Enhancing Internal Trip Capture Estimation for Mixed-Use Developments (2011)

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

F-1 Estimation Procedure The estimation procedure was applied to seven different developments for which at least land use information, peak hour cordon counts, and proximity information were avail- able. Four of these developments provided data for this study; the other three did not. The validation test was to see how well the estimation procedure could begin with ITE trip generation data and reproduce the external vehicular cordon volumes. Five of the developments had directional cordon traffic vol- umes available for both peaks. These developments included: • Mockingbird Station, • Legacy Town Center, • Atlantic Station, • Crocker Center (independent site, Boca Raton, Florida), and • Mizner Center (independent site, Boca Raton, Florida). Two developments had on non-directional P.M. peak period counts available. They were: • Boca del Mar and • Southern Village (independent site, Chapel Hill, North Carolina). The validation test compared four different estimation methods to determine which method produced the results closest to the cordon counts: • the estimator described in this report, • the estimator, but without the proximity adjustment, • the existing ITE estimation method, and • unadjusted ITE trip generation. Development data and approximations of surveyed mode split and vehicle occupancies were input to the estimation pro- cedure. Table F-1 shows the ITE land use codes used to esti- mate single-use vehicle trip generation for component land uses of the seven developments. Where businesses were closed during a peak period and there were no observed trips to or from the business (e.g., cinema during A.M. peak), no trips were included in the validation estimate. Southern Village had additional land uses (a school and park-and-ride lot) that were not included in the internal cap- ture estimate; those were handled as additional land uses. Information provided in the source document was used as the basis for the trip generation estimate. Table F-2 shows the results numerically. Figures F-1 through F-4 graphically com- pare the results for the five developments for which complete data were available. Error comparisons were also made and are shown in Table F-3. Table F-2 rows contain data as follows: 1. Counted at cordon: vehicles(persons) counted using site driveways; 2. Estimator output: directional volume of vehicles (persons) estimated with recommended estimation method: – First four columns: volumes as described, – Last four columns: percent internal trips; 3. From survey – directional volume of vehicles (persons) derived from survey: – First four columns: volumes as described, – Last four columns: percent internal trips; 4. Estimator/counted: ratio of estimated trips divided by counted trips in respective columns; and 5. Unadjusted/counted: estimate using raw ITE trip genera- tion divided by counted trips Table entries for Southern Village contain additional rows to account for land uses that do not qualify for internal cap- ture under the recommended procedure. Table F-2 shows comparison of external vehicle and person trips estimated by each method. Also shown are estimated internal capture percentages. The most important results are Validation of Estimation Procedure A P P E N D I X F

the external trip estimates. Figures F-1 through F-4 show the comparisons of vehicle trips for both A.M. and P.M. peak periods and both inbound and outbound directions. In Figure F-1, it is evident that for the A.M. peak hour inbound vehicle trips, the NCHRP estimation methods—both with and without the proximity adjustment—produce the best results for three of the five developments; the current ITE method is closest for one site and slightly better than the NCHRP method for another site. Atlantic Station is more closely estimated by both unadjusted trip generation and the current ITE method. The current ITE method is better than raw trip generation, but the method developed in this project is even closer to the counts. Figure F-2 shows similar results for A.M. peak hour out- bound vehicle trips with the recommended estimator (both with and without the proximity adjustment) producing the best results for four of the five developments. This time Mizner Center is better estimated by raw trip generation and the cur- rent ITE method. As with the previous comparison, the ITE method is an improvement on raw trip generation. The P.M. inbound comparison shown in Figure F-3 shows that the NCHRP method with proximity adjustment pro- duces the closest estimates for two sites, with the methods with and without proximity about equal for the two sites, and the raw ITE trip generation closest for one site. Again, Mizner Center was better estimated by another method (this time raw trip generation), but the other four are best estimated by the recommended method. Figure F-4 shows the comparison for P.M. peak hour in- bound trips. As for the other time periods and directions, one or the other of the NCHRP methods produces the closest esti- mates in four of the five cases. The methods with and without proximity adjustments are each best for one MXD while both yield approximately the same results for two MXDs. In this case, Boca Center is better estimated using the existing ITE method. In total, the recommended method—with or without the proximity adjustment—produces more reliable estimates for four of the five developments. The results for the other two developments—Boca del Mar and Southern Village—show two different patterns. For Boca del Mar, both the existing ITE and recommended methods produce significantly low estimates, but are closer than the rec- ommended method without proximity adjustments or the ITE method. The raw estimate is above the actual external count, but it and the ITE method are the closest of the estimates (about 4 percent closer than the recommended method with proximity adjustment). For Southern Village, the results are very different. The recommended method (both with and without proximity adjustments) produce estimates very close to the counts. Table F-3 may quantify the degree of accuracy or error more clearly, recognizing that the statistics presented represent the sum of combined results. The average error shown is the sim- ple sum of the percent deviations from the counts as derived in Table F-2. On average, as a group the estimates all exceed the counts (for example, the recommended method with proxim- ity adjustment is an average of 4 percent). This is very mislead- ing and not relevant for single developments because overesti- mates and underestimates tend to cancel each other out. What may be of value in those percentages is that they could result in the sum total trip generation of several developments in an area. However, that is not what is being validated here. More applicable is the absolute average error, which is the sum of the magnitudes of the errors averaged over the five F-2 Land Uses NCHRP Project 8-51 Classification Subgroup ITE Land Use Code Office - 710 Retail - 820 Quality sit down 931 High turnover 932 Fast food, no drive-through 933 Restaurant Fast food with drive-through 934 Cinema - 444 Hotel - 310 Single family detached 210 Apartments 220 Residential Townhomes 230 Additional Land Use Port and terminal Park-and-ride lot 090 Table F-1. ITE land use codes used in validation.

