Landslides pose significant threats to human safety due to their inherent unpredictability and potential for severe human and financial losses. Monitoring landslide-prone areas is crucial, and while satellite surveys provide extensive topography and elevation data, in-situ detection tools such as piezometers and strain gauges offer precise monitoring of internal pressures and movements in areas affected by cracks. However, relying solely on empirical monitoring is insufficient for effective hazard management, especially in a preventive capacity. The high economic cost and practical challenges of conducting real experiments underscore the growing need for numerical techniques capable of simulating landslide phenomena. This study introduces a two-dimensional particle numerical method for modeling flow-like landslides, presenting a semi-conservative variant of the Depth-Averaged Material Point Method (DAMPM). The mathematical model is based on the Shallow Waters equations, derived by depth-integrating the Navier-Stokes equations. The framework accounts for both bed friction and rheology, incorporating the Voellmy model and the depth-integrated Bingham visco-plastic stress model, respectively. After validating the numerical method's performance through various benchmarks and idealized settings, it is applied to a realistic scenario.