PLEASE SCROLL DOWN FOR ARTICLE

Transcription

1 This article was downloaded by:[coyne, Joseph T.] On: 31 March 2008 Access Details: [subscription number ] Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK International Journal of Aviation Psychology Publication details, including instructions for authors and subscription information: Pilot Weather Assessment: Implications for Visual Flight Rules Flight Into Instrument Meteorological Conditions Joseph T. Coyne a ; Carryl L. Baldwin b ; Kara A. Latorella c a U.S. Naval Research Laboratory, Washington, DC b Old Dominion University, Norfolk, VA c NASA Langley Research Center, Hampton, VA Online Publication Date: 01 April 2008 To cite this Article: Coyne, Joseph T., Baldwin, Carryl L. and Latorella, Kara A. (2008) 'Pilot Weather Assessment: Implications for Visual Flight Rules Flight Into Instrument Meteorological Conditions', International Journal of Aviation Psychology, 18:2, To link to this article: DOI: / URL: PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

2 THE INTERNATIONAL JOURNAL OF AVIATION PSYCHOLOGY, 18(2), Copyright 2001 Taylor & Francis Group, LLC ISSN: ISSN print / online DOI: / Pilot Weather Assessment: Implications for Visual Flight Rules Flight Into Instrument Meteorological Conditions Joseph T. Coyne 1, Carryl L. Baldwin 2, and Kara A. Latorella 3 1 U.S. Naval Research Laboratory, Washington, DC 2Old Dominion University, Norfolk, VA 3NASA Langley Research Center, Hampton, VA Visual flight rules flight into instrument meteorological conditions (IMC) account for over 10% of the fatalities in general aviation. Evidence suggests that pilots assessments of weather conditions are related to their decision to continue. This study investigated pilots ability to assess ceiling and visibility in a flight simulator. Assessment accuracy did not differ between instrument- and non-instrument-rated pilots for ceiling accuracy, but visibility accuracy was better for non-instrument-rated pilots. The data indicated pilots allowed their estimates of ceiling and visibility to influence each other. That is, pilots tended to judge a ceiling to be higher than it actually was when it was paired with a high visibility. This interaction may play a significant role in pilots decisions to continue into IMC. It is difficult to overstate the importance of weather in general aviation (GA) flight safety. Weather is a factor in 15% of all fatal GA accidents, and weather accidents are more likely to be fatal than any other type of GA accident (AOPA Air Safety Foundation, 2002). Inadvertent visual flight rules (VFR) flight into instrument meteorological conditions (IMC) is a significant contributing factor in 90% of all fatal weather-related accidents. The Federal Aviation Regulations (FARs) provide specific guidelines for pilots regarding the weather conditions necessary to conduct a VFR flight. The FARs also prohibit non-instrument-rated pilots from flying when conditions fall below these VFR minimums. However, 75% of the pilots involved Correspondence should be sent to Joseph T. Coyne, U.S. Naval Research Laboratory, Code 5511, 4555 Overlook Avenue Southwest, Washington, DC coyne@itd.nrl.navy.mil

