This study evaluates three real-time Kalman filter (KF) approaches—Mean Field (MKF), Site-Specific (SKF), and Lagrangian (LKF)—for radar rainfall bias correction in northern China. While SKF addresses spatial heterogeneity using independent gauge-location filters, LKF advects a gridded log-bias correction field with storm motion estimated from optical flow, reducing the erroneous transfer of large correction factors into trailing light-rain areas. Using 176,008 15-min validation radar-gauge samples from 449 held-out gauges, LKF reduced radar bias from − 35.0

Insights from Forecast Demonstration Projects and Research Development Projects, training workshops, and symposia, conducted between 2000 and 2024 are summarized. The projects were organized by the Nowcasting and Mesoscale Research Working Group of the World Weather Research Programme of the World Meteorological Organization. The objective was to advance, promote, and build capacity in nowcasting and very short-range forecasting. The projects were associated with the Olympic Games, emergency management, and aviation services. They brought international experts together to work in a collaborative fashion. Extensive interaction with end users and decision-makers expanded and extended the scope of services from traditional weather hazards (heavy rain, wind, hail, lightning) to include specific user needs (e.g., visibility in complex terrain or airport runways, periods of calm winds or light rain, heat stress). Substantial progress has been made in many areas including advanced radar nowcasting algorithms, stochastic nowcasts, kilometric and hectometric numerical weather prediction models, blending of observations and models, and multimodel systems. Verification was a key and valuable component of the projects quantifying the results. Also, the types of services have expanded to include both summer and winter services, complex terrain and urban environments, air transport, air quality, hydrology, and health. Insights are presented in all aspects of nowcasting and very short-range forecasting from end-user decision-making, critical role of the forecaster, forecast systems (models, heuristics, observations), to science and knowledge gaps.

This study presents the first evaluation of using commercial microwave link (CML) data for rainfall measurements in Australia, with the test site being the greater Melbourne Metropolitan area. More than 100 CMLs with microwave frequency ranging between 10 and 40 GHz have been used for the rainfall retrieval. The 15-minute received signal levels (RSLs) for each CML based on two sampling strategies (average and minimum/maximum) collected for 2 years provided a unique dataset to compare performances of rainfall retrievals. The open source algorithm RAINLINK was used for deriving rainfall from the 15-minute RSL data. From two years of data, a subset of 30 rainy days distributed across this period were used for calibrating the RAINLINK parameters, with the remaining data used for validation. For this study, only path-averaged rainfall intensities were validated based on a gauge-adjusted radar product serving as the reference. The result of the wet-dry classification showed that the minimum and maximum RSL data performed better, with lower probability of false detection and higher Matthews correlation coefficient than average RSL data. For the rainfall retrieval, both datasets showed similar correlation with the gauge adjusted radar product. However, based on other statistics (RMSE, bias and CV) minimum and maximum RSL data outperformed average for the rainfall retrieval. Overall, this study highlights the robust accuracy of commercial microwave links for rainfall retrieval while using only minimum and maximum RSL data.

Short-term precipitation forecast plays a vital role for minimizing the adverse effects of heavy precipitation events such as flash flooding. Radar rainfall nowcasting techniques based on statistical extrapolations are used to overcome current limitations of precipitation forecasts from numerical weather models, as they provide high spatial and temporal resolutions forecasts within minutes of the observation time. Among various algorithms, the Short-Term Ensemble Prediction System (STEPS) provides rainfall fields nowcasts in a probabilistic sense by accounting the uncertainty in the precipitation forecasts by means of ensembles, with spatial and temporal characteristic very similar to those in the observed radar rainfall fields. The Australian Bureau of Meteorology uses STEPS to generate ensembles of forecast rainfall ensembles in real-time from its extensive weather radar network. In this study, results of a large probabilistic verification exercise to a new version of STEPS (hereafter named STEPS-3) are reported. An extensive dataset of more than 47000 individual 5-minute radar rainfall fields (the equivalent of more than 163 days of rain) from ten weather radars across Australia (covering tropical to mid-latitude regions) were used to generate (and verify) 96-member rainfall ensembles nowcasts with up to a 90-minute lead time. STEPS-3 was found to be more than 15-times faster in delivering results compared with previous version of STEPS and an open-source algorithm called pySTEPS. Interestingly, significant variations were observed in the quality of predictions and verification results from one radar to other, from one event to other, depending on the characteristics and location of the radar, nature of the rainfall event, accumulation threshold and lead time. For example, CRPS and RMSE of ensembles of 5-min rainfall forecasts for radars located in mid-latitude regions are better (lower) than those ones from radars located in tropical areas for all lead-times. Also, rainfall fields from S-band radars seem to produce rainfall forecasts able to successfully identify extreme rainfall events for lead times up to 10 minutes longer than those produced using C-band radar datasets for the same rain rate thresholds. Some details of the new STEPS-3 version, case studies and examples of the verification results will be presented.

