Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.

Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 13

13
CHAPTER 3
Modeling RSA Risk
The analysis of RSA risk requires three models that con- in the terrain or into a water body adjacent to the RSA. In ad-
sider probability (frequency), location and consequences. The dition, it takes into consideration the type of obstacle and the
outcome of the analysis is the risk of accident during runway estimated collision speed to cause severe consequences. For
excursions and undershoots. The three model approach is example, an aircraft colliding with a brick building may result
represented in Figure 15. in severe consequences even at low speeds; however, the air-
The first model is used to estimate the probability that an craft must be at a higher speed when striking a Localizer an-
event will occur given certain operational conditions. This tenna mounted on a frangible structure for a similar level of
probability does not address the likelihood that the aircraft severity. The collision speed is evaluated based on the loca-
may strike an obstacle or will stop beyond a certain distance. tion of the obstacle and the typical aircraft deceleration for
The model uses independent variables associated with causal the type of RSA terrain. The ensuing sections provide details
and contributing factors for the incident. For example, under for each component of the risk model. The same model is used
tailwind conditions it is more likely that an overrun will occur, for all five types of events (LDUS, LDOR, LDVO, TOOR,
and this is one of the factors used in the models for overruns. and TOVO.)
The aircraft performance is represented by the interaction be- The remainder of this chapter discusses the probability
tween the runway distance required by the aircraft for the and location models. The consequence model is discussed in
given conditions and the runway distance available at the air- Chapter 4.
port. Although human and organizational factors are among
the most important causes of aircraft accidents, it was not pos-
Event Probability (Frequency Model)
sible to directly incorporate these factors into the risk models.
Since this model is specific for the event type, five different Similar to ACRP Report 3 model development procedures,
models are required, e.g., one for landing overruns and another backward stepwise logistic regression was used to calibrate
one for takeoff overruns. five frequency models, one for each type of incident: LDOR,
The second component is the location model. The analyst LDUS, LDVO, TOVO, and TOOR. Various numerical tech-
usually is interested in evaluating the likelihood that an air- niques were evaluated to conduct the multivariate analysis,
craft will depart the runway and stop beyond the RSA or and logistic regression was the preferred statistical procedure
strike an obstacle. The location model is used to estimate the for a number of reasons. This technique is suited to models
probability that the aircraft stops beyond a certain distance with a dichotomous outcome (accident and non-accident)
from the runway. As pointed out in ACRP Report 3 and by with multiple predictor variables that include a mixture of
Wong (2007), the probability of an accident is not equal for all continuous and categorical parameters. Also, it is particu-
locations around the airport. The probability of an accident in larly appropriate for case-control studies because it allows
the proximity of the runways is higher than at larger distances the use of samples with different sampling fractions, depend-
from the runway. Since this model is specific for the event ing on the outcome variable without giving biased results.
type, five different models are required, e.g., one for landing Backward stepwise logistic regression was used to calibrate
overruns and another one for takeoff overruns. the frequency models because of the predictive nature of the
The last part is the consequence model. This model uses research, and the technique is able to identify relationships
the location models to assess the likelihood that an aircraft missed by forward stepwise logistic regression (Hosmer and
will strike an obstacle or depart the RSA and fall into a drop Lemeshow 2000).

OCR for page 13

14
Three-Part Risk Model
Event Location Consequences
probability probability
operating conditions
(airplane performance, type of RSA characteristics, type, size and
operation, runway distance geometry, location of
available and elevation, presence of EMAS obstacles
weather conditions)
Risk Classification
Figure 15. Modeling approach.
To avoid the negative effects of multi-co-linearity on the the actual runway had not been identified for the NOD. The
model, correlations between independent variables were first research team has gathered information on the runways used,
tested to eliminate highly correlated variables, particularly if and the process allowed the calculation of the head/tailwind
they did not significantly contribute to explaining the varia- components to be included in the model.
tion of the probability of an accident. Another major improvement that has increased model ac-
The basic model structure selected is a logistic equation, as curacy was the inclusion of a runway criticality factor. The
follows: basic idea was to include a new parameter that could repre-
sent the interaction between the runway distance required by
1 the aircraft and the runway distance available at the airport.
P { Accident _ Occurence } = The log of the ratio between the distance required and the dis-
1+ e b0 +b1X1 +b2 X 2 +b3 X 3 + . . .
tance available was used. Positive values represent situations
where when the distance available was smaller than the distance re-
P{Accident_Occurrence} = the probability (0100%) of an quired, and in this case, risk will be higher. The greater the
accident type occurring given value is, the more critical is the operation because the safety
certain operational conditions; margin decreases.
Xi = independent variables (e.g., The distance required is a function of the aircraft perform-
ceiling, visibility, crosswind, ance under specific conditions. Therefore, every distance re-
precipitation, aircraft type, cri- quired under International Standards Organization (ISO)
ticality factor); and conditions (sea level, 15 degrees centigrade) was converted to
bi = regression coefficients. actual conditions for operations. Moreover, the distances
were adjusted for the runway surface condition (wet, snow,
Several parameters were considered for inclusion in the slush, or ice) and for the level of head/tailwind. The adjust-
models. The backward stepwise procedure helps identify those ment factors applied to the distance required are presented in
variables that are relevant for each type of event. Some of the Table 2. A correction for slope was not applied to adjust the
independent variables are converted to binary form to avoid total distance required.
spurious effects of non-linear relationships in the model ex- The use of NOD in the accident frequency model was a
ponent. These binary variables are represented by only two major improvement introduced by Wong et al. (2006) and
values, 0 or 1. In this case, the presence of the factor (e.g., rain) was maintained for this study. The analysis with NOD also
is represented by 1, and the absence of the factor (e.g., no rain) adds to the understanding of cause-result relationships of the
is represented by 0. five accident types.
One significant improvement relative to the models pre- The technique used to develop the models is able to iden-
sented in ACRP Report 3 is the use of tailwind and headwind. tify relationships missed by forward stepwise logistic regres-
These variables were not present in previous models because sion (Hosmer and Lemeshow 2000). The predictor variables