Use of sedimentary pigments to infer past phosphorus concentration in lakes

We propose a palaeolimnological method for inferring past total phosphorus (TP) concentrations in lake water from spectrophotometrically-measured sedimentary pigments, particularly total carotenoids (TC). Our approach is based on a highly significant statistical correlation (P < 0.0001) between pigment concentrations (total carotenoids) in the surface sediment of 28 Italian lakes (subalpine, large, deep, shallow, volcanic) and TP concentrations measured in these lakes at overturn when the core was collected. A transfer function was developed from this "training" set, and used to estimate past TP concentrations from pigment concentrations in sediment cores. The results generally agreed with TP values as measured by long-term water quality monitoring programs. Contrasting results were obtained by a comparison with diatom-inferred TP. While the diatom model showed a tendency to overestimate TP values higher than 100 mu g l(-1), the pigment model correctly estimated TP in lakes when TP was < 100 mu g l(-1), but not when lakes were rich in macrophytes. In fact, lakes with extensive populations of aquatic submersed macrophytes and epiphytes are outliers in terms of the TC versus TP relationship. The root mean square error of prediction of the pigment model is lower than those derived from certain diatom-based inference models. The predicted and residual values are not related to the estimated values and their average is not statistically different from zero. Errors were estimated via a 'leave-one-out' re-sampling technique. The proposed method permits rapid and relatively inexpensive determination of reference trophic conditions.