4.2 The problem of ROC accuracy of the maximum entropy model for
the Mamukao River
In this study, the AUC value of the maximum entropy model ROC curve for
spring in the Mamukao River was only 0.571, significantly smaller than
the AUC values obtained from the other three types of models. Chen et al
demonstrated a correlation between the sample size of the species
(feature) distribution and the AUC value, indicating that a larger
sample size corresponds to a higher AUC value and better accuracy of the
predictive model (Chen et al. 2012). However, if the sample size for the
presence of the species (feature) is excessively large, such as in the
case of presence-only samples, the ROC curve cannot be applied due to
the absence of negative examples for measuring specificity at the
present moment (Anderson et al. 2003,Phillips et al. 2006). The Mamukao
River exhibits a significant presence of stable snow during the spring
season (see Figure x). Out of the total 1682 raster samples, only 276
were found to be non-accumulating, while the remaining 85.6%
represented accumulating raster samples. However, the lack of
specificity in our data may have contributed to a lower accuracy in the
prediction model for the spring maximum entropy of the Mamukao River.