1
The Application of Robust Acoustic Disdrometers in Urban
Drainage Modelling
Ruben DAHM1*, Stijn DE JONG2,3, Jan TALSMA1, Rolf HUT2, Nick van de GIESEN2
1Deltares, Delft, The Netherlands
2Delft University of Technology, Delft, The Netherlands 3Disdrometrics, Delft, The Netherlands
*Corresponding author
Email: ruben.dahm@deltares.nl
ABSTRACT
This paper explores the application of a robust acoustic disdrometer in urban drainage modelling using the Delft-FEWS framework. We monitored rainfall over a period of 90 days and developed a standardised procedure to import the Delft-Disdrometer measurements in the FEWS platform and consequently simulate an urban drainage model. We demonstrated this workflow for the urban drainage network of Delft, The Netherlands, and compared the simulation results to radar-forced simulations. The case study shows a successful application of the robust acoustic disdrometers in urban drainage modelling and this gives rise to a range of ideas on improving the rainfall forcing onto urban drainage modelling using low-cost disdrometers.
KEYWORDS
Disdrometer, FEWS, radar, urban drainage modelling
INTRODUCTION
High resolution precipitation measurement is needed in urban areas to link rainfall to local flooding events (Berne et al., 2004). It is expected that due to climate change more intense and local rain events in urban areas will take place (IPCC, 2007). Rainfall radar can give a complete overview of a precipitation event over a city and is a common tool to use as input for operational systems monitoring and managing the urban drainage network, as shown in ISA-Hoekse Waard, The Netherlands (van Nooijen, 2011). Nevertheless, the pixel size of the most common weather radar (C-band) is often too scarce to capture the dynamics of a rainstorm (van de Beek et al., 2010, Jaffrain and Berne, 2012). Steiner et al., (2004) showed that weather radar also needs ground based measurements to calibrate the relation between backscatter and precipitation rate. Furthermore, the accuracy of rainfall estimates can be greatly improved if the weather radar is combined with ground based rain gauges (Seo and Breidenbach, 2002). Weighing gauges and tipping buckets are the most common ground based rain gauges. Because of high maintenance costs it is often not feasible to place these rain gauges in a grid that is dense enough to capture the spatial variability of a rainstorm (Pardo-Ig´uzquiza, 1998). In comparison to other rainfall measurement techniques, disdrometers have the advantage of providing the (rain) drop size distribution (DSD). Insight in the DSD and the variability of DSD is of upmost importance in different fields (Jaffrain et
al., 2011) such as numerical weather modelling (e.g. Michaelides et al., 2009), weather radar
application (e.g. Marshall and Palmer, 1948), and soil erosion (e.g. Salles et al., 2002). More insight in the variability of DSD in time and space can reduce the uncertainty affecting radar rain rate estimates significantly (Jameson and Kostinski, 2001, Jaffrain et al., 2011).
2
In recent years, Delft University of Technology developed an acoustic disdrometer. This ‘Delft-Disdrometer’ is specifically designed to be low cost (maximum of 500 euros) and low maintenance, thus allowing the installation of dense networks without incurring large upkeep costs (Bagree and de Jong, 2012). Hut (2013) describes the design, calibration and field evaluation of this disdrometer. In this paper we discuss the potential use of the acoustic Delft-Disdrometer in urban drainage modelling.
METHODS
General approach
To explore the application of robust acoustic disdrometers in urban drainage modelling we carried out a case study using the Delft-Disdrometer network available at Delft University of Technology. An urban drainage model including sewer components of the town of Delft was used to simulate and assess the application of rainfall forcing derived from Delft-Disdrometers. The workflow of the case study consisted of i) the process of data collection by the Delft-Disdrometer, ii) set up of a standardized import routine of this data to Delft-FEWS, and iii) simulation of the urban drainage system within the Delft-FEWS platform.
