Climatological Variability and Limitations of Forecasting Extreme Rainfall Statistics

Stormwater Conference

Rainfall frequency analyses play a vital role in water resources engineering for accurate prediction of flood risk, efficient stormwater management and design of flood control and drainage infrastructure. Limited rainfall data at some sites in the Auckland Region leads to difficulties in fitting a probability distribution and estimation of extreme rainfall event statistics. Climate change and phase changes in the Inter-decadal Pacific Oscillation (IPO) affect rainfall patterns and create more issues in estimating extreme rainfall.

Log-Pearson Type III (LPIII) rainfall frequency distribution analyses were carried out for 10 rainfall gauges within the Auckland Region using HYDSTRA software (HYIFD) to generate intensity-frequency-duration tabular data. Rainfall trend analyses were also carried out based on frequency analyses for moving 20-year interval of rainfall data bins at six rainfall gauges within the Auckland Region with rainfall records ranging from 34 to 72 years of data.

Rainfall trend analyses show higher design rainfall depths during the 1978-1999 period and lower design rainfall depths during the 2000-2017 period. These are because of phase changes in the IPO on rainfall trends i.e. higher rainfall in the positive phase (1978 – 1999) of the IPO and lower rainfall in the negative phase (2000 – 2017) of the IPO. It demonstrates that the natural climatological variability of rainfall extremes is much greater than climate change effects.

Rainfall data shows that 80% of the total rainfall record in the Auckland region has been recorded in the past 19 years during the negative IPO phase. So if these data were used in a regional rainfall frequency analysis (HIRDS V4) the extreme event statistics would be skewed by the preponderance of recent data.

During the past 20 years less extreme rainfall occurred in the Auckland Region than in the 20 years previous. LPIII rainfall frequency distribution analyses of four rainfall gauges with 18-19 years of data show very low design rainfall depths compared to TP108, HIRDS V3 and HIRDS V4 design rainfall depths. TP108 and HIRDS V3 frequency analysis did not have the last 20 years of rainfall data, as a result predict higher design rainfall depths.

Climatological variability is increasingly creating severe risks to New Zealand as a result of continuing sea level rise and the increased frequency and intensity of flood damage on our infrastructure. This paper demonstrates the importance of both the climate variability (IPO) and limited rainfall data availability in estimating the extreme rainfall event statistics. This is critical to accurately predict the flood risk and ensure infrastructure is capable of performing now and into the future.


1 MB
01 Oct 2019


33 MB
01 Oct 2019