Multivariate soft-modeling to predict radiocesium soil-to-plant transfer

Area of Science:

• Environmental Science
• Soil Science
• Radiochemistry

Background:

• Radiocesium contamination in agricultural soils poses risks due to Chernobyl accident fallout.
• Understanding soil-to-plant transfer is vital for managing radionuclide uptake in food chains.
• Previous models often lack comprehensive validation across diverse soil types and treatments.

Purpose of the Study:

• To develop and validate a multivariate soft-modeling approach for estimating radiocesium transfer to grass.
• To identify key soil characteristics influencing radiocesium bioavailability and plant uptake.
• To provide a robust predictive tool for assessing radionuclide contamination in agricultural environments.

Main Methods:

• Utilized principal component analysis (PCA) for exploratory data analysis of 145 soil samples.
• Employed partial least squares regression (PLS) to build a multivariate prediction model.
• Validated the model using a split-data approach (calibration and prediction sets) with 21 soil and plant parameters.

Main Results:

• PCA revealed distinct groupings of soil samples based on field plots and agricultural treatments.
• PLS model identified significant soil parameters influencing radiocesium transfer, including phyllosilicate content and ammonium (NH4+) status.
• The developed model demonstrated satisfactory prediction accuracy for radiocesium soil-to-plant transfer.

Conclusions:

• The multivariate PCA-PLS approach effectively models radiocesium transfer to grass.
• Key soil properties like mineralogy and nutrient status are critical determinants of radiocesium bioavailability.
• The validated model offers a reliable tool for assessing and managing radiocesium contamination in agricultural soils.