The forest floor C stock needs to be accurately estimated in order to quantify its contribution to nutrient cycling and other ecological processes as well as for reporting purposes under international agreements. Hence, a modelling approach was used which involved testing three different types of models (GLM, GAM and random forest) to determine which one provided the best estimates of forest floor C stocks. The dataset employed contained over 1650 observations from different available sources embracing different climatic, topographic and biotic variables to be tested in the model. The approach that provided the best estimation of forest floor C stock was the random forest method, with forest type, latitude, altitude, canopy cover, mean summer temperature, annual accumulated temperature, summer precipitation, water deficit and the normalized difference vegetation index (NDVI) as covariates. To obtain a robust forecast, several iterations of the model were performed to estimate forest floor C stocks from the mean of the predictions. The model estimated a forest floor C stock of 0.148 ± 0.081 Pg, equivalent to a biomass of 0.381 ± 0.214 Pg, for a wooded area of almost 184,000 km2 in peninsular Spain and the Balearic Islands. The predictions were also presented in the form of a map showing the spatial distribution of the forest floor C stock. The results revealed a mean forest floor C stock of 8 Mg C ha−1 for Spanish forests and identified differences between coniferous (10.1 Mg C ha−1) and hardwood forests (6.3 Mg C ha−1).