There are numerous techniques for doing this.

The following paper may help you to select the methods.

It is good to use more than one method to check the general consistency of the shift.

Best of luck for your research

I go through some paper determining the rainfall shift but I would suggest you to chose small area with same climatic conditions as if you chose relatively large study area then you may not able to conclude solid results due to inconsistencies caused by climate contradictions. Similarly also separate the research in to seasons rather than annual to understand seasonal shift then you can combine the over all impact.

Best Luck!

BFAST package in R is good for identifying breaks/shifts in time series datasets. I suggest you explore more at http://cran.r-project.org/web/packages/bfast/bfast.pdf

Gopala, I really think that when we are talking abou rainfall the answer to your question varies upon the time intervall which you use.

In daily format data, I think the randomness is very high and it is not easy to decide weather there is a shift or not. But in monthly or yearly format while system usually remembers the past data (persistance) you can simply plot the changes of max. Cross correlation between each months or years. I think this is a good start for testing your hypothesis. What is more, you can choose a reliable index like NAO, ENSO etc. to compare the periodicities and possible changes and shifts in the procedure.

If you are intending to homogenize a set of rainfall series from stations with relatively homogeneous climate variability, my R package Climatol (http://www.climatol.eu/) can do the job. But as Babak has pointed out, daily data are too noisy for that, and therefore Climatol is devised to work mainly with monthly data.

What are you wanting to look at depth, intensity, time period of consecutive rain days, time period between rain days.  The latter two can be analysed using Markov Chain, (see Gabriel, K.R., Neumann, J., 1962. A Markov chain model for daily rainfall occurrence at Tel-Aviv. Q. J. Royal Met. Soc. 88, 90–95). For the other variables the BFAST package would be a good way to go.

you can calculate the shift detection in rain data sets by the Pettitt test,  the Buishand U-statistics, Bayesian procedure of Lee and Heghinian and segmentation Hubert.

thank you sir

http://khronostat.software.informer.com/1.0/

Khronostat is a free software with several methods of detection of rutpures in times series

Here is a You tube 2 min video to explain how to prepare the files

https://www.youtube.com/watch?v=0CfPuwNvwIA

Best

Gil Mahe

Hydro Climatologist at IRD

Montpellier

France

thank you Gil mahe,Jose A. Guijarro,Naga Manohar Velpuri,Pankaj Dey . for sharing your valuable ideas.

Of the many possible options for such problems related to shift and change detection, another potential choice is a Bayesian changepoint method available in R, Matlab, and Python (https://github.com/zhaokg/Rbeast). It has been used for problems in dozens of fields. Maybe it helps with problems like yours, but it may be not.