Adaptive precipitation nowcasting using deep learning and ensemble modeling

Amini, Amirmasoud; Dolatshahi, Mehri; Kerachian, Reza

doi:10.1016/j.jhydrol.2022.128197

Cited by 29 publications

(8 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Third, in order to build models separately for each circulation pattern, we divide the input arrays into 6 groups based on the SOM results. vector that is used as the input to the LSTM layer (Amini et al 2022). Among them, the kernel size of the first convolutional layer is set to 32× 3 × 3, where 32 is the output channel number, and 3 × 3 is the size of the kernel.…”

Section: Cnn-lstm Modelmentioning

confidence: 99%

Statistical post-processing of precipitation forecasts using circulation classifications and spatiotemporal deep neural networks

Zhang

Liang

et al. 2023

Preprint

View full text Add to dashboard Cite

Abstract. Statistical post-processing techniques are widely used to reduce systematic biases and quantify forecast uncertainty in numerical weather prediction (NWP). In this study, we propose a method to correct the raw daily forecast precipitation by combining large-scale circulation patterns with local spatiotemporal information such as topography and meteorological factors. Particularly, we first use the self-organizing map (SOM) model to classify large-scale circulation patterns for each season, then build the convolutional neural network (CNN) model to extract spatial information (e.g., elevation, specific humidity, and mean sea level pressure) and long short-term memory network (LSTM) model to extract time series (e.g., t, t-1, t-2), and finally correct local precipitation for each circulation pattern separately. Furthermore, the proposed method (SOM-CNN-LSTM) is compared with other benchmark methods (i.e., CNN, LSTM, and CNN-LSTM) in the Huaihe River basin with a lead time of 15 days from 2007 to 2021. The results show that the proposed SOM-CNN-LSTM post-processing method outperforms other benchmark methods for all lead times and each season with the largest correlation coefficient improvement (32.30 %) and root mean square error reduction (26.58 %). Moreover, the proposed method can effectively capture the westward and northward movement of the western Pacific subtropical high (WPSH), which impacts the basin's summer rain. The results illustrate that incorporating large-scale circulation patterns with local spatiotemporal information is a feasible and effective post-processing method to improve forecasting skills, which would benefit hydrological forecasts and other applications.

show abstract

Section: Cnn-lstm Modelmentioning

confidence: 99%

Statistical post-processing of precipitation forecasts using circulation classifications and spatiotemporal deep neural networks

Zhang

Liang

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Those studies have proven the efficacy of deep learning models in generating fake satellite image series or single images that are manually hard to distinguish from the target image. Based on those findings, advancements have been made in predicting meteorological variables such as precipitation (Amini et al, 2022;Ko et al, 2022); deriving secondary products such as land cover and vegetation conditions (Chowdhury et al, 2022;Kladny et al, 2022;Toker et al, 2022); and generating high spatial and temporal resolution images through image fusion that involves multiple image sources .…”

Section: Introductionmentioning

confidence: 99%

MA-SARNet: A one-shot nowcasting framework for SAR image prediction with physical driving forces

Li¹,

Xiang²,

Demiray³

et al. 2023

Preprint

View full text Add to dashboard Cite

Remote sensing imagery is one of the most widely used data sources for large-scale earth observations with consistent spatial and temporal quality. However, the current usage scenarios of remote sensing images are largely limited to retrospective tasks as they can only capture existing phenomena. This study proposes MA-SARNet, a one-shot forecasting framework built with a modified MA-Net structure and ResNet50 as the backbone, to predict the backscatter values of Synthetic-Aperture Radar (SAR) images using the previous observations, meteorological, and geomorphic data layers as input. The model was trained, validated, and tested with SAR images collected during the catastrophic 2019 Midwest U.S. Floods that affected several states on the Missouri and Mississippi River tributaries. Compared to the benchmark performance, model predictions show an increase of 31.9% and 17.8% for the mean and median AAI (Assemble Accuracy Index) and an increase of 37.9% and 15.1% for the mean and median NSE (Nash-Sutcliffe Efficiency) on the test set. Results showed that the flood extent derived from backscatter predictions is more robust against misclassifications caused by pixel-level noise compared to the flood map derived using the real backscatters and those from two additional flood map repositories. Results from spatial and temporal robustness tests demonstrate that the model has sufficient generalization potential for real-time, physically informed, deep learning-based earth surface prediction tasks to facilitate fast response to and mitigation for future floods.

show abstract

“…Amini et al. (2022) employed DNNs to forecast rainfall with a lead‐time of 5 min using a univariate forecasting framework. The outputs of the DNNs are combined with the NWP models through three ensemble models to increase the accuracy of predictions.…”

Section: Introductionmentioning

confidence: 99%

“…They concluded that boosting approaches such as adaboost and high gradient boosting could outperform dagging, stacking, and bagging strategies in hydrological applications. Amini et al (2022) employed DNNs to forecast rainfall with a lead-time of 5 min using a univariate forecasting framework. The outputs of the DNNs are combined with the NWP models through three ensemble models to increase the accuracy of predictions.…”

mentioning

confidence: 99%

Effects of Automatic Hyperparameter Tuning on the Performance of Multi‐Variate Deep Learning‐Based Rainfall Nowcasting

Amini

Dolatshahi

Kerachian

2023

Water Resources Research

Self Cite

View full text Add to dashboard Cite

Climate change has increased the importance of rainfall forecasting, because it has changed the duration and frequency of floods and increased their magnitude Ashrafi, Khoie, et al., 2022). As a result of inundation, landslides, and debris flows in urban areas, heavy rainfall often leads to fatalities and significant financial damages (Seo et al., 2014). Furthermore, overflow resulting from heavy rainfall is a primary contributor to the transport of water pollutants (Lintern et al., 2018). Various preventive measures have been developed, including flood warning systems, to alleviate damages caused by rainfall and enable urban flood management infrastructures to operate more efficiently. Therefore, accurate and reliable rainfall forecasting is essential in urban areas (Poornima & Pushpalatha, 2019). Reliable rainfall nowcasting is a challenging and complex undertaking since rainfall features are influenced by various variables, such as pressure, temperature, relative humidity, and wind speed (Govindaraju, 2000). Due to the difficulty in obtaining meteorological data and the complexity of estimating relationships among variables, physical models are not usually suitable for rainfall nowcasting. An effective alternative is to use empirical models to determine the relationship between inputs like temperature and relative humidity and outputs like rainfall (Nasseri et al., 2008). To ensure optimal performance for urban drainage infrastructures, rainfall nowcasting frameworks should be able to make accurate forecasts with different lead times. Researches on rainfall forecasting and its influencing factors have been conducted for decades, and many approaches have been proposed. There are various traditional forecasting models in the literature, which have been developed using classic time series algorithms such as linear regression, autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA), support vector regression (SVR), artificial neural networks (ANNs), numerical weather predictions (NWPs), and Fuzzy logic algorithm (

show abstract

Adaptive precipitation nowcasting using deep learning and ensemble modeling

Cited by 29 publications

References 29 publications

Statistical post-processing of precipitation forecasts using circulation classifications and spatiotemporal deep neural networks

Statistical post-processing of precipitation forecasts using circulation classifications and spatiotemporal deep neural networks

MA-SARNet: A one-shot nowcasting framework for SAR image prediction with physical driving forces

Effects of Automatic Hyperparameter Tuning on the Performance of Multi‐Variate Deep Learning‐Based Rainfall Nowcasting

Contact Info

Product

Resources

About