Dear Bir Biswa,

I would like to recommend you to convert the rates into depths for a day (cm/day or mm/day) and then divide the converted depths by 24 hours so that you will know how much rainfall you will get in a hour. There is nothing difficult and complex in the calculation but a pencil and a sheet of paper for the calculation. I hope this would help you.

The question is very tricky, the rainfall is not uniformly distributed over a day or three hours, so this may be numerically possible by dividing by three (for three hours data, answer in mm/hr) or dividing by 24 (for the day value to have in cm/hr), but it is not realistic. May be in your situation, you can use three hours value and place your judgement on the basis of three hours value. You can also find out the period within a day, where most of the rain in three hours is received

regards

You can't.

You can assume that the rainfall will be the same over the three hours.

Or you can choose your stochastic rainfall model of choice and use that to guide in you breaking the three hour rainfall into hourly rainfall. Your rainfall will be STATISTICALLY correct - i.e. when averaged over a long period - but will be wrong should you ever get the hourly rainfall data; in other words, when using the results, do not expect perfect validation against any other data, unless you aggregate that data also to the three-hour level. And remember that the 'statistical correct' values will be affected by the choices made in choosing your stochastic rainfall model.

The hard part is then in choosing a suitable stochastic rainfall model. I have data only for the UK; you will need to do a literature search for a suitable model for wherever is closest to you. Since Bhutan is rather mountainous, I am not sure where would be best.

I agree with Dr Dudley, as rainfall is not uniformly distributed over time

regards

I did not expect that the simple question may lead to this highly complicated speculations and theory. ^^ Unless you have observations made at a temporal scale of interest (i.e. hourly), you will not able to perfectly reproduce data at the resolution from data prepared at a coarser resolution. However, this does not necessarily mean you "cannot" convert  daily data into hourly data. There are many ways to do that, such as assuming uniform distribution (as I did), employing techniques to impose typical hourly temporal distributions (Huff, SCS types, etc.) developed for your areas into your daily data, using stochastic weather generators (such as CLIGEN), other interpolation methods. Of course, you will need to assume something whenever you are using one of the estimation (or simulation) methods. There is no such good or bad model, but we can always find "useful" ones for solving your problems even though they may not be perfect! (Do we have one perfect??? ^^)

I agree that it is not straightforward. Assuming that the rainfall is uniform over three hours may be fine; I presume that it will subsequently be used assuming that it is uniform over one hour. I also assume that the original poster was hoping for something better than assume uniform over three hours - but 'better' is only an assumption, based on choice of an appropriate means for breaking down the rainfall.

Look up dissagregation and find a suitable technique.

I do not think "more complicated" always mean "better"; often the simplest thing works better even though it does much simplify the reality so that it does not look able to well represent the reality. When I was writing the first responses to the question, I was assuming that Biswa must be a student who is learning rainfall depth conversion in a class such as hydrology. (Biswa, I am sorry if your are not a student.) Still I am assuming he is a student since the question seems too complicated if he wanted to incorporate temporal variations into his conversion. ^^ Drs. Kalra and Dudley, thanks for making me realize there is the possibility that he may look for much more advanced answers (rather than the simple math). We will see what he really wanted.

If there is evidence from previous years on the nearest weather station, it is possible to build a long-term connection curve rainfall in 24 hours, 12 hours, 6 hours and 3 hours 1 hour    -   P = f (dT)

I think Bir Biswa asked this question because he was unsure about the uniformity. Otherwise he should not log in to Researchgate and ask a question about dividing one number to 3.

Yes, the intensity of the rainfall changes over time and cannot be calculated by dividing numbers. I agree with Michael and Jeremy.

Of course you can divide numbers and have another number to research about storm water management for fire fighting works in school and colleges, it may give you a result, however it would not be scientifically correct for storm water management.

Another thought is that his question might be about IDF (Intensity-Duration-Frequency) curve or relationship for his project site if I do not seriously take the term, "convert" into account. As Michael mentioned, he may be able to use a IDF curve developed for a weather station close to his project site. If the question is not about IDF but "converting" from one to another, I still would like to recommend to divide it by three (or 24) because it is the simplest way to do that even though probably it may not look like "scientifically correct" (I do not agree with this though).

