The continuous development and advancements made in the air transport have leads to considerable delays over the years, making it one of the major challenges in the air transport industry. One of the attributes of high operational performance indicator in air transport is punctuality. Punctuality corresponds to the percentage of flights arriving and leaving on time against those that don’t make it in time. The concept of delay in the air transport industry had a broad aspect of definitions, however, the standard as defined by the United States Department of Transport’s Bureau of Transportation Statistics as arrival of a flight at the opening fifteen minutes or even more after its scheduled arrival as reflected by the computer reservation system.
The prediction of flight delays is an already explored subject and it involves the use of information relating to weather conditions, airport congestions and other mechanical aspects to give quite accurate prediction s of possible future delays. While this subject has been explored largely with different models being developed to confirm the accuracy of prediction, there exist some random variables have been identified as possible hindrances to prediction of the flight delays. However, it is possible to predict flight delays and take precautions and measures to curb the possible implications that may come with it.
Delay prediction in flights
Different researchers have established possible ways to make this happen and have shown affirmative results by accurately forecasting the possible delays successfully. Through these research studies, different models have been developed for flight delay prediction. This essay will bring out a number of these models and the aspects they use to forecast on flight delays. At the same time, it will refute some of the parameters used in the prediction and how they may affect possible forecasting of flight delays.
Weather conditions being one of the factors that cause delays in flights can be used to accurately predict flight delays (Santos and Robin, 2010). The delays of flight and cancellations in any given season grow directly proportional to the increase in precipitation an airport receives or its outside temperatures, compared with the airport’s thirty year average for that season. This is the first establishment necessary for successful prediction of flight delays (Santos and Robin, 2010). Prediction in this aspect can be done looking at the weather forecast changes and looking out at the factors related to those factors hazardous to flights. Such factors can be defined as convective weather, visibility, wind and reduced ceiling in correspondence AAR/ADR reports at corresponding local and airports of destination airports respectively (Pejovic, Williams, Noland and Toumi, 2009). All the data regarding the weather conditions such as the Quarter Hour Airport Report found in the FAA database, ASPM which reports weather condition of the factors mentioned above can be used to forecast the probable weather conditions.
Airport congestion is one of the factors that cause flight delays. The likelihood of congestion can be established by calculating the operation demand ratio at airports and the capacity at every particular time of departure as scheduled. A number of variables can be used to represent the capacity, demand and the relationship between the two in NAS. By doing this, the scheduled operations are separated into scheduled departure and arrival operations, and the capability into both the departure and arrival throughputs. Once this is determined, it becomes easy to find out the operation demand of the scheduled time and hence possibility of a congestion which may lead to the delays. This way, through analysis of the figures and variables, the flight delay times can be established conveniently.
Carrier delay is another component of flight delays listed in the BTS database (Gilbo, 1993). Elements in this aspect includes cleaning of aircraft, connecting passenger arrival and crew, aircraft damage, strike, loading of cargo, computer outage equipment, legality of crew, lavatory servicing, slow boarding by travelers, weight and balance delays, engineering inspection and fueling. These aspects can be used to look into the possibility of departure delays and the come up with a live forecast of these delays (Gilbo, 1993).
Another very important aspect which makes the prediction possible is the scheduled turn-around time (Morisset and Odoni, 2011). Turn-around time can be defined as the time between the arrival of the aircraft and period it departs from the same airport. (Wang, Schaefer and Wojcik, 2003) established a relations and existence of slack and departure time grant in turn-around time as well as flight time which absorb most delays the preceding flights. The turn-around time is a very significant variable and can be used to establish all the possible subsequent delays in a flight other variables remaining constant (Wang, Schaefer and Wojcik, 2003).
The use of In-bound delay is of great significance in the development of forecast delays in airports. An in-bound delay is the accumulated delay from an upstream airports and en-route legs. For a multiple flight-leg aircraft, in-bound delay at any station is as a result from a delay in the previous legs (Santos and Robin, 2010). Through this variable, researchers have come up with a delay network model with which the impacts of specific delays are put together and predicted. With in-bound delay being the initial delay to identify how it affects down the line delays. Boswell and Evans, (1997) approximated the prospective correlation between the arrival delay of flight leg coming as well as the arrival delay of outbound leg without outbound leg operational delays. There’s an approximation of the coefficient of arrival delay from the past leg to the delay in departure of the present leg within the airport. Through these in-bound delay studies, the subsequent delays can be established and prediction made accurately.
A variety of prediction model research studies in delays of flights have used different factors to predict the delay of flights in different ways. Diana, (2011) made use of spatial analysis to come up with a delay prediction model. In her study, factors considered included the weather, the configuration of the runway, wind angle, arrival demand, departures and arrivals, time taken for departure delays and arrival delays. Regression technique was used to develop the spatial error model (Diana, 2011). The outcome of it established the effect of closely spaced airports on one another. The resultant establishment of the relationship between the different aspects of delays in airports close to one another is very important as it proves the possibility of prediction of flight delays. In the study for instance, there was close connection between the time of arrival demands at the New York airports and the period of delay at the Newark airport. This information is useful in the larger prediction model schema of flight delays.
Abdel-Aty et. al, (2007) developed a two-step approach model which identified two-step approach which first identified the common delay periods and then examined the relationship of frequencies in delay. This establishment accommodated seasonal, day of week, time of day and particular date delay commonalities and fluctuations. From this study, higher levels of study were perceived on two consecutive days, and in summer months. The data from this can be used to implicatively predict the flight delays by looking at the previous flight frequencies correlations.
