Premium
Modeling Temporal Changes in Structural Stability of a Clay Loam Soil
Author(s) -
Caron J.,
Kay B. D.,
Kachanoski R. G.,
Stone J. A.
Publication year - 1992
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/sssaj1992.03615995005600050043x
Subject(s) - loam , soil water , predictability , water content , soil science , environmental science , stability (learning theory) , mathematics , statistics , geology , geotechnical engineering , machine learning , computer science
Field measurements of short‐term changes in structural stability often include an important temporal variability that tends to mask treatment effects. Adequate modeling of this temporal variability is needed to increase precision in treatment comparisons. Previous work has shown that this variability can be related to water content. However, the relationship between structural stability parameters and water content is complex, because different processes linked to water content (slaking, swelling, age hardening) may govern the response of structural stability to a change in water content. This study used a time series analysis approach to describe the complex response of structural stability parameters to water content. Temporal changes in the wet aggregate stability (WAS) and the dispersible clay fraction (DCF) of a clay loam soil under four different cropping histories were modeled using gravimetric water content (θ) as an independent variable. The lowest predictability for WAS was observed for soil in bare plots and under continuous corn ( Zea mays L.) production. The amount of variability in WAS explained by θ increased as the number of years of grasses increased (to a maximum of 3 yr in this study). Regression models indicated that the low predictability of WAS using θ observed for bare soils and soils under continuous corn production was improved by including values of the antecedent water content before sampling. The predictability of DCF using θ was generally higher than the predictibility of WAS, did not include past values of θ, and was independent of cropping history.