Traffic models usually require extensive parameter definition and significant computing resources. Moreover, for localized road design, such as the reduction of lane numbers, large-scale models could exhibit shortcomings at the mesoscopic scale. This work presents a lattice Boltzmann method applied to various lane reduction scenarios and different traffic density hypotheses. First, the method is introduced and its relevance in modeling mesoscopic road traffic is verified against other methods, such as Greenshields, Drake, and Daganzo. Subsequently, the method is applied to two road design variants: the direct alteration of road geometry from 4 lanes to 2 lanes, or the alteration from 4 lanes to 3 lanes, and then from 3 lanes to 2 lanes. General parameters of the calculation include vehicle density, vehicle category in terms of size and desired speed, and some model variables such as time and space steps. Several design speeds are taken into account in these scenarios, ranging from 110 km/h to 70 km/h. Results demonstrate the relevance of this numerical method in anticipating various flow congestions related to road design and capacity. The method's advantage lies in its ability to easily change every parameter and quickly determine its impact on the infrastructure design process. In future work, the method could be applied to other traffic situations, such as driver speed variability induced in curves or speed variations on slopes.