U. S. Department of Transportation
Factors leading to Traffic Complexity
Download 2.66 Mb.
|
- Navigate this page:
- Part 139 Airport Status
- OEP 35 Airport Status
- Land and Hold Short Capability at Airport
- Daily Operations
- Percent of Operations that are Air Carrier / Air Transport
- Number of Runway Intersections
- Number of Runways
- Number of Hotspots
Factors leading to Traffic Complexity(ATQA OE) This is a set of variables describing the factors leading to the traffic complexity ranking given in the previous variable. As with Traffic Complexity Code, it only applies to OE incidents. Figure 17 provides the frequency of yes and no by each factor, as well as the number of missing values. 
Figure – Frequency of Factors Leading to Traffic Complexity In many of the cases, it is unclear how these factors may interact with traffic complexity, let alone severity. For example, “Experience” may indicate a lack of experience or that the controller’s higher level of experience reduced the complexity. Additionally, the quality of the data is called into question as the flag for “N/A” is indicated alongside other factors. No test statistics are reported for these variables and any interpretation of them is likely erroneous. They are reported here to bring to light the problems in the data that prevent additional analysis.
This variable indicates whether the airport at which the incursion happened is categorized as a Part 139 airport.46 Table 70 and Table 71 present the distribution of this variable by severity. Note that a significant Chi-Squared statistic is also reported indicating some relationship between the severity of the event and Part 139 statistics. This is likely due to the higher traffic at Part 139 airports in general compared to non-Part 139 airports. Figure 18 presents the overall distribution of this variable. Table 72 and Table 73 report the same results, but limited to only conflict events (categories A through C). After removing category D events from the comparison, the significant relationship is no longer detected, indicating that the result seen in Table 70 is likely driven by the disparity between conflict and non-conflict events, which is itself based on the activity level of the airport, rather than on a real relationship with severity.
Figure – Distribution of Part 139 Status Table – Observed Distribution of Part 139 Status by Severity
Table – Expected Distribution of Part 139 Status by Severity
Table – Observed Distribution of Part 139 Status by Severity, Conflict Only
Table – Expected Distribution of Part 139 Status by Severity, Conflict Only
Future Research
Table 74 and Table 75 reports the distribution of incident type by Part 139 status. The Chi-Squared statistic indicates that there is also a relationship between incident type and Part 139 status. The expected values indicate that this is likely due to an overrepresentation of OE and PD incidents and a corresponding underrepresentation of V/PD incidents among Part 139 airports. Table 76 and Table 77 reports the same information, excluding V/PDs. The reported Chi-Squared statistic indicates that the relationship detected in Table 74 is observed again. Here, it appears that PDs are observed less frequently than expected at Part 139 Airports and the opposite is true for OE incidents. It is unclear why this disparity exists among incident types; further research in the prevalence of different incident types by Part 139 status is required to understand what is reported in Table 74 and Table 76. Table – Observed Distribution of Part 139 Status by Incident Type
Table – Expected Distribution of Part 139 Status by Incident Type
Table – Observed Distribution of Part 139 Status by Incident Type, OE & PD
Table – Expected Distribution of Part 139 Status by Incident Type, OE & PD
Table 70 and Table 72 addressed the issue of severity and Part 139 status. The results presented in those two tables indicate that any relationship between Part 139 status and severity is a product of the conflict/non-conflict event dynamic. Therefore, due to the loss of information from combining the categories, it is unlikely that an effect would be detected related to severity in the logit framework. Table 74 and Table 76 addressed the issue of incident type and Part 139 status. The results presented in Table 78 indicate that there is a relationship with incident type and that incidents at Part 139 airports have twice the odds of being an OE as non-Part 139 airports. All incursions, and thus airports, included in this analysis are controlled. It is possible that the disparity between Part 139 and non-Part 139 airports may be related to the differing pilot populations between airport types. As noted earlier, further research into why OE incidents are more common at Part 139 airports is warranted. Table – Logit Estimate of Impact on Incident Type, Part 139 Status
OEP 35 Airport Status(Runway Incursion Database) This variable indicates whether or not the airport at which the incursion occurred is considered part of the OEP 35, the 35 busiest airports in the country in 2000. Though OEP 35 is used in this analysis, the same results hold for the Core 30, the 30 airports of interest to the FAA in 2011, a designation the FAA is using going forward. Figure 19 presents the overall distribution of OEP 35 status. 
