A. Amengual, R. Romero, M. Vich and S. Alonso

1. Universitat de les Illes Balears Inclusion of potential vorticity uncertainties into a hydrometeorological forecasting chain: Application to a flash-flood event over Catalonia, Spain A. Amengual, R. Romero, M. Vich and S. Alonso Grup de Meteorologia, Departament de Física, Universitat de les Illes Balears, Palma, Mallorca, Spain e-mail: arnau.amengual@uib.es

2. Inclusion of potential vorticity uncertainties into a hydrometeorological forecasting chain: Application to a flash-flood event over Catalonia, Spain • General aspects of flood events in the Western Mediterranean area • Hydrometeorological forecasting chain • The “Montserrat” flash-flood event • Hydrological and meteorological tools • Probabilistic versus deterministic QPFs: Inclusion of Potential Vorticity uncertainties • Probabilistic versus deterministic QPFs and driven runoff simulations • Conclusions and further remarks

3. 1. General aspects of flood events in the Western Mediterranean area Floods are hydrometeorological events: depend on both hydrological and meteorological factors 1. Meteorological conditions: 1.1 Incursion of cold air masses: increase in the probability of heavy precipitations 1.2 Closed and warm sea: - surrounded by coastal high mountains - extreme rainfalls favoured by high SST • Convective instabilities along the cold fronts producing interactions between frontal and orographic enhancement

4. 1. General aspects of flood events in the Western Mediterranean area • 2. Hydrological conditions: • 2.1 antecedent moisture of the soil: infiltration properties • 2.2 terrain and surface runoff characteristics: mountain systems and high urbanization rates near the coast • Flash-floods are often associated with quasi-stationary convective events: high precipitation rates + periods of several hours • These are experienced in urban areas frequently in time and randomly in space • Short temporal scale of these episodes (~ hours): occurrence too rapid to attempt damage mitigation • Flash-floods are dangerous in terms of human lives and properties

5. Hydrometeorological modeling chain: spatial and temporal scales of the numerical weather predictions • GCMs / NWP (~103-104 km2, several days) LAMs / NWP (~10 km2, 48 h) HMS 2. Hydrometeorological forecasting chain • Mesoscale models provide realistic rainfall distributions for heavy precipitation episodes • Supply a useful support for flood simulation and forecasting: further extension lead times • Coupled atmospheric-hydrologic system: advanced validation tool for the simulated rainfall amounts

6. 2. Hydrometeorological forecasting chain • Sources of uncertainty and limitations • Errors in initial and boundary conditions • Approximations in physical parameterizations • Models’ structures • impacts on the hydrometeorological simulations

7. Synoptic situation 1. Entrance of an Atlantic low- level cold front and an upper- level trough 2. Generation of a mesoscale cyclone in the Mediterranean Sea 3. Advection of warm and moist air toward Catalonia from the Mediterranean Easterly flow + Atlantic front + orographic enhancement quasi-stationary convective system 9 June 2000 00 UTC 10 June 2000 00 UTC 3. The “Montserrat” flash-flood event

8. 3. The “Montserrat” flash-flood event • Torrential precipitation took place on 10 June 2000 from 00 to 06 UTC: hourly quantities above 100 mm and 6 h maximum up to 224 mm • The Llobregat catchment is a medium-size basin with an area of 5040 km2 and a length close to 170 km. Accumulated rainfall reached over 200 mm inside this basin

9. 3. The “Montserrat” flash-flood event • Huge increase in the river flow, producing 5 fatalities, 500 evacuated people and material damage estimated at over 65 M€ • Return period: • Q= 1550 m3s-1 25 yrs

10. 4. Hydrological and meteorological tools • MM5 model set-up • Meteorological simulations have used the same model configuration as in the real-time operational version at UIB (http://mm5forecasts.uib.es) • Initial and boundary conditions: ECMWF forecasts (update 6h, 0.3º) • Two domains: 22.5 and 7.5 km, interacting with each other and 30 vertical σ-levels • Kain-Fritsch scheme is usedtoparameterise convectionfor both domains • The experiments consider a 54h period simulation (09/06/00 00 UTC - 11/06/00 06 UTC) • HEC-HMS model set-up • Loss rate: Soil Conservation Service Curve Number (SCS-CN) model • Transform: SCS Unit Hydrograph model • Flow routing: Muskingum method • The experiments consider a 96h period simulation (09/06/2000 00 UTC-13/06/2000 00 UTC)

11. Montserrat event NSE EV (%) EP (%) rain-gauges 0.84 1.1 -7.8 control -0.05 -99.7 -99.7 5. Probabilistic versus deterministic QPFs Spatial distribution of accumulated precipitation for the control simulation Difficulties to correctly forecast precise location and timing of convectively-driven rainfall system affecting small and medium size basins

12. 5. Probabilistic versus deterministic QPFs: Inclusion of Potential Vorticity uncertainties Initial conditions of the upper-level precursor trough Slight perturbations on the initial state: study of the spatial and temporal uncertainties of QPFs into a medium-sized catchment • To introduce realistic perturbations in the ensemble prediction system (EPS), a PV error climatology (PVEC) has been carried out. This allows to perturb the PV fields using the appropriated error range

13. 5. Probabilistic versus deterministic QPFs: Inclusion of Potential Vorticity uncertainties • The PVEC is calculated using a large collection of MEDEX cases (19 cases, 56 days of simulation; further information available at: http://medex.inm.uib.es/), and provides displacement and intensity errors of the PV fields in the study region • The displacement error (DE) corresponds to the displacement of the ECMWF 24 h forecast PV field showing local maximum correlation with the ECMWF analyses PV field Percentile levels of displacement errors at 300 hPa

14. 5. Probabilistic versus deterministic QPFs: Inclusion of Potential Vorticity uncertainties • Intensity error (IE) corresponds to the difference between the displaced ECMWF 24 h forecast and analyses PV fields • PVEC is used to implement the EPS by randomly perturbing the PV field • Perturbations are applied along the areas with the most intense PV values and gradients • PV Inversion Technique consistently perturb the mass and wind fields, but conserving the energetic balance Percentile levels of intensity errors at 300 hPa

16. 5. Probabilistic versus deterministic QPFs • MM5 ensemble comprises 21 elements (control + 20 PV-perturbed) • Important spread on rainfall values • Highly sensitive to PV perturbations • Essential role of atmospheric dynamical forcing Ensemble mean (in mm, shaded) and standard deviation (in mm, continuous line at 5 mm interval)

17. 6. Probabilistic versus deterministic QPFs driven runoff simulations • Elements of the driven runoff ensemble are considered equally-like • Cumulative distribution functions (CDFs) of driven runoff peak flows

18. 7. Conclusions and further remarks • MM5 control simulation is deficient for the Montserrat event: maximum precipitation amounts are obtained quite far away from the Llobregat basin • PQPFs reduce biases obtained for the deterministic forecast • For civil protection purposes, a hypothetical first warning for a peak flow exceeding • Qp (T = 5 yrs) would have produced a probability of exceedence over 0.25. This • fact points out the benefits of a probabilistic versus a deterministic prediction system • The performance of the hydrometeorological simulations strongly depends on the initial and boundary conditions of the databases and on the case under study • References: • Amengual et al. (2009): Inclusion of potential vorticity uncertainties into a hydro-meteorological forecasting chain: Application to a medium size basin of Mediterranean Spain. Hydrol. Earth Syst. Sci., 13, 793-811