Extreme rainfall is the main factor triggering flooding in various regions of the world including Indonesia. The increase in intensity and duration of current extreme rainfall is predicted as a result of global climate change. This paper aims to analyze the impact of extreme rainfall to the peak discharge of flood hydrographs at a watershed outlet in Palu, Sulawesi, Indonesia. Maximum daily rainfall data for the period 1990-1999 recorded at the Palu Meteorological Station, Central Sulawesi were selected using the Annual Maximum Series Method, and grouped into two types. Type I is the maximum daily rainfall data with extreme events and Type II is the maximum daily rainfall data without extreme events. Frequency analysis was applied to the two data groups using the best distribution method of: Normal, Normal Log, Pearson III Log, and Gumbel to obtain the design rainfall of each data group. In the next stage, the design rainfall transformation into a flood hydrograph is performed using the Nakayasu Synthetic Unit Hydrograph based on a number of return periods in one of the rivers flowing into Palu Bay, namely the Poboya River. The analysis results show that the design rainfall graphs with both extreme rainfall and without extreme rainfall are identical at the low return period and divergent at the high return period with a difference of up to 21.6% at the 1000-year return period. Correspondingly, extreme rainfall has a greater impact at the peak of the flood hydrograph with increasing return periods ranging from -1.28% to 26.81% over the entire return period.

Design rainfall is one of the hydrology quantities that can be used for analyzing design flood. This discharge can then be applied to design and manage civil engineering structures such as drainage channels in urban areas. Generally design rainfall is analyzed using maximum daily rainfall data with a minimum data length of 10 years obtained from a rainfall observation station. In various references, the use of rainfall data with a longer range is highly recommended. This paper intends to examine the use of variations in the length of rainfall data on the consistency of design rainfall. Rainfall data from the Mutiara Meteorological Station of Palu, Central Sulawesi (1990-2019) are grouped into five categories based on the length of the data: A (10 years), B (15 years), C (20 years), D (25 years) and E (30 years). Frequency analysis was performed to predict the design rainfall of each data group using four distribution methods: Normal, Normal Log, Gumbel and Log Pearson III. Design rainfall consistency is measured based on design rainfall deviations which are calculated using the longest data. The analysis shows that the consistency of design rainfall in the study area at various return periods for all statistical distribution tends to increase in proportion to the length of the data with high category. This consistency value indicates that the 10-year rainfall data is still sufficient for frequency analysis.