Analysis, characterization, prediction, and attribution of extreme atmospheric events with machine learning and deep learning techniques: a review

Theoretical and Applied Climatology, Год журнала: 2023, Номер 155(1), С. 1 - 44

Опубликована: Авг. 28, 2023

Abstract Atmospheric extreme events cause severe damage to human societies and ecosystems. The frequency intensity of extremes other associated are continuously increasing due climate change global warming. accurate prediction, characterization, attribution atmospheric is, therefore, a key research field in which many groups currently working by applying different methodologies computational tools. Machine learning deep methods have arisen the last years as powerful techniques tackle problems related events. This paper reviews machine approaches applied analysis, most important extremes. A summary used this area, comprehensive critical review literature ML EEs, provided. has been extended rainfall floods, heatwaves temperatures, droughts, weather fog, low-visibility episodes. case study focused on analysis temperature prediction with DL is also presented paper. Conclusions, perspectives, outlooks finally drawn.

Язык: Английский

---

Interpretable machine learning for weather and climate prediction: A review

Atmospheric Environment, Год журнала: 2024, Номер 338, С. 120797 - 120797

Опубликована: Сен. 12, 2024

Язык: Английский

---

Deployment of 3D-Conv-LSTM for Precipitation Nowcast via Satellite Data

Опубликована: Май 3, 2024

Язык: Английский

---

Advances and prospects of deep learning for medium-range extreme weather forecasting

Geoscientific model development, Год журнала: 2024, Номер 17(6), С. 2347 - 2358

Опубликована: Март 21, 2024

Abstract. In recent years, deep learning models have rapidly emerged as a stand-alone alternative to physics-based numerical for medium-range weather forecasting. Several independent research groups claim developed forecasts that outperform those from state-of-the-art models, and operational implementation of data-driven appears be drawing near. However, questions remain about the capabilities with respect providing robust extreme weather. This paper provides an overview developments in field scrutinises challenges events pose leading models. Lastly, it argues need tailor forecast proposes foundational workflow develop such

Язык: Английский

---

Rainfall Prediction Rate in Saudi Arabia Using Improved Machine Learning Techniques

Water, Год журнала: 2023, Номер 15(4), С. 826 - 826

Опубликована: Фев. 20, 2023

Every farmer requires access to rainfall prediction (RP) continue their exploration of harvest yield. The proper use water assets, the successful collection water, and pre-growth construction all depend on an accurate assessment rainfall. heavy rain provision information regarding natural catastrophes are two most challenging factors in this regard. In twentieth century, RP was methodically technically complicated issue worldwide. Weather may be used calculate analyse behaviour weather with unique features determine patterns at exact locale. To end, a variety methodologies have been intensity Saudi Arabia. classification methods data mining (DM) approaches that estimate both numerically categorically can achieve RP. This study, which DM approaches, achieved greater accuracy than conventional statistical methods. study conducted test efficacy several machine learning (ML) for forecasting rainfall, utilising southern Arabia’s historical obtained from live database comprises various meteorological variables. Accurate crop yield predictions crucial would undoubtedly assist farmers. While engineers developed analysis systems whose performance relies connected factors, these seldom despite potential precise forecasts. For reason, agricultural should make impact drought difficult forecast there is need careful preparation choice, planting window, motive, storage space. relevant characteristics required predict precipitation were identified ML approach utilised innovative method whether predicted will regular or heavy. outcomes different methodologies, including accuracy, error, recall, F-measure, RMSE, MAE, evaluate metrics. Based evaluation, it determined DT provides highest level accuracy. Function Fitting Artificial Neural Network classifier (FFANN) 96.1%, higher any other classifiers currently database.

