HS7.6

Disasters associated with heavy rainfall such as urban floods, riverine floods, and landslides often simultaneously occur while each of them sensitively reacts to rainfall variabilities at distinct ranges of time scales. Therefore, a stochastic rainfall model suitable for modeling compounding of disasters must be good at reproducing the rainfall variability across all timescales relevant to all types of disasters. This study proposes a point stochastic rainfall generator that can reproduce various rainfall characteristics at timescales between 5 minutes and one decade. The model generates the fine-scale rainfall time series using a randomized Bartlett-Lewis Rectangular Pulse (RBLRP) model. Then the rainstorms are shuffled such that the correlation structure between the consecutive storms is preserved. Finally, the time series is rearranged again at the monthly timescale based on the result of the separate coarse-scale monthly rainfall model. The method was tested using the 69 years of 5-minute rainfall data recorded at Bochum, Germany. The mean, variance, covariance, skewness, and rainfall intermittency were well reproduced at the timescales from 5 minutes to a decade without any systematic bias. The extreme values were also well reproduced at timescales from 5 minutes to 3 days. The past-7-day rainfall before an extreme rainfall event, which is highly associated with the extreme riverine flow and landslide was reproduced well too. Then, the model was extended to integrate the influence of climate change. For this, the model was re-parameterised in terms of parameters representing average magnitude and temporal structure of the rainfall time series. Then, the relationship between these new parameters and the covariates (e.g. monthly, weekly, daily temperature) were investigated. Lastly, the derived regression relationships were applied to adjust the duration and the magnitude of rain storms and cells that were generated by the stationary RBLRP model.

This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (No. NRF-2021R1A2C2003471).

How to cite: Kim, D., Onof, C., Park, J., and Wang, L.: A Stochastic Rainfall Generator Suitable for Modeling Future Compound Disasters Associated with Heavy Rainfall, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2595, https://doi.org/10.5194/egusphere-egu22-2595, 2022.

The representation of extreme precipitation at small spatio-temporal scales is of major importance in urban hydrology. The present study compares observations from tipping bucket gauges and a C-band radar to two sets of re-analysis climate model output data for a historic period of 14 years where there is full spatial and temporal overlap between datasets. The reanalysis data are based on models with different parametrizations and spatio-temporal resolutions, one being a “convective-permitting” model while the other uses a convective parametrization scheme to account for convective rainfall. The study focuses on an area of approximately 100 by 150 km.

The datasets are compared with respect to seasonality of occurrence, intensity levels and spatial structure of the extreme events. All datasets have similar seasonal distributions, and comparable intensity levels. There are, however, clear differences in the spatial correlation structure of the extremes. Seemingly, the radar data is best representation of a “real” spatial structure for extreme precipitation, even though challenges appear in data when moving far from the physical radar. The spatial correlation in point observations is a valid representation of the spatial structure of extreme precipitation. The convective-permitting climate model seem to represent the spatial structure of extreme precipitation much more realistically, compared to the coarser convective parameterized model. However, improvement could be made for the shortest durations and smallest spatial scales.

How to cite: Arnbjerg-Nielsen, K., Thomassen, E. D., Thorndahl, S. L., Andersen, C. B., Gregersen, I. B., and Sørup, H. J. D.: Comparing extreme precipitation between data from rain gauges, weather radar and high-resolution climate models, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2624, https://doi.org/10.5194/egusphere-egu22-2624, 2022.

Since Hinton et al. introduced Deep Learning (DL) in 2006 [1], DL methods have led to breakthroughs in various scientific fields, such as speech recognition, medical, materials, and many more. Various early attempts to apply DL to short-term rainfall forecasting (nowcasting) were also reported. However, these early models did not lead to significant improvements as compared to non-AI nowcasting models such as STEPS [2]. It was TrajGRU model [3] which first demonstrated the potential gains that may be achieved with DL-based nowcasting models. Since then, a variety of DL models have been proposed or applied to tackle rainfall nowcasting, with the most iconic ones including U-Net, MetNet and DGMR [4-6]. Similarly to the Trajectory GRU (TrajGRU) model, the U-Net and MetNet models show clear improvements in predicting the occurrence of rainfall at high spatial and temporal resolutions and with a longer lead time, as compared to non-AI models. However, the predicted rainfall images from these three models (and their variants) become overly smooth rather quickly (at lead times of 15-20 minutes); this is a common ‘feature’ of many other DL models [7]. This means that significant amount of spatial rainfall details is lost, which is undesirable for certain hydrological applications, such urban flow and flood forecasting where small-scale rainfall variability -in particular localised peaks- may have tangible impacts [8]. In 2016, DeepMind [6] proposed a new type of DL-based nowcasting model called the Deep Generative Model of Radar (DGMR), which is based upon a Generative Adversarial Network (GAN) framework. The DGMR successfully improves the aforementioned smoothing drawback of other DL-models by incorporating noise into the rainfall forecast generator such that small-scale rainfall details can be preserved and, consequently, localised peak intensities can be better predicted. DGMR thus shows great potential for hydrological applications.

