ML forecasts for a point, station, city or small area, mostly from observed time series (rainfall, temperature, wind, fog, visibility), without a gridded NWP-like model. Use post-processing instead when the model corrects NWP output.
Search and filter this topic Plain list on GitHub
Genetic-fuzzy GENOFIS model beats ANFIS and SVM for Kigali rainfall prediction using 1991–2023 gauge data, combined with downscaled climate-change scenarios.
Compares logistic regression, MLP and SVM for next-day rain prediction, showing how accuracy changes with training data volume.
Uses an artificial neural network on historical weather variables to forecast temperature, with a framework claimed to outperform traditional methods.
Applies SMOTE and GAN augmentation with federated learning to classify rare weather events in imbalanced tabular station data; no headline numbers given.
A neural network trained on tri-band VHF radiometric sounding data forecasts short-term ground-layer temperature, humidity and precipitation from daily measurements.
A Conv1D-Transformer predicts 3-hour rainfall levels in northern Xinjiang from 24 hours of station data, a regional semi-arid case study.
A neural network predicts parametric wind speed distributions, with the M-Rice distribution outperforming alternatives and ERA5 neighbour data improving 6-hour station forecasts.
IoT sensor networks with unsupervised learning give hyperlocal short-term forecasts of temperature, pressure and humidity, and flag anomalous weather patterns in real time.
Systematic review of 23 studies finds deep learning outperforms traditional methods for short-term wind speed and direction forecasting, with mean MAE of 0.56 m/s.
A branched deep CNN forecasts Paris haze and low-visibility occurrence from multi-decadal meteorological maps, with blind forecasts for 2021 and 2022 showing reasonable skill.
Machine learning models forecast short-term daily precipitation across contiguous US climate regions; the abstract is missing, so the method and gains are unverified.
Random forest and LSTM estimate Sofia airport visibility from 11 surface meteorological variables, with regression post-processing improving skill for fog detection.
Ant colony optimisation tunes decision-tree and neural data-mining models for weather prediction, claiming accuracy gains without specific quantitative results.
A variational mode decomposition plus LSTM hybrid forecasts next-hour wind speed in Saudi Arabia with 61–74% skill, outperforming other hybrid and deep models.
Stacks linear regression, ridge, SVR, random forest and LightGBM to forecast next-day temperatures at Cox's Bazar, Bangladesh, with R² up to 0.965.
MODWT wavelet decomposition combined with machine learning improves daily rainfall forecasts three days ahead at three Turkish stations, judged by MSE and Nash-Sutcliffe efficiency.
Live experiments with XGBoost, MLP, Transformer and LSTM nowcast temperature, wind and cloud cover; a variance-horizon method cuts compute by over 50% and concept-drift retraining cuts it by up to 90%.
Compares XGBoost, SVM and Random Forest for daily maximum and minimum temperature at Visakhapatnam; no single model dominates.
A stacking ensemble of machine learning models estimates annual rainfall for Karnataka state using 1901–2015 station records.
LSTM, ARIMA, SARIMA and Prophet are applied to historical temperature series for major cities to forecast future temperature trends; methods are standard and evaluation is limited.
CNN-LSTM model predicts rainfall from meteorological variables and is claimed to beat statistical methods, though the abstract gives no quantitative results.
Stacked LSTM forecasts hourly Denpasar weather (temperature, pressure, humidity, wind) from 20 years of station data, reporting average R² of 0.87 and RMSE of 0.035.
LSTM outperforms Dynamic Mode Decomposition for daily rainfall forecasting from 118 years of station records in North-East India.
Feature engineering and selection for airport wind nowcasts 2–20 min ahead; wind speed MAPE reaches 0.35–0.63%, but direction remains hard.
An Arduino and Android weather station uses a backpropagation neural network on temperature and humidity, reaching a 78.6% forecast success rate on 14 trials.
Stacked-LSTM beats Transformer and SVR for precipitation at 56 Jilin stations; Gaussian noise augmentation improves robustness and zero-rain prediction.
Compares genetic algorithm, differential evolution and particle swarm tuning of ANN, LSTM and GRU hyperparameters for short-term weather forecasting, with no headline numeric gain given.
Proposes a granular sigmoid extreme learning machine applied to weather forecasting; abstract unavailable, so details are unverified.
Proposes ANN-based weather forecasting apps claiming high accuracy up to 1460 days ahead, but methods and validation are weakly described.
Random Projection dimensionality reduction with Decision Tree and Naive Bayes on web-scraped weather data is claimed to improve forecast accuracy.
Framework combining random forest, linear and decision-tree regressors with feature selection for general weather forecasting; no quantitative results reported.
LSTM with a Weibull wind-speed distribution, combined with NWP, forecasts Montreal wind 48 h ahead, cutting RMSE 13–31% versus plain LSTM.
Reviews AI and deep learning rainfall forecasting studies for Thailand from 2000 to 2022, concluding hybrid ANN-wavelet models are most promising.
Neural network combined with an improved GPT3 model forecasts Japanese precipitation from GNSS-derived atmospheric data; abstract unavailable, so details are limited.
Genetic-algorithm feature extraction for Maharashtra rainfall prediction, with claimed efficiency gains over traditional systems but no quantitative results.
Rough set reducts and association rules create binary features that improve classifiers for Kenyan farm weather from Wunderground data; no numbers in abstract.
Informer transformer fuses station observations with LDAPS NWP for 6–336 h temperature forecasts, lowering RMSE by 0.25 °C versus CNN-BLSTM.
Multiple linear regression predicts Yogyakarta daily rainfall from temperature, humidity, wind and sunshine, reaching an R2 of only 12.99%.
Bi-LSTM models forecast London temperature and humidity 24–72 h ahead from station history, reaching about 1.45 °C RMSE at Kew Gardens.
Describes a sensor and machine-learning system for real-time local weather monitoring and forecasting; little technical detail or evaluation given.
Compares ARIMA, exponential smoothing, MLP and Gaussian processes for extreme temperature in five US cities; MLP performs best and projects temperatures to 2030.
AdaBoost ensemble classifies rainfall from 1901–2015 Indian station data, claiming better accuracy than SVM; no numbers given in abstract.
Backpropagation neural network predicts classified daily rainfall at a Lampung station, outperforming SVM and logistic regression on light-rain recall.
Compares ANN, SVM, random forest and KNN for weather-condition classification in Swedish cities; the ANN performs best and beats Weatherstack's own forecasts.
Applies machine learning to predict precipitation in several Chinese regions; no abstract was provided, so the details are unclear.
Proposes a hybrid deep learning weather prediction model; abstract unavailable, so details and reported gains could not be verified.
CNN-LSTM hybrid forecasts aviation visibility, reducing RMSE to 2.68 and MAPE to 13.41% from LSTM alone.
Reviews hybrid machine learning models for rainfall prediction, summarizing methods and their reported accuracy gains.
Boosting ensemble of machine learning models predicts multiple meteorological parameters from measurements; no quantitative results given.
A random forest tuned on station observations forecasts three-hour visibility with about 78% accuracy and a 6% false alarm rate, relevant to aviation.
GUMP combines ML sensor predictions with CFD flow fields to give hyper-local urban wind forecasts and a risk map for drone operations.
Radial basis function neural network forecasts temperature and humidity in Warangal, India, from one year of station data and outperforms other ML models.
IoT and fuzzy-logic system predicts 24-hour farm rainfall from sensors and cut water use 55% in a two-hectare maize trial; ML role is modest.
LSTM and deep LSTM models forecast daily air temperature in Tabriz, Iran, from 1,097 observations, with the deep version reaching RMSE 0.08 °C.
GRU and XGBoost, tuned and deployed on an IoT weather station, forecast local sensor readings and four-day rain; XGBoost wins regression, GRU wins classification.
Non-parametric trend tests and machine learning models analyze and forecast meteorological variables in the Thoubal river watershed, India; the abstract is missing.
GMDH model using lagged sea surface temperature predicts monthly rainfall in Pará, Brazil, with R 0.96 versus 0.82 for a multilayer perceptron.
Describes an IoT weather prediction device built on transformer models; the abstract is missing, so results cannot be verified.