Delft-Disdrometer
A disdrometer is an instrument to measure the drop size distribution of falling hydrometeors. There are four main type of disdrometers; video disdrometers, optical disdrometers, acoustic disdrometers and impact disdrometers. These types of disdrometers differ in principle of measurement and price levels, ranging from $3.000-$100.000. The Delft-Disdrometer is based on the principle of the acoustic disdrometer. The basic sensing principle of the Delft-Disdrometer is that of a drum excited by raindrops. The impact of a raindrop forces a piezoelectric disk to generate an electric signal. The amount of signal energy of this electric signal depends on the size of the drop (Bagree & de Jong, 2012). All drops are recorded individually and categorized per drop size. The final output is a drop size distribution per time step. For this case study a measurement frequency of 1 minute is used. The rain rate can be estimated using the DSD. The Delft-Disdrometers were situated on the roofs of several faculty buildings of the Delft University of Technology. The university is located in the western part of The Netherlands. The disdrometers send the collected data near real time with the use of a mesh network. From August 1st to October 30th 2013, Delft-Disdrometers collected precipitation and drop-size distribution data. The network of disdrometers is calibrated with respect to a cumulative rain gauge operated by the Royal Dutch Meteorological institute in the Botanical Gardens of Delft, situated 500 meter from the case study area. The calibration constants are determined in such a way that the cumulative rainfall amount of the test period is equal for the disdrometer and the rain gauge.
Urban modelling framework
The case study area covers 3.4 km2 within the city of Delft, the Netherlands. The sewer
system collects waste water from households and industries, and also rainfall runoff from streets and roofs. As the study area is rather flat, the sewer system is mainly horizontally oriented. Consequently, gravity discharge occurs only in small parts of the system, while an extensive network of pipes and sewer pumps is maintained to transport the sewer discharge to a waste water treatment plant. As these pumps have limited capacity, this system uses more than 130 external weirs to, in case of extreme events, spill towards the open water system. In this case study we analysed the simulated spilling of the Jaffalaan external weir, see Figure 1.
3
Figure 1. Urban drainage model used in the study. Blue triangle indicates the spill location of the external weir. The green circle indicates the location of the Delft-Disdrometer. We set up a standardized procedure to import Disdrometer measurements in the Delft-FEWS platform and made it available to the Delft-FEWS community. The Delft-Delft-FEWS system is a world-wide used platform developed for flood forecasting and flood early warning. Over the years it has broaden its field of application, frequently used also in droughts forecasts and real time management of water systems. In general, it is extremely useful when dealing with different source of data and models. The platform is used to be import and pre-process data, to run hydrological and hydraulic models and show results. Given its worldwide use and acceptance, Delft-FEWS is a global standard for this kind of water information systems (Werner et al. 2013). This assures generalization of the results and tools developed in this case study. Observations recorded by any Delft-Disdrometer world-wide can easily be implemented in any Delft-FEWS platform.
The Delft-Disdrometer measurements are compared with 5 minutes radar calibrated data delivered from the ‘Nationale Regenradar’ (http://nationaleregenradar.nl/). This dataset is based on the radar measurements of the Dutch, the German and the Belgian meteorological institutes, and calibrated with the measurement networks of the Netherlands. Both data sources are used as forcing data to the urban drainage model of the city of Delft. In this study we focused on the actual set-up of a framework to feed the urban drainage model with Delft-Disdrometer data instead of a sound calibration of the models. The urban model is operated by a stand-alone FEWS system. By developing real time import routines for Delft-FEWS platform, it is possible to upgrade it and use these measurements in real time (urban) flood modelling.
