The project is called DAUFIN (Data Assimilation within a Unifying Modeling Framework for Improved River Basin Water Resources Management). The DAUFIN project 
is part of the Fifth Framework Research Program of the EC. The focus of this project is a new watershed modelling approach developed by Paolo Reggiani (in his 
Ph.D. work at the University of Western Australia), Majid Hassanizadeh and Sivapalan (also from U. Western Australia ). This modelling approach has been well 
received in the hydrologic community and this project is an important step towards demonstrating the practical applicability of this approach. The main involvement 
of the TU Delft is the programming and testing of the REW-concept. The program Flowsim will also be applied. This simulation program has been, developed at the 
TU Delft for physically based Rainfall-Runoff modelling. The project DAUFIN started in April 2000 with a kick-off meeting in the town Nardin in Ardenne, Belgium 
(in the catchment of the Ourthe river). From the TU Delft Majid Hassanizadeh presented 'A new approach for rainfall-runoff modelling of catchments; REW-approach' 
and Tom Rientjes 'Optimum use of Hydrological state information in Physically-based runoff models'. More information can be found at the Daufin Homepage 
at the Wageningen University. 
            

The final report summarizes the main results of the 20 month R&D project DAUFIN aiming at the development of improved tools for hydrological modeling. The tools 
concentrate on a new catchment-based modeling approach-  the Representative Elementary Watershed model - and new data assimilitation techniques. The main 
strength of the REW approach is the combination of being general and physically based and at the same time operating at the sub-catchment level. Data assimilation 
techniques have the potential to integrate observations and models in an automated way, so that automatic model calibration as well as online prediction with 
state corrections is possible. Data assimilation can be useful to parameterize and correct the REW-model, while the REW-model has a size and the numerical 
properties that are ideal for data assimilation purposes. In this way both approaches capitalize each others strengths.   
   
In this report the coding of the REW model is explained, and it is shown in detail how it is applied to the Geer catchment, a sub-catchment of the Meuse. Next, 
a number of new datat assimilation techniques, specifically suited for hydrological problems, are described. In the first place a nudging technique, aimed 
at application in 2D or 3D Richards-based models, is explained and tested. Secondly, a technique based on cross-validation for application in a wide class 
of lumped or distributed hydrological models is described. Both techniques for the automatic calibration of hydrological models is studied in the Geer and 
the Ourthe (another tributary to the Meuse). Finally a statistical correction technique is applied to the assimilation of Meteosat-based evapotranspiration 
data in a distributed hydrological model. The efficiency of the last method is tested on data from the Mehaige, Ourthe Orientale, Eau Blanche and Loison (Meuse 
sub-catchments).   
The conclusion of the project are that both the REW modeling technique as  well as the newly deveoloped data assimilation techniques appear to be very useful 
already at an early stage development. The techniques need to be further developed and tested for operational use. In the future there should be particular 
attention to develop data assimilation algorithms that can handle problems typically encountered in hydrological applications: more robust handling of 
unobservable system parts and multi-scale heterogeneity. 
            

The focus of this project is a new watershed modelling approach developed by Paolo Reggiani (in his Ph.D. work at the University of Western Australia), Majid 
Hassanizadeh and Sivapalan (also from U. Western Australia ). This modelling approach has been well received in the hydrologic community and this project 
is an important step towards demonstrating the practical applicability of this approach.