If you have other nearby weather stations with full time series around the station where you have gaps, you can calculate the gaps by linear regression with the closest station or multi-regresion with two or more. This is a tradicional method. Other more complex but currently used methods are Kriging described in geostatics. Rainfall is not a continue parameter like temperature and is just probability the estimation.
As no well defined method is available, radius of influence is determined subjectively by personal experience varying from 150 km to 300 km for use in the objective analysis of rainfall (Roy Bhowmik et al., 2005; Mitra et al.,1997; 2003). The dimensionless weighting functions are suggested on the basis of the logical and geometrical conceptualizations. In reality, it is expected that the weights should reflect to a certain extent on the observational network and regional behavior in the occurrence of the phenomenon concerned.
When you say gaps are you referring to gaps between rainfall sensors or gaps in time. For gaps between sensors in the US radar is used to calculate rain estimates and in cases where no radar exist satellite is used by NOAA. Both are calibrated and corrected from ground sensors. Here is a page that describes what is done in both QPE and SPE. Hope this helps:
If you have other nearby weather stations with full time series around the station where you have gaps, you can calculate the gaps by linear regression with the closest station or multi-regresion with two or more. This is a tradicional method. Other more complex but currently used methods are Kriging described in geostatics. Rainfall is not a continue parameter like temperature and is just probability the estimation.
I completely agree with Raymundo and Ashim comments. If you don't want to use linear regression, I suggest you to use regional analyses to obtain the best equation between rainfall and other geomorphological parameters , such as Area, Slope, River length and ....
Moreover, by using Interpolation techniques like IDW, Kriging and other methods you can find the best method and estimate the rainfall values in desired locations...
I agree with Raymundo and Ashim comments. Also, you can use regional techniques with a set of rainfall stations in "an homogeneous region". You should be careful with the orography of the area. In my opinion, one of the most easiest and precise methods is to correlate the series among your gauge and other gauges with common data (one to one or one to the weighted mean of the others). If you obtain a good correlation among them, it is expected that the same correlation will be fine for the values belonging to the gap period.
Regional analysis is recommended to predict peak discharge in areas that have not enough data series for peak discharge and i couldnt find any article that use this method to predict missing data of daily rainfall . For some addition help i used langbein method to find homogenous basin
The method of nearby stations is very good. Calculate the correlations and choose stations about 10 km from your station. Your catchment should be relatively flat. You can also check the covariance of those nearby stations to see how variable your rainfall is
Hi, you may also want to check out surfer (goldensoftware.com) it automates spatial interpolation(s). Methods included are: Kriging, Local Polynomial, Nearest Neighbor, Radial Basis Function, Natural Neighbor and Triangulation with Linear Interpolation.
I'm currently working on a method that utilises reverse hydrology - that is estimating rainfall from streamflow. Unfortunately its not ready for publication yet.
Sangamreddi Chandramouli "take the average of the previous 35 to 50 years data". Data of the too long period can be misleading, because the climate may have been changed, however 50 years seems allowable.
50 year data is quite acceptable while double mass curve technique is available in all hydrology text book. In my region I used data for 25years for finding missing values, the result was reasonable.
try the double mass method using nearby stations, however it is still possible to complete or expand your data using remote sensing databases like TRMM. best of luck
I agree with Mr. Khattab using data from nearby stations with double mass curve and the possibility of using GLDAS (Global Land Data Assimilation System)
This technique is ideal to fill gaps, you can just give the available time-series and the missing locations will be simulated with a resampling technique that generates realistic rainfall patterns:
https://doi.org/10.5194/hess-18-3015-2014
Here you can find a tutorial to use the same technique with a free software: