Visibility Forecasts Research Articles

Improving visibility forecasting during haze-fog processes in shanghai and eastern China: The significance of aerosol and hydrometeor extinction

Aerosols and droplets are the main factors of visibility reduction by scattering and absorbing light. In this study we use the 3-D meteorology fields from the 9-km WRF-ADAS Real-Time Modeling System (WARMS) to drive the Community Multiscale Air Quality (CMAQ) model. A visibility forecasting framework is developed by combining extinction due to hydrometeors and aerosols, based on WARMS and CMAQ predictions. We have analyzed the results of a one-month forecasting period during the winter of 2021–2022 to assess the performance of the WARM-CMAQ model and to understand the impact of hydrometeor and aerosol extinction on operational visibility forecasting in Shanghai, the Yangtze River Delta (YRD), and part of the North China Plain (NCP). With our newly developed framework, we generally find improved or comparable results in visibility forecasts when compared to recent studies. We find that for the city of Shanghai, aerosol extinction has a minor impact on the model's performance when forecasting visibility below 1 km, but it becomes crucial for predictions spanning 1–10 km. Comparison against observations shows that the framework well captures the general contributions from various chemical constituents with nitrate as the most important factor in aerosol extinction (∼60%). Finally, our assessment of the YRD and part of the NCP highlights that in highly polluted areas aerosols could be significant for visibility below 1 km. This study emphasizes the necessity of integrating both aerosol and hydrometeor extinction in visibility forecasting during haze-fog processes, particularly for regions characterized by diverse pollutant levels and environmental conditions.

Improvement in the Forecasting of Low Visibility over Guizhou, China, Based on a Multi-Variable Deep Learning Model

High-quality visibility forecasting benefits traffic transportation safety, public services, and tourism. For a more accurate forecast of the visibility in the Guizhou region of China, we constructed several visibility forecasting models via progressive refinements in different compositions of input observational variables and the adoption of the Unet architecture to perform hourly visibility forecasts with lead times ranging from 0 to 72 h over Guizhou, China. Three Unet-based visibility forecasting models were constructed according to different inputs of meteorological variables. The model training via multiple observational variables and visibility forecasts of a high-spatiotemporal-resolution numerical weather prediction model (China Meteorological Administration, Guangdong, CMA-GD) produced a higher threat score (TS), which led to substantial improvements for different thresholds of visibility compared to CMA-GD. However, the Unet-based models had a larger bias score (BS) than the CMA-GD model. By introducing the U2net architecture, there was a further improvement in the TS of the model by approximately a factor of two compared to the Unet model, along with a significant reduction in the BS, which enhanced the stability of the model forecast. In particular, the U2net-based model performed the best in terms of the TS below the visibility threshold of 200 m, with a more than eightfold increase over the CMA-GD model. Furthermore, the U2net-based model had some improvements in the TS, BS, and RMSE (root-mean-square error) compared to the LSTM_Attention model. The spatial distribution of the TS showed that the U2net-based model performed better at the model grid scale of 3 km than at the scale of individual weather stations. In summary, the visibility forecasting model based on the U2net algorithm, multiple observational variables, and visibility data from the CMA-GD model performed the best. The compositions of input observational variables were the key factor in improving the deep learning model’s forecasting capability, and these improvements could improve the value of forecasts and support the socioeconomic needs of sectors reliant on visibility forecasting.

Visibility forecast in Jiangsu province based on the GCN-GRU model

Low visibility weather easily leads to traffic accidents, posing threats to human life and property. To accurately forecast visibility, we conduct an empirical study focusing on Jiangsu Province. Firstly, we collect the monitoring data from meteorological stations and environmental stations for 2017-2018. Secondly, we analyze the changes in visibility from both spatial and temporal perspectives. Next, the maximum Relevance Minimum Redundancy (mRMR) algorithm is employed to select factors affecting visibility, finding that humidity and PM2.5\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$PM_{2.5}$$\\end{document} concentrations are the primary factors. Finally, we propose GCN-GRU (Graph Convolutional Network and Gated Recurrent Unit) model for short-term visibility forecasting, which employs GCN to capture the interactions between stations and uses GRU to learn the interactions between times. Experimental results indicate that GCN-GRU outperforms the standalone GRU model and three machine learning models regarding 6-hour visibility forecasting. Compared to the best competitor, GCN-GRU achieves an average increase of 3.32% in Correlation Coefficient (CORR), a decrease of 17.52% in Root Mean Square Error (RMSE), a reduction of 26.62% in Mean Absolute Percentage Error (MAPE), and a decline of 16.53% in Mean Absolute Error (MAE).

