Nataša Atanasova

Automated discovery of a model for dinoflagellate dynamics

The aim of this paper is to discover a model equation for predicting the concentration of the algal species Peridinium gatunense (Dinoflagellate) in Lake Kinneret. This is a rather difficult task, due to the sudden ecosystem changes that occurred in the mid-1990s. Namely, the stable ecosystem (with regular Peridinium blooms until 1993) underwent changes and has transformed into an unstable system, with cyanobacterial blooms now occurring regularly. This shift in the algal succession is expected to influence attempts to model the lake ecosystem. Namely, the model structure before and after the change is likely to be different. Our modelling experiments were directed to discover a single model equation that can simulate dinoflagellate dynamics in both periods. We apply an automated modelling tool (Lagramge), which integrates the knowledge- and the data-driven modelling approach. In addition we include an expert visual estimation of the models discovered by Lagramge to assist in the selection of the optimal model. The dataset used included time-series measurements of typical data from the periods 1988 to 1992 and 1997 to 1999. Using the data and expert knowledge coded in a modelling knowledge library, Lagramge successfully discovered several suitable mathematical models for Peridinium. After the experts visual estimation and validation of the models, we propose one optimal model capable of long-term predictions.

Computational Assemblage of Ordinary Differential Equations for Chlorophyll-a Using a Lake Process Equation Library and Measured Data of Lake Kasumigaura

The software LAGRAMGE for computational assemblage and adaptation of ODE by using the expert knowledge and measured data has been applied for the simulation of chl-a in Lake Kasumigaura. As a result two types of chl-a models were discovered: (1) chl-a equations without considering zooplankton grazing assembled and trained by data of consecutive years were data of the last year was used for testing, and (2) chl-a equations considering zooplankton grazing assembled and trained by data of the years 1986 to 1989. The test results of the different models have demonstrated that LAGRAMGE can discover ODE that allow to simulate chl-a in Lake Kasumigaura for a variety of years. However the generalisation of discovered equations for unseen data of consecutive years was unsatisfactory, and the accuracy of calculated trajectories with regards to timing and magnitudes of peak events was moderate. The results have highlighted the importance of nutrients as growth limiting factors, and the need for considering functional algae groups in order to appropriately represent their selective grazing by zooplankton.

Development of a knowledge library for automated watershed modeling

In this work, we develop a library of components for building semi-distributed watershed models. The library incorporates basic modeling knowledge that allows us to adequately model different water fluxes and nutrient loadings on a watershed scale. It is written in a formalism compliant with the equation discovery tool ProBMoT, which can automatically construct watershed models from the components in the library, given a conceptual model specification and measured data. We apply the proposed modeling methodology to the Ribeira da Foupana catchment to extract a set of viable hydrological models. By specifying the conceptual model and using the knowledge library, two different hydrological models are generated. Both models are automatically calibrated against measurements and the model with the lower root mean squared error (RMSE) value is selected as an appropriate hydrological model for the selected study area.

Modelling dinoflagellate dynamics in Lake Kinneret

1992). Compared to simple linear regression, which calculates one equation (one weight vector) for the dependant variable that applies to the entire data set, piecewise or tree-structured regression divides the data set into several subsets on which a uniform value or linear equation can be applied. In this manner the piecewise linear parts can much better cover the nonlinear behaviour of the dependent variable.Regression tree model consists of nodes (branching points); branches, which connect the nodes; and leaves, which are ter-minal nodes where the dependant variable is predicted. The algorithm works recursively, starting with the entire set of ex-amples (S) and selecting the best attribute and the best split of that attribute according to the splitting criterion, deriving the most homogeneous subset regarding the class values or regres-sion model. Further details about the algorithm can be found in Q

Predicting chemical parameters of the water from diatom abudance in lake Prespa and its tributaries

In this work, we are modelling the physic-chemical parameters of water using bioindicator data (diatom taxa abundance data). Chemical status of the water (or water quality class) is defined by the values of measured physic-chemical parameters. Traditional approach to model these data is to learn a separate model for each parameter and then derive a global overview with some kind of summarization over the multiple models. Another approach is to learn a single model that describes all parameters (multi target approach). We explore these approaches and apply them on data from Lake Prespa and its tributary rivers. The obtained models revealed interesting connections between the diatom taxa and the water quality (i.e. the values of the chemical parameters).

