Extreme Precipitation in a Changing Climate: Ankara Case Study Middle - PowerPoint PPT Presentation

Extreme Precipitation in a Changing Climate: Ankara Case Study Middle East Technical University Graduate School of Natural and Applied Sciences Earth System Science (ESS) Interdisciplinary Program Sertaç Oruç 1 , İsmail Yücel 1, 2 and Ayşen Yılmaz 1, 3 1 Earth System Science (ESS) Interdisciplinary Program, Middle East Technical University 2 Dept. of Civil Eng., Water Resources Division, Middle East Technical University 3 Institute of Marine Sciences, Middle East Technical University

Outline  Motivation  Objectives  Methodology  Observed Precipitation Data Analyses  Projected Precipitation Data Analyses  Summary and Conclusions 2

Motivation: Climate Change Global Precipitation Change (For the last 2 decades and Projected Periods) Low (RCP 2.6) ensemble average (dark line) and High (RCP 8.5) ensemble average (dark line) and spread of ensemble members (shaded area). Values spread of ensemble members (shaded area). Values are for the model grid cell containing: 39.912 ° N are for the model grid cell containing: 39.912 ° N 32.84 ° E 32.84 ° E. 3 https://gisclimatechange.ucar.edu/inspector RCP : Representative Concentration Pathways

Motivation: Climate Change and Extreme Events Figure 1.2. Distribution of extreme events and their types in Figure 1.1. Annual count of extreme events in Turkey in the period of 1940-2017 Turkey in 2017 Annual count of extreme events in Turkey shows an increasing trend in 1940-2017 period (Climate Assessment 2017 Report, February 2018 – State Meteorological Service). During 2017 most hazardous extreme events were; heavy rain/floods (31%) , wind storm (36%), hail (16%), heavy 4 snow (7%), and lightning (4%)

Problem Statement • Climate change in Turkey has been evaluated in many different studies with its different aspects. Majority of analysis performed and the future estimation works were focused on temperature and precipitation changes which are the most important climate parameters causing the extreme events. • In the last decades, heavy rainfall and flash flooding caused various damages in Turkey; for example settlements were damaged, road transportation and vehicles are disrupted, and life was negatively affected in Ankara Objectives • To analyze the rainfall extreme value frequencies for stationary and nonstationary conditions in Ankara region, • To produce Return Levels in stationary and non-stationary conditions with observed data and future projections,. • To figure out the superiority of nonstationary and stationary models to each other, 5

Methodology and Data The methodology of precipitation analysis in this study consists of; (1) Trend analysis is carried out for observed (1950-2015) and projected data (2015-2098) (2) Projected data is disaggregated into finer scales (5 min) and then it is aggregated to next analysis time scales (10, 15, and 30 min , … ) (3) Stationary GEV (St) models are developed, return levels are derived for desired return periods considering single and multi-time periods for observed and single period for projected data (4) Non-stationary GEV (NSt) models are developed, return levels are derived for desired return periods for observed and projected data (5) Stationary and Non-stationary model results were compared • Observed Data for Ankara - 1950-2015 (State Meteorological Services) • Projected Data; Three global climate models (GCM) are used; namely HadGEM2-ES, MPI-ESM-MR and GFDL- ESM2M. These models are operated with the RCP 4.5 and RCP 8.5 emission scenarios - 2015-2098 (State Meteorological Services) 6

Methodology Rainfall Data Observed Projected Trend Analysis Obatining Blocks Disaggregation GEV Models Aggregation Obatining Blocks Stationary Nonstationary GEV Models Return Levels for Mean/Median of Return Desired Storm Levels for Desired Define Covariates Stationary Nonstationary Durations and Periods Storm Duration Return Levels for Mean/Median of Return Desired Storm Levels for Desired Define Covariates Durations and Periods Storm Duration Figure 1.3. Rainfall Data Analyses Framework 7

Observed Data Trends & Change Point 50 90 Precipitation Height (mm) Precipitation Height (mm) 45 80 40 70 35 60 30 50 25 40 20 30 15 20 10 10 5 0 0 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Years Years 1 hour 2 hours 3 hours 6 hours 5 min 10 min 15 min 30 min Figure 1.4. Sub-Hourly Time Series Trend Figure 1.5. Hourly Time Series Trend 90 Precipitation Height mm 80 70 Average 1950-1975 60 Maximum 1950-1975 50 40 Average 1976-2015 30 Maximum 1976-2015 20 Average 1950-2015 10 Maximum 1950-2015 0 5 min 10 min 15 min 30 min 1 hour 2 hours 3 hours 6 hours Storm Duration Figure 1.6. Average annual maximum rainfall intensities (mm) for sub-hourly and hourly storm durations 8

