Using the extreme value analysis tool

This section covers using the Extreme value analysis for Australia where observed data is recorded daily only.

1) Select extreme value analysis



2) Select the following in order (refer Figure 1)

a) Data type Daily

b) Select BoM site

c) Select date range (month) select all

d) Select climate variable

e) Historical period (keep this consistent with the baseline you are using in the analysis)

f) Leave event type as high (high gives maximum highs (hottest day maximum), low gives minimum highs (coldest day maximum)

g) Select number of consecutive days

i) A value of 1 gives the data for a single day event.

ii) If you select more than one day in step (g) you can choose between event definition “consecutive” or “average”. Consecutive looks for a temperature that exceeds a threshold each day. Average looks for the threshold temperature averaged across three days.

Anecdotally – for Sydney the return period using the average above was half (twice as frequent) of that given by using the “on consecutive” (absolute) button

h) Click Run


3) The table at the bottom of the form will populate with data (refer Figure 2)

a) Click the data to generate a chart

b) Change the extreme event value or the return period to see how they are related

c) Click on table to generate the table of the set return periods

d) Export the table to excel for comparison

e) Repeat using the different projected years by clicking the scenario button



How to read a table of return periods

This section sets describes the how to interpret the values in a table generated using the extreme event analysis tool in SimCLIM.

          1. Extreme event return periods


These tables set out temp and precip values against a return period

Moving left to right

RP = return period in years

Chance = chance of that event occurring in any given year

Bold value = baseline value (the temperature that corresponds to the return period in column 1 in the baseline (observed) climate.

25%, 50%. 75% - from the 25th, 50th (median) and 75th percentile of the GCM ensemble respectively

Paired columns - temperature value (°C) and RP BL

  • Temperature value is the projected temperature value for that future year and that percentile (from the ensemble result) for the return period in column 1
  • RP BL return period of the projected temperature value in the baseline climate.


e.g. Collinsville, 2050, 75% 5 year RP (column 1) – value is 43.5 degrees. In 2050 it will occur every 5 years on average. In the baseline it would have occurred every 95.8 years. This is almost a 20% increase in frequency.

Very Large RPBLs and -1.0s mean that that value is outside the historical values observed in that location.


Using the Power Weighted mean function

The power weighted mean field can be used to get a better model fit towards the extreme end, with underperformance away from this extreme end.


We did an experiment for BoM site Observatory Hill. RCP8.5, high sensitivity, 2030 ACCESS 1-0. Single day. 46 degrees.

There is a huge range in the results as shown in the table below.


Power weighted mean

Return period for 46 degree day

0

696652

1

886

2

253

3

156

4

127


This changes the RP tables that are generated within the Extreme Value analysis as well.