A random forest classifier predicts next-day rainfall in Mumbai from five years of local records, with no benchmark against other methods.
Ensembles combining GEP, SVR and GMDH improve hybrid precipitation forecasts at two Iranian stations, with MSE reduced by 67% at Rasht.
Random forest, Lasso, ridge, elastic net and ARIMA forecast daily rainfall in Beni Mellal-Khenifra, Morocco, with random forest giving the lowest error.
A probabilistic CNN-BiLSTM with multi-head attention forecasts day-ahead wind speed for energy use; the abstract was unavailable, so the claim is limited.
Federated learning with GAN-based oversampling is applied to imbalanced tabular weather data; gains are reported without numbers.
Spatiotemporal deep network for multi-horizon wind prediction; the abstract is missing, so architecture and results are unverified.
XGBoost, ARIMA and LSTM models forecast atmospheric visibility from station time series, with outputs reported as close to observed values.
Applies quantum classifiers (QSVM, QNN, VQC) to 2016–2022 station weather data from a Thai dam, with QSVM reaching 85.3% accuracy.
Uses machine-learning combinations of ENSO, IOD and MJO indices to predict Medan rainfall, finding the SOI plus MJO combination gives the best verification, with MAE of 15.0 mm.
Applies artificial neural networks to predict daily weather for several cities, claiming very low deviations, though without a quantified benchmark in the abstract.
An ANN trained on Nigerian station data reaches over 90% rainfall prediction accuracy, versus 70% for multiple linear regression.
Reviews IoT sensor-based weather monitoring and deep learning forecasting, including self-attention and hybrid CNN-LSTM models, with challenges and future directions.
Ensemble ML models forecast fog and dense fog at Patna airport from 6 h to 1 h ahead using historical synoptic station data for aviation early warning.
Hybrid SOPDEL model combining decomposition, DBN-ELM and LSTM improves long-time-series climate prediction; the abstract gives no quantitative result.
PTCN uses multi-scale causal dilated convolutions and a new MVLoss to improve prediction of small-variance weather variables on WFD2018.
LSTM beats ARIMA on most station weather variables such as temperature, humidity and radiation, and in computation time, though precipitation errors remain highest.
Linear regression on atmospheric state variables forecasts monthly precipitation and is claimed to beat established methods; evaluation detail is limited.
Review of linear regression as a machine learning approach to weather forecasting, with little technical depth in the abstract.
Rough set feature handling combined with deep learning is proposed for precipitation prediction; the abstract is unavailable, so details and results are unverified.
LSTM outperforms feed-forward networks for predicting monthly maximum temperature at Kaba Aye, Myanmar, with MAE of 1.33 degrees.
Nanjing University of Information Science & Technology (NUIST)
A hybrid deep learning model for rainfall in southern Iraq wetlands; the abstract is unavailable, so the details are limited.
Spatiotemporal graph neural network for rainfall forecasting; no abstract was provided, so the assessment rests on the title.
Survey of classical machine learning methods, from SVMs to neural networks, applied to weather prediction, to help researchers choose approaches.
An LSTM forecasts temperature, humidity, rainfall and wind speed from Serang maritime station data, with best RMSE of 0.37 for temperature.
Hybrid ARIMA-RNN models forecast monthly extreme weather in West Sulawesi, Indonesia, with the ARIMA-RNN-Adam model giving the lowest RMSE.
Bayesian-GRU encoder-decoder with multi-head attention forecasts station temperature and humidity for agriculture, reporting R of 0.80–0.90 at three sites.
Compares several machine learning classifiers and regressors for rainfall prediction in southern Saudi Arabia using station meteorological records, with crop-yield motivation.
Backpropagation neural network forecasts daily weather at Radin Inten II airport, Lampung, from three years of station data, reaching 74.36% best test accuracy.
Systematic review of IoT and deep learning weather monitoring and prediction, covering CNN, LSTM, transformer and hybrid models; no new results are reported.
Compares decision trees, random forest, SVM and ANN on historical meteorological variables to estimate rainfall; reports improved accuracy but no specific numbers in the abstract.
Proposes a perceptron feature selection with Kriging gradient-boost classifier for marine weather forecasting; the abstract gives no quantitative results.
Neural networks forecast Nigerian weather parameters from 19 years of NASA data to assess solar PV siting, reaching 96% training correlation.
Massachusetts Institute of Technology (MIT)University of Oklahoma (AI2ES)
Shows that entanglement and superposition layers improve variational quantum circuits for weather forecasting on small datasets, reducing the number of variational layers needed.
Combines hybrid decomposition with adaptive deep learning for weather forecasting; the abstract is unavailable so the classification relies on the title.
A rough-set feature-interaction method feeds a generalized linear model for interpretable station weather prediction at one Kenyan farm station.
Point-scale machine learning models predict frost events 6-48 h ahead in New Mexico, with 6-h RMSE of 1.53-1.72 °C, outperforming HRRR.
Builds a website weather forecast application for Medan City using a Hopfield neural network and BMKG data to disseminate forecasts.
Random forests, decision trees and regression forecast apparent temperature from Dark Sky API data, offering a generic comparison with limited novelty.
A modified fuzzy neural network with attention forecasts location-specific temperature from website data, reporting RMSE about three times lower than recurrent networks.
An LSTM with a site-embedding layer forecasts hourly wind speeds at multiple US East Coast buoys, with 2.09 m/s average RMSE, and works during hurricanes.
Stacked ensemble of ML models on IoT sensor data reports 99.25% accuracy for local weather parameters, and 85% on Australian rainfall. Limited generality.
Decision-tree and ML systems are proposed for smart-city weather modification forecasting, claiming 35% accuracy gains over traditional models on Shaanxi data with limited methodological detail.
LightGBM gradient boosting forecasts daily Seattle precipitation and temperatures from 1948–2017 station records, claimed to outperform other boosting methods.
Proposes an optimal KNN with data pruning for missing values to classify Seattle rainfall; reports better accuracy and lower runtime. Weakly described and narrow.
Compares regression models and ensembles on NASA POWER climate data for precipitation forecasts in southern India, with ensemble E-MLR-RFR performing best.
Wavelet-LSTM-ARIMA hybrid predicts GNSS precipitable water vapour, cutting RMSE by about 55% versus other models over 3 h horizons.
Hybrid CNN-CatBoost predicts solar radiation at 18 Korean locations from weather variables, and adding wind and precipitation improves results at 15 of them.
An LSTM with regression-based feature selection forecasts daily temperature from station variables for aviation safety, cutting RMSE from 4.03 to 2.22.
EEMD-LSTM-ARIMA coupled model forecasts monthly precipitation in Luoyang by modelling high- and low-frequency components separately, outperforming single and combined baselines.
Presents a generic machine learning classifier for weather prediction with a farmer-oriented regional guidance angle; no quantitative results reported.
CNN-Bi-LSTM with spatiotemporal clustering forecasts precipitation delta-18O at 10 German stations, performing best for multi-site multi-year forecasts.
Random forest forecasts and non-parametric trend tests for rainfall and temperature at one Indian tiger reserve; no abstract provided, so scope is limited to local station climate.
Describes a solar-powered IoT weather station with a machine learning model to predict local weather conditions in Mangalpalli; results are not quantified.
CEEMDAN-LSTM hybrid predicts annual precipitation in Zhengzhou city, reaching 2.69% MAPE and R² of 0.9863, outperforming five baseline models.
A hybrid rank-pooling and FFT model with MLP or LSTM heads improves multi-step station wind speed forecasts over single and hybrid baselines using 2010-2020 data.
An LSTM-style deep network forecasts heat index from IoT temperature sensor series and is compared with generalized linear regression; the methods are thinly described.
Survey of machine learning methods for rainfall and weather prediction, arguing that reported accuracy often fails when applied at large scale.
DWFH hybrid uses transductive LSTM for deep weather forecasting; abstract missing, so classification relies on the title alone.
Bayesian network predicts monthly maximum and minimum dustfall in Iraq from station weather data, reporting 75% and 89% precision.
Ensemble GRU-LSTM model predicts road surface temperature 1 to 6 hours ahead from station data, lowering MAE versus CNN-LSTM and SVR baselines.