The platform allows running the urban drainage model of the city of Delft under a comprehensive shell and forcing it with two sources of precipitation data: i. ‘Nationale Regenradar’ data, spatial resolution of 1km2, and ii. Point (TUDelft campus) precipitation
data collected by the Delft-Disdrometer. The ’Nationale Regenradar’ is public available as are the import routines for Delft-FEWS. The disdrometer data is delivered according to the
4
WaterML 2.0 standard (OGC, 2008). This is a generic standard for the representation of in-situ hydrological observations data. We improved and updated the Delft-FEWS general adapter to correctly validate and import Delft-Disdrometer data (including rain depth (mm) and drop size distribution, see Figure 2). Radar data is pre-processed before used in the urban drainage model. Two different processes are performed. In the first routine, the pre-processed radar data is spatially interpolated in order to define rainfall intensities fallen in each wastewater unit of Delft. In the second routine, only rainfall intensities in the cells around the disdrometer location are extracted from the radar. Their average is then copied for all the wastewater unit of Delft. The urban drainage model is run with both sets of pre-processed data. The first data set is common practice when downscaling radar data, while the second data set allows a more coherent comparison with Delft-Disdrometer data. Disdrometer rainfall intensities are uniformly copied for all wastewater units. Pre-processed rainfall data is filled with zeros at missing values, and made available for the urban drainage modelling. The urban drainage network is schematised in the integrated modelling suite SOBEK. The network is set up as a combined open channel–sewer flow 1D model including urban rainfall-runoff processes. A Delft-FEWS system runs the SOBEK model with the rainfall data. SOBEK describes cross-sectional averaged flow in a network of open and closed channels and pipes. The model framework is based upon the solution of the full de Saint Venant equations, thus including backwater and transient flow phenomena. It includes an automatic drying and pressurised procedure. In SOBEK the numerical solution is based upon an implicit formulation on a staggered numerical grid. On the staggered grid the dependent variables Q (discharge) and ζ (water level above horizontal plane of reference) are defined alternatingly at successive grid points along the x-axis. (Deltares, 2013, Stelling and Duinmeijer, 2003). The Delft-FEWS platform imports the Delft-Disdrometer data, provides it to the urban drainage model and starts a simulation. The simulation results are then imported to Delft-FEWS for locations of interest. In this case study, we imported water levels in the manholes, time-water on the streets, and sewage spilling towards the ditch at location Jaffalaan.
RESULTS AND DISCUSSION
Drop size distribution
The Delft-Disdrometer algorithm uses the DSD to determine the amount of rainfall. Insight in the DSD and the variability of DSD is of upmost importance in different fields as discussed in the introduction. Therefore, the import procedure from Delft-Disdrometer to Delft-FEWS covers this. Figure 2 shows an example of the DSD as monitored by the Delft-Disdrometer and visualised by Delft-FEWS. The drop sizes are divided in 10 classes and visualised as a percentage of the total amount of drops per time step. The classes of the drop size are uniformly distributed with exception of the lowest and highest class, respectively:
- DropClass (1,:) = all drops smaller than 1 mm, - DropClass (2,:) = drops between 1 and 1.5 mm, - Drop Class (3,:) = drops between 1.5 and 2 mm, - ...
- DropClass (9,:) = drops between 4.5 and 5 mm, - DropClass (10,:) = drops bigger than 5 mm.
As the distribution of drop sizes is related to the rainfall intensity (Uijlenhoet & Stricker 1999) storm events with high rain rates will have higher probability on bigger drops and therefore higher drop classes (e.g. event on September 8th, 2013). Events with hardly any rain will have relatively smaller drops compared to high intensity rain events. Note that the
5
DSD is visualised as a percentage of the total amount of drops and does not yet give information about the amount of drops. If for instance only a single drop is recorded in the lowest class, this will be presented such that 100% of the drops of the recorded drops are smaller than 1 mm. In Figure 2 this class is represented by the colour yellow.
Figure 2. Example of drop size distribution as measured by a Delft-Disdrometer and visualised in Delft-FEWS
Comparison of rainfall data
The Delft-Disdrometers recorded the DSD and consequently rainfall over a period of 90 days, starting from August 1st, 2013. Two extreme rainfall events occurred during the monitoring period. The first storm event started on September 6 and lasted for 6 consecutive days. The most intense rainfall occurred from the second to the fifth day (4 days in total). The second storm event started on October 11 and lasted for 3 days, with most of the rain during the last two days as presented in Table 1. KNMI (2004) derived return periods for multiple-day storm events in The Netherlands. The cumulative depth of both storm events and monitoring techniques is well within the 95% confidence interval of the shown return periods.
Table 1. Analysis of storm events during monitoring period
September event Delft-Disdrometer Radar
Cumulative rainfall depth 6 days (mm) 103.4 104.1 Cumulative rainfall depth 4 days (mm) 97.6 94.7
Return period 4 days (years) 25 25
October event Delft-Disdrometer Radar
Cumulative rainfall depth 3 days (mm) 59.9 107.9 Cumulative rainfall depth 2 days (mm) 50.9 87.2
6
The mismatch between the radar and the Delft disdrometer for the storm events from 11th till 13th October 2013 can be explained by a gap of 24hours in the disdrometer data on the 12th of October. This gap can be caused by loss of connection between the communication module and the gateway or a firmware update from the server side. On October 13th the disdrometer came back online and started measuring again. From this point onward the disdrometer and the radar measured comparable amounts of rainfall as is shown in Figure 3.
