There has been a gradual uptake of the new design rainfall intensity-frequency-duration (IFD) system provided by the Bureau of Meteorology (BoM) since their release in 2016. However, there remains widespread hesitancy to fully adopt the new temporal patterns and methodologies that are proposed by Australian Rainfall & Runoff (ARR, 2016) to accompany them.
What are regional temporal patterns and how are they used?
Temporal patterns are used in design flood estimations to describe the distribution of rainfall with time. After rainfall depth, the temporal pattern of rainfall may have the biggest influence on flood estimates.
Australian Rainfall & Runoff (ARR, 1987) originally recommended the use of the single burst Average Variability Method (AVM). This was a single temporal pattern for each rainfall duration traditionally adopted in designs. It assumed that every storm follows the same general pattern but we all know that’s not quite right!
ARR 2016 changes everything!
In developing ARR 2016, temporal patterns were reviewed for 35 catchments across Australia (including 9 catchments within Western Australia) based on different event types (burst, complete storm and pre-burst) extracted from the national events database.
To accurately account for diverse and different systems of rainfall Australia was divided into 12 regions. The Perth Metropolitan area falls into Flatland West Region as illustrated in the figure below.
10 different temporal patterns were developed for each of 24 durations (from 10 minutes to 168 hours) within four AEP probability ‘bins’ (Frequent, Intermediate, Rare and Very Rare). The recommended design events related to each probability bin are shown below.
|Probability Bins||Recommended Design Event|
|Frequent||50% – 20% AEP (2yr – 5yr ARI) Event|
|Intermediate||10% – 50% AEP (10yr – 20yr ARI) Event|
|Rare||2% – 1% AEP (50yr – 100yr ARI) Event|
|Very Rare||Bigger than 1% AEP (100yr ARI) Event|
This means that in place of one single rainfall pattern we now have thirty different patterns! This equates to ten patterns for each probability bin in each of the 12 regions.
Each of the patterns reflects real rainfall events within the given region. You can even find out the start and end time, and date of the bursts and the location (Longitude & Latitude) that the temporal patterns occurred in from the provided temporal pattern metadata.
The real number of temporal patterns per region has not been determined yet, but 10 patterns seems a good compromise between numerical overhead, pattern availability and result sensitivity. ARR Research Project 3: Temporal Patterns of Rainfall found that an ensemble regional pattern approach using ten patterns can capture much of the variability of temporal patterns and remove much of the variability of the flow estimates.
So why isn’t everyone doing it?
Given the long modelling run times, particularly in two dimensional hydraulic models, it can take much longer to run all 10 patterns for each duration. This could be a large part of the hesitancy. A more practical approach could be to run a separate hydrological modelling processes to determine the average pattern and then run this pattern through the hydraulic model. However, this should be determined by further assessments.
The updated Temporal Patterns certainly provide an opportunity to simulate more realistic conditions of the catchments and estimate more accurate design flows. Along with the new Intensity-Frequency-Durations (IFDs) published for the whole country, it might be a good time to use the current method and re-estimate some the major design floods and hydrology investigations within Western Australia and compare the results.
More information on ARR 2016 can be found in: