Real-time hydro-meteorological forecasting in the upper Po river basin

Transcription

1 Real-time hydro-meteorological forecasting in the upper Po river basin A. Ceppi 1, G. Ravazzani 1, A. Salandin 2, D. Rabuffetti 2, M. Mancini 1 Roma, 22 Marzo 212 1) Politecnico di Milano D.I.I.A.R. (Dipartimento di Ingegneria Idraulica, Ambientale, Infrastrutture viarie e Rilevamento), Milano, Italy 2) ARPA Piemonte, Torino, Italy

2 Aims and tools of the study Goal Assessing the reliability of a real time flood forecasting system, coupling meteorological and hydrological models, analysing the quantitative forecasting precipitation and temperature fields over mountain basins in different weather conditions Tools Set up an hydro-meteorological chain as an operative real time flood forecasting tool in mountain basins: MAP-D-PHASE Project Evaluation of the forecasted atmospheric forcing errors: in particular the key role of air temperature that can affect the Quantitative Discharge Forecast (QDF) and the whole hydro-meteorological alert system in the Alpine region 2

3 MAP-D-PHASE Project A Forecast Demonstration Project (FDP) of the WWRP (World Weather Research Programme of WMO). It aims at demonstrating some of the many achievements of the Mesoscale Alpine Programme (MAP), in particular the ability of forecasting heavy precipitation and related flooding events in the Alpine region. The MAP FDP addresses the entire forecasting chain ranging from limited-area ensemble forecasting, highresolution atmospheric modelling, hydrological modelling and nowcasting to decision making by the end users, by setting up an end-to-end forecasting system. The D-PHASE Operations Period (DOP) has been from 1 June to 3 November 27, but it is still working! The real-time end-to-end forecasting system includes: centralised Visualisation Platform (VP, and Data Archive (DA, cera- Source: Ravazzani, G., Ceppi, A., Rabuffetti, D., Mancini, M. 29; dphase_fest: hydrological model FEST run by Politecnico di Milano for the MAP D- PHASE project. World Data Center for Climate. [doi: /WDCC/dphase_fest] Analyzed watersheds by POLIMI: the Toce, Maggia, Ticino D-PHASE hydro-meteorological warning codes Visualisation platform for the Map-D-Phase No data No alert 6 times a year Twice a year Every 1 years 3

4 Area of study 1 km 11 BASINS Lag time [h] Area [km 2 ] Ticino basin (closed at Bellinzona) Maggia basin (closed at Solduno) Toce basin (closed at Candoglia) Sesia basin (closed at Palestro) Po basin (closed at Carignano) Stura basin (closed at Fossano) Tanaro basin (closed at Farigliano) Belbo basin (closed at Castelnuovo) Bormida basin (closed at Cassine) Orba basin (closed at Casal Cervelli) Scrivia basin (closed at Serravalle)

5 The POLIMI hydro-meteorological chain: the forecasting cascade system Weather forecasts are useful to predict possible extreme hydrological events, in order to active in advance mitigation measures and alert systems, above all over small-medium size mountain basins where lag times are low. In the present day, operational real time hydro-meteorological forecast systems are realized by use of one-way coupling, i.e. the meteorological output variables are driven into hydrological models. The Hydro- Meteorological chain Observations: Weather and Hydro data Local M ~al Modelling Basin H ~al Modelling Quality Control ARPA Piemonte COSMO-LEPS FEST-WB Skill scores Meteo Swiss MOLOCH Statistical analysis The initial hot start is sent daily by ARPA-Piemonte which runs the same hydrological model with weather observations. The current hydro-meteorological chain includes: a) probabilistic forecasts based on ensemble prediction systems with lead time of a few b) short-range forecasts based on high resolution deterministic atmospheric model The hydrological model used to generate the runoff simulations is the FEST-WB model, developed at Politecnico di Milano. Statistical analyses are used to calculate the skill scores for hydrological applications 5

6 Meteorological models: Cosmo-Leps and Moloch COSMO-LEPS Model (Marsigli et al., 25) Spatial Resolution: 1. km (.9 ) Temporal Resolution: 3 h Vertical levels: 4 (non-hydrostatic) Ensemble members: 16 nested on ECMWF EPS Forecast range: +132 h Run starting at: 12: UTC Owner: ARPA Emilia-Romagna MOLOCH Model (Malguzzi et al., 26) Different spatial resolutions used by the two weather models over the Maggiore Lake basin: a temperature field on 27 November 27 is shown in Celsius degrees Spatial Resolution: 2.3 km (.2 ) Temporal Resolution: 1 h Vertical levels: 5 (non-hydrostatic) Deterministic model, nested on BOLAM, nested on ECMWF Forecast range: +48 h Run starting at: : UTC Owner: ISAC-CNR 6

7 PoliMi - Distributed Hydrological Model: FEST-WB Atmospheric forcing Spatial interpolation: Thissen, IDW root zone transmission zone saturated zone The snow model includes the snow melt and the snow accumulation dynamics 1) snow accumulation dynamic Snow Model Drainage Routing: Linear reservoir Soil Water balance Outflow hydrograph Surface Runoff Routing: Muskingum Cunge Dinamic equation Full scheme of the rainfall-runoff distributed hydrological model FEST-WB, physically based (Mancini,199 Ravazzani et al., 21) Soil parameters Vegetation parameters DEM River network definition M The partitioning of total precipitation, P, in liquid, Pl, and solid, Ps, phase is a function of air temperature,ta (Tarboton et al., 1994). T inf and T sup are calibrated parameters (Corbari et al., 29). The snow melt simulation is based on the degree day concept (Martinec et al., 196) ( Ta Tb) 7 s = C m 1 T inf T sup 2) snow melt dynamic C C m is an empirical coefficient depending on meteorological conditions and geographic location The melt rate in ms -1, M s, is proportional to the difference between air temperature T a and a predefined threshold temperature, T b (Salandin et al., 24)

