Abstract. We consider the flow of marine-terminating outlet glaciers that are laterally confined in a channel of prescribed width. In that case, the drag exerted by the channel side walls on a floating ice shelf can reduce extensional stress at the grounding line. If ice flux through the grounding line increases with both ice thickness and extensional stress, then a longer shelf can reduce ice flux by decreasing extensional stress. Consequently, calving has an effect on flux through the grounding line by regulating the length of the shelf. In the absence of a shelf, it plays a similar role by controlling the above-flotation height of the calving cliff. Using two calving laws, one due to Nick et al. (2010) based on a model for crevasse propagation due to hydrofracture and the other simply asserting that calving occurs where the glacier ice becomes afloat, we pose and analyse a flowline model for a marine-terminating glacier by two methods: direct numerical solution and matched asymptotic expansions. The latter leads to a boundary layer formulation that predicts flux through the grounding line as a function of depth to bedrock, channel width, basal drag coefficient, and a calving parameter. By contrast with unbuttressed marine ice sheets, we find that flux can decrease with increasing depth to bedrock at the grounding line, reversing the usual stability criterion for steady grounding line location. Stable steady states can then have grounding lines located on retrograde slopes. We show how this anomalous behaviour relates to the strength of lateral versus basal drag on the grounded portion of the glacier and to the specifics of the calving law used.

with a planned heliocentric orbit that will carry it on a series of close passes by the Sun with perihelion distances that eventually will get below 10 solar radii. Among other in-situ and imaging sensors, the PSP payload includes the two-instrument "Integrated Science Investigation of the Sun" suite, which will make coordinated measurements of energetic ions and electrons. The high-energy instrument (EPI-Hi), operating in the MeV energy range, consists of three detector-telescopes using silicon solid-state sensors for measuring composition, energy spectra, angular distributions, and time structure in solar energetic particle events. The expected performance of this instrument has been studied using accelerator calibrations, radioactive-source tests, and simulations. We present the EPI-Hi measurement capabilities drawing on these calibration data and simulation results for illustrations.

Performing Bayesian inference via Markov chain Monte Carlo (MCMC) can be exceedingly expensive when posterior evaluations invoke the evaluation of a computationally expensive model, such as a system of partial differential equations. In recent work [10], we described a framework for constructing and refining local approximations of such models during an MCMC simulation. These posterior-adapted approximations harness regularity of the model to reduce the computational cost of inference while preserving asymptotic exactness of the Markov chain. Here we describe two extensions of that work. First, we prove that samplers running in parallel can collaboratively construct a shared posterior approximation while ensuring ergodicity of each associated chain, providing a novel opportunity for exploiting parallel computation in MCMC. Second, focusing on the Metropolis-adjusted Langevin algorithm, we describe how a proposal distribution can successfully employ gradients and other relevant information extracted from the approximation. We investigate the practical performance of our approach using two challenging inference problems, the first in subsurface hydrology and the second in glaciology. Using local approximations constructed via parallel chains, we successfully reduce the run time needed to characterize the posterior distributions in these problems from days to hours and from months to days, respectively, dramatically improving the tractability of Bayesian inference.

Scientists and engineers frequently rely on mathematical and numerical models to interpret observational data, forecast system behavior, and make decisions. However, unknown and neglected physics, limited and noisy data, and numerical error result in uncertain model predictions. The MIT Uncertainty Quantification library (MUQ) is a modular software framework for defining and solving uncertainty quantification problems involving complex models. MUQ is written in C++ but uses pybind11 (Jakob et al., 2017) to provide a nearly comprehensive Python interface. Users can access nearly all of MUQ's capabilities from either language.MUQ provides users many commonly used UQ tools and its modular design allows developers to easily modify, extend, and advance existing algorithms. For example, MUQ allows exact sampling of non-Gaussian distributions (e.g., Markov chain Monte Carlo and importance sampling), approximating computationally intensive forward models (e.g., polynomial chaos expansions and Gaussian process regression), working with integral covariance operators (e.g., Gaussian processes and Karhunen-Loève decompositions), and characterizing predictive uncertainties. The software is designed to support algorithm developers who want to easily construct new algorithms by exploiting a wide variety of existing algorithmic building blocks. Many UQ algorithms are model agnostic: Different physics-based or statistical models can be substituted into the algorithm based on the application. Therefore, MUQ enables users to quickly implement new models and exploit state-of-the art UQ algorithms.A suite of documented examples, including Gaussian process regression of Mauna Loa C02 observations, global sensitivity analysis of an Euler-Bernoulli beam, and a hierarchical Bayesian model of groundwater pump-test data, are provided to guide users through the process of implementing their own models and leveraging MUQ's UQ algorithms on quasi-realistic applications.

No abstract