Farm-scale calibrations of low-cost spectrometer data for in situ predictions of soil moisture and soil organic carbon

Soil organic carbon (SOC) is a key indicator of soil quality and contributes to mitigating atmospheric carbon dioxide. In situ soil spectrometry enables fast, inexpensive, and non-destructive SOC assessment, for example to monitor SOC accumulation in regenerative farming. Therefore, three miniaturized, portable, low-cost spectrometers (400–2400 nm) were tested both in the laboratory and under field conditions. Field-independent calibration models were developed to predict soil moisture (SM) and SOC directly on a farm. As a benchmark, performance was compared against a high-resolution benchtop spectrometer (325–2450 nm). In total, 74 samples, including pseudo-replicates with varying soil moisture and nutrients concentrations, were spectrometrically analysed, yielding 1003 spectra, about half acquired in situ. Results from the lab tests show that the three low-cost spectrometers accurately predicted SM (R 2 = 0.97, RMSEP = 1.5 %) and SOC (R 2 = 0.86, RMSEP = 0.21 %) while the high-resolution spectrometer performed only slightly better (SM: R 2 = 0.99, RMSEP = 1.0 %; SOC: R 2 = 0.89, RMSEP = 0.19 %). Also, during field application, the low-cost spectrometers achieved good predictions (SM: R 2 = 0.95, RMSEP = 0.96 %; SOC: R 2 = 0.85, RMSEP = 0.14 %). On average, best predictions were achieved using support vector machines (SVM) and combinatorial ratio (CoRa) transformations. In contrast, Kubelka-Munk-transformations and artificial neural networks (ANN) performed significantly worse. Moreover, external parameter orthogonalization (EPO) was tested in order to remove the disturbing effect from moisture variations. However, EPO did not lead to the expected improvements in prediction accuracy. Finally, the results indicate that one of the low-cost NIR spectrometers could potentially be omitted without substantially compromising prediction performance, which would be associated with a reduction in measurement efforts and equipment costs.

Authors

Institutions

Publication Details

Journal
Computers and Electronics in Agriculture
Published
2026-11-16
DOI
https://doi.org/10.1016/j.compag.2026.112655
Primary Topic
Soil Geostatistics and Mapping
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Farm-scale calibrations of low-cost spectrometer data for in situ predictions of soil moisture and soil organic carbon

Michael Horf, Sebastian Vogel
Computers and Electronics in Agriculture
Soil Geostatistics and Mapping
article

Farm-scale calibrations of low-cost spectrometer data for in situ predictions of soil moisture and soil organic carbon

Michael Horf, Sebastian Vogel
article en

Abstract

Soil organic carbon (SOC) is a key indicator of soil quality and contributes to mitigating atmospheric carbon dioxide. In situ soil spectrometry enables fast, inexpensive, and non-destructive SOC assessment, for example to monitor SOC accumulation in regenerative farming. Therefore, three miniaturized, portable, low-cost spectrometers (400–2400 nm) were tested both in the laboratory and under field conditions. Field-independent calibration models were developed to predict soil moisture (SM) and SOC directly on a farm. As a benchmark, performance was compared against a high-resolution benchtop spectrometer (325–2450 nm). In total, 74 samples, including pseudo-replicates with varying soil moisture and nutrients concentrations, were spectrometrically analysed, yielding 1003 spectra, about half acquired in situ. Results from the lab tests show that the three low-cost spectrometers accurately predicted SM (R 2 = 0.97, RMSEP = 1.5 %) and SOC (R 2 = 0.86, RMSEP = 0.21 %) while the high-resolution spectrometer performed only slightly better (SM: R 2 = 0.99, RMSEP = 1.0 %; SOC: R 2 = 0.89, RMSEP = 0.19 %). Also, during field application, the low-cost spectrometers achieved good predictions (SM: R 2 = 0.95, RMSEP = 0.96 %; SOC: R 2 = 0.85, RMSEP = 0.14 %). On average, best predictions were achieved using support vector machines (SVM) and combinatorial ratio (CoRa) transformations. In contrast, Kubelka-Munk-transformations and artificial neural networks (ANN) performed significantly worse. Moreover, external parameter orthogonalization (EPO) was tested in order to remove the disturbing effect from moisture variations. However, EPO did not lead to the expected improvements in prediction accuracy. Finally, the results indicate that one of the low-cost NIR spectrometers could potentially be omitted without substantially compromising prediction performance, which would be associated with a reduction in measurement efforts and equipment costs.

Computers and Electronics in AgricultureVol. 257
Leibniz Institute for Agricultural Engineering and Bioeconomy (DE)
Bundesministerium für Ernährung und Landwirtschaft
Zero hunger, Life on land
Soil Geostatistics and Mapping
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon