摘要
栖息地模拟法是根据指示物种所需的物理生境条件确定河流流量,评价栖息地适宜度,为水生生物提供一个适宜的物理生境。与其他方法相比,栖息地模拟法考虑了生物本身对物理生境的要求,需要建立物种-生境评价指标。文章总结了目前用于描述物种—生境关系的栖息地适宜度评价方法,包括栖息地适宜度指数、多元统计方法、模糊逻辑方法、人工神经网络、对多物种和群落的统计分析,归纳了各个方法的进展和应用情况,重点分析了统计学方法在评估栖息地适宜度时的优势和不足。栖息地适宜度指数有二元、单变量、多变量3种格式,前2种方法只考虑单一因子,实际应用中多变量格式应用较多。多变量格式方法直观、所需数据容易获取、实际操作性强,是栖息地定量化的经典方法,也是目前应用最多的方法,然而其对专家经验依赖较多,主观性较强。近些年来,多元统计方法在栖息地适宜度评价方面的应用不断增加,它考虑物理变量之间的相互作用和相关性。模糊逻辑法在处理栖息地模拟中的不确定性方面具有优势,能更好地利用专家知识,更合理的处理建模过程中测量的不准确性和不确定性,同时也考虑了多个变量之间的相互作用,但是当考虑的变量数增加时,模糊规则的数量会迅速增加,给计算带来不便。人工神经网络能够隐性地找出响应变量和环境变量之间的复杂关系,但是其解释能力不足,并需要大量的实测数据对其进行训练,实际应用受到限制。通过排序分析或梯度分析可以对多物种和群落进行统计分析。这些方法都各有优劣,在实际应用中,应结合实际情况选择最合适的栖息地适宜度评价方法。
Based on the requirement of indicator species on physical habitat conditions, habitat suitability simulation is considered to be a most reliable method to determine the environmental flow. The purpose of evaluating habitat suitability is to provide a suitable habitat for aquatic organisms. Compared with other evaluation methods of environmental flow, habitat suitability simulation consid- ers the species-habitat relationships, and quantifies habitat requirements by evaluation indices. This paper aims to provide an over- view of the current habitat suitability assessment methods for describing the relationship between species and habitat. Methods in- cluding habitat suitability index, multivariate statistical methods, fuzzy logic method, artificial neural networks, statistical analysis of multi-species and community are introduced. The development and application of these methods are described in this paper. At the same time, the advantages and disadvantages of statistical methods are compared. Habitat suitability index is a classical method for quantifying habitat Binary format, univariate format and multivariate format are the three formats of habitat suitability index. The first two methods only consider single factor. Because of the intuitive, required less field survey data, and can be easily used, multi- variate format is the most popular method at present. However, it strongly relies on the experience of expert and tends to be subjec- tive easily. Multivariate statistical methods include multiple linear regressions, ridge regression method, principal component regres- sion, logistic regression, generalized linear model approach, generalized additive model method, and so on. More and more applica- tions of multivariate statistical methods in habitat suitability assessment have appeared in recent years. Multivariate statistical meth- ods considered the interaction between the physical factors. The fuzzy logic method is better dealing with uncertainties in the process of habitat simulation. It uses expert knowledge and dispose inaccuracies and uncertainties of measurement in the modeling process. Fuzzy logic method takes into account the interactions between multiple variables, but the number of fuzzy rules will increase rapidly when the number of variables increases. Artificial neural network presents advantages in identifying complicated relationship be- tween response variables and environment variables implicitly, but it suffers from inadequate explanatory power. Statistical analysis of multi-species and community can be conducted by sequencing analysis or gradient analysis. These methods both have their ad- vantages and disadvantages. The most suitable habitat suitability evaluation method should be chosen by combining actual condi- tions.
出处
《生态环境学报》
CSCD
北大核心
2013年第5期887-893,共7页
Ecology and Environmental Sciences
基金
环境保护部环保公益项目(201209029-4)
国家自然科学基金项目(51279220)
国家科技支撑计划课题(2011BAC12B02)
关键词
栖息地适宜度
单变量
多变量
统计学方法
模糊逻辑
回归
人工神经网络
habitat suitability
tmivariate method
multivariate method
statistical method
fuzzy logic
regression
artificial neuralnetworks
