Predicting the risk of cucurbit downy mildew in the eastern United States using an integrated aerobiological model.

Cucurbit downy mildew caused by the obligate oomycete, Pseudoperonospora cubensis, is considered one of the most economically important diseases of cucurbits worldwide. In the continental United States, the pathogen overwinters in southern Florida and along the coast of the Gulf of Mexico. Outbreaks of the disease in northern states occur annually via long-distance aerial transport of sporangia from infected source fields. An integrated aerobiological modeling system has been developed to predict the risk of disease occurrence and to facilitate timely use of fungicides for disease management. The forecasting system, which combines information on known inoculum sources, long-distance atmospheric spore transport and spore deposition modules, was tested to determine its accuracy in predicting risk of disease outbreak. Rainwater samples at disease monitoring sites in Alabama, Georgia, Louisiana, New York, North Carolina, Ohio, Pennsylvania and South Carolina were collected weekly from planting to the first appearance of symptoms at the field sites during the 2013, 2014, and 2015 growing seasons. A conventional PCR assay with primers specific to P. cubensis was used to detect the presence of sporangia in rain water samples. Disease forecasts were monitored and recorded for each site after each rain event until initial disease symptoms appeared. The pathogen was detected in 38 of the 187 rainwater samples collected during the study period. The forecasting system correctly predicted the risk of disease outbreak based on the presence of sporangia or appearance of initial disease symptoms with an overall accuracy rate of 66 and 75%, respectively. In addition, the probability that the forecasting system correctly classified the presence or absence of disease was ≥ 73%. The true skill statistic calculated based on the appearance of disease symptoms in cucurbit field plantings ranged from 0.42 to 0.58, indicating that the disease forecasting system had an acceptable to good performance in predicting the risk of cucurbit downy mildew outbreak in the eastern United States.

A multi-criteria approach to identify favorable areas for goat production systems in Veracruz, México.

This research identifies favorable areas for goat production systems in the state of Veracruz, Mexico. Through the use of the analytic hierarchy process, layers of biophysical and soil information were combined to generate a model of favorability. Model validation was performed by calculating the area under the curve, the true skill statistic, and a qualitative comparison with census records. The results showed the existence of regions with high (4494.3 km2) and moderate (2985.8 km2) favorability, and these areas correspond to 6.25 and 4.15%, respectively, of the state territory and are located in the regions of Sierra de Huayacocotla, Perote, and Orizaba. These regions are characterized as mountainous and having predominantly temperate-wet or cold climates, and having montane mesophilic forests, containing pine, fir, and desert scrub. The reliability of the distribution model was supported by the area under the curve value (0.96), the true skill statistic (0.86), and consistency with census records.

The predictive skill of species distribution models for plankton in a changing climate.

Statistical species distribution models (SDMs) are increasingly used to project spatial relocations of marine taxa under future climate change scenarios. However, tests of their predictive skill in the real-world are rare. Here, we use data from the Continuous Plankton Recorder program, one of the longest running and most extensive marine biological monitoring programs, to investigate the reliability of predicted plankton distributions. We apply three commonly used SDMs to 20 representative plankton species, including copepods, diatoms, and dinoflagellates, all found in the North Atlantic and adjacent seas. We fit the models to decadal subsets of the full (1958-2012) dataset, and then use them to predict both forward and backward in time, comparing the model predictions against the corresponding observations. The probability of correctly predicting presence was low, peaking at 0.5 for copepods, and model skill typically did not outperform a null model assuming distributions to be constant in time. The predicted prevalence increasingly differed from the observed prevalence for predictions with more distance in time from their training dataset. More detailed investigations based on four focal species revealed that strong spatial variations in skill exist, with the least skill at the edges of the distributions, where prevalence is lowest. Furthermore, the scores of traditional single-value model performance metrics were contrasting and some implied overoptimistic conclusions about model skill. Plankton may be particularly challenging to model, due to its short life span and the dispersive effects of constant water movements on all spatial scales, however there are few other studies against which to compare these results. We conclude that rigorous model validation, including comparison against null models, is essential to assess the robustness of projections of marine planktonic species under climate change.

Landslide susceptibility prediction improvements based on a semi-integrated supervised machine learning model.