Figure 3. Double mass comparison between the Delft-disdrometer and the radar. Left: storm events from September 6th till September 12th, 2013. Right: storm events from October 11th till October 13th, 2013.
We selected the September 2013 storm event to illustrate the application of the robust and low-cost Delft-Disdrometer in urban drainage modelling. The urban drainage model is consequently forced with the rainfall recorded by both the Delft-Disdrometer and the ‘Nationale Regenradar’ data. The radar data is catchment averaged.
Urban drainage modelling using Delft-Disdrometer forcing
We used these two data sources to run the SOBEK model of the Delft urban drainage network and analyse the potential use of Delft-Disdrometer data in urban drainage modelling. As the Delft-Disdrometer shows equivalent total rain depth to the calibrated radar data of the ‘Nationale Regenradar’ for the selected period, we conclude that the simulation results will mainly be affected by possible differences in temporal distribution. Reduction of the spilling towards the open water system is often taken as an objective of urban drainage (sewer) analysis in The Netherlands for water quality reasons. Therefore, we choose to compare the spill discharge at location Jaffalaan, close to the university campus. Figure 4 shows that there is a sound agreement in the timing of the peak spill discharge. However, the simulated spilled volume differs quite a lot for both forcing data sources. Figure 3 shows that the Delft-Disdrometer measured less rain for most of the storm event. This could be an explanation to the differences. Additionally, the specific timing of a certain rainfall volume and the operation of the urban drainage system should be assessed to explain the differences in detail. As this case study focused on the framework instead of e.g. calibration that analysis was not carried out.
7
Figure 4. Simulated sewage spilling at location Jaffalaan, Delft Ideas on real world applications
The case study shows an example of how to use the Delft-Disdrometer in urban drainage modelling. How could the developed sensor and modelling framework be used in real world applications? Some examples on how this study could be applied in the ‘real world.’
- It could be used to calibrate an existing radar network by implementing a grid of Delft-Disdrometers. The observation of the spatial distribution and intensity of storm events might benefit from such a relative low-cost and robust monitoring network. - A network of Delft-Disdrometers can be beneficial for near real time asset
management, such as sewer management and gully pot maintenance management. - In catchments with a high correlation between rainfall and the occurrence of (urban)
flooding the low-cost Delft-Disdrometer could be used to set up a dense network and upgrade the quality of a flood early warning system.
- A Delft-Disdrometer network could easily be set up in ungauged or poorly gauged basins, as studied in the Trans-African Hydro-Meteorological Observatory (TAHMO) project. The goal is of the TAHMO project is to install a dense network of hydro-meteorological monitoring stations in Sub-Sahara Africa. This is only possible if robust monitoring equipment is available, such as the Delft-Disdrometer (http://www.tahmo.org).
- The development of an import routine to the Delft-FEWS platform provides many opportunities as the framework is used in a range of applications, e.g. for flood and drought forecasting, water quality, reservoir management and/or hydropower. The WaterML 2.0 import routines developed for Delft-FEWS allows users to readily access and transforms Delft-Disdrometer data as desired, use it as forcing for a set of models and start a simulation workflow.
CONCLUSION
We set up a framework to use acoustic disdrometer data as forcing to simulate urban drainage systems with hydrological and hydraulic models. The effect of using either calibrated ‘Nationale Regenradar’ or the Delft-Disdrometer as meteorological forcing for urban drainage modelling is compared and the two sources are in agreement in terms of cumulative rain depth. The simulated spilling discharge differs as the temporal component of the two forcing data sets vary. Given the results, we consider the Delft-Disdrometer a candidate for calibration of radar data, contributing to the accuracy of such measurements, and a useful
8
instrument to provide forcing data to urban drainage models. With this case, we demonstrated a successful application of robust acoustic Delft-Disdrometers in urban drainage modelling using Delft-FEWS. This gives opportunity to a low-cost increase of gauge density in urban areas which might lead to a more accurate estimation of the rain depth and thus increasing the likelihood to link rainfall to local flooding and its simulation with urban drainage models.
ACKNOWLEDGEMENTS
We would like to thank the Climate City Campus Project, an initiative of Delft University of Technology, for providing the necessary disdrometer data. Furthermore, we are grateful that Delft municipality shared their sewer model for the case study. This study was carried out within the framework of the European Institute of Innovation and Technology (EIT) Climate-KIC programme, Smart Urban Water project.
