Package: phylloptim 0.36.0

One surface, two consumers | Terms | The three calls | The solve path, and the eight kinds | Why the implicit function theorem | Why the kind is recorded, not inferred | Inside the surface | Extending it | What is tested | Deliberately not built

Simulated data, so the answer is known | The objective, and the loop that makes it affordable | Using the gradients | Faster or slower depends on which parameter count you mean | What reverses it is more fitted parameters per model parameter | Where the intercept goes | Fitting a different collection of parameters | Cost does not depend on the collection | A stem gradient means something different at a boundary | Not everything is reachable | What a new study actually has to write | Where a gradient-based fit goes wrong | Fitting to real data

What this model does differently | What is on this page | One solve | A drought response | The same sweep on a familiar scale | Light and VPD together | Traits and numerical settings are separate | Water potentials are positive magnitudes | Two supply paths, two routes | Multiple soil layers | The stateful interface | Trait gradients, and where they stop working | Choosing which gradients to compute | A collar potential you supply | A faster method: the closed form | When it applies, and when it refuses | What it costs, and what it buys | Accuracy, and one failure that is not a tolerance | The caveat that would stop you using this in a demographic model | What the model assumes | Where to go next

How to read this page | A leaf choosing where to operate | The problem being solved | The two supply paths | One problem, several coordinates | Solving for the optimum by its first-order condition | The scan, which does not run by default | When the answer is a bound, not a stationary point | The benefit, and the root-find inside it | The cost curves | The benefit link, and the proof it costs nothing | The wet-end price, and the one curve that has one | Pricing the drop is not pricing the potential | λ is the common currency | λ is not water use efficiency | λ is a common currency only across difference objectives | The two gates | Thermal damage | When the leaf cannot transpire | The variants, under one roof | Comparable at their defaults | The biological components | 1. Photosynthetic demand | 2. Temperature responses | 3. Diffusive supply | 4. Closing the loop: supply equals demand | 5. Hydraulic supply, and why the problem is non-trivial | 6. Optionally, energy balance and thermal damage | What the models disagree about | What the model assumes | Glossary | Where to go next | Appendix A — the (P50, c) parameterisation | Why not (P50, b) | Why c is not a constant | Converting from published numbers | psi_crit is a quantile, not a parameter | Three consequences for gradients | Appendix B — gradients | The implicit function theorem | The link makes one derivative serve every model | The envelope theorem, and where it stops | Prescribed versus optimised | A partial is not always a total | The energy-balance term | Appendix C — numerical practice | The maximum is flat, so its location is the delicate quantity | A finite difference through a solve has a noise floor, so sweep the step | A scan's argmax cannot feed a parameter gradient | References

The two umbrella headers, and why the extension matters | Building | A solve | Trait gradients from C++ | What is fast, and what is not | The golden file is the contract