Premium
Incorporating Spatial Trends and Anisotropy in Geostatistical Mapping of Soil Properties
Author(s) -
Crawford Carol A. Gotway,
Hergert Gary W.
Publication year - 1997
Publication title -
soil science society of america journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.836
H-Index - 168
eISSN - 1435-0661
pISSN - 0361-5995
DOI - 10.2136/sssaj1997.03615995006100010043x
Subject(s) - kriging , variogram , geostatistics , spatial variability , soil science , anisotropy , range (aeronautics) , isotropy , selection (genetic algorithm) , ridge , organic matter , environmental science , mathematics , statistics , geography , cartography , computer science , ecology , biology , physics , materials science , quantum mechanics , artificial intelligence , composite material
Abstract The spatial variation in soil parameters often differs with direction. These differences may occur naturally or may be due to management practices. Regardless of their origin, they present a challenge in geostatistical mapping of soil parameters. Recommendations pertaining to the selection of an appropriate geostatistical method based on the current literature are often incomplete or contradictory. The purpose of this investigation was to provide a unified description, comparison, and discussion of different geostatistical methods for handling trend and anisotropy that may be present in measured soil properties. Soil organic matter content of the 0‐ to 20‐cm depth from a field in continuous ridge‐tilled corn ( Zea mays L.) was used to compare five geostatistical methods: ordinary kriging with an isotropic semivariogram (OKI); ordinary kriging with an anisotropic semivariogram (OKA); ordinary kriging within local neighborhoods (OKN); universal kriging (UK); and median polish kriging (MPK). Organic matter maps produced from the five methods showed similar large‐scale features but marked differences in the finer features. A comparison of percentage of total area in each organic matter range among mapping methods also showed strong similarities; however, the proportion of the field assigned to each range differed by as much as 7%. Larger differences would be expected at large sample spacing. Although the five methods produced similar maps, selection of the “best” technique should be based on selection of an associated model that best accounts for and describes the nature of the cause of the variation.