U. S. Department of Transportation
Conclusions and Next Steps What have we learned?
Download 2.66 Mb.
|
- Navigate this page:
- Further Research Ideas
Conclusions and Next Steps
The research described in this report covered many different aspects of runway incursions. While some of the results were inconclusive, many provide specific steps for further research. Some of the insights were relevant to incident type distribution as well as severity. While not a central objective of this research, these conclusions regarding incident type can provide valuable insights. A summary of the results is contained in Appendix C:. One major conclusion is that OE incidents tend to be more severe than other incident types. The reasons for this are unclear at the moment. It is potentially a product of the nature of controller errors. Alternatively, current training practices may already be effective in preventing category D OE incursions. This disparity in incident severity has policy implications: policy directed at a particular incident type will not impact severity uniformly. Another strong conclusion is related to the regional distribution of incident type and severity. Both incident type and severity vary systematically between regions. Differences in pilot populations as well as traffic levels may impact this. This also indicates that any policy action will have disparate effects between regions, and that must be taken into account when crafting policy goals and responses. Commercial carrier status also has a clear impact on severity. Commercial carriers tend to be involved in less severe incidents. Despite this lower severity overall, commercial carriers are more likely to be involved in conflict events, potentially due to operating at busier airports. However, once the conflict versus non-conflict dynamic has been controlled for, commercial carriers are less likely to be involved in severe incidents. This relationship holds true even for OE incidents – suggesting that pilot skill and experience may play a role even when they are not responsible for the error. The phase of flight during which the incursion occurs appears to impact severity as well. Incursions during takeoff appear to be more likely to be severe than those when the aircraft at fault is taxiing or landing, once other controlling variables are included. The preliminary models described in this report also give no indication that controller age impacts incident severity. It may be that there is no effect of age or it may be that the increased experience associated with increased age counteracts any impacts. While still preliminary, it is encouraging that there is no effect, and the results do not suggest a change to current policies surrounding controller age. Controller workload – the number of aircraft a controller is responsible for – plays a significant role in severity. Increased workload is associated with higher probabilities of severe events. The positive relationship between severity and workload conforms to current expectations. Nonetheless, it is helpful to quantify that relationship – five additional aircraft increase the probability of a category A incursion by approximately .03 – and to provide statistical evidence to support intuition. Airport layout also appears to influence severity; this is intuitive, but with the analysis offers tangible statistical support. Evidence indicates that more runway intersections are associated with higher probabilities for severe events. There is also evidence that more runways (for any fixed number of intersections) help reduce the probability of severe events. These two results combined indicate that more parallel (or at least non-intersecting) runways may be a way to reduce the likelihood of severe events. In general, it does not appear that radar systems play a role in severity. There is some marginal evidence that STARS may help reduce severity, but it is tentative, at best. It is possible that radar still helps reduce the rate of runway incursions; however, such frequency models could not be run with the data provided for this study.
Finally, while not based on statistical tests, there are a series of observations about the general distribution of variables and severity that are informative. These insights may not be as useful from a policy perspective, but provide a richer context for understanding incursion severity:
In addition to the conclusions outlined in the paper and mentioned above, several questions arose during the research process. These questions fall into two major categories. The first category is additional research on how variables impact severity. The second category is extensions of this research into other areas. Throughout the paper, specific variables that would benefit from particular follow-up were identified. These can serve as springboards for more focused research into how severity may be impacted, likely with a combination of additional data, statistical analysis, and support from human factors and other safety experts. A full list of these variables and topics can be found in Appendix D: Future Research. Beyond those specific variables, there are three major classes of variables that require in-depth study. First, and perhaps the largest group, are pilot variables. Some of the pilot variables were addressed through cross tabulations. However, a modeling effort focused on pilot variables and PD incidents would be beneficial. The results in this paper pertain mostly to OE incidents, which represent the smallest absolute number of incidents. A better understanding of pilot incidents would help in minimizing the impact of that category. A similar suggestion holds true for V/PD incidents, although information surrounding them is less available. This suggestion cross-cuts all the below suggestions as they too may vary by type of incident. Secondly, the weather variables warrant further examination. It is clear that there is some relationship between severity and weather conditions, but it is unclear what the specific causes are. Likely, that relationship is being driven by underlying behavioral responses to weather rather than the weather itself. This preliminary research identified a need to understand these variables and a more focused examination might better explain their impact. Thirdly, the controller variables require a more thorough examination. The controller variables in particular were plagued by data problems and small sample sizes.100 It is surprising that no controller attributes contributed to the severity of controller incursions. A more focused examination – perhaps using more accurate controller data – might reveal some trends. Aside from specific variables to follow up on, another fruitful area for research would be frequency modeling. As noted throughout the research, these insights pertain only to severity, given that an incursion has already occurred. It is possible that many of these variables contribute to the underlying rate of incursions, but not their severity. Additionally, some variables may impact both severity and frequency. Frequency and severity are two sides of the same safety issue. To gain a complete understanding of the problem, frequency must also be modeled. This is the most beneficial next step, even if that frequency research were focused only on OE incursions.
|