Machine learning methods without tears: a primer for ecologists

Machine learning methods, a family of statistical techniques with origins in the field of artificial intelligence, are recognized as holding great promise for the advancement of understanding and prediction about ecological phenomena. These modeling techniques are flexible enough to handle complex problems with multiple interacting elements and typically outcompete traditional approaches (e.g., generalized linear models), making them ideal for modeling ecological systems. Despite their inherent advantages, a review of the literature reveals only a modest use of these approaches in ecology as compared to other disciplines. One potential explanation for this lack of interest is that machine learning techniques do not fall neatly into the class of statistical modeling approaches with which most ecologists are familiar. In this paper, we provide an introduction to three machine learning approaches that can be broadly used by ecologists: classification and regression trees, artificial neural networks, and evolutionary computation. For each approach, we provide a brief background to the methodology, give examples of its application in ecology, describe model development and implementation, discuss strengths and weaknesses, explore the availability of statistical software, and provide an illustrative example. Although the ecological application of machine learning approaches has increased, there remains considerable skepticism with respect to the role of these techniques in ecology. Our review encourages a greater understanding of machin learning approaches and promotes their future application and utilization, while also providing a basis from which ecologists can make informed decisions about whether to select or avoid these approaches in their future modeling endeavors.     

Distinguishing the roles of dispersal in diversity maintenance and in diversity limitation

Considerable recent research effort has gone into studying how dispersal might affect the diversity of local communities. While this general topic has received attention from theoretical and empirical ecologists alike, the research focus has differed between the two groups; theoretical ecologists have explored the role of dispersal in the maintenance of diversity within local communities, whereas empirical ecologists have sought to quantify the role of dispersal in limiting local diversity. We argue that there is no necessary relationship between these two components of diversity and we therefore need to develop empirical approaches to quantify the dispersal-maintained component of diversity, as well as the dispersal-limited component. We develop one such approach in this paper, based on a quantitative partitioning of the natural regeneration within intact communities onto different sources of recruits (local communityvs. dispersal across different spatial or temporal scales).     

The culture of marine ecology

Marine algal ecology today faces many of the same problems as ecology in general, e.g. lack of generality of experimental results, the difficulty of making long-term predictions, and an apparent lack of agreement as to what constitutes the proper or `acceptable' way of doing this particular component of science. These problems, if real, affect marine algal ecology everywhere but, in different geographical areas, specific problems also occur; science in parts of Asia has some problems different from those in other parts of the world. Since its inception, research in marine algal ecology has been motivated by many factors, ranging from traditional needs, to curiosity, to survival, to new technology, and economic needs. Each of these has shaped the questions that have been asked by, and the level of support society has been willing to supply to, ecology. For example the requisites of tradition pushed marine ecology to ask questions about food and ceremonial biota, and our fears today about loss of biota are pushing for answers to questions about the means of preserving biodiversity. The limitations of many marine ecological studies have been pointed out by different individuals. Their comments have been valuable in forcing us to examine what we are doing as marine ecologists, and how we are doing it. Ecology, and marine algal ecology with it, has been accused of carrying out small-scale studies that have no greater generality than the sites at which the studies were done, and of using statistical procedures that are wrong or inappropriate; also, there is disagreement within the ecological community of how to correct for these `faults'. Some of the problems arise due to the nature of our particular science, e.g. working with organisms with differing genetic makeup and sensitivity of experimental results to small changes in initial conditions. Other problems are more likely due to the individuals doing the science, e.g. an inability to be an `expert' on all areas of knowledge required for a modern ecologist (taxonomy, experimental design, data analysis, etc.), and perhaps an unwillingness to recognize that in some instances different methods of data analysis are applicable and valid. As ecologists, we must come to grip with these problems, both for the sake of our science, and for our own sake as practicing ecologists.     

Robert H. Whittaker (1920–1980): The man and his work

R. H. Whittaker enlivened many fields within ecology, systematics and evolution with his insights. Perhaps his most significant contributions to ecology lie in the development of the theories and methods of gradient analysis. Through the verification of the individualistic hypothesis with field data from many regions, and the subsequent development and dissemination of methods for studying species distributions along continua, he helped replace the Clementsian paradigm with a Gleasonian one. His extensive field data on primary production, nutrient cycling patterns and species diversity established new standards for documentation in synecology and helped clarify the basis for site-to-site variation in these variables. Through his broad command of the ecological literature, his writings and his contact with ecologists throughout the world he fostered international understanding of the diversity of approaches to vegetation study.     