REFERENCES
Bagree, R., de Jong, S., Precipitation measurement system and method for measuring precipitation, NL Patent 2008563, issued 29-03-2012
van de Beek, C.Z., Leijnse, H., Stricker, J.N.M., Uijlenhoet, R., Russchenberg, H.W.J. (2010). Performance of
high-resolution X-band radar for rainfall measurement in The Netherlands. Hydrology And Earth System
Sciences, Vol 14, 205–221.
Berne, A., Delrieu, G., Creutin, J.D., Obled, C. (2004). Temporal and spatial resolution of rainfall
measurements required for urban hydrology. Journal of Hydrology, Vol 299, 166–179.
Deltares (2013). SOBEK User Manual.
Hut, R. (2013). New Observational Tools and Datasources for Hydrology. PhD Thesis, Delft University of Technology. http://dx.doi.org/10.4233/uuid:48d09fb4-4aba-4161-852d-adf0be352227
IPCC AR4 WG1 (2007), Solomon, S.; Qin, D.; Manning, M.; Chen, Z.; Marquis, M.; Averyt, K.B.; Tignor, M.; and Miller, H.L., ed., Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change, Cambridge University Press.
Jaffrain, J., Berne, A. (2012). Quantification of the small-scale spatial structure of the raindrop size distribution
from a network of disdrometers. Journal of Applied Meteorology and Climatology 51, 941–953.
Jaffrain, J., Berne, A. (2012). Influence of the subgrid variability of the raindrop size distribution on radar
rainfall estimators. Journal of Applied Meteorology and Climatology 51, 780–785.
Jaffrain, J., Studzinski, A., Berne, A., (2011), A network of disdrometer to quantify the small-scale variability of
the raindrop size distribution. Water Resources Research, Volume 47, issue 3.
Jameson, A. R., and A. B. Kostinski (2001). What is a raindrop size distribution? Bull. Am. Meteorol. Soc., 82(6), 1169–1177
KNMI (2004). Statistics of extreme precipitation in The Netherlands (in Dutch).
Marshall, J. S., and W. M. Palmer (1948). The distribution of raindrops with size. J. Meteorol., 5, 165–166. Michaelides, S., V. Levizzani, E. Anagnostou, P. Bauer, T. Kasparis, and J. Lane (2009). Precipitation:
Measurement, remote sensing, climatology and modelling. Atmos. Res., 94, 512–533.
R.R.P. van Nooijen, A.G. Kolechkina, K-J. van Heeringen and E. van Velzen. (2011). Implementation of central
control for multiple sewer systems. 12th International Conference on Urban Drainage, Porto Alegre/Brazil OGC Document 09-124r2 (2008). Harmonising Standards for Water Observation Data (OGC Discussion
Paper) WMO-168: Guide to Hydrological Practices, Sixth edition.
Pardo-Ig´uzquiza, E. (1998). Optimal selection of number and location of rainfall gauges for areal rainfall
estimation using geostatistics and simulated annealing. Journal of Hydrology 210, 206–220.
Salles, C., J. Poesen, and D. Sempere-Torres (2002). Kinetic energy of rain and its functional relationship with
intensity. J. Hydrol., 257(1–4), 256–270.
Seo, D.J., Breidenbach, J.P. (2002). Real-Time Correction of Spatially Nonuniform Bias in Radar Rainfall Data
Using Rain Gauge Measurements. Journal of Hydrometeorology,Vol 3, 93–111.
Steiner, M.; Smith, J.A.; Uijlenhoet, R. A. (2004). Microphysical Interpretation of Radar Reflectivity–
Rain Rate Relationships. Journal of the atmospheric sciences. 61, 1114–1131.
Stelling G.S. and S.P.A. Duinmeijer. (2003). A staggered conservative scheme for every Froude number in
rapidly varied shallow water flows. International journal for numerical methods in fluids, Vol. 43 (2003),
12, pp. 1329-1354.
Werner, M., Schellekens, J., Gijsbers, P., van Dijk, M., van den Akker, O., Heynert, K. (2013). The Delft-FEWS
flow forecasting system. Environmental Modelling and Software
