According to World Wildlife Fund statistics, from 1970 to 2016, the global population of representative species decreased by 68%, and biodiversity continued to decline.
Protecting biodiversity requires accurate analysis of local ecological conditions and the formulation of reasonable ecological protection policies. However, ecological data are too complex and statistical standards are difficult to unify, making large-scale ecological analysis difficult to carry out.
Recently, researchers at Cornell University used deep learning to analyze 9 million sets of bird data and obtained the distribution data of warblers in North America, opening a new era of ecological data analysis. Chapter.
Author | Xuecai
Editor | Sanyang, Iron Tower
According to statistics from the World Wildlife Fund (WWF),From 1970 to 2016, the global average population of 4,392 representative species and 20,811 populations decreased by 68% , global biodiversity is declining.
Protecting biodiversity requires accurate large-scale analysis of species distribution in relevant areas. However, due to the large amount of data and the lack of unified statistical methods, researchers are currently unable to accurately count specific Regional biodiversity (species richness, population size, etc.) and biological composition data (the status of a species in the local ecosystem).
Traditional species richness statistics require overlaying distribution maps of different species for modeling and prediction, or prediction directly through macroecological models. No matter which method is used, the inference results will be affected by the accuracy of the model, and the former will also be affected by the accuracy of the map.
Moreover,the temporal resolution of this prediction method is very poor, making it impossible to accurately judge seasonal changes in species distribution, let alone Research on the connections between species is not conducive to the formulation of ecological protection policies.
Deep learning provides an effective means for large-scale spatiotemporal research on biodiversity. Researchers at Cornell University in the United States developed the DMVP-DRNets model by combining Deep Reasoning Network (DRN) and Deep Multivariate Probit Model (DMVP). The spatiotemporal distribution of Warbler in North America was analyzed from 9,206,241 sets of eBird data, and the connection between Warbler, the environment and other species was inferred. Relevant results have been published in "Ecology".
Paper link:
https://esajournals.onlinelibrary.wiley.com/doi/10.1002/ecy.4175
experiment procedure
Dataset: eBird with covariates
The researchers used eBird data from January 1, 2004 to February 2, 2019, between 170°-60° W, 20°-60° N, as the dataset for this study. After excluding duplicate data, there are a total of 9,206,241 sets of eBird data. Each set of eBird data includes time, date, location and all bird species observed.
The researchers also introduced 72 covariates, including 5 covariates related to the observer, such as activity status, number of observers, observation time, etc.; 3 covariates related to time Relevant covariates, mainly used to bridge the deviation between different time zones; 64 variables related to topography, such as altitude, coastline, islands, etc.
Model Framework:Decoder + Latent Space
This study uses DMVP-based DRN for data analysis and prediction. This model includes a 3-layer fully-connected network decoder for analyzing the correlation of input features, and two structured latent spaces. Used to express the relationship between species and species-environment.
Finally, the DMVP-DRNets model outputs 3 ecologically relevant results through an interpretable latent space:
1. Environment-related characteristics: Reflects the connection and interaction between different environmental covariates;
2. Species related characteristics: Reflect the connection between different species through the residual correlation matrix;
3. Biodiversity-related characteristics: Such as the abundance and distribution of a certain species, etc.
Model evaluation: comparison with HLR-S
Before putting the DMVP-DRNets model into large-scale use,researchers first compared it with the HLR-S model based on spatial Gaussian processes. HLR-S is one of the most commonly used models in ecology to study the joint distribution of multiple species.
Both models were first trained on 10,000 sets of eBird data. The HLR-S model took more than 24 hours to train, while the DMVP-DRNets model took less than 1 minute.
Subsequently, the eBird data of different scales were analyzed, and DMVP-DRNets model was better than the HLR-S model in 11 evaluation criteria , trailing the HLR-S model only in species richness calibration loss.
Experimental results
Distribution area:Appalachian Mountains
After analyzing eBird data, the DMVP-DRNets model outputs a monthly distribution map of the North American warbler with a spatial resolution of 2.9 km2. The distribution of different species of wood warblers in North America is very dynamic, with different distribution hotspots every month. After overlaying the monthly distribution maps, the researchers found that the Appalachian Mountains are the region with the highest species diversity for wood warblers.
a: Maximum species richness distribution of wood warblers across North America
b: The main distribution area of wood warblers in North America
At the same time, researchers also discovered distribution hotspots of warblers at different migration stages. During their pre-breeding migration, wood warblers are found primarily around the Appalachian Mountains in Ohio, West Virginia, and Pennsylvania. After breeding, the northern Appalachian Mountains are the area where the warblers are most abundant.
Warbler-Environment:Water, Land and Season Preference
Further, the researchers used the DMVP-DRNets model to analyze warblers-environment interactions in the northeastern United States.
First,researchers were able to roughly distinguish the preferences of different warblers for aquatic and terrestrial environments. Later, they found that different species of warblers have different preferences for the environment during the breeding season. The blue-winged yellow warbler, northern warbler and yellow-throated warbler, which prefer aquatic environments, live closer together during the breeding season, while the pine warbler will move closer to other species associated with pine forests, such as brown-headed warblers. Common turtles and red-headed woodpeckers.
As the seasons change, the distribution of different warblers also changes. During the post-breeding migration period, most warblers roost in groups, while palm warblers migrate later in the fall. Pine warblers and yellow-rumped white-throated warblers inhabit the northeastern United States year-round.
Interspecies Associations: Competition and Cooperation
Warblers exhibit different relationships with other species during the breeding, non-breeding, and migratory seasons.
During the breeding season, warblers mainly defend their own habitat and are less associated with other species. There is even a negative correlation between species that share similar habitats and are more aggressive, such as the Black-naped Wesson's Warbler and the Orange-tailed Warbler.
During the migration period, most warblers show strong positive correlations with each other and with other species in the forest. This is consistent with observations that wood warblers form mixed migratory groups with other species such as red-eyed green azaleas and black-crowned chickadees.
During this period, the relationship between warblers and predators such as giant-winged buzzards, striped eagles, chicken hawks, and red-shouldered buzzards was poor, and the negative correlation coefficient between the two was high.
The above results show thatDMVP-DRNets model can accurately judge the distribution of warblers in different periods, and can infer the connection between warblers, the environment and other species , providing a basis for formulating ecological policies.
AI "Bird Census"
In addition to data analysis, data collection is also an important part of ecological research. Unlike plants, birds are highly alert and move quickly. Some species are small and difficult to observe accurately.
Traditional methods rely on telephoto cameras, high-powered telescopes and stationary cameras to observe birds from long distances. Although this method avoids disturbing the birds, it requires a lot of manpower and material resources, and it also requires the observer to have considerable knowledge of ecology and taxonomy.
Through deep neural networks,AI can perform efficient image recognition and sound recognition, providing a new method for bird observation. Deploy audio and video recording equipment at the main activity areas of birds. The equipment can upload the recorded data to the server, and then analyze the data through AI to extract the information in the audio and video. Finally, Get the distribution of birds in this area. This method has been widely used by the National Forestry and Grassland Administration in parks, wetlands and ecological reserves.
At the same time, this skill of AI can also reduce the workload of scientific researchers. AI can eliminate the interference of background and noise, focus on the characteristics of the image, and quickly solve problems that are difficult for ecologists to make judgments. For example, in the photo below, without any knowledge about birds, it is difficult to quickly determine the number of chicks from the numerous feathers.
Reference links:
[1]https://www.worldwildlife.org/publications/living-planet-report-2020
[2]https://phys.org/news/2023-09-ai-birds-easier.html
[3]https://www.forestry.gov.cn/main/586/20230118/094644604451331.html