Observed Data NonStationary Stationary Models (NSt) Models (St) 10 15 30 10 15 30 5 Minutes 1 Hour 2 Hours 3 Hours 6 Hours Minutes Minutes Minutes 5 Minutes 1 Hour 2 Hours 3 Hours 6 Hours Minutes Minutes Minutes Figure 1.7. Storm Durations Used for Stationary Models 2-year 5-year 10-year 25-year 50-year 100-year 200-year Model Location Scale Shape Mean Value Change FiveMin -4% -4% -5% -9% -11% -15% -18% NStGEV1 𝜈 t = 𝛾 0 + 𝛾 1t 𝜏 (constant) 𝜊 (constant) TenMin -14% -13% -12% -9% -7% -5% -3% NStGEV2 𝜈 t = 𝛾 0 + 𝛾 1t 𝜏 t = 𝛾 0 + 𝛾 1t 𝜊 (constant) FifteenMin -1% -4% -6% -9% -12% -14% -17% ThirtyMin 0% -3% -6% -10% -13% -16% -19% NStGEV3 𝜈 (constant) 𝜏 t = 𝛾 0 + 𝛾 1t 𝜊 (constant) OneHour -7% -5% -3% 0% 3% 6% 10% TwoHours 0% -3% -4% -5% -6% -7% -7% NStGEV4 𝜈 t = 𝛾 0 + 𝛾 1temperature 𝜏 (constant) 𝜊 (constant) ThreeHours 0% -3% -5% -8% -11% -13% -16% SixHours 1% -1% -2% -3% -4% -5% -5% NStGEV5 𝜈 t = 𝛾 0 + 𝛾 1t 𝜏 t = 𝛾 0 + 𝛾 1exp(temperature) 𝜊 (constant) Median Value Change FiveMin -3% -2% -4% -7% -9% -12% -15% NStGEV6 𝜈 t = 𝛾 0 + 𝛾 1exp(temperature) 𝜏 t = 𝛾 0 + 𝛾 1exp(temperature) 𝜊 (constant) TenMin -13% -12% -10% -7% -4% -2% 0% FifteenMin -1% -2% -4% -7% -9% -12% -14% NStGEV7 𝜈 t = 𝛾 0 + 𝛾 1exp(temperature) 𝜏 t = (constant) 𝜊 (constant) ThirtyMin 1% -2% -5% -8% -10% -13% -16% OneHour -8% -5% -2% 2% 5% 8% 12% NStGEV8 𝜈 (constant) 𝜏 t = 𝛾 0 + 𝛾 1temperature 𝜊 (constant) TwoHours -1% -2% -3% -4% -4% -5% -5% ThreeHours -1% -2% -4% -7% -9% -11% -14% SixHours 0% -1% -2% -2% -3% -3% -4% Table 1.1. Non-stationary models with time and covariate (temperature) Table 1.2. Nonstationary GEV Best Fit Model Return Levels (mm) - Mean and 9 dependent location and scale parameters Median Value Change with Respect to Stationary GEV Model

Observed Data 20 30 40 70 StFifteenMin Return Level mm Return Level mm Return Level mm StFiveMin StTenMin Retun Level mm StThirtyMin 60 25 NStFifteenMin NStTenMin 15 30 NStFiveMin NStThirtyMin 50 20 40 10 15 20 30 10 20 5 10 5 10 0 0 0 0 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 Return Period (Years) Return Period (Years) Return Period (Years) Return Period (Years) 100 100 100 100 Return Level mm Return Level mm StOneHour StThreeHours Return Level mm StSixHours Return Level mm StTwoHours 80 80 80 80 NStOneHour NStThreeHours NStSixHours NStTwoHours 60 60 60 60 40 40 40 40 20 20 20 20 0 0 0 0 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 2 5 10 20 25 50 100 200 Return Period (Years) Return Period (Years) Return Period (Years) Return Period (Years) Figure 1.8. Stationary and Best Fit Nonstationary Model Return Level (mm) Comparison - Return Period vs. Return Level • The shorter the storm duration the larger the differences between the non-stationary and stationary extremes. • Among the storm durations, only one hour time series exhibit larger values for its nonstationary model return level values, however this is not valid for shorter return periods such as 5 years or 20 years • Sub-hourly storm durations indicate larger difference than hourly storm durations and non-stationary estimates are smaller than their corresponding stationary values 10

Projected Data (Non) Stationary ((N)St) 10 Minutes 15 Minutes 1 Hour 6 Hours MPI HG MPI HG MPI HG MPI-ESM-MR HADGEM2-ES (MPI) (HG) RCP RCP RCP RCP RCP RCP RCP RCP RCP RCP RCP RCP 4.5 8.5 4.5 8.5 4.5 8.5 4.5 8.5 4.5 8.5 4.5 8.5 RCP RCP RCP RCP 4.5 8.5 4.5 8.5 GFDL GFDL GFDL GFDL-ESM2M (GFDL) RCP RCP RCP RCP RCP RCP 4.5 8.5 4.5 8.5 4.5 8.5 RCP RCP 4.5 8.5 Figure 1.9. Projected Storm Durations Used for Stationary Models for 2015-2098 period 11