Comparison of the Visibility Grading Forecast Method Based on Meteorological Factors and Environmental Factors

The main visibility forecast factors were identified with the support of data from routine meteorological observations from the Mianyang Airport and the Mianyang Environmental Monitoring Station from 2015 to 2018, and a visibility grading forecast model was established and tested by dint of the multiple linear regression and the KNN algorithm based on big data mining technology, and the variation characteristics of visibility in winter at the Mianyang Airport were studied. The results showed that (1) in addition to having a significant positive correlation with wind speed, the visibility in winter at the Mianyang Airport has a significant negative correlation with relative humidity, dew point temperature, AQI, PM2.5 concentration, PM10 concentration, and CO, and it has the strongest correlation with relative humidity, and the correlation coefficient is −0.76. (2) The multivariate linear regression model and the KNN model were adopted for grading forecasting experiments on visibility, and the results revealed that both models could be used for visibility grading forecasts. The multiple regression model secures an accuracy of over 70% for forecasts of level 1–5 visibility. In terms of the KNN model, the forecast accuracy is the best when K = 3 or K = 5, notably for level-2, level-4, and level-5 visibility. The forecast accuracy rate is 100% for level-2 visibility, but the forecast for level-1 visibility is poor. (3) The minimum value of the average daily visibility of the Mianyang Airport in winter appeared at 09 : 00 and the maximum value appeared at 17 : 00. The level-1 visibility occurred and developed before 09 : 00 and faded and vanished between 08 : 00 and 15 : 00.

FOG/VISIBILITY FORECAST AND VERIFICATION AT IGI AIRPORT, NEW DELHI DURING THE WINTER SEASONS OF 2020-21 & 2021-22

The reduction in visibility due to fog leads to cancellation of flights, delays and diversions at IGI Airport, New Delhi during the winter season. Accurate prediction of fog/ visibility is required within sufficient lead time to make the flight operations safer, economical and more convenient. This study attempts to improve the fog/visibility forecast at IGI airport for the months of December and January during the winter seasons of 2020-21 and 2021-22. Various meteorological parameters required to predict the fog/visibility are obtained from IMD GFS forecasted meteorological information which are valid for next 24 and 48-hours. Forecast verification was performed using the METAR and synoptic observations. The results for 2020-21 fog/visibility forecast for 24-hours are promising with a POD of 0.92, critical success index as 0.68, and false alarm ratio as 0.28. Moreover, the 2021-22 results for 24 -hours fog /visibility forecast are also observed to be promising with a POD of 0.73, critical success index of 0.54 and false alarm ratio at 0.32. This method gives fairly accurate predictions in point locations and can also be used for a larger spatial area. However, the 48-hours forecast performance needs further improvement. The method also predicted the wind speed and relative humidity which were found to be in agreement with the observed data to a great extent. In the upcoming years, the fog/visibility forecast method will be more robust with better prediction accuracy of meteorological parameters from model outputs and observations.

MODELLING FOR FUEL POLICY IN LOW VISIBILITY OPERATIONS IN AIRLINE OPERATIONS MANAGEMENT

Purpose- It can be said that fuel costs have the highest share in airline operation costs. Therefore, fuel efficiency is important for cost optimization in airline operation management. Fuel efficiency has been studied in many ways in airline operation management in the literature. It can be reported that the cost optimization due to additional fuel transportation caused by meteorological conditions is also an important field. This study aims to provide data for making a decision to form a fuel policy for a flight schedule by determining the divert risk rate by analyzing previous meteorology reports when there is a risk of divert due to low visibility in meteorology reports. Methodology- Thus, by providing data for airline operation managers to make decisions that will provide fuel optimization, cost-effectiveness will also be contributed. Data for the study were taken from OGIMET (the Spanish meteorology website). A simulated flight was created for the study, and the past period forecast efficiency was calculated from previous meteorological data to determine the divert risk rate. Then, it was revealed how much the divert risk rate was reduced if additional fuel was carried for the flight according to the past period meteorological data to ensure that the aircraft waited for suitable conditions in the air at the landing aerodrome. Findings- The above results reveal the divert risk rate for the airline39 flights when additional fuel is loaded and without loading additional fuel in low visibility forecast operations for the three months of the year when low visibility is more common, based on past meteorological data. Conclusion- As a result of the study, it can be reported that a model has been created relating to how to determine the divert risk rate to make a fuel policy decision for a flight schedule when there is a risk of diverting due to low visibility in airline operation management. Keywords: Airline operation management, low visibility, fuel policy, airline fuel efficieny, divert risk ratio, JEL Codes: L93, M21, C44

