The Hydrology, Meteorology and Complexity Laboratory of École des Ponts ParisTech (
Numerous rainfall simulators have been developed and used primarily to study soil erosion as well as tillage techniques. Indeed the natural extreme variability of rainfall features (e.g. occurrence, intensity, duration, drop size distribution), makes such study under natural conditions more complicated. For a short historical review of rainfall simulators and their uses, the interested reader is referred to
Some authors analysed the rain simulated with the help of disdrometers, which are devices that give access to size and velocity of the falling drops. For example
In this paper we present disdrometer data collected during a measurement campaign aiming at testing a rainfall simulator installed in the climatic chamber of the Sense-City experiment. The campaign took place in September 2017. Before continuing, it should mentioned that the two disdrometers used here are already presented in
Data and methods are presented in Sect. 2 with a brief overview of the device functioning and available datasets as well a description of the rainfall simulator and the measurement campaign. Section 3 presents the database and the available tools to use it. Outputs of the campaign are discussed in Sect. 4 along with an illustrative comparison with some actual rainfall.
As pointed out in the introduction, the measurement campaign uses devices whose functioning has already been described in detail in
The output provided is actually not directly the features of each individual drop but rather a matrix containing the number of drops recorded during the time step
Definition of the classes of particle size and velocity for the Parsivel
Pictures showing an overview of the measurement campaign
The three quantities analysed in the paper and made available in the corresponding database are the rain rate
Finally it should be mentioned that no filter was implemented on the matrix for this specific implementation; i.e. the whole matrix is used. In some case authors introduced a filter to exclude drops whose measured fall velocity was too far from the theoretical expected terminal fall velocity and hence assumed to correspond to non-meteorological hydrometeors
Sense-City is a 400 m
The rainfall simulator does not cover all the areas but only a 25 m
The measurement campaign took place 26–28 September 2017 in the Sense-City climatic chamber.
Some pictures illustrating it can be found in Fig.
Scheme of the various locations tested over the area covered by the rainfall simulator.
Measurements were carried out for both light and heavy rainfall at five different locations within the area wet by the rainfall as shown in Fig.
In addition, a specific test keeping the disdrometers at location no. 1 while changing the input flow of water was carried out on 27 September 2017 between 09:45:00 and 11:20:00. Given that the rainfall simulator is not designed for such use, it resulted in a malfunctioning of the nozzles, notably with very large drops falling on the wet area. Hence measurement during this period is not discussed in this paper and should not be considered.
Start and end time of tests for each location in both light and heavy rainfall configuration. Local time is used.
Snapshot of the page “Calendar_Sense_City.html” of the database.
This section is actually quite similar to the corresponding one of disdrometers_data_base/
Raw_data_zip/
Pars1/ Pasr2/ Each folder contains the files for its disdrometers. The name is Raw_DisdroName_YYYYMMDD.zip (e.g. Raw_pars1_20170926.zip). Daily_data_python/
Pars1/ Pasr2/ Each folder contains the files for its disdrometers. The name is DisdroName_raw_data_YYYYMMDD.csv (e.g. Pars1_raw_data_20170926.npy). Exports/
Full_matrix/ KE/ R/ Each folder contains the files for all the disdrometers. The name is DisdroName_DataType_date.csv (e.g. Pars1_KE_30_sec_2017_09_26_00_00_00__2017_09_26_23_59_30). Calendars/
Data_5_min/ (one file per day;e.g. R_5_min_Sense_
City_2017_09_26_00_00_00__2017_09_26_23_59_30.csv Data_30_sec/ (one file per day;e.g. R_30_sec_Sense_City_2017_09_26_00_00_00__2017_09_26_23_59_30.csv) Quicklooks/ (one file per day and test;e.g. Quicklook_Sense_City_2017_09_26_00_00_00__2017_09_26_23_59_30.png) Calendar_Sense_City.html Python_scripts/
It contains the Python scripts (and associated files) to generate and use this database. Read_me.txt
It contains a short description of the Taranis Sense-City database.
Figure
Quicklook of the measurements at location no. 1 with light rainfall.
An example of quicklook can be found in Fig. the rain rate cumulative rainfall depth the DSD time series of missing time steps (a visible coloured bar for missing ones) (middle right); a visual representation of the matrix containing the number of drops according the velocity and size classes, with the classes of velocity represented vertically or diameter represented horizontally; the solid black line is the curve corresponding to the relation between the terminal fall velocity of drops as a function of their equivolumic diameter obtained by the temperature the kinetic energy density flux KE (in J m
The files for daily rainfall rate are named in a similar way as the quicklooks except that “Quicklook” is replaced by “R_5_min” or “R_30_sec”. They are CSV files with the following format:
(i) There is one line per time step (either 30 s or 5 min time step starting on YYYY-MM-DD 00:00:00 LT). (ii) In each line, values of rain rate (in mm h (iii) Missing data are denoted by “nan”.
This folder contains exports in a simple CSV format of the main outputs of the disdrometer that could be relevant for potential users of these data, i.e. the full matrix of drops according to size and velocity classes (subfolder “/Full_matrix”), kinetic energy density flux (subfolder “/KE”), and rain rate (subfolder “/R”).
