Predicting the risk of cobalt deprivation in grazing livestock from soil composition data

The objective was to derive predictive equations for acetic acid-extractable cobalt (A-ECo) in soils so that extensive national databases for total (T) Co in soils and stream sediments could be converted to 'plant available' concentrations for the purpose of predicting risk of Co deficiency in grazing livestock. Data on the chemical and physical properties of 103 soils from 15 different parent materials and 54 soil series in England and Wales were used. Ranges for the mean values for parent materials were: TCo, 5.0-20.4 and A-ECo, 0.20-1.30 mg kg(-1); percentage (P) A-ECo, 3.4-13.5; soil manganese (Mn) 268-1174 mg kgDM(-1); pH, 3.7-8.0. There were significant effects of parent material on all parameters with Chalks, Old and New Red Sandstones particularly low in A-ECo. Multiple linear regression yielded the following equation for predicting A-ECo, which accounted for 56% of the variance with 12 outliers, including the lowest pH values, omitted: A-ECo ( mg kg(-1)) = 1.4 - 0.0619TCo (mg kg(-1)) -0.432TMn (g kg(-1)) -0.171 pH Prediction of PA-ECo was less precise: PA-ECo (%) = 21.1 - 4.5TMn (g kg(-1)) - 1.77pH (r(2) = 44 8%; 88 d.f.) REML was used on the complete, unbalanced, log-transformed data set to fit a Generalised Mixed Model with parent material as random effect and soil Mn and pH as fixed effects; the effect of parent material was no longer significant. It was concluded that A-ECo can be satisfactorily predicted for most soils in England and Wales from TCo, TMn and soil pH.