Uncertainty in ensemble modelling of large-scale species distribution: Effects from species characteristics and model techniques

Guo, Chuanbo; Lek, Sovan; Ye, Shaowen; Li, Wei; Liu, Jiashou; Li, Zhongjie

doi:10.1016/j.ecolmodel.2014.08.002

Cited by 63 publications

(45 citation statements)

References 53 publications

(62 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unlike results for AUC, there was also a weak negative effect of prevalence on partial ROC. There may be two explanations for the weak strength of the effect of prevalence on model performance compared to the literature (Guo et al, ; Sor et al, ). First, as discussed above, our substitution of sample prevalence for landscape prevalence may have introduced noise that obscured the effect.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Deciphering ecology from statistical artefacts: Competing influence of sample size, prevalence and habitat specialization on species distribution models and how small evaluation datasets can inflate metrics of performance

Hallman

Robinson

2020

Diversity and Distributions

View full text Add to dashboard Cite

Aim Sample size and species characteristics, including prevalence and habitat specialization, can influence the predictive performance of species distribution models (SDMs). There is little agreement, however, on which metric of model performance to use. Here, we directly compare AUC and partial ROC as metrics of SDM performance through analyses on the effects of species traits and sample size on SDM performance. Location Three counties dominated by agricultural lands and coniferous forest in Oregon's Willamette Valley and Coast Range ecoregions. Methods We systematically reduced a large avian point count dataset to alter sample sizes of 22 species of songbird. We used boosted regression trees to run SDMs for each species, quantified habitat specialization, and used mixed effects models to compare the influence of sample size, prevalence, and habitat specialization on SDM performance, calculated as AUC and partial ROC, across species. We calculated AUC and partial ROC with subset and independent evaluation data separately to more comprehensively investigate differences in metrics. Results We found a positive quadratic effect of sample size and a strongly positive effect of habitat specialization on both metrics of model performance. We found a weak effect of prevalence on partial ROC and no effect in AUC. Contrary to expectations, when evaluated with a subset evaluation data, partial ROC was consistently highest in models with the smallest sample sizes. These small sample size models had correspondingly small sample sizes in subset evaluation datasets. Partial ROC evaluated with independent data and AUC evaluated with subset or independent data showed the expected positive correlation between sample size and model performance. Main Conclusions We found that small evaluation datasets can artificially inflate partial ROC. With literature recommended minimum SDM sample sizes as low as three, attention must be given to the effects of correspondingly low sample sizes in evaluation datasets.

show abstract

Section: Discussionmentioning

confidence: 99%

“…The negative relationship between species' prevalence and SDM performance is well documented (Guo et al, ; Sor et al, ). Species' prevalence on a landscape is driven by a combination of the degree of species' habitat specialization and the degree of landscape heterogeneity.…”

Section: Introductionmentioning

confidence: 94%

Deciphering ecology from statistical artefacts: Competing influence of sample size, prevalence and habitat specialization on species distribution models and how small evaluation datasets can inflate metrics of performance

Hallman

Robinson

2020

Diversity and Distributions

View full text Add to dashboard Cite

show abstract

“…There are several limitations in this study that should be clarified: (1) the relationship between frequency of occurrence (photograph rate at a given site) and animal density remains unclear (Carbone et al, 2001; Li et al, 2010, 2014; Liu et al, 2013); (2) the effectiveness of different survey methods can vary depending on species, particularly for small mammals and birds; (3) variation in microclimate on different time frames may cause spatial changes in potential distribution (e.g., we only created variables for annual or seasonal population changes) (Broennimann et al, 2012; Sedgeley, 2001); (4) obtaining high-resolution environmental variables may create potential bias (Wang et al, 2012); (5) modeling limitations (e.g., part of a species’ fundamental niche may not correspond to any combination of environmental variables; species interactions may cause bias) (Cruz-Cárdenas et al, 2014; Elith & Leathwick, 2009; Guo et al, 2015; Maiorano et al, 2013); (6) human influence may be important but is difficult to quantify on a local scale.…”

Section: Discussionmentioning

confidence: 99%

Gap analysis and implications for seasonal management on a local scale

Yang

Zhang

Wang

et al. 2018

PeerJ

View full text Add to dashboard Cite

BackgroundIdentifying biodiversity hotspots on a local scale, using multiple data sources, and ecological niche modeling, has the potential to contribute to more effective nature reserve management.MethodsIn this study, we used infrared-triggered camera trapping, field surveys, and interviews to create a dataset on the distribution of species (mammals and birds) in Hebei Wulingshan Nature Reserve (Hebei Province, China).ResultsWe identified 101 species (14 orders, 38 families), 64 of which (2,142 effective records) were selected for environmental niche modeling. All results were reclassified into three groups: “priority areas” (areas including the potential distributions of over 80% of species), “important areas” (those with 50% of species), and “normal areas” (all other areas). Our results show that priority areas (1.31–1.82 km2) and important areas (7.73–21.44 km2) for conservation were mainly covered by the core and experimental zones of the reserve; additionally, a kilometer-wide margin around the outside of the nature reserve seems to be important to maintaining biodiversity.DiscussionWe close by suggesting some actions for enhancing conservation of biodiversity in the reserve, including monitoring, strengthen law enforcements, introducing popular science, and co-operating with local people.

show abstract

“…There are many factors, such as lower prevalence, spatial autocorrelation, species attributes, environmental range size and the model technique, may result in the observed low accuracy of prediction for these three species. In the current case, the low predictive accuracy of the three species could mostly attribute to low occurrence and environmental range (Guo et al, 2015). Normally for rare species, poor performance measures would be obtained.…”

Section: Model Performance and Uncertaintiesmentioning

confidence: 99%

Predicting fish species richness and assemblages with climatic, geographic and morphometric factors: A broad-scale study in Chinese lakes

Guo

Lek

et al. 2015

Limnologica

Self Cite

View full text Add to dashboard Cite

a b s t r a c tThe present study was designed to investigate the relative importance of climatic (temperature and precipitation), geographic (altitude) and morphometric (lake area) factors in predicting fish species richness and assemblages in Chinese lakes at a large spatial scale. Two recursive partitioning tree-based approaches: Classification and Regression Trees (CARTs) and Multivariate Regression Trees (MRTs) were employed to generate predictive models respectively. Six fish assemblages were thus defined from the MRT model. The results indicated that lake altitude was the main determinant for predicting fish assemblages in Chinese lakes (30.43%), followed by precipitation of the driest month (10.47%), temperature annual range (3.62%) and annual mean temperature (3.15%). Validated CART model implied that precipitation of driest month, maximum temperature of warmest month and lake area were the main predictors in determining fish species richness patterns. Overall, our results indicated that the altitudinal extent and range of climatic variation was sufficient to overshadow the area effect in predicting fish species richness and assemblages in Chinese lakes. At the macroecological scale, the effect of temperature and precipitation on fish richness and assemblages also suggests future changes in fish diversity as a consequence of climate change.

show abstract

Uncertainty in ensemble modelling of large-scale species distribution: Effects from species characteristics and model techniques

Cited by 63 publications

References 53 publications

Deciphering ecology from statistical artefacts: Competing influence of sample size, prevalence and habitat specialization on species distribution models and how small evaluation datasets can inflate metrics of performance

Deciphering ecology from statistical artefacts: Competing influence of sample size, prevalence and habitat specialization on species distribution models and how small evaluation datasets can inflate metrics of performance

Gap analysis and implications for seasonal management on a local scale

Predicting fish species richness and assemblages with climatic, geographic and morphometric factors: A broad-scale study in Chinese lakes

Contact Info

Product

Resources

About