cesm_hawc.convergence¶

source

Convergence diagnostics for the L2 retrieval.

The retrieval’s output dataset carries no reliable convergence flag. The dependable source is the verbose=2 output that scipy.optimize.least_squares prints during the retrieval: capture stdout around the call (e.g. with contextlib.redirect_stdout) and pass it to parse_scipy_convergence().

parse_scipy_convergence¶

def parse_scipy_convergence(captured_stdout: str) -> dict

source

Read the convergence status from least_squares verbose output.

Parameters¶

• captured_stdout (str) — Everything scipy.optimize.least_squares(..., verbose=2) printed during one retrieval.

Returns¶

• dict — converged (bool), termination_reason ("ftol", "xtol", "gtol", "max_nfev" or "max_iter") and n_function_evaluations (int). Each value is None if no recognized message was found.

extract_l2_native_diagnostics¶

def extract_l2_native_diagnostics(l2_obj) -> dict

source

Read the iteration count and final cost from an L2 dataset.

Use these as a cross-check on parse_scipy_convergence().

Parameters¶

• l2_obj (xarray.Dataset or None) — The "l2" product from simulator.run().

Returns¶

• dict — l2_num_iterations (int) and l2_final_cost (float), each None if missing or unreadable.