Convolutional neural networks improve species distribution modelling by capturing the spatial structure of the environment

Convolutional Neural Networks (CNNs) are statistical models suited for learning complex visual patterns. In the context of Species Distribution Models (SDM) and in line with predictions of landscape ecology and island biogeography, CNN could grasp how local landscape structure affects prediction of species occurrence in SDMs. The prediction can thus reflect the signatures of entangled ecological processes. Although previous machine-learning based SDMs can learn complex influences of environmental predictors, they cannot acknowledge the influence of environmental structure in local landscapes (hence denoted punctual models). In this study, we applied CNNs to a large dataset of plant occurrences in France (GBIF), on a large taxonomical scale, to predict ranked relative probability of species (by joint learning) to any geographical position. We examined the way local environmental landscapes improve prediction by performing alternative CNN models deprived of information on landscape heterogeneity and structure (ablation experiments). We found that the landscape structure around location crucially contributed to improve predictive performance of CNN-SDMs. CNN models can classify the predicted distributions of many species, as other joint modelling approaches, but they further prove efficient in identifying the influence of local environmental landscapes. CNN can then represent signatures of spatially structured environmental drivers. The prediction gain is noticeable for rare species, which open promising perspectives for biodiversity monitoring and conservation strategies. Therefore, the approach is of both theoretical and practical interest. We discuss the way to test hypotheses on the patterns learnt by CNN, which should be essential for further interpretation of the ecological processes at play. Author summary Species distribution models aim at linking species spatial distribution to the environment. They can highlight the ecological preferences of species and thus predict which species are likely to be present in a given environment. These models are used in many scenarios such as conservation plans or monitoring of invasive species. The choice of model and the environmental data used have a strong impact on the model's ability to capture important information. Specificaly, state-of-the-art models generally use a punctual environment and do not take into account the environmental context or neighbourhood. Here we present a species distribution model based on a convolutional neural network that allows the use of large scale data such as spatialized environmental data including the environmental neighbourhood in addition to the punctual environment. We highlight the interests and limitations of this method as well as the importance of the environmental context in learning about species distributions.

Automated summary

CNN models are found to be crucial in improving predictive performance of CNN-SDMs by considering local landscape structure in predicting plant occurrences in France. They can identify the influence of local environmental landscapes and provide noticeable prediction gain for rare species, showing promising potential for biodiversity monitoring and conservation strategies.