Assimilation Results for Egan & Powell (TBD): Physics-Informed Neural Networks as Differentiable Surrogates for 4D-Variational Data Assimilation

This archive contains the assimilation results supporting the manuscript “Physics-Informed Neural Networks as Differentiable Surrogates for 4D-Variational Data Assimilation” (Egan & Powell, TBD). It includes 10,000 twin-experiment results across three configurations: baseline, irregular-sparse, and field-informed. The .pkl files contain 4D-Var performance metrics for both the Traditional-NPZ and PI-NPZ models. File naming conventions: two_samples_per_day → Twin Experiment 1 (Baseline) irregular_sparse → Twin Experiment 2 (Irregular-Sparse) field_informed → Twin Experiment 3 (Field-Informed) File suffixes: xb_data → Initial background states and perturbed samples assimilation_10000xb_estimates → 4D-Var assimilation results for Traditional-NPZ and PI-NPZ All files are packaged within a single compressed archive (assimilation_results.zip) for convenience.