Predicting the future spread of Alien Invasive Species (AIS) in the Mediterranean Sea is a critical challenge for marine risk assessment due to their significant ecological and economic impact. The aim of this thesis is to provide a unified theoretical framework for AIS risk assessment that integrates both correlative (Machine Learning) and mechanistic (process-based) models to predict suitable areas and simulate the invasion process. This thesis addresses the development of the three key components to achieve this goal: The first component is the preparation of a solid data basis. The dataset contains current and future key bioclimatic variables for the Mediterranean Sea. It covers the period 2005–2099 under the high emissions scenario RCP8.5 at high spatial and temporal resolution. The second component is a robust Machine Learning Ecological Niche Modelling (ML-ENM) method. It includes three major methodological advances: (i) an improved model tuning phase, (ii) a novel site weighting scheme to improve the validity of performance metrics, and (iii) a multi-criteria decision-making framework for unbiased model selection. These innovations address critical challenges of ML-ENM modelling related to: maximising model transferability, mitigating data bias and ensuring ecological validity. The third component is a spatial interaction model that explicitly simulates vessel-mediated invasions from known presence sites. The probability of invasion is modelled as the strength of the connection between “infected” and “exposed" (susceptible) areas, using a production-constrained gravity model. The model incorporates data on vessel traffic, suitability and substrate type. The framework is presented through an applied case study of Caulerpa cylindracea, one of the most dangerous invasive alien species in the Mediterranean. The results consist of annual suitability maps of C. cylindracea for the period 2000-2050 and the maps of simulated annual invasions for the period 2020-2023. Despite a significant negative trend in suitability, the species continues to expand its range due to increasing shipping activities, highlighting the detrimental impact of human activities on the basin. The framework is modular and easily transferable to allow for future extensions and applications to other target species, significantly expanding the reliability and depth of the AIS risk assessment. The dataset and scripts to reproduce the analysis are publicly available.

Towards an integrated ecological modelling framework to assess the risk of Alien Species spread in the Mediterranean Sea: Caulerpa cylindracea as a case study

FIANCHINI, MARCO
2025

Abstract

Predicting the future spread of Alien Invasive Species (AIS) in the Mediterranean Sea is a critical challenge for marine risk assessment due to their significant ecological and economic impact. The aim of this thesis is to provide a unified theoretical framework for AIS risk assessment that integrates both correlative (Machine Learning) and mechanistic (process-based) models to predict suitable areas and simulate the invasion process. This thesis addresses the development of the three key components to achieve this goal: The first component is the preparation of a solid data basis. The dataset contains current and future key bioclimatic variables for the Mediterranean Sea. It covers the period 2005–2099 under the high emissions scenario RCP8.5 at high spatial and temporal resolution. The second component is a robust Machine Learning Ecological Niche Modelling (ML-ENM) method. It includes three major methodological advances: (i) an improved model tuning phase, (ii) a novel site weighting scheme to improve the validity of performance metrics, and (iii) a multi-criteria decision-making framework for unbiased model selection. These innovations address critical challenges of ML-ENM modelling related to: maximising model transferability, mitigating data bias and ensuring ecological validity. The third component is a spatial interaction model that explicitly simulates vessel-mediated invasions from known presence sites. The probability of invasion is modelled as the strength of the connection between “infected” and “exposed" (susceptible) areas, using a production-constrained gravity model. The model incorporates data on vessel traffic, suitability and substrate type. The framework is presented through an applied case study of Caulerpa cylindracea, one of the most dangerous invasive alien species in the Mediterranean. The results consist of annual suitability maps of C. cylindracea for the period 2000-2050 and the maps of simulated annual invasions for the period 2020-2023. Despite a significant negative trend in suitability, the species continues to expand its range due to increasing shipping activities, highlighting the detrimental impact of human activities on the basin. The framework is modular and easily transferable to allow for future extensions and applications to other target species, significantly expanding the reliability and depth of the AIS risk assessment. The dataset and scripts to reproduce the analysis are publicly available.
15-apr-2025
Inglese
Alien Species; ENM; MaxEnt; invasion dynamics; shipping traffic
CANU, DONATA
Università degli Studi di Trieste
File in questo prodotto:
File Dimensione Formato  
fianchini_3110_final.pdf

accesso aperto

Dimensione 4.92 MB
Formato Adobe PDF
4.92 MB Adobe PDF Visualizza/Apri

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.14242/202391
Il codice NBN di questa tesi è URN:NBN:IT:UNITS-202391