Aviation visibility forecasting by integrating Convolutional Neural Network and long short-term memory network

Weather forecasts are essential to aviation safety. Unreliable forecasts not only cause problems to pilots and air traffic controllers, but also lead to aviation accidents and incidents. To enhance the forecast accuracy, an integrated model comprising a convolutional neural network (CNN) and long short-term memory (LSTM) network is developed to achieve improved weather visibility forecasting. In this model, the CNN acts as the precursor of the LSTM network and classifies weather images to increase the visibility forecasting accuracy achieved with the LSTM network. For a dataset with 1500 weather images, the training, validation, and testing accuracy achieved with the integrated model is 100.00%, 97.33%, and 97.67%, respectively. On a numerical dataset of 10 weather features over 10 years, the RMSE and MAPE of an LSTM forecast can be reduced by multiple linear regression from RMSE 12.02 to 11.91 and 44.46% to 39.02%, respectively, and further by the Pearson’s correlation coefficients to 10.12 and 36.77%, respectively. By using CNN result as precursor to LSTM, the visibility forecast by integrating both can decrease the RMSE and MAPE to 2.68 and 13.41%, respectively. The integration by deep learning is shown an effective, accurate aviation weather forecast.

Implementation and application of ensemble optimal interpolation on an operational chemistry weather model for improving PM2.5 and visibility predictions

Abstract. Data assimilation techniques are one of the most important ways to reduce the uncertainty in atmospheric chemistry model input and improve the model forecast accuracy. In this paper, an ensemble optimal interpolation assimilation (EnOI) system for a regional online chemical weather numerical forecasting system (GRAPES_Meso5.1/CUACE) is developed for operational use and efficient updating of the initial fields of chemical components. A heavy haze episode in eastern China was selected, and the key factors affecting EnOI, such as localization length scale, ensemble size, and assimilation moment, were calibrated by sensitivity experiments. The impacts of assimilating ground-based PM2.5 observations on the model chemical initial field PM2.5 and visibility forecasts were investigated. The results show that assimilation of PM2.5 reduces the uncertainty in the initial PM2.5 field considerably. Using only 50 % of observations in the assimilation, the root mean square error (RMSE) of initial PM2.5 for independent verification sites in mainland China decreases from 73.7 to 46.4 µg m−3, and the correlation coefficient increases from 0.58 to 0.84. An even larger improvement appears in northern China. For the forecast fields, assimilation of PM2.5 improves PM2.5 and visibility forecasts throughout the time window of 24 h. The PM2.5 RMSE can be reduced by 10 %–21 % within 24 h, and the assimilation effect is the most remarkable in the first 12 h. Within the same assimilation time, the assimilation efficiency varies with the discrepancy between model forecasts and observations at the moment of assimilation, and the larger the deviation, the higher the efficiency. The assimilation of PM2.5 further contributes to the improvement of the visibility forecast. When the PM2.5 increment is negative, it corresponds to an increase in visibility, and when the PM2.5 analysis increment is positive, visibility decreases. It is worth noting that the improvement of visibility forecasting by assimilating PM2.5 is more obvious in the light-pollution period than in the heavy-pollution period. The results of this study show that EnOI may provide a practical and cost-effective alternative to the ensemble Kalman filter (EnKF) for the applications where computational cost is the main limiting factor, especially for real-time operational forecast.

Kratkoročna prognoza vidljivosti određena metodom slučajne šume

Accurate visibility forecasting is essential for safe aircraft operations. This study examines how various configurations of the Random Forest model can enhance visibility predictions. Preprocessing techniques are employed, including correlation analysis to identify fundamental relationships in weather observations. Time-series data is transformed into a regular Data Frame to facilitate analysis. This study proposes a classification framework for organizing visibility data and phenomena, which is then used to develop a visibility forecast using the Random Forest method. The study also presents procedures for hyperparameter tuning, feature selection, data balancing, and accuracy evaluation for this dataset. The main outcomes are the Random Forest model parameters for a three-hour visibility forecast, along with an analysis of errors in low visibility forecasts. Additionally, models for one-hour forecasts and visibility forecasting under precipitation are also examined. The resulting models demonstrate a deterministic forecast accuracy of approximately 78%, with a false alarm rate of around 6%, providing a comprehensive overview of the capabilities of the Random Forest model for visibility forecasting. As anticipated, the model demonstrated limitations in accurately simulating fast radiative cooling or abrupt decreases in visibility caused by precipitation. Specifically, in relation to precipitation, the model achieved an accuracy of 79%, yet exhibited a false alarm rate of 19%. Additionally, this method sets a foundation for enhancing prediction accuracy through the inclusion of supplementary forecast data, while its implementation on real-world datasets expands the reach of machine learning techniques to the members of the meteorological community.

