Abstract:Biological records are often the data of choice for training predictive species distribution models (SDMs), but spatial sampling bias is pervasive in biological records data at multiple spatial scales and is thought to impair the performance of SDMs. We simulated presences and absences of virtual species as well as the process of recording these species to evaluate the effect on species distribution model prediction performance of (1) spatial bias in training data, (2) sample size (the average number of observ… Show more
“…The 22 species in LS revealed similar frequencies of phenological responses to latitude, but only on a subset of species’ LHTs. We agree with Gaul et al (2020) that to improve understanding, it is better to tolerate small imprecisions over information loss. We are grateful to LS for pinpointing errors in our paper and stimulating further analyses.…”
Larsen & Shirey (2020) criticised our analysis of latitudinal changes in butterfly phenology on the grounds of improper data management. We admit some imprecisions, but show that stringent reanalyses did not change the overall results. We also show that unreasonable treatment of data may result in critical information loss.
“…The 22 species in LS revealed similar frequencies of phenological responses to latitude, but only on a subset of species’ LHTs. We agree with Gaul et al (2020) that to improve understanding, it is better to tolerate small imprecisions over information loss. We are grateful to LS for pinpointing errors in our paper and stimulating further analyses.…”
Larsen & Shirey (2020) criticised our analysis of latitudinal changes in butterfly phenology on the grounds of improper data management. We admit some imprecisions, but show that stringent reanalyses did not change the overall results. We also show that unreasonable treatment of data may result in critical information loss.
“…This is in part because there are automated pipelines for estimating IUCN Red List status from distribution data (e.g., Dauby et al, 2017) but also because the measures of range size favoured by the IUCN is intentionally insensitive to data quantity (IUCN, 2022b). We would expect even more pronounced effects if we carried out analyses more sensitive to data quantity (e.g., species distribution models; Gaul et al, 2020).…”
Section: Small Herbaria Are Essential For Accurate Threat Assessmentsmentioning
Societal Impact StatementHerbaria can be considered plant libraries, each holding collections of dried specimens documenting plant diversity in space and time. For many plant species, these are our only evidence of their existence and the only means of assessing their conservation status. Specimens in all herbaria, especially those in small and often under‐resourced herbaria in megadiverse countries, are key to achieving accurate estimates of the conservation status of the world's plant species. They are also part of a country's shared heritage and critical contributions to knowledge of the world's diversity.Summary
Internationally agreed targets to assess the conservation status of all plant species rely largely on digitised distribution data from specimens held in herbaria.
Using taxonomically curated databases of herbarium specimen data for the mega‐diverse genera Begonia (Begoniaceae) and Solanum (Solanaceae) occurring in Peru, we test the value added from including data from local herbaria and herbaria of different sizes on estimations of threat status using International Union for Conservation of Nature (IUCN) Red List criteria.
We find that the Global Biodiversity Information Facility (GBIF) has litter data from Peruvian herbaria and adding these data influences the estimated threat status of these species, reducing the numbers of Critically Endangered and Vulnerable species in both genera. Similarly, adding data from small‐ and medium‐sized herbaria, whether in‐country or not, also improves the accuracy of threat assessments.
A renewed focus on resourcing and recognising the contribution of small and in‐country herbaria is required if we are to meet internationally agreed targets for plant conservation. We discuss our case study in the broader context of democratising and increasing participation in global botanical science.
“…VS have been widely used to study various aspects of species distribution models (Miller, 2014), such as: testing various sampling designs (Albert et al, 2010), methods for sampling bias corrections (Fourcade et al, 2014; Inman et al, 2021; Ranc et al, 2017; Stolar & Nielsen, 2015; Varela et al, 2014), different modelling techniques (Elith & Graham, 2009; Hirzel et al, 2001; Qiao et al, 2015), combinations of sampling design and modelling algorithms (Fernandes et al, 2018; Gaul et al, 2020), testing approaches to account for spatial autocorrelation (Dormann et al, 2007), to deal with multicollinearity (Dormann et al, 2013), estimating the effect of collinearity (De Marco & Nóbrega, 2018) or error types in abundance trends (Nuno et al, 2015).…”
The virtual species (VS) and virtual ecologist (VE) approaches are useful tools that allow testing different methodological aspects of species distribution modelling. However, methods used to generate VS so far lack solutions that can ensure a high degree of biological realism, taking into account spatial and temporal variability of population densities.
We have developed a method for generating dynamic VS that can reconstruct their living prototypes in a realistic way. The framework consists of fitting a spatiotemporal model to real abundance data, generating a VS population from that model over the entire study area and spanning the whole study period, calibrating the VS, and obtaining the VE data by sampling from the VS. The effectiveness of the developed approach has been illustrated by data from large‐scale and long‐term bird abundance monitoring, using the whinchat Saxicola rubetra as a study system. We evaluated how well the spatiotemporal model can reconstruct the ‘true’ system by comparing response curves and population trends between those used to generate the VS (i.e. what constitutes the ‘truth’) and those estimated from the replicated instances of VE data. In addition, we performed a sensitivity analysis to test how the varying sampling effort affects the accuracy of trend estimation.
The synthetic VE data thoroughly reconstructed the real monitoring data. Response curves from generalized additive mixed models (GAMMs), fitted to these two types of data, showed high concordance, as indicated by the 95% confidence intervals of coverage probability of 87.7%–99.8% (mean 96.9%). The population trend estimated from the VE data accurately reconstructed the ‘true’ trend calculated from VS (coverage probability: 82.3%).
The proposed method for generating VS and VE data by reverse engineering of the spatiotemporal ecological model reproduces well the properties of the original system, substantially increasing the ecological realism of simulation results. The method may have further applications in evaluating various modelling techniques used to study species range dynamics, where real‐world properties are of particular importance, like conservation and invasion biology or climate change impact assessment.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.