Update: We are happy to announce that all available recordings from the conference can now be watched and are listed here
Flash floods in small to meso-scale catchments and intense precipitation over cities from severe local storms pose increasing threats to our society. For the timely prediction of such events, the value of high-resolution and high-quality quantitative precipitation estimation and corresponding forecasts cannot be overrated.
Seamless predictions harmonizing nowcasting and NWP across forecast lead times from minutes to days would greatly help to improve the value and efficiency of warnings. We invite contributions of recent ideas and achievements along the process chain from quantitative precipitation estimation (QPE), observation-based nowcasting (QPN), combination of nowcasting with NWP and flash-flood prediction (FFP). Area-covering and high-resolution polarimetric weather radar observations currently provide core information for QPE/QPN like precipitation intensity, hydrometeor types, and wind gusts augmented with satellite and lightning observations.
With the perspective to enable/improve seamless predictions of surface precipitation and flash floods from the time of observation to days ahead, potential topics of this session are advances in quantitative high-resolution precipitation retrieval including promising sensor synergies, advances in nowcasting including life cycle information, advances in NWP data assimilation for short-term forecasts, concepts for merging observation-based QPN with NWP, and concepts to transfer the uncertain precipitation estimates over the forecast range into river runoff and flash-flood predictions.
The RealPEP online conference will be carried out using the German Research Network DFN.
Talks are by invitation only and we are currently in a process of putting together an exciting programme
with contributions by both international experts and young researchers.
The talks will be streamed, while discussions take place in a range of targeted discussion rooms,
moderated by the RealPEP research group principle investigators and research scientists.
Please register if you intend to take part to be provided with links and further information.
Quantitative precipitation estimation
Numerical weather prediction
Flash flood prediction
The HRRR model is an operational storm-scale numerical weather prediction forecast system that has been updated approximately every 2 years since it was first introduced in 2014. This seminar gives a brief update on the evolution of the HRRR, with a focus of what is new for version 4 which will become the operational model in December 2020. I will also give an update of the future of operational storm-scale modeling in the US as NOAA moves towards a unified forecasting system.
There are different "optimal" forecast methods for different forecast lead times and different weather phenomena. Focusing on precipitation and convective events up to some hours ahead, radar extrapolation techniques (Nowcasting) show good skill up to about 2 h ahead (depending on the situation), while numerical weather prediction (NWP) outperforms Nowcasting only at later hours. Ensembles of both Nowcasting and NWP help to assess forecast uncertainties. "Optimally" combining precipitation forecasts from Nowcasting and NWP as function of lead time leads to seamless forecasts.
However, very often separate groups are working on these topics independently, even at the same center. Progress on all these fields could be enhanced by strengthening the coordination, feedback and exchange among the groups.
In 2017, DWD has taken this road when setting up a project to research, develop and establish its future Seamless INtegrated FOrecastiNg sYstem (SINFONY), which is intended to be the basis for DWD's future severe weather warning process from minutes to 12 h.
Different teams work closely together in developing
a) Radar Nowcasting ensembles for precipitation, reflectivity (ideas from STEPS) and convective cell objects (KONRAD3D),
b) hourly NWP rapid-update-cycle ensemble prediction system on the km-scale (SINFONY-RUC-EPS), assimilating high-resolution observations of 3D radar volume scans (radial wind, reflectivity, cell objects), Meteosat VIS channels and lightning,
c) optimal combination of Nowcasting and NWP ensemble forecasts in observation space (precipitation, radar reflectivity and cell objects),
d) systems for common Nowcasting and NWP verification of precipitation, reflectivity and objects. In particular the cell object based verification will provide new insights into the representation of deep convective cells in the model.
For b), New innovative and efficient forward operators for radar volume scans and visible satellite data enable
direct operational assimilation of these data in an LETKF framework.
For c), the SINFONY-RUC-EPS outputs simulated reflectivity volume scan ensembles of the entire German radar network every 5' online during its forecast runs. Ensembles of composites and cell object tracks are generated by the same compositing and cell detection- and tracking methods/software packages which are applied to the observations.
To help evolve DWD's warning process for convective events towards a flexible "warn-on-objects", our Nowcasting- and NWP cell object ensemble forecasts are blended into a seamless forecast ("probability objects") in a pragmatic way. The gridded combined precipitation and reflectivity ensembles are targeted towards hydrologic warnings.
The RealPEP research group funded by the German Research Foundation (DFG) was established at the beginning of 2019 and motivated by the prediction of floods especially in small to meso-scale catchments to mitigate risks to society and ecosystems.