3 154 COYNE, BALDWIN, LATORELLA in VFR into IMC accidents do not hold an instrument rating (AOPA Air Safety Foundation, 1996). Because non-instrument-rated pilots are prohibited from flying in IMC this suggests that these pilots either disregarded the FARs or were unaware of the weather conditions. The serious nature of the problem has prompted several empirical investigations (Goh & Wiegmann, 2001, 2002a, 2002b; O Hare, Owen, & Wiegmann, 2001; O Hare & Smitheram, 1995; Wiegmann, Goh, & O Hare, 2002). This research has investigated a number of variables that influence pilots decisions to continue into deteriorating weather conditions. These can be broken into two general categories: poor situation assessment and improper motivation. The situation assessment hypothesis suggests that pilots continue into IMC because they do not completely realize that the weather conditions no longer support a VFR flight (Goh & Wiegmann, 2001). The alternative hypothesis is that pilots realize conditions have deteriorated but continue because of some other type of external motivation, such as financial incentive or social pressure. The situation assessment hypothesis implies that if pilots are able to recognize the weather cues or have accurate information indicating they are flying into IMC, they will divert (Wiegmann et al., 2002). If this hypothesis is incorrect, then a pilot s decision to continue into IMC is a willful disregard of the FARs and the weather cues that would have indicated a safer course. Support for the situation assessment hypothesis comes from both accident data (AOPA Air Safety Foundation, 1996) and empirical investigations (Goh & Wiegmann, 2001; Wiegmann et al., 2002). The strongest support for the situation assessment hypothesis comes from a simulator study conducted at the University of Illinois (Goh & Wiegmann, 2001). Weather conditions at takeoff were VFR, but after 45 min flight conditions deteriorated to instrument flight rule (IFR) conditions (ceilings were 1,500 ft mean sea level [MSL] and visibility was 2 statute miles. Participants had a 5-min window from when the conditions deteriorated below VFR minimums until the experiment was terminated. If a decision was not made to divert during this time, participants were considered to have made a decision to continue. Pilots were asked to estimate ceiling, visibility, and distance to the airport at the time the simulation ended. Of the 32 pilots, only 10 made the decision to divert using the experimenter s cutoff. The most important factor in predicting this dichotomy was the pilot s estimation of visibility at the time the scenario ended. The pilots who decided to continue overestimated the visibility (conditions were worse than the pilots believed). Pilots who decided to divert had a mean error of 0 miles for visibility estimates compared to a mean error of 1.4 miles for the pilots who decided to continue. Pilots in both groups tended to overestimate the ceilings by about 2,200 ft. The study both revealed the importance of situation assessment in predicting VFR flight into IMC and demonstrated pilot s difficulty with interpreting out-the-window weather cues.

4 PILOT WEATHER ASSESSMENT 155 Additional support for a situation assessment hypothesis comes from research on how pilots integrate different aspects of the weather. According to FARs, a low ceiling or poor visibility alone can categorize conditions as IMC. Research sponsored by the Federal Aviation Administration (FAA) addressed the question of how pilots mentally combine different ceilings and visibilities (Driskill et al., 1997; Hunter, Martinussen, & Wiggins, 2003). Pilots were provided with textual information about cloud ceiling, visibility, and precipitation. This information was given within 81 textual scenarios consisting of three sets of 27. Sets were divided by terrain (water, mountainous, and nonmountainous). Each set included three levels of ceiling, visibility, and precipitation (each containing a high, medium, and low). Results from both FAA studies (Driskill et al., 1997; Hunter et al., 2003) indicated that pilots used a compensatory decision-making strategy. That is, pilots would mathematically combine the different aspects of the weather and create an overall assessment. Averaging severity levels of different weather conditions (as these pilots appeared to do) instead of making a judgment on the worst condition can result in an improper diagnosis. A VFR ceiling cannot compensate or average out an IFR visibility. Compensatory models of decision making are an efficient use of the available information, but they are not the optimal decision-making strategy. In fact compensatory models can place inexperienced pilots at a greater risk of being involved in a weather accident (Hunter et al., 2003). For example, inexperienced pilots cannot allow for good ceilings to compensate for poor visibility. If non-instrument-rated pilots cannot see where they are flying, they should not be flying. Indeed, for conditions to be VFR they must meet both ceiling and visibility requirements stipulated in the FARs. Despite a growing body of research, it remains unclear why pilots continue into deteriorating conditions. The situation assessment hypothesis provides one of the most plausible explanations, but it currently has only limited support (Goh & Wiegmann, 2001). The strongest support for the theory comes from a single estimation of weather conditions at the end of a simulation. This study takes a systematic approach to investigate pilots ability to estimate weather conditions. Based on previous research it is expected that pilots will be inaccurate and demonstrate a tendency to overestimate weather conditions (Goh & Wiegmann, 2001). Instrument-rated pilots represent half of the GA population, but only account for 25% of the VFR into IMC accidents (AOPA Air Safety Foundation, 1996). There is inadequate data to determine if the additional training received in the process of gaining an instrument rating results in more accurate weather assessments or if instrument-rated pilots simply have more experience making weather judgments as well as flying in IMC. This study compares a group of instrument-rated and non-instrument-rated pilots of similar cross-country experience levels on their ability to estimate simulated weather conditions. The in-

