Validation of a diatom-phosphorus calibration set for Sweden

1. A weighted averaging (WA) regression and calibration model for diatoms and total phosphorus (TP) was developed from a dataset of 45 surface-sediment samples from Swedish lakes. Jack-knifed error statistics were comparable with those for similar diatom-TP datasets: r(jack)(2) = 0.47, root mean squared error of prediction (RMSEP) = 0.24 log(10) mug TP L-1 and mean bias = -0.002 log(10) mug TP L-1 for the simple WA model; r(jack)(2) = 0.36, RMSEP = 0.27 log(10) mug TP L-1 and mean bias = 0.017 log(10) mug TP L-1 for WA with tolerance downweighting. 2. The model was used to estimate TP concentrations for the Ekoln basin of Lake Malaren using a Pb-210-dated sediment core record. Highly eutrophic conditions developed in the basin in the 1960s as a result of nutrient inputs from cultivated land and the city of Uppsala. A reduction in the supply of phosphorus from sewage outlets in the late 1960s resulted in less eutrophic conditions. 3. The model results indicated levels of 50-60 mug TP L-1 prior to 1900. The rapid eutrophication of the lake basin after the 1950s and the subsequent recovery were evident from the diatom data. 4. Diatom-inferred TP (DI-TP) values were validated by comparison with monitored data for the period 1966-95. The diatom model tended to underestimate TP at high levels (> 80 mug L-1) but overestimate at lower concentrations. 5. A good agreement was observed between the trends in TP concentration and the DI-TP concentration and the timing of the maximum was well reflected by the diatom-based reconstruction. A significant correlation (r(2) = 0.69, P < 0.01) was found between DI-TP and measured TP at this site.