Different approaches have been found to be effective for identifying different types of changes so the answer will partly depend on the nature of changes that you suspect may be an issue for your data set and the type of changes that exist in the data set.  Hence, applying many of the techniques suggested so far will be useful to help learn more about your data.  One additional technique to consider is Bayesian Change Point Analysis, which is available through the BCP package in R.  For any changes that are found, it is important to examine the metadata for your time series to look for any changes in measurement protocols or equipment that may explain the changes as well as any known physical changes that could explain the change (i.e., a forest fire on a watershed can result in changes to the streamflow time series).

Hi

You may want to try Rebecca Killick's Pruned Exact Linear Time (PELT) algorithm. It's avaliable in R.

Thank you José M.C. Pereira

Double mass curve technique with a statistical test for significance.

This may be useful:

Thank you Sir 

to start with a field visit to the stream gage station

and review the station history

Ing. Marcelo Jorge Pérez 

I think in order to make robust conclusion, you had better use two different methods. For example, Double mass curve and Mann Kendal Rank Test. Please refer to this paper. 

Wei, X., and M. Zhang (2010), Quantifying streamflow change caused by forest disturbance at a large spatial scale: A single watershed study, Water Resour. Res., 46, W12525, doi:10.1029/2010WR009250.

Thank you Sir

I can suggest three methods for you:

- Comparison with Double mass curves if you have more than two stations in order to verify the quality of the data and also if the change is natural and not do to biases in the measurements

- Petit test as suggested by others, it detect only one change point

- The segmentation procedure of Hubert which helps to detect more than one change point

Hubert P. (2000). The segmentation procedure as a tool for discrete modeling of hydrometeorological regimes. Stochastic Environmental Research and Risk Assessment 14(4)–(5):297–304.

I think it is better to use Double mass curve which is used to check the consistency of many kinds of hydrologic data bycomparing data for a single station with that of a pattern composed of the data from several other stations in the area.

Different approaches have been found to be effective for identifying different types of changes so the answer will partly depend on the nature of changes that you suspect may be an issue for your data set and the type of changes that exist in the data set.  Hence, applying many of the techniques suggested so far will be useful to help learn more about your data.  One additional technique to consider is Bayesian Change Point Analysis, which is available through the BCP package in R.  For any changes that are found, it is important to examine the metadata for your time series to look for any changes in measurement protocols or equipment that may explain the changes as well as any known physical changes that could explain the change (i.e., a forest fire on a watershed can result in changes to the streamflow time series).

Here is the paper for detecting the changing point using Modified double mass curve and ARIMA model to detect the change point for slope of DMCs.

Quantifying streamflow change caused by forest disturbance at a large spatial scale: A single watershed study. Mingfang zhang and xiaohua wei