Hyperspectral close-range and remote sensing of soils and related plant associations – Spectroscopic applications in the boreal environment

Maarit Middleton

    Research output: ThesisDoctoral ThesisCollection of Articles

    Abstract

    Hyperspectral close-range and remote sensing techniques have been available to the research community since the 1980’s but applications have focused on forestry and land use. The objective of the study was to explore relevant applications of visible and short wavelength infrared spectroscopy (350−2500 nm) for detection of physical and chemical properties of glacial till soils and plant species communities related to the soil properties in the boreal environment of northern Finland. Empirical single and multivariate regression techniques (MVR) were applied for predicting glacial till soil dielectric permittivity (ε, i.e. soil moisture) and till elemental concentrations from close-range spectrometry. Predictive kernel and neural network based fuzzy classification approaches were applied for classification of data acquired with AISA and HyMap airborne imaging spectrometers. Ordination techniques were used for revealing plant community structures and optimizing the thematic class hierarchal level. The till soil ε was well predicted from VSWIR spectra with the exponential single-spectral variate but also with MVR techniques. The most accurate results were gained with relevance vector machines. Prediction of till soil chemical element concentrations of Al, Ba, Co, Cr, Cu, Fe, Mg, Mn, Ni, V, and Zn was also statistically valid. Soil moisture based site suitability for Scots pine (Pinus sylvestris) from imaging spectroscopic data was moderately successful as the highest area under the receiver operating characteristics curve (AUC) value was 0.741. Site type mapping of aapa peatlands with support vector machines was highly successful with AUC values 0.946−0.999 for bog, sedge fen, and eutrophic fen. Understanding the ε-reflectance relationship would be evident when artificial regeneration to Scots pine, intolerant of wet soils, is considered on clear-cuts with high soil moisture variability. The site suitability on site prepared forest compartments could be predicted using exposed soil pixels in high spatial resolution imagery but also with indirect imaging of soil moisture through understory species patterns. The high success of the peatland site type mapping was attributed to optimization of class hierarchal levels with a constrained ordination based approach which was used to test the spectral and ecological class separability prior to classification. These novel applications of imaging spectroscopic data can readily be applied in practice once cost-effective satellite based data is available. Further research is required to make the close-range spectroscopy operational for quantification of element concentrations to serve forest soil research and mineral potential mapping.
    Translated title of the contributionMaaperän ja siihen liittyvien kasviyhdyskuntien hyperspektrinen etätunnistus ja kaukokartoitus - Spektroskopian sovelluksia boreaalisessa ympäristossä
    Original languageEnglish
    QualificationDoctor's degree
    Awarding Institution
    • Aalto University
    Supervisors/Advisors
    • Haggren, Henrik, Supervising Professor
    • Sutinen, Raimo, Thesis Advisor, External person
    Publisher
    Print ISBNs978-952-217-281-5
    Electronic ISBNs978-952-217-282-2
    Publication statusPublished - 2014
    MoE publication typeG5 Doctoral dissertation (article)

    Keywords

    • remote sensing
    • airborne methods
    • hyperspectral analysis
    • spectroscopy
    • dielectric properties
    • chemical elements
    • multivariate analysis
    • till
    • biotopes
    • vegetation
    • forest soils
    • peatlands
    • classification

    Fingerprint

    Dive into the research topics of 'Hyperspectral close-range and remote sensing of soils and related plant associations – Spectroscopic applications in the boreal environment'. Together they form a unique fingerprint.

    Cite this