Conducting robust ecological analyses with climate data

Although the number of studies discerning the impact of climate change on ecological systems continues to increase, there has been relatively little sharing of the lessons learnt when accumulating this evidence. At a recent workshop entitled ‘Using climate data in ecological research’ held at the UK Met Office, ecologists and climate scientists came together to discuss the robust analysis of climate data in ecology. The discussions identified three common pitfalls encountered by ecologists: 1) selection of inappropriate spatial resolutions for analysis; 2) improper use of publically available data or code; and 3) insufficient representation of the uncertainties behind the adopted approach. Here, we discuss how these pitfalls can be avoided, before suggesting ways that both ecology and climate science can move forward. Our main recommendation is that ecologists and climate scientists collaborate more closely, on grant proposals and scientific publications, and informally through online media and workshops. More sharing of data and code (e.g. via online repositories), lessons and guidance would help to reconcile differing approaches to the robust handling of data. We call on ecologists to think critically about which aspects of the climate are relevant to their study system, and to acknowledge and actively explore uncertainty in all types of climate data. And we call on climate scientists to make simple estimates of uncertainty available to the wider research community. Through steps such as these, we will improve our ability to robustly attribute observed ecological changes to climate or other factors, while providing the sort of influential, comprehensive analyses that efforts to mitigate and adapt to climate change so urgently require.     

Towards a behavioral ecology of ecological landscapes

Recent developments in landscape-level ecological modeling rest upon poorly understood behavioral phenomena. Surprisingly, these phenomena include animal movement and habitat selection, two areas with a long history of study in behavioral ecology. A major problem in applying traditional behavioral ecology to landscape-level ecological problems is that ecologists and behaviorists work at very different spatial scales. Thus a behavioral ecology of ecological landscapes would strive to overcome this inopportune differential in spatial scales. Such a landscape-conscious behavioral undertaking would not only establish more firmly the link between behavior and ecological systems, but also catalyze the study of basic biological phenomena of Interest to behaviorists and ecologists alike.     

Alternative statistical approaches to the use of data as evidence for hypotheses in human behavioral ecology

In their ambitious Evolutionary Anthropology paper, Winterhalder and Smith 1 review the history, theory, and methods of human behavioral ecology (HBE). In establishing how HBE differs from traditional approaches within sociocultural anthropology, they and others laud its hypothetical‐deductive research method. 1 - 3 Our aim is to critically examine how human behavioral ecologists conduct their research, specifically how they analyze and interpret data as evidence for scientific hypotheses. Through computer simulations and a review of empirical studies of human sex ratios, we consider some limitations of the status quo and present alternatives that could strengthen the field. In particular, we suggest that because human behavioral ecologists often consider multiple hypotheses, they should use statistical approaches that can quantify the evidence in empirical data for competing hypotheses. Although we focus on HBE, the principles of this paper apply broadly within biological anthropology.     

β-多样性的研究：应用多元回归和典范分析研究生态方差的分解

 β-多样性刻画了地理区域中不同地点物种组成的变化,是理解生态系统功能、生物多样性保护和生态系统管理的一个重要概念。该文介绍了如何从群落组成,相关环境和空间数据角度去分析β-多样性。β-多样性可以通过计算每个地点的多样性指数,进而对可能解释点之间差异的因子所作的假设进行检验来研究。也可以将涵盖所有点的群落组成数据表看作是一系列环境和空间变量的函数,进行直接分析。这种分析应用统计方法将多样性指数或群落组成数据表的方差进行关于环境和空间变量的分解。该文对方差分解进行阐述。方差分解是利用环境和空间变量来解释β-多样性的一种方法。β-多样性是生态学家用来比较不同地点或同一地点不同生态群落的一种手段。方差分解就是将群落组成数据表的总方差无偏分解成由各个解释变量所决定的子方差。调整的决定系数提供了针对多元回归和典范冗余分析的无偏估计。 方差分解后,可以对感兴趣的方差解释部分进行显著性检验,同时绘出基于这部分方差解释的预测图。     

