Using data calibration to reconcile outputs from different survey methods in long-term or large-scale studies.

Understanding the impact of management interventions on the environment over decadal and longer timeframes is urgently required. Longitudinal or large-scale studies with consistent methods are best practice, but more commonly, small datasets with differing methods are used to achieve larger coverage. Changes in methods and interpretation affect our ability to understand data trends through time or across space, so an ability to understand and adjust for such discrepancies between datasets is important for applied ecologists. Calibration or double sampling is the key to unlocking the value from disparate datasets, allowing us to account for the differences between datasets while acknowledging the uncertainties. We use a case study of livestock grazing impacts on riparian vegetation in southeastern Australia to develop a flexible and powerful approach to this problem. Using double sampling, we estimated changes in vegetation attributes over a 12-year period using a pseudo-quantitative visual method as the starting point, and the same technique plus point-intercept survey for the second round. The disparate nature of the datasets produced uncertain estimates of change over time, but accounting for this uncertainty explicitly is precisely the objective and highlights the need to look more closely at this very common problem in environmental management, as well as the potential benefits of the double sampling approach.

Towards open, reliable, and transparent ecology and evolutionary biology.

Unreliable research programmes waste funds, time, and even the lives of the organisms we seek to help and understand. Reducing this waste and increasing the value of scientific evidence require changing the actions of both individual researchers and the institutions they depend on for employment and promotion. While ecologists and evolutionary biologists have somewhat improved research transparency over the past decade (e.g. more data sharing), major obstacles remain. In this commentary, we lift our gaze to the horizon to imagine how researchers and institutions can clear the path towards more credible and effective research programmes.

Misinformation, internet honey trading and beekeepers drive a plant invasion.

Biological invasions are a major human induced global change that is threatening global biodiversity by homogenizing the world's fauna and flora. Species spread because humans have moved species across geographical boundaries and have changed ecological factors that structure ecosystems, such as nitrogen deposition, disturbance, etc. Many biological invasions are caused accidentally, as a byproduct of human travel and commerce driven product shipping. However, humans also have spread many species intentionally because of perceived benefits. Of interest is the role of the recent exponential growth in information exchange via internet social media in driving biological invasions. To date, this has not been examined. Here, we show that for one such invasive species, goldenrod, social networks spread misleading and incomplete information that is enhancing the spread of goldenrod invasions into new environments. We show that the notion of goldenrod honey as a "superfood" with unsupported healing properties is driving a demand that leads beekeepers to produce goldenrod honey. Social networks provide a forum for such information exchange and this is leading to further spread of goldenrod in many countries where goldenrod is not native, such as Poland. However, this informal social information exchange ignores laws that focus on preventing the further spread of invasive species and the strong negative effects that goldenrod has on native ecosystems, including floral resources that negatively impact honeybee performance. Thus, scientifically unsupported information on "superfoods" such as goldenrod honey that is disseminated through social internet networks has real world consequences such as increased goldenrod invasions into novel geographical regions which decreases native biodiversity.

Useful surrogates of soil texture for plant ecologists from airborne gamma-ray detection.

Plant ecologists require spatial information on functional soil properties but are often faced with soil classifications that are not directly interpretable or useful for statistical models. Sand and clay content are important soil properties because they indicate soil water-holding capacity and nutrient content, yet these data are not available for much of the landscape. Remotely sensed soil radiometric data offer promise for developing statistical models of functional soil properties applicable over large areas. Here, we build models linking radiometric data for an area of 40,000 km2 with soil physicochemical data collected over a period of 30 years and demonstrate a strong relationship between gamma radiometric potassium (40K), thorium (²³²Th), and soil sand and clay content. Our models showed predictive performance of 43% with internal cross-validation (to held-out data) and ~30% for external validation to an independent test dataset. This work contributes to broader availability and uptake of remote sensing products for explaining patterns in plant distribution and performance across landscapes.

Quantifying ecosystem quality by modeling multi-attribute expert opinion.

The evaluation of ecosystem quality is inherently subjective, requiring decisions about which variables to notice or measure, and how these variables are integrated into a coherent evaluation. Despite the central role of human judgment, few evaluation methods address the subjectivity that is inherent in their design. There are, however, advantages to directly using opinion to create an expert system where the metric is constructed around opinion data. These advantages include stakeholder inclusion and the encouragement of a dialogue of data-driven criticism rather than subjective counter-opinion. We create an expert system to express the quality of a grassland ecosystem in Australia. We use an ensemble of bagged regression trees trained on calibrated expert preference data, to model the perceived quality of this grassland using a set of eight site variables as inputs. The model provides useful predictions of grassland quality, producing predictions similar to real expert evaluations of independent synthetic test sites not used to train the model. We apply the model to real grassland sites ranging from pristine to highly degraded, and confirm that our model orders the sites according to their degree of modification. We demonstrate that the use of too few experts produces relatively poor results, and show that for our problem the use of data from over twenty experts is appropriate. The scaling approach we used to calibrate between-expert data is shown to be an appropriate mechanism for aggregating the opinions of multiple experts. The resultant model will be useful in many contexts, and can be used by managers as a tool to evaluate real sites. It can also be integrated into ecological models of change as a means of evaluating predicted changes, for example, as a measure of utility when combined with cost estimates. The basic approach demonstrated here is applicable to any ecosystem, and we discuss the opportunities and limitations of its wider use.

