Lima holds ~30% of Peru's entire population. It is listed among the 30 most populated cities of the world, and since the past decade, it faces an increasing water stress. In this context, evaluating future precipitation patterns in the Rimac Basin, which runs along Lima and provides for 60% of the drinking water to its population, is crucial. In this paper, we use statistical downscaling (SD) to correct and evaluate 22 years (Set/1982-Ago/2004) of historical daily precipitation data from the HadGEM2 general circulation models (GCM) to assess a near future time frame (Dec/2015-Nov/2045) being centered in 2030. To perform SD, we apply quantile-mapping techniques that uses a non-parametric transformation function being represented by the Bernoulli-Gamma mixture probabilistic model from eight ground weather stations. HadGEM2 models were evaluated by considering two model efficiency metrics, namely: bias and relative root mean squared error (RRSME). Our results show that the bias and RRSME were reduced to the ranges 0.7-1.6 units and-2.3%-8.3%, respectively. Likewise, we determined that the HadHGEM2-ES and HadHGEM2-CC models performed markedly better during dry and wet seasons.