An evaluation of the state of spatial point pattern analysis in ecology

Over the last two decades spatial point pattern analysis (SPPA) has become increasingly popular in ecological research. To direct future work in this area we review studies using SPPA techniques in ecology and related disciplines. We first summarize the key elements of SPPA in ecology (i.e. data types, summary statistics and their estimation, null models, comparison of data and models, and consideration of heterogeneity); second, we review how ecologists have used these key elements; and finally, we identify practical difficulties that are still commonly encountered and point to new methods that allow current key questions in ecology to be effectively addressed. Our review of 308 articles published over the period 1992–2012 reveals that a standard canon of SPPA techniques in ecology has been largely identified and that most of the earlier technical issues that occupied ecologists, such as edge correction, have been solved. However, the majority of studies underused the methodological potential offered by modern SPPA. More advanced techniques of SPPA offer the potential to address a variety of highly relevant ecological questions. For example, inhomogeneous summary statistics can quantify the impact of heterogeneous environments, mark correlation functions can include trait and phylogenetic information in the analysis of multivariate spatial patterns, and more refined point process models can be used to realistically characterize the structure of a wide range of patterns. Additionally, recent advances in fitting spatially‐explicit simulation models of community dynamics to point pattern summary statistics hold the promise for solving the longstanding problem of linking pattern to process. All these newer developments allow ecologists to keep up with the increasing availability of spatial data sets provided by newer technologies, which allow point patterns and environmental variables to be mapped over large spatial extents at increasingly higher image resolutions.     

What can observational data reveal about metacommunity processes?

A key challenge for community ecology is to understand to what extent observational data can be used to infer the underlying community assembly processes. As different processes can lead to similar or even identical patterns, statistical analyses of non‐manipulative observational data never yield undisputable causal inference on the underlying processes. Still, most empirical studies in community ecology are based on observational data, and hence understanding under which circumstances such data can shed light on assembly processes is a central concern for community ecologists. We simulated a spatial agent‐based model that generates variation in metacommunity dynamics across multiple axes, including the four classic metacommunity paradigms as special cases. We further simulated a virtual ecologist who analysed snapshot data sampled from the simulations using eighteen output metrics derived from beta‐diversity and habitat variation indices, variation partitioning and joint species distribution modelling. Our results indicated two main axes of variation in the output metrics. The first axis of variation described whether the landscape has patchy or continuous variation, and thus was essentially independent of the properties of the species community. The second axis of variation related to the level of predictability of the metacommunity. The most predictable communities were niche‐based metacommunities inhabiting static landscapes with marked environmental heterogeneity, such as metacommunities following the species sorting paradigm or the mass effects paradigm. The most unpredictable communities were neutral‐based metacommunities inhabiting dynamics landscapes with little spatial heterogeneity, such as metacommunities following the neutral or patch sorting paradigms. The output metrics from joint species distribution modelling yielded generally the highest resolution to disentangle among the simulated scenarios. Yet, the different types of statistical approaches utilized in this study carried complementary information, and thus our results suggest that the most comprehensive evaluation of metacommunity structure can be obtained by combining them.     

Diversity and biogeography of marine actinobacteria

The actinomycetes, although not all the Actinobacteria, are easy to isolate from the marine environment. However, their ecological role in the marine ecosystem is largely neglected and various assumptions meant there was little incentive to isolate strains for search and discovery of new drugs. However, the marine environment has become a prime resource in search and discovery for novel natural products and biological diversity, and marine actinomycetes turn out to be important contributors. Similarly, striking advances have been made in marine microbial ecology using molecular techniques and metagenomics, and actinobacteria emerge as an often significant, sometimes even dominant, environmental clade. Both approaches - cultivation methods and molecular techniques - are leading to new insights into marine actinobacterial biodiversity and biogeography. Very different views of actinobacterial diversity emerge from these, however, and the true extent and biogeography of this are still not clear. These are important for developing natural product search and discovery strategies, and biogeography is a hot topic for microbial ecologists.     

Illustrations and guidelines for selecting statistical methods for quantifying spatial pattern in ecological data

This paper aims to provide guidance to ecologists with limited experience in spatial analysis to help in their choice of techniques. It uses examples to compare methods of spatial analysis for ecological field data. A taxonomy of different data types is presented, including point‐ and area‐referenced data, with and without attributes. Spatially and non‐spatially explicit data are distinguished. The effects of sampling and other transformations that convert one data type to another are discussed; the possible loss of spatial information is considered.
Techniques for analyzing spatial pattern, developed in plant ecology, animal ecology, landscape ecology, geostatistics and applied statistics are reviewed briefly and their overlap in methodology and philosophy noted. The techniques are categorized according to their output and the inferences that may be drawn from them, in a discursive style without formulae. Methods are compared for four case studies with field data covering a range of types. These are: 1) percentage cover of three shrubs along a line transect; 2) locations and volume of a desert plant in a 1 ha area; 3) a remotely‐sensed spectral index and elevation from 10⁵ km² of a mountainous region; and 4) land cover from three rangeland types within 800 km² of a coastal region. Initial approaches utilize mapping, frequency distributions and variance‐mean indices. Analysis techniques we compare include: local quadrat variance, block quadrat variance, correlograms, variograms, angular correlation, directional variograms, wavelets, SADIE, nearest neighbour methods, Ripley's (t), and various landscape ecology metrics.
Our advice to ecologists is to use simple visualization techniques for initial analysis, and subsequently to select methods that are appropriate for the data type and that answer their specific questions of interest. It is usually prudent to employ several different techniques.     

Putting the "landscape" in landscape genetics

Landscape genetics has emerged as a new research area that integrates population genetics, landscape ecology and spatial statistics. Researchers in this field can combine the high resolution of genetic markers with spatial data and a variety of statistical methods to evaluate the role that landscape variables play in shaping genetic diversity and population structure. While interest in this research area is growing rapidly, our ability to fully utilize landscape data, test explicit hypotheses and truly integrate these diverse disciplines has lagged behind. Part of the current challenge in the development of the field of landscape genetics is bridging the communication and knowledge gap between these highly specific and technical disciplines. The goal of this review is to help bridge this gap by exposing geneticists to terminology, sampling methods and analysis techniques widely used in landscape ecology and spatial statistics but rarely addressed in the genetics literature. We offer a definition for the term "landscape genetics", provide an overview of the landscape genetics literature, give guidelines for appropriate sampling design and useful analysis techniques, and discuss future directions in the field. We hope, this review will stimulate increased dialog and enhance interdisciplinary collaborations advancing this exciting new field.     

Persistent Controversy in Statistical Approaches in Wildlife Sciences: A Perspective of Students

ABSTRACT The controversy over the use of null hypothesis statistical testing (NHST) has persisted for decades, yet NHST remains the most widely used statistical approach in wildlife sciences and ecology. A disconnect exists between those opposing NHST and many wildlife scientists and ecologists who conduct and publish research. This disconnect causes confusion and frustration on the part of students. We, as students, offer our perspective on how this issue may be addressed. Our objective is to encourage academic institutions and advisors of undergraduate and graduate students to introduce students to various statistical approaches so we can make well-informed decisions on the appropriate use of statistical tools in wildlife and ecological research projects. We propose an academic course that introduces students to various statistical approaches (e.g., Bayesian, frequentist, Fisherian, information theory) to build a foundation for critical thinking in applying statistics. We encourage academic advisors to become familiar with the statistical approaches available to wildlife scientists and ecologists and thus decrease bias towards one approach. Null hypothesis statistical testing is likely to persist as the most common statistical analysis tool in wildlife science until academic institutions and student advisors change their approach and emphasize a wider range of statistical methods.     

多样性排序:方法与实例

多样性已经成为生态学理论与实践中的一个很重要的概念 ,很多多样性指数被引进或设计出来 ,并被广泛地应用于生态系统管理以及环境监测与评价之中。然而 ,很多学者对其也提出了疑议 ,因为当用不同的多样性指数对一组群落进行比较时会得出不同的顺序。Patil和Taillie等人提出用含有一个参数的多样性指数族可以潜在地解决这个问题。本文对这个类型的 1 4种方法做了评述 ,并将其分为 4组。证明了组内各方法间是等价的 ,而不同组的各方法间是不等价的。并认为在进行多样性排序时 ,Patil和Taillie的“右尾和”方法应属首选 ,因为只需计算有限个数值 ,并且该方法判断了群落内多样性的高低。其次 ,Hulbert_Smith_Grassle的方法也值得推荐 ,因为该方法有直观而明确的意义 ,它是一个随机样本中期望包含的物种数目。作为实例 ,还将这些方法应用于北京东灵山地区的 3个森林群落多样性的比较     

New approaches to analyzing microbial biodiversity data

Modern molecular techniques have revealed an extraordinary diversity of microorganisms, most of which are as yet uncharacterized. This poses a major challenge to microbial ecologists: how can one compare the microbial diversity of different environments when the vast majority of microbial taxa are usually unknown? Three statistical approaches developed by ecologists and evolutionary biologists--parametric estimation, nonparametric estimation and community phylogenetics--are proving to be promising tools to meet this challenge. The combination of these tools with molecular biology techniques allow the rigorous estimation and comparison of microbial diversity in different environments.     

