The effects of logging and hunting on tropical rainforest soundscapes in Gabon

Fig. 1: BAR-LT deployment in the field (Photo provided by Walter Mbamy). Pictured are: Walter Mbamy, Medard Obiang Ebanega, park ranger Charles Bikong, and Tatiana Satchivi.

Project Outline

Intact tropical forests have enormous value to both humans and wildlife. The forests of Gabon are no exception. Forests cover over 90% of the country and supply locals with subsistence (hunting) opportunities as well as supporting countless species, including the critically endangered forest elephant and western lowland gorilla and over 418 bird species. Contrarily though, over 50% of Gabon’s forests are leased by logging companies, most being externally owned by foreign companies, and hunting is poorly regulated.

Walter and the Sound Forest Lab team wanted to understand the collective impacts of logging and hunting by measuring and comparing soundscape saturation, and the rate of gunshots detected, in the protected and logging forests of Gabon. They also wanted to know the difference between logging concessions with sustainability certification, and those without it.  The graphic below is describing soundscape saturation in undisturbed, formerly logged, logged at reduced impact, and conventionally logged patches of rainforest.  

Fig. 2: Main results of the study.

To achieve this, the team recorded 15,265 hours of audio across 109 sites using BAR-LT recorders, over the course of 16 months from February 2021 and June 2022. Specifically, measuring the soundscape saturation and the gunshot rates between:

• protected and unprotected sites.

• certified and non-certified logging sites.

• forest and forest mosaic habitat.

• different seasons.

Fig. 3: Map of the recording locations within the study region in Gabon, Central Africa (inset), within each of the color-coded forest designations. The blue point represents the proposed community conservated area (PCCA) within a non-certified logging concession.

This is a timely study, with increasing logging concerns from locals becoming more prevalent in the media and  pushes for  self-determined creation of community conserved areas. Many communities across Gabon have witnessed logging or are being directly affected by logging, and have voiced their concerns of pollution, damage to the environment and sacred sites, blocked waterways and more.

Fig. 4: BAR-LT deployment on the field (Photo provided by Walter Mbamy). Pictured is Walter on the right, training Ivindo National Park rangers in BAR-LT deployment.

Methods

Study area/location and categories

This study took place across four provinces in Gabon. A total of 109 sites were selected in tropical forests and monitored during both the dry and rainy period for comparing seasonality. Sites fell into the categories: protection level, logging concession status, and habitat type. Logging history was also considered.

BAR-LT setup:

BAR-LTs were deployed in each site to record the soundscape continuously across the human audible range. Each recorder had a single downfacing omnidirectional microphone, gain set to 40dB, sampling rate of 44.1 kHz (ksps), and were mounted 1.8 m above the ground. In terms of scheduling, each BAR-LT recorded in 30-minute intervals for a minimum of seven days per site.  

Fig. 5: Verifying BAR-LT parameters on the field (Photo provided by Walter Mbamy). Pictured is Tatiana Satchivi on the right, ensuring that the recorder is set up correctly before deployment.

Hunting pressure:

In the past, quantification of hunting pressure via remote sensing or participatory surveys was, unsurprisingly, inaccurate, time-consuming and heavily biased. For this study, the team developed a machine learning detector to retrieve gunshot frequency from the recordings to estimate hunting pressure. Site accessibility was also considered.

Logging disturbance:

Selective logging disturbance was assessed with three measures: forest extent from satellite data, site-level logging history (logged <2 years ago, logged 20+ years ago, never logged), and mean canopy height from remote sensing. Only logging history and canopy height were used in final models to represent logging impacts.

Fig. 6: Fieldnote-taking training at Moukala-Doudou National Park (Photo provided by Walter Mbamy). Pictured is Walter Mbamy (middle) training members of the research team (left and right)

Results

There was a notable difference in sound saturation between the different seasons, logging regimes, hunting pressures and habitat type across the 109 sites. Some of the main findings include:

• Lower sound saturation levels corresponded with higher rates of gunshots.

• Non-certified logging sites had the lowest sound saturation and on average the highest gunshot rates regardless of season.

• Sites that had never been logged had the highest sound saturation levels and unique daily patterns.

  
  

FSC certified logging sites had comparable sound saturation levels to that of National Parks during dawn and dusk periods, and sometimes, even greater levels. For the complete results, read their paper here. 

Fig. 7 and 8: Spectrogram and recording samples: before selective logging (top), during selective logging (bottom)

Summary

This project informs us that the few remaining unlogged forests are precious. Without protection from future logging, these remaining areas, their biodiversity, and their soundscapes could be forever changed. Additionally, an emphasis needs to be placed on sustainable hunting, unregulated hunting was also demonstrated to have a negative impact on overall soundscape saturation. The team is next working on using machine learning to classify the individual species that make up the soundscape at each site.

Fig. 9: Walter Mbamy (pictured on the left) explaining to Duke University research team how to deploy and revisit BAR-LT at Ivindo National Park (Photo provided by Walter Mbamy).

Acknowledgement

Authors: Natalie Yoh, Walter Mbamy, Benjamin L. Gottesman, Graden Z.L. Froese, Tatiana Satchivi, Médard Obiang Ebanega, Lauren Carlson, Serge Ekamza Koto, Mutlu Özdoğan, Dave J.I. Seaman, Vincent Maicher, Halina Malinowski, John Poulsen, Alex Ebang Mbélé, Zuzana Buřivalová