Pesticides are chemicals used in various agricultural practices to control pests, weeds and plant diseases. They are responsible for many negative impacts on the environment and human health. In addition, they are linked to the intensive use of agricultural machinery and equipment, and to mass production, often associated with monoculture. These non-ecological practices contribute to the contamination of freshwater by pesticides, as well as to the continuous soils degradation and the erosive processes increase. In this context, Rouzies et al. (2019) developed the dynamic and continuous model PESHMELBA with the aim of simulating pesticide transfers at the watershed scale, and comparing scenarios by explicitly taking into account the landscape spatial organisation. This model has a modular structure, which allows to improve the representation of some processes or landscape elements (such as plots, vegetative filter strips, ditches, rivers, etc.), or to add new ones. The current version of the model only estimates the pesticides transfer in solution (water), which may underestimate the impact of these solutes, as the part of them absorbed in the soil particles is not yet taken into account in the simulations. Thus, an erosion module was developed in order to integrate the transfer of pesticides in the particulate phase into the PESHMELBA model, and to quantify the soil loss and the particles transfer at the watershed scale. This model differs from other existing erosion models by performing continuous dynamic simulations, i.e. it takes into account the variation of sediment concentrations in space and time, and the continuity between precipitation events. The developed model simulates soil detachment by rainfall (interrill erosion) and by runoff (rill erosion), transfer of particles by laminar and concentrated surface runoff, and deposition of particles when transport capacity is exceeded, in particular inside landscape elements as vegetated strips or hedges, or where slope decreases. This deposition consolidates along time between two surface runoff events. This module integration inside PESHMELBA allows the comparison of different scenarios of agricultural practices and land use, by taking into account different configurations of the landscape elements. This first version of the erosion model is still in the testing phase, where the first simulations show consistent results and illustrate the interest of a continuous dynamic simulation, essential to represent pesticide transfer. Despite the current constraints and limitations of this model, it shows great potential to represent erosive processes in a more realistic way, by further refining and improving this first version in subsequent contexts.