Mid-infrared spectroscopy outperforms visible–near infrared spectroscopy for predicting soil stable carbon isotopes

Published: 8 July 2026| Version 1 | DOI: 10.17632/bskdgjk2nt.1
Contributor:
Gizachew Tiruneh

Description

This dataset contains soil physicochemical, stable isotope, and spectroscopic data collected from five Amazonian floodplain ecosystems on Ilha das Cinzas, Amazon River estuary, Brazil: high-floodplain forest, low-floodplain forest, regenerated forest, mangrove forest, and açaí agroforestry. Soil samples were collected from 0–2 m depth and analyzed for total carbon (TC), total nitrogen (TN), stable carbon (δ¹³C) and nitrogen (δ¹⁵N) isotopic compositions, and complementary soil properties. The dataset also includes visible–near-infrared–shortwave infrared (Vis–NIR–SWIR; 350–2500 nm) and mid-infrared (MIR; 4000–650 cm⁻¹) spectral measurements, together with calibration and validation results from Partial Least Squares Regression (PLSR), Random Forest (RF), and ensemble models used to predict soil isotopic composition. Supporting files include spectral variables, model performance metrics (R², RMSE, bias, and RPIQ), and Random Forest feature importance analyses. The dataset supports research on soil carbon sequestration, nutrient cycling, digital soil assessment, spectroscopy, machine learning, and sustainable management of Amazonian floodplain ecosystems.

Files

Categories

Reflectance Modeling, Carbon Isotope, Isotope Biogeochemistry, Absorption Spectra

Funders

Licence