- General Characteristics
The GPUFLOW module allows the modeling and simulation of geophysical flows. GPUFLOW is a 2D cellular automaton for the modeling of Bingham fluids with or without temperature-dependent rheological parameters. The computational domain is a rectangular grid covering the area of interest, divided into square cells. Each cell holds information about the local topography and the amount of fluid present in the cell; in the case of temperature-dependent rheology with phase transition (i.e., lava), the cell information also includes the temperature and heat for the fluid, as well as the amount of fluid that has solidified. The topographical information is taken from a digital surface model (DSM), with a resolution matching that of the cellular automaton. The initial conditions depend on whether a lava flow is being simulated. For lava flows, one or more cells can be marked as eruptive vents, with an associated mass flux rate (possibly varying in time), such that at the beginning of each time-step the amount of fluid in the cell is increased according to the current flux rate. It features a physical model for the thermal and rheological evolution of lava flows (including temperature-dependent emissivity and cooling by radiation and air convection), a parallel implementation on graphic processing units (GPUs), and a simpler and computationally more efficient solution to the grid bias problem.
- Specific Products Description
The module requires a number of input parameters: the DSM of the area of interest (defining the simulation domain and resolution), a description of the physical parameters of the fluid being simulated (density, rheological parameters, thermal parameters if appropriate), and the location of the vents and their mass flux rates over time (in the case of lava flows). The model produces as output snapshots of the automaton state, providing information about the fluid thickness (and temperature, if appropriate) in each cell of the domain. By combining different lava flow simulations (using different input data), the module also allows producing hazard maps showing the probability of lava flow inundation (see fig.1).
Figure 1: Short-term lava flow hazard map obtained by GPUFLOW during the first phase of the 2021 Tajogaite eruption
- Products Availability (spatial resolution, spatial coverage, temporal resolution, delay, error etc.)
Lava flow simulations and inundation maps have a spatial resolution of 10 meters (or less, depending on data input) and are generated on demand. The spatial coverage is automatically derived from the lava flow emplacement of the lava flow. The outputs provide continuous data (as the input data changes) to enable the development of updated scenarios and therefore detailed analyses. The delay in producing lava flows scenarios depends on the availability of input data and the execution of numerical simulations. However, the GPU implementation of the module provides speed-ups of two orders of magnitude over its serial CPU counterpart, allowing 7-day forecasts for lava flow to be obtained in a few seconds and a month’s simulation in a few minutes.