MaxEnt is a popular species distribution model which has been widely used in the last few years. Some case studies on endangered species, invasive plants, and simulated data, reported that the MaxEnt model is capable of accurately predicting species distribution even when trained with a small sample of occurrence data. However, only few studies focused on the effect of sample size on the prediction distribution of well-studied species. Moreover, the variation of study range size might further affect the accuracy of the model. The black-and-white snub-nosed monkey (Rhinopithecus bieti) occurs in a narrow region between the Yangtze River and Mekong River in southwest China. There are about 18-20 groups of R. bieti, living in the region and the population distribution of 16 of them is fairly well-known from long-term field observations performed by the Institute of Eastern-Himalaya Biodiversity Research team. A minimum bounding rectangle covering the occurrence data from these 16 groups R. bieti was defined as the minimum study range. Buffer zones of 25 to150 km at 25 km intervals from the minimum study range were drawn and used as alternative study range. In order to examine the effect of sample size and study range on the MaxEnt model prediction capabilities, we compiled different training datasets using occurrence data from 11 groups with sample sizes ranging from 1 to 10 groups, and background data sampled within each of the seven buffer ranges. Occurrence data from the remaining 5 groups were used as presence data for independent test. The Area Under the Curve (AUC) value of each resulting model increased with increasing sample size and study range. Our findings demonstrate how sample size influences the MaxEnt model prediction of species distribution. However, AUC values might be overestimated by an enlarged study range. Results indicated that background data should be sampled from the neighborhood of presence data. Furthermore, to predict the whole potential habitat of a species, presence data should cover the ecological gradient of this species.

MaxEnt模型是过去几年最为流行的物种分布预测模型之一。针对一些濒危物种、入侵种和模拟数据的研究表明，MaxEnt模型均能在小样本的分布数据下得到较准确的预测结果。此外，研究范围的变化也会影响MaxEnt模型的构建。 然而，基于动物的实际分布数据来评估MaxEnt模型的研究甚少。 我们以黑白仰鼻猴 (Rhinopithecus bieti)为例，以11个猴群的分布数据为训练数据（样本量从1到10个猴群），在不同研究范围内构建MaxEnt模型，通过其它5个的猴群分布数据验证，分析样本量和研究范围变化对模型准确度产生的影响。 结果表明，随样本量和研究范围增大，MaxEnt模型准确度及稳定性都有增加。 此外，研究范围变化对模型准确度有一定影响。 应用Maxent进行物种分布预测时，训练数据应尽可能涵盖该物种可能出现的全部环境梯度。构建模型所需的背景数据点选择，应与建模使用的物种出现点形成有效对照。