5 156 COYNE, BALDWIN, LATORELLA strument-rated pilots are expected to have less error when judging ceiling and visibility. METHODS Participants Twenty-four GA pilots, ages 19 to 76, participated in the experiment. One participant was female. Pilots were recruited, scheduled, and compensated through a contract with Lockheed-Martin. The contractor at Lockheed-Martin provided a database of pilots without military or commercial experience from the Mid-Atlantic region. Pilots were then selected based on their cross-country hours and instrument ratings. Instrument-rated and non-instrument-rated pilots were matched as best as possible on their cross-country hours. In addition to the pre-experiment data from Lockheed-Martin, pilots also supplied information on experience during the experiment at NASA Langley. There was some variation in the consistency of the pilot information obtained at NASA and from Lockheed-Martin. Several pilots cited that they used their logbooks to provide the information to Lockheed-Martin, but did not have the exact information available at the time of the experiment. In general, pilots reported more hours on the day of participation (which could also reflect additional hours since providing their information to Lockheed-Martin). The mean experience data provided on the day of the experiment are provided in Table 1. Apparatus and Materials All experimental data were collected in a laboratory chamber at NASA Langley. Two computers linked via a local area network were used to run the experiment. The first computer drove the out-the-window weather depiction. This computer had a Pentium GHz processor, 1 GB of RAM, and a GeForce FX 5950 Ultra video card with 256 MB of video RAM. The computer was linked to a Dell 3300 MP projector with a maximum brightness of 1,500 ANSI lumens, and a native res- TABLE 1 Mean and Standard Deviation Data for Pilot Experience Age Total Flight Hours Last 90 Days Cross-Country Hours M SD M SD M SD M SD Instrument-rated pilots Non-instrument-rated pilots

6 PILOT WEATHER ASSESSMENT 157 olution of pixels. The projected image was in. 26 in. The second computer displayed a static representation of the flight instruments and collected the participant s responses. The second computer had a Pentium Xeon 1.8 GHz processor, and 512 MB of RAM. The information was presented on a 17-in. Dell flat-panel display. Experimental Instructions Pilots were told that their objective was to view a series of simulated out-the-window video clips and answer several questions regarding the scene. The questions were to be answered as if there was preflight information obtained 2 hr ago that indicated a warm front might be coming through the area and there was a possibility of reduced ceilings and visibility. Weather Generation Program The out-the-window weather conditions were generated through Microsoft Flight Simulator The videos were augmented with satellite imagery and terrain overlays from MegaScenery and displayed via an overhead projector. The video resolution was pixels. Due to a limitation of the software, all of the clouds rendered were stratus clouds with a height of 900 ft. All of the videos taken from Flight Simulator used the same location over Brookhaven Airport (KHWV) in Shirley, NY. This location was selected based on several criteria. First the airport is approximately at sea level. This alleviates any confusion between using MSL and above ground level (AGL) when pilots estimate cloud ceiling. Second, the area had satellite overlays included in the MegaScenery software to provide a more realistic depiction. Third, the specific region had several features that could help pilots estimate distance, including an interstate (approximately 2 miles from the aircraft s position), and Calverton Naval Base (a former air station), and an operational very high frequency omnidirectional range (VOR) beacon (approximately 5 miles away). All of the videos collected were 5 sec in duration and start from the same location just east of KHWV. The accuracy of the depicted visibilities was verified against the distance to known landmarks in the simulation software. Procedure Pilots were briefed on the experiment intent and schedule for the day. They then read and signed a written informed consent document. The training included familiarization with the test area including images taken from sectional charts and a depiction of the area with no ceiling and high visibility. A screen shot depicting

