Reporting on combined benthic and fish observations can be one of the more challenging tasks for a coral reef scientist. Typically, benthic and fish data are collected by different observers using different methods. With the increasing use of artificial intelligence (AI)-driven image classification tools, such as CoralNet and ReefCloud, often result in data being stored across multiple platforms.
How can you use tools like photo quadrats, AI, and MERMAID together? In Mozambique, WCS staff are monitoring a series of locally managed marine areas near Nampula, along the northern coast of the country. Here, we share an example overview of how the team uses image classification tools with MERMAID to integrate photo quadrat and other coral reef survey methods (including reef fish).
Step 1: Data collection
Erwan Sola, a marine scientist with WCS Mozambique, uses photo quadrats with his team to assess coral communities. Since time underwater can be scarce and costly, finding ways to optimize data collection during dives is critical. Taking quick photos underwater can increase the efficiency of data collected during survey dives, even if the tradeoff is more time spent later processing and analyzing images after the dive.
Step 2: Image classification
After the dive, Erwan uploads his data into CoralNet and annotates each image. Erwan saves a lot of time by using CoralNet’s automatic image classification, powered by machine learning technology. To use CoralNet effectively, Erwan stresses the importance of having as many photos as possible to ‘teach’ the machine and increase its prediction accuracy. He also recommends classifying at least 25 points per photo and taking the time to review each classified image to ensure the predictions are correct.
Step 3: Export CoralNet results to Excel
Once the photo quadrat images are classified, Erwan then exports his results in a spreadsheet and wants to compare these results with the fish surveys he’s entered into his MERMAID project. One option is to create a new Collect record of a ‘Benthic photo quadrat’ in MERMAID to manually enter each of the observations and number of points per observation from the CoralNet spreadsheet; check out MERMAID’s documentation on photo quadrats. But this approach requires a lot of re-typing numbers from an Excel file and CoralNet has already exported a spreadsheet of the results.
Step 4: Prepare data in Excel
Instead of manually typing in observations, Erwan uploads the CoralNet results directly into MERMAID using the mermaidr R package.
- Erwan prepares the CoralNet data for uploading into MERMAID by creating a pivot table in Excel to estimate the number of points for each benthic observation in each photo quadrat.
- He maps the CoralNet labelset to the accepted MERMAID benthic observations, using a formula he created in Excel to save time.
- Finally, he adds the required metadata to complete the MERMAID upload, ensuring the entries are available in the correct MERMAID project. See an example of the workflow here.
Step 5: Ingest Excel file into MERMAID
And voilà (or, “e pronto, já está!” as they say in Mozambique!): Each transect with photo quadrat classifications is a new MERMAID collect record.
Step 6: Validate and submit in MERMAID
Erwan can now validate and submit the collect records in MERMAID, and can check out his results in the MERMAID Global Dashboard – head to the Mozambique project dashboard link to see the results for yourself
In conclusion, sharing your data as Public Summary or Public in MERMAID allows you to immediately view and analyze all of your data in one place, and easily download or share your results for reporting purposes. MERMAID also helps you visualize and analyze all of your data together, including benthic surveys from photo quadrats and transect methods, reef fish, habitat complexity, and coral bleaching.
With MERMAID used by over 1,000 coral reef scientists worldwide, collaboration and analysis of your data has never been easier. Let's work together #ForCoral to better understand and conserve coral reefs. Finally, a big thank you to Erwan for sharing your data and inspiring us all to share our knowledge and collaborate for a better future. E pronto, já está!