8 1-5 November 28: stratiform event Palestro South-East winds Radar accumulation (mm) over Piemonte and South Switzerland Days Toce Ticino Maggia Sesia Po Stura Tanaro Belbo Bormida Orba Scrivia 1 November November November November November Cumulated Precipitation [mm]

9 The role of atmospheric forcing: precipitation 9 Q [m 3 s -1 ] Discharge h False Alarm h Missed Alarm BS November output run: h before the main peak flow Sesia basin BS for alert code.14 Cumulated Precipitation [mm] 3 25 Precipitation ini 1 nov ini 2 nov ini 3 nov ini 4 nov Brier Score Lead Time 1 BS = n ( y k o k n = number of forecasting istances k = 1 y k = the probability that an event was forecasted o k = the actual outcome of the event at instance k ( if it doesn't happen and 1 if it happens) Best score = The BS is essentially the mean-squared error of the probability forecasts, considering that the observation is o=1 if the event occurs and o= if the event does not occur. The score averages the squared differences between pairs of forecast probabilities and the subsequent observations (Wilks, 26). n 2 ) 9

10 The role of atmospheric forcing: precipitation Discharge 4 November output run: h before the main peak flow Q [m 3 s -1 ] Sesia basin.77 BS for alert code Cumulated Precipitation [mm] Precipitation BS Hit ini 1 nov ini 2 nov ini 3 nov ini 4 nov Lead Time min-max ensemble FEST-WB Observed Median Q25-Q75 1

11 The reliability of the hydro-meteorological chain: brief summary Sesia basin November event BS for alert code hours before the main peak Hit Toce basin November event BS for alert code hours before the main peak Why does not this reliability subsist over Toce (and Stura) basins? False Alarm 11

12 The role of atmospheric forcing: temperature Discharge 4 november Temperature Q [m 3 s -1 ] Temperature [ C] Cumulated Precipitation [mm] Precipitation Toce basin Which is the interaction between forecasted temperature and precipitation errors affecting the peak discharge in a mountain basins during Autumn/Winter period? 12

13 Sensitivity analysis at finite changes: the key role of temperature Q [m 3 s -1 ] FEST-WB Observed Median Q25-Q75 Observed Forecasted Precipitation Forecasted Observed Temperature f = fp + ft + fp, T Source: Borgonovo, Q max observed = 916 m 3 s -1 Q max FEST-WB = 992 m 3 s -1 median(q max P) = 966 m 3 s -1 median(q max T) = 1678 m 3 s -1 Finite changes sensitivity indices f median(q max P and T) Q max FEST-WB = +849 m 3 s -1 f P f T median(q max P) Q max FEST-WB = -26 m 3 s -1 median(q max T) Q max FEST-WB = +686 m 3 s -1 median(q max P and T) = 1841 m 3 s -1 f P, T marginal effects of interaction = +189 m 3 s -1 13

14 Effects of temperature on the peak discharge: quantification of errors Which is the acceptable temperature error in the discharge forecast over mountain basins? Set the precipitation input using the observed precipitation field (raingauges) I calculated the Q differences, changing the temperature input only, keeping the other hydrometeorological variables unchanged (observed precipitation and initial conditions) Quantify discharge errors 14

15 Effects of temperature on flood contributing area: Toce basin Z Altitude [m a.s.l.] Q [m 3 s -1 ] % 1% 2% 3% 4% 5% 6% 7% 8% 9% 1% Toce basin Area [%] Snow limit: m a.s.l. Ipsographic curve A SWE [mm] SWE sim SWE sim (+.5) SWE sim (+1.) SWE sim (+1.5) SWE sim (+2.) SWE sim (+2.5) Evaluation Discharge error +.5 C +1. C +1.5 C +2. C +2.5 C T [ C] +8% +16% +23% +3% +39% Error 15

16 Effects of temperature on flood contributing area: Sesia basin Z Altitude [m a.s.l.] Q [m 3 s -1 ] % 1% 2% 3% 4% 5% 6% 7% 8% 9% 1% Area [%] Sesia basin Snow limit: m a.s.l. Ipsographic curve A SWE [mm] SWE sim SWE sim (+.5) SWE sim (+1.) SWE sim (+1.5) SWE sim (+2.) SWE sim (+2.5) Evaluation Discharge error +.5 C +1. C +1.5 C +2. C +2.5 C T [ C] +1% +3% +5% +5.8% +6.1% Error

17 Conclusions 1) The hydro-meteorological chain is a very useful tool to predict in real time (generally with hours before the main peak discharge) possible river floods in advance over mountain basins, where lag times are generally lower. 2) Precipitation is not the only atmospheric forcing to be considered. The quantitative discharge forecast (QDF) is influenced by temperature errors and it is related to the basin ipsographic curve, therefore to the percentage of area that contributes with more liquid water (rain) over watershed. 17

18 Last flood in Milan urban area: the river Seveso 16 September 21 output run: h before the main peak flow Area: 195 km2 2 Cumulated Precipitation [mm] 2 Discharge Seveso basin 1 5 : 2 / 9/ : 19 / 9/ 2 1 : 19 / 9/ : 1 18 / 9/ 2 1 : 18 / 9/ 2 1 Q [m3s-1] 15 Precipitation 15 5 Ensemble Emean Forecast Reliability: 37.5% 22/9/21 : 21/9/21 12: 21/9/21 : 2/9/21 12: 2/9/21 : 19/9/21 12: 19/9/21 : 18/9/21 12: 18/9/21 : 17/9/21 12: 17/9/21 : 16/9/21 12: E median P observed Q sim Fest-WB alarm code Ensembles 18