Arbimon December Newsletter
Included in this newsletter
White paper & case study highlights
Check out our recent white paper
Arbimon and the Rainforest Connection (RFCx) team has recently launched an enlightening white paper titled “Harnessing the Power of Sound and AI to Track Global Biodiversity Framework (GBF) Targets“
Read to discover:
- The potential of acoustic & AI technology to monitor biodiversity
- 11 real-world case studies that align with GBF targets and their candidate indicators
- The power of PAM and AI for Biodiversity Monitoring to inform conservation action
- Collaborative projects showcasing a wide range of ecoacoustic applications
Case study #1
Location: Atlantic Forest
GBF Target: 2 (restore 30% of degraded lands)
Finding: There are positive effects of restoration on biodiversity, with restored sites able to host a similar number of species as protected forests.
Case study #2
GBF Target: 4 (halt species extinctions)
Focus: We are surveying the Critically Endangered mangrove finch to help maintain the population.
Case study #4
GBF Target: 6 (mitigate invasive species)
Focus: We've established a real-time acoustic detection pipeline to monitor the spread of invasive grey squirrels.
Developing AI models to detect species across disturbance gradients
The Biological Dynamics of Forest Fragments Project (BDFFP), co-run by the Smithsonian Institute, George Mason University, and Instituto Nacional de Pesquisas da Amazônia, is the world’s largest and longest-running study of habitat fragmentation. Located 70 km from Manaus in the Brazilian central Amazon, the BDFFP’s overarching goal is to assess biodiversity along various fragmentation and disturbance gradients.
Collaborators from the project deployed passive acoustic recorders at 43 sites across differently-sized fragments, primary forest and secondary forest. Using Arbimon’s Pattern Matching tool, we detected 200+ species of birds, monkeys, and frogs! We then used the validated presence and absence clips from Pattern Matching to train an AI model (convolutional neural network, or CNN) to automatically classify 98 of these species. The model performed very well in our evaluations, achieving an average precision across all classes of 0.94.
Now that we already have a strong trained model, new recordings can be passed through it to automatically detect species calls, facilitating long-term monitoring and efficient analyses moving forward!
Interested in partnering on an ecoacoustic monitoring project?
Email us: email@example.com
Tips for using arbimon
Did you know…
There are lots of ways to download your results from Arbimon:
Want to download all the validations for a Pattern Matching job?
Go to the Analysis tab, Pattern Matching menu, click on a job, then click the ‘Export PM data’ to get a spreadsheet of each region of interest (ROI), its coordinates, and validation status.
Want a spreadsheet of all the recordings with a certain set of species?
Go to the Data tab, Recordings menu, and click ‘Export’ to download species detections, summarized across sites, days, and hours
Want to share figures and maps from your project Insights page?
You can download each visualization as an image file, or click ‘Download Source’ to get a spreadsheet of the data being used to generate those figures.
Review our extensive support documentation and video tutorials.
PRO TIP: You can also click our help button “?” icon in the bottom right of the Arbimon platform) for chat support.