in attached file you find my article : intitulated : Spatial Interpolation of Annual Precipitation in Annaba- Algeria - Comparison and Evaluation of Methods.

If you have any questions Im here.

Hi Deepak,

In ArcGIS, there are several methods available for interpolation, spanning from deterministic (e.g. Inverse Distance Weighting---IDW) method and Stochastic (e.g. all sorts of kriging) methods. However, usually kriging gives good result in spatial interpolation. Since kriging considers variogram for computing its weight factors, it has dependency on correct variogram estimation. ArcGIS 9.3 or earlier verson does not have option for correct variogram modelling except doing manual trial and error calibration. I didnt try with ArcGIS10. So, in kriging you need to first identify a correct variogram model to interpolate your data in kriging. You can try this variogram modelling with another commercial package GS+ (a geostatistical software package) or ISATIS. If you would like to go free variogram modelling software you can try GSLIB out. Hopefully, you can then use those variogram parameters in your ArcGIS to interpolate using kriging. If you need details information, you can refer to the variogram modelling part of the following paper.

http://onlinelibrary.wiley.com/doi/10.1002/hyp.10389/pdf

I found spline as the best method, when your data are not densed. Also regression equations could be more relevant when you apply them in raster calculator. A step-wise regression to select independent variables which affect rain value also is advised.

If you have a dtm model, you can use cokriging (general formulation), colocated cokriging or kriging with external drift (see International Journal of Climatology, 29(14): 2156-2170). I think this last one isn' implemented in ArcGis

Note that there is no theory supporting IDW, in most software it assumes isotropy. It is somewhat similar to kriging but does not incorporate all the information that kriging does. Kriging is equivalent to using Radial Basis functions (which includes thin plate splines. There are many papers on the use of kriging to interpolate rainfall

What about Regression Kriging proposed by Odeh, Hengl etc

I think Kinging could be used if you have at least 30 data locations to be able to get a reliable variogram estimation. If the amount of data is less you may use IDW.  

See the paper below on Kinging but for mapping groundwater quality it may help.

https://www.researchgate.net/publication/258225101_Geostatistical_analysis_using_GIS_for_mapping_groundwater_quality_case_study_in_the_recharge_area_of_Wadi_Usfan_western_Saudi_Arabia?ev=prf_pub

Article Geostatistical analysis using GIS for mapping groundwater qu...

You should first analyze the variogram of your data, if a model can be adjusted, then you can use kriging, If it does not splines or radial basis.

You also have to take into account that for kriging you assume the normality of the data, and, generally rainfall data do not have a normal distribution.

the use of cokriging with elevation is also possible but being careful especially in tropical domain.

The first decision is incorporating or not the elevation data in the estimation process. You must analyse carefuly the (sdpatial) correlation between both variables. Resulting maps are quite diferent and reflects this decision (see my previous reference given before) What method to use in each case is a second order decision. If you decide to include elevation in the process, kriging with external drift, simple kriging with varying mean or cokriging are the common techniques to use.

sir, do you know how to do interpolates temporally precipitation daily data in ArcGIS,?, so I can use it in ArcSWAT Input, it is like what we do in SWAT_PCP, unfortunately SWAT_PCP not available for ArcGIS 10x.

A,J. Sousa gave you good advice but I would add a couple of comments:

1. What do you mean by "suitable", presumably you mean for your data set and for your intended use of the results. There is no theoretical answer to your question.

2. Note that precipitation is not a one time event so you might want to think about how time affects the data. Is your data hourly accumulation, daily accumulation, weekly accumulation, time interval averages. Is there a seasonal effect that should be incorporated?

3. ArcGIS spatial analyst includes interpolation by Radial Basis functions as though it were a totally different method than splines or kriging but that is not true.

If you are interested in runoff then you need to be interested in spatial averages, there was some work done on this dating back to the early 1980's by people at Grenoble (they wanted to be able to predict runoff in real time)

If we use the spline method, sometimes it generates some minus values for rainfall, how to overcome this drawback in Arcgis, is there some suggestion? Thanks in advance. I try some methods, find that Spline and natural neigghobur are a bit better.

Some good examples here: http://smrf.readthedocs.io/en/develop/

There are many research outputs on the selection of better techniques of interpolation in ArcGIS. The logic is more mathematical. You better check the why one is better and the other not? For soil mapping and spatial interpolations, I prefer cokriging - kriging point iteration. See the bellow links.

https://www.esri.com/news/arcuser/0704/files/interpolating.pdf

It may not be the right time to answer, still I would like to recommend all the methods you should run and compare. The input data can be divided into training and testing data points randomly, where the interpolated data with training points can be validated with testing points.

I prefer cokriging - kriging point iteration.

The Kriging and the IDW (Interpolated Distance Weighted) methods are very good and suitable for rainfall data interpolation. However, their applications depend on the available number of sample points within the grid plots.

These articles will be helpful to you for further reading.

Hi, anyone could help me? I wouls like to ask, what kind of image data should I use for interpolate rainfall data? Besides that I need the rainfall data value.