Spatial prediction of forest soil properties using digital soil mapping approaches

Vincent Yaw Oppong Sarkodie, MSc

Disertační práce

Abstract:

Abstract Forests are estimated to cover about one-third of the landmass for the European continent. These forests have been established on areas such as steep slopes, relatively shallow and/or stony or very wet landscapes. Forests on the continent, including those of Czechia, suffered atmospheric acid pollution from sulphur and nitrogen compounds discharged from anthropogenic activities of the continent before the end of the 20th century. Interventions on Czech atmospheric acid deposition, for example, and many studies have only focused on the most affected forest areas. Forest soil characteristics are influenced by their environment, thus reliable knowledge on forest soil environment’s influence is important for their sustainable management. This thesis is focused on predicting forest soil properties of Czech forest areas. The thesis employs digital soil mapping approaches, which resolves the difficulty of acquiring detailed soil information in required structure, spatial density and in some instances non-availability. A database of soil samples collected between the years 2000 and 2021 from coniferous, broadleaved and mixed forest stands from several surveys conducted across the entire forest areas of the Czech Republic was used. A comparative test was conducted to harmonize the data from different determination methods. Random forest, cubist and random forest - kriging were employed for the spatial distribution predictions. Principal component analysis, kriging, semivariograms and cross variograms were utilised for assessing multivariate associations and spatial variability of forest soil properties and their predictors. The predicted maps were developed in R-studio and ArcGIS. Models developed for the spatial predictions of soil properties in this study were validated for their accuracy using root mean square error (RMSE), mean absolute error (MAE), ratio of performance to inter-quartile distance (RPIQ) matrix, and for their explained variance with the coefficient of determination (R2) and Akaike information criterion (AIC) matrix. It was found that the influence environmental factors have on SOC stock prediction is determined by forest region and altitude. High SOC stocks for the organic, mineral topsoil and subsoil horizons are spatially concentrated along the mountainous border regions of Czechia. The Doupovské Mts. showed the highest predicted values for the pseudototal content of Ca, Mg and K. Forest types have a greater influence on soil reaction and base saturation.

Keywords

forest soil, digital soil mapping, environmental covariates, soil properties, soil organic carbon

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 4. 1. 2025

Obhajoba závěrečné práce

• Vedoucí: prof. Dr. Ing. Luboš Borůvka
• Oponent: Radim Vácha, externi, Martin Valtera, externi, Piotr Gruba, externi

Česká zemědělská univerzita v Praze

Fakulta agrobiologie, potravinových a přírodních zdrojů

Doktorský studijní program / obor:
Agricultural Specialization / Exploitation and Protection of Natural Resources