RealPEP addresses the full process chain from Quantitative Precipitation Estimation (QPE), Quantitative Precipitation Nowcasting (QPN), Quantitative Precipitation Forecasting (QPF) by Numerical Weather Prediction (NWP) to Flash Flood Prediction (FFP). Major objectives include improved accuracy of radar-based QPE, the use of convective initiation and storm development in nowcasting, bridging the gap between observation-based nowcasting and NWP, and the combination of advanced QPE/QPN/QPF with physics-based hydrological models to improve flood forecasting.
After >1.5 years of joint research, RealPEP organizes this conference to present first results summarized in this presentation, but also to exchange with the community and to reflect and address the state-of-the-art questions of current research on Precipitation and Flash Flood Prediction from Minutes to Days.
Significant progress has been recently made for estimation of rainfall at relatively short distances from the radar where the radar samples pure rain below the melting layer (ML). The R(A) methodology based on the use of specific attenuation A proves to be superior compared to the standard QPE algorithms utilizing radar reflectivity Z, specific differential phase KDP, and differential reflectivity ZDR (at least at S band). However, several challenges still remain including (1) the impact of the rain type variability within the radar coverage area, (2) possible hail contamination, (3) strong vertical gradients of rain rate below the ML, and (4) utilization of the R(A) approach for QPE at C band. These will be discussed in the talk.
At longer distances from the radar, contamination from the ML and overshooting of rain are the principal problems which are poorly resolved with existing methods for the vertical profile of reflectivity (VPR) correction. A novel technique, polarimetric VPR (or PVPR), is introduced. It utilizes measured radial profiles of cross-correlation coefficient CC which are coupled with the corresponding radial dependencies of the Z bias caused by the ML contamination and overshooting. The PVPR technique takes into account statistical correlations of CC and Z within the ML and explicitly treats the effects of radar beam broadening on the radial profiles of CC and Z. Multiple examples of the PVPR performance at S and C bands will be presented.
Polarimetric weather radar have provided improved depictions of the size, shape, phase and concentration of various scatterers in the atmosphere over the single-polarized weather radar and a more accurate quantitative precipitation estimation (QPE). A number of studies using disdrometer data indicated that specific attenuation, a polarimetric radar derived variable, was more linearly related to the liquid water content than reflectivity and was less sensitive to systematic biases in reflectivity and differential reflectivity observations. A specific attenuation-based QPE (“RA”) was implemented across the United States NEXRAD network in a real-time experimental system at the National Severe Storms Lab in Oct. 2016. A key parameter in the RA scheme, , was assumed to be a relatively uniform in space but dynamically updated to reflect the temporal change of drop size distributions (DSD). The RA QPE was evaluated and refined for 2 years and a relatively stable version was implemented in mid-Sept. 2018.
The new version of RA was evaluated against quality controlled gauge observations and compared with a reflectivity-based QPE (“RZ”) for over 50 events from different regions of CONUS. RA provided consistent improvements over RZ for moderate to heavy rain. RA also showed less sensitivity to partial beam blockages and calibration biases than RZ. For light to very light stratiform rain, RA had a dry bias due to weak attenuation signal. RA scheme also had some local under- and overestimation errors where the spatial variations of DSD were large. A spatial adjustment of RA rates was developed and was shown to further reduce the local biases.
The inherent uncertainty in radar snow estimates comes from variability in snow size distributions, diversities among snow growth habits, and changes in particle densities. As a consequence, radar snow measurements are very challenging. However, the utilization of polarimetric radar data can address some of these problems. A novel polarimetric method for quantification of snowfall rate, based on the joint use of specific differential phase KDP and horizontal reflectivity factor Z, is introduced. Large 2D-video disdrometer snow dataset from central Oklahoma is utilized to derive polarimetric bivariate power-law relation for snowfall rate, S(KDP, Z) = γKDP^αZ^β. The relation is generalized for the range of particle aspect ratios from 0.5 to 0.8 and the width of the canting angle distribution from 0 to 40 degrees, and validated via analytical/theoretical derivations and simulations. The relation’s multiplier is sensitive to variations in particles’ aspect ratios and the width of the canting angle distribution, whereas the exponents are practically invariant to these changes. The novel approach is tested with operational polarimetric WSR-88D data. Radar estimates of S are compared to the ground measurements in several United States locations, exhibiting small bias. The results are encouraging and show good potential for the improvement in radar QPE of snow.
