Statistical models for the estimation of net phosphorus sedimentation in lakes

Abstract

The empirical adequacy of four phosphorus mass-balance models is evaluated with respect to how the prediction error variance of the corresponding net sedimentation parameters is propagated in the steadystate equations. Using the criterion of minimum propagation error variance (PEV), different groups of lakes can be distinguished for which different empirical equations are used to predict net phosphorus sedimentation. The classification reduced prediction error significantly and also reflected different patterns of sedimentation. Application of this criterion to time-series of individual lakes shows that it is possible to determine a priori whether net annual sedimentation will be better correlated to the annual loading or to the lake content. The correlations depended also on the load/lake content ratio, suggesting that net sedimentation is best viewed as the sum of the partial sedimentation of the load and of the partial sedimentation of the lake content. On average, 25% of the load and 18% of the lake content are sedimented annually. Viewing net phosphorus sedimentation as a function of both the load and the lake content can also explain and predict the well-known cross-sectional correlation between phosphorus retention and water residence time.