RLMD-BiLSTM hybrid forecasts solar irradiance at two Indian sites, cutting RMSE by 59–89% against benchmark models; the study is station-based and regional.
ANN classifies monthly and daily rainfall in Java with up to 89% accuracy, then Ordinary Kriging maps six-month rainfall estimates.
Purely machine-learning forecasts of optical turbulence and astroclimate parameters 1-2 hours ahead above the Paranal VLT, benchmarked against autoregressive approaches.
Classification algorithms detect and predict Alpine foehn events from GNSS troposphere products at Altdorf, reaching 70-85% detection and alarm measures.
SVM, tree and KNN models using camera-derived visibility predict 30-minute fog onset at Poprad-Tatry airport with a 0.89 probability of detection.
Generic report applying various machine learning algorithms to improve forecast accuracy for temperature, rain and humidity; no specific method or quantitative result is given.
Deep learning on an edge computing platform with self-powered sensors predicts urban block microclimate indices in two cities, with small errors reported.
Compares decision trees, LMT, M5 model trees, gradient boosting and NARX for rainfall classification from three Kashmir stations; gradient boosting reaches 87.5% accuracy.
Compares LSTM, BiLSTM and GRU for long-term offshore wind speed and wave height forecasts from FINO3 station data, reporting better seasonal response and reliability.
Adaptive deep learning models for weather forecasting; abstract unavailable, so the target scale and gains cannot be verified.
Ensemble of ten regressors forecasts daily rainfall, humidity and temperatures in Senegal, with R² of 0.82–0.93 for the ensemble model.
A backpropagation neural network with a LabVIEW interface is built for weather condition forecasting; the abstract reports no quantitative results.
Random Forest and SVM predict 12-hour snowfall occurrence from meteorological and terrain data, with RF RMSE of 3.2 cm versus 5.1 cm for regression.
Compares linear regression, SVM and decision trees with empirical models for station precipitation, humidity, wind and radiation forecasts in India, reporting MSE 0.1397.
Compares decision tree, k-NN and logistic regression classifiers for weather prediction on a Kaggle dataset, with a Flask demo application; small-scale evaluation.
EMD-decomposed LSTM forecasts daily minimum and maximum weather variables for Shenzhen from 2015–2019 station data, reporting better precision than traditional models.
XGBoost with SHAP predicts forest-interior temperatures from open-field weather data for Brandenburg sites, supporting climate-change ecosystem assessment.
Shows that adding neighbouring station wind data improves 1-6 hour wind speed forecasts by up to 20% RMSE across linear and ML models.
Six machine learning methods predict visibility from ground observations; PCA improves accuracy, and neural networks perform best among the methods tested.
LSTM predicts atmospheric wind and temperature profiles up to 30 km from US radiosondes, matching or beating reanalysis in some cases for infrasound modeling.
Random forest, decision tree, naive Bayes and boosting classify Seattle daily weather from 2012-2015 data; Naive Bayes is best at 84.2% accuracy.
Decision tree regression is applied to weather prediction, with a claimed long-term trend in forecast accuracy. The contribution is thin and not well documented.
Hybrid CNN-LSTM model forecasts temperature, humidity, pressure and wind speed time series, reporting lower errors than machine learning and deep learning baselines.
Bayesian neural networks forecast offshore wind speed and direction at FINO1 from ocean variables, giving calibrated uncertainty for North Sea wind-resource prediction.
Hybrid deep learning model gives multi-step-ahead forecasts of daily maximum temperature and heatwaves; no abstract was provided, so skill figures are unknown.
Hybrid CNN-REGST model predicts daily solar radiation at six Queensland farms from GCM and observation features, reported as more accurate than DL and ML baselines.
Stacked ensemble with CEEMD decomposition is proposed for rainfall prediction; the abstract is missing, so the method and results cannot be verified.
Study of exogenous atmospheric variables as machine learning predictors of wind speed for energy applications; abstract unavailable, so scope is inferred from the title.
Reviews 2013-2017 machine learning studies for rainfall prediction, summarizing data mining approaches and suggesting future directions.
Low-cost station sensors (temperature, humidity, rain and others) feed an ANN and decision tree that classify rain with scores of 0.92 and 0.96.
Reviews machine learning rainfall prediction techniques and public datasets, classifying forecasting models and suggesting future research directions.
Compares regression-style machine learning models on historical station data to predict minimum, maximum and average temperature and precipitation for different locations.
Fuzzy-logic fusion of decision tree, naive Bayes, KNN and SVM classifiers predicts rainfall for Lahore from 12 years of station data.
Hybrid EMD-LSTM with time-dependent cross-correlation forecasts monthly rainfall; no abstract was provided, so skill figures are unknown.
Hybrid flexible rough neural network with genetic algorithm and emotional learning forecasts Tehran station meteorological variables from 2008 to 2012.
Compares logistic regression, tree ensembles and boosting for next-day rainfall classification on ten years of Australian station data with imbalance handling.
Ensemble-learning neural networks interpolate neighbouring station data, terrain and NDVI to predict next-hour minimum temperature and frost at sites without local sensors, detecting 92.55% of frost events.
Collaborative regression and neural forecasts from multiple Mauritius stations via OpenWeather API data reduce MAPE by about 5% versus non-collaborative setups.
Hybrid empirical wavelet transform with Markov-chain-incorporated LSTM for precipitation forecasting; no abstract available to confirm the results.
Hybrid ICEEMDAN decomposition, transformer and MLP error correction for multi-horizon wind speed forecasts at an Indian wind farm; reports very small error metrics.
A tower network fusing multiple data streams gives 11% lower temperature forecast RMSE than yr.no, beating CNN and convLSTM baselines.
XGBoost radiation fog nowcasts are evaluated with persistence and guessing baselines, showing high validation scores can hide operational failure and that blockwise CV inflates skill.
Compares shallow supervised learners for 10 m wind speed forecasting at 32 Italian anemometer sites, finding well-designed shallow models competitive with deep networks.
Multiple linear regression predicts rainfall from Indian weather parameters and claims to beat textbook algorithms, but gives no quantitative results.
Machine learning model for short-term weather prediction in Sri Lanka; the abstract is only journal boilerplate, so the details are unverified.
C4.5 decision tree classifies next-day weather as no rain, rain, light rain or heavy rain from weather variables; a simple classification study.
SSA-tuned least-squares SVM forecasts Amman temperature, pressure and soil humidity 24 to 72 h ahead for smart irrigation; a single-city study.
Multiple linear regression predicts a day's average temperature from past station meteorological data with a reported error of 2.8 °C.
Multi-site collaborative deep learning for wind speed forecasting in valley complex terrain; abstract unavailable, judged from title only.
Applies a neural network autoregression on lagged station rainfall series to forecast monsoon rainfall in Maharashtra, India.
Hybrid M5P-SVR model forecasts station precipitation in northern Bangladesh, reaching R2 up to 0.92 at Sylhet.
Trains a neural network on Viking and Mars Science Laboratory measurements to predict next-day Martian temperature and pressure within 0.5 to 1% accuracy.
BLSTM-attention and BLSTM-Transformer models predict wind speed at seven Qingdao stations, cutting RMSE by about 44% versus ECMWF forecasts.
Nanjing University of Information Science & Technology (NUIST)
Proposes an optimized deep recurrent network for predicting atmospheric and oceanic parameters; the abstract is missing, so the claims cannot be verified.
HiSTGNN uses hierarchical spatio-temporal graphs over station meteorological variables, reducing forecast errors by 4.2% to 11.6% versus baselines.
An ANN trained on Kalpana IR and water-vapour satellite pixel values forecasts 24-hour precipitation at six Western Himalaya stations with a 0.78-0.95 probability of detection.
Applies Bayesian-regularized neural networks to weather forecasting, reporting improved accuracy with more training data but no quantified headline result.
Reviews and proposes a hybrid CNN-LSTM pipeline for multi-step meteorological forecasting, reporting that it beats baselines on MAE, RMSE and R2.
Compares random forest, decision tree and gradient boosting for Iraq weather variables from NASA data; random forest reaches 92% accuracy.
Encoder-decoder LSTM outperforms attention LSTM for daily temperature, humidity and sunshine forecasts in Narmadapuram using 1977-2022 station data.
