Multiple X-band radardata for operational use in public water utility sector - new update report

1. Climate challenges, mitigation and adaptation
Seyyed Hasan Hosseini¹ , Hossein Hashemi¹, Rolf Larsson¹, Ronny Berndtsson¹, Andreas Bengtsson², Sven Bengtsson², Sofia Dahl², Jonas Olsson³, Remco van de Beek³, Nicholas South⁴, Henrik Aspegren⁵, Simon Granath⁶, Susanne Steen Kronborg⁶, Emma Falk^6
1 Lund University/LTH
² NSVA
³ SMHI
⁴ Tyréns
⁵ Sweden Water Research
⁶ VA SYD

Abstract text
In this study, X-band Radar-data (XR) from NSVA and VA SYD XR units in Scania/Sweden (is linked together. The purpose with this report is to integrate all XR-data, from double-polarization and several levels, in the overlapping zone from the two XR facilities in Scania. Results are concluded both from an empiric linear regression model (REG) and from a data driven numeric calculation model which uses an artificial neural network (ANN) to calculate 2D rain results with XR technology. Estimated 2D rain results were evaluated with two different methods:

• Direct validation with measured values with stationary rain gauges (tipping buckets)
• Indirect validation with a runoff model for two moderate sized catchments, Ellinge and Lundåkra WWTP catchment, with rain data both from ANN and REG-models and from stationary rain gauges in the area (indirect validation)

The results from the direct validation were:

• XR can be used as a method to describe rain over time with adequate quality.
• In general, the precision was improved with data overlap from the two XR-data sets
• A linear multi-regression- and an ANN model gives similar results
• The precision of the rain product can be improved by using more than one radar level and by increasing the measurement range to 70km.

The results from the indirect validation showed that the potential of using rain data from ANN and REG- models. In the Ellinge WWTP catchment were the runoff simulations conducted in an adapted model for representation of rain data, used with improved precision in comparison with stationary rain gauges. In the Lundåkra WWTP catchment the runoff simulations were conducted in an environment which was partially adapted to the local geographic conditions, and this probably had an impact on the results. For the Lundåkra WWTP catchment, the precision during the simulation was better in comparison to the different data sets for Ellinge WWTP catchment.

The research community and WWTP utilities must improve their understanding of the runoff process with new best practice methods. Two sets of error sources were presented in the report and have an impact on the results were:

• A description of the extraneous water as a factor. To obtain a representative overview of flow levels/rates in the sewage network, both a fast and slow extraneous water component must be integrated in the runoff model. In this study, the sewage network model has integrated this component differently.
• Pumping stations- it is complicated to study the runoff results where there are pumps/basins in the sewage network which creates a pulse bias that in turn makes statistical analysis more difficult.

In conclusion, overlapping analysis with two XR datasets enables improved rain estimations and possibilities to improve the rain products for applications in the WWTP sector such as runoff simulations.