Performance of different modeling techniques in testing the impact of environmental variables on eel landing in Ichkeul Lake, a RAMSAR Wetland and UNESCO biosphere reserve

• DOI: 10.1016/j.rsma.2024.103587
• Publication Date: 2024-05-26T14:22:03Z
• Authors: Sabrine Sahbani; Béchir Béjaoui; Ennio Ottaviani; Sihem Benabdallah; Eva Riccomagno; Enrico Prampolini; Donata Melaku Canu; Hechmi Missaoui; Cosimo Solidoro
• Keywords: Ramsar site

Advanced modeling techniques, including Random Forest (RF) and Cubist model (CB), were used to assess the relationship between environmental factors European eels (Anguilla anguilla) abundance provide insights into lake's ecological status while considering climate change anthropogenic influences. A comprehensive dataset, attained through extensive biological monitoring for period 2010-2020, was employed. The performance of models is carried out using key metrics root mean square error (RMSE), coefficient determination (R²), estimation (MAE). In addition, a sensitivity analysis conducted ascertain relative significance thirteen input variables in shaping predictions models. precision CB RF predicting eel landings surpassed that Multiple Regression. training achieved R2=0.55, RMSE=7.68 tons, MSE=6.20 R2=0.56, RMSE=7.20 MSE=5.56 tons. High accuracy maintained on testing with achieving R2=0.73, RMSE=5.13 MSE=5.89 RMSE=5.81 MSE=4.67 scatter plot predicted measured indicated tends overestimate lower values underestimate higher landings, gave better this context. Further, unequivocally identified three pivotal – water level, salinity, turbidity level as most influential determinants governing landing ecosystem. Thus, considered be more promising interpreting parameters which could by managers an effective lake management strategy.