Fit many curves at once

Several files

Give several files to one pk fit. Each is fitted on its own:

pk fit photosynthesis examples/aci.csv examples/messy_aci.csv --model empiricalA

The result lists each fit under results. Files that couldn't be read or fitted are listed under errors and don't stop the others. Add --format csv for one row per file, with the error in the last column.

One file, split by a column

--group-by <column> splits one file into a fit per value of that column, in the order the values first appear. Rows with a blank value are dropped.

pk fit hydraulics examples/pv.csv --model pvcurve --group-by Leaf
pk fit soil examples/soil.csv --units kPa --group-by sample --format csv

Group-by or CurveID?

For photosynthesis, --group-by CurveID fits each curve separately. Without it, the curves in a file are fitted together as one dataset that shares parameters, with the CurveID column used only for the per-curve constraints. Choose based on whether the curves should share parameters.

Thousands of datasets in Python

fit_many fits a list of datasets in parallel on all CPU cores. Use it when the data are already in memory, for example from pandas:

import phytkit as pk

datasets = []
for ci, q, t_kelvin, a in curves:   # one entry per curve
    datasets.append(({"Ci": ci, "Q": q, "T": t_kelvin}, a))

results = pk.fit_many("fvcb", datasets, init={"gm": 1e6},
                      fit_params=["Vcmax25", "Jmax25", "TPU25", "Rd25", "alpha", "theta"], threads=8)
for r in results:
    print(r["params"]["Vcmax25"], r["rmse"], r["converged"])

Results come back in the same order as datasets. threads defaults to every core. The fits run outside Python's global lock, so other Python threads keep running while they do.