Compares ridge, random forest, linear regression and decision tree for temperature prediction; the decision tree gets the lowest RMSE of 0.036.
Combines DBSCAN clustering with an ANN to group meteorological parameters and forecast weather in Annaba, Algeria, with no quantified headline result.
Uses artificial neural networks on INMET station climate records to forecast temperature and precipitation, linking forecasts to crop planting schedules.
Two-layer LSTM predicts Bangladesh rainfall from 10 selected features with 97.14% accuracy, outperforming conventional ML baselines; the evaluation is limited.
ANFIS models forecast typhoon rainfall at a Tamsui station up to three hours ahead and are packaged in a mobile app, with best RMSE of 7.9 mm.
Comparison of 12 regression methods for next-day Seoul maximum and minimum temperature finds XGBoost and LightGBM most accurate.
An attention-based encoder-decoder LSTM forecasts short-term surface temperature from multiple meteorological variables, with learned spatial feature attention weights validated against correlations.
LSTM and CNN models using GNSS radio occultation profiles forecast regional wind fields over Beijing-Tianjin-Hebei, with about 1.4 m/s speed error and 1 s inference.
Ensemble of gradient boosting, random forest, Bernoulli Naive Bayes and KNN reaches 95% weather prediction accuracy and runs in under 40 s for 5,000 nodes.
Bird-mating-optimizer-tuned bidirectional GRU for weather forecasting reports RMSE 0.173, better than MLP-ANN at 0.202.
Applies multiple linear regression to Intel lab wireless sensor data to predict humidity, reporting about 11% minimum error.
An artificial neural network using five factors including previous rainfall classifies monthly weather in Bogor, reaching 83.33% accuracy on 2017-2018 test data.
Uses an LSTM to predict temperature, humidity, sunshine and rainfall for Köppen-class agricultural yield estimation, reporting 3.29% MAPE.
Machine learning regression on ten years of airport station data nowcasts visibility and ceiling height up to 1 h, with MAE of about 324 m and 167 m.
Deep ensemble of self-normalizing networks with Gaussian outputs wins top-3 in Yandex Shifts weather temperature regression challenge under distribution shift.
Wavelet packet denoising with CatBoost combines station observations and NWP priors to beat LSTM and random forest for short-term Beijing weather forecasts.
Radial basis function network predicts Malaysian rainfall for tree management, reaching 57% accuracy; a small study with limited scope.
A 1D and 2D CNN combining station observations with NWP output forecasts harbor wind speed one hour ahead, with MAE of 0.3945 m/s.
Ensemble neural networks trained on perturbed observations nowcast wind speed and direction for wind farm output, with better accuracy and runtime than alternatives.
Dense neural networks predict daily max and min 2 m temperatures from radiosonde profiles up to 500 days ahead, beating persistence but not climatology beyond day 2-3.
A CNN classifies rain events and a GRU forecasts hourly rainfall in Thailand, with focal and autoencoder losses. It outperforms ARIMA and CNN-GRU on RMSE in most regions.
Decision tree methods, including M5P, REP Tree and Random Forest, forecast daily maximum and minimum temperature in Delhi from 17 months of data. Random Forest is best.
Fitness-tuned adaptive neural network forecasts Karnataka rainfall and temperature for 2021-2030 and maps effects on five crops; validation is limited.
A machine learning approach to precipitation level prediction and mapping is proposed on a cloud big-data environment. The abstract is truncated and gives no quantitative results.
Compares time-series regression, SARFIMA and neural networks for monthly weather forecasts in Lahore. The ANN with moving-average preprocessing has the lowest RMSE.
Whale Optimization tuning of k-nearest neighbours for weather prediction, reporting better accuracy and lower run time than plain K-NN on an unspecified dataset.
Shallow backpropagation neural networks are tuned to forecast a year of daily weather from historical records, mainly comparing hidden-layer and neuron choices.
Uses machine-learning feature selection on five years of single-station meteorological data to rank which parameters most influence rainfall.
PCA-reduced AdaBoost on station records predicts mid-term temperature with R2 of 0.992, reported as beating other tested models.
An LSTM plus Chebyshev-polynomial network forecasts Yibin City precipitation with lower error than LSTM, SVR, ARIMA and MLP baselines.
Nonlinear autoregressive and NARX neural networks forecast 12 years of Marrakesh temperature, with NARX giving better training and test performance.
A hybrid MCEEMD decomposition and RBF neural network forecasts Taipei daily maximum temperature 7 days ahead, aimed at heatwave risk under warming.
A general review of ANN and CNN approaches to weather prediction, with qualitative claims that ANNs are more efficient and little quantitative evidence.
Data-mining workflow trains an artificial neural network on preprocessed weather records to predict rainfall, with no quantitative results in the abstract.
Feedforward backpropagation networks forecast next-day maximum temperature and sunshine hours in Debre Markos, Ethiopia, with regression above 0.98.
Multi Scale Graph WaveNet combines dilated convolutions with graph convolutions to forecast wind speed at Danish cities, outperforming baselines by 4-5% across horizons.
Compares statistical, machine learning and deep learning models for rainfall forecasting at four Maharashtra talukas using local station records.
UKF-aided LSTM improves short-term aviation wind nowcasts over plain LSTM and quantifies uncertainty with Monte Carlo simulation.
Graph attention network with a learnable adjacency and per-variable attention, combined with LSTM, forecasts wind speed at several Danish and Dutch cities from multivariate weather data.
Regularized classifiers in a Spark streaming pipeline predict winter precipitation type (rain, freezing rain, snow) from ASOS station data.
GRU-CNN models forecast typhoon winds and wave heights 1 to 6 h ahead at Taiwan buoys using station data and radar reflectivity images.
Generic overview of data mining and machine learning for weather prediction; it claims simpler models can match physical methods but gives no quantitative results.
A CEEMD-LSTM hybrid predicts monthly precipitation in Zhengzhou with NSE 0.95, beating LSTM and BP-based baselines on station data.
Integrates observed cloud characteristics with machine learning to improve prediction of surface solar irradiance variability at a site.
Retracted article proposing an improved neural network for coastal rainfall forecasting; no abstract is available and the reported results cannot be relied on.
A wavelet packet decomposition plus extreme learning machine hybrid (WPD-ELM) forecasts annual rainfall better than ARIMA and BPNN baselines, though the study is regional and modest in scope.
A bagging framework combining KNN, SVM, DNN and LSTM engines forecasts regional rainfall in Upstate New York, with SVM and KNN best for classification.
Compares KNN, SWAT and RCP weather generators for station temperature, precipitation and wind; KNN is more robust to noisy data.
Hybrid LSTM and genetic programming model forecasts short-term global solar radiation from weather station data, with improved accuracy over stacking baselines.
Twelve AI models predict wind speed from meteorological data, with LSTM best at 97.8% accuracy, and copulas model joint wind speed dependence across sites.
Artificial neural network predicts rainfall rate 30 seconds ahead at Durban from three prior measurements, for link-quality planning; the abstract is garbled and the study is small.
Smartphone app predicts five-day rainfall with ANN and Gaussian process regression using disaster-authority data; limited methodological novelty.
Compares LASSO, backpropagation and linear regression for Indian rainfall prediction using past rainfall records, with metrics but no novel method.
Tensorized transformer forecasts temperature from weather station records in USA, Canada and Europe, outperforming CNN, LSTM and ConvLSTM baselines.
Deep neural network with particle swarm feature selection predicts daily minimum and maximum temperature from local weather variables.
A basic neural network predicts daily rainfall for Yogyakarta Province from 2016–2021 BMKG station data; the study is small and regional.
Compares ARIMA, feedforward nets and LSTM for one-day-ahead solar irradiance forecasts from 12 weather stations, with imputation for missing data; LSTM is best under clouds.
Compares neural networks, Naive Bayes, random forest and KNN for classifying weather states from a Kaggle synoptic dataset; random forest reaches 89%.
Combines CEEMDAN decomposition, permutation entropy and ConvGRU to forecast daily South Asian High intensity from NCEP geopotential height, reporting higher accuracy than classical deep learning baselines.
An MLP predicts next-day minimum and maximum temperature and pressure for Szczecin, Poland, with 96% accuracy for temperature.
