I have hourly rainfalls of a station for the last 7 years and daily rainfalls (9 to 9 am) before that (for 25 years) and want to compute 24 hour rainfall of 50-year Return Period. I noticed significant difference in annual maximum values estimated from hourly data (for consecutive 24 hours) and daily rainfalls (at fixed time interval) for the last 7 years .
What is the best method (to use total record) for computing my design rainfall, at 50-year return period, using above data?
I think this might help. https://www.engr.colostate.edu/~ramirez/ce_old/classes/cive322-Ramirez/IDF-Procedure.pdf
This is standard hydrology method. Basically you have 32 years with an annual daily maximum each year. Order these from low To high, and plot the lowest in year 1 on xAxis ... then each ordered year to highest year 32 on x axis, with y axis representing the annual max rainfalls in desired units. Then extend (extrapolate) the resultant curve to year 50. Some may decide to develop a logarithmic or exponential equation with the data to enter the 50 year and compute. One can do this pretty well in MS Excel, or plot by hand. With flood data, log-log to multiple log paper or log-log probability paper is sometimes used.
I would probably disregard that mixing records with 25 year record may from 0900 hours to 0900 next day hours, while 7 years are hourly, and to remain consistent, you could use the same timing for the daily record. Theoretically, this might cause some daily total differences from 0000 hours to 2400 hours, which is commonly used or standard. I would just report the timing used for the daily totals when reporting.
As a first rough estimate, you could aggregate the hourly data to daily data, starting/ending at 0900. This makes for a total of 32 years of daily data. Then you could use standard statistical software like R and an appropriate package like ismev or extRemes to compute 50 year return levels.
"I noticed significant difference in annual maximum values estimated from hourly data (for consecutive 24 hours) and daily rainfalls (at fixed time interval) for the last 7 years ."
So the last 7yeas (24h 9-9) AMS is significantly different from the old data AMS, right? Maybe, when updating the station to automatic, the placement was changed? Perhaps you can make a double mass curve comparing this station with the nearby stations to assess if there is some significant slope change? Maybe you can make corrections before calculating the RP