Estimating Nutrients and Chlorophyll a Relationships in Finnish Lakes

Abstract

We model the response of chlorophyll a-a surrogate for the phytoplankton community volume-to variations in lake total phosphorus (TP) and total nitrogen (TN) concentrations. The model is fitted to a large cross-sectional data set from the Finnish Lake monitoring network. The objective is to support the Finnish Government in identifying management actions to achieve compliance of the chlorophyll a concentration standard with a given confidence level and to provide tools for the estimation of critical (target) loads for nutrients in monitored lakes. We develop a Bayesian hierarchical linear model which combines advantages of both the currently preferred non-hierarchical lake-type-specific linear model and lake-specific linear model fitted separately using data from a single lake. The hierarchical model is less biased at lake-level compared to the lake type model. In contrast to the lake model, it predicts the lake specific chlorophyll a response to nutrients outside the lake specific observational range. The hierarchical model is used to calculate probabilities of chlorophyll a concentration exceeding the standard under different nitrogen and phosphorus concentration combinations. These probabilities can be used to estimate acceptable nitrogen-phosphorus concentration combinations by a lake manager. We discuss how our study can be useful in implementing the European Water Framework Directive.

Bayesian Hierarchical/Multilevel Models for Inference and Prediction Using Cross-System Lake Data

“…Multilevel/hierarchical models are similar in concept to random coefficient and empirical Bayes models which have been previously demonstrated with cross-system lake data (Reckhow 1993, Reckhow 1996. They can be very useful for organizing ecological data and have been used to synthesize information in a cross-system data set of Finnish lakes (Malve and Qian 2006).…”

“…A popular strategy to increase the likelihood of similar behavior is to group lakes based on common attributes. Features used to categorize lakes have included lake type (natural vs. reservoir), geographic setting, and geomorphology (Canfield and Bachman 1981, Reckhow 1988, Malve and Qian 2006 and models based on such groupings are then used for individual lake forecasts (Hession et al 1995). The goal of this strategy is to ''borrow'' information from other similar lakes to increase the accuracy of prediction for a particular lake, so that the risk of making a bad decision is acceptably low.…”

Causal analysis of phytoplankton development in a small humic lake using structural equation modelling

“…For instance, the different characteristics (e.g. area, depth, water colour, residence time) of a lake partly determine its water quality and response to elevated nutrient levels (Malve and Qian 2006, Jackson et al 2007, Nõges 2009). Besides lake characteristics, the different ecoregions (Lamon and Qian, 2008), landscapes (Wagner et al 2011) have also been used as hierarchy levels.…”

“…Despite this clear advantage over simple regression, for the review only seven LMM papers for phytoplankton was found, of which two were actually using a non-linear, so called general additive method (Carvalho et al 2011, Salmaso et al 2012). The hierarchical model for chlorophyll a introduced by Malve and Qian (2006) was used as a basis for the LLR modelling tool, which has been developed to ease the use of models in WFD-related management of lakes. LLR gives estimates of the loading reduction that is required to have the concentrations of total phosphorus, total nitrogen and chlorophyll a under the lake type specific class limits for good water quality (see Table 1 in IV).…”

“…The assumption of a continuously stirred tank reactor (CSTR) is applied, hence all the values are given as averages for the lake retention time (in years). To improve the model in Malve and Qian (2006), more recent data from Finnish lakes were used, with almost double the number of observations. In addition, the lake types in the model were updated.…”