A Moroccan soil spectral library use framework for improving soil property prediction: Evaluating a geostatistical approach

Summary

This research developed a spatially optimised Moroccan Soil Spectral Library (MSSL) using stratified spatially balanced sampling across environmental covariates and FAO soil units. The authors evaluated multiple calibration sample selection strategies—including spatial autocorrelation of spectral principal component scores, spectra similarity memory-based learning, and environmental covariate clustering—to optimise predictions of soil properties from NIR and MIR spectroscopy. The spatial calibration sample selection approach demonstrated distinct precision improvements, offering a practical framework for leveraging large spectral libraries for farmland-specific soil property prediction without requiring costly local calibration.

UK applicability

The methodological framework for spatially optimised spectral library development and calibration sample selection is potentially transferable to UK soil contexts, particularly where national or regional soil spectral libraries exist. However, direct applicability of the Moroccan calibration models would be limited; UK adoption would require developing equivalent spectral libraries calibrated to UK soil types, environmental conditions, and spectrometer configurations.

Key measures

Prediction accuracy of twelve soil properties using NIR and MIR spectral ranges; spatial autocorrelation of principal component scores; calibration sample selection criteria performance

Outcomes reported

The study evaluated calibration sample selection strategies for a Moroccan Soil Spectral Library (MSSL) to predict twelve soil properties using NIR and MIR spectroscopy. Distinct precision improvements were observed from spatial autocorrelation-guided calibration sample selection compared to alternative methods.

Topic tags