7 158 COYNE, BALDWIN, LATORELLA landmarks and their approximate distances was given to each pilot. On completing the training, pilots began the main experimental trials. Each trial presented pilots with the 5-sec out-the-window video on a projector screen in front of them. The second, flat-screen display provided pilots with the primary flight instruments (altimeter, attitude indicator, and airspeed indicator) and the moving map display. On several trials the moving map display also contained a graphical representation of ceiling and visibility conditions. Only trials where no weather information was presented are included in this article. For information on the trials with graphical weather data please see Coyne, Baldwin, and Latorella (2005). The videos looped until the pilots answered a set of questions regarding the current situation including what they estimated the ceiling to be (height AGL) and what they estimated the visibility to be (statute miles). Pilots experienced 32 trials representing two complete blocks of each combination of the ceiling (4) and visibility (4) factors. The different out-the-window videos were presented via an overhead projector. Ceiling was manipulated by adjusting the base layer of the clouds. The definition of ceiling followed the definition of ceiling provided in the FARs, which is also used in aviation weather reports. Specifically, a ceiling is the lowest layer of clouds or obscuring phenomenon that is reported as broken, overcast, or an obscuration. Ceiling is reported as AGL. Cloud cover was presented at four different base levels (400, 900, 2,900, and 4,500 ft AGL). VFR conditions were those above 3,000 ft, marginal visual flight rules (MVFR) ceilings were between 1,000 and 3,000 ft, IFR conditions were between 500 and 1,000 ft, and low instrument flight rules (LIFR) conditions were below 500 ft. Visibility took on one of four distances, expressed in statute miles (2, 3, 5, and 10 miles). These four levels represented one VFR, two MVFR, and one IFR. RESULTS The ceiling and visibility estimates were used to create a ceiling proportional error (CPE) and visibility proportional error (VPE). CPE was computed as {(Estimated Ceiling Actual Ceiling)/Actual Ceiling}, negative values thus reflect a ceiling estimated below the actual depicted ceiling. The VPE was computed in the same manner. The CPE and VPE data were analyzed using a three-way, ratings (2) ceiling (4) visibility (4) mixed repeated measures analysis of variance. All significant main effects were further addressed by Tukey post-hoc analyses. CPE</CPE> There was no significant main effect of pilot rating on the CPE data, F(1, 22) = 1.126, p =.300. There was a significant main effect of ceiling on the CPE data, F(3,

8 PILOT WEATHER ASSESSMENT ) = , p <.001 (Figure 1). The proportional error for the 400-ft ceiling was significantly larger than all of the other ceilings. The 900-ft ceiling also had a significantly larger error than the 4,500-ft ceiling. There were no other significant differences between ceiling condition pairs. There was a significant main effect of visibility on the CPE data, F(3, 66) = , p <.001 (Figure 2). Post-hoc analysis revealed that the CPE for the 10-mile visibility conditions was significantly larger than all other visibilities. Additionally, the CPE data for the 5-mile visibility condition were significantly larger than for the 2-mile condition. There were no other significant differences between the visibility condition pairs. There was a significant interaction of ceiling and visibility on the CPE data, F(9, 198) = 6.136, p <.001 (Figure 3). A test of the simple main effect of visibility FIGURE 1 Main effect of ceiling on the ceiling proportional error data. Error bars reflect the standard error of the mean. FIGURE 2 Main effect of visibility on the ceiling proportional error data. Error bars reflect the standard error of the mean.

9 160 COYNE, BALDWIN, LATORELLA FIGURE 3 The interaction of ceiling and visibility on the ceiling proportional error data. Error bars reflect the standard error of the mean. Note. X axis represents discrete points and is not continuous. at each ceiling revealed a significant effect of visibility at the 400-, 900-, and 4,500-ft ceilings but not the 2,900-ft ceiling. The strength of the effect of visibility was the largest at the 400-ft ceiling. At the 400-ft ceiling, the 2-mile condition had significantly less error than the 5- and 10-mile conditions. Additionally at the 400-ft ceiling the 3-mile visibility was significantly less than the error at 10 miles. At the 900-foot ceiling the only significant difference was that the 3-mile visibility had significantly less error than the 10-mile condition. At the 4,500-ft ceiling the only significant difference was between the 2- and 10-mile visibility conditions. In the 4,500-ft and 2-mile visibility conditions CPE was negative, such that pilots estimated the ceiling to be lower than the depicted ceiling; however, at the 4,500-ft ceiling and 10-mile visibility condition pilots estimated ceilings to be higher than the depicted ceiling. There were no other significant differences between visibilities in the 4,500-ft ceiling condition. VPE There was a significant main effect of rating on the VPE data, F(1, 22) = 4.959, p =.037. The mean VPE for the instrument-rated pilots (0.121) was significantly higher than that of the non-instrument-rated pilots ( 0.066). There was a significant main effect of ceiling on the VPE data, F(3, 66) = , p <.001 (Figure 4). Post-hoc analysis revealed that the VPE was significantly lower at the 400-ft ceiling compared to all of the other ceilings. Additionally the VPE at the 900-ft ceiling was significantly lower at the 900-ft ceiling compared to the 2,900-ft and 4,500-ft ceilings. There were no other significant differences between the different ceiling condition pairs on the VPE data.