Ammonium removal in landfill leachate using SBR technology: dispersed versus attached biomass

Large concentrations and oscillations of ammonium nitrogen (NH4+-N) in municipal landfill leachate pose considerable constraints to its further treatment in central wastewater treatment plants. The aim of this study was to evaluate and optimize two technologies for the pre-treatment of 600 L/day of landfill leachate: in particular, to optimize their operational conditions for NH4+-N removal up to a level appropriate for discharge to sewers, i.e. <200 mg/L. Both technologies were based on a sequencing batch reactor (SBR), with two different biomass processes: (A) SBR with dispersed/flocculated biomass and (B) SBR with biomass attached to carriers. The results revealed that both technologies successfully reduced the NH4+-N from 666 mg/L (on average) at the inflow to below 10 mg/L at the outflow with alkalinity adjustment in a 12-hour cycle. Both technologies achieved 96% removal efficiencies for NH4+-N. However, SBR with dispersed biomass showed higher flexibility under varying conditions due to the shorter adaptation time of the biomass.

Modeling the Kinneret Ecosystem

Modeling of the Lake Kinneret ecosystem and its various components has developed greatly since the first effort in 1980. Modeling studies have included a range of approaches, some of which have focused on the entire ecosystem, while others only on certain components. The modeling approaches that have been applied to the Lake Kinneret ecosystem range from statistical approaches, data mining, and machine-learning techniques to flux models, bioenergetics, nutri- ent-phytoplankton-zooplankton (N-P-Z)-type models, and complete ecosystem models. Models have been used to enhance our understanding of key limnological and food-web processes. The models, however, have also been used as a means for providing resource managers with improved management tools. This has included, in some cases, integrating model output and a quantified water quality (WQ) sys- tem as the basis for establishing the relationships between management measures and water quality. Thus, highlighting the role of ecosystem modeling as a critical management tool. The use of models, especially complex ecosystem models, requires constant updating and ongoing development along with a wide range of information and data. In most cases, the available information and data do not satisfy model needs thereby introducing model uncertainty. In order to improve model reliability, and, as a result, its applicability, it is important to increase the overlap between routine monitoring and model data requirements.

Modeling Of Waste Water Treatment Plant With Regression Trees

Simulation of wastewater treatment plants (WWTP) is a difficult task, due to the complex and mostly dynamic behaviour of the WWTP system. Regression trees are presented as an useful simulation/modelling tool for making predictions on WWTP operation given measured data at the input. Crucial step in the construction of such models is data preparation. Two data sets measured on two different WWTP are used in this paper. Both data bases are composed of data that are usually measured on a WWTP and characterise the WWTP operation. The main difference between them is in data presentation. In the first data set (WWTP1) data are presented as a one-day situation of the plant operation, i.e. daily averaged values of the measured data (attributes) are given. Second data set (WWTP2) is composed of actual values of the attributes measured in one hour intervals. Regression tree models, that predict outflow attributes according to inflow attributes are constructed for both data sets and compared in their performance. Assumption that data presentation and further preparation have a big influence on the results was confirmed. Program package WEKA, which includes most of popular machine learning algorithms, was used for constructing the models.

Multi-criteria Evaluation of Sustainable Urban Drainage Systems

This paper evaluates the use of Sustainable Urban Drainage Systems (SUDS) measures with hydrological-hydraulic modelling and multi-criteria analysis on a case study of Girona, Spain. To assess their effectiveness for rainwater runoff reduction and consequently the reduction of combined sewer overflow (CSO) the software Giswater was employed for development of the urban catchment model. This software couples spatial data from QGIS with EPA Storm water management model (SWMM). In accordance with the conditions in the urban catchment, five scenarios were developed consisting of following SUDS measures: infiltration basins, infiltration trenches, green roofs and their combinations. These scenarios were evaluated with multi-criteria analysis based on CSO reduction, CAPEX, OPEX, amenity, biodiversity, and feasibility regarding ownership. According to the results, the scenario that included only infiltration basins was most favourable (average grade: 4.1/5), followed by the scenario which combined infiltration basins and tranches (average grade: 3.5/5).