Deep learning framework (DSSAE-BBOA) for weather big-data analysis and visualization; no abstract is available, so the contribution is unclear.
CNN-based system predicts rainfall from climatological variables such as temperature and humidity; the abstract gives no quantitative results.
Deep learning (LSTM-style) predicts heat index from IoT temperature sensor time series and is compared with generalized linear regression; the study is low-rigour.
Deep learning model (DLztd) predicts daily GNSS zenith tropospheric delay at four West African stations from surface meteorology, filling data gaps.
Compares neural networks, random forest, SVM and kNN for weather prediction from government data, with a chatbot front end; the abstract reports no results.
Random forest predicts daily solar radiation and wind speed for Tamil Nadu from station data, reaching R2 0.97 and beating regression and SVM baselines.
Low-cost sensor weather station feeds machine learning to predict local rainfall from temperature, pressure, humidity and wind data.
Deep learning nowcasts evaporation duct height in the Yellow Sea at 5-minute resolution for 0–2 h, outperforming traditional time-series methods.
Retracted paper proposing a C5.0 decision tree with K-means for short-range weather prediction on MERRA data, claiming 90.18% accuracy.
Compares backpropagation, RBF, GRNN and deep networks on ten years of Kanyakumari district data, finding deep learning most accurate.
LSTM with correlation-based feature selection predicts hourly rainfall at Axim, Ghana, using ECMWF and station data, with MSE of 0.002.
Nanjing University of Information Science & Technology (NUIST)
Multilayer perceptrons with ENSO indices forecast weekly precipitation at six Tocantins, Brazil stations, with 33.38 mm/week test RMSE.
An artificial neural network predicts hourly solar radiation from station weather variables such as temperature, humidity, wind and pressure, reporting MAE and MSE.
Compares classical and deep learning models for regional rainfall forecasting in Upstate New York catchments; KNN best at handling precipitation uncertainty.
Wavelet neural networks forecast monthly rainfall at Queensland stations up to 12 months ahead, reducing three-month RMSE to 38.6 mm versus 72.2 mm for plain ANN.
Uses Extreme Learning Machine with quantum delta PSO to predict Denpasar weather; reports 100% accuracy on a small, unvalidated dataset.
Learning vector quantization neural network for weather forecasting reaches 72% accuracy; a small, locally scoped study with limited novelty.
RBF neural networks forecast monthly temperature and precipitation for Paraná, Brazil, with best skill for temperature and 3-month smoothed series.
Applies the equation learner symbolic regression to derive an interpretable analytical formula for short-term wind speed forecasting using few features.
Compares kNN, random forest, SVR, Holt-Winters and ARIMA for univariate hourly wind speed forecasting at Yanbu, Saudi Arabia.
Applies an artificial neural network to forecast air temperature in the Ararat valley from station observations.
Basic AI models trained on multi-station historical observations forecast near-term weather; neighbouring-region data is found to improve accuracy.
Supervised machine learning applied to wind speed forecasting at one site in Uruguay to support grid energy management; the abstract gives no quantitative results.
Spatio-temporal graph convolution with a learnable adjacency matrix forecasts wind speed across weather stations in Denmark and the Netherlands, outperforming baselines.
Compares LSTM and Facebook Prophet on five years of daily air temperature in Bandung; each performs better on different variables, with small RMSE differences.
Compares ANN, DNN, ELM, LSTM and LSTM with peephole connections for temperature and humidity at ten South Korean cities; LSTM gives the lowest errors.
Recurrent neural networks generate localized synthetic weather data from airport and on-site measurements, improving test accuracy over standard weather files by over 100%.
Seasonally-integrated LSTM autoencoder forecasts daily precipitation, raising horizon-3 forecast-observation correlation by 37 points over baselines.
Neural network autoregression forecasts daily maximum temperature in Coimbatore from 1991-2017 station records at medium range.
CLCRN, a graph-based recurrent network with local conditional convolutions, forecasts meteorological sensor time series and reports state-of-the-art results on weather benchmarks.
Applies a multilayer perceptron trained on Malaysian Meteorological Department station data to forecast daily weather conditions for Batu Pahat.
Compares decision trees, random forests and gradient boosting for rainfall prediction from historical records, reporting better accuracy than traditional statistical methods.
BL-FC, a bidirectional LSTM with fully connected layers, forecasts hourly station temperature in Ireland and beats LSTM, GRU and RNN, especially for abnormal temperatures.
Generic study of machine learning for weather forecasting; no abstract is available, so methods and results cannot be verified.
Builds an LSTM-based surface weather forecast model; no abstract is available to confirm the results.
Stacked LSTM forecasts day-ahead hourly temperature at about 150 m resolution on NYC IoT sensors, beating ARIMA and FNN, with Weather Underground history improving accuracy.
Learning vector quantization tuned by a genetic algorithm reaches 73% accuracy for categorical weather forecasts in Citeko, Indonesia.
MasterGNN jointly predicts air quality and weather at monitoring stations using heterogeneous recurrent graph networks with multi-adversarial training, beating seven baselines.
Compares CNN, MLP and hybrid CNN-MLP for air temperature prediction at a Padang station; CNN gives the highest R2.
Compares ARIMA with a hybrid 1D-CNN plus LSTM network for hourly temperature forecasting in Szeged, finding the deep model outperforms ARIMA.
Naive Bayes classifier predicts daily precipitation for Srinagar from five selected weather attributes, reported as accurate against observed station data.
Hybrid wavelet-packet and RNN models nowcast 100–150 m wind speed at three Brazilian tower sites; an RNN with discrete Meyer wavelet performed best.
Garbled, low-quality abstract on machine learning for temperature and rainfall prediction; no clear method, data or quantitative result.
Wavelet neural network forecasts drought indices for Fiji tropical stations, applying a hybrid wavelet-ANN approach to seasonal drought prediction; abstract not provided.
LSTM forecasts month-wise temperature and rainfall in Bangladesh using 115 years of station data, with a mean temperature error of -0.38 C over two years.
Deep-and-wide network with convolutional features and latitude/longitude inputs predicts rainfall across Rajasthan from historical time series, compared against MLP, LSTM and CNN baselines.
Applies a temporal convolutional network with transfer learning to forecast offshore visibility at Haikou Station, improving 24 h forecast scores and beating the CUACE model on TS score by 0.16.
Nanjing University of Information Science & Technology (NUIST)
Studies how ANN parameter count and hidden-layer depth affect one-day-ahead maximum temperature forecasts at South Korean stations, with a five-layer, 49-parameter model giving the lowest RMSE.
Proposes a streaming pipeline in which an LSTM continuously learns from live OpenWeatherMap feeds via Kafka to forecast temperature and precipitation for agriculture; no quantitative results reported.
Explores a generic regression model for weather forecasting over long periods; the abstract reports it is outperformed by professional services.
Applies an ANN to simulate monthly precipitation in Semnan, Iran from temperature, humidity, wind and pressure inputs, with the best scenario reaching R² of 0.86.
Recurrent-convolutional networks with self-attention forecast daily weather at 18 European cities, with occlusion analysis explaining which features and cities matter.
NARX neural networks trained with Levenberg-Marquardt and Bayesian regularization forecast short-term solar irradiance and other weather variables from hourly station data.
Two spatio-temporal extensions of functional neural networks are shown in simulations and applied to precipitation prediction, with the meteorological case as the main demonstration.
LSTM and MLP models use city-bus sensor data to monitor and forecast temperature, humidity and pressure one day ahead.
Tutorial introducing deep learning for weather time-series forecasting from station data; abstract not available, so classification is low confidence.
Compares k-NN, decision tree, random forest and linear regression for weather prediction, reporting accuracy with no new method.
Random Subfeature Ensemble Learning (RSEL) combines random feature subsets and ensembling to improve wind speed and direction forecasts over traditional methods, with claimed robustness.
Institute of Atmospheric Physics, Chinese Academy of Sciences
Simple ML models trained on multi-station historical data give short-range local weather forecasts in Tennessee, and adding neighboring stations improves accuracy.
RNN and LSTM models predict station precipitation, reaching 85.71% accuracy for weekly predictions versus 83.33% for monthly on validation data.