10 PILOT WEATHER ASSESSMENT 161 FIGURE 4 Main effect of ceiling on the visibility proportional error data. Error bars reflect the standard error of the mean. There was a significant main effect of visibility on the VPE data, F(3, 66) = , p <.001 (Figure 5). Post-hoc analysis revealed that at 10 miles visibility, VPE was significantly lower than all of the other visibilities. There were no other significant VPE differences between any of the different visibility condition pairs. There was a significant interaction of ceiling and visibility on the VPE data, F(9, 198) = , p <.001 (Figure 6). A test of simple main effects was performed on ceiling at each level of visibility. In the 2-mile visibility condition the VPE significantly increased from the 400-ft ceiling to all of the other ceilings, and from the 900-ft ceiling to the 2,900-ft and 4,500-ft ceilings. In the 3-, 5-, and 10-mile visibility conditions the 400-ft and 900-ft ceilings had a significantly lower VPE than the 2,900-ft and 4500-ft ceilings. FIGURE 5 Main effect of visibility on the visibility proportional error data. Error bars reflect the standard error of the mean.

11 162 COYNE, BALDWIN, LATORELLA FIGURE 6 The interaction of ceiling and visibility on the visibility proportional error data. Error bars reflect the standard error of the mean. Note. X axis represents discrete points and is not continuous. DISCUSSION Overall, the results of this study reveal valuable insight into pilot weather judgment. The results provide additional support for the situation assessment hypothesis as an explanation of VFR flight into IMC. The current data revealed that pilots have problems estimating weather conditions. In addition to being fairly inaccurate in their estimation, pilots also had a tendency to overestimate conditions (i.e., estimate that conditions were better than those presented). The instrument-rated pilots did not outperform the non-instrument-rated pilots. Although there were no significant differences in the CPE, there was a significant difference in the visibility error as a function of pilot rating. The results were such that the non-instrument-rated pilots had less error and tended to underestimate conditions, whereas the instrument-rated pilots overestimated conditions. The hypothesis that instrument-rated pilots would estimate ceiling and visibility more accurately than non-instrument-rated pilots was therefore not supported. The hypothesis emerged from the observation that instrument-rated pilots are involved in only 25% of VFR into IMC accidents (AOPA Air Safety Foundation, 1996). Previous research has found that the number of cross-country hours is a predictor of the cues a pilot uses to estimate weather conditions (Wiggins & O Hare, 2003). It was suggested that the extra training required to obtain an instrument rating accounted for the differences between the incidents in which the two groups are involved in VFR into IMC accidents. It is most likely that the lack of differences between the instrument-rated and non-instrument-rated pilots on ceiling estimation and the tendency for non-instrument-rated pilots to underestimate visibility is a result of the matching of the two groups on cross-country hours. That is, these results suggest that it is more likely that any difference between the groups that appears in

