关键词: MaxEnt模型, 模型复杂度, 参数优化, 柽柳属, 物种潜在分布, 生态恢复

Abstract: MaxEnt model is one of the most powerful tools for predicting and explaining species distribution under different environmental conditions due to its openness and minimal parameter requirements. The complexity of a distribution model is a key factor affecting the accuracy of model predictions. Therefore, optimizing the parameters that influence model complexity is of great significance for reducing model uncertainty and enhancing prediction accuracy. Tamarix is a shrub genus widely distributed in northern China and serves as flagship species in “Southern Red (Carpinus) and Northern Willow (Tamarix)” ecological restoration project along coasts of China. With the genus Tamarix as the research object, we optimized the complexity of the MaxEnt model by combining the Akaike Information Criterion (AIC) with accuracy validation, screened out the optimal parameter combination, and simulated and evaluated the potential distribution pattern of Tamarix in China. The results showed that: (1) A frequency multiplication (RM) of 1.6 and a feature combination (FC) of LQPT constituted the optimal parameter combination for complexity optimization in fitting the potential distribution of Tamarix. The prediction results showed high credibility, with an average Area Under Curve (AUC) value of 0.82. (2) Considering the tradeoff between mean AUC and time duration, 0.5 was chosen as the step length for the optimization of the regularization multiplier. (3) Comparing with the default model, the optimized model resulted in a reduction in the area of unsuitable and lowly suitable regions for the distribution of Tamarix, while the area of moderately suitable and highly suitable regions expanded. This change is attributed to the reduced sensitivity of the optimized model to the training data. This study can provide a valuable reference for the parameter optimization of Tamarix distribution model, as well as the research on its potential distribution and the management of ecological restoration and conservation.

Key words: MaxEnt model, model complexity, parameter optimization, Tamarix, potential species distribution, ecological restoration