Short-term Rainfall Prediction Research Articles

TrajPredRNN+: A new approach for precipitation nowcasting with weather radar echo images based on deep learning

：Short-term rainfall prediction is a crucial and practical research area, with the accuracy of rainfall prediction, particularly for heavy rainfall, significantly impacting people's lives, property, and even their safety. Existing models, such as ConvLSTM, TrajGRU, and PredRNN, exhibit limitations in capturing fine-grained appearances due to insufficient memory units or addressing positional misalignment issues, thereby compromising the accuracy of model predictions. In this study, we propose trajPredRNN+, an innovative approach that integrates the trajectory segmentation model and the PredRNN deep learning model to address both limitations in nowcasting precipitation using weather radar echo images. By incorporating attention mechanisms, the model demonstrates an enhanced focus on short-term and imminent heavy rainfall events. To ensure improved stability during training, a residual network is introduced. Lastly, a more rational and effective training loss function is proposed, encompassing weight mechanism, SSIM index, and GAN loss. To validate the proposed model, we conducted a comparative experiment and an ablation experiment using the radar echo map dataset obtained from the Shenzhen Meteorological Bureau. The results of these experiments demonstrate that our model has achieved significant improvements across multiple key performance indicators.

Advancing very short-term rainfall prediction with blended U-Net and partial differential approaches

Accurate and timely prediction of short-term rainfall is crucial for reducing the damages caused by heavy rainfall events. Therefore, various precipitation nowcasting models have been proposed. However, the performance of these models still remains limited. In particular, the current operational precipitation nowcasting method, which is based on radar echo tracking, such as the McGill Algorithm for Precipitation Nowcasting by Lagrangian Extrapolation (MAPLE), has a critical drawback when predicting newly developed or decayed precipitation fields. Recently proposed deep learning models, such as the U-Net convolutional neural network outperform the models based on radar echo tracking. However, these models are unsuitable for operational precipitation nowcasting due to their blurry predictions over longer lead times. To address these blurry effects and enhance the performance of U-Net-based rainfall prediction, we propose a blended model that combines a partial differential equation (PDE) model based on fluid dynamics with the U-Net model. The evaluation of the forecast skill, based on both qualitative and quantitative methods for 0–3-h lead times, demonstrates that the blended model provides less blurry and more accurate rainfall predictions compared with the U-Net and partial differential equation models. This indicates the potential to enhance the field of very short-term rainfall prediction. Additionally, we also evaluated the monthly-averaged forecast skills for different seasons, and confirmed the operational feasibility of the blended model, which contributes to the performance enhancement of operational nowcasting.

Assessment of Deep Learning-Based Nowcasting Using Weather Radar in South Korea

This study examines the effectiveness of various deep learning algorithms in nowcasting using weather radar data from South Korea. Herein, the algorithms examined include RainNet, ConvLSTM2D U-Net, a U-Net-based recursive model, and a generative adversarial network. Moreover, this study used S-band radar data from the Ministry of Environment to assess the predictive performance of these models. Results show the efficacy of these algorithms in short-term rainfall prediction. Specifically, for a threshold of 0.1 mm/h, the recursive RainNet model achieved a critical success index (CSI) of 0.826, an F1 score of 0.781, and a mean absolute error (MAE) of 0.378. However, for a higher threshold of 5 mm/h, the model achieved an average CSI of 0.498, an F1 score of 0.577, and a MAE of 0.307. Furthermore, some models exhibited spatial smoothing issues with increasing rainfall-prediction times. The findings of this research hold promise for applications of societal importance, especially for preventing disasters due to extreme weather events.

An improved GEV boosting method for imbalanced data classification with application to short-term rainfall prediction

This paper considers the imbalanced binary classification problem by focusing on the application of the short-term rainfall forecasting in arid and semi-arid regions. Specifically, we present a novel boosting-type method by utilizing the generalized extreme value (GEV) distribution as the link function and applying a gradient tree boosting algorithm to capture complex interactions among covariates. The proposed method has several appealing advantages such as, it can identify rare rainfall events as well as quantifying the uncertainties; it is data-driven that without any assumption on the relationship between the covariates and the rainfall event; the fitted model is highly interpretable, making it a useful tool for studying the rainfall mechanisms in arid and semi-arid regions. Experiments on two real-world datasets show that our approach outperforms its competing methods.