Figure – Distribution of OEP 35 Status Table 79 presents the estimated effect on the odds of being severe if an incident occurs at an OEP 35 airport. Table – Logit Estimate of Impact on Severity, OEP 35 Status
The increase in the odds of a severe incident is moderate compared to some of the other variables examined. Given that OEP 35 airports are extremely busy, it is possible that this relationship is merely a product of the higher likelihood of conflict events at a busy airport. Table 80 presents the same estimate, excluding category D incursions. Not surprisingly, the previous relationship is now not detected, indicating that OEP 35 status is likely a better indicator of conflict versus non-conflict rather than severity. Table – Logit Estimate of Impact on Severity, OEP 35 Status, Conflict Only
Table 81 presents a look at the impact on the odds of being an OE. Interestingly, the impact on OEs is fairly strong, increasing the odds by around 170%. Given the relationship between OE incidents and severity, it is prudent to check if the impact on severity is an independent effect. That is, given that OEP 35 incidents are more likely to be OEs and that OEs are also likely to be more severe, it is not surprising that OEP 35 incidents are more severe. Table 82 presents a multivariate logit that controls for this relationship and examines the impact on severity. Table – Logit Estimate of Impact on Incident Type, OEP 35 Status
Table – Logit Estimate of Impact on Severity, OEP 35 Status and Incident Type
Future Research
The results indicate that not only is there an independent impact on severity, there is an interaction between OEP 35 status and incident type. The results indicate that incidents at OEP 35 airports tend to be more severe, OE incidents tend to be more severe, but OE incidents at OEP 35 airports are less severe than the combination would suggest – there is a mitigating factor in the interaction of OEP 35 status and incident type. Table 83 presents the same results but excludes category D incidents. Here, the independent OEP 35 impact is no longer detected, but the interaction is still detected – though just barely. It is possible that this mitigating factor is related to controller experience or skill (broadly defined). That hypothesis would indicate that only the most skilled controllers are at the OEP 35 airports and they make less severe mistakes than their non-OEP 35 counterparts. That is only one hypothesis and is difficult to test. A deeper understanding of the differences in controllers between OEP 35 airports and non-OEP 35 airports is required to formulate better hypotheses and to test them adequately. Table – Logit Estimate of Impact on Severity, OEP 35 Status and Incident Type, Conflict Only
Land and Hold Short Capability at Airport(Airport Database) This variable indicates if an airport is capable of LAHSO operations. This is in contrast to the variable described previously in Table 33, which indicates if one of the aircraft involved was performing a LAHSO. Figure 20 contains the overall distribution for this variable. 
Figure – Distribution of LAHSO Capability at Airport Table 84 and Table 85 present the distribution by incident type. Table 84 indicates that OE and PD incidents are observed more frequently than expected with a corresponding underrepresentation of V/PD incidents. This relationship is found to be statistically significant. It may be possible that LAHSO capability is correlated with other factors such as Part 139 status and overall traffic levels. Thus, it may be that LAHSO capability is correlated with something that creates this disparity in incident types, rather than LAHSO capability being the driving factor behind the disparity. Table – Observed Distribution of LAHSO Capability by Incident Type
Table – Expected Distribution of LAHSO Capability by Incident Type
Table – Correlation of LAHSO Capability and Other Airport Characteristics
Table 86 presents the correlation of LAHSO Capability with other relevant airport characteristics. Interestingly, it is not highly correlated with any of these factors. The relationships seen in Table 84 and Table 85 cannot be attributed to that correlation. Further research is warranted to better understand how LAHSO capability is correlated with incident type. Table 84 indicated a significant relationship between incident type and LAHSO capability at an airport, while Table 87 provides an estimate of the impact LAHSO capability has on the odds of being an OE (i.e., an OE has odds 71% higher under LAHSO capability). Table – Logit Estimate of Impact on Incident Type, Land and Hold Short Capability at Airport
Table 88 and Table 89 present the distribution by severity. There is no relationship between severity and LAHSO capability, as indicated by the insignificant Chi-Squared statistic. Combining this result with previous results raises some interesting questions. The logic chain is as follows: Future Research
The results in Table 84 indicate that OEs are more common than expected at LAHSO capable airports.
Further research into severity, incident type, and LAHSO capability might help clarify this surprising (lack of) relationship. Table – Observed Distribution of LAHSO Capability by Severity
Table – Expected Distribution of LAHSO Capability by Severity
Daily Operations(OPSNET) As noted in Section 3.1.6, operations are available on a variety of time scales: hourly, daily, and annually. The ideal operations measure is both granular and accurate. The hourly counts provided by ETMSC are the most granular option available, but due to the way VFR operations are allocated to hours of the day, the accuracy of the data is questionable, at best. In fact, the allocation procedure indicates that some incursions happened in hours with zero operations, which is extremely unlikely. Yearly operations are much more stable, but do not offer the granularity that may be important as operations vary throughout the year. Daily operations offer a good mix of granularity and accuracy. Figure 21 presents the distribution of this variable overall, and by severity. Table 90 presents the median daily operations by severity while Table 91 presents the results of a Kruskal-Wallis test. 
