138-6 Digital Soil Mapping Using Geomorphon As a Predictive Terrain Attribute.

Poster Number 922

See more from this Division: SSSA Division: Pedology
See more from this Session: Scaling Soil Processes and Modeling: II (includes student competition)
Monday, November 3, 2014
Long Beach Convention Center, Exhibit Hall ABC
Share |

Helena Pinheiro1, Phillip R. Owens2, Lucia Helena Cunha dos Anjos3, Cesar da Silva Chagas4 and Waldir de Carvalho Júnior4, (1)Purdue University, West Lafayette, IN
(2)915 W. State St, Purdue University, West Lafayette, IN
(3)Universidade Federal Rural do Rio de Janeiro, Seropédica, Brazil
(4)Embrapa Solos, Rio de Janeiro, Brazil
Poster Presentation
  • ASA_PosterHelena_102614_JMAproedits_Rev.pdf (2.1 MB)
  • Abstract: Soils and related properties repeat across landscapes as patterns related to topography. The goal was to evaluate the GRASS Add on- Geomorphon tool as a terrain co-variate for digital soil mapping. Geomorphons is a novel method for classification and mapping of landform elements from DEM based on the principle of pattern recognition rather than differential geometry. To generate landforms with Geomorphon analyses are necessary values of two parameters, which are defined as search radius (L), and relief threshold (d). Thereby, Geomorphons are calculated as from flexible procedure that making possible recognition of the same types of landforms in different sizes. The digital mapping co-variate was evaluated to predict soil classes in Guapi-Macacu watershed, in Rio de Janeiro State (Brazil), through supervised classification applying Artificial Neural Networks (ANNs). Models were generated for continuous landscape attributes related to pedogenesis, such as altimetry, slope, curvature, combined topographic index, euclidean distance, clay minerals, iron oxide, normalized difference vegetation index, geology and different sizes of Geomorphons. The predominant soils were: Oxisols, Ultisols, Inceptisols, Aquents and other Entisols. The appropriated Geomorphon was selected considering the spatial resolution of the other attribute models, reference literature, statistical indexes and consistency with the scale of presentation for the final soil map generated with Artificial Neural Networks (ANNs). The approach was based on pedogenic knowledge for choosing attributes that represent the variability of the soil forming factors in the area. To accomplish the objectives, there were sixteen different sets of ANNs, which contained all of the terrain attributes, but with a Geomorphon calculated from different sizes of search radius. For comparative purposes about the influence of Geomorphon in pattern recognition, one of the sets included no Geomorphon. The results showed that the best classification was obtained from the set in which  a Geomorphon was used that had a search radius equivalent to forty five pixels.

    Keywords: Digital Soil Mapping, Geomorphology, Terrain Patterns

    See more from this Division: SSSA Division: Pedology
    See more from this Session: Scaling Soil Processes and Modeling: II (includes student competition)