As part of the contract, the following work is planned in the Rhine catchment area:
1) Development of Multivariate Model Conditional Processor MCP (Coccia & Todini, 2011; Todini, 2008) on the consideration of spatial and temporal dependencies and variable dependencies on each other. Here, the uncertainty of the ensemble predictions is to be taken into account in the method (univariate variant see Biondi & Todini, 2018). It should be noted that the prediction length of the predictions used varies. With regard to the marginal distribution, it must be taken into account that zero values can occur during precipitation (see e.g. Biondi et al., 2021; Reggiani & Boyko, 2019). In the case of marginal distributions for water level and discharge, it must be ensured that no unrealistic quantile values result from the uncertainty distribution (see, e.g., Hemri & Klein, 2017; Hemri et al., 2015).
2) Development of parametric and non-parametric (Schefzik et al., 2013) methods for the generation of realistic ensemble progressions from the uncertainty distributions, taking into account the spatial and temporal dependencies and variable dependencies among each other.
3) Application and verification of the developed multivariate MCP for the pre-processing of precipitation and temperature for the 375 sub-areas of the hydrological model of the BfG for the Rhine basin (see above). The MCP is to be applied for daily values with and without taking into account the variable dependencies among themselves and for hourly values without taking into account the variable dependencies. The quality of the predictive uncertainties are to be compared with each other and with the quality of the "raw" ensemble predictions.
4) In order to use the pre-processed precipitation and temperature forecasts for the calculation of runoff forecasts with the hydrological model, parametric and non-parametric (ensemble copula coupling) methods are used to generate realistic spatio-temporal ensemble curves for precipitation and temperature from the predictive uncertainties and, in addition to univariate methods, to verify them with multivariate verification measures and to compare them with the "raw" ensembles.
5) Application and verification of the developed multivariate MCP for the post-processing of the water level forecasts of the BfG for 2 selected Rhine gauges. The method is to be applied to daily averages and hourly values and verified for daily averages and compared with the EMOS method currently used in operational forecasting (Hemri & Klein, 2017). With the developed multivariate MCP, the total probability of exceeding or falling below defined low water and flood thresholds for the next 7, days 8 to day 14 and the entire 14 days is to be determined and verified.
Pos. 1 Preparation, schedule of project work (content, time and resources)
Pos. 2: Development of a multivariate MCP method taking into account ensemble predictions
Pos 2.1: Creation of concept
Pos 2.2: Objektorientiertes Programmdesign
Pos 2.3: Implementation
Pos. 3 Application of multivariate MCPs for the pre-processing of low impact and temperature
Pos 3.1: Determination of predictive uncertainty of daily mean values of precipitation and temperature without taking into account the dependence of the variables on each other
Pos 3.2: Determination of predictive uncertainty of daily mean values of precipitation and temperature, taking into account the dependence of the variables on each other
Pos 3.3: Determination of predictive uncertainty of hourly mean values of precipitation and temperature without taking into account the parameter dependence of the variables on each other
Pos 3.4: Verification of predictive uncertainties
Pos. 4 Generation of realistic spatio-temporal ensemble curves for precipitation and temperature
Pos 4.1: Application of non-parametric methods to account for spatio-temporal dependencies
Pos 4.2: Development and application of parametric methods for the consideration of spatio-temporal dependencies
Pos 4.3: Verification of spatio-temporal ensemble curves for precipitation and temperature
Pos. 5 Application of multivariate MCPs for post-processing of water level forecasts
Pos 5.1: Determination of predictive uncertainty of water level taking into account the temporal dependence
Pos 5.2: Verification of the predictive uncertainties of the water level forecasts
Pos 5.3: Determination of the overall probability of threshold value being exceeded / undershot within defined time periods
Item 6 Final report
Item 7 Project meetings, final presentation
Please refer to the tender documents for a detailed description of the services to be provided.