Probabilistic Forecast of Visibility at Gimpo, Incheon, and Jeju International Airports Using Weighted Model Averaging

In this study, weighted model averaging (WMA) was applied to calibrating ensemble forecasts generated using Limited-area ENsemble prediction System (LENS). WMA is an easy-to-implement post-processing technique that assigns a greater weight to the ensemble member forecast that exhibits better performance; it is used to provide probabilistic visibility forecasting in the form of a predictive probability density function for ensembles. The predictive probability density function is a mixture of discrete point mass and two-sided truncated normal distribution components. Observations were obtained at Gimpo, Incheon, and Jeju International Airports, and 13 ensemble member forecasts were obtained using LENS, for the period of December 2018 to June 2019. Prior to applying WMA, a reliability analysis was conducted using rank histograms and reliability diagrams to identify the statistical consistency between the ensembles and the corresponding observations. The WMA method was then applied to each raw ensemble model, and a weighted predictive probability density function was proposed. Performances were evaluated using the mean absolute error, the continuous ranked probability score, the Brier score, and the probability integral transform. The results showed that the proposed method provided improved performance compared with the raw ensembles, indicating that the raw ensembles were well calibrated using the predicted probability density function.

Analog Ensemble Forecasting System for Low-Visibility Conditions over the Main Airports of Morocco

Low-visibility conditions (LVC) are a common cause of air traffic, road, and sailing fatalities. Forecasting those conditions is an arduous challenge for weather forecasters all over the world. In this work, a new decision support system is developed based on an analog ensemble (AnEn) method to predict LVC over 15 airports of Morocco for 24 forecast hours. Hourly forecasts from the AROME model of eight predictors were used to select the skillful analogs from 2016 to 2018. The verified hourly observations were used as members of the ensemble. The developed ensemble prediction system (EPS) was assessed over 1 year (2019) as a single-value forecast and as a probabilistic forecast. Results analysis shows that AnEn outperforms persistence and its best performances are perceived generally during night and early-morning lead times. From continuous verification analysis, AnEn forecasting errors are found to be location- and lead-time-dependent and become higher for low-visibility cases. AnEn draws an averaged Centered Root Mean Square Error of about 1500 m for all visibilities, 2000 m for fog and 1500 m for mist. As an EPS, AnEn is under-dispersive for all lead times and draws a positive bias for fog and mist events. For probabilistic verification analysis, AnEn visibility forecasts are converted to binary occurrences depending on a set of thresholds from 200 m to 6000 m by a step of 200 m. It is found that the averaged Heidke Skill Score for AnEn is 0.65 for all thresholds. However, AnEn performance generally becomes weaker for fog or mist events prediction.

Visibility Prediction Based on Machine Learning Algorithms

In this study, ground observation data were selected from January 2016 to January 2020. First, six machine learning methods were used to predict visibility. We verified the accuracy of the method with and without principal components analysis (PCA) by combining actual examples with the European Centre for Medium-Range Weather Forecast (ECMWF) data and National Centers for Environmental Prediction (NECP) data. The results show that PCA can improve visibility prediction. Neural networks have high accuracy in machine learning algorithms. The initial visibility data plays an important role in the visibility forecast and can effectively improve forecast accuracy.

Deep learning models for visibility forecasting using climatological data

Low visibility conditions affect safety and traffic operations, leading to adverse scenarios that often result in serious accidents. Due to the complexity and variability associated with modeling weather variables, visibility forecasting remains a highly challenging task and a matter of significant interest for transportation agencies nationwide. Given that the literature on single-step visibility forecasting is very scarce, this study explores the use of deep learning models for single-step visibility forecasting using time series climatological data. Five different deep learning models were developed, trained, and tested using data from two weather stations located in the US state of Florida, which is one of the top states nationwide dealing with low visibility problems. The authors provide discussions of the models’ results and areas for future research.

