Use of Meteorological Forecast Data and Products as Input into Hydrological Models Jožef Roškar, Enviromental Agency of the Republic of Slovenia Branka Ivančan-Picek, Meteorological and Hydrological Service of Croatia Regional Workshop on Hydrological Forecasting and Real Time Data Management 11 – 13 May 2009, Park Hotel, Dubrovnik, Croatia
39
Embed
Use of Meteorological Forecast Data and Products as Input into Hydrological Models Jožef Roškar, Enviromental Agency of the Republic of Slovenia Branka.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Use of Meteorological Forecast Data and Products as Input into Hydrological Models
Jožef Roškar, Enviromental Agency of the Republic of SloveniaBranka Ivančan-Picek, Meteorological and Hydrological Service of Croatia
Regional Workshopon
Hydrological Forecasting and Real Time Data Management
11 – 13 May 2009, Park Hotel, Dubrovnik, Croatia
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Hydrological Model
Input = Precipitation (Air temperature,
Evapotranspirationetc.)
Input = Precipitation (Air temperature,
Evapotranspirationetc.)
Matematical description of complex hydrological systemincluding characteristics of the watershed,
evapotranspiration, infiltration etc.
Matematical description of complex hydrological systemincluding characteristics of the watershed,
evapotranspiration, infiltration etc.
Output = Discharge(Soil Moisture, etc.)
Output = Discharge(Soil Moisture, etc.)
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Rainfall Estimation – Example from Nile Forecast Centre
IR/T ~ CCD ~ R
Method useful only in areas with predominant convective raifall
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Accumulated rainfall 29 March 2009 06 UTC – 30 March 2009 06 UTC estimated by Radar and observed rainfall at some stations (figures)
Example of by radar estimated rainfall
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Accumulated rainfall 29 March 2009 06 UTC – 30 March 2009 06 UTC estimated by NMM over Slovenia, model run start at 29 March 00 UTC and observed rainfall at some stations (figures)
LAM - Example
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Regression models are used in some Regression models are used in some NMHS’s using observed precipitation NMHS’s using observed precipitation at some stations.at some stations.
Some of them cooperate with JRC in Some of them cooperate with JRC in the EFAS (European Flood Alert the EFAS (European Flood Alert System) project, designed for System) project, designed for simulation of rainfall-runoff processes simulation of rainfall-runoff processes in in large large catchments (Danube, Drava, catchments (Danube, Drava, Sava).Sava).
FactsFactsMajority of present countries doesn’t use hydrological conceptual or Majority of present countries doesn’t use hydrological conceptual or dynamic modells for a real time hydrological forecast.dynamic modells for a real time hydrological forecast.
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
To overcome the existent deficiencies and to improve the flood warning capabilities To overcome the existent deficiencies and to improve the flood warning capabilities in the Sava Basin the “Sava Project” - Development and Upgrading of in the Sava Basin the “Sava Project” - Development and Upgrading of Hydrometeorological Information and Forecasting System for the Sava River BasinHydrometeorological Information and Forecasting System for the Sava River Basin[Albania, Bosnia and Herzegovina, Croatia, Montenegro, Slovenia and Serbia] [Albania, Bosnia and Herzegovina, Croatia, Montenegro, Slovenia and Serbia] developeddeveloped
FactsFacts
There is relatively scarcenetwork of real-time rainfallstations for flash-floodwarning
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Numerical Weather Prediction Models - NWP
Weather Observing System
Weather Observing System
Data AnalysisData Asimilation
Initialization
Data AnalysisData Asimilation
Initialization
Integration
Atmospheric PhysicsSurface Physics
Surface Processes
Numerical Methods
Simulated FieldsSimulated Fields
PostprocessingVizualization
PostprocessingVizualization
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Numerical Weather Prediction Models - NWP
Simulated/Predicted
AthmosphericVariables/Fields
Initial andBoundaryConditions
(data archive)
Integration
Numerical resolving of mathematical equations describing development of athmospheric variables in time
Air PressureWindTemperatureHumidityCloudinessPrecipitationEvaporationSoil Moisture
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Real continous space is in a model presented in equidistant discrete grid formSpace resolution defines the smallest structures seen by a
NWP:
Low resolution (~200 km) – simulation of basic structures (planetary waves, big frontal systems) – used for climate modeling and studies of global mechanisems;
Medium resolution (50 - 10 km) – simulation of sinoptic and mesoscale systems – used for general weather forecast;
High resolution (< 10 km) – simulation of local systems (wind, fog, tunderstorms, etc.)
