DAN Retrieval Data Release User Guide

Overview

This archive accompanies the paper Computationally Efficient Parameter Retrievals for Active, Time-resolved, Neutron Albedo Measurements on Mars. It contains processed retrieval products used for comparison studies based on active measurements from the Dynamic Albedo of Neutrons (DAN) instrument on the Mars Science Laboratory (MSL) rover. The release is organized around two analysis sets:

- first_1900m_comparison/: products used for comparison against previously published results from the first 1900 m of the Curiosity traverse.
- marias_pass_comparison/: products used for comparison against prior Marias Pass analyses, including one-layer and layered retrieval cases.

These files are primarily intended as a reproducibility companion to the paper. They include processed observation inputs, retrieval intermediates, posterior summaries, and diagnostic plots generated by the retrieval pipeline. The package is not a raw telemetry release.

What the data represent

The retrieval workflow compares processed DAN active measurements against a library of MCNP forward-model outputs, then uses Markov Chain Monte Carlo (MCMC) parameter estimation to infer subsurface properties. Depending on the retrieval configuration, the fitted parameters can include:

- water-equivalent hydrogen (WEH)
- macroscopic neutron absorption cross section ($\Sigma_{abs}$)
- top-layer and bottom-layer values for layered models
- layer interface depth

The supplied code shows that the retrieval uses selected DAN time bins from background-subtracted observation data, constructs a simulation grid, estimates priors from the best-fit grid model, runs MCMC, and then saves posterior diagnostics and Gaussian-mixture summaries.

Directory layout

1. first_1900m_comparison/

Top-level contents:

first_1900m_comparison/
├── MCMC_ref_sims.csv
├── my_grid.csv
├── mitrofanov_results.csv
└── parameter_estimation/
    ├── sol00011/
    ├── sol00012/
    ├── ...
    └── sol00361/

Within parameter_estimation/, each solXXXXX/ directory contains one or more individual DAN observations. Each observation is stored in a folder named with its DAN product identifier, for example:parameter_estimation/sol00012/DNA_398555462EAC00120030004_______M1/

This level typically contains the observation-specific retrieval inputs and outputs.

2. `marias_pass_comparison/`

Top-level contents:

marias_pass_comparison/
├── site_1/
├── site_2/
├── ...
└── site_16/

Within each site, observations are grouped into a coadded regional product, for example:

site_10/
└── sol_1053-1056_Murray/
    ├── bg_dat.npy
    ├── coadded_raw.txt
    ├── coadded_bg_subtracted.txt
    ├── label.txt
    ├── raw_dat.npy
    ├── times.npy
    └── results/
        ├── one_layer/
        ├── two_layer_4_param/
        └── two_layer_5_param/

The results/ subdirectories separate retrieval products by model type.

Retrieval model folders

Where present, retrieval outputs are grouped by model type:

- one_layer/: homogeneous-material retrieval.
- two_layer_4_param/: layered retrieval with four free physical parameters.
- two_layer_5_param/: layered retrieval with five free physical parameters.

The supplied retrieval code defines representative parameter sets for one-layer, full two-layer, and reduced layered configurations, including WEH, $\Sigma_{abs}$, and depth. 

Typical analysis path

A typical path in this release follows the pattern:

[comparison set] / [sol or site] / [observation or coadd] / [results] / [model type] / [files]

Examples:

first_1900m_comparison/parameter_estimation/sol00012/DNA_.../
marias_pass_comparison/site_10/sol_1053-1056_Murray/results/two_layer_5_param/

File summary

Observation Level Data Files

Source Files

Retrieval Setup and Intermediate Files

MCMC output files

Diagnostic Plots

Gaussian-Mixture Summary Files

Top-Level Comparison Files

Notes for users

- This release is organized around processed retrieval products, not mission raw data.
- For most reuse cases, the most immediately useful files are bg_dat.npy, times.npy, apriori.npy, parameters.npy, results.npy, MCMC.h5.zip, corner.png, and all_walker_histories.png.
- Folder names encode the analysis context: either individual observations grouped by sol, or coadded regional products grouped by site.
- Where descriptions are blank, the attached paper and supplied code did not provide enough information to document the file confidently.

Suggested citation

Please cite the associated paper when using this dataset.