QPE Evaluation for the 2013 Colorado Flood Paul Kucera1, Andrew Newman1, Christian Klepp2 and Jörg Burdanowitz2 Longmont 1National Center for Atmospheric Research 2University of Hamburg/Max Planck Institute for Meteorology Evacuating Jamestown residents Photo: Dennis Pierce, AP Jamestown Lafayette Jamestown Portions of I-25 closed 19 November 2014 Photo: Matthew Gurnsey Flooding in CU dorms Photo: Cliff Grassmick, Daily Camera, via AP Boulder County Aurora 1 Outline • Lifecycle of the flood event • Operational QPE estimation results (radar and satellite) • Disdrometer and storm structure QPE estimation results • QPE Impacts on streamflow prediction • Summary and future work IPWG7 November 19, 2014 2 Time-line of Flood Event • The historical flood occurred over 10-16 September 2013 Gochis et. al. (2014) 3 Northeastern Colorado Flash Floods 9-13 September 2013 ü Upper level closed Low Pressure system located over SW United States ü Funneled SW monsoon moisture over mountains and NE Colorado ü Low-level, moist SE flow created upslope flow along the foothills bringing additional moisture and rainfall to NE Colorado Boulder/Denver NWS office GOES Water Vapor and 500 hPa Heights/Vorticity Movie Loop: 9-13 September 2013 Boulder CountyAurora Courtesy of Eric Nelson, NCAR/RAL 4 NCAR’s Marshall Field Site Rain Gauge Measurements Accumulation (mm) Courtesy of Roy Rasmussen ~10 inches accumulation Intensive Period #2 Intensive Period #3 10 Sept 10 Sept 06 UTC 00 UTC 13 Sept 02 UTC Intensive Period #1 11 Sept 21 UTC 12 Sept 18 UTC Three main intensive precipitation periods were observed: - Initial convective period: 00-06 UTC Sep 10 - Second precipitation phase: 21 UTC 11 Sep – 18 UTC Sep 12 - Third precipitation phase: 18 UTC 12 Sep – 02 UTC Sep 13 5 Radar Observations (1st Wave) 00-06 UTC 10 September Marshall 6 Radar Observations (2nd Wave) 21 UTC 11 Sep – 18 UTC 12 Sep Marshall 7 Radar Observations (3rd Wave) 18 UTC Sep – 02 UTC 13 Sep Marshall 8 Event Precipitation Totals • Total Precipitation map for storm event (9-17 Sep 2013) From the Colorado Climate Center • Max accumulations - Boulder: 429.3 mm (16.9 in) - Front range 530 mm (21 in) • Max precipitation located along the foothills from S. Boulder to Fort Collins Gochis et. al. (2014) Marshall Field Site 9 Operational QPE Evaluation • A variety of operational QPE were evaluated: – – – – – Default NEXRAD Z-R “Tropical” Z-R NSSL Multi-sensor precipitation estimates (MPE) Polarimetric precipitation estimates Stage IV Typical NEXRAD Z – R Relationship Z = 300 R1.4 Boulder 60 mm Large underestimate Tropical Rainfall Z-R relationship Z = 32.5 R1.65 Boulder 230 mm Only slightly High 10 QPE Comparison • Comparisons at high quality precipitation sites • 11-13 Sep accumulation period • Underestimation of rainfall – Default NEXRAD – Dual-Pol • Overestimation of rainfall – Default Tropical Z-R Gochis et. al. (2014) 11 QPE Bias Comparison • Default Z-R and dual-pol QPE significant underestimation bias • MPE best estimates • Stage IV mixed Gochis et. al. (2014) 12 Satellite Precipitation Estimation • This study evaluated CMORPH (CPC MORPHing technique) product at the produces global precipitation analyses at 8 km, 30 min resolution for the three main periods and for the total accumulation over the 3 day period Ft Collins Boulder Marshall DIA Denver Ft Collins Ft Collins Boulder Marshall Boulder DIA Denver Marshall DIA Denver 13 Satellite Precipitation Estimation Ft Collins Boulder Marshall Erie DIA Denver CMORPH underestimation bias is 13 times lower than observations and