The niche_analysis project develops a scientific prototype to reconstruct the ecological niche of phytoplankton species or functional groups and estimate bloom-favouring conditions. It combines PCA and KDE to model the environmental space historically associated with bloom initiation, using bloom events as practical proxies for competitive ecological conditions. Based on this framework, the project defines prototype indicators such as the Suitability Index, Prevalence Index, and Environmental Niche Index. Current results, especially for neighbourhood 2, show promising ecological differentiation and interpretable temporal signals. However, further refinement is needed in functional grouping, bloom-point extraction, KDE validation, and operational testing before routine use.
By Frédéric Haffter.
Phytoplankton Niche Project: Development of an Ecological Indicator
The niche_analysis project develops a scientifically grounded prototype to reconstruct the ecological niche of phytoplankton species or functional groups and estimate the environmental conditions that may favour bloom development. The workflow is based on the ecological niche concept and combines principal component analysis (PCA) with kernel density estimation (KDE) to represent the environmental conditions historically associated with bloom initiation. Bloom events are used as practical proxies for situations in which a functional group is ecologically competitive, allowing bloom inflection points from historical abundance time series to be projected into a shared PCA-based environmental space. KDE is then applied to quantify the density of these bloom-related observations, creating a niche representation that supports prototype indicators such as the Suitability Index (SI), the Prevalence Index (PI), and the Environmental Niche Index (ENI).
At the current stage, neighbourhood 2 provides the most advanced implementation of the prototype, and the first results suggest that the workflow can distinguish functional groups in environmental space while producing interpretable temporal signals in the index series. These results indicate that the approach can already generate useful ecological insights and may become valuable for operational monitoring. However, several methodological limitations remain, including uncertainty in the functional grouping strategy, ambiguity in bloom inflection-point extraction under noisy real-world conditions, and limited validation of KDE-based niche estimates. The next priorities are therefore to refine the unit of comparison, improve and partially validate the bloom-point extraction workflow, strengthen confidence in the niche estimates, and test the indicator under operational monitoring conditions before considering more routine use.
