Extensive H2O degassing in deeply erupted submarine glasses inferred from Samoan melt inclusions: The EM2 mantle source is damp, not dry

Submarine glasses erupted at intraplate volcanic hotspot settings sampling enriched mantle (EM)-characterized by high 87Sr/86Sr-exhibit lower H2O/Ce than glasses representing less enriched mantle domains, leading to the interpretation that the EM mantle is H2O-poor ("dry"). We test whether low H2O/Ce observed in pillow glasses of EM lavas resulted from degassing of higher H2O/Ce primary melts by measuring H2O/Ce and 87Sr/86Sr in olivinehosted melt inclusions from two deeply-erupted (3950 and 2190 m below sea level (mbsl)) Samoan submarine lavas from Vailulu'u and Malumalu seamounts. Vailulu'u (H2O/Ce = 161-275) and Malumalu (180-215) melt inclusions with 87Sr/86Sr data have an average H2O/Ce (197 +/- 58 2SD, N = 15) that is nearly twice as high as H2O/Ce in pillow glasses from these two seamounts (average H2O/Ce = 106 +/- 51, N = 65), and comparable to pillow glasses from non-EM hotspots. We show that lower H2O/Ce in submarine Samoan glasses compared to melt inclusions results from greater closed-system degassing, and concomitant loss of H2O, because EM melts have higher initial concentrations of CO2. We show that the lower H2O/Ce in global EM pillow glasses compared to non-EM pillow glasses can be modeled to be the result of more extensive degassing of H2O in EM melts, which owes to higher CO2 in primary melts (20,000-90,000 ppm) of EM sources compared to non-EM melts (300-50,000 ppm CO2). Instead of originating from a dry mantle, we conclude that EM lavas derive from a damp mantle, but EM melts lose more H2O by degassing than non-EM melts.