In a given folder, a file is typically called “Pars1_KE_30_sec_2017_09_26_00_00_00__2017_09_26_23_59_30.csv”, meaning the disdrometer name, the data type, and start and end of the period corresponding to the data are easily visible for the user. In the file, the format is the following: (i) one line per time step; (ii) for each line, date (YYYY-MM-DD HH:MM:SS); data. For
This folder contains the daily file for both disdrometers in their own subfolder. Each file is stored in NPY format and requires Python 3 to be read. They contain all the collected data stored as a list. It is these files that are read by the Python scripts described in the corresponding subsection.
This folder is actually very similar to the corresponding one of
Average rain rate expressed in millimetres per hour for each location (#) for both light and heavy rainfall simulations.
Quicklook of the measurements at location no. 1 with heavy rainfall.
Again this section is very similar to the corresponding one of Quicklook_and_R_series_generation_Sense_City_without_PWS generates a quicklook image and the corresponding 30 s and 5 min rain rate time series for a given rainfall event for the Sense_City campaign. extracting_one_event_Sense_City reads daily.npy files and generates three lists (one for each disdrometer) containing all the data that can be analysed for the Sense_City campaign. exporting_full_matrix reads daily NPY files and exports the full matrix in CSV files for a given disdrometer and event. exporting_R readings daily NPY files and exports exporting_T reads daily NPY files and exports exporting_KE reads daily NPY files and exports KE in CSV files for a given disdrometer and event.
Commented examples of use of the functions can be found in the scripts: “Example_of_use_data_base_sense_city.py”. Note that Python 3 (
Quicklook of the measurements carried out on the roof of the Carnot Building of ENPC campus on 10 February 2019 between 07:00:00 and 10:00:00 UTC.
Figure
Figures
The first point is the steadiness of the features of the simulated features both in terms of rain rate (upper left) or DSD (middle left), which is radically opposed to what is found in natural rainfall. It is actually a property that is wanted for a rainfall simulator. Let us simply mention that it takes a few minutes to reach a “permanent” regime in the functioning of the nozzles.
A closer look at the DSD (middle left and lower left on the quicklooks) reveals that the drops generated by the rainfall simulators are smaller than actuals ones. More precisely the DSD is much thicker and centred on smaller drops. A common indicator is the mass-weighted diameter
In addition, the maps basically displaying the number of drops according to the classes of velocities and sizes (middle right in the quicklooks) show that drops tend to reach the disdrometers with lower velocities than expected. Indeed for actual rainfall, measured distribution is roughly scattered around the expected theoretical relation between the terminal fall velocity and the equivolumic diameter (solid black line). Such measured distribution is shifted toward smaller velocities. It can be noticed that this issue is more pronounced for larger drops than smaller ones, which is expected. This is due to the fact that the height of the rainfall simulator of 8 m is not sufficient to enable the drops to reach their terminal fall velocities
As a result of both the absence of large drops and lower fall velocity, the kinetic energy of the simulated rainfall is strongly underestimated with the regards to the expected values for such rain rates. This feature is visible on the lower right panel of the quicklooks of Figs.
The database presented in this paper has been made available by the Hydrology, Meteorology and Complexity laboratory of École des Ponts ParisTech (HM&Co-ENPC) and the Sense-City consortium at The following citation should be used for this paper: The following citation should be used for the database:
This dataset is available for download free of charge. Licence terms apply.
Finally it should mentioned that these disdrometers and others have been and are used in other measurement campaigns by HM&Co.
Regular updates of their status along with updates of the database are to be provided through the lab's website (
The 30 s disdrometer data from a measurement campaign beneath a rainfall simulator are presented in this paper. Raw data as well as Python-formatted data with the associated scripts for basic manipulation are described and made available to the community for further use.
In order to discuss the features the rainfall generated by the simulator, an illustrative comparison is made with actual rainfall. It appears the properties of the rainfall generator remain steady over time, which is the desired quality. In terms of more refined properties, the drop size distribution generated is thinner than actual rainfall and centred on smaller drops. In addition the height of the simulator is not sufficient for larger drops (
The research grants obtained respectively by DS and AR permitted the acquisition of the scientific equipment and realization of experimental facilities. IT and AG designed the study. AG and PB carried out the measurement campaign. AG wrote the paper. The joint discussion and analysis by all the authors of the obtained data and results shaped the paper into its actual form.
The authors declare that they have no conflict of interest.
The authors greatly acknowledge partial financial support from the Chair of Hydrology for Resilient Cities (endowed by Veolia) of the École des Ponts ParisTech, and the Île-de-France region RadX@IdF Project. We also want to thank ANR, the French Funding Agency, for supporting Sense City Equipex.
This study received financial support from the Chair of Hydrology for Resilient Cities (endowed by Veolia) of the École des Ponts ParisTech, and the Île-de-France region RadX@IdF Project.
This paper was edited by Scott Stevens and reviewed by two anonymous referees.