Recent advances have been made to demonstrate the benefits of radar-derived horizontal specific attenuation (AH) for Quantitative Precipitation Estimation (QPE) at S and X band, however, to date the methodology has not been adapted and optimized for C-band radars in widespread use in Europe. Simulations based on a large dataset of Drop Size Distributions (DSDs) measured in Germany are performed to investigate the DSD dependencies of the attenuation parameter αH, mandatory to first derive AH. The normalized raindrop concentration (Nw) and the slope of differential reflectivity (ZDR) versus reflectivity (ZH) are used to categorize radar observations and corresponding optimized αH are applied. For heavier continental rain dominated by large raindrops originating from hail or contaminated with hail, the R(AH) algorithm is further combined with the rainfall retrieved from proxy specific differential phase R(KDP). Also, the performance of retrievals based on vertical specific attenuation R(AV) are tested. Finally, the adapted hybrid QPE algorithms are applied to five diverse case study days monitored by the nation-wide C-band radar network in Germany and compared to rain gauge measurements and the operational German RAdar-OnLine-ANeichung (RADOLAN) RW product, which is based on reflectivities only but adjusted to rain gauges. Rainfall retrievals using AH/V with optimized αH/V combined with R(KDP) show a smaller bias and outperform traditional R(Z) algorithms when evaluated via rain gauges. Moreover, R(AH/V)-based algorithms are consistent with the RW product and additionally bear the advantage of immunity to the partial beam blockage. An optimization of net αH/V along the radial or within a segment is suggested for future work.
Hydrometeorologists have traditionally relied on dedicated measurement equipment to do their business. Such instruments are typically owned and operated by government agencies and regional or local authorities. Installed and maintained according to (inter)national standards, they offer accurate and reliable information about environmental states and fluxes. Such standard instruments are often further developments of novel measurement techniques which have their origins in the research community and have been tested during dedicated field campaigns. One drawback of the operational measurement networks available to the hydrometeorological community today is that they often lack the required coverage and spatial and/or temporal resolution for high-resolution real-time monitoring or short-term forecasting of rapidly responding systems (e.g. urban areas). Another drawback is that dedicated networks are often costly to install and maintain, which makes it a challenge for nations in the developing world to operate them on a continuous basis, for instance.
Yet, our world is nowadays full of sensors, often related to the rapid development in wireless communication networks we are currently witnessing (including 5G). Let us try to make use of such opportunistic sensors to do our science and operations. They may not be as accurate or reliable as the dedicated measurement equipment we are used to working with, let alone meet official international standards, but they typically come in large numbers and are accessible online. Hence, in combination with smart retrieval algorithms and statistical treatment, opportunistic sensors may provide a valuable complementary source of information regarding the state of our environment.
The presentation will focus on some recent examples of the potential of opportunistic sensing techniques for rainfall monitoring, using microwave links from cellular communication networks (in Europe, South America, Africa and Asia) as well as personal weather stations.
Urban catchments are characterized by diverse land cover with large ratio of impervious surfaces on which rainfall-runoff is generated extremely fast. This possess high requirements on spatio-temporal resolution of rainfall data needed for reliable runoff predictions. The contribution summarizes results from three-years experiment evaluating reliability of rainfall-runoff simulations at small urban catchment using rainfall observations from Commercial Microwave Links (CMLs). In addition, system for providing real-time CML-based rainfall product at city scale is presented. Finally, sustainability of the CML-based rainfall observation system is discussed with focus on challenges related to the fast development of communication networks.
Commercial microwave links (CMLs) can be used for quantitative precipitation estimation (QPE) by exploiting the close to linear relationship between path-integrated attenuation and path-averaged rain rate. Currently, attenuation data from 3904 CMLs in Germany is obtained in real-time, with 3 years of collected data. We have carried out a first large scale analysis of a full year of countrywide CML observations. Building on top of this analysis we were able to significantly improve the robustness of our CML QPE by detecting rainfall induced attenuation patterns with a neural network approach.
In parallel we investigate the synergetic use of CMLs and weather radars. Polarimetric weather radars can estimate rainfall in a similar fashion as CMLs by relating rainfall and attenuation via the R(A) algorithm. Within RealPEP we aim to use the synergies that these two approaches offer. Weather radars estimate accumulated attenuation along the radar ray which can lead to uncertainties in areas far from the radar or behind hail cores. Our aim is to complement weather radars with in situ attenuation measurements from CMLs.
