Presence-only data from museum and herbarium collection records are being used increasingly to model the world’s biodiversity patterns and to make important decisions about conservation priorities. Results are also used to test theoretical predictions about factors influencing diversity patterns and to target future field programs.

“Managing the use of the world’s resources and related conservation and biosecurity issues are closely entwined in social, economic and political frameworks.”

Despite this wide application, there has been no clear consensus on which analytical methods work best. Achieving such a consensus is difficult because of the effort involved in collating the extensive data sets required to test methods thoroughly.

The research in our paper resulted from a large international collaboration, where we were able to collate data for more than 200 species across six regions of the world. These data included not only species records and environmental data for modeling, but also data that had been independently collected to evaluate model predictions.

We think that part of the attraction of this paper is that it presents a very thorough and robust test of most currently used modeling methods on a scale that few if any researchers could carry out on their own.

Another important feature is that the methods tested include several only recently introduced to ecology. Some of these considerably out-performed more traditional techniques. Together, these features appeal to a wide range of people, including those working in both pure and applied branches of ecology.

It certainly is a new synthesis of knowledge, because of the breadth of the study and its robust comparison of both traditional and relatively new statistical modeling techniques. The knowledge we gained will encourage new research agendas as researchers across a broad spectrum of applied and basic sciences use these modeling techniques more routinely.

Many people in the world are concerned about changes in biodiversity and particularly those that are caused by human activities. However, there are limits to the time and money available to collect new data to describe species’ distributions. Given this tension, predictive models linking species records with relevant environmental or geographic data play a useful role in enabling us to map distributions across sometimes large areas of land and to predict distributions reliably in places we haven’t had time to visit.

To create useful predictions, it is important that we use modeling methods that suit the data, that make sense ecologically, and that are well-tested. Our testing of 16 modeling methods with diverse species data from six regions around the globe provides some clear guidance on which modeling methods perform best. The results will promote additional research to help managers and policymakers make better use of data as they address the biodiversity crisis.

This research was funded by NCEAS, the National Center of Ecological Analysis and Synthesis in California, in a working group proposed by A. Townsend Peterson and Craig Moritz. Some authors took part in the working groups that developed the experimental design, and others supplied data, developed models, or gave advice.

The biggest obstacles were coming to a common understanding of what we needed to achieve. The other hurdles were practical ones—finding appropriate data, identifying modeling experts, and other difficulties associated with organizing a large collaborative research program with 27 co-authors.

Data preparation is an enormous task! By the time we had finished, each modeler had produced about 750 sets of predictions per method, so the final analysis was a huge undertaking.

Members of the group have widely differing interests. Many of us are involved in the practical application of these models in supporting conservation management in our various countries or internationally. Some primarily work on the research side; the ecologists and evolutionary biologists are building knowledge about why species exist where they do.

Others are developing new modeling tools relevant not only to species modeling but also to a broad range of analytical problems. Our priorities include understanding the historic biogeography of species, conserving current biodiversity, and predicting and ameliorating the effects of invasive species and climate change on species distributions.

Many of us aim to keep working on methods we consider most promising, perhaps teaching others how to use them or applying them to problems we are asked to help solve. Our illustrations provide several examples.

Managing the use of the world’s resources and related conservation and biosecurity issues are closely entwined within social, economic, and political frameworks. It is important that we, as scientists, present methods that are as reliable, unbiased, and efficient as possible, so we can underpin decisions with the best estimates and most realistic assessments of uncertainties that we can produce. Our research is one small contribution to this goal.