Two-Stage UA-GAN for Precipitation Nowcasting

Short-term rainfall prediction by radar echo map extrapolation has been a very hot area of research in recent years, which is also an area worth studying owing to its importance for precipitation disaster prevention. Existing methods have some shortcomings. In terms of image indicators, the predicted images are not clear enough and lack small-scale details, while in terms of precipitation accuracy indicators, the prediction is not accurate enough. In this paper, we proposed a two-stage model (two-stage UA-GAN) to achieve more accurate prediction echo images with more details. For the first stage, we used the Trajectory Gated Recurrent Unit (TrajGRU) model to carry out a pre-prediction, which proved to have a good ability to capture spatiotemporal movement of rain field. In the second stage, we proposed a spatiotemporal attention enhanced Generative Adversarial Networks (GAN) model with a U-Net structure and a new deep residual attention module in order to carry out the refinement and improvement of the first-stage prediction. Experimental results showed that our model outperforms the optical-flow based method Real-Time Optical Flow by Variational Methods for Echoes of Radar (ROVER), and some well-known Recurrent Neural Network (RNN)-based models (TrajGRU, PredRNN++, ConvGRU, Convolutional Long Short-Term Memory (ConvLSTM)) in terms of both image detail indexes and precipitation accuracy indexes, and is visible to the naked eye to have better accuracy and more details.

Short-Term Rainfall Prediction Based on Radar Echo Using an Improved Self-Attention PredRNN Deep Learning Model

Accurate short-term precipitation forecast is extremely important for urban flood warning and natural disaster prevention. In this paper, we present an innovative deep learning model named ISA-PredRNN (improved self-attention PredRNN) for precipitation nowcasting based on radar echoes on the basis of the advanced PredRNN-V2. We introduce the self-attention mechanism and the long-term memory state into the model and design a new set of gating mechanisms. To better capture different intensities of precipitation, the loss function with weights was designed. We further train the model using a combination of reverse scheduled sampling and scheduled sampling to learn the long-term dynamics from the radar echo sequences. Experimental results show that the new model (ISA-PredRNN) can effectively extract the spatiotemporal features of radar echo maps and obtain radar echo prediction results with a small gap from the ground truths. From the comparison with the other six models, the new ISA-PredRNN model has the most accurate prediction results with a critical success index (CSI) of 0.7001, 0.5812 and 0.3052 under the radar echo thresholds of 10 dBZ, 20 dBZ and 30 dBZ, respectively.

Short-term rainfall prediction using MLA based on commercial microwave links of mobile telecommunication networks

Short-term rainfall prediction using MLA based on commercial microwave links of mobile telecommunication networks

Machine Learning-based Short-term Rainfall Prediction from Sky Data

To predict rainfall, our proposed model architecture combines the Convolutional Neural Network (CNN), which uses the ResNet-152 pre-training model, with the Recurrent Neural Network (RNN), which uses the Long Short-term Memory Network (LSTM) layer, for model training. By encoding the cloud images through CNN, we extract the image feature vectors in the training process and train the vectors and meteorological data as the input of RNN. After training, the accuracy of the prediction model can reach up to 82%. The result has proven not only the outperformance of our proposed rainfall prediction method in terms of cost and prediction time, but also its accuracy and feasibility compared with general prediction methods.

Daily Maximum Rainfall Forecast Affected by Tropical Cyclones using Grey Theory

This research aims to develop a model for forecasting daily maximum rainfall caused by tropical cyclones over Northeastern Thailand during August and September 2022 and 2023. In the past, the ARIMA or ARIMAX method to forecast rainfall was used in research. It is a short-term rainfall prediction. In this research, the Grey Theory was applied as it is an approach that manages limited and discrete data for long-term forecasting. The Grey Theory has never been used to forecast rainfall that is affected by tropical cyclones in Northeastern Thailand. The Grey model GM(1,1) was analyzed with the highest daily cumulative rainfall data during the August and September tropical cyclones of the years 2018–2021, from the weather stations in Northeastern Thailand in 17 provinces. The results showed that in August 2022 and 2023, only Nong Bua Lamphu province had a highest daily rainfall forecast of over 100 mm, while the other provinces had values of less than 70 mm. For September 2022 and 2023, there were five provinces with the highest daily rainfall forecast of over 100 mm. The average of mean absolute percentage error (MAPE) of the maximum rainfall forecast model in August and September is approximately 20 percent; therefore, the model can be applied in real scenarios. Doi: 10.28991/CEJ-2022-08-08-02 Full Text: PDF

Very Short-Term Rainfall Prediction Using Ground Radar Observations and Conditional Generative Adversarial Networks

