Abstract
The estimation of rainfall variability, especially in arid regions, represents a major element for flood prediction and water resources development design works. Such a task presents a major challenge to water resources engineers and hydrologists in arid regions due to the extreme random and erratic nature of rainfall events, which is further compounded by climate change impact. In arid region that covers major portion of Saudi Arabia, the length of rainfall and runoff records is usually short and sometime with information gaps for undertaking proper design work. Such constraints present a difficulty in the application of rainfall and runoff frequencies. In this study the application of different frequency for stations with missing rainfall record in Al-Madinah region, located in western Saudi Arabia was addressed. The analysis first used the record of 10 rainfall stations over the period of 1970-2015 (46-years), to fill the missing information through the application of Inverse Distance Weighted and Kriging techniques and later applied Gamma and GEV probability density distributions. The two distributions parameters were estimated and tested by the K-S and Chi square tests. The two distributions predication estimated the maximum annual rainfall depths for 100, 200, 300, 500 and 1000 years return periods. The Gamma distribution provided a better fits for 3 stations with values ranging from 64 to 92 mm depths for the 100 year return period while GEV for the remaining 7 stations ranging from 55 to 107 mm depth. In addition, the Probable Maximum Precipitation (PMP) technique was applied to estimate the different return periods based on 24-h maximum annual amounts. The proposed approach may provide a mean to estimate the design rainfall depth for stations with missing records, however, it may require further evaluation for other regions of Saudi Arabia or other areas of similar characteristics.