Rainfall-induced landslides cause many victims each year in mountainous and tropical environments. In the last century, more than a half million families in the tropical Andean region were affected by shallow landslides usually caused by heavy rainfall. Early warning systems that may reduce losses caused by rainfall induced landslides are commonly activated by rainfall thresholds. These thresholds can be obtained by statistical analysis when sufficient data is available concerning rainfall history and landslides inventory. However, when these data is scarce, physically-based models of slope stability become an attractive method for landslide susceptibility assessment (LSA). In the present study a comprehensive methodology is proposed to calibrate the input parameters for rainfall induced LSA . The proposed methodology consists in three stages: (i) model calibration using several key indexes to increase reliability, (ii) model validation with an inventory of landslides, and (iii) a sensitivity analysis to evaluate the changes in the geotechnical parameters, the water table position, and the rainfall return period on the LSA for different rainfall duration. The proposed methodology and key indexes are applied in San Antonio de Prado, an emerging neighborhood at the outskirts of Medellín, Colombia. Transient Rainfall Infiltration and Grid-Based Regional Slope-Stability (TRIGRS) model is used. The results of the calibration and validation stage suggest that TRIGRS can be used to assess the landslide susceptibility in tropical regions, to provide information for risk mitigation, and warning systems. However, the sensitivity analysis found that the effect of the rain duration is significant for most cases and is more pronounced when the geotechnical parameters are reduced by one standard deviation and when the groundwater table is one meter below the sliding surface. For these cases, the fraction of failed cells increases from 3\% to almost 19\% after 24 hours of rainfall. In addition, the fraction of failed cells does not increase linearly with respect to the rainfall duration. The sensitivity analysis highlights the importance of accurately measure key indexes, the geotechnical parameters for the different geological units and the water table position. Rainfall thresholds obtained with physically-based models should account for these factors variability for early warning systems.