Uses deep neural networks and multiple linear regression on Spark-based big-data infrastructure to predict typhoon rainfall at four Taiwan stations, with faster computation than standalone systems.
A multilayer perceptron outperforms ARIMA for daily rainfall forecasting at a single site, with lower MAE and RMSE and R of 0.94.
Stacked stateful GRU models forecast solar irradiance at four West Bengal regions, with stateful designs improving prediction accuracy.
SVM classifies sequences of daily US rainfall maps to predict regional heavy rain 1 to 30 days ahead; central region beats baseline but corner regions do not.
Wavelet-decomposed neural networks predict monthly precipitation at sixteen Andalusian stations, reaching correlations of 0.76 to 0.90.
A multidimensional CNN forecasts wind speed from multivariate station weather histories in Danish and Dutch cities, outperforming 2D and 3D CNN baselines.
A didactic machine learning approach classifies rainfall categories near Etna and Stromboli from a weather station, with Stau and Foehn effects considered.
Lightweight LSTM and TCN models forecast surface weather from station records and are reported to beat classical methods and the WRF NWP model.
Applies linear regression to weather forecasting; no abstract is available, so scope and results rest on the title alone.
Applies machine learning methods to rainfall prediction for weather forecasts; no abstract is available, so details are limited.
A lightweight regression model runs on a microcontroller beside a weather station to forecast wind speed and direction offline, tested on real maritime data.
Applies artificial neural networks to predict drought (SPI at 3 to 48 months) in Tabriz, Iran, using station precipitation and other hydro-meteorological variables from 1992 to 2010.
Compares machine learning and neural network approaches for rainfall prediction from past rainfall records and selects the more accurate one; no quantitative results given.
Compares twelve AI algorithms for wind speed prediction from meteorological station data, with LSTM reaching 97.8% accuracy for wind farm planning.
Compares classical machine learning classifiers and regressors on NCDC station weather data, finding CART, XGBoost and AdaBoost most accurate for classification.
Combines artificial bee colony optimization with a backpropagation network to forecast basin precipitation from 17 stations over two decades, linked to wavelet periodicity and drought/flood years.
Reviews neural network regression for weather time-series forecasting and assesses ANN performance on country site data; methodological contribution is limited.
LSTM trained on NOAA station parameters (temperature, humidity, pressure, wind) on AWS predicts good or bad weather conditions; a small-scale applied study.
Arduino-based sensor system with machine learning to predict rainfall from local temperature and humidity records; limited methodological novelty.
Feature selection study for machine-learning weather forecasting; abstract unavailable, so contribution is unclear and likely incremental.
Temporal convolutional network applied to local weather station time series outperforms LSTM and classical ML for short-range forecasts of farm-scale weather.
PCA/ICA with BP network and LSTM model GPS zenith tropospheric delay in West Antarctica; six-hour forecasts reach 7.2 mm RMS.
Compares RNN, LSTM, GRU, linear regression and random forest for station relative-humidity prediction from date and time inputs, ranking which gives best accuracy under limited resources.
Reviews 2004–2016 literature on wind speed forecasting models, classifying regression, neural network, time series, fuzzy logic and Gaussian process approaches; ANN and fuzzy logic are flagged as future directions.
Compares MLP, RNN and CNN on hourly inputs for daily temperature forecasts at three sites; hourly data beats daily inputs and CNN performs best.
DEL-Jet ensemble of CNNs, autoencoder and nonlinear PCA forecasts wind speed for renewable energy, offering a more robust ensemble than single models.
Neuro-fuzzy model forecasts monthly temperature, humidity, wind and rainfall for a region of Egypt using station records; modest gains, regional scope.
Proposes a transductive LSTM for time-series prediction, applied to weather forecasting from observed meteorological series.
Decision tree with stepwise regression predicts weather from temperature, pressure and humidity; a basic station-based application with little novelty.
Compares data mining and machine learning methods for predicting next-year precipitation from weather records; largely a comparative study.
Deep LSTM networks trained on 15 years of hourly Brazilian station data for 1-hour and 24-hour weather prediction; two hidden layers balance accuracy and runtime.
Compares ANN, Naive Bayes and SVM classifiers for rainfall prediction in Sulaymaniyah, Iraq, using station weather records; no single headline number given.
Hybrid Conv1D-MLP model predicts daily rainfall 1 to 5 days ahead at twelve sites using GCM meteorological inputs, outperforming MLP and SVR.
Random Forest models predict daily maximum and minimum temperatures for Tamil Nadu cities from a month of prior weather data, beating linear regression.
A multilayer perceptron forecasts precipitation for Managua from three weather stations and teleconnection indices (NAO, SOI, ONI and others), for climate-sensitive planning.
A neural network predicts GNSS zenith wet delay up to six hours ahead from in-situ meteorological data, reaching 1.5 cm RMSE at two Austrian stations.
Describes an Arduino and WiFi weather monitoring setup with a claimed deep-learning forecaster; the ML description is vague and gives no quantitative results.
Applies WEKA classification algorithms, including an optimized KNN, to past weather datasets to predict rainfall, humidity and temperature.
Compares RNN and LSTM models for weather prediction from radar, ship and ground observations, reporting the RNN as more accurate than other neural networks.
Hybrid MLP and SVR models optimized with a flow regime algorithm forecast monthly precipitation in East Azarbaijan, Iran, outperforming other models.
LSTM with a missing-data refinement step forecasts station temperature at 6 to 24 hours and 7 to 14 days ahead, with lowest RMSE among compared models.
Compares random forest and gradient boosting for monthly rainfall prediction on Tenerife using station and reanalysis predictors over four decades.
Lightweight perturbed-observation ensemble neural network nowcasts wind speed at masts, cutting NRMSE by 20.9% and running an order of magnitude faster than rivals.
Neural network applied to maritime weather prediction; the abstract is missing, so the scope and results could not be verified.
Machine learning on Paranal monitoring data forecasts optical turbulence and seeing over the next two hours, a turbulence nowcasting study for VLT scheduling.
A comparison of standard machine learning classifiers predicts next-day rain in major Australian cities from station weather records, assessing inputs and preprocessing choices.
Deep recurrent networks in a two-step model predict typhoon wind speed and wave height at a Taiwan buoy, beating a one-step wave-height baseline.
Multiple linear regression feeds a backpropagation neural network to forecast low, average and heavy rainfall at Bengaluru from five years of station data.
LSTM with multivariate meteorological inputs forecasts one-hour global solar radiation, outperforming earlier univariate models.
LSTM networks trained on NCDC station variables to predict weather conditions, with no quantitative skill reported in the abstract.
Feed-forward neural network predicts rainfall from time series for agriculture and drought or flood warning, with no quantitative results given.
Hybrid clustering and neural-network model claims superiority over MLP feed-forward networks for rainfall prediction, with no quantitative results given.
Proposes a fuzzy-transform seasonal forecaster for station heat index in Campania, Italy, outperforming ARIMA and ANN baselines; the ML contribution is modest.
Apriori association-rule mining on Chennai rainfall data claims about 95% accuracy for rainfall occurrence, with no independent validation described.
Linear regression on temperature, humidity and wind predicts monthly rainfall from one area's historical records, claimed to beat traditional systems without numbers.
Instance-based transfer learning borrows records from neighboring cities to extend scarce station data for temperature and precipitation forecasts in Zhaoqing, China.
Identifies PWV, solar radiation and seasonal/diurnal features as key predictors and uses them in an ML classifier, reaching 79.6% accuracy for rainfall prediction.
Decision-tree model predicts five-day weather using machine learning as an alternative to physics-based forecasting; little methodological novelty.
Hybrid GSTAR-SUR with neural network residual model improves 10-day precipitation forecasts at four West Java stations, reducing mean error from 4.33 to 3.89 mm.
Python workflow pulls open weather API data to forecast five days ahead at three-hour intervals; the machine learning contribution is minimal.
Probabilistic quantitative precipitation forecasting combining machine learning with probable maximum precipitation; abstract not available, so classification is low-confidence.
Stacked denoising autoencoder features with LSTM improve South African wind speed forecasts by 17% over benchmarks, with NWP data cleaned by clustering.