Regardless the resolution, there are Global models covering entire globe and Limited Area Models simulating weather over choosed
smaller area
Numerical Weather Prediction Models - NWP
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Very important over montainous areas
Relief presentation depends on horizontal resolution
Example: slope and precipitation:
Wind
Numerical Weather Prediction Models - NWP
This is why majority of NWP models underestimate precipitation
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Numerical Weather Prediction Models - NWP
• To run a LAM, access to global archives of weather patterns is needed• Data in regular grid, internally consistent, without errors (+)• Not directly related to actual situation “on ground” (-)
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
DATA SOURCE:
European Centre for Middle-range Weather Forecasts (ECMWF)
Available data sets:
1. Operational data set;2. ERA-Interim - Daily re-analyses of weather patterns for last 20 years (1989 – 2008)3. ERA-40 - Daily re-analyses of weather patterns for time period (1957 – 2001)
Numerical Weather Prediction Models - NWP
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Limited Area Model
Main Idea:
•Take data from global archive •Choose area and grid points •Re-simulate weather patterns from global model to obtain more details on a regional scale
Numerical Weather Prediction Models - NWP
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
(Non)Predictability
Physical Laws Co-existence of various scales Interaction of all variables Exchange of energy among various scales
Chaos
Numerical Weather Prediction Models - NWP
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Numerical Weather Prediction Models - NWP
(Non)Predictability Physical Laws
Co-existence of various scales Interaction of all variables Exchange of energy among various scales
Discretization of continous space Limited computing power Incomplete knowledge of initial state
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Numerical Weather Prediction Models - NWP
Uncertainty
Taking into account (Non)Predictability, we have to consider that by the model simulated parameters and fields are not directly related to actual parameters, in particular to the situation “on ground”.
How to reduce uncertainty of the NWP products?
1. to choose the model and setup it in the way that the output best fits the observations (might be difficult for precipitation!);
2. to calibrate the hydrological model with the model simulated precipitation (problem of distribution in space);
3. to use multi-model or assembley prediction (extended streamflow prediction);
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
LAM – Example: Flood in Železniki on
18 September 2007
WRF-ARW, Horizontal resolution app. 1 km Acc. Precipitation 20070918-19,06-06
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
LAM – Example: Flood in Železniki on
18 September 2007
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
LAM – Example: Flood in Železniki on
18 September 2007
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
LAM – Example: Flood in Železniki on
18 September 2007
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
LAM – Example: Flood in Železniki on
18 September 2007
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
Simulation using operational data fromECMWF
Usual process in operational weather forecasting Does it have any potential for flood and drought monitoring?
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
In addition to short middle and/or long term forecasting the model could be used as an analytical tool.
Goal:
To re-compute re-analyses data over limited area in dense grid to obtain “model climatology”for flood and drought situations interpretation.
ECMWF ERA – Interim1989 - 2008
Limited Model IntegrationArea
Limited Area ModelNNM (NCEP)
Application of NWP for flood and drought monitoring
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood/drought monitoring
LimitedArea
NMM (NCEP)Non-Hydrostatic
Meso-scaleModel
• Area: 461 x 289 x 92 = 12.257.068 points (133.229 points “on ground”)• Top Level: 2 hPa (~ 60 km)• Horizontal resolution: 8.5 km• Vertical levels: 91• Integration Time: 36 h• Time Step: 30 sec.• No. days in re-analyse: 7305
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
HorizontalResolution – Grid density
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
OUTPUT:
Simulated and averaged variables (air and soil) – daily aggregates 1989-2008
Application of NWP for flood and drought monitoring
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
FLOOD/DROUGHT RELATED VARIABLESSoil moisture?Water balance?Temperature?Evapotranspiration?
FLOOD/DROUGHT RELATED TIME SCALE Not daily! Decade?
FLOOD/DROUGHT RELATED INTERPRETATIONNot absolute values, deviation from normals,percentils …
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
STATISTICS:
Model climatologybased on ERA – Interim (1989 – 2008) re-analyses
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
STATISTICS:
Model climatologybased on ERA – Interim (1989 – 2008) re-analyses
Application of NWP for flood and drought monitoring
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
STATISTICS:
Model climatologybased on ERA – Interim (1989 – 2008) re-analyses
Application of NWP for flood and drought monitoring
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
STATISTICS:
Historical 40 % Percentile of Soil moisture index of upper 10 cm layer averaged over 10 days
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
POSSIBLE
PRODUCT:
Anomaly of mean 10 days mean temperature, based on re-analyses statistics
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Application of NWP for flood and drought monitoring
POSSIBLE
PRODUCT:
Accumulated Water Balance from 20 Februaryto 30 April 2009In percentil classes
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009
Conclusions:• With growing computer power and latest generation of
atmospheric modells uncertainty of the NWP products gradually reduces;
• Uncertainty of the simulated precipitation amounts over a limited area is reducing with growing horizontal grid density, but in the same time it increase in term of spatial distribution (important specially for flash flood warning);
• In spite of NWP outputs uncertainty, use of NWP outputs is the best we can use for the hydrological forecasting;
• Taking into account the fact that a certain uncertainty of the NWP output used as input into the hydrological model multiplies the uncertainty of hydrological model outputs by more than 1.5, one might consider to calibrate the hydrological model by NWP outputs – use of re-analyses;
• NWP model climatology based on re-analyses is useful for flood and drought monitoring.
Regional Workshop on HFRTDM, Hotel Park, Dubrovnik, 11 - 13 May 2009