An inductive approach to the modeling of lake water quality in dimictic, glacial/boreal lakes

Sammanfattning: To explore the interaction between elements of the watershed and lake water quality, a database large enough to approximate the generic population of dimictic, glacial/boreal lakes is assembled and analyzed with multivariate statistical methods in an inductive top-down approach. Utilizing the analytical capabilities of geographical information systems, the spatially distributed characteristics of land cover, soil and bedrock elements are identified and parameterized with respect to element morphology and hydrology. Together with additional parameters of catchment/lake morphology, hydrology and climate, the watershed descriptors serve as potentially explanatory variables in assessing the inter-lake variance of water quality in 87 monitored watersheds. With lake water quality defined as the composition of conductivity, hardness, Secchi depth, color, alkalinity, total phosphorus and pH, the following is concluded:It is possible to mathematically transform the probability density functions of water quality constituents to approximately normal distribution. With the monitored watersheds being lime stone treated, it is concluded that the transformation functions derived approximate the probability density functions of water quality constituents in the generic class of dimictic, glacial/boreal lakes, and that they should be well suited as transformation standards in natural, as well as in limestone treated lakes. This implies that limestone treatment does not significantly alter the shape of the probability density functions analyzed, and that the effects of limestone treatment can be neglected whenever analysis of variance is focused upon.The total variance of the water quality constituents may be unambiguously approximated in a two-dimensional Cartesian base with transformed variables of lake water hardness and color as principal variance representatives. The base is suggested as a lake classification standard, wherein the inorganic and organic characteristics of subset lake populations (such as single lakes) can be combined and referenced to the standard of the generic population.Introducing descriptors of element morphology and hydrology does significantly improve the performance of basic regression models in explaining the inter-lake variance of water quality. This verifies that morphology and hydrology are generally important in determining the element contribution to the chemical composition of downstream waters. With additional descriptors of lake/catchment climate and morphology included, optimally performing models do significantly improve the performance of existing regression models in explaining the inter-lake variance of water quality. Lake water hardness, color, conductivity and Secchi depth are simulated with precision enough for predictive purposes, whereas the models of lake water alkalinity, total phosphorus concentration and pH provide a rather qualitative - although highly significant - estimate. Fifteen landscape elements, and seven additional descriptors, are identified as significantly contributing to the chemical composition of downstream waters.With watershed descriptors being designed to carry process information, the derived models assess the physical processes that govern the inter-lake variance of water quality. Since parameterized upon the watershed scale, they should be applicable whenever watershed management is implemented into the conservation - or manipulation - of water quality in dimictic, glacial/boreal lakes.