In spite of the success, the model structure of DGMR is complex and hard to digest by someone without proper training in DL. Therefore, even though the model structure has been published, it remains a mystery for most hydrologists, thus hindering its application.

In this work, we explore the success of DGMR with an in-depth analysis of its model structure. More specifically, through the process of re-constructing the DGMR model, we have developed a short tutorial on the different model components, in plain language and with example images and intermediate analyses. This will enable better understanding of the features and behaviour of the DGMR model and of the implications for hydrological applications. Additionally, a better understanding of the DGMR model components may instigate further improvements.

References:

[1] Hinton, G.E., et al., Neural Comput., 18 (7), 1527-1554, 2006.

[2] Bowler, N.E., et al., Q. J. Roy. Meteor. Soc., 132(620), 2127-2155, 2006.

[3] Shi, X., arXiv preprint arXiv:1706.03458, 2017.

[4] Agrawal, S., et al., arXiv preprint arXiv:1912.12132, 2019.

[5] Sønderby, C.K., et al., arXiv preprint arXiv:2003.12140, 2020.

[6] Ravuri, S., et al., Nature, 597, 672-677, 2021.

[7] Ayzel, G., Geosci. Model Dev., 13(6), 2631-2644, 2020.

[8] Ochoa-Rodriguez, S., et al., J. Hydrol., 531, 389-407, 2015.

How to cite: Heh, Y.-T. and Wang, L.-P.: Unraveling the mystery of DeepMind’s rainfall nowcasting: a step-by-step tutorial for hydrologists, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-9054, https://doi.org/10.5194/egusphere-egu22-9054, 2022.

Rainfall data collected by citizen scientists is typically regarded as low quality and therefore remains underused in hydrological applications. Conversely, official data collected by professional organisations is often treated as more reliable than it really is. Here we demonstrate that value can be extracted from citizen science rainfall data by applying automated statistical quality control combined with manual checks. We also consider the pros and cons of citizen science rain observations.

Carefully selected rain data from official and citizen science gauges have been blended with radar.  Examples of how rainfall depths vary depending on the data inputs are presented, highlighting the benefit of incorporating all available data sources. This approach is particularly important when determining rainfall during spatially and temporally variable convective storms. The research is concerned with convective storms that resulted in pluvial flooding in urban areas of the UK between 2014 – 2018, however, the methodology could be implemented in any location where hourly (or shorter interval) rain gauge data and radar are available.  

How to cite: O'Hara, T., Lewis, E., McClean, F., Fowler, H., and Parkin, G.: Selecting Good Quality Official and Citizen Science Rain Gauge Data and Blending with Radar for More Accurate Rainfall Representation, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-7879, https://doi.org/10.5194/egusphere-egu22-7879, 2022.

This study proposes a survey methodology to identify areas for combined Sustainable Drainage Systems (SUDS) and Aquifer Storage and Recovery (ASR), (Dearden et al. 2013, Sharp Jr., 1997); these techniques exploit the hydrogeological and geomorphological characteristics of an area, to increase the natural capacity of water to infiltrate the ground.

The target area of this case study is the city of Rome and the aim of such techniques is to reduce the problems related to rainwater which, in case of extreme events, struggles to infiltrate the ground, overloads the undersized hydraulic systems and floods the urban space.

The proposed method involves GIS geospatial analysis of various data: the permeability of outcropping lithology, the piezometric level of the aquifer, hydrogeological units, geomorphology and land use.

In this aim zones characterised by high permeability and a piezometric level that would confer a volumetric capacity to possibly store even large quantities of water without triggering possible problems associated with fluctuations in the water table, have been identified.

The data have been divided into classes and indexed for comparison and overlap them. Finally, hydrogeological units were also taken into account (by analysing their depth trend) in order to identify areas with similar characteristics of permeability with respect to depth. The latter will also be compared with the previous data to identify the areas suitable for SUDS and ASR.

The final product of the suitable areas from a hydrogeological point of view will be compared with the land use map in order to exclude those areas that, for administrative and other legislative reasons, cannot be used for such activities.

