Sensor fusion and fitness-for-purpose screening for SOC prediction in vis–NIR spectral libraries, insights from vis–NIR plus XRF, local calibration, and MRV thresholds

- 336951
Posters
Favorite this paper
How to cite this paper?
Abstract

Soil organic carbon (SOC) monitoring requires analytical tools that are both scalable for measurement, reporting and verification process (MRV). We assessed whether fusing X-ray fluorescence (XRF) information with vis–NIR spectra improves SOC prediction when models are trained on large spectral libraries and transferred to local samples, and we translated model performance into practical recommendations using a fitness-for-purpose framework. A national spectral library of was used to build a global Cubist model, while nine local spectral libraries supported local partial least squares models. Two sensing scenarios were compared, vis–NIR alone and vis–NIR+XRF data fusion. Overall accuracy gains from sensor fusion were modest under global modelling but more evident under local modelling. Globally, vis–NIR+XRF achieved RMSE of 1.56 g kg⁻¹ versus 1.57 g kg⁻¹ for vis–NIR. Locally, fusion reduced RMSE from 1.13 to 0.95 g kg⁻¹ and decreased the error profile. Fitness-for-purpose evaluation revealed that suitability depends on within-farm SOC variability. All sensing scenarios were not recommended in low-variability farms. In medium variability farms, vis–NIR with local models achieved a “use with caution” classification, while fusion using local model was recommended. In high variability farms, all scenarios were recommended regardless of calibration strategy. For temporal monitoring, the small annual SOC change leads to low analytical tolerance, making sensor-based inference unsuitable over short windows. Temporal monitoring only becomes acceptable from 14 years onward and only for high sequestration potential under local calibration. These results clarify when sensor fusion and modelling strategy translate into actionable SOC monitoring services helping to support adoption.

Share your ideas or questions with the authors!

Did you know that the greatest stimulus in scientific and cultural development is curiosity? Leave your questions or suggestions to the author!

Sign in to interact

Have a question or suggestion? Share your feedback with the authors!

Institutions
  • 1 ESALQ SP, “Luiz de Queiroz” College of Agriculture, Center for Carbon Research in Tropical Agriculture, University of São Paulo (CCARBON/USP), Piracicaba, SP, Brazil
  • 2 ESALQ - USP
  • 3 ESALQ/USP
Track
  • Monitoring, Reporting and Verification (MRV) protocols
Keywords
SOC monitoring
sensor-based inference
fitness-for-purpose interpretation
data fusion