Selection of a suitable probability model for the analysis of biochemical data from soil

Abstract

Compacted soil in Argentine and other countries causes serious loss of agricultural production. In previous studies we used biological indicators of soil quality to describe the soil in different systems of agricultural management. In this study, we studied the microbial activity from soil based on enzyme activities measurements at agricultural soil located in the Northeast of the Province of Buenos Aires. This followed an experimental design consisting of three blocks completely randomized with three treatments: decompacted soil (DECOMP), compacted soil (COMP) and non-cultivated soil (CONTROL). The size of the experimental unit was 4.200 m. The number of variable continuous quantitative biochemical measures was 10. Soil samples were taken with 15-days frequent biochemical analyses during the period June 2010 to April 2013. Multivariate time series analysis was used to study the behavior of soil enzyme activities. A time series algorithm was used to study the biochemical profile of the soil subjected to three cycles of agriculture with the ts library in the R environment under Debian GNU/Linux system. The exact likelihood was computed via a state-space representation of the ARIMA process, and the innovations and their variance found by a Kalman filter. We demonstrate that the biochemical characterization of the soil is related to the state of its compaction using the ARIMA model. These data lead to the idea that the interrelation of biochemical variables from soil could be used as an accessor of the degree of current compaction of soil. Finally the effect of mechanical decompaction of soil on biochemical variables for 3 years under no-tillage as a management system was tested.