Uncertainty in IDF Curves Rami Mansour, Fahad Alzahrani and Donald H. Burn Department of Civil & Environmental Engineering University of Waterloo Waterloo ON CANADA
Introduction Work has focused on quantifying the uncertainty in IDF curves for a single site for current conditions Rami has led this work (January to April) The next steps will involve looking at two additional issues (work by Fahad): Uncertainty in regional estimates of IDF curves Uncertainty in IDF estimates under climate change
Methodology The starting point for the work is the weather generator, which was used to obtain daily precipitation values Used the 7 variable version of the model Used stations that are close to London (<50 km) Results in 9 stations not all have hourly daily Turned off the perturbation feature Gives better agreement with historical data
Methodology Goal was to evaluate the capability to reproduce the historical (observed) IDF curve for London Requires disaggregation of generated daily data to hourly data A model based on the method of fragments was developed to do this Similar events are identified based on similarity in daily precipitation and hourly precipitation for the last hour of the previous day Simpler version of the model developed by Karen Hofbauer (Wey)
Results First step was the calibration and testing of the disaggregation model Single parameter to adjust in this model Weight to apply to similarity in daily versus hourly values Model results were found to not be overly sensitive to the parameter Testing was done using data from London Other sites have not been tested
Results
Results Disaggregation model was then used to get hourly precipitation values from each daily value generated by the weather generator These were used to extract, for each year, the values required to create an IDF curve for each rainfall duration of interest 1, 2, 6, 12 and 24 hour Since we have 27 years of input data for the weather generator, we created sequences of length 27 years This has been done 50 times and IDF curves obtained for each sequence
Results Percentage Error in IDF Values (based on 50 sequences) Return Period Rainfall Duration (hours) 1 2 6 12 24 2 0.35% 5.95% 12.70% 6.78% 0.20% 5 1.11% 4.28% 7.62% 4.31% -3.76% 10 1.44% 3.58% 5.44% 3.19% -5.49% 25 1.74% 2.94% 3.43% 2.12% -7.11% 50 1.92% 2.58% 2.29% 1.50% -8.03% 100 2.06% 2.29% 1.37% 0.98% -8.78%
Results
Results
Results
Results
Results
Conclusions from this phase Disaggregation model creates reasonable hourly data Agreement between generated and historical IDF data is good Better for shorter durations and shorter return periods Approach can be used to quantify the uncertainty in IDF curves under current conditions
Next Steps Multiple sites and regional estimates of IDF IDF data have to be extracted for all sites (not just London) Data from multiple sites to be combined to estimate IDF for London (regional estimate) Future work on climate change scenarios The uncertainty in estimates of IDF will be analysed
Con t By using Gumbel extreme values type 1 distribution, the function quantile is calculated based on return period (year) and average of the intensity values, standard deviation for different duration of event. The return periods that are used are 2, 5,10, 25, 50, 100 years The event durations are 1, 2, 6, 12, 24 hours.
Con t We have 9 climate stations that are located around 50 km from London, ON. As mentioned in the beginning, the hourly data are used through WG to create the intensity value for each station for different duration of event.
Table of 9 climate stations A Table of nine stations that are used in WG perturbation removed close. Station s name Dorchester 1 Embro 2 Exeter 3 Folden 4 Stratford 5 St. Thomas 6 Woodstock 7 London 8 Ilderton 9 No.
Results After running the model, IDF values have been received and created the curves. IDF Curves show us the extreme rainfall of each station for different return period. There are differences in values of function quantile WG perturbation removed because the distance between each stations that are located around London
The IDF curves from WG perturbation removed close.
Issues and Challenges Weather generator model is slow and the procedure is computationally intensive May need to improve the weather generator performance Need to determine if we should use the perturbation feature for the climate change scenarios May need to address the uncertainty in IDFs that arises from the limited record length This may be fairly easy to do, but computationally intensive