Influence of climate change on diatoms diversity indices in Lake Prespa

Abstract

Applying machine learning techniques into ecology have proven to be useful into obtaining knowledge for certain problems. Using these diversity indices (DIs) it will be very useful to model the specific diatom communities which are known to exist only in definite environmental conditions. This property is used to model the abiotic environment influence on diatoms. Using the diatom relative abundances into account, a diversity index depends not only on species richness but also on the evenness, or equitability, with which individuals are distributed among the different species. Diversity indices provide important information about rarity and commonness of species in a community. The ability to quantify diversity in this way is an important tool for biologists trying to understand community structure (River diatoms: a multiaccess key). Understanding how these indexes interact with physical-chemical parameters of the given environment is very useful to know. Parameters like temperature, dissolved oxygen, pH, ammonia and others are one of the few that are vital for diatom survival (Reynolds. C. S., 1998). This is why we build models to see how diatoms diversity indices response on the changes of these parameters. In order to extract this knowledge from the ecological data we use machine learning techniques. The most researched type of machine learning is inductive machine learning, where the experience is given in the form of learning examples. Machine learning (and in particular predictive modelling) is increasingly often used to automate the construction of ecological models (Džeroski, 2001), (Joergensen, 2001). Most frequently, models with regression trees of diversity indices and population dynamics are constructed from measured data by using machine learning techniques. The most popular machine learning techniques used for modelling diversity indices include decision tree induction and rule induction. In this paper, we focus on applications of machine learning in ecological modelling, more specifically, applications of modelling diversity indices. We will use a dataset, which has been collected from different measurement stations placed in Lake Prespa, as a part of the EU project TRABOREMA (TRABOREMA Team, 2005-2007). Several important parameters are measured, which reflect the physical, chemical and biological aspects of the water quality of the lake. From these measurements, several diatoms (algae) belonging to the group Bacillariophyta) will be considered for estimating a relationship between their relative abundance, and then calculated their diversity indices and the abiotic characteristics of the habitat. Diatoms are known to be almost ideal bio-indicators of the environment in several studies (Van Dam H., 1994). The paper is organized as follows. Section 1 introduces with idea of the diversity indices modelling and the main purpose of this paper, the diversity indices models for Lake Prespa, in Section 2 we give an overview of the diversity indices modelling and introduction to machine learning, with a briefly description of the approach to machine learning that is often used in this kind of modelling: decision tree induction and rule induction. The measured data, the main diatom bio-indicators and data collection procedures are presented in Section 3, while Section 4 describes the diversity indices models which were built for several diatoms from lake and rivers measurements. Section 5 concludes and gives directions for future work on this subject.