Deep shared-representation network for jointly forecasting several weather elements; abstract unavailable, so scope is inferred from the title.
Uses clustering strategies in machine learning for weather prediction; no abstract is available to confirm the approach or results.
LSTM sequence-to-sequence models forecast solar irradiance 1 to 24 hours ahead, outperforming GBRT and feed-forward baselines using neighbouring-site wind and irradiance features.
An MLP neural network forecasts monthly SPEI drought index at 17 Pakistani climate stations, showing usable skill by MAE, RMSE and correlation.
Japanese conference paper on machine-learning spot weather forecasting; abstract unavailable, so details are unverified.
Applies neural networks and decision trees to 2000–2009 Ibadan, Nigeria station data for weather and extremes prediction; a low-novelty applied study.
A wavelet-denoised multicellular gene expression programming model (WTGEPRP) forecasts precipitation at Zhengzhou, Nanning and Melbourne, outperforming several baseline hybrid models.
A neural network nowcasts downvalley wind speed and direction from routine station observations, with a -0.28 m/s speed bias and 84% of directions within 45 degrees.
Ordinal neural network refines a persistence baseline to predict airport low visibility, outperforming persistence and autoregressive models at longer horizons on station data.
Compares neural networks, regression trees and Nexting for short-term local weather trend forecasts from historical data; no single method clearly wins, with Nexting good for slow signals.
Surveys machine learning algorithms applied to weather forecasting for agriculture in India; the abstract is truncated, so the contribution is low-confidence.
Linear and logistic regression applied to open government and NOAA data predict weather at several locations, with no quantified skill reported.
Investigation of clustering and classification ML and ensemble techniques for weather forecasting; the abstract gives no quantitative results.
Compares linear regression, CART, MLP and SVM on station temperature, pressure and humidity data; no new method is proposed.
Uses upstream lidar inflow measurements with classification and probabilistic regression for minute-scale wind and power forecasts.
ANN models relate monthly temperature and rainfall in Colombo, Sri Lanka, to humidity and wind, with temperature predicted well and rainfall poorly.
An LSTM forecasts 5-minute solar irradiance minutes ahead from one year of plant radiometer data, beating standard time-series baselines.
Bayesian network learned from four years of station data forecasts wind speed and solar irradiance classes for renewable energy resource evaluation.
CNN trained on past wind, humidity, temperature and visibility predicts runway visual range at a local airport from station time series.
Multi-variable stacked LSTM forecasts short-term wind speed at West Texas sites from meteorological inputs, with competitive accuracy versus baselines.
Two-layer spatio-temporal stacked LSTM predicts temperature across multiple locations, with spatial information improving accuracy in most cases.
Neural network models (RBF, fuzzy C-means, NARX) forecast weather variables in Nineveh, Iraq, to 2050, with a GUI for decision-makers; no quantitative accuracy given.
Neural-network rainfall forecasting expert system estimates local geostrophic wind direction and magnitude from surrounding pressure and temperature observations in Valladolid, Spain.
Simplified deep extreme learning machines are proposed for Indonesian rainfall prediction to support planting calendars; no quantitative accuracy is reported in the abstract.
Title describes deep-learning weather prediction with a sliding-window method, but the abstract is boilerplate and reports no method details or results.
ANFIS and multiple linear regression predict monthly station temperature, humidity, wind and pressure from 2008–2012 data; ANFIS gives lower RMSE.
Applies deep belief networks to precipitation forecasting, claiming better results than conventional forecasting methods; quantitative results are not given in the abstract.
Nanjing University of Information Science & Technology (NUIST)
VMD decomposition with per-component extreme learning machines forecasts 60 years of monthly precipitation at two Shaanxi sites, outperforming BP, Elman and EMD-ELM baselines.
Stacked auto-encoder with dynamic fine-tuning for weather forecasting; no abstract was provided, so the scope and scale are judged from the title only.
Backpropagation neural network forecasts 12 and 24 h wind speed at La Serena, Chile, with RMSE about 0.7 m/s, and links it to wind turbine power output.
Neural networks trained on pattern-based meteorological data forecast precipitation volumes, reporting up to 97% success for one-day and 90% for two-day forecasts.
Artificial neural network predicts monthly rainfall in Kano, Nigeria from lagged ENSO indices, achieving correlation 0.73 versus 0.70 for a linear model.
Deep belief network forecasts next-24-hour precipitation at four Chinese regions from environmental factors, outperforming traditional machine-learning baselines in accuracy.
Applies artificial intelligence approaches to rainfall prediction in Kerala, India; the abstract is missing, so methods and results cannot be confirmed.
Proposes a binary multilayer perceptron method for rainfall prediction; no abstract was available, so reported skill is unconfirmed.
Hybrid EEMD and wavelet neural network for wind speed prediction; abstract not available, classified from title.
Clustering-based multi-model machine learning forecasts one-hour-ahead solar irradiance, outperforming single-model baselines by about 20%.
Neural network weather forecasting study; the abstract is unavailable, so scope and results are unclear.
Artificial neural networks using new Botswana automated weather station data predict daily precipitation at four stations, but struggle with extreme events.
Backpropagation neural networks forecast typhoon wind speeds at three western Taiwan stations, outperforming linear regression across different typhoon tracks.
Marginalized Gaussian CRF predicts continental US precipitation from partially observed climate-variable graphs and identifies weather stations to cut data-collection costs.
Multilayer perceptron trained with Levenberg-Marquardt predicts whether it will rain from temperature, pressure and humidity, reaching 99.75% accuracy.
Generalized regression neural network nowcasts low-level wind profiles up to 45 minutes ahead at Guarulhos Airport using SODAR data, with near-zero bias.
Ensemble data-mining framework of ANNs, random forests and SVMs forecasts 10 m wind speed for Austria from 1 to 48 hours using station observations.
MATLAB back-propagation neural network applied to one month of weather data; a small, basic application with limited novelty.
Tsukamoto fuzzy inference system predicts rainfall in South Lampung, Indonesia from ten years of data to support agricultural planning.
NARX recurrent network forecasts temperature and wind speed with MAPE under 20%, applied to meteorology for nuclear release assessment.
Backpropagation neural network and ANFIS models forecast daily precipitation for two Texas cities using weather variables and lagged precipitation values.
Genetic algorithm and harmony search tune neural networks to forecast daily mean temperature in six Turkish cities, improving on plain ANN baselines.
Survey of artificial neural network methods for rainfall prediction, arguing they outperform NWP and statistical models for nonlinear precipitation dynamics.
Nanjing University of Information Science & Technology (NUIST)
Backpropagation neural network predicts daily maximum temperature and classifies weather types from 1997-2015 records; a basic application study.
Neural network predicts GNSS zenith tropospheric and wet delay from GNSS and meteorological data, reaching 5 mm (ZTD) and 3.6 mm (ZWD) error for water-vapour estimates.
Proposes an explainable two-stage classification-then-regression ML post-processor for daily station precipitation, improving on single-stage black-box baselines.
A hybrid dynamic-factor, random forest and neural network model predicts greenhouse temperature, humidity, PAR and CO2 at multiple horizons, beating LSTM and BPNN baselines.
AdaBoost with MLP base predicts India's March–June Tmax anomalies 10 days ahead, matching CFS reforecast skill in April–May but not in March or June.
Pairs LSTM-RNN weather forecasting with a Random Forest for crop selection, reporting 5-8% RMSE on temperature and rainfall and 97% crop-selection accuracy.
LSTM predicts local heatwave occurrence from historical temperature, humidity and wind data, outperforming traditional statistical models in accuracy and F1.
LightGBM interpolates plains forecasts and observations to improve temperature and precipitation forecasts in Japanese mountain areas, with partial RMSE gains.
Machine learning predicts maximum air temperature to define heatwaves in Rajasthan and Karnataka, India; the abstract is unavailable, so details are from the title only.
Ensemble learning with focal loss improves 60-hour sea fog forecasts over WRF and NOAA FSL baselines off the Yangtze estuary, raising detection and cutting false alarms.
Post-processing & calibrationTsinghua University (NowcastNet)
Introduces a neural-network stochastic precipitation generator for New Zealand that reproduces daily and hourly statistics and captures climate-change drift; runs orders of magnitude faster than an RCM.
