Beef cattle feedlot operations can have a negative impact on the environment because feedlots can be sources of several contaminants to soil and water, including microorganism, nutrients, heavy metals, and hormones. Nutrients have been reported as contaminants of major concern, causing problems of eutrophication of water bodies, impairment of the quality of drinking water, and loss of fertility in soils. Therefore, the assessment of nutrients coming from feedlots is quite important to preserve the natural resources. Traditional ways to assess nutrient in soils by coring is tedious and time consuming, which led to the application of alternative techniques such as electromagnetic induction (EMI). This technique is fast and provides useful data set for plotting and interpretation. The general objective of this research was to use electromagnetic induction to describe nutrient distribution. Six feedlot areas were surveyed in Manitoba, samples were acquired using a response surface design and analyzed for electrical conductivity (ECe), nitrate, and phosphate, and calibration models for EMI readings were built using two different types of multiple linear regression (MLR) — soft and spatial MLR. Soft MLR used "easy-to-acquire" information, while spatial MLR used trend surface parameters. These models were built for simulations by layer and for the whole soil profile (i.e. composite). Models with better prediction capabilities were employed to depict nutrient distribution by mapping modelled readings. Interactions of feedlot and physiographic features were used to propose feedlot design criteria based on nutrient distribution. Comparison of the models showed that both soft and spatial MLR gave similar results because predictors included in the models were the same in most cases. Prediction of electrical conductivity was fairly good in both MLR, but prediction of nitrate and phosphate was usually poor in this study. However, good prediction of these nutrients by EMI was shown by other studies. Composite methods seemed more appropriate than by layer method for the prediction because they take into account the whole soil profile and overcome the limitation of non-significant depth intervals observed in the by-layer analysis. Models for vertically-weighted profiles gave the best results, but they could only be employed when the whole profile was sampled. Underestimated values were observed when these models were used in partially-sampled profiles. Predictors of major influence were both primary and secondary decorrelated principal component scores (i.e. z1 and z2), derived from EMI readings. The maps created from modelled values of profile-weighted ECe showed that feedlots 1 through 4 had ECe ≤ 3.5 dS m-1, but feedlots 5 and 6 exceeded this threshold and reached maximum values of 5.5 and 7.0 dS m-1, respectively. The distribution of this variable was mainly affected by topography in the field, with low elevations corresponding to high values of ECe. However, features promoting preferential surface flow (i.e. drainage ditches) strongly affected migration pattern in the field. Also, feedlots located in areas with a rocky layer beneath the soil layer tended to show higher values of this variable probably due to nutrients accumulation. Feedlot design criteria based on the results suggest locating the pens upslope with an impermeable drain to direct runoff into storage or treatment areas. For soils with large hydraulic conductivity, pens should be provided a liner.
展开▼