12 PILOT WEATHER ASSESSMENT 163 published accident data is the result of flight experience rather than IFR training. However, the data required to verify this suggestion are not available at present. One of the major objectives of the experiment was to provide some empirical evidence of pilots ability to estimate weather conditions. Previous research has suggested that pilots are poor at estimating weather in flight (Goh & Wiegmann, 2001b). However, this evidence was taken from a single IFR weather estimation. The study reported here set out to systematically examine how pilots estimated different weather conditions. The data demonstrate pilots difficulty in estimating weather conditions and reveal that their errors are typically made in the direction of overestimating weather conditions; that is, believing the conditions are better than they actually are. Although it was not anticipated, there was a significant impact of ceiling on visibility error and of visibility on ceiling error. The nature of these effects was such that pilots demonstrated difficulty independently judging the different weather components. The CPE increased as the visibility increased. This suggests that pilots tended to estimate the ceilings as being higher as visibility improved. Although even at the poorest visibility (2 miles) pilots still typically judged the ceiling as being higher than it actually was. The trend for this error to increase as visibility increases is problematic and should be further explored in future research. A similar trend was found with ceiling influencing visibility estimates. When the ceiling was below 1,000 ft pilots estimates of visibility tended to be below the actual visibility, whereas when the ceiling was 2,900 ft and above pilots estimates of visibility were above the actual visibility. That is, when ceilings were low, pilots thought the visibility conditions were poorer than they actually were, but when the ceilings were high pilots estimated visibility to be better than it actually was. It is not clear why pilots estimations of ceiling would be influenced by visibility, or why ceiling would influence visibility estimates. One potential explanation is that pilots use a compensatory decision-making model as suggested by previous research (Driskill et al., 1997; Hunter et al., 2003). For example, the pilots might be using the high visibility to compensate for the low ceilings. The compensatory model evidenced by both Driskill et al. (1997) and Hunter et al. (2003) suggested that when a pilot knew both the ceiling and visibility he or she would use a mathematical combination or average of the two values to determine their overall comfort. However in this study, the exact ceilings and visibility are not known and the pilot must make an estimate of both. It could be that the pilot evaluates one dimension, and this in turn might impact his or her perception of the second dimension. That is, before pilots even have the two pieces of data necessary to make an overall decision they might have already been biased in their estimation. Alternatively pilots might make an initial assessment of the overall weather conditions, and then bias their judgment of the individual components based on the whole. The bias suggested in this experiment does not contradict the compensatory decision

13 164 COYNE, BALDWIN, LATORELLA model suggested by previous research (Driskill et al., 1997; Hunter et al., 2003) but, rather may suggest an additional perceptual element. Limitations and Directions for Future Research Although the use of a two-dimensional display is a limiting factor in this experiment, the inclusion of motion and the use of a projected image aided in pictorial realism. Motion was included at the recommendation of several pilots during a pretest phase. The inclusion of terrain features based on satellite imagery also eliminated problems based on indistinct terrain features. The pilot population selected represents a relatively homogenous group of GA pilots. This homogeneity resulted from matching of instrument-rated and non-instrument-rated pilots on cross-country hours. Pilots were intentionally matched to isolate the impact of instrument training. This typically resulted in the selection of lower hour instrument-rated pilots and higher hour non-instrument-rated pilots. None of the pilots selected for the study had more than 1,000 flight hours. Based on cross-country hour cutoffs used in previous research (Wiggins & O Hare, 2003) all of the pilots were novices. Future research should investigate these phenomena using a wider range of pilot cross-country experience. The overrepresentation of non-instrument-rated pilots in IMC accidents may be a result of IFR pilots having the tendency to have more cross-country experience. In addition to the IFR rating requiring additional training, it would also allow those pilots to fly in a wider range of conditions and therefore have greater opportunity to fly. Unfortunately the previous literature (AOPA Air Safety Foundation, 1996) does not discuss cross-country hours within the different ratings. Separately addressing the issue of ratings and experience would require a large number of participants, and would require a considerable database to match pilots within both groups on experience. Future research should continue to examine the issues of pilot ratings and experience on weather judgment. The interaction of the different weather conditions should be further investigated. It is unclear if pilots make an initial assessment of the overall situation that then biases estimates of the individual components, or if assessment of one condition is biased by another weather condition. From a safety perspective, the influence of high values or good conditions in one weather dimension influencing judgment of the other dimension is a potentially serious problem. Conditions are IFR if the visibility is below a certain minimum, regardless of whether or not the ceilings are high. Summary and Conclusions Previous research cites the difficulty that pilots have estimating weather conditions as a possible cause for their decision to continue into IMC (Goh & Wiegmann,

