On the rise: using reentrants to extract magma ascent rates in the Bandelier Tuff caldera complex, New Mexico, USA

The Bandelier Tuff is the result of two subsequent caldera-forming eruptions of similar volume and composition which produced the Otowi (1.61 Ma) and Tshirege (1.26 Ma) members. Their remarkably similar characteristics and shared caldera boundaries provides a unique platform to investigate whether magma ascent is affected by the presence of a preexisting caldera boundary. Here, we present decompression rates for discrete layers within the initial plinian phase of each member by modeling volatile gradients (H 2 O, CO 2 absent) in quartz-hosted reentrants (unsealed melt inclusions). Successful best-fit 1D diffusion models for the lower ( n = 4/9) and upper ( n = 11/13) units resulted in average decompression rates of 0.041 MPa/s and 0.026 MPa/s, respectively. Strong overlap between rates extracted from the two eruptions suggests there was no significant change in ascent dynamics. However, the older Otowi member contains a larger number of reentrants that cannot be modeled adequately, suggesting a more complicated path than can be reconstructed with our constant decompression approach. In contrast, reentrants from the Tshirege can be readily modeled from storage depths, an observation that suggests conduit formation was more efficient in the second eruption. To further evaluate the robustness of these extracted rates, we then applied a 2D diffusion model, which considers various reentrant geometries; surprisingly, we find little alteration to 1D-derived rates. By contrast, incorporating the uncertainty in Bandelier temperature (~ 130 °C) shifts rates by 340–440%. However, we argue that the largest source of variation from decompression rates extracted from reentrants lies in the extreme range preserved within each fall deposit, each spanning three orders of magnitude, suggesting extreme conduit dynamic shifts, and emphasizing that petrologic-based ascent rates may vary widely, even within a single-sampled layer. Finally, the lack of detectable CO 2 concentrations in measured profiles is at odds with the amounts detected in sealed melt inclusions (