New paper: Modelling the effect of wet chemistry measurement error on pedotransfer functions

_{Scatter density plot of measured CEC values versus the ‘true’ CEC (with 1:1 line in black)}

A new open access paper titled "Effect of measurement error in wet chemistry soil data on the calibration and model performance of pedotransfer functions", with ISRIC-World Soil Information colleague Cynthia van Leeuwen as main author, was recently published in Geoderma.

Access the paper here

The authors found that measurement error in calibration data reduces model performance of pedotransfer functions (PTFs) to predict the soil's ability to exchange cations, known as cation exchange capacity (CEC), especially when using small calibration datasets. The impact of multiplicative measurement error was evident, as PTFs calibrated with soil samples that had a large measurement error variance showed a greater reduction in their model performance.

Cynthia and her co-authors investigated the effect of additive and multiplicative errors present in the wet chemistry calibration data on the predictions made by multiple linear regression and random forest PTFs.

“The research emphasizes the need for repeated soil measurements for precise quantification of measurement error during routine laboratory analyses”, said Cynthia van Leeuwen, lead author on the paper and PhD candidate at ISRIC and the Soil Geography and Landscape Group at Wageningen University.

The reduction of model performance after fitting PTFs with erroneous data showed that accounting for this error source when calibrating and validating new PTFs is of great importance.

Full citation: van Leeuwen, C. C., Mulder, V. L., Batjes, N. H., & Heuvelink, G. B. (2024). Effect of measurement error in wet chemistry soil data on the calibration and model performance of pedotransfer functions. Geoderma, 442, 116762.

If you are unable to access the full text, email Cynthia van Leeuwen at the email address below.