A Review of the Florida Keys Carrying Capacity Study (2002)

The Human Infrastructure Module focuses on two major connected issues: traffic on U.S. Route 1 and hurricane evacuation. The traffic component of the module uses level of service (LOS) as an indicator based on the speed of traffic moving along U.S. Route 1. Hurricane evacuation clearance times are independently estimated and are based solely on population using linear extrapolation of estimates from a previous study (Miller Consulting, 2001). Comments on each component are presented below and detailed comments are reported in Appendix D of this review.

Given that LOS for U.S. Route 1 has a mandated role in Florida’s concurrency requirements (Florida Administrative Code §9J-5.0055[2][a][1]), the indicator is an important indicator to be included in the model. Though the LOS component appears to need only minor documentation revisions, it is not possible to fully validate the methods used to calculate it given its current presentation. The use of non-local data and 10-year-old coefficients warrants some sensitivity testing.

Land-use-specific trip generation rates are not available for Monroe County, therefore the consultants used national data from the Institute of Transportation Engineers trip generation manual (Institute of Transportation Engineers, 1998).

The committee therefore recommends that an “Assumptions and Uncertainty” subsection be added to Section 3.4.1 of the Final CCAM report. That subsection should provide a statement that characterizes the probable margin of error associated with using non-local coefficients. In addition, sensitivity analysis for LOS estimates should be conducted using an appropriate range of variation for the national trip generation values.

The percent of trips generated from outside the Florida Keys is based on the Monroe County Long-Range Transportation Plan (Draft CCAM Section 3.4.1). It is not clear how or when that coefficient was calculated, or what range of uncertainty should be associated with it in projecting LOS for different future scenarios. It also is not clear from the Draft CCAM report nor from the consultant’s response to the NRC Committee’s Question 3 (Appendix D) how this factor is used to estimate traffic volumes and calculate LOS within planning units. The “Assumptions and Uncertainty” subsection provides a statement that acknowledges a probable margin of error associated with using this dated coefficient for the percent of trips originating outside the Florida Keys. In addition, the committee recommends including a description of how this variable is used in estimating traffic volumes and calculating levels of service within planning units. As with the trip generation rates discussed above, it is important to conduct sensitivity analysis for LOS estimates using an appropriate range of variation for this coefficient.

It is unclear how trip lengths are estimated for calculating internal-internal, internal-external, external-internal, and external-external trips for planning units. Average trip length estimates (Draft CCAM Table 3.4) are generated from 1992 data for which no source is given. It is not clear if the average of the six sites in Table 3.4 is used for all planning units, or if trip lengths for some planning units are based on one of the six sites. The Committee recommends that the contractors provide a description in the text of how trip lengths are estimated for individual planning units and are partitioned among internal-internal, internal-external, external-internal, and external-external trips in different planning units. The “Assumptions and Uncertainty” subsection recommended above, should provide a statement that trip lengths are based on 1992 data and a characterization of the uncertainty resulting from the 10-year difference in land use and socioeconomic conditions that presumably determine trip lengths. As recommended above, a sensitivity analysis should be conducted for LOS estimates using an appropriate range of variation for the trip length values.

Florida’s local planning requirements (Florida Administrative Code §9J-.12[3][b][7]; Florida Administrative Code §9J-.12[3][c][3]) stipulate that local plans must include objectives and policies for maintaining or reducing hurricane evacuation clearance times, making it important to include that indicator in the mode. The Draft CCAM, however, exhibits a paucity of information pertaining to this important issue of personal safety and public policy. The evacuation clearance time is presented as accurate based on linear extrapolation of estimates from a separate study called the Florida Keys Hurricane Evacuation Study, Final Report (Miller Consulting, Inc., 2001). No documentation is provided for the methods, assumptions, limitations, or constraints of the primary model upon which the estimates are based or of the margins of error that result from the simplistic extrapolation used to predict the effects of future development on evacuation clearance times. This topic begs for more background information; better documentation of how important variables are treated such as evacuation behavior, road elevations, and storm surge flooding; and a sensitivity analysis centered on major assumptions and uncertainties.

The Draft CCAM estimates evacuation clearance times by assuming a linear relationship between total population within the Florida Keys and the aggregate clearance time for the islands (Draft CCAM Section 3.4). These estimates do not account for the impact of growth on clearance times within the seven individual evacuation analysis zones used in the Miller Study (Miller Consulting Inc., 2001), or on the effects of that growth on the five bottlenecks along U.S. Route 1 identified by Miller Consulting, Inc. As a result, the projected evacuation clearance times are invalid estimates of aggregate clearance times for the Florida Keys. Furthermore, the Draft CCAM does not identify the population base upon which this linear extrapolation is based, nor does it explain the basis for the three “threshold” population estimates reported in Equations #95–97 of the Hurricane Evacuation section in Appendix C of the Draft CCAM report.

In addition, the Draft CCAM presents none of the Miller (2001) study information required to interpret the evacuation clearance time estimates. In the absence of such information, the estimates seem to be based entirely on the capacity of the single road connecting the Florida Keys. This assumption fails to acknowledge that bottlenecks at intersections and along segments of U.S. Route 1 constitute critical factors that determine evacuation clearance times. The draft report also fails to address the assumptions and uncertainties implicit in the use of only two hurricane intensity scenarios as to (Category 1-2 and Category 3) to compare evacuation times. As a result, it is unclear which bottlenecks are the critical determinants of aggregate evacuation clearance time under different storm scenarios. Furthermore, the Draft CCAM is silent about where U.S. Route 1 is most vulnerable to flooding from hurricanes of different intensities and the

associated probabilities of evacuation being terminated under different storm scenarios.⁵

The margin of error resulting from the linear extrapolation used in the Draft CCAM cannot be estimated without testing the Test Carrying Capacity Model (CCAM)(C. Miller, pers. com.). The Committee recommended, therefore, that this element of the CCAM be deleted or that the Miller (2001) model be incorporated as a sub-model in the Final CCAM. The Miller model is a spreadsheet that calculates clearance times based on estimates of the numbers of dwelling units in each analysis zone as differentiated into three categories: (1) mobile home dwelling units, (2) other residential dwelling units, and (3) tourist residential units (Miller Consulting Inc., 2001). It seems straightforward to produce such inputs from the CCAM Scenario Generator, which makes it possible to produce evacuation clearance time estimates that can be properly compared to those produced by previous evacuation clearance studies conducted for Monroe County.

If the Miller model is incorporated into the CCAM, the Graphical User Interface (GUI) should be redesigned to allow the user to specify important starting conditions for running the model, including hurricane category and evacuation response assumptions (Miller Consulting, Inc., 2001). The inputs, outputs, and mechanics of the Miller model should be summarized in the CCAM, including documentation of how bottlenecks, low-lying areas, and alternative storm scenarios are accounted for. All assumptions embedded in the Miller model should be explicitly described in an “Assumptions and Uncertainty” subsection.