How the choice of safety performance function affects the identification of important crash prediction variables.

Across the nation, researchers and transportation engineers are developing safety performance functions (SPFs) to predict crash rates and develop crash modification factors to improve traffic safety at roadway segments and intersections. Generalized linear models (GLMs), such as Poisson or negative binomial regression, are most commonly used to develop SPFs with annual average daily traffic as the primary roadway characteristic to predict crashes. However, while more complex to interpret, data mining models such as boosted regression trees have improved upon GLMs crash prediction performance due to their ability to handle more data characteristics, accommodate non-linearities, and include interaction effects between the characteristics. An intersection data inventory of 36 safety relevant parameters for three- and four-legged non-signalized intersections along state routes in Alabama was used to study the importance of intersection characteristics on crash rate and the interaction effects between key characteristics. Four different SPFs were investigated and compared: Poisson regression, negative binomial regression, regularized generalized linear model, and boosted regression trees. The models did not agree on which intersection characteristics were most related to the crash rate. The boosted regression tree model significantly outperformed the other models and identified several intersection characteristics as having strong interaction effects.

Intelligent geocoding system to locate traffic crashes.

State agencies continue to face many challenges associated with new federal crash safety and highway performance monitoring requirements that use data from multiple and disparate systems across different platforms and locations. On a national level, the federal government has a long-term vision for State Departments of Transportation (DOTs) to report state route and off-state route crash data in a single network. In general, crashes occurring on state-owned or state maintained highways are a priority at the Federal and State level; therefore, state-route crashes are being geocoded by state DOTs. On the other hand, crashes occurring on off-state highway system do not always get geocoded due to limited resources and techniques. Creating and maintaining a statewide crash geographic information systems (GIS) map with state route and non-state route crashes is a complicated and expensive task. This study introduces an automatic crash mapping process, Crash-Mapping Automation Tool (C-MAT), where an algorithm translates location information from a police report crash record to a geospatial map and creates a pinpoint map for all crashes. The algorithm has approximate 83 percent mapping rate. An important application of this work is the ability to associate the mapped crash records to underlying business data, such as roadway inventory and traffic volumes. The integrated crash map is the foundation for effective and efficient crash analyzes to prevent highway crashes.

Directed site exploration for permeable reactive barrier design.

Permeable reactive barriers (PRBs) are being employed for in situ site remediation of groundwater that is typically flowing under natural gradients. Site characterization is of critical importance to the success of a PRB. A design-specific site exploration approach called quantitatively directed exploration (QDE) is presented. The QDE approach employs three spatially related matrices: (1) covariance of input parameters, (2) sensitivity of model outputs, and (3) covariance of model outputs to identify the most important location to explore based on a specific design. Sampling at the location that most reduces overall site uncertainty produces a higher probability of success of a particular design. The QDE approach is demonstrated on the Kansas City Plant, Kansas City, MO, a case study where a PRB was installed and failed. It is shown that additional quantitatively directed site exploration during the design phase could have prevented the remedial failure that was caused by missing a geologic body having high hydraulic conductivity at the south end of the barrier. The most contributing input parameter approach using head uncertainty clearly indicated where the next sampling should be made toward the high hydraulic conductivity zone. This case study demonstrates the need to include the specific design as well as site characterization uncertainty when choosing the sampling locations.

MODFLOW 2000 head uncertainty, a first-order second moment method.

A computationally efficient method to estimate the variance and covariance in piezometric head results computed through MODFLOW 2000 using a first-order second moment (FOSM) approach is presented. This methodology employs a first-order Taylor series expansion to combine model sensitivity with uncertainty in geologic data. MODFLOW 2000 is used to calculate both the ground water head and the sensitivity of head to changes in input data. From a limited number of samples, geologic data are extrapolated and their associated uncertainties are computed through a conditional probability calculation. Combining the spatially related sensitivity and input uncertainty produces the variance-covariance matrix, the diagonal of which is used to yield the standard deviation in MODFLOW 2000 head. The variance in piezometric head can be used for calibrating the model, estimating confidence intervals, directing exploration, and evaluating the reliability of a design. A case study illustrates the approach, where aquifer transmissivity is the spatially related uncertain geologic input data. The FOSM methodology is shown to be applicable for calculating output uncertainty for (1) spatially related input and output data, and (2) multiple input parameters (transmissivity and recharge).