The habitat template of phytoplankton morphology-based functional groups

Abstract

The identification of the main factors driving phytoplankton community structure is essential to understand and adequately manage freshwater ecosystems. We hypothesize that differences in morphological traits reflect phytoplankton functional properties that will be selected under particular environmental conditions, namely their habitat template. We apply a morphology-based functional groups (MBFG) approach to classify phytoplankton organisms and define each group template. We use machine learning techniques to classify a large number of phytoplankton communities and environmental variables from different climate zones and continents. Random forest analysis explained well the distribution of most groups' biovolume and the selected variables reflected ecological preferences according to morphology. By means of a classification tree it was also possible to identify thresholds of the environmental variables promoting groups dominance in different lakes. For example group III (filaments with aerotopes and high surface/ volume including potentially toxic species) was dominant when light attenuation coefficient was[3.9 m -1 and total nitrogen was[2,800 lg l -1 . We demonstrate that morphology captures ecological preferences of phytoplankton groups and provides empirical values to describe their habitat template.

Morphology-based classification of functional groups for potamoplankton

“…2). In contrast to the previous MBFG findings (Kruk and Segura, 2012), the MBFG R model emphasized the importance of flow velocity and suspended sediment. These results showed that the dominance of potamoplankton functional groups depends first on hydrological conditions and second on nutrients.…”

“…These species were from more than 200 lakes situated in climate zones that ranged from subpolar to tropical ones. Differences in simple morphological traits among MBFGs well reflected the significant differences in functional characteristics and habitat templates independently from the geographical location (Kruk et al, 2010;Kruk and Segura, 2012).…”

“…binary partitioning of the response variable into regions that are increasingly homogeneous (i.e. nodes) until no improvement is possible (Kruk and Segura, 2012). The final nodes are called leaves.…”

“…This process continues until reduction of SSE is no longer achieved. The methods are easily interpretable and provide simple yes (>) or no (<) decision trees (Kruk and Segura, 2012). After constructing the full tree, this tree was pruned back by minimizing the cross-validated error to avoid overfitting (De'ath and Fabricius, 2000).…”

“…The MBFG R well reflected the different habitat templates. The main ecological processes in lakes are usually evasion of loss processes (mixing and zooplankton), water temperature conditions, and resources acquisition (nutrients and light) (Kruk and Segura, 2012). In addition to these processes, the unique effect from river flow was also represented in the results.…”