In this presentation we show our recent advancements, preliminary results and also the challenges that arise when comparing different sources of rainfall information like e.g., the spatio temporal mismatch of high resolution QPE products.
The simplest and most effective way to nowcast precipitation in the next 2 hours is to extrapolate composite radar images along the estimated motion field (optical flow), commonly referred to as nowcasting by Lagrangian persistence. In recent years, there has been a growing amount of studies exploiting machine learning to extract useful (predictable) information from radar archives for precipitation nowcasting. These new methods have great potential, but suffer from the same limitations of statistical and analogue-based nowcasting approaches, for example the inevitable smoothing arising from error minimization and the finite size of archives.
In this contribution, we review recent advances in precipitation nowcasting using machine learning at MeteoSwiss. The talk will present a selection of studies that include both nowcasting of convective cells and orographic precipitation in mountainous regions. Through the presentation, we will answer to the following questions, among others:
• Which are the sources of predictability of precipitation growth & decay?
• How to go beyond a smooth deterministic machine learning prediction?
• How to perform a fair comparison of machine learning and extrapolation-based nowcasts?
• How to reduce the number of predictors by feature selection?
• How to generate new stochastic cells in radar images?
• How to combine machine learning and stochastic approaches for uncertainty quantification?
Finally, we will define a possible way forward both from a research (generative machine learning) and a collaboration perspective (open-source projects, e.g. pysteps).
The ability to predict extreme rain and flooding over the next 6 hours is one of the major challenges facing meteorological services worldwide. Currently, two main approaches are available to forecasters: a) data-driven predictions based on the extrapolation of radar observations and b) physically-based numerical weather prediction models (NWPs). Radar-based forecasts offer high spatial resolutions and update times but lack the ability to anticipate crucial dynamic processes such as the growth of new rain cells or sudden changes in rain cell motion or intensity. Mesoscale numerical weather prediction models, on the other hand, are computationally more intensive and tend to have lower spatial resolution and accuracy than radar. However, because they include more physics, they are better at anticipating future changes in storm dynamics beyond a few hours.
In this talk, I propose a new approach to rainfall forecasting based on machine learning that combines the accuracy and speed of radar extrapolation techniques with the physical knowledge and contextual understanding of a numerical weather prediction model. The basic idea is to teach machine learning models to anticipate crucial changes in rain cell properties (e.g., size, motion, intensity, etc..) based on the coarse-scale guidance from a numerical weather prediction model. Different models and approaches are discussed and some real-life examples for heavy convective rain events in the Netherlands are shown.
The most widely used technique for nowcasting of quantitative precipitation in operational and research centers is the Lagrangian extrapolation of the latest radar observations. However, this technique has a limited forecast skill because of the assumption made on its formulation. These assumptions are constant motion vectors and neglection of growth or decay in the precipitation field. In this work, the McGill Algorithm for Precipitation Nowcasting by Lagrangian Extrapolation (MAPLE) errors have been computed for 10 yr of radar composite data over the continental United States. The study of these errors shows systematic bias depending on the time of day. This effect is related to the solar cycle, whose heating energy increases the average rainfall in the afternoon. This external forcing interacts with the atmospheric system, creating local initiation and dissipation of convection depending on orography, land use, cloud coverage, etc. The signal of the diurnal cycle in MAPLE precipitation forecast has been studied in different locations and spatial scales as a function of lead time to recognize where, when, and for which spatial scales the signal is significant. This information has been used in the development of a scaling correction scheme where the mean errors due to the diurnal cycle are adjusted. The results show that the developed methodology improves the forecast for the spatial scales and locations where the diurnal cycle signal is significant.
Conventional extrapolation nowcasting methods assume that the precipitation fields do not evolve during the lead-time period. In recent years, nowcasting approaches have been developed addressing this limitation. One of those approaches is the short-term ensemble prediction system (STEPS). Based on the assumption that smaller precipitation structures have a shorter lifetime and lower predictability than larger structures, STEPS decomposes the precipitation field into different spatial scales and filters those having a short lifetime and low predictability. The latter is achieved by controlling the observation memory using an auto-regressive (AR) model. Next, STEPS replaces the filtered scales associated with low predictability by a realization of a stochastic noise field.
