Performance Comparison between Deep Learning and Optical Flow-Based Techniques for Nowcast Precipitation from Radar Images

Marrocu, Marino; Massidda, Luca

doi:10.3390/forecast2020011

Cited by 23 publications

(18 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Among various network architectures applied to the weather nowcasting problem, we opted for the PredNet model. This paper is the evolution of a previous study [22] in which we analyzed the performance of the PredNet architecture, highlighting its excellent performance in merely deterministic terms, but also its limitations and the need for a revisiting in probabilistic terms. As specified, the architecture is borrowed from computer vision, likewise the optical flow methodology, which represents the state-of-theart architecture in the field.…”

Section: Prednetmentioning

confidence: 99%

“…The basic idea of this work is to verify the feasibility and effectiveness of an approach that, by overcoming these limitations, manages to fully exploit the unquestionable predictive potential demonstrated by the neural networks, and by the PredNet [21,22] in particular. The chosen operational model is based on the interpretation of the neural network prediction as an optimal estimate of the mean of an ensemble of nowcasts which spread in turn is representative of the associated uncertainty.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Coupling a Neural Network with a Spatial Downscaling Procedure to Improve Probabilistic Nowcast for Urban Rain Radars

Marrocu

Massidda

2022

Forecasting

Self Cite

View full text Add to dashboard Cite

Rainfall forecasting plays a key role in mitigating environmental risks in urban areas, which are subject to increasing hydrogeological risk due to transformations in the urban landscape. We present a new technique for probabilistic precipitation nowcasting by generating an ensemble of equiprobable forecasts, which is especially useful for weather radars with limited spatial range, and that can be used operationally on devices with low computational capacity. The ensemble members are obtained by a novel stochastic noise generation process, consistent with the spatial scales not resolved by the prediction model, which allows continuous downscaling of the output of a deep generative neural network. Through an in-depth analysis of the results, for precipitation accumulated in the first hour, measured by all the most robust skill indicators, extended to an entire year of data at 5-min time resolution, we demonstrate that the proposed procedure is able to provide calibrated, reliable, and sharp ensemble rainfall forecasts with a quality comparable or superior to the state-of-the-art classical alternative optical flow technique. The ensemble generation procedure we propose is sufficiently general to be applied in principle to other deterministic architectures as well, thus enabling their generalization in probabilistic terms.

show abstract

Section: Prednetmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Coupling a Neural Network with a Spatial Downscaling Procedure to Improve Probabilistic Nowcast for Urban Rain Radars

Marrocu

Massidda

2022

Forecasting

Self Cite

View full text Add to dashboard Cite

show abstract

“…is the number of the next iteration, and W r is the weight of the r ∈ {OP, B3D, f c} component. From Equations (8) and (9), the weight is updated by taking the partial derivative of the loss. Next, we introduce the partial derivative processes of several modules, including the f c, OP , and B3D modules, which are expressed as follows:…”

Section: A Sopnetmentioning

confidence: 99%

“…Moreover, satellite images are commonly used, and various CNNs have been applied to satellite images to identify and forecast precipitation intensity. For example, variants of popular CNNs, such as ResNet [6], U-Net [7], and GAN [8], have been used to analyze satellite and infrared images. Although standard CNNs can employ a 2D convolutional operator to extract and generate features of a series of images, standard CNNs cannot identify time-series relationships between images.…”

Section: Introductionmentioning

confidence: 99%

SOPNet Method for the Fine-Grained Measurement and Prediction of Precipitation Intensity Using Outdoor Surveillance Cameras

Lin

Yang

2020

IEEE Access

View full text Add to dashboard Cite

Surveillance cameras have been widely used in urban environments and are increasingly used in rural ones. Such cameras have mostly been used for security, but they can be applied to the problem of furnishing fine-grained measurements and predictions of precipitation intensity. In this study, we formulated a stacked order-preserving (OP) learning framework to train a network using time-series data. We constructed an OP module, which uses a three-dimensional (3D) convolution operation to extract features with spatial and temporal information and that are associated with ConvLSTM; this feature extraction is used to learn the short-term and OP time-series relationships between features. Furthermore, the OP modules are stacked to form a stacked OP network (SOPNet), which strengthens the relationship between features in long-term time-series image sequences. This SOPNet can be use to obtain fine-grained measurements and predictions of precipitation intensity from images captured by outdoor surveillance cameras. Our main contributions are threefold. First, the SOPNet strengthens the short-term and longterm time-series relationship between features. Second, the SOPNet simultaneously examines spatial and temporal information to measure and predict precipitation intensity. Third, we constructed a precipitation intensity database based on optical images captured by outdoor surveillance cameras. We experimentally evaluated our proposed architecture using our self-collected data set. We found that SOPNet yields better performance and greater accuracy relative to its well-known state-of-the-art counterparts with respect to various metrics.

show abstract

“…CNN is responsible for extracting spatial information. The deep learning extrapolation approaches usually perform better than the optical flow methods [13], because they do not have the unreasonable constant assumption and they can effectively leverage the valuable historical observations.…”

Section: Introductionmentioning

confidence: 99%

A Novel LSTM Model with Interaction Dual Attention for Radar Echo Extrapolation

Luo

Wen

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

The task of precipitation nowcasting is significant in the operational weather forecast. The radar echo map extrapolation plays a vital role in this task. Recently, deep learning techniques such as Convolutional Recurrent Neural Network (ConvRNN) models have been designed to solve the task. These models, albeit performing much better than conventional optical flow based approaches, suffer from a common problem of underestimating the high echo value parts. The drawback is fatal to precipitation nowcasting, as the parts often lead to heavy rains that may cause natural disasters. In this paper, we propose a novel interaction dual attention long short-term memory (IDA-LSTM) model to address the drawback. In the method, an interaction framework is developed for the ConvRNN unit to fully exploit the short-term context information by constructing a serial of coupled convolutions on the input and hidden states. Moreover, a dual attention mechanism on channels and positions is developed to recall the forgotten information in the long term. Comprehensive experiments have been conducted on CIKM AnalytiCup 2017 data sets, and the results show the effectiveness of the IDA-LSTM in addressing the underestimation drawback. The extrapolation performance of IDA-LSTM is superior to that of the state-of-the-art methods.

show abstract

Performance Comparison between Deep Learning and Optical Flow-Based Techniques for Nowcast Precipitation from Radar Images

Cited by 23 publications

References 22 publications

Coupling a Neural Network with a Spatial Downscaling Procedure to Improve Probabilistic Nowcast for Urban Rain Radars

Coupling a Neural Network with a Spatial Downscaling Procedure to Improve Probabilistic Nowcast for Urban Rain Radars

SOPNet Method for the Fine-Grained Measurement and Prediction of Precipitation Intensity Using Outdoor Surveillance Cameras

A Novel LSTM Model with Interaction Dual Attention for Radar Echo Extrapolation

Contact Info

Product

Resources

About