Search

This paper aims to explore machine learning techniques for post‐processing high‐resolution Numerical Weather Prediction (NWP) products for the early detection of convection. Data from the Arome Ensemble Prediction System and satellite observations from the Rapidly Developing Thunderstorm (RDT) product by Météo‐France are used to train a recurrent neural network model to predict areas of total convection and moderate convection. The learning task is formulated as a binary classification problem using a long short‐term memory (LSTM) network architecture. Results from the LSTM model are compared with an object‐based probabilistic approach to forecast convection using metrics such as a receiver operating characteristics (ROC) curve, the Brier score and reliability. Results indicate that the LSTM model performs similarly to the object‐based probabilistic benchmark when classifying moderate convection areas and shows improved skill when classifying areas of total convective. Finally, the LSTM model results are presented within an air traffic management context to showcase the potential use of machine learning models within an operational application. [ABSTRACT FROM AUTHOR]

Small-scale, fast-evolving airflow disturbances, possibly due to upstream building clusters, are known to affect aircraft landing at corridor 25RA of Hong Kong International Airport (HKIA), China, under southerly or southwesterly winds. The paper presents a new algorithm, based on the concept of coherent integrated differential velocity (CIDV), for the automatic alerting of such hectometric (scale of hundreds of metres) turbulence features using short-range LIDAR (SRL) scans at 30 m radial resolution and an observation frequency of around every 20 s by Hong Kong Observatory (HKO). The CIDV identifies the cumulative effect of coherent patches of radial velocity changes occurring along the landing flight path by comparing two consecutive subminute SRL scans. Verification was conducted between June and September 2017 against a selection of relevant pilot reports, with a positive skill level observed over the whole range of the CIDV thresholds considered. When simulating a realistic setting where the SRL would issue alerts in conjunction with HKO's operational Windshear and Turbulence Warning System (WTWS), the CIDV algorithm would bring a gain in hit rate between 5% and 25%, with a corresponding increase in alert duration between 0.4% and 20%. The chosen CIDV threshold of 0.4 m·s-1 nautical mile (nm) can be physically interpreted as an average radial velocity change of 4.0 m·s-1 sustained over a distance of < 200 m (or between 2 and 3 s of flight), which agrees with reported values of crosswind changes considered significant for a landing aircraft and also pilots' perception of the characteristics of such events. [ABSTRACT FROM AUTHOR]

In the evening of 11 April 2011, vortex/wave shedding from a mountain to the south of the Hong Kong International Airport (HKIA) appeared to occur following the establishment of fresh to strong easterly airflow in Hong Kong. This paper documents the observations by the various meteorological instruments of the event and discusses the possibility of forecasting the vortex/wave shedding using a high resolution numerical weather prediction (NWP) model. The vortex/wave shedding was well observed by the Light Detection and Ranging (LIDAR) system at HKIA. The vortex/wave appeared as an area of reversed flow against the background east to southeasterly airflow in the airport area. It was detached from the mountain and moved to the northwest with the background southeasterly flow, with a gradual decrease of the spatial extent in the area of reversed flow. The vortex/wave brought about significant wind shear to both arrival flight paths to the west of HKIA, as evidenced from the headwind profile data measured by the LIDARs and the aircraft data. It had rather large vertical extent, reaching the top of the boundary layer, but did not extend downward to the surface, as shown from the vertical scans of the LIDAR and the surface wind observations. The shedding event is successfully captured by high-resolution NWP model with a spatial resolution of 50 m. The model output has a lead time of 6-7 h ahead of the event, indicating the possible occurrence of terrain-induced wind shear in the form of vortex/wave shedding from a mountain. As a result, the numerical simulation could be useful in giving an early alert to the aviation weather forecaster about the significant wind shear event, which could be hazardous to the landing aircraft. Copyright © 2011 Royal Meteorological Society [ABSTRACT FROM AUTHOR]

ABSTRACT At airports with significant economic growth, more and more buildings are being constructed inside and around the airport premises. These buildings may become a source of low level wind disturbances to the aircraft operating at the airport. In assessing such low level wind effects of buildings, objects and structures, a 7 kn crosswind change criterion can be used. This criterion restricts the wind variations along the glide path and above the runway to a crosswind change of less than 7 kn. It was developed in and based on the experience in the Netherlands over the past 20 years. Therefore, the question arises whether this criterion is applicable to airports and buildings elsewhere in the world, and whether it is consistent with other hazardous wind metrics available in the literature. Based on two cases of pilot reported building induced wind disturbances at Hong Kong International Airport, computer simulated wind fields and a flight simulator were employed to study the validity of the 7 kn criterion. It was found that the criterion is in general consistent with the other hazardous wind metrics available in the literature for the cases under consideration. The number of cases considered in the present study is still limited and further studies into the matter should be considered. [ABSTRACT FROM AUTHOR]

ABSTRACT A novel system for real-time prediction of wind speed and direction in a critical region of complex terrain is developed by combining measurements and analytically based fast computation of the flow field. The new system is applied to the aircraft approach flight paths of Hong Kong International Airport ( HKIA), where orographically generated turbulence downwind of a mountain ridge in high-wind conditions causes disruptions to aircraft movements regularly. By using wind profiler and radiosonde measurements to drive the analytically based ' FLOWSTAR' model, good agreement is demonstrated between modelled and observed wind flow data from a network of ground-based monitors and LIDAR scans in the vicinity of the airport for a historical wind event. The methodology forms the basis of a system that can be used to forecast extreme wind events. [ABSTRACT FROM AUTHOR]

ABSTRACT A theoretical and computational study is presented of the wind flow over the mountain ridges to the south of Hong Kong International Airport given the upwind profiles of 'mean' velocity and temperature. A sensitivity study shows how large variations in wind speed, wind shear and wind direction occur on the approach path when the upwind flow from the southeast is stably stratified above the boundary layer with a significant inversion no more than a few hundred metres above the ridge (height about 450 m). The Froude number is close to unity. The effects of vertical wind shear across the inversion layer reduce the speed-up, and directional wind shear changes the surface flow direction. The observed horizontal length scales along the approach are about equal to the projected widths of the predominant features of the ridges and valleys. For an extreme wind event there was found to be satisfactory agreement between the wind speed measured by a landing aircraft and the predictions of the fast quasi-linear analytically based computer model FLOWSTAR. This suggests that near real time application of the model and near real time monitoring with remote and in situ instruments can be used to predict these extreme flow events and hence give warnings about them. Conclusions are drawn about how such predictive modelling could be developed. [ABSTRACT FROM AUTHOR]