LeafPhotosynthesis.jl
A small Julia package for composable leaf photosynthesis. Students and coupled models call the same process functions. Parameters and leaf environments use FT; the implementation is tested on CPU and NVIDIA CUDA in Float32 and Float64.
Scope: C3 biochemistry (both Farquhar–von Caemmerer–Berry and Johnson & Berry 2021 Cytochrome LeafModel composes concrete photosynthesis, stomatal, and oxygen components under abstract interfaces. NoOxygenExchange skips the oxygen calculation; NoStomatalDynamics keeps prescribed conductance fixed. There are no runtime dependencies or hidden scientific parameters. C4, TPU, prognostic NPQ relaxation memory, leaf energy balance, Metal execution, and ClimaLand integration remain future work.
The PSII fluorescence component and the coupled Cytochrome
Installation
The package is not registered yet. Add it from GitHub:
using Pkg
Pkg.add(url = "https://github.com/RSFraLab/LeafPhotosynthesis.jl")A first leaf
julia> using LeafPhotosynthesis
julia> temperature = C3TemperatureResponse(; # rates at 25 °C; kc, ko and Γ* in Pa
vcmax = ConstantResponse(60.0), jmax = ConstantResponse(120.0),
respiration = ConstantResponse(0.9), kc = ConstantResponse(40.5),
ko = ConstantResponse(27840.0), gamma_star = ConstantResponse(4.28));
julia> transport = ElectronTransportParameters(;
psii_fraction = 0.5, psii_yield = 0.85, limitation = QuadraticLimitation(0.7));
julia> leaf = LeafModel(; photosynthesis = C3Photosynthesis(; temperature,
electron_transport = transport,
stoichiometry = C3Stoichiometry(; electrons_per_carboxylation = 4.0, oxygenation_cost = 2.0),
limitation = MinimumLimitation(), assimilation_sign = SignedAssimilation()));
julia> environment = LeafEnvironment(298.15, 1000.0, 28.0, 21000.0);
julia> result = photosynthesize(leaf, environment);
julia> round(result.net_assimilation; digits = 2) # μmol CO₂ m⁻² leaf s⁻¹
13.47The arguments of LeafEnvironment are leaf temperature (K), absorbed PAR (μmol photons m⁻² leaf s⁻¹), and CO₂ and O₂ partial pressures at the biochemical site (Pa). The values above are illustrative and temperature-independent. The Bonan 11.1 walkthrough builds the same kind of leaf with Arrhenius temperature responses from a parameter file with units and provenance.
Where to go next
| Question | Page |
|---|---|
| What do the examples produce? | Example gallery |
| How do empirical and mechanistic SIF models compare on Flexas et al. (2002)? | Flexas explorer, walkthrough |
| How do the lake-model relations of Gu et al. (2019) compare with Johnson & Berry (2021)? | Lake-model fluorescence |
| What belongs in this package, and how will coupling work? | Design decisions |
| Which Bonan exercises fit this package? | Bonan examples |
| How do the optional core components fit together? | Core gas exchange |
| How is coupled A–Ci equilibrium solved, and how do fluxes respond to stomata? | Equilibrium and sensitivities |
| How are gross turnover and realized electron flow diagnosed? | Biochemical turnover |
| How does Γ* respond to oxygen? | Oxygen response |
| How do external optics supply PSII photons? | Excitation |
| How do Cytochrome | Cytochrome |
| Where is every type and function documented? | API reference |
| Where are the scientific values, units and sources? | References and examples/parameters/*.toml |
Development scope
Models are immutable parameter containers. Scalar evaluation returns named values; the caller owns arrays, devices, and state evolution. Equilibrium iteration has explicit settings and convergence diagnostics. Broadcast can apply the same function to independent leaves. A small number of concrete components should express physical choices without a large hierarchy of abstract containers.
Examples here cover problems solvable within leaf physiology: Bonan 11.1 is implemented, and prescribed-temperature gas exchange in 12.1 is a candidate as the necessary leaf components are added. A separate example repository can later combine the physics packages for broader exercises. Whole-book coverage is not a requirement to expand this package into soil, canopy transport, or ecosystem dynamics.
This is a new implementation organized by biophysics, not a file-by-file port of Photosynthesis.jl or Emerald. Scientific references and source provenance are in References.
Optional NVIDIA validation
CUDA is isolated in a test environment; it is not a core dependency. From the package root, on a GPU-enabled host:
julia --project=test/gpu -e 'using Pkg; Pkg.instantiate()'
julia --project=test/gpu test/gpu/runtests.jlThe test disables scalar GPU indexing, evaluates every response-curve environment plus mixed dark/lit leaves in both orders, and compares all diagnostics with CPU results for FP32 and FP64. Its batch checks also run on CPU in the regular suite. An unavailable device causes failure, not a skipped test that could look like success. The core and no-op components pass on an NVIDIA L40S with Julia 1.12.6 and CUDA.jl 6.5.1. Older CUDA 5.11.3 failed compilation in the tested Julia environment; support for every version in the optional environment's compatibility range is not implied. Apple Metal remains untested.
The CUDA tests are not run in continuous integration. How to run the CPU test groups and build this manual locally is described in the README.
License
LeafPhotosynthesis.jl is licensed under the Apache License, Version 2.0. The vendored Chart.js and KaTeX files used by the interactive charts are MIT licensed; their licence texts are kept next to them in docs/src/public/vendor/.