In this work, the configuration, implementation, and performance of STEPS are studied for its radar-based nowcasting application in Germany. Attention is given to the parameters that control the spatio-temporal evolution of precipitation. Furthermore, the spatial variability of the precipitation field is considered to locally weight the stochastic noise field. Preliminary analysis shows that the performance of STEPS is in particular sensitive to the chosen memory order of the AR model, the resolution scales at which observed precipitation is decomposed, and the spatial integration of predicted precipitation at different scales during the lead-time period. It is also seen that by weighting the noise field according to the local precipitation variability, the skills of the nowcast improve. Moreover, the spatial and temporal structures of the nowcast fields are consistent with observed precipitation fields.
Our preliminary results highlight aspects needed in the configuration of STEPS such as the spatio-temporal properties of the nowcast fields, thereby serving as a basis for an improved nowcasting system in Germany.
This presentation will provide an overview of the activities related to the combination and analysis of lightning and radar data conducted at MeteoSwiss. The first part of the presentation will be devoted to the analysis of the performance of the lightning jump algorithm that is implemented in real-time for nowcasting severe thunderstorms. It uses the total lightning rate from the operational network to identify the occurrence of lightning jumps inside each convective cell identified automatically by the TRT nowcasting system. In the second part, we will present an analysis of a large dataset of lightning and polarimetric weather radar data collected in the course of a lightning measurement campaign. The campaign took place in the summer of 2017 in the area surrounding the Säntis mountain, in the northeastern part of Switzerland. The area is a hotspot for lightning activity in Switzerland. For this campaign and for the first time in the Alps, a lightning mapping array (LMA) was deployed. In this study, we relate LMA very high frequency (VHF) sources data with collocated radar data in order to characterise the main features (location, timing, polarimetric signatures, etc.) of both the flash origin and its propagation path. We provide this type of analysis first for the entire dataset and then we stratify it into intra-cloud and cloud-to-ground flashes (and within this category positive and negative flashes) and also upward lightning. We show that polarimetric weather radar data can be helpful in determining regions where lightning is more likely to occur, namely in areas with a high extend in the vertical of hail and graupel, but that lightning climatology and/or knowledge of the orography and man-made structures is also relevant.
Highly spatial and temporal variable atmospheric water vapor largely determines whether clouds, and subsequent precipitation, evolve and how they develop differently in certain regions. The spatial variability of atmospheric water vapor fields at very small scales, but for larger regions, is currently only observable from polar-orbiting satellites.
We present our newly developed and evaluated total column water vapor retrievals from measurements of the passive imager OLCI, onboard Sentinel-3 satellites, based on extensive experience with TCWV retrievals for MERIS and MODIS. While precipitation itself is mostly invisible to passive VIS/NIR/TIR radiation, past case studies have shown the potential to identify small-scale convective structures in satellite-based TCWV fields. In the context of improving the characterization of the pre-convective environment using satellite observations of water vapor and clouds to potentially improve Quantitative Precipitation Nowcasting in Germany, an assessment is made in a statistical manner as well.
By combining the high spatial resolution TCWV observations from OLCI (250m) with high temporal resolution (15 min.) TCWV time series from the dense German GPS network and cloud observations from the passive imager SEVIRI on the geostationary satellite MSG, observed measures of spatial and temporal TCWV variabilities are related to various cloudy conditions at later time steps. The match-up of the various observational datasets allows us to mimic at least partly the significantly improved observational capabilities of future MTG, with which the monitoring of the pre-convective environment as well as cloud and precipitation development is expected to improve significantly. To this end, also an algorithm framework was set up to investigate the potential of TCWV retrievals from MSG-SEVIRI observations and, on a test-basis, from future MTG-FCI observations.
In countries, usually very large or small, there is usually the need to use QPE both from radar and satellite. The combination of these two sensors is not straight forward as there are usually inconsistencies between them. Within the Nowcasting Satellite Application Facility (NWC SAF) from EUMETSAT, AEMET has started to make some efforts in understanding the source of these inconsistencies. This work is still ongoing.
As an alternative solution, AEMET is trying to combine in a unique QPE product measurements from radar, satellite and rain gauges. This QPE will be used operationally in future FFPs in some Spanish regions.
The NWC SAF is also preparing its next generation precipitation estimation using meteorological knowledge, past experience and machine learning techniques. This product will improve on the previous existing one.
Another line of work along these lines is the combined representation of different products on the same geographical projection. The NWC SAF is currently exploiting ADAGUC.
All these subjects will be presented in this talk.