Weather radars play an important role in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">in situ</i> rainfall monitoring owing to their ability to measure instantaneous rain rates and rainfall distributions. Currently, the Korea Meteorological Administration (KMA) provides instantaneous radar observation data and predictions based on the McGill algorithm for precipitation nowcasting by Lagrangian extrapolation (MAPLE) for up to 6 h, for short-term forecasting. This study presents a conditional generative adversarial network (CGAN)-based radar rainfall prediction method for very short-range weather forecasts from 10 min to 4 h. The CGAN-predicted model was trained and tested using KMA’s constant altitude plan position indicator (CAPPI) observation data. The qualitative comparison between the radar observation and the CGAN-predicted rain rates displayed high statistical scores, such as the probability of detection (POD) = 0.8442, false alarm ratio (FAR) = 0.2913, and critical success index (CSI) = 0.6268, in the case of a 1-h prediction for rainfall on September 5, 2019, 15:20 KST. This study demonstrates the capability of the CGAN model for short-term rainfall forecasting. Consequently, the CGAN-generated radar-based rainfall prediction could complement the KMA MAPLE system and be useful in various forecasting applications.

Demonstration of Stormwater Management Technology by Short-Term Rainfall Prediction and Real-Time Runoff Analysis System Using Small X-Band Radar

We developed a system that combines urban area rainfall radar (small X-band, dual-polarization radar), short-term rainfall prediction model, and real-time runoff analysis technology, and the demonstration study was conducted on the drainage districts in Fukui City and Toyama City. We demonstrated the effectiveness of the flood damage, by providing the real-time information on rainfall prediction, water level in sewerage pipes, and inland flood prediction to the operators of drainage pump of stormwater storage pipes, and residents in flood-prone areas. During the study for about two years, it was confirmed that the accuracy of the radar rainfall observation was comparable to that of the X-band dual-polarization Doppler weather radar managed by the Ministry of Land, Infrastructure, Transport and Tourism. In the operation of the drainage pump for the Tsukimiminori Stormwater Storage Pipe in Fukui City, we were able to secure the storage capacity for the next rainfall based on the forecast information by maximizing the drainage capacity of the discharge destination. In addition, it was also confirmed that the residents themselves could secure the lead time for setting up water-stop sandbags and moving their vehicles to higher ground.

Prediction of Short-Time Rainfall Based on Deep Learning

Short-time heavy rainfall is a kind of sudden strong and heavy precipitation weather, which seriously threatens people’s life and property safety. Accurate precipitation nowcasting is of great significance for the government to make disaster prevention and mitigation decisions in time. In order to make high-resolution forecasts of regional rainfall, this paper proposes a convolutional 3D GRU (Conv3D-GRU) model to predict the future rainfall intensity over a relatively short period of time from the machine learning perspective. Firstly, the spatial features of radar echo maps with different heights are extracted by 3D convolution, and then, the radar echo maps on time series are coded and decoded by using GRU. Finally, the trained model is used to predict the radar echo maps in the next 1-2 hours. The experimental results show that the algorithm can effectively extract the temporal and spatial features of radar echo maps, reduce the error between the predicted value and the real value of rainfall, and improve the accuracy of short-term rainfall prediction.

Very short-term rainfall prediction based on radar image learning using deep neural network

Very short-term rainfall prediction based on radar image learning using deep neural network

An empirical ensemble rainfall nowcasting model using multi-scaled analogues

Short-term rainfall prediction with high spatial and temporal resolution is of great importance to rainfall-triggered natural hazards such as landslide, flood, and debris flow. An analogue-based forecasting method may be able to provide short-term radar rainfall predictions, or “nowcasts,” in which the probability distribution of the atmospheric state at a given location in the future is estimated based on a set of past observations. Expensive computations and diverse atmospheric interactions across multiple spatial scales, however, limit the application of this method. This study employs a physically based, empirical ensemble rainfall nowcasting model to obtain ensembles from historical rainfall fields. To do this, we consider meteorological factors, rainfall spatial distributions, and the temporal evolution of rainfall patterns. Cross-correlation is used as a measure of the similarity between the spatial distributions of two rainfall patterns, and the movement rate of rain, with respect to both direction and distance, is used to compare the patterns temporally. A moving window scheme is used to reduce the computational load, with rainfall data providing most reference values. The relationship between different scales and rainfall forecasts is further explored by cascade divisions. While smaller analogue scales indicate better forecasts in this study, the optimal analogue scale will vary for different rainfall events. This study is only a start in incorporating multi-scaled analogue into rainfall nowcasting analysis, and a great deal of more effort is still needed to build a realistic and comprehensive analogue-based rainfall prediction model.