Figure – Distribution of Daily Operations Table – Percentiles of Daily Operations
Table – Kruskal-Wallis Test Results for Daily Operations
[D]aily operations are likely a better determinant of conflict versus non-conflict event rather than… severity The results of the Kruskal-Wallis indicate that daily operations jointly differ across severity categories. Category D appears to have many fewer median daily operations than any of the other categories. The pairwise comparison tests indicate that categories A, B, and C can all be distinguished from D statistically. However, categories A, B, and C are pairwise indistinguishable. This indicates that daily operations are likely a better determinant of conflict versus non-conflict event rather than contributing to severity. Percent of Operations that are Air Carrier / Air Transport(Airport Database) This variable indicates the average percent of traffic at an airport that is categorized as Air Carrier or Air Transport.47 Figure 22 presents the distribution of this variable by severity, while Table 92 presents the percentiles of the distribution. Table 93 reports the results of a Kruskal-Wallis test by severity. 
Figure – Distribution of AC/AT Percent of Operations Table – Percentiles of AC/AT Percent of Operations by Severity
Table – Kruskal-Wallis Test Results for AC/AT Percent of Operations
[P]olicy interventions need to account for traffic mix at an airport Interestingly, all severity levels appear to have similar medians, with the values for category C tending to be a bit higher. Additionally, the interquartile range for category D incursion appears to be smaller, indicating a more narrowly distributed variable (especially given the overwhelming prevalence of category D). The result of the Kruskal-Wallis test supports the conclusion that the categories are jointly different, but offers little information for the pairwise comparisons. Category C can be distinguished from category D, but no other pairs are significantly different. This may indicate that high percentage AC/AT airports are also very busy and are thus unlikely to commit an error in the absence of another aircraft. Further exploration will need to control for the operations at the given airport to disentangle the two effects. Table 94 and Table 95 examine the percent of operations that are AC/AT by incident type. All three incident types appear to have very different distributions. OE incidents have a higher median percentage while V/PD incidents have the lowest. The results of the Kruskal-Wallis test corroborate this, indicating that the three incident types are jointly different as well as all pairwise different from each other. This suggests that policy interventions need to account for traffic mix at an airport. That is, any policy intervention targeted predominately at one kind of airport will have differing impacts on severity and incident types across airports. Table – Percentiles of AC/AT Percent of Operations by Incident Type
Table – Results of Kruskal-Wallis Test for AC/AT Percent of Operations by Incident Type
Number of Runway Intersections(Airport Database) This variable measures the number of runway intersections at the airport where the incursion occurred. Figure 23 and Table 96 gives the distribution of this variable. Table 97 gives the results of a Kruskal-Wallis test by severity. 
Figure – Distribution of Number of Runway Intersections Table – Percentiles of Number of Runway Intersections by Severity
Table – Kruskal-Wallis Test Results for Number of Runway Intersections
On a pairwise basis, only categories C and D can be considered different. Table 98 presents the results of a Kruskal-Wallis test, examining conflict only events. The three severity categories can no longer be considered jointly different. This indicates that number of runway intersections is helpful for identifying conflict or non-conflict events but not severity among conflict events. Table – Kruskal-Wallis Test Results for Number of Runway Intersections, Conflict Only
Number of Runways(Airport Database) This variable indicates the total number of runways at the airport where the incident occurred. Note that this is not the number of runways in operation at the time of the incident. A measure of the number of operating runways was unavailable and future research may want to explore how that impacts severity. Figure 24 and Table 99 present the distribution of the number of runways. Table 100 presents the results of a Kruskal-Wallis test by severity. The results of the Kruskal-Wallis test indicate that there is a difference in number of runways between the severity categories. Examining the distribution indicates that category D appears the most different in terms of percentiles. Additionally, only categories C and D can be considered pairwise different. It is likely that the observed relationship will be no longer significant once category D incursions are excluded from the analysis (controlling for the conflict versus non-conflict dynamic).
Figure – Distribution of Number of Runways Table – Percentiles of Number of Runways by Severity
Table – Kruskal-Wallis Test Results for Number of Runways
Table 101 presents the results of a Kruskal-Wallis test examining conflict events only. As expected, the relationship between number of runways and severity is no longer significant. It is likely that number of runways is a proxy for overall traffic levels and likelihood of two planes conflicting. A similar argument may hold for the number of runway intersections. Table – Kruskal-Wallis Test Results for Number of Runways, Conflict Only
Number of Hotspots(Airport Database) This variable indicates the number of hotspots identified at an airport. A hotspot is defined as “a location on an airport movement area with a history of potential risk of collision or runway incursion, and where heightened attention by pilots and drivers is necessary.”48 Table 102 and Figure 25 present the distribution of this variable while Table 103 presents the results of a Kruskal-Wallis test. The severity categories are jointly different while only categories C and D can be considered pairwise different. As with total runways and runway intersections, it is instructive to examine conflict events only. The evidence is weaker for conflict only events, as seen in Table 104. However, the change is not as dramatic as for number of runways or number of runway intersections. Thus, number of hotspots appears to be most useful in identifying conflict versus non-conflict events but may also provide some additional information regarding severity for conflict events.
Figure – Distribution of Number of Hotspots Table – Percentiles of Number Hotspots by Severity
Table – Kruskal-Wallis Test Results for Number of Hotspots
Table – Kruskal-Wallis Test Results for Number of Hotspots, Conflict Only
|
