Skip to content

Diffusion-length / matrix-τ identifiability diagnostic + prior #32

Description

@awickert

A pre-calibration diagnostic (and a matrix-τ prior) based on the diffusion length L_D = √(K·H·t/S) over the calibration window: matrix recession timescale is only identifiable when L_D is large relative to the basin over the record. Below K ≈ 1e-6 m/s the matrix is hydraulically silent at decadal scales.

Cross-basin evidence: low-K basins (BLR/LSR/Wild Rice/Cottonwood) park matrix τ at bounds; Cannon (productive aquifers, K~1e-4) shows an identifiable matrix-τ decadal trend. Surface this before calibration and feed a prior on whether matrix τ is constrainable.

Depends on #30 (K, H). Full write-up: notes/HANDOFF_K_threshold_diffusion_length.md.

🤖 Generated with Claude Code

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    priors-from-dataDeriving model parameters/priors from external data (terrain, soil, land cover, remote sensing)

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions