ONR Workshop on "PIPS 3.0"
7-9 July 1998, Monterey,CA
by
Erland Schulson(Leader) and Bob Pritchard(Rapporteur)
Dempsey reminded the audience that the resistance to crack propagation in sea ice increases with increasing size of the feature, from around 100 kPam0.5 at lab scale (0.1 m) to around 250 kPam0.5 on the larger scale (80 m). This effect, he suggested, should be incorporated in PIPS 3.0. He noted that at the moment geophysical scale modelers don't think they need fracture mechanics, and added that intersecting lead patterns will not be determined using slip line field theory. He asked if echelon cracking (under compressive and shear loading) explains the rectilinear lead patterns onserved in satellite imagery.
Hopkins presented his discrete element model (DEM) of ice floe dynamics, limited to the 10 km scale. It consists of a set of interacting polygons(multi-year ice) bordered by first-year ice and by new ice. Failure occurs in the border by tensile rupture. Animated videos of the evolution of failure under shear loading appeared to capture some of the elements of lead formation, including their spacing which may be a function of tensile strength. He suggested that his DEM could be nested within a regional model to obtain better resolution at certain loacations, for both ice behavior and noise source generation. Mcphee noted that the DEM always splits the thinnest floes,but in fact thick ice often fails.
Hibler presented new modeling of the flow and fracture of an ice sheet, based upon anisotropic mechanical behavior. Starting with randomly oriented lines of weakness (i.e., potential leads) within each element (10 km on side), he studied the behavior of the composite of thin ice (refrozen lead) and thick ice. He assumed that the applied strain rate is the sum of the strain rates in the thin and thick ice in proportion to their area. He showed that the actual leads that are expected to fail depend upon the ratio of the applied stresses (biaxial loading). The model shows that the acute angle defining intersecting leads increases from around 18o to 40o as the ratio of the minor/major compressive stress increases from 0.025 to 0.20. Acute angles of this magnitude are seen in satellite images. He also offered the process of kinematic inertial coupling to explain the rectilinear lead patterns observed in satellite images.
Schulson noted the similarity in the appearance of fracture patterns in the field and in the lab. He noted wing cracks and interescting leads/"minileads" on both scales. Acute angles of lead intersection in the lab range from 38-45o , compared with 20-40o in the field, in both cases different from the 90o angle predicted from slip line field theory for pressure-insenstive flow. C-axis alignment and/or different stress ratios may explain the difference between the lab and field orientations and frictional effects may explain the large departure from orthogonality. He suggested that similar failure mechanisms operate on both scales, and then presented a new mechanism for the triggering of brittle compressive shear faults. He argued that PIPS 3.0 should include a Coulombic type of failure criterion that takes explicit account of fracture, friction and crack size. Pritchard said that we have begun to use a shorthand description over the years, and that sea ice plasticity is the most general possible definition. The difference in opinion may be one of semantics, "plasticity" to some people (incldg Schulson) meaning crack-free deformation.
Overland presented SAR images of deformation fields. He showed that the "slip lines" that constitute leads are actually composed of regions of damage several km wide. He said strongly that the "Aidjex model got it right." He wants to consider wind-forced problems, especially when atmospheric conditions are similar enough that the ice moves such that lead patterns are relatively fixed in space. He showed a case during SIMI when buoys displayed lead opening and shearing. He noted that GPS buoys on a 10 km ring, stress measurement within that ring, and SAR imagery will provide data to test new ice models. He then argued that on the scale of interest the ice cover should be regarded as a granular material. Schulson noted that the 10 km damage region bordering leads is similar to the damage zone that constitutes compressive shear faults in the lab, and then suggested that failure within the damage zone may be the trigger that sets off the lead/fault. He also suggested that it is within the damage region that the ice may undergo a transition from a continuum to a granular material.
Zhang addressed differerent methods of running models. He argued for a more efficient numerical scheme to integrate the Hibler model. The one he proposes is based on an ADI approach. He compared four different approaches: PIPS 2 using a PSR that converges slowly; Zhang/Hibler introduced a fast LSR in 1997; Hunke/Dukawicz EVP model has a non-physical elastic wave; and Zhang/Rothrock ADI scheme is not iterative and is therefore fast, efficient, stable and good for parallel computers. Simulations using ADI show that most stress states lie on the yield envelope. Pritchard questioned this result, saying that he expects roughly linearly varying stress states in a balance between wind stress and stress divergence. Zhang uses different yield envelopes and likes the Hibler-Schulson teardrop. Pritchard pointed out that the teardrop cannot sustain the unconfined compressive stress states needed to simulate arching observed across thre Bering and other Straits. The teardrop should be modified to include uniaxial compressive and tensile strength.
Parkington described the NIC 5-year plan. He argued that global
products (daily ice charts) must be completely automated to allow
fous on local charts. He noted that the quality of the data input to
PIPS should be improved. For example, the SSM/I underestimated
the ice edge when the ice is wet. Without better data, the
improvement in physics will be wasted. They want a specific plan for
assimilation, not pie-in-the-sky research ideas, even if the scheme is
simplistic. The NIC recommends several improvements: improve
assimilation of SSM/I data; use 85 GHz motion data from SSM/I; use
the 0-hr PIPS 3.0 forecast along with assimilated SSM/I data to give
the daily nowcast. Preller suggested that the nowcast should be
simulated by beginning with a minus 48-hr initial condition. the 0-hr
forecast using PIPS would then include these data and PIPS physics.
Curtin suggested that buoy data be assimilated with SSM/I data. In
discussion, it was noted that SSM/I "data" are really products of
image analysis and so some caution is appropriate.
Pritchard described a complete elastic-plastic constitutive law. It is
composed of a yield surface defined in stress space, a normal flow
rule, a stiff linear elastic closure, and a kinematic relationship
between elastic strain rate, stretching, spin and plastic stretching.
Ice conditions are described by an oriented thickness distribution. He
assumes that isotropy is a limit in which equal fractions of ice having
all orientations exist. The yield surface is then a body of revolution in
stress invariant space , and it must be composed of a conical cap
lying along the tensile cutoff cone (45o line in shear stress-normal
stress space), a compressive cap that is also a 45o cone, and pressure
insensitive (flat line) shear strength surface. This surface could take
any shape in invariant space (Mohr-Coulomb, frictional, etc.).Schulson
noted in discussion that within the context of a Coulombic failure
criterion the 45o slope and the 0o slope correspond to friction
coefficients of infinity and zero, respectively.