Investigating rainfall trend over India using the wavelet technique Ranjit Kumar Paul and P. S. Birthal ABSTRACT In this paper, using the wavelet technique we analysed rainfall behaviour in the country across different agro-climatic zones over a century. Findings indicate that at the national level there is no signicant trend in rainfall in the long run, but there are pockets of change in the rainfall pattern. There was a signicant increase in the rainfall in the arid zone, whereas in the humid, semi-arid tropics and semi-arid temperate zones the trend was downward but insignicant. The behaviour of rainfall was different during this period. Except in the arid zone, we nd a similar trend in other zones increasing initially, tapering off in the middle and then declining but with some difference in time intervals. In the arid zone, the behaviour of rainfall had been erratic. In the short run, the direction of change in trend remains the same as in the long run but the change is statistically signicant. Ranjit Kumar Paul (corresponding author) ICAR Indian Agricultural Statistics Research Institute, New Delhi 110012, India E-mail: ranjitstat@gmail.com P. S. Birthal ICAR National Institute of Agricultural Economics and Policy Research, New Delhi 110012, India Key words | agro-climate, India, rainfall trend, wavelets INTRODUCTION Indian agriculture is rain-dependent, with almost two-thirds of the net cropped area being rain-fed. Therefore, a regular rainfall pattern is crucial for agricultural growth; too much or too little rain can have a devastating effect on agriculture. A regular pattern in rainfall may not be observed all the time. The series may behave erratically, with highs and lows. Recent evidence shows monsoon rainfall in India is becoming more erratic (Manton et al. ; Kumar et al. ; Goswami et al. ; Mall et al. ; Ghosh et al. ; Paul et al. ). Auffhammer et al. () have shown that the Indian climate has become drier and warmer, adversely affecting the crop yields. India is a large country, and regional patterns in rainfall and temperature are more important for agriculture. A few studies have analysed behaviour of rainfall at disaggregated spatial units, at the level of meteorological sub-division, state and district. Guhathakurta & Rajeevan () found a sharp decrease in the number of rainy days and amount of rainfall in many sub-divisions of India. They reported a decline in rainfall in the months of June, July and September and an increase in August in a few sub-divisions. Goswami et al. () studied rainfall behaviour over central India for the period 19512000 and observed a signicant increase in heavy rainfall events and a decrease in moderate rainfall events. They also reported a rise in the frequency of extreme rainfall events such as droughts and oods. Krishnakumar et al. () studied variations in monthly, seasonal and annual rainfall in the state of Kerala for the period 1871 2005, and found a signicant decline in the south west mon- soon (JuneSeptember) rains, especially in the months of June and July; and an upward trend in winter rainfall in the months of January, February and April. Soltani et al. () computed rainfall and rainy day trends in Iran. Some studies have also attempted to predict rainfall be- haviour and estimation of trends using recently developed nonparametric techniques like wavelets. Kallache et al. () applied discrete wavelet transform (DWT) analysis to river run-off data from gauges in Southern Germany to separate the variability of small and large time-scales, assum- ing that the trend component is part of the latter. Adamowski et al. () proposed a method for trend detec- tion and estimation based on continuous wavelet transform 353 © IWA Publishing 2016 Journal of Water and Climate Change | 07.2 | 2016 doi: 10.2166/wcc.2015.079