Below is the research proposal that I wrote before the start of my research this summer. Reflecting on this at the end of the process, I can see how my methods and plans changed as my knowledge of the field grew. I ended up using Acoustic Indices to quantify overall vocal activity, and the Machine Learning model BirdNET for individual species classifications, which I hadn't come across as options before I started my project. I also learnt that a large part of the data analysis process is the cleaning and processing stages before the analysis can actually start!
The research project:
We will be investigating the avian biodiversity of mangrove forests by undertaking novel acoustic biodiversity monitoring of species along the Suriname coastline. This area currently has low biodiversity knowledge, and the mangrove forests there are likely to include some key species that are used to trigger KBA (Key Biodiversity Area) designation. Automatic recording units (ARUs) have been deployed to collect data on biodiversity soundscapes, in healthy and degraded areas of three types of forest over a 2 year period. We will classify this data using AI algorithms to extract information on the influence of sediment variation and mangrove types & size on bird biodiversity and behavioral patterns in Suriname. The project will assess the benefits mangrove forests and avian biodiversity provide to the ecosystem, and inform coastal management strategies.
The research complements and is supported by the GCBC ENHANCES project that focuses on understanding the coastal erosion and flood risk mitigation potential of mangrove forests in Suriname and Guyana.
My research:
I will be using MonitoR (a package for use in R) and/or the software applications Raven-Pro, Song Scope and Kaleidoscope Pro to classify the calls of birds that inhabit a range of different areas of the mangrove forests: The Rufous Crab-Hawk, Arrowhead piculet, Blood-coloured woodpecker (all endemic and/or near-threatened mangrove specialists) and a type of coastal wader (wetland specialist). This will involve ‘teaching’ the software the qualities of the different calls, to gain an output of the times in the recordings when a call was detected. These outputs are then weighted, according to the accuracy of each software application that they came from, into one ‘ensemble’ model. This will enable me to see patterns between bird activity and the variation in types and size of mangroves, which I will statistically test for correlation.
My research project will extract data on avian vocalization, which is a vital component of the wider project and can be used to optimise mangrove forest management strategies and reduce biodiversity loss. I will produce a report of my findings, detailing my methods of birdsong classification and visualizations of the data that I find. I will discuss possible correlations between the data and mangrove types and size, determining how mangrove forest characteristics influence avian activity.