Sliding-window DNN bias correction of 100 m LDAPS forecasts at Pyeongchang reduces RMSE by 33% for temperature and 45% for wind speed.
Analyzes 1979–2022 Dongjiang basin extreme precipitation trends and uses a BP neural network to simulate and predict them.
A deep network combining satellite imagery with station series improves day-ahead solar irradiance forecasts and outputs per-step distributions, with zero-shot tests at unseen stations.
Machine learning lightning prediction model trained on ten years of Lokoja station weather records to give advance lightning warnings.
Gradient-boosted models forecast short-term local wind for a Moroccan chemical plant, cutting errors 38-52% versus official forecasts and enabling pollution-aware production decisions.
Applications & impactsMassachusetts Institute of Technology (MIT)
Stacking ensemble of neural networks corrects ECMWF-IFS maximum temperature forecasts at about 2400 stations, cutting RMSE by 52% versus IFS.
Post-processing & calibrationChina Meteorological Administration (CMA)
Describes a Microsoft AI system that predicts on-farm micro-climate conditions from farm sensor data to support precision agriculture decisions.
Compares five ML anomaly detectors for cleaning noisy IoT sensor data to improve local weather forecast reliability; the abstract gives no headline numbers.
Random forest and XGBoost ensembles predict site-specific solar radiation from station weather forecast reports, then convert it to PV power output.
A neural ML method predicts pentad water-surface evaporation for rice fields from forecast temperature, wind and precipitation, aimed at medium-term agricultural planning.
Neural network predicts lightning occurrence in South Africa from meteorological station data; no abstract is available.
CatBoost best predicts solar radiation in Korea from temperature, humidity, cloudiness and extraterrestrial radiation, outperforming LSTM and GRU variants.
Compares uncertainty estimation methods for deep learning forecasts of meteorological time series, testing whether predicted uncertainty behaves as expected across horizons.
Hybrid CNN-MLP model predicts solar radiation for solar energy; the abstract is missing, so the contribution is inferred from the title only.
Tree-based XGBoost trained on LDAPS NWP output beats raw LDAPS for Seoul visibility (MAE 2.04 km vs 4.66 km, R 0.88 vs 0.40).
Post-processing & calibrationKorea Meteorological Administration (KMA)
STCNet combines 2D and 1D convolutions to post-process NWP for station weather, outperforming a Seq2Seq RNN and yielding hourly forecasts from 3-hourly input.
Transformer post-processing of NWP 2 m temperature, trained on crowd-sourced stations, cuts STD to 1.13 °C versus 1.96 °C for raw NWP.
Data-driven forecasting of drought indicators for Jaisalmer, India; the abstract is unavailable, so the method and results are unverified.
Neural network using global climate indices forecasts monthly and seasonal precipitation in Crimea up to six months ahead, with about 55–62% forecast quality.
Uses a deep RNN weather forecast with Gaussian process uncertainty to schedule seed planting, cutting required capacity by up to 69%.
Back-propagation neural network combines GNSS tropospheric delay with meteorological data for short-term rainfall warning in Ningbo, with 100% true detection rate.
Random Forest-based solar irradiance forecasting at 288 San Antonio sites from weather data, achieving R2 of 0.85 to 0.91.
Neural network system with time-frequency decomposition forecasts tropical cyclone wave height in real time; the abstract is missing, so the classification relies on the title.
Compares linear regression and neural networks for filling gaps in micrometeorological station data in complex terrain, finding significant wind differences only in synoptic flow.
A stacked LSTM autoencoder with feature selection gives daily solar radiation prediction intervals at six Queensland farms, outperforming tree ensembles and persistence.
Random Forest and XGBoost ensembles trained on automatic weather station data predict 2-4 hour rainfall occurrence with about 80% accuracy.
LSTM model using multi-factor typhoon time series predicts storm-surge water level at a Yangtze Estuary station, giving 1 h accuracy and 15 h predictions, 41% better than traditional ML.
Physics-informed CNN estimates pixel-wise fog visibility from single images via Koschmieder's law, trained on a 3,000-image synthetic AirSim dataset.
DeepSTF combines CNN and attention encoder-decoder to post-process multi-site forecasts at 226 Beijing stations, improving on MOS benchmarks.
Post-processing & calibrationInstitute of Atmospheric Physics, Chinese Academy of Sciences
A deep neural network retrieves near-surface air temperature over South Korea from satellite land surface temperature and vegetation indices, reaching R of 0.98 and RMSE of 2.19 K.
AutoML post-correction of WRF forecasts at 13 Xinjiang stations cuts temperature RMSE from 3.24 to 2.34 °C and humidity RMSE from 19.54% to 11.54%.
Post-processing & calibrationChina Meteorological Administration (CMA)
Gaussian process model fuses station observations with NWP forecasts to improve local short- and long-term weather forecasts for dispersion modelling.
Neural network uses SST and SLP indices to forecast monthly precipitation for the Ai-Petri region up to 6 months ahead, with about 23–28% mean absolute deviation.
FogNet, a 3D CNN with attention, post-processes NAM forecasts to predict fog visibility at a US airport, outperforming HREF on eight metrics.
Post-processing & calibrationNOAA (incl. GFDL, PSL, GSL, NCEP/EMC)
LSTM and other ML wind forecasts drive a scaffold collapse warning system, giving 1–6 hour advance alerts for typhoon winds at Taichung stations.
Compares 1D CNN, MLP and Wide-Deep networks for summer monsoon rainfall in Rajasthan using IMD gridded and gauge data; the Wide-Deep model performs best.
BLSTM and CNN fuse station observations with RDAPS model images to predict Korean temperatures up to 14 days ahead, beating single-source models.
A hybrid model selects the best climate model forecast with a probabilistic MLP, then uses a deep LSTM to produce next-day rainfall in Australian climate zones.
Stacked LSTM and DRNN models using WRF outputs predict typhoon wind speeds in northeastern Taiwan for 1 to 6 hours, outperforming memoryless MLPs.
LSTM and Markov chain models forecast offshore wind and wave conditions from FINO3 station measurements to support turbine maintenance planning; the study compares the two models.
Uses linear regression and support vector regression to attribute and predict Canberra's seasonal precipitation and temperature from ENSO, SAM and IOD drivers, with SVR reaching r=0.737 for precipitation.
ZiANN combines feature selection, ensemble learning and neural networks to post-process NWP output for six-hour nowcasts and 40-hour point wind forecasts in Austria, outperforming operational INCA and AROME.
A CNN trained on 35 years of meteorological maps and visibility data forecasts severe haze occurrence in Singapore, beating a no-skill baseline but missing many haze events.
Machine-learning regression diagnoses horizontal visibility from mesoscale NWP forecasts; abstract not available, so the scope is inferred from the title.
Gaussian process regression with Matern kernel predicts short-term PV plant output from irradiance, temperature and solar angles, with a 95% error range of about -1.6% to 1.4%.
PSO-optimised neural network forecasts cloudburst rainfall in India from historical records; accuracy figures are not given in the abstract.
Artificial neural networks using ENSO and IOD indices forecast seasonal spring rainfall at three Western Australian stations, with testing correlations up to 0.96.
Hybrid ANN and multiple regression forecasts typhoon total rainfall and groundwater-level change in Taiwan's Zhuoshui basin, raising accuracy to about 80% from station data.
Hybrid clustering, ARIMA and neural network model forecasts subseasonal station precipitation, reducing RMSE by 24–48% versus seasonal ARIMA in Iran.
PVNet, a convolutional LSTM over NOAA NWP fields, forecasts aggregated German PV output 24 and 48 hours ahead, outperforming persistence.
Coupled Gaussian process priors improve multi-site solar power and wind speed forecasts, matching or beating benchmarks with better uncertainty quantification.
Self-adaptive differential evolution tuned extreme learning machines predict long-term solar radiation in solar-rich cities using MODIS and reanalysis predictors.
A differential polynomial network corrects 24-hour numerical relative humidity forecasts using local observations, improving forecasts in most cases.
CNN and RNN models estimate ambient air temperature from outdoor images alone, with the RNN outperforming prior methods on image sequences.