Public investment does not crowd out private supply of environmental goods on private land.

In landscapes where private land tenure is prevalent, public funds for ecological landscape restoration are sometimes spent subsidising the revegetation of cleared land, and the protection of remnant vegetation from livestock. However, the total area treated may be unclear because such projects are not always recorded, and landholders may undertake similar activities without subsidisation. In the absence of empirical data, in the state of Victoria, Australia, a reporting assumption has been employed that suggests that wholly privately funded sites match publicly subsidised sites on a hectare for hectare basis (a so-called "x2" assumption). Conversely, the "crowding out" theory of investment in public goods such as environmental benefits suggests that public investment may supplant private motivation. Using aerial photography we mapped the extent of revegetation, native vegetation fencing and restoration on 71 representative landholdings in rural south-eastern Australia. We interviewed each landholder and recorded the age and funding model of each site. Contrary to the local "x2" reporting assumption, about 75% of the total area of the 412 sites was from subsidised sites, and that proportion was far higher for the period after 1997. However, rather than displacing unsubsidised activity, our modelling showed that landholders who had recently been subsidised for a project were more likely to have subsequently completed unsubsidised work. This indicates that, at least in terms of medium-term economic impact, the large increase in public subsidies did not diminish privately funded activity, as might be expected according to the theory of crowding out.

Biocrust morphogroups provide an effective and rapid assessment tool for drylands.

Biological soil crusts (biocrusts) occur across most of the world's drylands and are sensitive indicators of dryland degradation. Accounting for shifts in biocrust composition is important for quantifying integrity of arid and semi-arid ecosystems, but the best methods for assessing biocrusts are uncertain. We investigate the utility of surveying biocrust morphogroups, a reduced set of biotic classes, compared to species data, for detecting shifts in biocrust composition and making inference about dryland degradation.We used multivariate regression tree (MRT) analyses to model morphogroup abundance, species abundance and species occurrence data from two independent studies in semi-arid open woodlands of south-eastern Australia. We advanced the MRT method with a 'best subsets' model selection procedure, which improved model stability and prediction.Biocrust morphogroup composition responded strongly to surrogate variables of ecological degradation. Further, MRT models of morphogroup data had stronger explanatory power and predictive power than MRT models of species abundance or occurrence data. We also identified morphogroup indicators of degraded and less degraded sites in our study region.Synthesis and applications. Sustainable management of drylands requires methods to assess shifts in ecological integrity. We suggest that biocrust morphogroups are highly suitable for assessment of dryland integrity because they allow for non-expert, rapid survey and are informative about ecological function. Furthermore, morphogroups were more robust than biocrust species data, showed a strong response to ecological degradation and were less influenced by environmental variation, and models of morphogroup abundance were more predictive.

Examining change over time in habitat attributes using Bayesian reinterpretation of categorical assessments.

Prospects for evaluating effects of vegetation restoration have long been limited by availability of appropriately sensitive baseline data. Data that are typically collected to justify investment in restoration are rarely suitable for estimating subsequent change over time, but given how commonly such data are collected, can they contribute something to learning about ecological change over time? We compared vegetation and habitat data from a quantitative reassessment of 25 habitat restoration sites seven years after they were initially assessed using a semiquantitative, categorical scoring system. Our aim was to estimate the change at sites between the first, semiquantitative survey and a second, quantitative survey. We treated the initial values as effectively unknown and used Bayesian models to infer plausible values using three different informative prior distributions, variously comprising the initial site assessments and modeled values from a statewide data set. We successfully constructed models of change over time between the two surveys, and regardless of which prior model was implemented, our data analysis suggested that cover of exotic species was reduced, but canopy cover, the cover of organic litter, and the length of fallen logs were all increased after the seven-year period. A small increase in the mean number of large-diameter trees was likely due to initial measurement error. Site fertility and canopy cover were important covariates in explaining the magnitude of change in total log length. Sites with higher canopy cover decreased more in weed cover and increased more in litter cover. Our approach could be used to retrospectively analyze any ordinal data set where there is a scoring logic that can be interpreted quantitatively. Data sets where treatment contrasts and untreated controls exist will be particularly valuable for testing the utility of our approach. While this novel approach should prove a useful analytical complement to genuine longitudinal monitoring and space-for-time surveys, it is no substitute for initiation of learning about management effectiveness using data from purposefully designed and measured surveys.