Abstract
An extensive soil condition assessment should include some appropriate tools to measure crop management influence on the soil condition. Univariate (partial regressions) and multivariate (canonical correspondence analysis, partial canonical analysis of correspondence and the method of variation partitioning) were used to determine the independent and joint effects of crop management and the spatial structure of the soil condition. Soil bulk density (BD), aggregate stability (AS), soil organic carbon (SOC) content and carbon lability (LAB) were measured in crop fields as soil condition variables. Four crop management descriptors were used: 1) an indicator ( T) for assessing the tillage effect on retention of crop stubble on soil surface and soil aggregate stability; 2) the direct energy consumption [Mj ha − 1 year − 1 ] during tillage operations ( T energy); 3) the number of annual crops during the continuous cropping cycle (CROPS); and 4) the maize crop frequency of the continuous cropping cycle (MZ). Geographical location of the crop fields was included in order to estimate the spatial autocorrelation of the crop management descriptors. Results from partial correlations showed that SOC content was negatively correlated with tillage energy consumption ( r = − 0.51). LAB and AS were negatively correlated ( r = − 0.52 and − 0.78, respectively) with the number of annual crops (CROPS). All these correlation remain significant when the site location effect were removed. When the marginal (i.e. individual) effects of crop management descriptor on the soil condition were analyzed using multivariate methods, T explained 27.6% of the overall soil variability whereas CROPS and MZ explained 25.3% and 23%, respectively. The variance partitioning procedure showed that when the autocorrelated fraction of the crop management was isolated, all the significant indicators explained 23% of the overall variability of the soil quality properties analyzed. Additional explanatory power of the above is due to autocorrelation of the management variables as this fraction explained increased from 23% to 55.2% when the autocorrelated spatial component of the management influence was added. It means that 32.2% (55.2%–23%) is the amount of the explained variation that comes from sampling nearby crop fields. Most of the pure management variation was explained by CROPS, as it explained a significant portion of soil variation (16.1%) when all other confounding variables were removed. From the unexplained variance, 10.7% of total variability was autocorrelated and could be explained solely by the geographic coordinates. Finally, 34.1% of the soil matrix variance remained unexplained by either the management descriptors or by the geographical coordinates we considered. The variance partition procedure proved to be a useful analytic approach for detecting some agricultural intensification effects, by allowing the separation of pure management effect from the spatial structure of soil variability.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have