ashiklom/rrtm

1. What this repository is

- rrtm is an R package for vegetation radiative transfer modeling. Its core purpose is to compute wavelength-resolved reflectance/transmittance spectra for leaves, canopies, and soils, and to combine those into higher-level remote-sensing models.

2. Primary user-facing layer

- The main API lives in R/*.R and is exported through NAMESPACE.

- Key entry points include prospect*, foursail, pro4sail_*, liberty, gpm, edr, hapke_soil, and spectral-response helpers. These are the functions a caller uses directly from R.

3. Model subsystem boundaries

- The code is organized by physical model rather than by application layer.

- Leaf optics are handled by PROSPECT-family code; canopy radiative transfer is handled by SAIL/4SAIL; soil/background reflectance is handled by Hapke-based logic; additional vegetation models include LIBERTY, GPM, and EDR-related routines.

4. Composition of models into end-to-end workflows

- The package’s main architectural pattern is model chaining: leaf spectra feed canopy scattering/transfer models, which then combine with soil/background reflectance.

- The README’s pro4sail_5() example illustrates this: PROSPECT-derived leaf spectra are coupled with 4SAIL canopy calculations and a soil model to produce multiple directional reflectance outputs.

5. Compiled numerical core

- The src/ directory contains the C++ implementation, with solvearray.cpp as the main compiled numerical routine and RcppExports.cpp/RcppExports.R providing the R-to-C++ bridge.

- This is the package’s performance boundary: higher-level model orchestration is in R, while the heavy numerical solving is pushed into compiled code.

6. Shared utilities and internal math

- R/utils.R, R/lidf.R, R/sail-volscatt.R, and related files hold reusable mathematical helpers and geometry/scattering calculations.

- These functions support multiple models rather than representing separate public workflows.

7. Data and parameter assets

- The package ships with precomputed/internal datasets in data/ and sysdata.rda, including sensor response data.

- extdata/ contains fixed reference tables used by models, such as soil and wood reflectance files.

8. Data preparation pipeline

- data-raw/ is the reproducible source for generated package data.

- Scripts there convert raw spectra, sensor descriptions, and reference tables into packaged .RData assets, separating ingestion/preprocessing from runtime model execution.

9. Validation and regression coverage

- tests/testthat/ contains model-specific tests for PROSPECT, LIBERTY, EDR, SAIL, spectral response, and an integration-style pecanrtm test.

- This indicates the package is validated both at the individual model level and across model combinations.

10. Documentation and reproducibility surface

- README.Rmd, the generated README.md, man/ Rd files, and the vignettes/inversion.Rmd vignette form the main documentation layer.

- The vignette suggests at least one higher-level workflow around parameter inversion, not just forward simulation.

11. Build and quality tooling

- The package uses standard R package tooling plus compiled-code build files (src/Makevars, src/Makevars.win).

- CI is run via GitHub Actions (.github/workflows/R-CMD-check.yaml), and .lintr indicates linting/static style checks are part of the repo’s quality boundary.

12. Likely architectural diagram framing

- A useful diagram would show: external user input parameters and sensor/soil reference data on the left; R model modules in the center; compiled C++ solver underneath as the performance layer; packaged datasets as a supporting store; and outputs as spectra/reflectance products returned to R callers.