The development and the calibration of a microscopic traffic simulation model, using MITSIMLab, for the entire metropolitan area of Des Moines, Iowa, are presented. The primary contributions include the applica- tion of a microscopic model on such a large-scale network and an effort for joint calibration of the model parameters and estimation of origin–destination flows. The application of microscopic traffic simu- lation models to very large networks such as this poses a number of methodological and practical challenges that are not faced with smaller applications. Solutions to these problems are both heuristic and analyt- ical. The solutions presented are generic and hence applicable to any large-scale microscopic traffic modeling. Microscopic traffic simulation models have drawn significant attention from both practitioners and academicians in recent years. However, their applications are limited to small to medium-sized networks. Furthermore, the calibration of the simulation model is limited to ad hoc changes in a few driving parameters to match field conditions. Although such calibration methods often result in satisfactory per- formance for small networks, a much more thorough calibration that includes both estimation of origin–destination (O-D) flows and route choice and driving behavior parameters is needed for large-scale applications. This paper presents the development and calibration of a large-scale microscopic traffic simulation model using MITSIMLab (1, 2) for the metropolitan area of Des Moines, Iowa, and derives insights from this application. Simulation models have been applied to perform operational analy- sis of highways for a number of decades. However, their applica- tion to complex networks is fairly recent. With the development of new traffic simulation models such as AIMSUN (3), MITSIMLab, PARAMICS (4), and VISSIM (5 ), it is now possible to simulate increasingly larger networks with complex scenarios that involve intelligent transportation system (ITS) elements, incident scenar- ios, highway construction, and such. Even though the simulation of large networks is similar to that of small ones at the abstract level, it poses a number of practical (and sometime theoretical) difficulties concerning the development and calibration of such models. Some of these difficulties have not been addressed in the literature so far and are therefore a significant obstacle to the application of microscopic traffic simulation models to large-scale networks. Researchers have long been concentrating their efforts toward the calibration of microscopic simulation tools to match the field condi- tions. Most studies have focused on either parameter calibration or O-D estimation, but not both. Some of the methodologies adopted for calibrating parameters include simple search techniques (6 ), genetic algorithms (7, 8), and a simplex-based approach (9). Approaches that have been adopted for O-D estimation include generalized least squares (GLS) (10, 11), maximum likelihood (12, 13), and entropy maximization or information minimization (14). It is only recently that O-D estimation and parameter calibra- tion are being done jointly. Liu and Fricker (15) sequentially esti- mate O-D flows and route choice parameters for uncongested networks by first fixing route choice parameters and estimating O-D flows and then using the estimated O-D flows as inputs to estimate the route choice parameters. Toledo et al. (16 ) propose an iterative approach to calibrate model parameters jointly and estimate O-D flows with aggregate data and apply the method to calibrate MITSIMLab for a test network in Stockholm, Sweden, under congested traffic conditions. This approach is also applied in Darda (17 ) for a network in Irvine, California. The rest of this paper is organized as follows: the next section describes the project and input development followed by a brief description of our methodology for calibration and O-D estimation. Practical challenges that were faced in the development and cali- bration of large-scale models are described next, followed by pre- sentation of calibration and validation results. Finally, we provide some concluding remarks concerning the future applications of such models. PROJECT DESCRIPTION The Des Moines Area Metropolitan Planning Organization (MPO) and Iowa Department of Transportation (DOT) jointly decided to develop a large-scale microscopic traffic simulation model using MITSIMLab for the entire Des Moines area. This model is intended to complement the existing regional planning model and would enable the agencies to perform detailed operational analyses of traf- fic ranging from studying the impact of a planned reconstruction project that would cause significant route diversions to evacuation planning. Traditionally, only regional models are used for both short- and long-term policy decisions. In the immediate application the Development and Calibration of a Large-Scale Microscopic Traffic Simulation Model Mithilesh Jha, Ganesh Gopalan, Adam Garms, Bhanu P. Mahanti, Tomer Toledo, and Moshe E. Ben-Akiva Transportation Research Record: Journal of the Transportation Research Board, No. 1876, TRB, National Research Council, Washington, D.C., 2004, pp. 121–131. M. Jha and G. Gopalan, Jacobs Civil, Inc., 222 South Riverside Plaza, Chicago, IL 60606. A. Garms, Des Moines Area Metropolitan Organization, Merle Hay Centre, Urbandale, IA 50322. B. P. Mahanti and T. Toledo, NE20-208, and M. E. Ben-Akiva, 1–81, Center for Transportation and Logistics, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge MA 02139. 121