Configuration and Evaluation of Models for Ecological Systems the Case of Distribution of Koala

Due to environmental issues, species distribution modeling (SDM) has been applied in environment protection strategy decisions. Configurations and evaluation models are considerable challenges in SDM. There has been limited literature discussing the configuration of parameters in detail for ecological models. With the increased use of SDM, there is a need for more research on parameter configurations and evaluation methods in these systems. It is difficult to simulate a “best” model if the parameter configurations of models are not taken into consideration. This study aims to analyze the configuration influences on the Koala distribution model. The Koala distribution is a case study in this paper; several algorithms, namely Artificial Neural Networks (ANNs), Multivariate Adaptive Regression Splines (MARS), and Maxent, are used for Koala distribution modeling. Four evaluation methods, namely TSS, PPV, Kappa, and AUC, are used in this research study. This study analyzes the results between configuration setting and evaluation methods and discusses the selected evaluation methods’ performance, to discover their effect and results to improve modeling of the Koala distribution. Findings demonstrate that the parameter setting and evaluation methods of SDM are essential.