Uni-Bayreuth grafik-uni-bayreuth



Landform Classification for Functional Upscaling

TERRECO Cluster F-03

Von 09/2012

Projektleiter: Mareike Ließ, Bernd Huwe, Björn Reineking, Soo Jin Park
Mitarbeiter: Gwan Yong Jeong

Abstract: Information on the spatial variability of soils is essential for sustainable land management. Site characteristics such as nutrient availability or drainage determine forest growth and management as well as agricultural production. Digital soil maps (DSM) of the Soyang Lake Watershed shall be developed for important chemical soil parameters. The early conceptual models have resulted in quantitative soil-landscape models, which do not only make the spatial prediction of continuous soil properties possible, but include model uncertainty as an important attribute of soils themselves as the interaction of the various soil processes is still not completely understood.

Supervised learning methods (e.g. random forest, boosting, regression trees or support vector machines) will be used to relate soil properties to environmental predictors. Hence, soil properties of interest at “unknown” locations are predicted by using the relationship between soil and environmental variables at “known” locations under similar environmental conditions to develop digital soil maps (DSM) for the Soyang Lake Watershed. Furthermore, high precision DSM will be generated for a subarea of Soyang due to the additional availability of LIDAR data of 1 m resolution.

Spectrometric soil functions, which relate particular soil properties to VIS/NIR measurements, will be established with a data subset to make the rather time consuming laboratory measurements for the remaining samples dispensable.

Pedodiversity and DSM derived soil functions such as soil fertility, carbon stocks and potentials for ecosystem services will be investigated at various scales from catchment to watershed level to provide environmental models such as LandClim, MMF, PIXGRO, and SWAT with reasonable input values. The soils’ relation to biodiversity and forest structure shall be investigated to understand landscape functions, the impact of land use on soil fertility and ecosystem services.

Key words: digital soil mapping, supervised learning methods, chemical soil properties, soil fertility, site characteristics 


Poster January 2013

Letzte Änderung 18.03.2015