Good short-term predictions of rainfall over a few hours can be made through advecting the current radar image. Numerical Weather Prediction (NWP) extrapolates high resolution models of the atmosphere. Advection forecasts are useful for a range of 0-3 hours and NWP forecast are generated up to days in advance. The question is to combine the two to optimize the forecast for the 3-24 hour period when information from the initial radar field may still usefully correct the NWP. To achieve this blending, several questions need to be addressed. Firstly, the reliability of both types of the forecasts needs to be estimated. The reliability of advection forecasts is, to some degree, answered by Short-Term Ensemble Prediction Systems (STEPS) through creating ensembles of forecasts. This can also be applied to NWP’s though the size of the datasets involved in this makes it unwieldy. Furthermore, NWP forecast rainfall has systemic biases, underestimating the area of rainfall and skewing the probability distribution of rainfall rates at each pixel to the right, overestimating the maximums. Post processing of the NWP rainfall is done so the structure more accurately represents real rain fields. Even with a post-processed NWP there remains the smoothing issue: if the advection and NWP forecasts locate the storm front at different places then blending is smoother than either, decreasing the variance in rainfall across the domain. Thus, we also consider how the real time radar image may be used to correct the NWP forecast in space and time to mitigate this smoothing effect.

A novel proposal to create probabilistic attenuation nowcasting as a by-product from ensembles of rainfall forecasts is presented in this study. These attenuation nowcasts may eventually be used by mobile network operators to dynamically adjust their wireless network operations in advance and during heavy and extreme rainfall events. It may also facilitate mobile network operators to see a direct benefit of widely sharing its received power level data of their backhaul towers for 'opportunistic' rainfall estimation in real-time in urban areas becoming a clear win-win situation for telecom operators and hydrologists. It is proposed here that probabilistic attenuation forecasts can be derived from the ensembles of high-resolution forecast rainfall fields with lead times of 15 to 90 minutes generated from weather radar using the Short-Term Ensemble Prediction System (STEPS). The ensembles of rainfall predictions can be easily converted to attenuation for specific operating frequencies. This study used 109 microwave links ranging from 15 to 40 GHz to verify the results of this probabilistic attenuation forecast. Results suggest that the STEPS-based attenuation forecast was within the narrow span of the 90 percent confidence region for all microwave links tested, with up to 30-minute lead time, and was found to be skilful for lead times of up to 30-45 minutes.

In this letter, we propose a novel approach to produce attenuation forecasts for microwave links using a probabilistic approach. It uses ensembles of forecast rainfall fields to easily derive attenuation forecasts for specific frequencies. The proposed approach uses the short-term ensemble prediction system (STEPS) to generate ensembles of high, spatial and temporal, resolution forecast rainfall fields based on observed weather radar fields with lead times of 15-90 min. Attenuation forecasts could eventually be used by telecommunication operators to drive the operation of wireless networks and ensure their maintenance during severe and extreme rainfall events. This study used 109 microwave links ranging from 15 to 40 GHz to verify the results of this probabilistic attenuation forecast. Results suggest that the STEPS-based attenuation forecasts were within the narrow span of the 90% confidence region for all microwave links tested up to 30 min lead time, decreasing for longer lead times. Examples of how the proposed approach can be used to derive a detailed probabilistic attenuation forecast for multiple lead times within a domain of few kilometers, as well as probability of attenuation maps for large areas are shown.

