Title: Inter- and intra-laboratory replicate measurements of soil organic carbon for LUCAS
Description: Analytical soil data from conventional methods is generally assumed to be error free. However, conventional methods also have associated errors due to different laboratory conditions, protocols, operators, instruments and random variation. These factors contribute to variation in analytical measurements from a single laboratory (the intra-laboratory error) as well as between different laboratories (inter-laboratory error).
The relative differences between the intra- and inter-laboratory analytical measurement error depend on the soil property being measured and the protocols and instruments used. Therefore, any methods that seek to replace conventional analytical methods, such as visible near-infrared spectroscopy (VNIRS), need to be assessed in context of these error variances. This allows for a better quantification of the value of VNIRS predictions for subsequent analysis and decision-making.
Breure et al. (2026) provided two different methods in which the intra- and inter-laboratory from LUCAS measurements can be used to interpret soil property predictions, either from VNIRS or other methods. We found that using the laboratory error changed relative differences between models compared to mean prediction metrics commonly used in the literature (e.g. RMSE, R2), given that mean predictions are more dependent on the underlying data distribution and have a higher sensitivity to outliers.
The inter-laboratory replicate dataset: Results from the International Soil-Analytical Exchange Program (WEPAL, 2024), where a set of LUCAS 2009 samples were analysed for SOC in multiple laboratories (a.k.a. ring-test). For each LUCAS sample in the ring-test, the mean and standard deviation between laboratories is given. This dataset was used to quantify the inter-laboratory error variance in Breure et al. (2026).
- Original: Reported SOC content in the LUCAS 2009 survey (g/kg)
- Mean_participants: Mean SOC content measured in ring-test (g/kg)
- SD_participants: Standard deviation of SOC content measured in ring-test (g/kg)
- N_participants: Number of laboratories that participated in the ring-test
The intra-laboratory replicate dataset: Replicate SOC measurements of LUCAS 2009 and 2015 samples during the analytical work on the LUCAS 2015 samples (Jones et al. 2015). This dataset was used to quantify the intra-laboratory error variance in Breure et al. (2026).
- Original: Reported SOC content either in the LUCAS 2009 or 2015 survey (g/kg)
- REP: SOC content for replicate measurement during the LUCAS 2015 analysis (g/kg)
- CaCO3: Reported calcium carbonate content in the LUCAS 2009 or 2015 survey (g/kg)
References:
Breure, T.S., Jones, A., Panagos. P. 2026. Evaluation of visible near-infrared spectroscopy in context of a repeated sampling survey across the European Union. Geoderma, 465: 117647, https://doi.org/10.1016/j.geoderma.2025.117647
Jones, A, Fernández-Ugalde, O., Scarpa, S. LUCAS 2015 Topsoil Survey. Presentation of dataset and results, EUR 30332 EN, Publications Office of the European Union: Luxembourg. 2020, ISBN 978-92-76-21080-1, doi:10.2760/616084, JRC121325.
WEPAL (International Soil-Analytical Exchange Programme). 2024. www.wepal.nl/en/wepal/home/proficiency-tests/soil/ise.htm.
