Autoregressive state spatial modeling of soil bulk density and organic carbon in fields under different tillage system

• Spectra analysis identified cyclic patterns related to uncontrolled traffic and harvesting practices.
• Cospectra and coherence showed that BD, WC and OC were spatial related at 2.5, 4 and 9 m.
• Spatial relationships among soil variables allow predicting BD and OC with state space models.
• Despite WC was spatial correlated to BD and OC no model could properly predicting its behavior.

Uncontrolled transit operations could be responsible of repetitive patterns of soil properties. Cyclic spatial structure patterns of soil bulk density (BD), soil organic carbon (OC) and water content (WC) in farm fields were assessed with spectra and cross spectra analysis in a 100 m-long transect sampled at 1 m apart, on Vertic and Typic Argiudol soils under no tillage (NT) from a commercial farm in San Antonio de Areco, 34° 15′32.42″ S, 59°25′19.93″ W, Ondulate pampa region, Argentina. Although spectrum and cospectrum analysis showed several cyclic locations, only sites at approximately 2.6 m, 4 m and 9 m were significant (p < 0.05). These distances are clearly related with tractors axles and combine harvester paths, thus suggesting the importance of those operations on soil variables. An autoregressive state space modeling approach was used to integrate the spatial information and to model BD, WC and OC in different transects at 10 m and 30 m apart. With the spatial relationship between BD and OC we create predictive models that explain 63% of the OC data and 54% of the BD data over a 2740 m-long transect. However, it was not possible to predict WC, despite the spatial correlation observed among the soil variables. The low importance of WC in the modeling process of BD and OC, and the large dominance of the autoregressive part in the final model are pointing out that an important surrogate variable is missing, which could be the key for modeling soil variables at different scale.