Emil Juni

Regression Tree Models to Predict Winter Storm Costs

Winter maintenance can consume one-third or more of highway maintenance budgets. Tools for estimating winter maintenance costs can enhance allocation, accountability, and management of expenditures. Historical weather forecasts and associated maintenance resources are used to create statistical models to estimate county-level resources to fight a forecast snow or freezing rain event. County-level analysis allows for model refinement for slightly different business practices and areas small enough to assume uniform weather effects. The statistical models are organized as regression trees that accommodate variables for operation of winter maintenance, such as service level expectations, range of county size, and weekend and overtime events. The regression trees fit subsets of data to form families of multiple linear models. In this way, models can be refined for important categorical variables such as service level and county size. The models presented here estimate labor, equipment, and material resources required to fight a storm in counties having 87 to 1,460 lane mi to maintain. The models estimate resources, not cost. Accordingly, the models are independent of unit costs of labor, equipment, and material that change over time and vary from county to county. Unit costs for labor, material, and equipment at each county are needed to convert resource estimates to resource costs. Statewide or regional storm costs can be computed by summing county-level costs.

Relating Cost to Condition in Routine Highway Maintenance

When transportation agencies prepare a design for new highway construction or major improvements to existing highways, the life-cycle, agency, and user costs are considered in project design decisions. However, after highway projects are completed, maintenance budgets often do not keep pace with maintenance needs since they are rarely adjusted to accommodate the routine maintenance of new lane miles. Instead, maintenance budgets per lane mile remain relatively constant, regardless of increase in the number of vehicle miles traveled per highway mile. Thus, disparity increases between maintenance budgets and maintenance requirements, leading to difficult choices for maintenance priorities. Concerns about safety and mobility tend to trump preservation of capital investment. A study is presented of the relationship between maintenance cost and condition. With use of regression tree analysis, the study identified physical, environmental, operational, and socioeconomic parameters that influence maintenance costs for asphalt and concrete pavements, shoulders, litter pickup, vegetation control, and ditches. As a result, valid model equations will show just what kind of effect a certain investment has on the level of service. Other models will show how much investment is needed to get to a certain level of service. These models will provide information to maintenance managers about the trade-offs they will get based on the relationship of cost and maintenance, so they can make the best decision in allocating available funds.

Integrated Field and Office Tools for Bridge Management

The Wisconsin Department of Transportations (WisDOT) Bureau of Structures Data Consolidation Project brings together all bridge-related data into a single data warehouse, the Highway Structures Information System (HSIS), which includes inventory, inspection and maintenance records, and maintenance costs. With the new HSIS, WisDOTs interactive bridge inspection forms and data procedures became incompatible. This conflict presented a timely opportunity to investigate alternative data collection tools that would be practicable in the field. This paper investigates the use of alternative field data collection technologies for bridge inspectors. The participation of bridge inspectors in identifying functional requirements led to an assembly of field tools that included the tablet PC with a shoulder carrying case, a noise-canceling headset microphone, and a tethered pen. Bridge inspectors from six districts volunteered to train on the handwriting and speech recognition technologies and then to test the tools ...

Part 4: Winter Maintenance: Regression Tree Models to Predict Winter Storm Costs

Winter maintenance can consume one-third or more of highway maintenance budgets. Tools for estimating winter maintenance costs can enhance allocation, accountability, and management of expenditures. Historical weather forecasts and associated maintenance resources are used to create statistical models to estimate county-level resources to fight a forecast snow or freezing rain event. County-level analysis allows for model refinement for slightly different business practices and areas small enough to assume uniform weather effects. The statistical models are organized as regression trees that accommodate variables for operation of winter maintenance, such as service level expectations, range of county size, and weekend and overtime events. The regression trees fit subsets of data to form families of multiple linear models. In this way, models can be refined for important categorical variables such as service level and county size. The models presented here estimate labor, equipment, and material resources ...

Development of a Carrier Selection Method for Intrastate Compliance Review in Wisconsin

A method was developed for ranking intrastate carriers in Wisconsin to enhance the selection of carriers for Wisconsin’s intrastate compliance (comprehensive safety) review program. The federal SafeStat methodology for identifying interstate motor carriers that have poor safety records cannot be used in Wisconsin because of the lack of carrier size data. Missouri’s intrastate SafeStat methodology, which does not require carrier size data, was used as the starting point in developing Wisconsin’s methodology. The methodology involves merging a ranking of carriers based on historical inspection data (the total out-of-service rate) with a ranking based on total crashes. The validation based on crashes per carrier for a subsequent period showed a similar pattern for the before and after crash rates by percentile group. The overall level of crashes for the after period, however, was substantially lower because of the statistical phenomenon of regression to the mean (RTM). The methodology is also limited by the lack of exposure data for the carrier crash data. To account for RTM, an empirical Bayesian method for estimating RTM was proposed and applied to a before-and-after analysis of intrastate carrier crashes in Wisconsin. The method provided estimates of the after-period crash rates that were much better than the naive assumption of no change in the crash rate. An improved fit for the RTM estimates is likely if more homogeneous groups of carriers were created on the basis of grouping by carrier size.

Development of Intrastate Inspection Selection System for Wisconsin

Roadside safety inspections have been an important component of federally supported programs to improve motor carrier safety. The inspection selection system (ISS) and its successor, ISS-2, were designed to help roadside inspectors target the vehicles of Interstate carriers with the worst past safety performance. ISS was used as a starting point for developing an intrastate ISS for Wisconsin. Logistics regression was used to identify the relevant independent variables. The logit model estimates the probability that a vehicle will be found to have an out-of-service (OOS) safety violation. An alternative to the logit model, a simple direct-estimation technique that uses the total OOS rate for a motor carrier, was found to produce the highest success rates in predicting inspections that result in an OOS violation. The model can be implemented as a field inspection tool and as a means to generate a list of intrastate carriers ranked by their historical OOS violation record.