Preface

Guy Woodward*; Alex J. Dumbrell┼; Donald J. Baird╬; Mehrdad Hajibabaei§, * Imperial College London, United Kingdom, ┼ University of Essex, United Kingdom, ╬ Environment Canada, Canada, § University of Guelph, Canada

Ecology is entering previously uncharted waters, in the wake of the huge growth in "Big Data" approaches that are beginning to dominate the field. Previously, the rate at which ecology advanced, especially when dealing with large scales and multispecies systems, was limited by the paucity of empirical data, which was often collected in a painstaking and labour-intensive manner by a few dedicated individuals. We are now entering a phase where the polar opposite situation is the norm and the new rate-limiting step is the ability to process the vast quantities of data that are being generated on an almost industrial scale and, more importantly, to interpret their ecological significance. This ecoinformatics revolution is happening simultaneously on many fronts: from the exponential increases in sequencing power using novel molecular techniques, to the increased capacity for remote sensing and high-resolution GIS, and the marshalling of huge volumes of metadata collected by both the scientific community and the rapidly swelling ranks of Citizen Scientists. This latter group will account for a sizeable portion of the Big Data that needs to be handled in future: Citizen Scientists are already starting to eclipse the capacity of official bodies to carry out large-scale and long-term routine data collection and biomonitoring, as the traditional boundaries between natural and social sciences and data ownership become evermore blurred. This democratisation and sharing of data among scientists, across disciplines, and with the lay public that has gone hand in hand with Big Data approaches is altering the very nature of scientific discourse in a profound manner, and in ways that we do not yet fully comprehend. This volume highlights three examples of some of the main Big Data trends and their potential to address the big questions in ecology in this new multidisciplinary era.

In addition to geospatial data series and large federated databases that are becoming commonplace, particularly in the field of biomonitoring and remote sensing, ecogenomics represents both one of the greatest informatics resources and one of the biggest emerging challenges in ecology. This is a rapidly growing field, and the recent explosion of molecular ecology embraces a plethora of terms that were barely on the horizon a decade ago, including metasystematics, metranscriptomics, and functional genomics, among others. These terms are entering the day-to-day lexicon of ecologists at an accelerating rate, and they are now frequently seen in both grant proposals and peer-reviewed publications. Even so, most ecological studies that use such approaches are still restricted to descriptive "fishing expeditions", rather than being used for explicit hypothesis generation or testing. Thus, although countless recent papers have revealed previously unguessed-at levels of biodiversity in even the most remote and hostile environments, particularly in the microbial world, very few have been couched in the rigorous hypothetico-deductive framework that is the bread and butter of the more established fields of mainstream ecology. In the light of this, it is critically important that in the heady rush to adopt Big Data approaches, we must take care to corroborate them with more traditional techniques, if only to enable a degree of handshaking before jettisoning obsolete technologies: otherwise, we run the risk of creating a schism in ecology that could lead to huge inefficiencies in the future, where we simply end up asking the same old questions but with different data, rather than truly advancing the field.

Before ecogenomics techniques and data are widely applied, they must therefore first provide credible evidence that they can do at least what existing techniques can do, but with added value. In the paper by Dafforn et al. (2014), the authors describe a case study that applies a metagenomics approach in estuarine ecosystems in Australia, while comparing the results with a parallel approach using traditional taxonomic analysis. The authors demonstrate convincingly that, despite the bioinformatics challenges, the ecogenomics approaches clearly provide data far more rapidly and efficiently, with benthic assemblages resolved at higher levels of taxonomic resolution. Perhaps even more importantly, though, they provide far stronger insights into the major environmental drivers of composition across a range of contrasting estuarine ecosystem conditions. In the second paper in the volume, by Gardham et al. (2014), a comparable metagenomics approach is applied to analyse mesocosm experiments studying the effects of metal pollution on freshwater benthic assemblages. When focused on the microbial community in particular, the exploratory power of multivariate approaches is greatly enhanced, in terms of exploring assemblage pattern-driver relationships, and this offers a huge new potential means of ecological indicator development. While metagenomics approaches are now being more widely applied in ecosystem research, both studies illustrate the opportunities created through the application of these new techniques, and also the emergence of the new generation of studies that are starting to embed Big Data into more explicitly hypothesis-focused frameworks. They also illustrate how Big Data processing requirements make it more crucial than ever to understand the complex analytical pathways that turn terabytes of DNA sequence into trustworthy ecological information.

The mushrooming of such sequence-based databases provides a vast and potentially invaluable resource for current and future generations of ecologists (Fig. 1), but increasingly concerns have been raised about the stringency of quality assurance and ground-truthing of the underlying data, which could seriously undermine the field if errors are being propagated unwittingly and repeatedly and on a potentially grand scale: i.e. there can be a world of difference between Big Data and Good Data. Notwithstanding these underlying issues, the rate of data generation that can now be achieved at relatively little cost is breathtaking and would have been inconceivable just a few years ago. It is also the sophistication of the data and the fact that multiple forms of information are being synthesised and compiled simultaneously that form the hallmarks of the most recent advances in this area. Collated databases containing outputs from multiple ecological studies will soon surpass single studies in terms of data breadth, and emerging molecular (e.g. next-generation sequencing) approaches will dwarf other ecological data in terms of depth and breadth of coverage of multispecies systems: in fact, it could be argued that this revolution has already happened (Fig. 1).

Figure 1 The number of DNA sequences contained within the GenBank database (the principal non-NGS sequence repository) as a function of time (open symbols). This acts as a proxy for publication quantity as you can't publish DNA sequences without first providing them to GenBank. These data include non-ecological DNA sequences. The solid symbol at the top is current number of DNA sequences contained within the MG-RAST repository, which only stores metagenome data, i.e. whole-community ecological data. Note the change in axis scales and how metagenomic approaches over the course of a couple of years has now produced more DNA sequences than the entire GenBank collection.

There is another major source of large ecological datasets that are becoming increasingly prevalent, which also present associated Big Data challenges, and this comes in the form of the outputs of large-scale multi-institutional (often multi-national) research programmes. Within the UK, the Natural Environment Research Council recently launched the Biodiversity and Ecosystem Service Sustainability Programme (BESS; 2011-2017), a multimillion pound investment that represents a UK-wide effort to characterise the links between biodiversity stocks and flows of ecosystem services across a broad spectrum of terrestrial and aquatic landscapes (http://www.nerc-bess.net/). This ambitious programme is led by Professor Dave Raffaelli (University of York), and the paper he leads in this volume (Raffaelli et al., 2014) highlights the Big Data challenges faced by BESS and the approaches being used to overcome these. Raffaelli et al. begin with lessons that can be learnt from history and draw the readers' attention to the pioneering International Biological Programme (IBP), which ran from 1964 to 1974 and was one of the first to attempt what we now call Big Data ecology. The IBP was in many ways too far ahead of its time, and it was beset by numerous problems resulting from its own huge complexity and scale of ambition, and it was abandoned long before its full potential could be realised. Raffaelli et al. highlight how half a century later we are only now finally starting to be able to deal with the size and scope of this style of research programme. It is only in the last few years that we have been able to wield the necessary tools for such a complex and challenging undertaking, and these were unfortunately lacking in the 1960s. To illustrate this, Raffaelli et al. explore the different approaches taken by the four main projects within BESS, which work to answer similar ecological questions,...