A diatom-based inference model for autumn epilimnetic total phosphorus concentration and its application to a presently eutrophic boreal lake

A diatom transfer function to infer epilimnetic total phosphorus (TP) concentration was derived using surface sediment diatom data from 68 medium-sized (10-1000 ha) lakes in Southern Finland. Publicly available monitoring records were used in lake selection to avoid gradients caused by pH and humic substances. Constrained and partially constrained ordinations indicated that TP was an important variable influencing diatom assemblages. A long floristic gradient in relation to TP was also apparent and therefore an inference model was developed for TP using unimodal-based regression and calibration methods. The final model included 61 lakes with epilimnetic TP concentrations between 3 and 89 mug P l(-1), measured during the autumnal circulation period. It has a jackknifed-estimated root mean squared error of prediction of 0.16 log mug P l(-1), a maximum bias of 0.28 log mug P l(-1), and an r(j)(2)ack of 0.76. The model was tested in the presently eutrophic Lake Valkjarvi (epilimnetic [TP] 60-85 mug P l(-1)), located in Southern Finland. It successfully predicted the measured autumnal epilimnetic TP concentration for the past twenty years and the changes in inferred [TP] reflected disturbances known to have occurred before that time. The diatom-based inferences show that Lake Valkjarvi was oligo-mesotrophic as late as the 1930's and has become eutrophic because of nutrient inputs from agriculture and, especially, municipalities. However, epilimnetic TP concentration has not increased further.