Estimation of phytoplankton photosynthetic parameters

Description

The pre-trained PI-estimator can be used to estimate photosynthetic (photosynthesis-irradiance) parameters of phytoplankton in clear to turbid waters.

Input

• Water temperature (T, °C)
• Chlorophyll-a concentration (Chl-a, mg/m3)
• Depth-averaged diffuse attenuation coefficient of PAR (KdPAR, 1/m)
• Light transmittance of downwelling PAR (rPAR, unitless)

Output

• Assimilation number (PBmax, mg C/mg Chl/h)
• Light saturation parameter (Ek, μmol photons/m2/s)

Reference

"Remote estimation of phytoplankton primary production in clear to turbid waters by integrating a semi-analytical model with a machine learning algorithm". Zhaoxin Li, et al. (2022). Remote Sensing of Environment. 113027. 10.1016/j.rse.2022.113027.

Usage

The testing data is available to test the code. The values of Chl-a and KdPAR should be log-transformed.

import pandas as pd
X = pd.read_csv('testing_data.txt', sep = ' ')
for i, ip in enumerate(X.columns):
    if ip in ['Chl-a', 'KdPAR']:
        X.iloc[:, i] = np.log10(X.iloc[:, i])
X = X.to_numpy()

Then, we can restore the pre-trained model to estimate PBmax and Ek:

from PI_estimator import PI_Estimator
Model = PI_Estimator(do_restore = True, restore_path = './Model', model_name = 'PI_Estimator_PBmax_Ek_20220412_v1.0.joblib')
Y = Model.predict(X)
PBmax = Y[:, 0]
Ek = Y[:, 1]

We can also set "n_jobs" to enable parallel computation:

Model.reg_2[-1].n_jobs = -1

(see sklearn.ensemble.RandomForestRegressor)

If you would like to train the PI-estimator using your own data, you can do it by:

Model = PI_Estimator(do_save = True, save_path = '<your path>', model_name = '<your model name>')
Model.fit(X_train, Y_train)

For details, you can see PI_estimator.py.