Brief periods of intense rainfall can lead to flooding with the potential to cause damage to property and to threaten lives. To improve the prediction of these events, more accurate forecasts of convective rainfall are needed, and these can then be used to inform flood guidance and warning systems. In this talk we will review data assimilation for convection-permitting numerical weather prediction and discuss the main unsolved issues still facing this area, including forecast error covariance modelling, nonlinearity, microphysics and multiscale assimilation. We will highlight recent progress in understanding observation uncertainty and how better characterization of observation uncertainty can lead to better observation impacts in the assimilation. We will discuss novel observation types arising from datasets of opportunity. We will end by discussing the assimilation and observing systems for the future of high impact weather prediction.
Prediction of severe precipitation events and flash floods happens at timescales of minutes to hours, which is also the intersection of nowcasting (NWC) and numerical weather prediction (NWP). At most centers, NWC and NWP have thus far comprised very separate systems, with not just different prediction methods but also different observation types used to constrain the predictions. One aim of DWD's SINFONY project is to bring NWC and NWP together by taking the data assimilation cycle of the NWP system from 3-hourly to 1-hourly updates, while also bringing in new observation types that contain information about clouds, precipitation, and convection. These observation types include radar, lightning, and cloud images, and are, at least in part, already used for nowcasting. This talk will discuss recent progress in making the assimilation and rapid update cycling of these observations a reality, identifying the particular difficulties of assimilating cloud- and precipitation-related observations while also outlining the possibilities that these new observation types present for better short-term weather prediction.
There is an ever-increasing demand for high-resolution weather prediction, which in the UK has an emphasis on improved forecasts/nowcasts of the types of convective storms that, in recent years, have led to serious flooding. In this talk, we will present some of the latest contributions made by the Met Office Reading group to the field of convective data assimilation and the use of radar observations to improve precipitation forecasts. This will include a comparison between a 4DVar hourly cycling NWP and the Met-Office operational nowcasting system, a pragmatic approach to account for Doppler radial wind correlated observation error and results from a new scheme for the direct assimilation of radar reflectivity observations.
Dual-polarization radar data can be used to infer a wealth of information about precipitation microphysics. The development and coupling of forward polarimetric radar operators to models has emerged as a popular approach for relating and studying the sensitivity of observed polarimetric signatures to underlying microphysical processes, with the goal of improving microphysical parameterization schemes and nowcasting abilities. This talk will present the latest progress of developing one-dimensional microphysical models informed by polarimetric radar data. Results from a one-dimensional snow model initialized with polarimetric microphysical retrievals demonstrate the potential for estimating melting-layer cooling rates from specific differential phase (KDP), as well as simulating snow sublimation and the accurate prediction of snow start times at the surface in the presence of dry air. In addition, preliminary results indicate the successful incorporation of snowflake aggregation and breakup processes constrained by observed polarimetric radar variable profiles. Finally, applications of a one-dimensional model of melting graupel/hail and comparison with observations will be presented that suggest the potential for early detection of wet microbursts and nowcasting of hailfall swaths.
The assimilation of dual-polarimetric radar observations in NWP models is promising especially for short-term forecasts of quantitative precipitation. However, directly assimilating polarimetric moments is still a challenge due to, e.g., the rather rudimentary appreciation of particle size and shape distributions in the NWP models. Consequently, pioneering studies started assimilating synthetic model state variables derived from radar polarimetric observables, such as hydrometeor mixing ratios. This study evaluates hydrometeor mixing ratio retrievals already published, and improves and adapts them for applications to the C-band dual-polarimetric radar network of the German national weather service (DWD) and later assimilation of retrievals in the ICON model. A large data set of drop-size-distributions (DSD) measured by DWD in Germany including a large variety of different rain types is investigated based on T-matrix simulations. Results show that existing, mostly simple power-law retrieval relations derived for other climate regimes are not appropriate for Germany, which underlines a set of local retrieval relations is needed for assimilation of hydrometeor mixing ratios in the ICON model. Suitable local power-law, polynomial and also rational retrievals are derived from the DSD data set. Application of the newly introduced relations to both stratiform and convective rainfall events monitored by DWD’s radar network shows encouraging results.
Recently, the assimilation of radar reflectivities has become operational at the DWD. However, the assimilation of radar data still comes along with some unique challenges and shows several deficiencies in practice.
For addressing some of these deficiencies, we present two approaches for further improving the assimilation of this new kind of data within the framwork of the localized ensemble transform Kalman filter (LETKF).
First, we investigate the impact of an update of additional hydrometeors like, e.g., qr, qg, qs during the assimilation which, eventually, aims for improving the model physics.
