Anna Kauffeldt, in her study "Imbalanced land surface water budgets in a numerical weather prediction system" (Geophysical Research Letters, doi: 10.1002/2015GL064230) states that:

This study shows that there are strong biases in key hydrological variables for all of the global model outputs. In addition, the data assimilation process causes water budget issues which are seen across scales (grid cell, basin, and global). Since they are persistent over long time periods, this indicates that the soil moisture and snow increments compensate for systematic deficiencies in the models. The reason for using data assimilation, to improve the forecast performance for atmospheric variables, is at the same time reducing the usability of hydrological model components. Better process descriptions and higher resolution may help to alleviate these issues in the forecasting system, and bias correction will be a necessity for hydrological applications until they are resolved. 

This study used ERA reanalysis data but since reanalysis is carried out with the goal to close the energy budget, not the water budget, it probably holds for most, if not all reanalysis products.

Another reason to be careful withreanalysis data in hydrological applications has to do with the often differing spatial scales. Reanalysis data give averages over global grid cells, whereas hydrological applications commonly are relevant in smaller scales. Examples from e.g. Central America points at large discrepancies between local observations and reanalysis data. 

Dear Pradeep

Generally, reanalysis data are suggested for areas where data is not available. There are number of products that can be use beside NCEP precipitation. You can find lots of data products here: 

http://apdrc.soest.hawaii.edu/datadoc/ncep_mon.php

it is often recommended to check the reliability of such data by comparing with some existing available data. It is better you check it first then use it..

Thanks

NCEP and NCAR data are too coarser to get any meaningful signal over river basin for extreme rainfall. try to get data set of fine resolution (preferably 0.5 degree of less). for daily extreme rainfall analysis you may use freely available satellite derived rainfall data of fine resolution. for e.g. you may try TRMM, GPM, IMSRA etc.

Pradeep,

The NCEP/NCAR Reanalysis is fine for assessing the large-scale atmospheric pattern.  However, it is very coarse compared with the scale of precipitation systems. I'd recommend a higher resolution data set that incorporates rain gauge observations.  For instance, Stage IV precipitation is a 4-km radar-derived, rain gauge corrected data set (http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4/) for the continental United States.

You could search through the IRI Data library for a precipitation data set in your region of interest: http://iridl.ldeo.columbia.edu/docfind/facetedbrowser.html?Set-Language=en&taxa=iridl%3AMaproom_Search&itemClass=iridl%3Amap_room

Anything less than 0.25 degree horizontal resolution could be adequate, depending on your research question.  CHIRPS is newer global data set that may be of interest.  I haven't worked with it, so I can't personally comment on its strengths/limiations. I definitely recommend delving into the literature and/or documentation to learn about the limitations of the data/model! 

Best!

Stephanie

http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4/

http://iridl.ldeo.columbia.edu/docfind/facetedbrowser.html?Set-Language=en&taxa=iridl%3AMaproom_Search&itemClass=iridl%3Amap_room

Thank you Kamal Ahmed Sir, Charu Singh Sir and Stephanie Mam. Thank you for your Suggestions...

Current atmospheric models aren't capable of representing the high spatial resolution and extreme rainfall events required for intensity, frequency, duration curves used by Civil Engineers. It is better to use any observations you can lay your hands on.

The Regional Frequency Analysis approach of Hosking and Wallis is a better way to handle this. Their approach is the basis for NOAA Atlas 14 (NA14), Precipitation Frequency Atlas of the Unites States. The approach is particularly useful in data sparse areas.

In some cases it's possible to use observations from areas with similar extreme precipitation characteristics. NA14 Vol 5 for selected pacific islands is a good example of this. Some characteristics of extreme precip in NA14 Vol 2 were used in developing NA14 Vol 3 for Puerto Rico. 

http://www.amazon.com/dp/0521019400/ref=rdr_ext_tmb