What statistical tools are best analysing rainfall data (overall trends, decadal trends, intensity distribution, extremes, etc.)?
1$\begingroup$ My question is how analyze it, if I have daily rainfall data for, say, 30 years over Indian region. What statistical methods are apt (regression, pdf etc) for analysing it. $\endgroup$– ajileshFeb 26, 2017 at 11:17
1$\begingroup$ Daily rainfall data like measurements from one or more stations or like gridded data? $\endgroup$– FuzzyLeapfrogFeb 26, 2017 at 22:21
1$\begingroup$ Gridded data 0.25 degree resolution $\endgroup$– ajileshFeb 27, 2017 at 2:25
1$\begingroup$ I know this paper, where they created such a gridded data set. Additionally, they analyzed the data set and compared it with other existing data sets. Maybe this will give you some ideas: dx.doi.org/10.1127/0941-2948/2013/0436 $\endgroup$– FuzzyLeapfrogFeb 27, 2017 at 10:18
1$\begingroup$ One consideration is that rainfall data are frequently non-normal, so you cannot apply methods that assume normality (e.g. most regression methods) without some type of transformation (e.g. logit). $\endgroup$– El NiñoMar 1, 2017 at 0:22
It depends on your data, if the rainfall data is in a grid with temporal and spatial dimension, you can use a NetCDF format to analyse the data via R, matlab, IDL or Python. NetCDF is Network Common Data Form, which is a very common format of grided and temporal data. In Python, you can simply do several 10-year average map to see the long term decadal trend. Or you could plot a transient period of a smaller region with average rainfall to see monthly or seasonally change. To find out the intensity of distribution you could plot a map with a logarithmic scale of colour. Use a threshold to identify extreme rainfall event that is above the threshold. I use a Python package called iris which can easily plot graphs using NetCDF data. You can look at some examples here http://scitools.org.uk/iris/docs/latest/gallery.html
For getting the rainfall data, you can download the rainfall data from the CRU-NCEP database https://crudata.uea.ac.uk/cru/data/hrg/ CRU stand for Climate Research Unit and it has the most accurate historical data available to the public. They are reanalyzed data from meteorological stations and ships. The highest resolution available is 0.5 latitude * 0.5 longitude degree grid. And for temporal resolution they have 6-hourly and daily. The data is in NetCDF format which is a format most map processing data use. If you don't know how to handle ncdf you could use Google Earth as well https://crudata.uea.ac.uk/cru/data/hrg/cru_ts_3.23/ge/ .
To know more how the data is derived you can read this paper. Hope it helps. http://onlinelibrary.wiley.com/doi/10.1002/joc.3711/abstract