Streamlining the World Bank's Highway Design and Maintenance Standards model (HDM-III) for network-level application.

摘要

The World Bank's Highway Design and Maintenance Standards model (HDM-III) is increasingly becoming a popular economic analysis "engine" with wide applications within pavement management systems, in particular, the network-level priority programming sector in developing countries. The major disincentives, however, for widespread application of the model in low-income road agencies include the excessive data needs and lack of effective guidelines on local adaptation.; The thesis addresses streamlining the application of the HDM-III model at the network-level priority programming by reducing the model data needs. This was done by screening out the insensitive input variables with respect to an application-specific output.; An advanced statistical experimental design based on the Latin hypercube sampling was formulated to investigate the sensitivity of the link characterization input factors upon several HDM-III outputs--the net present value of net benefits (NPV), the agency life-cycle costs, and the road users' (mainly vehicle operating costs, VOCs) life-cycle costs.; The boundaries of the input space (factor ranges) investigated in this study were determined from field data collected in Tanzania in 1994.; The plausibility of extending the current pool of "default inputs" for task-specific applications was examined. The inactive factors (as determined by the sensitivity study) were replaced by constant values reflective of typical levels for the case study region. The HDM-III model predictions from using the full set of inputs, and those obtained by using default values in place of the inactive factors, were statistically compared.; The research findings confirm the suspected nature of factor sparsity (few active factors) of the HDM-III model. The agency and road users' life-cycle costs were found to be dominated by very few input factors. More importantly, the factor sensitivity is specific to the R&M strategy employed.; Statistical comparison between the HDM-III life-cycle predictions based on default inputs and the predictions based on the full data set found that there was no significant difference. The study demonstrated the possibility of extending the current pool of default inputs for application-specific model output.