Predicting soil organic carbon stocks in different layers of forest soils in the Czech Republic

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F60460709%3A41210%2F23%3A97370" target="_blank" >RIV/60460709:41210/23:97370 - isvavai.cz</a>

Alternative codes found

RIV/00216224:14310/23:00131498 RIV/00020702:_____/23:N0000078

Result on the web

<a href="https://www.sciencedirect.com/science/article/pii/S2352009423000548" target="_blank" >https://www.sciencedirect.com/science/article/pii/S2352009423000548</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1016/j.geodrs.2023.e00658" target="_blank" >10.1016/j.geodrs.2023.e00658</a>

Result language

angličtina

Original language name

Predicting soil organic carbon stocks in different layers of forest soils in the Czech Republic

Original language description

Carbon dioxide, the most produced anthropogenic greenhouse gas, could be moderated by sequestering carbon in forest soils. Forest soils store more carbon than there is in the atmosphere. Thus, the smallest variation in soil carbon levels could trigger a significant change in atmospheric carbon. This study focused on predicting the spatial distribution of carbon stocks within surface organic and mineral topsoil and subsoil layers of the forty-one natural forest areas of the Czech Republic. Cubist and Random Forests machine learning algorithms were employed with a grid search hyper tuning to improve prediction accuracy. We used the five-fold cross-validation to verify the model accuracy using Root Mean Square Error (RMSE), coefficient of determination (R2), and Mean Absolute Error (MAE). Random Forests yielded lower RMSE of 1.10 kg/m2, 3.85 kg/m2, and 4.77 kg/m2 in the surface organic horizon (F + H layer), mineral topsoil (0-30 cm layers) and subsoil horizons (30-80 cm layers), respectively, compared to the RMSE values of Cubist, which were 1.14 kg/m2, 3.90 kg/m2, and 4.91 kg/m2 in the surface organic, mineral topsoil and subsoil horizons, respectively. R2 values of both models were low for all three horizons considered. Random Forests were the preferred algorithm for SOC stock prediction in all layers of the forest soils. Cubist predicted the spatial distribution of SOC stocks with more covariates than Random Forests. Altitude was the most important covariate for the spatial distribution of SOC stocks for both Random Forests and Cubist in all soil horizons considered. High SOC stocks for all soil horizons are spatially concentrated in soil horizons along the country borders in the mountaineous natural forest areas.

Czech name

—

Czech description

—

Type

J_imp - Article in a specialist periodical, which is included in the Web of Science database

CEP classification

—

OECD FORD branch

40104 - Soil science

Project

Result was created during the realization of more than one project. More information in the Projects tab.

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year

2023

Confidentiality

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

Name of the periodical

GEODERMA REGIONAL

ISSN

2352-0094

e-ISSN

2352-0094

Volume of the periodical

34

Issue of the periodical within the volume

SEP 2023

Country of publishing house

NL - THE KINGDOM OF THE NETHERLANDS

Number of pages

9

Pages from-to

—

UT code for WoS article

001029542300001

EID of the result in the Scopus database

2-s2.0-85161722925