The local expansion field is not isotropic: directional variation of H0 explains ~50% of the Hubble tension

The Hubble tension -- a 5-sigma discrepancy between local (H_0 = 73.04 +/- 1.04 km/s/Mpc) and CMB-inferred (H_0 = 67.36 +/- 0.54 km/s/Mpc) values -- remains one of the most pressing problems in cosmology. Using 2,098 Type Ia supernovae from the Zwicky Transient Facility Data Release 2 and 38,053 peculiar velocities from Cosmicflows-4, we decompose the local H_0 field into real spherical harmonics up to l_max = 5 and detect directional variation at p = 1.7 x 10⁻¹⁹ via a line-of-sight transfer function model. The shell model beats a simple dipole void by Delta chi^2 = 55 (p = 10⁻¹³) and the isotropic null by Delta chi^2 = 94 (p = 10⁻¹⁹), with a characteristic scale of ~222 Mpc consistent with the Keenan-Barger-Comerford (KBC) void. The quadrupole amplitude exceeds the dipole (A_2/A_1 = 1.02), confirming the kinematic backreaction prediction of Heinesen & Macpherson (2022) from exact general relativity. A Monte Carlo footprint test with 1,000 isotropic mocks at the exact ZTF sky positions confirms the signal is real: 0 out of 1,000 mocks reproduce the observed anisotropy (p < 0.001). Cross-correlation with Cosmicflows-4 peculiar velocities reveals a sign reversal at ~250 Mpc (negative inside, positive outside), independently validating the shell topology. A combined fit with 184 DES-SN5YR supernovae at 0.12 < z < 0.30, using a free calibration offset to avoid systematics, yields H_0^local = 69.9 +/- 0.4 km/s/Mpc (weighted least squares with intrinsic scatter, chi^2/dof = 1) and an effective underdensity delta = -0.22 +/- 0.03, consistent with KBC (delta = -0.30) at 1.4 sigma. The local void structure accounts for 45-56% of the Hubble tension (WLS vs OLS; 95% CI: [37%, 63%]), reducing the discrepancy from 5.0 sigma to approximately 2.2-2.8 sigma. No new physics is required for this reduction.