How to cite: Meddi, E., Lentini, A., Galve, J. P., Papiccio, C., and La Vigna, F.: Mitigation of stormwater flooding by identifying areas suitable for Sustainable Drainage Systems and Aquifer Storage and Recovery (case study: Rome, Italy), EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-12694, https://doi.org/10.5194/egusphere-egu22-12694, 2022.

Urban drainage is subject to a variety of external influencing factors that can have a negative impact on hydraulic system performance. These include changing precipitation characteristics due to climate change [1], an increase in sealed surfaces due to advancing urbanisation [2], but also further failures and malfunctions [3] in the technical grey and green infrastructure. With an increasing share of decentralised urban stormwater measures and uncertainties regarding responsibility, care and maintenance of these facilities, an increase in malfunctions can be assumed [4]. In this work, we are investigating common malfunctions in urban drainage systems with 1D/2D urban flood model in a virtual urban study site. The goal is to highlight differences between the failures and malfunctions in both grey and blue-green infrastructures for different design rainfall events (Type Euler II) and compare them to other possible scenarios like climate change and urbanization.

For the research, a model of a small virtual urban study site (1.5ha) is developed with the commercial software PCSWMM2D [5], which represents a small part of an urban catchment. It includes the following sub-structures and assets: i) combined sewer system, ii) urban stream, iii) urban structures including buildings, marketplaces, streets, bridges, pathways, and underbridges and iv) four sustainable urban drainage system (SUDS) structures (green roofs, permeable pavements, swales and bioretention cell). Connected to the combined sewer system are three border areas (30ha, 10ha, 10ha), representing inflows from outside. The urban drainage infrastructures and SUDS were designed based on a design rainfall (Euler Typ II) event with a 5-year return period and 1-hour event duration.

In total 12 different scenarios were designed for the virtual urban study site i) the SUDS-base-scenario which includes four different green infrastructure assets, ii) three reference scenarios with climate change, urbanisation and wet preconditions and iii) the malfunction-scenarios with seven single malfunction scenarios and one worst case which is all of the single scenarios combined. Each scenario was run with design rainfall events with an Euler II distribution and interval lengths between 15 minutes and 24 hours as well as return periods between 1 and 100 years.

To compare the different scenarios and assess their severity for the urban area we used 3 different objective values. i) The maximum water depth in the vulnerable infrastructure (underbridge), ii) the flooded area with water depths > 10cm and iii) the total combined sewer overflow emissions released into the urban stream.

Results show a clear difference between the different malfunction scenarios, with a higher influence of malfunctions in grey than in green infrastructures. In most cases, the reference scenarios climate change, urbanisation and wet preconditions show higher values than the malfunctions scenario. All scenarios are highly dependent on the rainfall event characteristics, with no differences in the objective values compared to the base case for low return periods and rising differences for medium to high return periods.

How to cite: Funke, F., Reinstaller, S., and Kleidorfer, M.: Impact of malfunctions on urban drainage for different design rainfall events, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-6652, https://doi.org/10.5194/egusphere-egu22-6652, 2022.

Weather radars have become a valuable tool for urban hydrological studies because they capture the rainfall intensities at a high spatial and temporal resolution. However, radar products are affected by objects or phenomena not of meteorological interest, making it necessary to apply various algorithms to correct and improve their rainfall estimation. In addition, multiple methods for merging rain-gauge and radar data are presented in the literature, which combines the advantages of high spatial resolution of radar products with the measurement accuracy of rain gauge stations. While merging methods are commonly validated on rain-gauge measurements, little has been discussed in the literature about the influence of such techniques on urban hydrological models. Therefore, this study investigates the use and selection of gauge-radar merging methods as input for urban hydrological modeling.

This work studies the influence of different precipitation products (rain-gauge stations, radar, and radar-gauge merged products) on flow rates simulated with a hydrodynamic model in two cities: Hildesheim and Osnabrück, Germany. Sewer pipe measurements at least every 2 minutes for several discharge events within 2020-2021 are available and used to evaluate different rainfall products. The techniques to be assessed are temporal and spatial smoothing and radar merging methods such as external drift kriging, quantile mapping, and conditional merging. This study will allow identifying if, in general, there is a single product that presents the best results for urban flow simulations or if, on the contrary, it depends on the type of rainfall event. Additionally, since the study areas are located at different distances from the Hannover radar station, it will be possible to analyze the influence of the attenuation correction on the improvement of the radar product.

How to cite: Echeverri-Salazar, E., Shehu, B., Verworn, A., and Wallner, M.: Effect of spatially distributed radar-gauge rainfall products on simulated urban flows, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-10291, https://doi.org/10.5194/egusphere-egu22-10291, 2022.