Second, we address one of the limitations of the LETKF itself which occurs in the case where simulated reflectivities exhibit a comparatively small spread leading to small increments even for situations where observed and simulated reflectivities show large discrepancies. For overcoming this issue, we present the additive covariance inflation approach which exploits an additional correlation between the reflectivity and humidity qv for artificially increasing the spread of simulated reflectivites if necessary.
Microphysical schemes in numerical weather prediction models are based on the
basic equations that govern the growth and interaction of liquid drops and ice
particles (hydrometeors) in the atmosphere. Unfortunately, our knowledge of
these microphysical processes is still incomplete. In addition, quite drastic
simplifications are necessary to make the problem tractable in numerical
weather prediction. Hence, we need observations to test, validate and improve
the microphysical parameterizations.
Here we give two examples of this validation and development process. First, a
deficiency in the properties of unrimed snowflakes is identified using Doppler
spectra of a vertically pointing radar. The snowflake properties are improved
based on aggregation modeling, and the revised microphysical model is
validated using multi-wavelength radar data. Second, the use of a one- vs
two-moment microphysics is compared for forecasts of deep convection. The
two-moment scheme has several advantages, but shows a bias in the diurnal
cycle of the forward-simulated radar cell-tracking statistics. This can be
traced back to a delayed initiation of convection, which is caused by a bias
in the cloud-radiation interaction of the two-moment microphysics. After this
issue is fixed, several biases of the NWP system are improved, e.g. radar cell
statistics and near-infrared brightness temperature histogram.
Several efforts have been made to enhance our knowledge on how flash floods occur and to develop forecasting techniques that anticipate their impacts. Short-term ensemble forecasts have been considered crucial to assess forecast uncertainty and improve flood warning. This presentation focuses on the challenges of evaluating the performance of ensemble forecasts of flash floods. It is based on ongoing work carried out within the PICS national research project in France (“Towards integrated nowcasting of flash floods impacts”; https://pics.ifsttar.fr/). The project’s goals is to develop and evaluate integrated short-range nowcasting chains that combine (i) high resolution quantitative precipitation estimates and short range (0-6h) precipitation forecasts, (ii) distributed rainfall-runoff models designed to simulate river discharges in gauged and ungauged conditions, (iii) DTM-based hydraulic models for the delineation of potentially flooded areas, and (iv) impacts models aiming to represent a range of socio-economic effects. After an overview of the PICS project, the presentation will focus on the evaluation of ensemble flood forecasts in the Aude River Basin in France. The GRSD semi-distributed hydrological model is first adapted to be flexible enough to simulate two key types of flood events observed in the study region: winter and spring floods, often occurring during or after wet periods, and autumn floods, often occurring after long and dry summer periods. It is then used to illustrate the performance of three ensemble precipitation forecast products recently developed by the French meteorological service (Météo-France) for the 14-15 October 2018 flood event. The three products are based on the ensemble forecasts from the AROME NWP model, on a combination of the AROME forecasts with nowcasting products, and on the introduction of perturbations to this combined forecast set. Precipitation forecasts up to six hours ahead are used to force the semi-distributed hydrological model at the hourly time step. The challenges of event-based ensemble forecast evaluation are discussed.
The changing frequency and severity of extreme flood events are becoming increasingly apparent over multi-decadal timescales at the global scale, even though confidence in climate risk scenarios is clouded by the confounding effects of hydrological and landscape system dynamics and time-varying factors such as land use changes. Improved flood risk management builds upon disentangling climate change impacts from other controlling factors, thereby contributing to the debate over the need for societal adaptation to extreme events. In this work we focus on the coupled hydrologic and geomorphic controls of flood risk at the catchment scale. Sediment (and large wood) transport events, of both high and low magnitude, have the potential to reshape channel and floodplain topography, thus introducing an additional source of uncertainty in the quantification of flood hazard. However, determining the extent to which such events are actually able to modify channel geometry is rather complex. Indeed, not all the large floods cause major reshaping of the river corridor, whereas relatively low-magnitude, high-frequency floods may result in major morphological changes.
Post-flood surveys designed to integrate interlinked observations of hydrologic response together with sediment and large wood transport provide key data for better understand and predict extreme floods and their morphological responses. In turn, the knowledge of flood runoff response and its morphological effects may inform improved flood risk management strategies and interventions at the basin scale, especially in poorly gauged basins. However, integrated observations of hydrologic and geomorphic response are difficult, and even more is the identification of the their controlling factors, as large geomorphic changes introduce large uncertainties in the post-flood estimation of peak discharges.