DEVELOPMENT SIMULATION OF AN UNSEASONAL HEAVY RAINFALL EVENT OVER SOUTHERN THAILAND BYWRFROMS COUPLINGMODEL

: A coupled regional climate among atmosphere-ocean coupled model systems is developed using the regional model Weather Research and Forecasting (WRF) and Regional Ocean Modelling System (ROMS). In process of atmospheric, the parameters are used to force the ocean model that included the wind data (Uwind, Vwind), atmospheric pressure data (Patm), relative humidity data (RH), atmosphere surface temperature data (Tair), cloud fraction data (Cloud), precipitation data (Rain), short wave data (SW), long wave data (LW). On the other hand, in process of oceanic, the only parameter of process the oceanic model is used to force the atmosphere model that is Sea Surface Temperature (SST). In this study, to investigate the seasonal prediction and short-term rainfall prediction using WRF and coupled WRFROMS model. The resolution of term rainfall prediction resolution was 36 km× 36 km (domain 1), 12 km × 12 km (domain 2), and 4 km × 4 km (domain 3). The two different microphysics schemes in both cases were Lin scheme and WSM6 scheme. In conclusion, short term rainfall prediction, WRFROMS with WSM6 scheme simulated lower values (RMSE, MAE) and highest value (CORR) than the other experiment. However, the results of the WRFROMS with microphysics improved trend rainfall and SST better than simulation by only WRF model. So, this study found microphysics scheme and SST result from ROMS model that have affected to change results of rainfall prediction. Furthermore, the result of WRFROMS with WSM6 scheme was able to simulate rainfall more accurately when compared with the other experiments by consideration from statistical and quantitative results.

GNSS-based PWV Application for Short Term Rainfall Prediction in Mountainous Region

Rainfall in the mountainous region has unique characteristics related to time-varying and spatial distribution. In Mt. Merapi region, located at the border of Special Region of Yogyakarta Province and Central Java Province, Indonesia, rainfalls are typically classified as the deep-convective type, which occurs in a short period with high intensity. Therefore, short term rainfall predictions in a proper way remain challenging tasks. The use of remote monitoring instruments such as the GNSS (Global Navigation Satellite System) is believed to provide a better measurement accuracy through the identification of water vapor variation in the process of deep convection weather. GNSS observes the geodetic position of the GNSS antenna or receiver while it broadcasts microwave signals continuously through the atmosphere to the ground-based receivers. As they travel through the atmosphere, the microwave signals are mostly influenced by ionospheric and neutral atmospheric effects, which cause some delays. By using a sufficiently dense network of GNSS receivers, the impact of the neutral atmosphere delay can be estimated as a by-product of the geodetic processing. These delays can be regarded as an integrated water vapor along the path, namely Precipitable Water Vapor (PWV), which can be indirectly measured by Zenith Total Delay (ZTD). By studying the relationship between time-varying PWV and rainfall, it can be found that the PWV level increases sharply before raining. Through the deployment of GNSS receivers, the spatial feature of rainfall characteristics is also depicted. The initial results showed that the increase of PWV is strongly correlated to rainfall occurrence based on the rain gauge measurement around Mt. Merapi region. The results show that the correct forecasted rate is about 47%-62% with the PWV increment time is three hours.

Fundamental Study of Real-time Short-term Rainfall Prediction System in Watershed: Case Study of Kinu Watershed in Japan

This study focuses on the improvement of the accuracy and construction of a real-time, short-term rainfall prediction system in watershed, comparing the prediction results with radar observed rainfall intensity, to provide a reliable tool for disaster and water resource management. The short-term prediction of rainfall is very important for hydrologie forecasting for watershed with a short response time, especially under the global warming and extreme weather. Last September 9th (2015), heavy rain came a day after Tropical Storm Etau and triggered widespread flooding in Kinu Watershed. We use the Weather Research and Forecasting Model (WRF) to confirm the uncertainty of the rain prediction in Kinu watershed first, then will modify the WRF to suit the 6h, 12h, 24h rainfall prediction in Kinu watershed. Also, radar rainfall data will be employed to correct the WRF prediction in real-time meaning. Finally, we will check the possibility of general application of the system in other watersheds. We now are working on reproducing the heavy rain in Kinu watershed last September. We tried different microphysics, which describes the formation of cloud, to find one to reproduce the heavy rain in this case. The coverage, location of rain bands and distribution of heavy rainfall are applied to comparisons between calculations and observations by C-band radar. We found no Microphysics does well in this case and the area of heavy rain is underestimated by WRF, but the accumulated precipitation is overestimated. Detail insight at microphysics and modification are required.