Rainfall data is collected in a variety of ways and frequencies.  Assuming standard methods were used, if you collect data on a three hour frequency, you cannot with accuracy divide by 3 to get hourly intensity.  However, that may be what you have to do if you are comparing storm events of various studies.  In some areas and storms, you might get away with it, as rainfall intensities are widespread and gentle.  However, for intense and variable storms such as thunderstorms, tropical or hurricanes, monsoon wind driven events, sometimes even 15 minute intensities are not adequate to describe rainfall intensity due to variable conditions.  If you must report intensity in hour with three hour data, I would suggest you divide the 3 hour data by 3 and report it or asterisk it as an average hourly intensity based on three hour sampling frequency.  I remember one storm in Missouri that I was in the field and collecting data with recorders with 5.5 inches of rainfall in 30 minutes collected in the weighing bucket, Fisher Porter Raingauge set up with Weather Service specs.  Reporting this intensity on a hourly or daily basis may miss the point for some types of studies, but perhaps not for others.  If you have a recording station nearby which had similar totals during the storm, you might be justified in using that, but the other reviewers are right, rainfall can be very site specific as seen on Doppler radar for very broad events to localized cells of intensity.  And the assumption with standard methods is the 90 degree from the raingauge top to the surrounding forest or other features is adequate.  But in a hurricane with 100 mile per hour plus winds, there could be some unexpected errors with these conditions.  If nothing else, in your methods, describe what was done.

Opps, I forgot to mention the metric units for you, as you may be more familiar than I.  cm/day is 100 mm/day, so in converting that to mm/hour, take the cm/day and divide by 24.  If you have mm/3 hours, then divide by 3 and see my previous message on the limits associated in doing this.  Just earmark what you did in the conversion or your methods as basically in either case, you are more or less averaging the intensity based on 24 hours or 3 hours by reporting it as an hourly rate.  I hope my talking about inches per hour did not confuse you.  So the storm I mentioned at 5,5 inches in a 30 minute period of time has to have the entire hour to report intensity, but if the 5.5 inches was the entire storm, then that would be 139.7 or 140 mm/hour.  In that instance, it is not an average number but it would be the entire storm rate.  The true instantaneous intensity for the 30 minute storm with 5.5 inches would be 279 mm/hour based on the 30 minute intensity.  In this instance, the hourly storm rate differs from the instantaneous hourly rate.  I hope this is not causing further confusion.  In most storms, if you are analyzing the data, you would take the one hour period with the most rain and report that as the peak storm intensity in inches per hour.  However, for erosion and storm peak flow responses, the peak instantaneous rate may have meaning.  You can always footnote that if you think you need to, especially if you think the averaging is not telling the full story of your data relative to other studies you may be comparing.  By the way, I am retired USFS, so I see I need to qualify that on my data.

if the rainfall depth of the 3 hrs is uniformly distributed. You can simply divide the rainfall depth collected over 3 hrs. by 3. So you get the rainfall depth by 1 hr. 

 I agree with your answers that 3hr rainfall data can be converted into 1hr data. It would solve my problem and doubts if someone could attach few related papers on it.

Is there any literature (journal, etc) to support the answer? The one with dividing the rainfall data by 24 hr. Cause I have the same problem too. I have the hourly rainfall data for 24 hours and now I need to convert it into rainfall intensity, mm/hr. Thanks in advance.

Aileen, the simple answer is no. You can create synthetic hourly rainfall, but that is all.

Your synthetic rainfall can be uniform, or disaggregated to be statistically representative of known rainfall patterns - uniform rainfall across a day is statistically unlikely. To disaggregate you need data on the typical hourly rainfall pattern for the nearest location to which you can get such data, and even then you need to ensure it is appropriate - if you are lowland and the nearest hourly dataset is highland it would be inappropriate to use that as the pattern for disaggregation.

As to how to subsequently process the data, I suggest that you use a webs search (try the keywords disaggregation daily rainfall). You should find a lot of (detailed statistical) theory in the PhD thesis of Paul Cowpertwait, which can be freely downloaded from the British Library Electronic Thesis archive. 

Also try

http://researchonline.ljmu.ac.uk/2344/3/Disaggregation%20Paper-revised%5B1%5D.pdf

http://arizona.openrepository.com/arizona/handle/10150/191279

http://www.mssanz.org.au/modsim09/I8/pui.pdf

http://www.eng.uwo.ca/research/iclr/fids/publications/cfcas-climate/reports/kmwey2006.pdf

hi, i have been following the responses of people about the question. my question is how to move 5mins average rain rate data to mm/hour and rainfall depth of mm/day.

@Sharafadeen Shehu, yours is easier. Ideally your 5 minute data is collected 00:00, 00:05, 00:10, ... so just SUM each set of 12 values for the hourly rainfall. The daily one then falls out from the hourly values.

If the data is not quite so straightforward - say, 00:04, 00:09, ... then for the two values that cross the hour assume uniform rainfall intensity and assign to each hour pro-rata. Thus, the 00:004 data is 4/5 in hour N+1 and 1/5 in hour N-1 and again for the 01:04 data value.

There will be a slight bias, but compared to other problems with rainfall data this will be the easy one.

@Jeremy Dudley, Thanks a lot for the guide. It really help and will make the work faster.