Hierarchical data fusion for mapping soil units at field scale

• Published in: Geoderma Vol. 112; no. 3; pp. 179 - 196
• Main Authors: Sommer, M, Wehrhan, M, Zipprich, M, Weller, U, zu Castell, W, Ehrich, S, Tandler, B, Selige, T
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Amsterdam Elsevier B.V 01.03.2003; Elsevier
• Subjects: Areal geology. Maps; Earth sciences; Earth, ocean, space; EM38; Exact sciences and technology; Expert systems; Geologic maps, cartography; Relief analysis; Remote sensing; Soil pattern; Soils; Surficial geology; Central Europe; Germany; Soil pattern; Relief analysis; EM38; Expert systems; Remote sensing; models; biomass; expert systems; maps; reflectance; Europe; multispectral remote sensing; digital elevation models; accuracy; remote sensing; vegetation; image analysis; electromagnetic induction; scanner methods; airborne methods; radiometry; alluvium; relief; electrical conductivity; soils; Pixel
• ISSN: 0016-7061; 1872-6259
• DOI: 10.1016/S0016-7061(02)00305-1

We analyzed a highly complex soilscape of fluvial sediments by a hierarchical expert system. Using (i) inquiries, (ii) relief analysis on basis of a DEM 5, and (iii) soils' apparent electrical conductivity (EM38) as a database, we first defined zones of identical pedogenic context. Next, multi-temporal remote sensing data of winter wheat were obtained by an airborne multi-spectral scanner (Daedalus-ATM), which gives radiometric information with a geometric (ground) resolution of 1 m 2 (pixel size). Leaf area index (LAI) was semi-physically modeled using red and near-infrared canopy reflectances and related to above-ground biomass. Further, the resulting spatial patterns of vegetation parameters were processed by image analysis methods, i.e. an opening–closing procedure using a circular element with a radius of 5 m. These coarser patterns of LAI and biomass, respectively, were interpreted as patterns of site quality within each zone of pedogenic context. By our multi-temporal approach we were able to distinguish between stationary and time-variant pattern. Combined with point calibration on basis of a 50-m raster we identified available water capacity (AWC) and O 2 deficiency due to stagnant water as the most important soil properties constituting site quality for plant growth. Our results will be used for precision agriculture practices in future.