Modeling Spatial Patterns of Recreational Boaters: Vessel, Behavioral, and Geographic Considerations

Abstract

A regression-based framework for modeling recreational boating patterns and estimating preferred on-water destinations was introduced. A survey of 500 boaters provided model input regarding vessel, behavioral, and geographic characteristics. This information was used to construct a travel network within a Geographic Information System to identify major network intersections (pivots) and to calculate the distance traveled along network segments between intersections, as derived model input. Model estimates of preferred destinations and use intensity were compared to mail survey results for validation. The average error between reported and estimated boating destinations was 4.3 miles for a regional application and 3.0 miles for a sub-regional application. In addition to vessel and behavioral considerations, the results highlighted the significance of geographic and network variables in modeling the spatial patterns of recreational boaters and destination choice.

Keywords:

• recreational boating
• boating patterns
• geographic information systems

Notes

¹The boat owner's home was not necessarily the departure site. Roughly 40% of the boaters surveyed either launched from a boat ramp or departed from a marina.

^*Variables, coefficients, and test statistics are displayed in the order in which they were introduced into the model by the forward stepwise procedure.

Summary statistics R² = .855 adjusted R² = .850 std. error of estimate = 23.001 F = 182.242 (sig. = .000) n = 384 observations (accounting for 14 outliers that were removed prior to re-estimating the model) k = 12 explanatory variables entered (zero removed) Shapiro Wilks normality test (fail to reject normality at 95% confidence level)

^*Variables, coefficients, and test statistics are displayed in the order in which they were introduced into the model by the forward stepwise procedure.

Summary statistics R² = .643 adjusted R² = .629 std. error of estimate = 18.104 F = 46.742 (sig. = .000) n = 379 observations (accounting for 19 outliers that were removed prior to re-estimating the model) k = 14 explanatory variables entered (2 removed: npx, p ² y) Shapiro Wilks normality test (fail to reject normality at 95% confidence level)

^*Refer to Figure 1 for place names.

^**Refer to Figure 3 for SOR descriptions.