A conceptual guide to measuring species diversity

Three metrics of species diversity – species richness, the Shannon index and the Simpson index – are still widely used in ecology, despite decades of valid critiques leveled against them. Developing a robust diversity metric has been challenging because, unlike many variables ecologists measure, the diversity of a community often cannot be estimated in an unbiased way based on a random sample from that community. Over the past decade, ecologists have begun to incorporate two important tools for estimating diversity: coverage and Hill diversity. Coverage is a method for equalizing samples that is, on theoretical grounds, preferable to other commonly used methods such as equal-effort sampling, or rarefying datasets to equal sample size. Hill diversity comprises a spectrum of diversity metrics and is based on three key insights. First, species richness and variants of the Shannon and Simpson indices are all special cases of one general equation. Second, richness, Shannon and Simpson can be expressed on the same scale and in units of species. Third, there is no way to eliminate the effect of relative abundance from estimates of any of these diversity metrics, including species richness. Rather, a researcher must choose the relative sensitivity of the metric towards rare and common species, a concept which we describe as ‘leverage.' In this paper we explain coverage and Hill diversity, provide guidelines for how to use them together to measure species diversity, and demonstrate their use with examples from our own data. We show why researchers will obtain more robust results when they estimate the Hill diversity of equal-coverage samples, rather than using other methods such as equal-effort sampling or traditional sample rarefaction.     

A Natural View of Microbial Biodiversity within Hot Spring Cyanobacterial Mat Communities

This review summarizes a decade of research in which we have used molecular methods, in conjunction with more traditional approaches, to study hot spring cyanobacterial mats as models for understanding principles of microbial community ecology. Molecular methods reveal that the composition of these communities is grossly oversimplified by microscopic and cultivation methods. For example, none of 31 unique 16S rRNA sequences detected in the Octopus Spring mat, Yellowstone National Park, matches that of any prokaryote previously cultivated from geothermal systems; 11 are contributed by genetically diverse cyanobacteria, even though a single cyanobacterial species was suspected based on morphologic and culture analysis. By studying the basis for the incongruity between culture and molecular samplings of community composition, we are beginning to cultivate isolates whose 16S rRNA sequences are readily detected. By placing the genetic diversity detected in context with the well-defined natural environmental gradients typical of hot spring mat systems, the relationship between gene and species diversity is clarified and ecological patterns of species occurrence emerge. By combining these ecological patterns with the evolutionary patterns inherently revealed by phylogenetic analysis of gene sequence data, we find that it may be possible to understand microbial biodiversity within these systems by using principles similar to those developed by evolutionary ecologists to understand biodiversity of larger species. We hope that such an approach guides microbial ecologists to a more realistic and predictive understanding of microbial species occurrence and responsiveness in both natural and disturbed habitats.     

The spatial scaling of beta diversity

Beta diversity is an important concept used to describe turnover in species composition across a wide range of spatial and temporal scales, and it underpins much of conservation theory and practice. Although substantial progress has been made in the mathematical and terminological treatment of different measures of beta diversity, there has been little conceptual synthesis of potential scale dependence of beta diversity with increasing spatial grain and geographic extent of sampling. Here, we evaluate different conceptual approaches to the spatial scaling of beta diversity, interpreted from ‘fixed’ and ‘varying’ perspectives of spatial grain and extent. We argue that a ‘sliding window’ perspective, in which spatial grain and extent covary, is an informative way to conceptualize community differentiation across scales. This concept more realistically reflects the varying empirical approaches that researchers adopt in field sampling and the varying scales of landscape perception by different organisms. Scale dependence in beta diversity has broad implications for emerging fields in ecology and biogeography, such as the integration of fine‐resolution ecogenomic data with large‐scale macroecological studies, as well as for guiding appropriate management responses to threats to biodiversity operating at different spatial scales.     

Advancing ecological understandings through technological transformations in noninvasive genetics

Noninvasive genetic approaches continue to improve studies in molecular ecology, conservation genetics and related disciplines such as forensics and epidemiology. Noninvasive sampling allows genetic studies without disturbing or even seeing the target individuals. Although noninvasive genetic sampling has been used for wildlife studies since the 1990s, technological advances continue to make noninvasive approaches among the most used and rapidly advancing areas in genetics. Here, we review recent advances in noninvasive genetics and how they allow us to address important research and management questions thanks to improved techniques for DNA extraction, preservation, amplification and data analysis. We show that many advances come from the fields of forensics, human health and domestic animal health science, and suggest that molecular ecologists explore literature from these fields. Finally, we discuss how the combination of advances in each step of a noninvasive genetics study, along with fruitful areas for future research, will continually increase the power and role of noninvasive genetics in molecular ecology and conservation genetics.