Comparison of geostatistics, machine learning algorithms, and their hybrid approaches for modeling soil organic carbon density in tropical forests

Journal of Soils and Sediments, Год журнала: 2025, Номер unknown

Опубликована: Апрель 5, 2025

Язык: Английский

---

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

Geoderma Regional, Год журнала: 2023, Номер 34, С. e00658 - e00658

Опубликована: Июнь 8, 2023

Язык: Английский

---

Importance of Terrain and Climate for Predicting Soil Organic Carbon Is Highly Variable across Local to Continental Scales

Environmental Science & Technology, Год журнала: 2024, Номер 58(26), С. 11492 - 11503

Опубликована: Июнь 21, 2024

Soil organic carbon (SOC) plays a vital role in global cycling and sequestration, underpinning the need for comprehensive understanding of its distribution controls. This study explores importance various covariates on SOC spatial at both local (up to 1.25 km) continental (USA) scales using deep learning approach. Our findings highlight significant terrain attributes predicting concentration with terrain, contributing approximately one-third overall prediction scale. At scale, climate is only 1.2 times more important than distribution, whereas structural pattern 14 2 vegetation, respectively. We underscore that attributes, while being integral all scales, are stronger predictors scale explicit arrangement information. While this observational does not assess causal mechanisms, our analysis nonetheless presents nuanced perspective about which suggests disparate scales. The insights gained from have implications improved mapping, decision support tools, land management strategies, aiding development effective sequestration initiatives enhancing mitigation efforts.

Язык: Английский

---

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

The Science of The Total Environment, Год журнала: 2023, Номер 883, С. 163677 - 163677

Опубликована: Апрель 25, 2023

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 (SOC) content can reliably be estimated in laboratory conditions, but more cost-efficient and mobile techniques are needed for large-scale monitoring SOC e.g. remote areas. We demonstrate the capability a hyperspectral camera operating visible-near infrared wavelength range practical estimation soil nitrogen content, to support efficient properties. 191 samples were collected Taita Taveta County, Kenya representing an altitudinal gradient comprising five typical use types: agroforestry, cropland, forest, shrubland sisal estate. imaged using Specim IQ under controlled their was determined with combustion analyzer. machine learning estimating N based on spectral images, studying also automatic selection informative wavelengths quantification prediction uncertainty. Five alternative methods all found perform well cross-validated R2 approximately 0.8 RMSE one percentage point, demonstrating feasibility proposed imaging setup computational pipeline.

Язык: Английский

---

Aggregated database of forest soil chemical properties in the Czech Republic based on surveys from 2000 to 2020

Annals of Forest Science, Год журнала: 2024, Номер 81(1)

Опубликована: Апрель 17, 2024

Abstract Key message The dataset includes data from forest soil surveys conducted in the period 2000–2020. It provides and site variables 8269 locations. Data are aggregated three basic layers: upper organic horizon (FH, 6875 locations), mineral layer 0–30 cm (M03, 8051 locations) deeper 30–80 (M38, 2260 locations). is available at https://doi.org/10.5281/zenodo.10608814 , access to metadata https://metadata-afs.nancy.inra.fr/geonetwork/srv/fre/catalog.search#/metadata/38f24573-3c0d-469a-a66a-7060ce082155 .

Язык: Английский

---

Digital soil mapping using machine learning-based methods to predict soil organic carbon in two different districts in the Czech Republic

Soil and Water Research, Год журнала: 2024, Номер 19(1), С. 32 - 49

Опубликована: Янв. 16, 2024

Soil organic carbon (SOC) is an important soil characteristic as well a way how to mitigate climate change. Information on its content and spatial distribution thus crucial. Digital mapping (DSM) suitable evaluate of properties thanks ability obtain accurate information about soil. This research aims apply machine learning algorithms using various environmental covariates generate digital SOC maps for mineral topsoils in the Liberec Domažlice districts, located Czech Republic. The class, land cover, geology terrain extracted from elevation model remote sensing data were used modelling. was predicted based relationships with random forest (RF), cubist, quantile (QRF) models. Results RF showed that cover (vegetation) most variables prediction both districts. had better efficiency accuracy than cubist QRF predict greatest R² value (0.63) observed district model. However, appropriate performance too.

Язык: Английский

---

Assessing and mapping of soil organic carbon at multiple depths in the semi-arid Trans-Ural steppe zone

Geoderma Regional, Год журнала: 2024, Номер 38, С. e00855 - e00855

Опубликована: Авг. 30, 2024

Язык: Английский

---

Assessment of multivariate associations and spatial variability of forest soil properties and their stand factors in the Czech Republic

Soil and Water Research, Год журнала: 2025, Номер unknown

Опубликована: Янв. 3, 2025

Knowing the relationship between forest soil properties and their stand conditions is relevant for sustainable exploitation management of soils. This study examines influence environmental factors on within environments. We further assessed spatial variability these controlling factors. A harmonised database entire areas Czech Republic was considered; however, only 851 sampling points with complete data used out more than 8 thousand in database. The topsoil mineral layer 0–30 cm analysed. Principal component analysis to determine relationships nugget ratios semivariograms cross-variograms were evaluate dependence properties, Forest types reaction availability cations topsoils. Phosphorus influenced by aluminium cation exchange capacity. There are higher concentrations total phosphorus under broadleaved forest.

Язык: Английский

---

Spatial modelling and drivers of soil organic carbon across successional communities in tropical deciduous forests: insights from Northwest Himalayan foothills

Environmental Monitoring and Assessment, Год журнала: 2025, Номер 197(5)

Опубликована: Апрель 7, 2025

Язык: Английский

---

Optimized modelling of countrywide soil organic carbon levels via an interpretable decision tree

Smart Agricultural Technology, Год журнала: 2022, Номер 3, С. 100106 - 100106

Опубликована: Авг. 10, 2022

There are relatively few studies that explicitly evaluate the performance of machine learning algorithms (MLAs) such as decision trees while varying conditions like data splitting strategies and feature selection methods in digital soil mapping (DSM). Since several more powerful black-box models Random Forest (RF) exist, regular Classification Regression Tree (CART) least applied despite being intelligible than former. We demonstrate a simple yet relevant way to improve CART model for DSM still benefiting from its intelligibility, interpretability intuition potential. Soil organic carbon (SOC) levels across Czech Republic predicted at 30 m × resolution using selected covariates coupled with respective models. For this work, 440 topsoils (0–20 cm) were retrieved LUCAS database. Regarding distinct models, (Random, SPlit Conditional Latin Hypercube Sampling: cLHS) 7 varied. Meanwhile, overall results compared accuracy metrics including root mean square error (RMSE). One satisfactory SOC validation based on has (RMSE) 17.30 g/kg coefficient determination (R2) 0.52. The cLHS proves robust splitting. Feature Stepwise (SWR), Recursive Elimination (RFE) Genetic Algorithm (GA) considered computationally efficient identifying covariates. Generally, study demonstrates relevance effectiveness improving modelling via tree (CART).

Язык: Английский

---