The prediction of mineral deficiencies in grazing livestock requires good correlations between convenient markers of mineral status and animal health or productivity. Correlations are likely to become weaker in moving from animal to pasture to soil in pursuit of a predictor because of the many factors which influence mineral uptake at each interface. However, soils are the easiest to characterize and correlations might be improved by removing the effects of known sources of variation. The influence of botanical (pasture species), geographical (altitude) and pedological (bedrock type, soil pH and extractable mineral concentration) factors on mineral concentrations in dry season pasture was therefore assessed. Samples of topsoil and herbage were gathered from 135 sites on 84 farms in the Mt Elgon region of W. Kenya between January and March, 1987. The underlying parent bedrock was determined from 1:125,000 Geological survey maps and altitude from topographical maps. Botanical composition of the pasture sample was recorded. Soil pH and total (Se) or extractable (not Se) mineral concentrations were determined by standard methods as were total mineral concentrations in unwashed herbage. Distribution of principle botanical species and all bedrock types amongst the sample sites are indicated in Tables 1 and 2 respectively, together with the analytical results. Statistical analysis used a residual maximum likelihood (REML) model for unbalanced data sets.
Pasture concentrations of Ca, P and Cu were generally below the requirement of ruminants, Zn and Se were marginal while Co, Fe and Mn were adequate. Soil bedrock had little influence on herbage composition. Of the four macro-elements, only S was affected by geology, low values being found above TV and MS bedrock. By contrast, only P was not affected by species, Ss being low and PC usually high in macro minerals. Of the seven trace elements analysed, geology influenced only one (Cu); low values were again found above TV and Mfi but Cu availability to grazing ruminants would be relatively high because of the associated low S values. By contrast, only Se was unaffected by species, PC being rich in all but Mn. Soil bedrock had a greater influence on soil composition but correlations between soil and herbage usually accounted for less than 10% of the variation in pasture composition (max r value 0.5 for P): correlations within species were equally poor. The influence of species on herbage Co disappeared if herbage Fe was used as a covariate, suggesting that contamination by soil Co varied between species: however, the correlation with soil Co remained weak. Soil pH was generally low and its use as a covariate did little to improve soil/plant relationships. Herbage Cu increased and Se decreased in curvilinear relationships with altitude.
Mineral deficiencies were therefore likely to occur in grazing livestock, risk being influenced by botanical and topographical but not pedological factors.