Язык: Английский

---

Enhanced Troposphere Tomography: Integration of GNSS and Remote Sensing Data With Optimal Vertical Constraints

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Год журнала: 2024, Номер 17, С. 3701 - 3714

Опубликована: Янв. 1, 2024

This paper explores the enhancement of Global Navigation Satellite Systems (GNSS) tropospheric tomography by integrating remote sensing data and employing various vertical constraints. Wet refractivity modeling, critical for understanding atmospheric dynamics, has shown promising advancements. Leveraging from Ocean Land Color Instrument (OLCI), this research addresses issue empty voxels that impede GNSS-based due to satellite receiver geometries. Incorporating sensors mitigates voxels, enhancing retrieval accuracy water vapor. study evaluates constraint functions in tomography, presenting eight schemes utilize GNSS OLCI data, highlighting their capacity fill without relying on empirical horizontal Results highlight superiority using observations accuracy. Validation against radiosonde measurements Weather Research Forecasting (WRF) model outputs affirms reliability approach. Integrating with reduces average Root Mean Square Error (RMSE) approximately 27%, Gaussian function exhibiting superior performance.

Язык: Английский

---

Hybrid neural network-aided strong wind speed prediction along rail network

Journal of Wind Engineering and Industrial Aerodynamics, Год журнала: 2024, Номер 252, С. 105813 - 105813

Опубликована: Июль 4, 2024

Язык: Английский

---

Machine Learning-Based Wet Refractivity Prediction through GNSS Troposphere Tomography for Ensemble Troposphere Conditions Forecasting

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 18

Опубликована: Янв. 1, 2024

Язык: Английский

---

Time-Series Embeddings from Language Models: A Tool for Wind Direction Nowcasting

Journal of Meteorological Research, Год журнала: 2024, Номер 38(3), С. 558 - 569

Опубликована: Июнь 1, 2024

Abstract Wind direction nowcasting is crucial in various sectors, particularly for ensuring aviation operations and safety. In this context, the TELMo (Time-series Embeddings from Language Models) model, a sophisticated deep learning architecture, has been introduced work enhanced wind-direction nowcasting. Developed by using three years of data multiple stations complex terrain an international airport, incorporates horizontal u (east–west) v (north–south) wind components to significantly reduce forecasting errors. On day with high variability, achieved mean absolute error values 5.66 2-min, 10.59 10-min, 14.79 20-min forecasts, processed within swift 9-ms/step timeframe. Standard degree-based analysis, comparison, yielded lower performance, emphasizing effectiveness components. contrast, Vanilla neural network, representing shallow-learning approach, underperformed all analyses, highlighting superiority methodologies efficient capable accurately air traffic operations, less than 20° 97.49% predictions, aligning recommended thresholds. This model design enables its applicability across geographical locations, making it versatile tool global meteorology.

Язык: Английский

---

A Data‐Driven Approach to Assess the Impact of Climate Change on a Tropical Mangrove in India

Journal of Geophysical Research Biogeosciences, Год журнала: 2024, Номер 129(8)

Опубликована: Авг. 1, 2024

Abstract As a potential carbon sink, mangroves play an important role in climate mitigation. India houses several major global mangrove patches, which remain vulnerable to change. The ecosystem‐atmosphere CO 2 exchange is most accurately measured by the eddy covariance method, whereas satellites provide biophysical parameters for wider area. In present study, Sentinel‐2 satellite data used map land cover types Pichavaram forest and identify two dominant species ( Rhizophora spp. Avicennia marina ), indicated more than 95% classification accuracy. We years (2017 2018) of situ gross primary productivity (GPP) leaf area index (LAI) measurements rectified Moderate Resolution Imaging Spectroradiometer (MODIS) GPP LAI products from 2010 2018. modified MODIS were develop machine learning models, that is, Random Forest (RF) Extreme Gradient Boosting (XGBoost) study influence on productivity. RF model R = 0.85 root mean square error (RMSE) 0.2) outperformed XGBoost 0.75 RMSE 0.26) was project impact change extreme scenarios, namely SSP1‐1.26 SSP5‐8.5. increases decreases future during wet dry periods, respectively. Overall, projected reduction 3.73%–20.3% 2050 2060 4.82%–28.15% 2090 2100, compared its current average (from 2018).

Язык: Английский

---