Research on the usability of different machine learning methods in visibility forecasting

Haze pollution, mainly characterized by low visibility, is one of the main environmental problems currently faced by China. Accurate haze forecasts facilitate the implementation of preventive measures to control the emission of air pollutants and, thereby mitigate haze pollution. However, it is not easy to accurately predict the low visibility events induced by haze, which requires not only accurate prediction for weather elements, but also refined and real-time updated source emission inventory. In order to obtain reliable forecasting tools, this paper studies the usability of several popular machine learning methods, such as support vector machine, k-nearest neighbor, random forest, as well as several deep learning methods, on the visibility forecasting. Starting from the main factors related to visibility, the relationships between wind speed, wind direction, temperature, humidity, and visibility are discussed. Training and forecasting were performed using the machine learning methods. The accuracy of these methods in visibility forecasting was confirmed through several parameters (i.e., root-mean-square error, mean absolute error, and mean absolute percentage error). The results show that: (1) Among all meteorological parameters, wind speed was the best at reflecting the visibility change patterns; (2) RNN LSTM, and GRU methods performs almost equally well on short-term visibility forecasts(i.e. 1h, 3h, and 6h); (3) A classical machine learning method (i.e. the SVM) performs well in mid- and long-term visibility forecasts; (4) The machine learning methods also have a certain degree of forecast accuracy even for long time periods (e.g. of 72h).

Delhi Model with Chemistry and aerosol framework (DM‐Chem) for high‐resolution fog forecasting

AbstractWe introduce here the Delhi Model with Chemistry and aerosol framework (DM‐Chem), a high‐resolution (330 m) visibility and fog forecasting model with the UK Chemistry and Aerosol scheme. DM‐Chem is developed at the National Centre for Medium‐Range Weather Forecasting (NCMRWF) in collaboration with UK Met Office and Carnegie Mellon University. We present a revised visibility scheme here which uses prognostic aerosols (mass and number concentration) from the two‐moment prognostic Global Model of Aerosol Processes (GLOMAP) ‐mode aerosol scheme. This scheme represents aerosol microphysical processes and the internal mixing of chemical components within five log‐normal size modes. The cloud microphysics parameterization also makes use of the GLOMAP scheme to predict cloud droplet number concentration (). Therefore we are able to evaluate the performance of the scheme indirectly by comparing the predicted with that derived from Moderate‐Resolution Imaging Spectroradiometer (MODIS) retrievals over the Indo‐Gangetic Plain. Aerosol emissions for the inner (330 m) nest are prescribed with a high‐resolution inventory for Delhi, from the System of Air Quality and Weather Forecasting project (SAFAR). Winter‐time stubble burning is represented using the MODIS‐based Fire Energetics and Emissions Research inventory, updated daily. In this article we consider case studies spanning both foggy and haze days of winter 2019–2020. The model compares well with the observed visibility and its spatial and diurnal patterns. The SAFAR emissions help improve the near‐surface model fields. For the “haze” case, the single scattering albedo (SSA) is high along the regions of high AOD, and the surface radiative cooling is enhanced. During the “foggy” case, the SSA is not clearly related to AOD, but aerosol absorption does lead to a minor surface warming via the semi‐direct effect at the time of onset of fog. The DM‐Chem entered operational mode at NCMRWF from winter 2020 and provides high‐resolution visibility forecast for Delhi and the adjoining region.

Application of Machine-Learning-Based Fusion Model in Visibility Forecast: A Case Study of Shanghai, China

A visibility forecast model called a boosting-based fusion model (BFM) was established in this study. The model uses a fusion machine learning model based on multisource data, including air pollutants, meteorological observations, moderate resolution imaging spectroradiometer (MODIS) aerosol optical depth (AOD) data, and an operational regional atmospheric environmental modeling System for eastern China (RAEMS) outputs. Extreme gradient boosting (XGBoost), a light gradient boosting machine (LightGBM), and a numerical prediction method, i.e., RAEMS were fused to establish this prediction model. Three sets of prediction models, that is, BFM, LightGBM based on multisource data (LGBM), and RAEMS, were used to conduct visibility prediction tasks. The training set was from 1 January 2015 to 31 December 2018 and used several data pre-processing methods, including a synthetic minority over-sampling technique (SMOTE) data resampling, a loss function adjustment, and a 10-fold cross verification. Moreover, apart from the basic features (variables), more spatial and temporal gradient features were considered. The testing set was from 1 January to 31 December 2019 and was adopted to validate the feasibility of the BFM, LGBM, and RAEMS. Statistical indicators confirmed that the machine learning methods improved the RAEMS forecast significantly and consistently. The root mean square error and correlation coefficient of BFM for the next 24/48 h were 5.01/5.47 km and 0.80/0.77, respectively, which were much higher than those of RAEMS. The statistics and binary score analysis for different areas in Shanghai also proved the reliability and accuracy of using BFM, particularly in low-visibility forecasting. Overall, BFM is a suitable tool for predicting the visibility. It provides a more accurate visibility forecast for the next 24 and 48 h in Shanghai than LGBM and RAEMS. The results of this study provide support for real-time operational visibility forecasts.

