2.5.0: High fidelity drag modeling

[ChristopherRabotin]

Summary

Nyx 2.5.0 brings major enhancements to atmospheric drag modeling alongside key fixes for CCSDS measurement processing, tracking data persistence, and expanded Python bindings.

---

The defining addition in this release is expanded atmospheric modeling, led by Nyx's first high-fidelity drag model: NRLMSISE-00 (#580).

Rather than wrapping legacy C or FORTRAN implementations, NRLMSISE-00 is written in 100% safe Rust. This eliminates foreign function interface (FFI) overhead and memory safety risks, allowing thread-safe parallel evaluation across Monte Carlo runs and estimation filters. The implementation has been independently verified against standard benchmarks for both localized density computations and integrated trajectory propagation.

• Space Weather Ingestion: Driven directly from CelesTrak CSV space weather files or via static space weather profiles for deterministic testing.
• Enum-Based Model Configuration (#586): Exposes all internal NRLMSISE-00 switches via strongly-typed enums, providing fine-grained control over specific physical effects (e.g., anomalous oxygen, geomagnetic storm terms, daily and annual density variations).
• Dimensional Consistency: The SI-versus-CGS switch is intentionally fixed to SI units ($kg/m^3$, $m$) to guarantee consistency with Nyx's equations of motion.
• Harris-Priester Model Integration (#583): Added the classic Harris-Priester empirical atmosphere model, offering a lightweight, low-altitude density calculation alternative.

---

Orbit Determination & Tracking Data

• Parquet Rejection State Persistence (#578): TrackingDataArc export and import routines now preserve measurement rejection flags in Parquet files, keeping estimation state histories intact across post-processing pipelines, and allowing for manual editing of the file to set the rejection status
• Rigorous TDM Range Unit Parsing (#581): Implemented strict parsing and physical unit conversions for RANGE_UNITS within CCSDS Tracking Data Messages (TDM), resolving physical scale discrepancies in range observations.
• PositionDevice Tracking & Python OD Bindings (#590): PositionDevice trackers now enforce explicit frame definitions to eliminate coordinate ambiguity during measurement modeling. Additionally, full PositionDevice orbit determination capabilities are now exposed through Python bindings.

---

Quality of Life & Infrastructure

• Transparent Gravity Field Compression (#587): GravityFieldData now auto-detects gzip compression on field files, removing the redundant manual configuration flag.
• Python Type Stubs (#588): Updated .pyi type hints for the 2.5.0 API surface, ensuring proper static analysis and autocompletion support in Python IDEs.

---

Pull requests

New Features

• Add Naval Research Lab MSISE-00 drag model by @ChristopherRabotin in Add Naval Research Lab MSISE-00 drag model #580
• Add harris priester drag model by @ChristopherRabotin in Add harris priester drag model #583
• Allow configuration of NRLMSISE-00 switches via enums by @ChristopherRabotin in Allow configuration of NRLMSISE-00 switches via enums #586

Improvements

• Add tracking frame support and Python bindings for PositionDevice by @ChristopherRabotin in Add tracking frame support and Python bindings for PositionDevice #590
• feat: Export/import rejected status in TrackingDataArc Parquet by @ChristopherRabotin in feat: Export/import rejected status in TrackingDataArc Parquet #578
• fix: implement rigorous RANGE_UNITS parsing with correct physical conversions in TDM files by @ChristopherRabotin in fix: implement rigorous RANGE_UNITS parsing with correct physical conversions in TDM files #581
• Remove gunzipped config field and auto-detect gzip in gravity field loader by @ChristopherRabotin in Remove gunzipped config field and auto-detect gzip in gravity field loader #587

Fixes

• Fix pyi 2.5.0 by @ChristopherRabotin in Fix pyi 2.5.0 #588

Full Changelog: 2.4.2...2.5.0

---

This discussion was created from the release 2.5.0: High fidelity drag modeling.