Freight transportation in the United States (U.S.) faces persistent challenges from highway congestion and adverse environmental effects. Inland waterway transportation (IWT), particularly Container-on-Barge (COB), offers a potential solution to reduce congestion and emissions within multimodal freight networks. Although extensive IWT infrastructure is available, COB remains underutilized in U.S. freight corridors. In particular, the South Texas freight corridor, connecting Houston, Galveston, Corpus Christi, and San Antonio, currently has limited containerized barge service. To examine the potential of COB in the South Texas corridor, this study implements a conventional SimPy-based discrete-event simulation (DES) model to evaluate the impacts of COB integration on fuel costs, greenhouse gas (GHG) emissions, and system time. The analysis considers four scenarios. The first is a road-rail-only baseline. The other three introduce COB at 5-15% modal shares under standard, demand-growth, and congestion-delay conditions, respectively. Results show that a 10% COB share under congestion delays reduces fuel costs by up to 6.94% and CO2 emissions by 6.07% relative to the non-COB baseline. However, system time increases by 5.56% under demand growth conditions. The estimated cost and emission benefits can support port authorities and terminal operators in evaluating investment decisions for barge-handling operations along the South Texas IWT corridor.
The behaviors of five random number generators (RNGs) available for the C programming language (rand, random, ran1, drand48, and Mersenne Twister) are investigated with spatial population models with global dispersal and with local dispersal. It is shown that unusual behavior in RNGs which would seem to be of little or no practical significance can in fact be greatly amplified when the RNG is used to simulate spatial population models. More specifically, a relative error of 0.12% in the frequency of random values below a specified cutoff is amplified by a factor of more than 400, leading to a relative error of 48.9% in the equilibrium density for a stochastic two-dimensional spatially explicit continuous-time population model. Another implementation of the same model, which uses one fewer random value when handling reproduction events, and which in theory should yield identical results, gives a relative error of -18.6% in the equilibrium density. This amplification of anomalous behavior based on generated random values can be exacerbated when the value of a generated random number affects how many additional random values will be subsequently generated.
