Dear Canute, 

I have sent you the procedures of a word document, let me know if you have any question. 

hello Mr.  Canute Hyandye

For rainfall satellite data  and before use must be calibrated with rainfall data to find the correction coeficient, so you can do a linear regression between the rainfall satellite data (axis orderly) and rain gauge data in (abcises axis).

for more information I am here.

Thanks Mehdi for initial highlights,

I would like to ask If you have any worked example i can mimic, you can use my private email [email protected]

I know running Linear regression is easy, but how to use the results to correct the Satellite data is what i am not sure, kindly be my tutor Mehdi

thank you sir, for a good words

I have some papers , I  sending you in this weekend

greeting

You can try through a simple method by multiplying the satellite precipitation data with the ratio of the mean of observed precipitation to that of satellite precipitation of the baseline period.  I

Thank you Ahuti Shrestha  for your answer, but the rainfall data from the Satelite images is an indirect measure, so you have to calibrate this data using the data from rain gauge stations .

Probably the best choice is to perform a linear regression, following the suggestion by Mehdi Kablouti. After that, you are going to obtain 2 parameters, the slope 'a' and the bias at 0 'b', so that, y=ax+b. With your choice being y=gauge and x=satellite, your calibrated data becomes after apply calibrated_satellite=a*raw_satellite + b

My advice is that you must take special care with the simultaneity in the data. Firstly, group the data if necessary for matching the sampling (for precipitation data, the numbers usually come as a daily, monthly,... cumulative, so maybe it is not needed here). After that, reject not-simultaneous values.

If you have an estimation of the experimental error in the gauge data, it would be the best choice to include them by using a linear regression algorithm that takes them into account. The result is going to be an error also in the a,b parameters that you must propagate in the final calibrated data.

Regards,

Thank you all(Julio,Mehdi and Ahuti) for your valuable contribution.

I am attaching the Excel file with the data from one of 35 stations vs 1 Met station data, i would appreciate Julio A. Castro-Almazán do/demonstrate  your  procedure on the same file so i can see the analysis,results and final corrected  data on the same excel file when you attach it back to me? Please. Thank you once again.

DEAR Canute Hyandye ;

I see your data

you have more data, so you can not use all that :

1. there are large gaps between rainfall values and satellite exmpel (46; 0.4)

2. These ecards be associated with little rainfall or satélaitaire data (we not know?)

For this you can use  only the closest data as (23.68; 11.8) (16.98; 10.4) ...... etc.

good luck

greeting

Hi Canute, 

I had some time to see your data and play with it. The date has vital missing information, though I did my experiment assuming the beginning date is Jan 1. You have few missing data and some of them seem to be in a dry season.

Bais can be corrected in multiple ways: by using a constant multiplication factor, if the satellite rainfall has captured the pattern very well. By multiplying the mean of observed divided by the mean of satellite on monthly bases to correct the first moment and the other more complicated is by histogram matching, this method requires more than 20 years climate data.

But when I came to you data: it can be corrected by a simple correction coefficient, since the CFSR data has captured the gauged pattern very well. I have tried to calculate the long-term monthly rainfall and I found that observed and simulated data have a higher correlation coefficient. Which means the satellite rainfall has captured the observed rainfall pattern very well, but if you see the volume difference, it overprotected by a couple of hundred percent.

Look the speed sheet attached, I have come with a correction coefficient of 0.217599 and see the corrected rainfall.

Please understand me, this method can be applied only for this station since it has captured the pattern, if it didn’t then you have to use other methods. I hope this helps.

Dear Abeyou,

Your answer and calculations procedure you have shown is real amazing! Thanks.

I have gone through your  formula one after another, i have grasped them except for K. You said "I have come with a correction coefficient of 0.217599 and see the corrected rainfall".  Its true K= 0.217599 have worked well with the Satellite data(daily data), but how did you calculate it? Its calculation formula is missing.I see the Correlation coefficient of daily data(Satellite and Met_station is 0.267422276) any use of it in the correction of satellite data?

Please help. Otherwise your  work is very useful and amazing!

Dear Canute, 

I have sent you the procedures of a word document, let me know if you have any question. 

Thanks a lot Abeyou

I m interested in this question as well Canute, may you share your experience with me? I obtainned daily precipitation data via satellite(TRMM 3B42) and ground stations. I have correlation coeeficient of them , after this step ı dont know what ı should do for bias correction on satellite data. I would be happy with ur helps.

Thank you.

 Hi Abeyou Wale and Canute !

Can you share me how to analysis the k-value? and i want to know that which kind of method in this bias correction

Thank you

I have same problem. would you please share your method with me also.

thank you

Hi Abeyou

Can you also share the procedure with me. I have difficulty correcting bias in my data

Esther

Hi Abeyou

I have same problem.I have difficulty correcting bias in my satellite data. Can you  share the procedure with me. 

email address: [email protected]

Efrem

I figured out how the bias correction factor, K is determined in using the simple correction method suggested by Dr. Abeyou Wale Worqlul. K is the ratio of the mean of the observed rainfall to that of the mean of the satellite rainfall.

Hei Efrem

I have not gotten any feedback regarding bias correction. Do you have the procedure they are referring to and If you have, can you share that with me?

Esther

Hi everyone, 

There would be different ways of bias correction as there are different biases. For a linear consistent bias (which has high R2) but high volume error, you can use linear bias correction by applying a linear bias correction. here below i have attached the bias correction procedure: 

Abeyou Worqlul 

Hi Dear Abeyou

Thank you very much for sharing us the procedure for bias correction for satellite  rainfall. 

Efrem

How can I do bias correction of satellite temperature data for use in filling long duration missing values in observed data.

I have encountered four consecutive years and several consecutive months of missing observed temperature data in a data obtained from National Meteorological Agency. How can fill the missing data using the satellite temperature data? Please help me in this regard.

Efrem

Hello Abeyou, Many thanks for the bias correction tool!

Hello Abeyou, Thanks for the tool. I have figure out how you've got the value for K. I just want to ask, What is this kind of method? Its name I mean.

 Thanks

 Hi Indie, 

It is called a linear bias correction. You will apply it when the satellite data captures the gauged variable very well (that means when R2 is high but when it overestimates or underestimates the satellite data). 

Regards, 

Abeyou 

 Thanks Abeyou. It is well noted. It will be of great help for my study on streamflow modelling using SWAT  in data scarce and ungaged watershed.

Regards,

Indie

Take courage !