The carbon cycle displays strong sensitivity to short term variations in environmental conditions, and it is key to understand how these variations are linked with variations in CO2 fluxes. Previously, atmospheric observations of CO2 have been sparse in many regions of the globe, making it challenging to evaluate these relationships. However, the OCO-2 satellite, launched in July 2014, provides new insight into global CO2 fluxes, particularly in regions that were previously difficult to monitor. In this study, we combine OCO-2 observations with a geostatistical inverse model to explore data-driven relationships between inferred CO2 flux patterns and environmental drivers. We further use year 2016 as an initial case study to explore the applicability of the geostatistical approach to large satellite-based inverse problems. We estimate daily, global CO2 fluxes at the model grid scale and find that a combination of air temperature, daily precipitation, and photosynthetically active radiation (PAR) best describe patterns in CO2 fluxes in most biomes across the globe. PAR is an adept predictor of fluxes across mid-to-high latitudes, whereas a combined set of daily air temperature and precipitation shows strong explanatory power across tropical biomes. However, we are unable to quantify a larger number of relationships between environmental drivers and CO2 fluxes using OCO-2 due to the limited sensitivity of total column satellite observations to detailed surface processes. Overall, we estimate a global net biospheric flux of -1.73 ± 0.53 GtC in year 2016, in close agreement with recent inverse modeling studies using OCO-2 retrievals as observational constraints.