I would like to say you may have a theory indicates that which factor will influence the rainfall. If you want to estimate, the simplest way is, applying linear regression model in this case. Of course, further analysis is required, at least you should know what factors you can use. 

First estimate the features that effect the rainfall for ex. Temperature, Humidity, Cyclone impact etc. The number of features you identify, that robust will be the prediction.

You can use either a Multi layer feed forward nerual network model or an Adaptive neuro fuzzy inference system as the predictor.

feed these features to the predictor as training samples and give the actual rainfall as the target, in the sequence 2015, 2010, 2005 and 2000.

While testing the network, input the features corresponding to the year 1995 and 1990. You will get the approximate rain estimate.

The years are equi-spaced. Fit linear regression models for the January data, then estimate the 1990 and 1995 January data on that basis. Similarly do so for the February data, for the March data, and so on. This is a possibility.