Here we revisit the lessons learnt from several integrated post-flood surveys carried out in Italy in the last five years in gravel bed rivers draining areas up to approximately 1000 km$^2$. This provides also the opportunity to discuss how to translate the knowledge gained through post-flood surveys into flood risk management and basin management practice.
We present a framework for flash flood nowcasting using the partial differential equation (PDE)-based ParFlow hydrologic model forced with quantitative radar precipitation estimates and nowcasts. The prelimiar results for a small 18.5 km 2 headwater catchment in Germany is chosen to illustrate the application of such framework for 2 aims: the first is to verify the applicability of PDE-based models in the context of flash floods of poorly gauged watersheds and the second is to use the framework to evaluate improvements of precipitation products, within the Research Unit RealPEP. For comparison, a commonly used conceptual model (HBV) was applied over the same catchment. Finally, the model lead time improvements when forced with precipitation nowcasts will be presented.
Forecasting flash floods some hours in advance is
still a challenge, especially in environments made up of many
small catchments, where uncertainties in magnitudo and spatial location of forecast rainfall are generally high. The scope of this work is to exploit both observations and modelling sources to improve the discharge prediction in
small catchments with a lead time of 2–8 h.
We used a nowcasting model and a meteorological where a data assimilation technique is applied with high frequency, to generate rainfall fields which are merged by a blending techinque. These latter are use tu feed a hydrological model and produce streamflow predictions.
Results seem to evidence that the implemented approach is quite promising in order to improve flood forecast
Reliable and timely flood forecasts help improve flood preparedness and recovery. Unlike riverine and coastal flooding where forecasting methods are well established, surface water and flash flood forecasting presents a unique challenge due to the high uncertainties around predicting the location, timing and impact of what are typically localised events.
Thanks to the recent rapid development of convection-permitting numerical weather prediction, it is now theoretically feasible to develop operational surface water forecasting systems. However there remains a scientific limit to the predictability of convective rainfall. To overcome this challenge a re-thinking of the established role of flood forecasting is needed, alongside the development of interdisciplinary solutions for communicating uncertainty.
For scientists this means considering what information end users need to inform flood warnings, and then developing innovative solutions to meet those needs. Effectively closing the circle of the end-to-end forecasting chain while building understanding of uncertainty propagation. For example at what spatial scale to decision makers need warnings?, How much lead time do they need to take effective action?, What level of uncertainty are they prepared to accept in the forecast?
Based on experiences from the Flooding from Intense Rainfall (FfIR) programme and the development of operational surface water flood forecasts in Scotland, this presentation will categorise different approaches to surface water flood forecasting into three groups; empirical-based thresholds, hydrological forecasts linked to pre-simulated impact scenarios and hydrological forecasting linked to real-time hydrodynamic simulations at the urban scale. Using operational and near operational examples of such systems, it will consider how these approaches can make the best use of emerging scientific capability and meet the different needs of end users.
Direct current meter measurements are rarely available for extreme flash floods. Corresponding discharges are generally estimated using so-called “indirect” techniques such as the slope – area method. These methods are based on empirical hydraulic formulae that typically use Manning’s equation, and have been calibrated and also widely tested for flow conditions that differ significantly from those encountered during flash floods. Recent work conducted in Europe, as part of the HYDRATE research project and other studies, has shown that the use of these formulae and their associated tabulated roughness values available in current guidance documents, without further verification, can lead to over-estimates of peak discharges in the case of flash floods. After having discussed the limitations of indirect methods based on Manning’s formula, the presentation describes how the uncertainty in indirect discharge estimates can be reduced through the analysis of various types of data that can be collected during post-event surveys and through consistency checks. Based on a review of current literature and on recent flash flood studies, we suggest simple guidelines to assist practitioners in estimating extreme discharges during post-event surveys.
Around the world, such as in southeastern Brazil, extreme rainfall events cause
several socioeconomic damages. Due to some of its characteristics, as the high
space-time variability, these events are difficult to predict. Methods based on
deep learning can be one of the solutions for this type of issue. These methods
have a high capacity to learning through historical events. This research aims to
develop a model able to generate rainfall nowcasting based on neural networks
using radar data. For study case, fifteen rainfall events were selected in the
region of Campinas (state of São Paulo-Brazil). The results show that the
approach can be a possible solution for nowcasting of extreme rainfall events.
Update your browser to view this website correctly. Update my browser now