Unit Scaling
import numpy as np
import pandas as pd
import pyrolite.geochem
pd.set_option("display.precision", 3) # smaller outputs
Here we create an example dataframe to work from, containing some synthetic data in the form of oxides and elements, each with different units (wt% and ppm, respectively).
from pyrolite.util.synthetic import normal_frame
df = normal_frame(
columns=["CaO", "MgO", "SiO2", "FeO", "Ni", "Ti", "La", "Lu"], seed=22
)
df.pyrochem.oxides *= 100 # oxides in wt%
df.pyrochem.elements *= 10000 # elements in ppm
In this case, we might want to transform the Ni and Ti into their standard oxide equivalents NiO and TiO2:
df.pyrochem.convert_chemistry(to=["NiO", "TiO2"]).head(2)
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[1235.19505234 1301.09283284 1297.67084898 1332.78130758 1231.28287236
1223.09485893 1213.13011719 1220.86333288 1269.99495233 1193.1419757 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[2388.2200631 2164.62285792 2286.2934774 2443.58980729 2245.75825742
2269.68007197 2130.01547469 2282.00522338 2257.57443799 2515.3361174 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
But here because Ni and Ti have units of ppm, the results are a little non-sensical, especially when it’s combined with the other oxides:
df.pyrochem.convert_chemistry(to=df.pyrochem.list_oxides + ["NiO", "TiO2"]).head(2)
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[1235.19505234 1301.09283284 1297.67084898 1332.78130758 1231.28287236
1223.09485893 1213.13011719 1220.86333288 1269.99495233 1193.1419757 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[2388.2200631 2164.62285792 2286.2934774 2443.58980729 2245.75825742
2269.68007197 2130.01547469 2282.00522338 2257.57443799 2515.3361174 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
There are multiple ways we could convert the units, but here we’re going to first
convert the elemental ppm data to wt%, then perform our oxide-element conversion.
To do this, we’ll use the built-in function scale():
from pyrolite.util.units import scale
df.pyrochem.elements *= scale("ppm", "wt%")
We can see that this then gives us numbers which are a bit more sensible:
df.pyrochem.convert_chemistry(to=df.pyrochem.list_oxides + ["NiO", "TiO2"]).head(2)
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[0.12351951 0.13010928 0.12976708 0.13327813 0.12312829 0.12230949
0.12131301 0.12208633 0.1269995 0.1193142 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[0.23882201 0.21646229 0.22862935 0.24435898 0.22457583 0.22696801
0.21300155 0.22820052 0.22575744 0.25153361]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
Dealing with Units in Column Names
Often our dataframes will start containing column names which pyrolite doesn’t recognize
natively by default (work in progress, this is an item on the roadmap). Here we can
create an example of that, and go through some key steps for using this data in
pyrolite:
df = normal_frame(
columns=["CaO", "MgO", "SiO2", "FeO", "Ni", "Ti", "La", "Lu"], seed=22
)
df.pyrochem.oxides *= 100 # oxides in wt%
df.pyrochem.elements *= 10000 # elements in ppm
df = df.rename(
columns={
**{c: c + "_wt%" for c in df.pyrochem.oxides},
**{c: c + "_ppm" for c in df.pyrochem.elements},
}
)
df.head(2)
If you just wanted to rescale some columns, you can get away without renaming your columns, e.g. converting all of the ppm columns to wt%:
df.loc[:, [c for c in df.columns if "_ppm" in c]] *= scale("ppm", "wt%")
df.head(2)
However, to access the full native capability of pyrolite, we’d need to rename
these columns to use things like convert_chemistry():
units = { # keep a copy of the units, we can use these to map back later
c: c[c.find("_") + 1 :] if "_" in c else None for c in df.columns
}
df = df.rename(
columns={c: c.replace("_wt%", "").replace("_ppm", "") for c in df.columns}
)
df.head(2)
We could then perform our chemistry conversion, rename our columns to include units, and optionally export to e.g. CSV:
converted_wt_pct = df.pyrochem.convert_chemistry(
to=df.pyrochem.list_oxides + ["NiO", "TiO2"]
)
converted_wt_pct.head(2)
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[0.12351951 0.13010928 0.12976708 0.13327813 0.12312829 0.12230949
0.12131301 0.12208633 0.1269995 0.1193142 ]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
/home/docs/checkouts/readthedocs.org/user_builds/pyrolite/checkouts/main/pyrolite/geochem/transform.py:351: FutureWarning: Setting an item of incompatible dtype is deprecated and will raise in a future error of pandas. Value '[0.23882201 0.21646229 0.22862935 0.24435898 0.22457583 0.22696801
0.21300155 0.22820052 0.22575744 0.25153361]' has dtype incompatible with int64, please explicitly cast to a compatible dtype first.
_df.loc[:, targetnames] = subsum.values[:, np.newaxis] @ coeff[np.newaxis, :]
Here we rename the columns before we export them, just so we know explicitly what the units are:
converted_wt_pct = converted_wt_pct.rename(
columns={c: c + "_wt%" for c in converted_wt_pct.pyrochem.list_oxides}
)
converted_wt_pct.head(2)
converted_wt_pct.to_csv("converted_wt_pct.csv")
Total running time of the script: (0 minutes 0.330 seconds)
