The gridded yearly climate data (5-year moving averages) were developed from weather station observations (from IDEAM and Cenicafe) for the period 1980 to 2010, at 500 m spatial resolution, for monthly precipitation, monthly minimum temperature and maximum temperature, following the method of Hijmans et al. (2005). In areas with low station density and low interpolaton quality, weather observations were complemented with pseudo-stations from AgMERRA for temperature (Ruane et al., 2015) and CHIRPS (Funk et al., 2015) for precipitation. The results were aggregated into decadal and 30-yr climatology data. Future climate data was developed by downscaling CMIP5 projections from 14 General Circulation Models (GCMs) for all four RCPs (RCP 2.6, 4.5, 6.0 and 8.5; IPCC, 2013) following the method of Ramirez-Villegas and Jarvis (2010). Data were for 2020s and 2030s, which reflect the 10 and 15-year coffee planning cycles. Using the monthly data, we developed climate indicators for modelling following expert knowledge (from Cenicafe) and literature (CENICAFE, 2016) for specific seasons (DJF, MAM, and SON) associated with phenological events: mean temperature range (thermal amplitude), thermal time, mean temperature, temperature seasonality, precipitation seasonality and reference evapotranspiration.