Description: Highlights: • Three different types of pCO2 sensors detected sedimentary artificial CO2 leaks in the water column. • Distribution of leaked CO2 in the water column featured high temporal and spatial heterogeneity. • Clear effect of CO2 leakage on the water column was visible only at high flow rates and low tides. • Fast recovery of the water column pCO2 was observed after the CO2 release was stopped. • Multivariate statistics can help to distinguish between anthropogenic and natural CO2 sources. Abstract: This work is focused on results from a recent controlled sub-seabed in situ carbon dioxide (CO2) release experiment carried out during May–October 2012 in Ardmucknish Bay on the Scottish west coast. Three types of pCO2 sensors (fluorescence, NDIR and ISFET-based technologies) were used in combination with multiparameter instruments measuring oxygen, temperature, salinity and currents in the water column at the epicentre of release and further away. It was shown that distribution of seafloor CO2 emissions features high spatial and temporal heterogeneity. The highest pCO2 values (∼1250 μatm) were detected at low tide around a bubble stream and within centimetres distance from the seafloor. Further up in the water column, 30–100 cm above the seabed, the gradients decreased, but continued to indicate elevated pCO2 at the epicentre of release throughout the injection campaign with the peak values between 400 and 740 μatm. High-frequency parallel measurements from two instruments placed within 1 m from each other, relocation of one of the instruments at the release site and 2D horizontal mapping of the release and control sites confirmed a localized impact from CO2 emissions. Observed effects on the water column were temporary and post-injection recovery took 〈7 days. A multivariate statistical approach was used to recognize the periods when the system was dominated by natural forcing with strong correlation between variation in pCO2 and O2, and when it was influenced by purposefully released CO2. Use of a hydrodynamic circulation model, calibrated with in situ data, was crucial to establishing background conditions in this complex and dynamic shallow water system.