数据（1851-2006 年）季节性和年度尺度：（i）线性回归，（ii）多重分形去趋势波动分析（MFDFA）和（iii）贝叶斯算法。线性回归技术用于短期（30 年）和长期（156 年）降雨数据的趋势分析。MFDFA 揭示了小规模和大规模的波动，而贝叶斯算法有助于量化降雨时间序列断点检测的不确定性。通过贝叶斯算法确定的主要断点年份是 1888 年、1904 年和 1976 年。 MFDFA 技术确定高波动年份在 1871-1890 年、1891-1910 年和 1951-1970 年之间。基于线性回归的分析显示，1881-1910 年和 1971-2006 年是印度北部山区的断点时期。对整个印度及其七个气候带进行了类似的分析。