Differences in model application effectiveness, insufficient numbers of disaster samples, and unreasonable selection of non-hazard samples are common problems in landslide susceptibility studies. Therefore, in this paper, we propose a semi-integrated supervised approach to improve the prediction performance of machine learning (ML) models in landslide susceptibility studies. First, taking the lower reaches of the Jinsha River as the study area, a geospatial dataset consisting of 349 landslides, an equal number of randomly selected non-landslide points, and 12 environmental factors were randomly divided into training (70%) and testing (30%) datasets. Then, K-nearest neighbors (KNN), random forest (RF), and Bayesian-regularized neural network (BRNN) models were built. Second, the three models were combined to form an integrated weighted model. Very high- and low-prone areas were selected and, combined with the prediction results and remote sensing images, landslide and non-landslide samples were identified. The identified samples were then combined with the original samples to form new samples, which were used to sequentially construct the ensemble-supervised K-nearest neighbors (ESKNN), ensemble-supervised random forest (ESRF), and ensemble-supervised Bayesian-regularized neural network (ESBRNN) models. Finally, the area under the curve (AUC), true skill statistic (TSS), and frequency ratio (FR) values were used to test the accuracy of each model. The traditional ML model results and accuracy were improved by the semi-integrated supervised method. The ESRF model had the best prediction effect (AUC = 0.939, TSS = 0.440, and FR = 95.8%). The proposed semi-integrated supervised ML model solved the problems observed in traditional landslide susceptibility studies and provided insights for reducing variations in model applications, expanding landslide data sources, and improving non-landslide sample selection.

Simultaneously estimating food web connectance and structure with uncertainty.

Food web models explain and predict the trophic interactions in a food web, and they can infer missing interactions among the organisms. The allometric diet breadth model (ADBM) is a food web model based on the foraging theory. In the ADBM, the foraging parameters are allometrically scaled to body sizes of predators and prey. In Petchey et al. (Proceedings of the National Academy of Sciences, 2008; 105: 4191), the parameterization of the ADBM had two limitations: (a) the model parameters were point estimates and (b) food web connectance was not estimated.The novelty of our current approach is: (a) We consider multiple predictions from the ADBM by parameterizing it with approximate Bayesian computation, to estimate parameter distributions and not point estimates. (b) Connectance emerges from the parameterization, by measuring model fit using the true skill statistic, which takes into account prediction of both the presences and absences of links.We fit the ADBM using approximate Bayesian computation to 12 observed food webs from a wide variety of ecosystems. Estimated connectance was consistently greater than previously found. In some of the food webs, considerable variation in estimated parameter distributions occurred and resulted in considerable variation (i.e., uncertainty) in predicted food web structure.These results lend weight to the possibility that the observed food web data is missing some trophic links that do actually occur. It also seems likely that the ADBM likely predicts some links that do not exist. The latter could be addressed by accounting in the ADBM for additional traits other than body size. Further work could also address the significance of uncertainty in parameter estimates for predicted food web responses to environmental change.

Climate change diminishes the potential habitat of the bont tick (Amblyomma hebraeum): evidence from Mashonaland Central Province, Zimbabwe.

BACKGROUND: Understanding the response of vector habitats to climate change is essential for vector management. Increasingly, there is fear that climate change may cause vectors to be more important for animal husbandry in the future. Therefore, knowledge about the current and future spatial distribution of vectors, including ticks (Ixodida), is progressively becoming more critical to animal disease control. METHODS: Our study produced present (2018) and future (2050) bont tick (Amblyomma hebraeum) niche models for Mashonaland Central Province, Zimbabwe. Specifically, our approach used the Ensemble algorithm in Biomod2 package in R 3.4.4 with a suite of physical and anthropogenic covariates against the tick's presence-only location data obtained from cattle dipping facilities. RESULTS: Our models showed that currently (the year 2018) the bont tick potentially occurs in 17,008 km2, which is 60% of Mashonaland Central Province. However, the models showed that in the future (the year 2050), the bont tick will occur in 13,323 km2, which is 47% of Mashonaland Central Province. Thus, the models predicted an ~ 13% reduction in the potential habitat, about 3685 km2 of the study area. Temperature, elevation and rainfall were the most important variables explaining the present and future potential habitat of the bont tick. CONCLUSION: Results of our study are essential in informing programmes that seek to control the bont tick in Mashonaland Central Province, Zimbabwe and similar environments.