Branched glycerol dialkyl glycerol tetraethers (brGDGTs) are a family of bacterial lipids which have emerged over time as robust temperature and pH paleoproxies in continental settings. Despite improvements in brGDGT analytical methods and development of refined models, the root-mean-square error (RMSE) associated with global calibrations between brGDGT distribution and MAAT in soils and peats remains high (~ 5 °C). Here we proposed to extend the global brGDGT terrestrial dataset previously proposed (n = 663; Dearing Crampton-Flood et al., 2019) with 112 soil samples from 6 altitudinal transects located in France (n = 49), Italy (n = 24), Tibet (n = 17), Chile (n = 8) and Peru (n =14). The transects were selected to take into account as much climatic and environmental variability as possible. All of these surficial soil samples (0 -10 cm depth) cover a wide range of temperatures (0°C to 26°C) and pH (3 to 8) and are representative of a wide diversity of environmental variables, vegetation and soil type. These new data were combined with previously published ones. This allowed the development of a new global terrestrial brGDGT temperature calibration from a worldwide extended dataset (i.e. 775 soil and peat samples) using a machine learning algorithm. This new model, called random Forest Regression for PaleOMAAT using brGDGTs (FROG), represents a refined brGDGT temperature calibration (R² = 0.8; RMSE = 4.01°C) for soils and peats, more robust and accurate than previous global soil calibrations while being proposed on an extended dataset.