Flow velocities and stresses within a glacier are determined by invert
ing known surface velocities with a specified glacier geometry. The su
rface velocities depend only weakly on the unknown velocities at the b
ed of a glacier, so the inversion is ill-posed and unstable. This inst
ability causes both numerical computation errors and data errors to gr
ow dramatically with depth, usually masking the actual velocity and st
ress solutions. To control the numerical errors, an analytical modelin
g scheme is presented which modifies the method of mean weighted resid
uals (used in finite element techniques). The resulting scheme impairs
convergence by producing power-series solutions, bur in an advantageo
us trade-off, the coefficients to the power series can be determined a
nalytically rather than numerically. This leads to arbitrary order ana
lytical power-series solutions to the internal stress state of glacier
s. The symbolic power-series solutions cart be evaluated at any point
in the glacier with negligible round-off and discretization errors. An
alytical model accuracy is confirmed with known stress solutions for s
everal widely used constitutive relations for ice.