Application of an X-Band Multiparameter Radar Network for Rain-Based Urban Flood Forecasting

AbstractUrban floods can occur immediately after heavy rainfall depending on the characteristics of the urban environment. Therefore, earlier warnings, even by 5–10 min, are crucial in terms of human safety. In this case study, an X-band multiparameter (X-MP) radar network, which has a 1-min temporal and 250-m spatial resolution, was employed to forecast urban floods in Kobe, Japan. The X-MP radar data can be used for flood prediction when coupled with urban flood alert criteria nomographs. Thus, a specific objective of this study was to develop a flood alert criteria nomograph for the Toga River in Kobe, Japan. To evaluate its applicability, flood events were reproduced using the nomograph and the data were compared with observed rain gauge and radar rainfall data of actual flooding events. Additionally, flood occurrences were predicted using the nomograph and forecasted X-MP radar rainfall data using a short-term rainfall prediction method. It was demonstrated that the method is accurate and can provide...

Improvement of rainfall and flood forecasts by blending ensemble NWP rainfall with radar prediction considering orographic rainfall

Summary Many basins in Japan are characterized by steep mountainous regions, generating orographic rainfall events. Orographic rainfall may cause localized heavy rainfall to induce flash floods and sediment disasters. However, the accuracy of radar-based rainfall prediction was not enough because of the complex geographical pattern of the mountainous areas. In order to reduce damage due to localized heavy rainfall, characteristics of orographic rainfall must be identified into a short-term rainfall prediction procedure. The accuracy of radar-based rainfall prediction performs best for very short lead time, however the accuracy of radar prediction rapidly decreases with increasing lead times. At longer lead times, higher accuracy QPFs are produced by Numerical Weather Prediction (NWP) models, which solve the dynamics and physics of the atmosphere. This study proposes hybrid blending system of ensemble information from radar-based prediction and numerical weather prediction (NWP) to improve the accuracy of rainfall and flood forecasting. First, an improved radar image extrapolation method, which is comprised of the orographic rainfall identification and the error ensemble scheme, is introduced. Then, ensemble NWP outputs are updated based on mean bias of the error fields considering error structure. Finally, the improved radar-based prediction and updated NWP rainfall considering bias correction are blended dynamically with changing weight functions, which are computed from the expected skill of each radar prediction and updated NWP rainfall. The proposed method is verified temporally and spatially through a target event and is applied to the hybrid flood forecasting for updating with 1 h intervals. The newly proposed method shows sufficient reproducibility in peak discharge value, and could reduce the width of ensemble spread, which is expressed as the uncertainty, in the flood forecasting. Our study is carried out and verified using the largest flood event by typhoon ‘Talas’ of 2011 over the two catchments, which are Futatsuno (356.1 km2) and Nanairo (182.1 km2) dam catchments of Shingu river basin (2360 km2), which is located in the Kii peninsula, Japan.

Understanding, management and modelling of urban hydrology and its consequences for receiving waters: A state of the art

Urban hydrology has evolved to improve the way urban runoff is managed for flood protection, public health and environmental protection. There have been significant recent advances in the measurement and prediction of urban rainfall, with technologies such as radar and microwave networks showing promise. The ability to predict urban hydrology has also evolved, to deliver models suited to the small temporal and spatial scales typical of urban and peri-urban applications. Urban stormwater management increasingly consider the needs of receiving environments as well as those of humans. There is a clear trend towards approaches that attempt to restore pre-development flow-regimes and water quality, with an increasing recognition that restoring a more natural water balance benefits not only the environment, but enhances the liveability of the urban landscape. Once regarded only as a nuisance, stormwater is now increasingly regarded as a resource. Despite the advances, many important challenges in urban hydrology remain. Further research into the spatio-temporal dynamics of urban rainfall is required to improve short-term rainfall prediction. The performance of stormwater technologies in restoring the water balance and in removing emerging priority pollutants remain poorly quantified. All of these challenges are overlaid by the uncertainty of climate change, which imposes a requirement to ensure that stormwater management systems are adaptable and resilient to changes. Urban hydrology will play a critical role in addressing these challenges.