Commercial microwave links (CMLs) have proven useful for providing rainfall information close to the ground surface. However, large uncertainties are associated with these retrievals, partly due to challenges in the type of data collection and processing. In particular, the most common case is when only minimum and maximum received signal levels (RSLs) over a given time interval (hereafter 15 min) are stored by mobile network operators. The average attenuation and the corresponding rainfall rate are then calculated based on a weighted average method using the minimum and maximum attenuation. In this study, an alternative to using a constant weighted average method is explored, based on a machine learning model trained to produce actual attenuation from minimum/maximum values. A rainfall retrieval deep learning model was designed based on a long short‐term memory (LSTM) model architecture and trained with disdrometer data in a form that is comparable to the data provided by mobile network operators. A first evaluation used only disdrometer data to mimic both attenuation from a CML and corresponding rainfall rates. For the test data set, the relative bias was reduced from 5.99% to 2.84% and the coefficient of determination (R2) increased from 0.86 to 0.97. The second evaluation used this disdrometer‐trained LSTM to retrieve rainfall rates from an actual CML located nearby the disdrometer. A significant improvement in the overall rainfall estimation compared to existing microwave link attenuation models was observed. The relative bias reduced from 7.39% to −1.14% and the R2 improved from 0.71 to 0.82.

Pysteps is an open-source and community-driven Python library for probabilistic precipitation nowcasting, that is, very-short-range forecasting (0–6 h). The aim of pysteps is to serve two different needs. The first is to provide a modular and well-documented framework for researchers interested in developing new methods for nowcasting and stochastic space–time simulation of precipitation. The second aim is to offer a highly configurable and easily accessible platform for practitioners ranging from weather forecasters to hydrologists. In this sense, pysteps has the potential to become an important component for integrated early warning systems for severe weather. The pysteps library supports various input/output file formats and implements several optical flow methods as well as advanced stochastic generators to produce ensemble nowcasts. In addition, it includes tools for visualizing and post-processing the nowcasts and methods for deterministic, probabilistic and neighborhood forecast verification. The pysteps library is described and its potential is demonstrated using radar composite images from Finland, Switzerland, the United States and Australia. Finally, scientific experiments are carried out to help the reader to understand the pysteps framework and sensitivity to model parameters.

1. Colorado State University, Fort Collins, Colorado, United States, 2. Finnish Meteorological Institute, Helsinki, Finland, 3. Federal Office of Meteorology and Climatology MeteoSwiss, Locarno, Switzerland, 4. Institute for Atmospheric and Climate Science, ETH Zurich, Switzerland, 5. Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Canada, 6. Bureau of Meteorology, Melbourne, Australia

A multiplicative cascade model called High‐resolution Downscaling of Rainfall Using Short‐Term Ensemble Prediction System (HiDRUS) is developed and tested in the greater Melbourne region (Australia) by downscaling coarse‐resolution ERA‐I rainfall to 1‐km horizontal and 6‐min temporal resolutions. The parameters required for the cascade model are computed from radar observations of rain events during 2008–2015, and a library of rainfall events and their associated synoptic conditions created. Each day, the area‐averaged rainfall and synoptic conditions are taken from ERA‐I and compared with the library. From the library, similar days are chosen randomly and downscaled using the cascade model. Ensembles of 100 realizations per day are produced for the period 1995–2004. The downscaled rainfall is compared with 6‐min rain gauges and daily gridded rain gauge data at four locations in the greater Melbourne region. HiDRUS reproduces the monthly variability of rainfall, frequency distribution of daily and 6‐min rainfall, and the autocorrelation function satisfactorily. Changes in heavy rainfall are also captured by HiDRUS but with increasing uncertainty as the intensities increase.