14 PILOT WEATHER ASSESSMENT b; Wiegmann et al., 2002). The evidence from this experiment demonstrates that pilots do indeed have problems accurately estimating the weather conditions. That the instrument-rated pilots did not outperform the non-instrument-rated pilots can likely be attributed to the approximate equivalence of the two groups on cross-country hours. This suggests that the additional IFR training undertaken by instrument pilots does not aid in their judgment of weather. It is unclear why the non-instrument-rated pilots in this study were more accurate in estimating visibility, or why they had a greater bias toward safer estimates; that is, estimating visibility to be worse than it was. It may be that instrument-rated pilots rely less on their perception of out-the-window conditions, or that non-instrument-rated pilots remember to bias their estimates in the appropriate directions. This investigation identified a new possible variable that influences pilot weather judgment: the interaction of the different elements of weather. Attributing a VFR into IMC accident to simply one of the factors would be a mistake. This investigation raises some new concerns for research in pilot weather judgment. It is important for research to continue to examine how the two dimensions of weather interact with pilot judgment, and to determine the potential for this interaction to push pilots to continue a VFR flight into IMC. ACKNOWLEDGMENTS This work was part of the Aviation Weather Information (AWIN) project at NASA Langley Research Center and was funded under a NASA graduate student fellowship (Grant NGT ). Thanks to the members of the Crew Systems Branch of NASA Langley Research for their assistance in the experimental design, particularly Paul Stough and Jim Chamberlain. Thanks to Regina Johns of Lockheed- Martin for her work in subject procurement. REFERENCES AOPA Air Safety Foundation. (1996). Safety review: General aviation weather accidents. Frederick, MD: Author. AOPA Air Safety Foundation. (2002) Nall report: Accident trends and factors for Frederick, MD: Author. Coyne, J. T., Baldwin, C. L., & Latorella, K. A. (2005). Influence of graphical METARs on pilot s weather judgment. In Proceedings of the Human Factors and Ergonomics Society 49th annual meeting (pp ). Santa Monica, CA: Human Factors and Ergonomics Society. Driskill, W. E., Weismuller, J. J., Quebe, J., Hand, D. K., Dittmar, M. J., & Hunter, D. R. (1997). The use of weather information in aeronautical decision making (Tech. Rep. No. NTIS DOT/FAA/AM- 97/3). Washington, DC: Federal Aviation Administration.

15 166 COYNE, BALDWIN, LATORELLA Goh, J., & Wiegmann, D. A. (2001). Visual flight rules flight into instrument meteorological conditions: An empirical investigation of the possible causes. International Journal of Aviation Psychology, 11, Goh, J., & Wiegmann, D. A. (2002a). Human factors analysis of accidents involving visual flight rules flight into adverse weather. Aviation Space & Environmental Medicine, 73, Goh, J., & Wiegmann, D. A. (2002b). Relating flight experience and pilot s perceptions of decision-making skill. In Proceedings of the Human Factors and Ergonomics Society 46th annual meeting (pp ). Santa Monica, CA: Human Factors and Ergonomics Society. Hunter, D. R., Martinussen, M., & Wiggins, M. W. (2003). Understanding how pilots make weather-related decisions. International Journal of Aviation Psychology, 13, O Hare, D., Owen, D., & Wiegmann, D. A. (2001). The where and why of cross-country VFR crashes: Database and simulation analysis. In Proceedings of the 45th annual meeting of the Human Factors and Ergonomics Society (pp ). Santa Monica, CA: Human Factors and Ergonomics Society. O Hare, D., & Smitheram, T. (1995). Pressing on into deteriorating conditions: An application of behavioral decision theory to pilot decision making. International Journal of Aviation Psychology, 5, Wiegmann, D. A., Goh, J., & O Hare, D. (2002). The role of situation assessment and flight experience in pilots decisions to continue visual flight rules flight into adverse weather. Human Factors, 44, Wiggins, M. W., & O Hare, D. (2003). Expert and novice pilot perceptions of static in-flight images of weather. International Journal of Aviation Psychology, 13, Manuscript first received: January 2007