Impact of weather on a lake ecosystem, assessed by cyclo-stationary MCCA of long-term observations

Temperate lake ecosystems are generally characterized by a strong annual cy cle, and the relationships between observations of such ecosystems and exte rnal forcing variables can exhibit a complex structure. Furthermore, the ob servational data record is often short. This makes it difficult to assess t he relationships between external forcing factors and their impact on the b iological succession. Cycle-stationary maximum cross-covariance analysis (M CCA) allows the effects of seasonality to be modeled in a flexible way, and we describe this statistical technique in detail. MCCA offers an objective method to approximate the high-dimensional total cross-covariance structur e by defining "weighting" patterns. With a predictor set of reduced dimensi on, a suitable regression between forcing variables and ecological response variables can be set up. Cyclo-stationary MCCA is used here to analyze the influence of meteorologic al variables (air temperature, wind speed, global radiation, humidity, and precipitation) on 13 biological and biogeochemical indicator variables of P lussee, a small lake in northern Germany. The main weather influence on the indicator variables was found to be connected to winter temperature. From the covariance structure the following major signals were detected to be re lated to higher winter temperature: a more intense spring algal maximum, a higher zooplankton biomass during the algal maximum, a less intense loss of nutrients to the hypolimnion, a higher summer bloom together with changes in the nutrient concentrations, and stronger oxygen consumption in autumn.