190 likes | 200 Views
Explore how uncertainties in extreme weather projections impact risk management for changing climate. Learn about methodologies, study areas, results, and discussions on future IDF curves.
E N D
Christopher Davidson, P.Eng. Hong Liu, Ph.D. Sean Capstick, P.Eng. Janya Kelly, Ph.D. Kevin Mackenzie, P.Eng. Managing Climate Change Risk by Uncertainty Analysis
Outlines Introduction Methodology Study Area and Periods Results Discussions
Introduction Changes in extreme weather and climate events have significant impacts and are among the most serious challenges to society in coping with a changing climate (CCSP, 2008, IPCC, 2007) Extreme rainfall events are often expressed in the Intensity-Duration-Frequency (IDF) statistics Development of future IDF curves depends on various inputs and tools that produce uncertainties (RCPs, GCMs, downscaling approaches) Uncertainty is risk: to manage, we need to quantify
Extreme Events FLOODING VIDEO REMOVED TO REDUCE FILE SIZE
Methodology - General Inputs RCPs GCMs Observations Downscaling BCSD BCCAQ EQM RM Future IDF Ensemble Uncertainty Analysis
Methodology – Inputs Data source: Pacific Climate Impacts Consortium Statistically Downscaled Climate Scenarios (https://pacificclimate.org/data/statistically-downscaled-climate-scenarios) Three RCPs (2.6, 4.5 & 8.5) RCP2.6 – stringent mitigation scenario to keep global warming below 2C above pre-industrial temperatures RCP4.5 – intermediate (comparable to SRESB1) RCP8.5 – very high GHG emissions (comparable to SRES A2) 12 GCMs Two Downscaling Approaches Bias Correction Spatial Disaggregation (BCSD, Wood et al., 2004) Bias Correction / Constructed Analogues with Quantile mapping reordering (BCCAQ, Werner and Cannon, 2015) Daily precipitation Gridded resolution of 300 arc-seconds (roughly 10 km) Observation – 24-hour rainfall
Methodology – Development of Future IDFCurves Equal Quantile Matching (EQM) Ratio Method (RM) (Srivastav, et al., 2014)
Methodology – Uncertainty Analysis Compare Results Using Spread = (95th%tile – 5th%tile) / Ensemble Mean Uncertainty = Standard Deviation / Ensemble Mean Total Uncertainty = Resampling vs. Central Limit Theorem
NITCHEQUON Study Area & Periods NITCHEQUON, QC Baseline Period: 1968 - 1985 Future Period: 2021 – 2050
Results –2030s IDF Curves by Individual GCMs Different GCMs could result in the projected 100-yr 24-hour rainfall in 2030s varying by a factor of 3. Individual GCMs could bring extreme uncertainties depending the GCM selected.
Results – Sub-ensemble 2030s IDF Curves Different combinations of RCP, GCM, and downscaling could result in the projected 100-yr 2030 rainfall depth varying by a factor of 2. RCP8.5 does not necessarily result in highest projections. EQM+BCCAQ result in the highest projections – double quantile matching / re-ordering might be an overkill. RM+BCSD result in the lowest projections – using simple ratio and bias correction approaches might miss some of the extreme rainfall events
Results – Spreads within Sub-ensemble Groups Different combinations of RCP, GCM, and downscaling approaches could result in projected 2030 rainfall spread over 1.5. EQM+BCCAQ result in the highest spreads, double quantile might be overkill RM+BCSD result in the lowest spreads, might miss some of the extreme rainfall events
Results – Uncertainties within Sub-ensemble Groups Different combinations of RCP, GCM, and downscaling approaches could result in projected 2030 rainfall uncertainty over 60%. EQM+BCCAQ result in the highest uncertainty, double quantile might be overkill RM+BCSD result in the lowest uncertainty, might miss some extreme rainfall events
Results – Selected Ensemble Members Selected ensemble members include; All three RCPs All 12 GCMs EQM+BCDS & RM+BCCAQ Uncertainties are in the range of 20-30% for different return periods
Results – Uncertainty Analysis Propagation of Uncertainty (Clifford, 1973) For a scalar-valued function The uncertainty for f is expressed as Where ij is the correlation coefficients The second term can be ignored with independent variables xi Can we combine individual uncertainties to quantify the total uncertainty assuming independence?
Results – Uncertainty Analysis Combining individual uncertainties assuming independence apparently underestimates the uncertainty due to correlations The central limit theorem (CLT, Rice 1995) establishes that the sum tends toward a normal distribution even if the original variables are not normally distributed when independent variables are added. The chart on the right side shows the dependence of all variables. Therefore, using ensemble approach is a better way to quantify the uncertainty and risk.
Discussions The projection of future extreme rainfall events could vary significantly depending on the selection of input data (RCPs, GCMs), therefore, the future projections are less meaningful without proper uncertainty estimates. Downscaling approaches are necessary due to the limitations of GCMs projections (e.g., spatial and temporal resolutions). However, selection of downscaling approaches needs to be carefully examined to minimize the uncertainty. When combing the individual uncertainties to quantify the total uncertainty, the dependence of different combinations / scenarios has to be considered. Using resampling technique should be a better choice to estimate the exceedance probability, instead of CLT. Using ensemble approaches will make the results more robust. The extreme value distributions could add another dimension to the total uncertainty, which has not been considered in this study.
References CCSP, 2008. Weather and Climate Extremes in a Changing Climate. Regions of Focus: North America, Hawaii, Caribbean, and U.S. Pacific Islands. A Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. Clifford, A A,1973. Multivariate error analysis: a handbook of error propagation and calculation in many-parameter systems. John Wiley & Sons. IPCC, 2014. Fifth Assessment Synthesis Report Climate Change 2014 Synthesis Report Long Report. Adopted 1 November 2014. Liu, H, Donahue, PE and J Lee, 2015. Quantifying Uncertainties in the Process of Developing Future IDF Curves. A&WMA Addressing Climate Change, Oak Brook, IL, September 2015. Rice, J, 1995.Mathematical Statistics and Data Analysis (2nd ed.). Duxbury Press. ISBN0-534-20934-3. Srivastav, R.K., Schardong, A. and S.P. Simonovic, 2014, Equidistance Quantile Matching Method for Updating IDF Curves under Climate Change, Water Resource Manage., 28: 2539-2562, DOI 10.1007/s11269-014-0626-y. Werner, A.T. and A.J. Cannon, 2015: Hydrologic extremes – an intercomparison of multiple gridded statistical downscaling methods. Hydrology and Earth System Sciences Discussion, 12, 6179-6239, doi:10.5194/hessd-12-6179-2015. WilbyRL, SP Charles, E Zorita, B Timbal,, P Whetton and LO Mearns, 2004, Guidelines for Use of Climate Scenarios Developed from Statistical Downscaling Methods, Intergovernmental Panel on Climate Change. Wilks, D.S., Statistical Methods in the Atmospheric Sciences, Academic Press, 1995. Wood, A.W., L.R Leung, V. Sridhar, and D.P. Lettenmaier, 2004: Hydrologic implications of dynamical and statistical approaches to downscaling climate model outputs.Climatic Change, 62, 189–216, doi:10.1023/B:CLIM.0000013685.99609.9e.
Christopher_Davidson@golder.com Thank You!