Prediction of flight delays is made possible by the Xu, Sherry & Laskey, (2008) model which made use of Multivariate Adaptive Regression splines way to put together a broader variables. In the model, several weather associated aspects particularly those convective activities with significant influence on the movements of aircrafts. The model utilized the Convective Weather Detection database to improve the delay prediction power (Xu, Sherry & Laskey, 2008). Other key variables the model made use of include the ratio of operational demand against the airport facility in-take, en-route weather, aircraft swap, time of day and the carrier delays. The use of historical data provided very accurate prediction of delays within five minutes. In the same way, historical index data can be used to make delay predictions in airlines.
Another key study used to explain the predictability of flight delays is the dynamic model. This model considered severe and particularly convective weather for reliable forecasting of flight delays. In the study, Sridhar and Chen, (2009) brought pointed out that up to 60% of flight delays was as a result of convective weather activity such as thunderstorm. The dynamic model, predicted-WITI was developed into an autoregressive model using exogenous inputs (Sridhar and Chen, 2009). This is the most important study in which update is done frequently with forward-looking functions and little retrospective views making it a viable tool for use in the prediction of flight delays.
The use of information regarding weather conditions, congestion in airports and the current flight delays allows quite accurate possible prediction of delays in the future because some of the parameters influencing these delays are known, despite having a random component (Mueller and Chatterji, 2002). Some sites have consulted several information necessary to establish the probabilities of being on-time, less than an hour late and even more than one hour to travelers. In many occasions, travelers use the information found on these sites have been found to be very helpful as they help identify the possible delays of flights in particular airports and at a particular time.
While the above arguments have pointed out to the possibility of developing accurate delay predictions of airports, there are some aspects which work against the predictions. In the case of weather for instance, there is a tremendous consensus of its influence on the airport arrival delays. This exposure is nothing new; it does not bring out the accurate prediction weather factors which mostly influence the delay of airport arrivals and forthcoming forecasts. Of course such model as visual conditions, wind and their influence on the airports have been identified but do not investigate further on how they affect the distinct weather phenomenon of particular airports and in explicit terminals. Conclusion from the many literatures above can be drawn that each airport has its particular distinctive issues that influence delays in and in the region within the facility and most of which cannot be detailed using a generic delay-prediction models. This makes it difficult to determine the delay patterns of a particular flight as several weather conditions come into play.
Other aspects that complicate flight delay prediction include the route taken by aircraft and the taxi conditions. Airplane arriving from particular directions throughout the runway configurations and time of day may earn diverse delay levels. This makes it hard to identify the pattern which produces a more precise delay profile. Similarly, it is true regarding the taxi conditions- taxi-in and out delays which may obviously affect the capacity to efficiently move the aircraft within the airport. This therefore hinders an accurate development of predictive aircraft delays.
All the literatures discussed and many others provided a major approach into the causes of delays in air traffic and the possible ways to predict them, however, numerous critical concerns have been left out. One of these considerations is that air traffic management principles can either influence in a positive or negative the air traffic delays. Besides, other facets such as the NextGen and SESAR which play significant roles have been left out. While these two aspects are of great importance, many studies have ignored their impacts and the possible ways to use them in predictions making it hard to accurately forecast the same.
Another constrain that makes it hard to successfully predict the delay patterns of aircraft is the absence of a coordinated model taking into account the predictions from previous prediction research studies and combining them with the input from various air traffic management experts (Abdel-Aty, 2007).. The absence of considerations required to strike a balance between including a sufficient number of feasible variables into a model with nod, inflated variance would create unwanted over-saturation of a delay forecast with too many variables. This would render the aircraft delay prediction less feasible (Abdel-Aty, 2007).
In order for the SESAR, NextGen and any other imminent developments to air transport management to thrive, premium, real-time delay prediction information is required to ensure that optimum efficiency in networks are established. The present prediction models observed above are either too general or inadequately condensed into intricacy of air traffic flows and delays. It is also of great importance to give considerations to other aspects beyond the weather as well as the infrastructural limitations. Through such, a wider perspective of delay in aircraft predictions can be brought into the table and be used to successively predict the delay patterns of flights. Additionally, the constrains that limited the success of predictions of flight delays can be dealt with by designing a more detailed study and investigation into the production of an airport arrival delay prediction model. Additionally, identification of the need to customize the different delay models for entity airports as well as terminal environments for more precise patterns in development of successful delay predictions.
Often times, many researchers in their studies have attempted to explore the full delay distribution probability. While these attempts have seen success in different aspects and failures in others, a simpler predictor probability provider can be used to determine if there will be delay or not. This is what is referred to as binary classifier, diving a prediction of utilizing a set of parameters associated to the observations. Binary classifier is discrete, returning only False (in the case of no delay) and true (in the case of delay). Through this model, the performance can easily be measured with the classes of test samples on delay predictions.
Essentially, there exist several models that can be used in the prediction of flight delays. Some of these models include the unified models, route-based models, the kernel models and other combinations. All of these have attempted to predict the flight delays using different categorical parameters and taking into account the diverse data set from different geographical areas and airports across the globe. Some of the parameters used in the prediction of flight delays are the weather (convective weather) including wind, thunderstorms and smoke, congestion capacity at the airports, GDP holding time at the airport for the outbound flights, scheduled turn-around time, carrier delay, aircraft substitution, en-route severe weather report, in-bound delay and scheduled departure. The data collected from the dataset collected from the previous observations dataset can be used to accurately predict flight delays and its subsequent delays. This however can be challenged by random variables which may make it difficult for accurate prediction to be achieved. While this may happen, it does not paralyze the possibility of making accurate prediction of flights.