Knowledge of the full rainfall drop size distribution (DSD) is critical for characterising liquid water precipitation for applications such as rainfall retrievals using electromagnetic signals and atmospheric model parameterisation. Southern Hemisphere temperate latitudes have a lack of DSD observations and their integrated variables. Laser-based disdrometers rely on the attenuation of a beam by falling particles and are currently the most commonly used type of instrument to observe the DSD. However, there remain questions on the accuracy and variability in the DSDs measured by co-located instruments, whether identical models, different models or from different manufacturers. In this study, raw and processed DSD observations obtained from two of the most commonly deployed laser disdrometers, namely the Parsivel1 from OTT and the Laser Precipitation Monitor (LPM) from Thies Clima, are analysed and compared. Four co-located instruments of each type were deployed over 3 years from 2014 to 2017 in the proximity of Melbourne, a region prone to coastal rainfall in south-eastern Australia. This dataset includes a total of approximately 1.5 million recorded minutes, including over 40 000 min of quality rainfall data common to all instruments, equivalent to a cumulative amount of rainfall ranging from 1093 to 1244 mm (depending on the instrument records) for a total of 318 rainfall events. Most of the events lasted between 20 and 40 min for rainfall amounts of 0.12 to 26.0 mm. The co-located LPM sensors show very similar observations, while the co-located Parsivel1 systems show significantly different results. The LPM recorded 1 to 2 orders of magnitude more smaller droplets for drop diameters below 0.6 mm compared to the Parsivel1, with differences increasing at higher rainfall rates. The LPM integrated variables showed systematically lower values compared to the Parsivel1. Radar reflectivity–rainfall rate (ZH–R) relationships and resulting potential errors are also presented. Specific ZH–R relations for drizzle and convective rainfall are also derived based on DSD collected for each instrument type. Variability of the DSD as observed by co-located instruments of the same manufacturer had little impact on the estimated ZH–R relationships for stratiform rainfall, but differs when considering convective rainfall relations or ZH–R relations fitted to all available data. Conversely, disdrometer-derived ZH–R relations as compared to the Marshall–Palmer relation ZH=200R1.6 led to a bias in rainfall rates for reflectivities of 50 dBZ of up to 21.6 mm h−1. This study provides an open-source high-resolution dataset of co-located DSD to further explore sampling effects at the micro scale, along with rainfall microstructure.

A new quality control system, primarily using a naive Bayesian classifier, has been developed to enable the assimilation of radial velocity observations from Doppler radar. The ultimate assessment of this system is the assimilation of observations in a pseudo-operational numerical weather prediction system during the Sydney 2014 Forecast Demonstration Project. A statistical analysis of the observations assimilated during this period provides an assessment of the data quality. This will influence how observations will be assimilated in the future, and what quality control and errors are applicable. This study compares observation-minus-background statistics for radial velocities from precipitation and insect echoes. The results show that with the applied level of quality control, these echo types have comparable biases. With the latest quality control, the clear air observations of wind are apparently of similar quality to those from precipitation and are therefore suitable for use in high-resolution NWP assimilation systems.

Distributions of rainfall with the time and space resolutions of minutes and kilometers, respectively, are often needed to drive the hydrological models used in a range of engineering, environmental, and urban design applications. The work described here is the first step in constructing a model capable of downscaling rainfall to scales of minutes and kilometers from time and space resolutions of several hours and a hundred kilometers. A multiplicative random cascade model known as the Short‐Term Ensemble Prediction System is run with parameters from the radar observations at Melbourne (Australia). The orographic effects are added through multiplicative correction factor after the model is run. In the first set of model calculations, 112 significant rain events over Melbourne are simulated 100 times. Because of the stochastic nature of the cascade model, the simulations represent 100 possible realizations of the same rain event. The cascade model produces realistic spatial and temporal patterns of rainfall at 6 min and 1 km resolution (the resolution of the radar data), the statistical properties of which are in close agreement with observation. In the second set of calculations, the cascade model is run continuously for all days from January 2008 to August 2015 and the rainfall accumulations are compared at 12 locations in the greater Melbourne area. The statistical properties of the observations lie with envelope of the 100 ensemble members. The model successfully reproduces the frequency distribution of the 6 min rainfall intensities, storm durations, interarrival times, and autocorrelation function.

Merging radar and gauge rainfall estimates is an area of active research. Since rain gauges alone are often limited at representing the complete spatial distribution of rainfall, a combination of radar-derived rainfall with spatially interpolated gauge estimates using alternate weighting approaches is investigated. This paper examines several merging methods that differ in the consideration of correlation among the estimation errors, their distribution, and the application of dynamic and static weighting. The merging process has been applied to the radar data from Terrey Hills radar located in Sydney, Australia, and spatially interpolated gauge rainfall on the same area. The performance of the merging methods is assessed by comparing the combined estimate with the gauge observation. It is however clear from our findings that rainfall estimation from any of the combination approaches assessed contains less error than any of the noncombination approaches. The results show that the correlation between these two rainfall estimation errors plays a significant role in the performance of the merging methods. The combination method should be chosen depending on the purpose, accuracy of the estimate, and complexity of the method.