F-3 Vehicle Trip (Person Trips) Percent Internal Trips (Peak Period) A.M. Peak Hour P.M. Peak Hour A.M. P.M. Development/data In Out In Out In Out In Out Mockingbird Station Counted at cordon 272(385) 128(213) 367(595) 353(586) Estimator output 259(329) 107(165) 422(565) 412(588) 19% 32% 33% 33% From survey 35% 46% 36% 42% Estimator/counted 0.95(0.85) 0.84(0.77) 1.15(0.95) 1.17(1.00) Without proximity adjustment Estimator output Same Same 422(563) 411(586) Same Same 33% 33% Estimator/counted Same Same 1.15(0.95) 1.16(1.00) With ITE Trip Gen Handbook data Estimator output 322(409) 156(242) 537(715) 523(745) No data No data 15% 15% Estimator/counted 1.18(1.06) 1.22(1.14) 1.46(1.20) 1.48(1.27) Unadjusted ITE Trip Generation report Estimator output 399 233 798 832 0% 0% 0% 0% Unadjusted/counted 1.47 1.82 2.17 2.36 Atlantic Station With proximity adjustment Counted at cordon 962(1012) 455(502) 1023(1396) 1038(1260) Estimator output 796(843) 252(308) 962(1126) 1151(1342) 17% 37% 36% 34% From survey 40% 30% 41% 42% Estimator/counted 0.83(0.83) 0.55(0.61) 0.94(0.81) 1.10(1.07) Without proximity adjustment Estimator output Same Same 938(1097) 1124(1310) Same Same 38% 36% Estimator/counted Same Same 0.91(0.79) 1.08(1.04) With ITE Trip Gen Handbook data Estimator output 952(1130) 398(484) 1232(1445) 1604(1750) No data No data 16% 13% Estimator/counted 0.99(1.11) 0.87(0.96) 1.29(1.04) 1.55(1.39) Unadjusted ITE Trip Generation report Estimator output 1122 473 1690 1992 0% 0% 0% 0% Unadjusted/counted 1.17 1.03 1.65 1.92 Legacy Town Center Counted at cordon 734(819) 641(779) 933(1187) 955(1122) Estimator output 736(906) 690(850) 1003(1236) 912(1123) 15% 16% 34% 36% From survey 32% 25% 48% 44% Estimator/counted 1.00(1.11) 1.08(1.09) 0.95(1.04) 0.95(1.00) Without proximity adjustment Estimator output Same Same 923(1136) 831(1023) Same Same 39% 42% Estimator/counted Same Same 0.98(0.96) 0.87(0.91) With ITE Trip Gen Handbook data Estimator output 864(1065) 821(1009) 1231(1516) 1413(1740) No data No data 27% 24% Estimator/counted 1.18(1.30) 1.28(1.30) 1.32(1.28) 1.48(1.55) Unadjusted ITE Trip Generation report 909 862 1598 1502 0% 0% 0% 0% Unadjusted/counted 1.24 1.34 1.71 1.57 Boca (ex-Crocker) Center Counted at cordon 488 219 281 532 Estimator output 525 189 342 461 13% 26% 32% 31% From survey No data No data No data No data Estimator/counted 1.08 0.86 1.22 0.87 Without proximity adjustment Estimator output Same Same 342 461 Same Same 32% 31% Estimator/counted Same Same 1.22 0.87 With ITE Trip Gen Handbook data Estimator output 617 271 385 502 No data No data 26% 33% Estimator/counted 1.26 1.24 1.37 0.94 Unadjusted ITE Trip Generation report 655 295 566 678 0% 0% 0% 0% Unadjusted/counted 1.34 1.35 2.01 1.27 Table F-2. Summary of estimator validation comparisons. (continued on next page)

