TRIX-XGBoost Trophic Model for assessing the trophic status of Ichkeul Lake, a Ramsar sanctuary in Northern Tunisia

A combined approach, the TRIX-XGBoost Trophic Model, was developed by integrating the Multimetric Trophic Index (TRIX) with the Extreme Gradient Boosting (XGBoost) algorithm to enhance the assessment of Lake Ichkeul's trophic status. The XGBoost technique was used to construct predictive models for chlorophyll-a concentrations in the lake and its associated rivers. Two datasets were assembled for the lake and its rivers to achieve these objectives, leveraging 13 environmental parameters as predictors. The study also investigated the connection between chlorophyll-a and the observed environmental parameters using a multivariate analysis, which show the spatio-temporal and seasonal variations of the parameters. The findings reveal that the observed TRIX values ranged from 3 to 5 in Ichkeul Lake, indicative of waters spanning from moderate to poor quality, while in Ichkeul Rivers, these values fluctuated between 4 and 7, signifying eutrophic conditions with a high trophic level. The XGBoost models, however, reveal a pivotal insight: the most influential factors affecting chlorophyll-a variations are nutrients and temperature. This discovery underscores the potential for developing a more streamlined model for this variable.