Tropical altitudinal gradient soil organic carbon and nitrogen estimation using Specim IQ portable imaging spectrometer

The largest actively cycling terrestrial carbon pool, soil, has been disturbed during the last centuries by human actions through decreasing woody land cover. Soil organic carbon (SOC) content can reliably be estimated in laboratory conditions, but more cost-efficient and mobile techniques are needed for large-scale monitoring of SOC e.g. in remote areas. We demonstrate the capability of a mobile hyperspectral camera operating in the visible-near infrared wavelength area for practical estimation of soil organic carbon (SOC) and nitrogen content, to support efficient monitoring of soil properties. The 191 soil samples were collected in Taita Taveta County, Kenya representing an altitudinal gradient comprising of five typical land use types: agroforestry, cropland, forest, shrubland and sisal estate. The soil samples were imaged using Specim IQ hyperspectral camera in controlled laboratory conditions, while the carbon content of was determined with a dry-oxidization analyzer. We empirically evaluate five multivariate regression methods, two of which are novel in this context, and validate a technical inference pipeline to provide reliable estimates of SOC. We investigate different ways of using the recorded spectral images for training machine learning models for estimating SOC and nitrogen content, covering both alternative strategies of extracting spectral information from the image as well as different predictive methods. Of the five multivariate regressors, two were novel in the context of SOC prediction and enable uncertainty quantification and band selection. Three of the regressors in effect performed equally well with cross-validated R2 scores of .8 and RMSE of 1% SOC. This demonstrates the feasibility of the proposed imaging configuration and computational pipeline in estimating SOC and justifies future research. The carbon and nitrogen content of the soil samples correlated strongly. Both variables were increasing together with the altitude indicating dependency on the climatic conditions from the hot lowlands to the cool highlands. In addition, the SOC and N increased with tree canopy height, canopy density and above ground biomass indicating the impact of the land cover to the soil variables.The largest actively cycling terrestrial carbon pool, soil, has been disturbed during latest centuries by human actions through reduction of woody land cover. Soil organic carbon (SOC) content can reliably be estimated in laboratory conditions, but more cost-efficient and mobile techniques are needed for large-scale monitoring of SOC e.g. in remote areas. We demonstrate the capability of a mobile hyperspectral camera operating in the visible-near infrared wave-length range for practical estimation of soil organic carbon (SOC) and nitrogen content, to support efficient monitoring of soil properties. The 191 soil samples were collected in Taita Taveta County, Kenya representing an altitudinal gradient comprising five typical land use types: agroforestry, cropland, forest, shrubland and sisal estate. The soil samples were imaged using a Specim IQ hyperspectral camera under controlled laboratory conditions, and their carbon and nitrogen content was determined with a combustion analyzer. We use machine learning for estimating SOC and N content based on the spectral images, studying also automatic selection of informative wavelengths and quantification of prediction uncertainty. Five alternative methods were all found to perform well with a cross-validated R-2 of approximately 0.8 and an RMSE of one percentage point, demonstrating feasibility of the proposed imaging setup and computational pipeline.Non peer reviewe