Identifying critical factors that control the off-site transport of pyrethoids in the urban environment is critical to the safe and effective use of pyrethroids in the control of insects for home and business owners. This work uses a data mining approach to extract critical event variables from an urban study site that had been operational for a year (August, 2011 - August, 2012). Six applications occurred for four surfaces (driveway, garage door, grass perimeter, and house wall) and one application to the grass lawn following historic and revised practices. A Multivariate Adaptive Regression Spline (MARS) modeling approach was used to statistically model the percent of pyrethroid mass applied (percent washoff) from all surfaces. This approach yielded accurate models for all surfaces, with the driveway surface having the simplest model of percent washoff. The MARS modeling approach allows very dynamic changes in variables to represent complex behavior at the sites-integrating many variables to calculate percent washoff. For all surfaces, a near-post application period (around 14 days for all surfaces except the grass lawn, which had an extended multiple month period post application) controlled washoff particularly during low intensity lawn sprinkler events. During natural and simulated rainfall events, the dynamics of washoff included multiple types of characterizing runoff factors (from 10, 20, 30, and 60 min maximum runoff rates), the rainfall amounts, days since the previous application of a pyretheroid, among other factors. In addition, a number of other often minor factors were included by the MARS models for each surface for the calculation of percent washoff that warrant further investigation.