A comparison of methods to relate grass reflectance to soil metal contamination
| Autor: | Rob S. E. W. Leuven, Piet H. Nienhuis, Lutgarde M. C. Buydens, Lammert Kooistra, Ron Wehrens |
|---|---|
| Rok vydání: | 2003 |
| Předmět: |
index
Multivariate statistics selection Soil science Lolium perenne Grassland Analytical Chemistry Laboratory of Geo-information Science and Remote Sensing Partial least squares regression Laboratorium voor Geo-informatiekunde en Remote Sensing river floodplains Hydrology geography geography.geographical_feature_category Radiometer biology leaf Molecular Materials Vegetation Contamination biology.organism_classification PE&RC Soil contamination field Environmental science General Earth and Planetary Sciences band Environmental Sciences |
| Zdroj: | International Journal of Remote Sensing, 24, 24, pp. 4995-5010 International Journal of Remote Sensing, 24(24), 4995-5010 International Journal of Remote Sensing, 24, 4995-5010 International Journal of Remote Sensing 24 (2003) 24 |
| ISSN: | 1366-5901 0143-1161 |
| DOI: | 10.1080/143116031000080769 |
| Popis: | Grass-dominated vegetation covers large areas of the Dutch river floodplains. Remotely sensed data on the conditions under which this vegetation grows may yield information about the degree of soil contamination. This paper explores the relationship between grassland canopy reflectance and zinc (Zn) contamination in the soil under semi-field conditions. A field radiometer was used to record reflectance spectra of perennial ryegrass (Lolium perenne) in an experimental field with Zn concentrations in the soil ranging from 32 to 1800 mg kg-1. Several spectral vegetation indices (VIs) and a multivariate approach using partial least squares (PLS) regression were investigated to evaluate their potential use in estimating Zn contamination levels. Compared to the best PLS model (RMSEP=181.4 mg kg-1), the narrow band vegetation index MSAVI2mm performed better (RMSEP=162.9 mg kg-1). Both MSAVI2mm and PLS gave a high user accuracy for the strongly contaminated soil class (100% and 91%, respectively), while the total accuracy was satisfactory (60% and 55%, respectively). Results from this feasibility study indicate the potential of using remote sensing techniques for the classification of contaminated areas in river floodplains. But as the results from this study may be both resolution- and location-dependent, research on field and image scale is now required to test the established relations and to assess their susceptibility to seasonal influences, species heterogeneity, and increased levels of spectral noise. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|