Pseudo-absence selection for spatial distribution models (SDMs) is the subject of ongoing investigation. Numerous techniques continue to be developed, and reports of their effectiveness vary. Because the quality of presence and absence data is key for acceptable accuracy of correlative SDM predictions, determining an appropriate method to characterise pseudo-absences for SDM’s is vital. The main methods that are currently used to generate pseudo-absence points are: 1) randomly generated pseudo-absence locations from background data; 2) pseudo-absence locations generated within a delimited geographical distance from recorded presence points; and 3) pseudo-absence locations selected in areas that are environmentally dissimilar from presence points. There is a need for a method that considers both geographical extent and environmental requirements to produce pseudo-absence points that are spatially and ecologically balanced. We use a novel three-step approach that satisfies both spatial and ecological reasons why the target species is likely to find a particular geo-location unsuitable. Step 1 comprises establishing a geographical extent around species presence points from which pseudo-absence points are selected based on analyses of environmental variable importance at different distances. This step gives an ecologically meaningful explanation to the spatial range of background data, as opposed to using an arbitrary radius. Step 2 determines locations that are environmentally dissimilar to the presence points within the distance specified in step one. Step 3 performs K-means clustering to reduce the number of potential pseudo-absences to the desired set by taking the centroids of clusters in the most environmentally dissimilar class identified in step 2. By considering spatial, ecological and environmental aspects, the three-step method identifies appropriate pseudo-absence points for correlative SDMs. We illustrate this method by predicting the New Zealand potential distribution of the Asian tiger mosquito (Aedes albopictus) and the Western corn rootworm (Diabrotica virgifera virgifera).

This study provides a bio-economic assessment of the global climate suitability and probabilistic crop-loss estimates attributable to wheat leaf rust. We draw on a purpose-built, spatially explicit, ecoclimatic suitability model for wheat leaf rust to estimate that 94.4% of global wheat production is vulnerable to the disease. To reflect the spatiotemporal variation in leaf rust losses, we used a probabilistic approach to estimate a representative rust loss distribution based on long-term, state-level annual U.S. loss estimates. Applying variants of this representative loss distribution to selected wheat production areas in 15 epidemiological zones throughout the world, we project global annual average losses of 8.6 million metric tons of grain for the period 2000 to 2050 based on a conservative, baseline scenario, and 18.3 million metric tons based on a high-loss scenario; equivalent to economic losses ranging from $1.5 to $3.3 billion per year (2016 U.S. prices). Even the more conservative baseline estimate implies that a sustained, worldwide investment of $50.5 million per year in leaf rust research is economically justified.

Johansson, M. U., S. D. Senay, E. Creathorn, H. Kassa, and K. Hylander. 2019. Change in heathland fire sizes inside vs. outside the Bale Mountains National Park, Ethiopia, over 50 years of fire-exclusion policy: lessons for REDD+. Ecology and Society 24(4):26. https://doi.org/10.5751/ES-11260-240426

We assembled 3,175 geo-tagged occurrences of fall armyworm worldwide and used that data in conjunction with information about the physiological requirements of the pest to spatially assess its global climate suitability. Our analysis indicates that almost the entire African maize crop is grown in areas with climates that support seasonal infestations of the insect, while almost 92% of the maize area supports year-round growth of the pest. In contrast, rich-country maize production largely occurs in temperate areas where only 2.3% of the area may allow the pest to survive year-round, although still subject to worrisome seasonal risks. This means the African maize crop is especially susceptible to damaging infestation from fall armyworm, on par with the risk exposure to this pest faced by maize producers throughout Latin America. We show that the maize grown in Africa is also especially vulnerable to infestations from a host of other crop pests. Our multi-peril pest risk study reveals that over 95% of the African maize area deemed climate suitable for fall armyworm, can also support year-round survival of at least three or more pests. The spatial concurrence of climatically suitable locations for these pests raises the production risk for farmers well above the risks posed from fall armyworm alone. Starkly, over half (52.5%) of the African maize area deemed suitable for fall armyworm is also at risk from a further nine pests, while over a third (38.1%) of the area is susceptible to an additional 10 pests. This constitutes an exceptionally risky production environment for African maize producers, with substantive and complex implications for developing and implementing crop breeding, biological, chemical and other crop management strategies to help mitigate these multi-peril risks.

Correlative species distribution models (SDMs) are increasingly being used to predict suitable insect habitats. There is also much criticism of prediction discrepancies among different SDMs for the same species and the lack of effective communication about SDM prediction uncertainty. In this paper, we undertook a factorial study to investigate the effects of various modeling components (species-training-datasets, predictor variables, dimension-reduction methods, and model types) on the accuracy of SDM predictions, with the aim of identifying sources of discrepancy and uncertainty. We found that model type was the major factor causing variation in species-distribution predictions among the various modeling components tested. We also found that different combinations of modeling components could significantly increase or decrease the performance of a model. This result indicated the importance of keeping modeling components constant for comparing a given SDM result. With all modeling components, constant, machine-learning models seem to outperform other model types. We also found that, on average, the Hierarchical Non-Linear Principal Components Analysis dimension-reduction method improved model performance more than other methods tested. We also found that the widely used confusion-matrix-based model-performance indices such as the area under the receiving operating characteristic curve (AUC), sensitivity, and Kappa do not necessarily help select the best model from a set of models if variation in performance is not large. To conclude, model result discrepancies do not necessarily suggest lack of robustness in correlative modeling as they can also occur due to inappropriate selection of modeling components. In addition, more research on model performance evaluation is required for developing robust and sensitive model evaluation methods. Undertaking multi-scenario species-distribution modeling, where possible, is likely to mitigate errors arising from inappropriate modeling components selection, and provide end users with better information on the resulting model prediction uncertainty.