Applying a Physically Based Blowing Snow Diagnostic Parameterization to Improve Wintertime Visibility Forecasts in the WRF Model

AbstractBlowing snow presents substantial risk to human activities by causing severe visibility degradation and snow drifting. Furthermore, blowing snow presents a weather forecast challenge since it is not generally simulated in operational weather forecast models. In this study, we apply a physically based blowing snow model as a diagnostic overlay to output from a reforecast WRF simulation of a significant blowing snow event that occurred over the northern Great Plains of the United States during the winter of 2019. The blowing snow model is coupled to an optics parameterization that estimates the visibility reduction by blowing snow. This overlay is qualitatively evaluated against false color satellite imagery from the GOES-16 operational weather satellite and available surface visibility observations. The WRF-simulated visibility is substantially improved when incorporating blowing snow hydrometeors. Furthermore, the model-simulated plume of blowing snow roughly corresponds to the blowing snow plumes visible in the satellite imagery. Overall, this study illustrates how a blowing snow diagnostic model can aid weather forecasters in making blowing snow visibility forecasts, and demonstrates how the model can be evaluated against satellite imagery.

The quality of modern numerical visibility forecast methods

The quality of modern numerical visibility forecast methods

Application of Offshore Visibility Forecast Based on Temporal Convolutional Network and Transfer Learning.

Visibility forecasting in offshore areas faces the problems of low observational data and complex weather. This paper proposes an intelligent prediction method of offshore visibility based on temporal convolutional network (TCN) and transfer learning to solve the problem. First, preprocess the visibility data sets of the source and target domains to improve the quality of the data. Then, build a model based on temporal convolutional network and transfer learning (TCN_TL) to learn the visibility data of the source domain. Finally, after transferring the knowledge learned from a large amount of data in the source domain, the model learns the small data set in the target domain. After completing the training, the model data of the European Mid-Range Weather Forecast Center (ECMWF) meteorological field were selected to test the model performance. The method proposed in this paper has achieved relatively good results in the visibility forecast of Qiongzhou Strait. Taking Haikou Station in the spring and winter of 2018 as an example, the forecast error is significantly lower than that before the transfer learning, and the forecast score is increased by 0.11 within the 0-1 km level and the 24 h forecast period. Compared with the CUACE forecast results, the forecast error of TCN_TL is smaller than that of the former, and the TS score is improved by 0.16. The results show that under the condition of small data sets, transfer learning improves the prediction performance of the model, and TCN_TL performs better than other deep learning methods and CUACE.

Machine-learning regression applied to diagnose horizontal visibility from mesoscale NWP model forecasts

Low-visibility conditions are a weather hazard that affects all forms of transport, and accurate forecasting of their spatial coverage is still a challenge for meteorologists, particularly over a large domain. Current predictions of visibility are based on physical parametrizations in mesoscale models and are thus limited with respect to accuracy. This paper examines the use of supervised machine-learning regression techniques (tree-based ensemble, feed-forward neural network and generalized linear methods) to diagnose visibility from operational mesoscale model forecasts over a large domain. To achieve this, hourly forecasts of meteorological parameters in the lower levels of the atmosphere have been used. In the short-range forecasting framework, the machine-learning algorithms were developed to provide hourly forecasts up to 24 h. To assess the performance of the developed models, hourly observed data, collected at 36 synoptic land stations over the northern part of Morocco, have been used. This region is characterized by a heterogeneous topography. The tree-based ensemble methods have shown some improvement in visibility forecasting in comparison with the operational visibility diagnostic scheme based on Kunkel’s formula and also with persistence. It is also found that this machine-learning technique performs better when the forecast depends on multiple predictors instead of only a few with very high importance. In addition, their performance is very sensitive to the disproportionality of data availability between daytime and night-time. Furthermore, it is found that the performance decreases when principal components are used instead of raw correlated data.