F-4 Vehicle Trip (Person Trips) Percent Internal Trips (Peak Period) A.M. Peak Hour P.M. Peak Hour A.M. P.M. Development/data In Out In Out In Out In Out Mizner Center Counted at cordon 220 145 547 328 Estimator output 239 99 417 388 13% 25% 29% 35% From survey No data No data No data No data Estimator/counted 1.09 0.68 0.76 1.18 Without proximity adjustment Estimator output Same Same 412 383 Same Same 30% 35% Estimator/counted Same Same 0.75 1.17 With ITE Trip Gen Handbook data Estimator output 267 134 425 402 No data No data 27% 32% Estimator/counted 1.21 0.99 0.78 1.23 Unadjusted ITE Trip Generation report 272 137 613 585 0% 0% 0% 0% Unadjusted/counted 1.24 0.94 1.12 1.78 Boca del Mar With proximity adjustment Counted at cordon - - 2187 2-way Estimator output - - 915 895 - - 26% 28% From survey No data No data 7% 8% Estimator/counted - - 0.83 2-way Without proximity adjustment Estimator output - - 689 676 - - 44% 47% Estimator/counted - - 0.62 2-way With ITE Trip Gen Handbook data Estimator output - - 839 831 - - 33% 35% Estimator/counted - - 0.76 2-way Unadjusted ITE Trip Generation report - - 1241 1209 - - 0% 0% Unadjusted/counted - - 1.12 2-way Southern Village Counted at cordon - - 1336 2-way Estimator output - - 546 438 Additional trips for non MXD uses - - 97 290 Total estimated - - 645 731 - - 11% 13% From survey No data No data No data No data Estimator/counted - - 1.03 2-way Without proximity adjustment Estimator output - - 537 429 No data No data N/Aa N/Aa Additional trips for non MXD uses - - 97 290 Total estimated - - 637 722 Estimator/counted 1.01 2-way With ITE Trip Gen Handbook data Estimator output 574 466 - - 6% 8% Additional trips for non MXD uses - - 97 290 Total estimated - - 671 756 Estimator/counted 0.99 2-way Unadjusted ITE Trip Generation report 633 512 - - 0% 0% Additional trips for non MXD uses - - 97 290 Total estimated - - 730 802 Unadjusted/counted 1.15 2-way a Person trips not known for non-MXD uses Table F-2. (Continued).

F-5 Figure F-1. Comparison of estimates to cordon counts: A.M. peak-hour inbound direction. Figure F-2. Comparison of estimates to cordon counts: A.M. peak-hour outbound direction. developments. This shows more clearly what deviations— above or below actual—were found. Clearly, by examining the figures and Table F-3, it is easy to determine that the raw trip generation greatly overestimates external vehicle trip generation for the validation sites. The existing ITE method is a major improvement from raw trip generation. The rec- ommended method brings the estimates significantly closer to actual. Note that the difference between the actual and absolute value of the errors shows that there are both over- estimates and underestimates occurring. The standard deviation shown in Table F-3 better repre- sents the estimated probable magnitude of error that might

F-6 Figure F-3. Comparison of estimates to cordon counts: P.M. peak-hour inbound direction. Figure F-4. Comparison of estimates to cordon count: P.M. peak-hour outbound direction. occur using these estimation methods. Again, the relative magnitudes of error among the methods place them consis- tently in the same order. It is clear that the recommended method provides more accurate estimates. Since the existing ITE method was devel- oped from data from three of the six developments used in this NCHRP project, the recommended method can only be viewed as being a further improvement. The standard deviations for the recommended method, both with and without proximity adjustment, are about 20 percent of the actual external inbound and outbound volumes. This is less than the variations in the raw ITE nondirectional trip gen- eration rates for the component land uses. For example, for the land uses listed in Table F-1, the standard deviations for their A.M. and P.M. peak hour trip generation rates are all in excess of 50 percent of the mean.

Not clear, however, is whether or not the proximity adjust- ment adds any current value. The validation results show no significant statistical benefit. It has sufficient data only for the P.M. peak period (and less of that than would be desired). There is no A.M. proximity adjustment recommended at this time. On the other hand, the only examples for which the results were better without the proximity adjustment was when both variations of the new method were overestimat- ing. In all cases the proximity adjustment either has no sig- nificant effect or renders the estimate more conservative (higher). Conclusions The validation supports two principal findings: 1. The recommended method does produce noticeably more accurate results than either raw ITE trip generation esti- mates from the ITE Trip Generation report or the existing method described in the Trip Generation Handbook. This is true with or without the proximity adjustment. 2. The proximity adjustment, available at this time for the P.M. peak period, tends to make slightly more conservative estimates but overall does not, at this time, improve accu- racy over a group of estimates. It can produce significant effects for larger developments. It would be logical for ITE to consider the recommended method for inclusion in the next edition of its Trip Generation Handbook. The researchers recommend this since it could increase trip generation estimation accuracy. The advisory committee that ITE uses to review potential new material may wish to test further both the existing method and the recom- mended method with more MXDs for which it can obtain the needed data. This could help to determine if the proximity adjustment shows enough added value in its current form to be included in the next edition. In addition, the research team confirmed the desirability and need for more surveys to expand the database. Six sam- ples are far better than three. Addition of several more could possibly provide the basis for confirming the value of the proximity adjustment. F-7 Recommended NCHRP Method Error Type Raw ITE Trip Generation Existing ITE Method With Proximity Adjustment No Proximity Adjustment Explanation Average error +55% +26% -4% 7% Average error for sum of all sites Absolute average error 55 28 17 17 Average magnitude of error per site Standard deviation 68 34 20 19 Expect two-thirds of site estimates within this error range Table F-3. Comparison of error statistics.