precipitation shifted eastward compared max precipitation area 14 Why are there significant underestimation biases in QPE products? • The precipitation observed during the second and third IP’s was not “typical” of high plains convection – Efficient warm rain processes – “Tropical” like drop size distributions • Large concentration of small drops • A new storm-based Z-R relationships were based on drop size distributions observations from disdrometers located at the Marshall research field site located south of Boulder 15 (#/m3 mm1) • IP2 and IP3 DSDs differ substantially from the rest of the event • IP2 shows increase in drop number of 1 to 3 mm diameter • IP3 shows strong increase in small drops between 1 and 2 mm diameter and overall larger drop numbers IP3: large number of 1-2 mm drops IP2: increased number of 1-3 mm drops All 3454 1-min spectra outside IP2 & IP3 time periods show small increase in the 1-3 mm size drops IP2: 11 SEP 21 UTC to 12 SEP 18 UTC; 1103 minutes (spectra) IP3: 12 SEP 18 UTC to 13 SEP 02 UTC; 491 minutes (spectra) 16 Reflectivity-Rainfall Relationship 1st (Intensive Period) IP had a continental Z-R • Z-R’s for IP2 and IP3 were more typical tropical Z-R relationships • – Short and Kucera (1997) – Z=120R1.43 • • The default NEXRAD Z-R fit the DSD in IP1 The default tropical ZR is not a good fit to the observed DSD 17 IP1 Z-R • Z-R: Z=305R1.49 • Convection in IP1 typical, deeper continental convection • Default NEXRAD Z-R worked well in this initial regime 18 QPE Comparison at Marshall • • • • • • Storm total Rainfall: 285 mm DSD Z-R performed well: 5 mm overestimation NEXRAD Z-R: 118 mm underestimation Tropical Z-R: 224 mm overestimation CMORPH precipitation: 220 mm underestimation However, CMORPH agreed well for the initial deep convection system 19 QPE Impacts on Streamflow Prediction WRF-Hydro SIMULATED Streamflow from NEXRAD (Tropical Z-R) Little Thompson St. Vrain James/Lefthand Fourmile Canyon Cr. Valid: Sep 12 1:15 a.m. LT Fourmile Cr. Boulder Cr. South Boulder Cr. NCAR Seminar Oct. 4, 2013 Coal Cr. Streamflow Courtesy of D. Gochis 20 QPE Impacts on Streamflow Prediction Validating Streamflow Simulations: Tropical Z-R 21 Courtesy of D. Gochis QPE Impacts on Streamflow Prediction Modeled Streamflow at St. Vrain at Lyons Forced by: NEXRAD Z=300R^1.4 Forced by: Tropical Z-R No Flood Predicted with default Z-R Courtesy of D. Gochis 22 Future Work – Storm Type Z-R QPE • The DSD based Z-R’s could be further stratified by rain type and storm structure • Friedrich et al. (2014) estimated precipitation by storm type – identified 4 rain regimes 12 Sep 13 Sep 23 Friedrich et al. (2014) Summary • The Colorado Flood had large societal impacts on the Front Range • Unique precipitation conditions – First deep convective system was typical for Eastern plains of Colorado – Second precipitation transitioned to a more “tropical” environment – Third intensive period was “tropical like” – large concentration of small drops, shallow convection, efficient rain processes • Some operational QPE products underestimated storm totals • Knowledge of DSD provided a better QPE products • Significant impact on flood forecast and early warning – Many people were stranded in the foothills • Follow on work – Evaluate the DSD Z-R QPE over the entire domain – Stratify DSD Z-R further by storm structure – Develop tools to provide a dynamic Z-R for operational QPE products 24 Colorado Flood Photos Boulder Flooding – Table Mesa Drive below NCAR Table Mesa Facilities - Photos courtesy of R. Cifelli 25
© Copyright 2026 Paperzz