Three multi-scale experiments are performed within WPs 3 and 4, focusing on observations of extreme rainfall and urban flood forecasting. Different case study catchments, spatial and temporal scales and forcing data have been tested and compared. The main objective was to evaluate the performance of rainfall observations and flood models at multiple scales as well as to integrate models operating at different scales. Key issues included the following:

    • Rainfall forcing. What is the impact of (temporal and spatial) resolution in rainfall observations and forecasts? How well is extreme rainfall described in different radar observation systems? How strong is the link between rainfall intensity and urban flooding?
    • Hydrological simulation. What is the impact of land-use information in urban flood modelling? What is the value of high-resolution multi- basin hydrological forecasting in an urban context? Can large-scale (hydrological) information support local (hydro-dynamic) flood models?
    • End-user communication.Which information is relevant for which end-user? How can we increase the relevance and ”sharpness” of the information? How should ensemble forecasts be used and presented for optimum value?

The three multi scale flood forecasting experiments are:

    • Experiment 1: Urban flood risk assessment: high-resolution hydrology vs. hydro-dynamics.
    • Experiment 2: Multi-scale assessment of high-intensity rainfall events in radar observation
    • Experiment 3: Coupled multi-scale forecasting in Aalborg

Each experiment is explained below and concluded in Table 1.

Experiment 1

Objective and Method

The objective was to evaluate the correspondence between urban flood risk as estimated by a high-resolution hydrological model and a hydrodynamic model respectively. Can the hydrological model improve the urban flood risk estimation of the hydrodynamic model? Aalborg in Denmark and Helsinki in Finland were the two study locations. A hydrological model (HYPE) model and a hydrodynamic model (SWMM in Helsinki and MIKE in Aalborg) have been set up for each city, see Figure 1.

All four models have been driven with a common forcing data set consisting of 95 selected high-intensity rainfall events from the period 1979-2017. In the hydrodynamic models, simple relationships between rainfall/runoff and flood risk have been derived to establish a link between runoff-flood response and runoff estimated by the hydrological model. The estimated runoff from the hydrological and hydrodynamic models has been compared to investigate if results were in agreement, and finally the flood risk correspondence was evaluated.

Figure 1. The study cases in Aalborg in Denmark (Kaerby) and in Helsinki I Finland (Pakila).


The expected value of the results

If the high-resolution hydrological model can contribute to improved flood risk estimation, the results will be useful for urban flood risk simulation and forecasting in a multi-basin hydrological framework.

Experiment 2

Objective and Method

How well do European radars capture heavy rainfall events? Radar data and rain gauge data of the top 50 rainfall events in each of the four countries - Netherlands, Denmark, Finland and Sweden -, have been compared in order to evaluate the correspondence between radar and rain gauges in periods of heavy rainfall as a function of temporal resolution. Are there any differences between countries, radars or rain events? How does performance vary with different scales (temporal and spatial)? What are the implications for flood forecasting and hydrological modelling?

The rain events from rain gauges have been selected based on rainfall intensity. There was a dry period of at least six hours before and after each event. Radar data during the same period have been collected. All countries have had C-band radar data but the Netherlands, which has X-band radar data.

Figure 2. Radar snapshots of the top 3 events in Denmark at the time of peak (rain gauge is in the middle of the figure).

The expected value of the results

Improved knowledge about the accuracy of high-intensity rainfall in radar and rain gauge data of different resolutions, for development of flood forecasting and hydrological modeling.

Experiment 3

Objective and Method

Hydrological models are often used for flood modelling in rural areas whereas hydrodynamic models are often used in urban areas. To improve flood modelling in urban areas by taking advantages of the two model types, an integrated urban flood forecasting model has been developed. A coupled semi-distributed hydrological model (HYPE), which estimates surface runoff and discharge from rural catchments, and a hydraulic 1D-2D urban drainage model (MIKE), including 1D sewer river network and 2D overland flow, have been set up for Aalborg in Denmark. The HYPE model will produce boundary conditions of surface runoff and catchment discharge to the MIKE model for 10 high-intensity rainfall events, see Figure 3.

The flow simulation results of the integrated flood model (HYPE/MIKE) will be compared with the flow simulation results based only on the hydrodynamic model (MIKE). The 2D overland flood response will be visualized.


Figure 3. To the left is the model area of the hydrological model (HYPE) and to the right the model area of the hydrodynamic model (MIKE). HYPE produces boundary conditions to the MIKE model. Study area is Aalborg in Denmark.

The expected value of the results

When coupling a hydrological model (HYPE) with a hydrodynamic model (MIKE), the boundary conditions from HYPE to MIKE are expected to improve the urban flood modelling. 

Table 1. Short information of each joint experiment concerning objective, method and expected outcome.



Objective Method Expected outcome
1 The results of simulated urban flood risk from HYPE Aalborg (Denmark) and MIKE Aalborg are compared.


The results of simulated urban flood risk from HYPE Helsinki (Finland) and MIKE Helsinki are compared.

All four models are run with common forcing data set of 95 rainfall events from 1979-2017.


A new method is developed to derive criteria for a flood threshold in urban catchments, which is correlated to output simulated using a hydrological model. It can be used across catchments and similar combinations of hydrological and hydrodynamic models.


Assessment whether large-scale high

resolution hydrological models can to some extent substitute local

hydrodynamic models in real-time urban flood forecasting.

2 Compare the correspondence of high-intensity rainfall between radar and rain gauges as function of temporal resolution. Are there any differences between:

·        Countries?

·        Types of radar?

·        Rain events?

·        Temporal and spatial scales?


Radar data and rain gauge data from the top 50 heavy rain events in Denmark, Finland, Sweden and the Netherlands, respectively, have been compared. Increased knowledge about extreme rainfall radar and rain gauge data at different resolutions for improvement of flood forecasting and hydrological modelling.
3 Develop an integrated urban flood model of HYPE and MIKE in Aalborg, Denmark.


Flood simulation results from HYPE/MIKE are compared with simulations results from only MIKE.


Ten high-intensity rainfall events are run in HYPE and the output data are used as boundary conditions in MIKE.


Ten high-intensity rainfall events are run in MIKE without input from HYPE.

Improved flood modelling and forecasting in urban areas, with the potential to be used in many European cities. New real-time visualization of discharge and flood risk.