Modelling species presence-only data with random forests

The Random Forest (RF) algorithm is an ensemble of classification or regression trees, and is a widely used and high-performing machine learning technique. It is increasingly used for species distribution modelling (SDM). Many researchers use implementations of RF in the R programming language with default parameters to analyse species presence-only data together with background samples. However, there is good evidence that RF with default parameters does not perform well with such species “presence-background” data. This is often attributed to the typical disparity between the number of presence and background samples also known as class imbalance, and several solutions have been proposed.Here, we first set the context: the background sample should be large enough to represent all environments in the region. We then aim to understand the drivers of poor performance of RF with presence-background data, and explain, test and evaluate suggested solutions. Using simulated and real species data, we compare performance of default RF with other weighting and sampling approaches.We show that class overlap is an important driver of poor performance, alongside class imbalance. The results demonstrate clear evidence of improvement in the performance of RFs when class imbalance is explicitly managed by sampling methods or when the overfitting commonly associated with overlapping classes is avoided by forcing shallow trees.Presence-background data is a particular version of class imbalance in which class overlap is highly likely and extreme imbalance exists. Without compromising the environmental representativeness of the sampled background, we show several approaches to fitting RF that ameliorate the effects of imbalance and overlap, and allow excellent predictive performance. Understanding the problems of RF in presence-background data allows new insights into how best to fit models, and should guide future efforts to best deal with such data.

Opportunities drive the global distribution of protected areas

BackgroundProtected areas, regarded today as a cornerstone of nature conservation, result from an array of multiple motivations and opportunities. We explored at global and regional levels the current distribution of protected areas along biophysical, human, and biological gradients, and assessed to what extent protection has pursued (i) a balanced representation of biophysical environments, (ii) a set of preferred conditions (biological, spiritual, economic, or geopolitical), or (iii) existing opportunities for conservation regardless of any representation or preference criteria.MethodsWe used histograms to describe the distribution of terrestrial protected areas along biophysical, human, and biological independent gradients and linear and non-linear regression and correlation analyses to describe the sign, shape, and strength of the relationships. We used a random forest analysis to rank the importance of different variables related to conservation preferences and opportunity drivers, and an evenness metric to quantify representativeness.ResultsWe find that protection at a global level is primarily driven by the opportunities provided by isolation and a low population density (variable importance = 34.6 and 19.9, respectively). Preferences play a secondary role, with a bias towards tourism attractiveness and proximity to international borders (variable importance = 12.7 and 3.4, respectively). Opportunities shape protection strongly in “North America & Australia–NZ” and “Latin America & Caribbean,” while the importance of the representativeness of biophysical environments is higher in “Sub-Saharan Africa” (1.3 times the average of other regions).DiscussionEnvironmental representativeness and biodiversity protection are top priorities in land conservation agendas. However, our results suggest that they have been minor players driving current protection at both global and regional levels. Attempts to increase their relevance will necessarily have to recognize the predominant opportunistic nature that the establishment of protected areas has had until present times.

Opportunities drive the global distribution of protected areas

Background. The establishment of protected areas, regarded today as a cornerstone of nature conservation, resulted from a multiple array of motivations and opportunities. We explore at global and regional levels, the current distribution of protected areas along biophysical, human, and biological gradients, and assess to what extent protection pursued (i) a balanced representation of biophysical environments, (ii) a set of preferred conditions (biological, spiritual, economical, or geopolitical) or (iii) lands where conservation opportunities exist. Methods. We described with histograms the distribution of terrestrial protected areas along physical, human, and biological independent gradients. Linear and non-linear regressions and correlations led to a general description of sign, shape, and strength of the relationships; a random forest analysis led to a importance ranking of the variables related to individual preferential motivations and opportunistic forces; and an evenness metric seized the strength of variables related to individual representative motivations. Results. We found that globally protection was primarily driven by the opportunities provided by isolation and a low population density (variable importance = 34.6 and 19.9, respectively). Preferences played a secondary role, with a bias towards tourist attractions and international borders (variable importance = 12.7 and 3.4, respectively). Opportunities shaped protection strongly in North America & Australia–NZ and Latin America & Caribbean, while the representativeness of biophysical environments was higher in Sub-Saharan Africa (1.3 times the average of other regions). Discussion. Our results suggest that in spite of being a top priority in land conservation agendas, environmental representativeness and biodiversity protection are still minor players driving protection at global and regional levels, and that the attempts to increase their relevance will necessarily have to recognize the predominant weight that other drivers are playing in the protected areas deployment.

Opportunities drive the global distribution of protected areas

Background. The establishment of protected areas, regarded today as a cornerstone of nature conservation, resulted from a multiple array of motivations and opportunities. We explore at global and regional levels, the current distribution of protected areas along biophysical, human, and biological gradients, and assess to what extent protection pursued (i) a balanced representation of biophysical environments, (ii) a set of preferred conditions (biological, spiritual, economical, or geopolitical) or (iii) lands where conservation opportunities exist. Methods. We described with histograms the distribution of terrestrial protected areas along physical, human, and biological independent gradients. Linear and non-linear regressions and correlations led to a general description of sign, shape, and strength of the relationships; a random forest analysis led to a importance ranking of the variables related to individual preferential motivations and opportunistic forces; and an evenness metric seized the strength of variables related to individual representative motivations. Results. We found that globally protection was primarily driven by the opportunities provided by isolation and a low population density (variable importance = 34.6 and 19.9, respectively). Preferences played a secondary role, with a bias towards tourist attractions and international borders (variable importance = 12.7 and 3.4, respectively). Opportunities shaped protection strongly in North America & Australia–NZ and Latin America & Caribbean, while the representativeness of biophysical environments was higher in Sub-Saharan Africa (1.3 times the average of other regions). Discussion. Our results suggest that in spite of being a top priority in land conservation agendas, environmental representativeness and biodiversity protection are still minor players driving protection at global and regional levels, and that the attempts to increase their relevance will necessarily have to recognize the predominant weight that other drivers are playing in the protected areas deployment.