Variability of physical, chemical and hydraulic parameters in soils of Mt. Kilimanjaro across different land uses

The heterogeneity of soils is a key to biological processes, carbon turnover and water storage. Climate change and anthropogenic land use changes often decrease important functioning of soil systems. In order to understand the soil complexity, sophisticated technologies to measure soil properties and processes at high spatial resolutions are needed. This study addresses the heterogeneity of various soil physical, chemical and hydraulic properties in different land uses at Mt. Kilimanjaro and its implications for carbon and water storage of the soils. The objectives were to apply visible to near infrared diffuse reflectance spectroscopy in-situ to gain information about the intact soil. The first three studies address the challenges that arise from in-situ spectral measurements. We could show that regression models calibrated with a regional spectral database fail to predict clay and carbon content from in-situ soil spectra. The incorporation of additional samples that were scanned in the field into a spectral calibration database increased prediction accuracies. As the collection of soil information with traditional laboratory methods is time demanding, often only a limited amount of additional samples are available. Therefore, we used the synthetic minority oversampling technique and demonstrated its potential for generating new soil spectra, which can be used to balance a calibration database regarding in-situ spectral characteristics. Including these new spectra into calibration models improved prediction accuracies substantially. Based on these findings we propose a framework for modelling with limited in-situ spectra. Consequently we applied this framework for the prediction of soil organic carbon, nitrogen, clay, silt and sand content in soil profiles in a high spatial resolution with predicted values every 3 cm. Soil hydraulic properties lack a direct physical basis, which could be reflected in the soil spectra. Furthermore, the creation of a spectral database would require a huge effort. Therefore, we used pedotransfer functions to predict soil hydraulic properties in the soil profiles. Soils of the four different land uses, homegarden (a traditional agroforestry system at Mt. Kilimanjaro), coffee plantation, maize field and savannah were thus thoroughly described regarding physical, chemical and hydraulic parameters. Soil heterogeneity of the less intensively managed land uses, homegarden and savannah was much higher, than those of coffee plantation and maize field. With our sampling design, it is unfortunately difficult to differentiate between pedogenic and land use effects on the soil properties. However, with our proposed framework, additional soil information can be derived rapidly to characterise further sites. In a further study, using basic soil properties derived from spectral measurements, we could improve spatial predictions for microbial biomass in two savannah plots at a spatial scale of several meters. Soil microbial biomass was strongly related to organic C and N content of the soil. We could further show, that its spatial distribution is related to vegetation and surface morphology. In order to better understand implications of land use changes for the whole Mt. Kilimanjaro, further research regarding the soils of the forest zones is needed. As spectral characteristics of the volcanic soils in the forest differ from those in the lower zones of the mountain, these should be studied comprehensively. A characterisation of the water and carbon storage potential of the different soils of Mt. Kilimanjaro would then be possible. We conclude, that by integrating visible to near infrared spectroscopy into additional prediction methods like geostatistics or pedotransfer functions, various soil physical, chemical, biological and hydraulic parameters can be derived rapidly and accurately.

Level thesis:degree_level

thesis.doctoral

Grantor dc:publisher

Universität Bayreuth

Year

2015

Author dc:creator

• Kühnel, Anna

Contributors dc:contributor

• Huwe, Bernd

Repository record source_url

https://epub.uni-bayreuth.de/id/eprint/2071/

OAI identifier oai:identifier

oai:epub.uni-bayreuth.de:2071