Vertical distribution and variability of soil organic carbon and CaCO3 in deep Colluvisols modeled by hyperspectral imaging

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00027049%3A_____%2F25%3AN0000013" target="_blank" >RIV/00027049:_____/25:N0000013 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.sciencedirect.com/science/article/pii/S0016706124003756" target="_blank" >https://www.sciencedirect.com/science/article/pii/S0016706124003756</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.geodrs.2024.e00874" target="_blank" >10.1016/j.geodrs.2024.e00874</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Vertical distribution and variability of soil organic carbon and CaCO3 in deep Colluvisols modeled by hyperspectral imaging

Popis výsledku v původním jazyce

The acceleration of soil erosion in undulating landscapes due to human activities has led to a larger area of land being affected by intensive soil redistribution. Colluvisols, sedimentary soils formed on concave slope positions, are considered to be important indicators of soil-landscape processes and soil organic carbon pools. In this study, we investigated the effectiveness of hyperspectral imaging in visible and near-infrared range to assess the detailed variability (both vertical and within each colluvial layer and in-situ soil horizon) of soil organic carbon (SOC) and CaCO3 concentrations in three deep Colluvisols developed on loess and located at different slope positions in southeast Czechia, and evaluate whether this in-detail mapped microvariability can be used as a proxy to assess the dynamics and history of colluvial sedimentation. A variety of nonlinear machine learning techniques such as cubist regression tree (Cubist), random forest (RF), support vector machine regression (SVMR) and one linear technique partial least square regression (PLSR) were compared to determine the most suitable model for the prediction of SOC and CaCO3 content in each profile. The results showed that RF provided the best performance for both SOC (R2 = 0.75) and CaCO3 (R2 = 0.76) contents. The maps depict significant differences in the vertical variability of the predicted properties in the profiles depending on the different intensity, form and period of sedimentation resulting from the slope position. The within-horizon/layer variability of SOC proves to be a suitable indicator of the character of deposition. High variability has been shown mainly in the medieval layers, where it reflects high-energy material redeposition, while low variability in the oldest and youngest parts of the profiles is probably associated with the type of deposited material and frequent pedoturbation, respectively. The within-horizon/layer variability of CaCO3, on the other hand, is independent of the dynamics of deposition. The study showed that imaging spectroscopy is a suitable tool to capture the detailed pattern of the colluvial matrix and, with appropriate sampling and processing, is applicable even in very deep soil profiles.

Název v anglickém jazyce

Vertical distribution and variability of soil organic carbon and CaCO3 in deep Colluvisols modeled by hyperspectral imaging

Popis výsledku anglicky

The acceleration of soil erosion in undulating landscapes due to human activities has led to a larger area of land being affected by intensive soil redistribution. Colluvisols, sedimentary soils formed on concave slope positions, are considered to be important indicators of soil-landscape processes and soil organic carbon pools. In this study, we investigated the effectiveness of hyperspectral imaging in visible and near-infrared range to assess the detailed variability (both vertical and within each colluvial layer and in-situ soil horizon) of soil organic carbon (SOC) and CaCO3 concentrations in three deep Colluvisols developed on loess and located at different slope positions in southeast Czechia, and evaluate whether this in-detail mapped microvariability can be used as a proxy to assess the dynamics and history of colluvial sedimentation. A variety of nonlinear machine learning techniques such as cubist regression tree (Cubist), random forest (RF), support vector machine regression (SVMR) and one linear technique partial least square regression (PLSR) were compared to determine the most suitable model for the prediction of SOC and CaCO3 content in each profile. The results showed that RF provided the best performance for both SOC (R2 = 0.75) and CaCO3 (R2 = 0.76) contents. The maps depict significant differences in the vertical variability of the predicted properties in the profiles depending on the different intensity, form and period of sedimentation resulting from the slope position. The within-horizon/layer variability of SOC proves to be a suitable indicator of the character of deposition. High variability has been shown mainly in the medieval layers, where it reflects high-energy material redeposition, while low variability in the oldest and youngest parts of the profiles is probably associated with the type of deposited material and frequent pedoturbation, respectively. The within-horizon/layer variability of CaCO3, on the other hand, is independent of the dynamics of deposition. The study showed that imaging spectroscopy is a suitable tool to capture the detailed pattern of the colluvial matrix and, with appropriate sampling and processing, is applicable even in very deep soil profiles.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

40104 - Soil science

Projekt

—

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Geoderma

ISSN

0016-7061

e-ISSN

1872-6259

Svazek periodika

453

Číslo periodika v rámci svazku

leden 2025

Stát vydavatele periodika

NL - Nizozemsko

Počet stran výsledku

15

Strana od-do

117146

Kód UT WoS článku

001403135200001

EID výsledku v databázi Scopus

2-s2.0-85213241696