Passive acoustic monitoring of biodiversity

KEY DETAILS

Principal Investigator: Dr. Matthew Rogan
Date; 4 November 2024
Version; 0.3.0
Programme; Rangelands Biodiversity Project (RBP)
Study Site; Lewa-Lolldaiga-Borana-Ngare Ndare (LLBN)
Key partners; Contact email; mrogan@naturalstate.org

1. PREAMBLE

Natural State’s objectives and activities are detailed by a set of accepted Standard Operating Procedures (SOPs). These documents describe the steps involved in all Natural State projects, from data collection to data processing and storage. Each SOP documents key project details and provides methodological details specific to the project. The objectives and background of the project, features of the study area, and details on survey and sampling design may be found in the project Design Document (DD) which is available in the Related Documents section below.

1.1 SOP PURPOSE

To provide a clear step by step guide to the methods implemented in the project, therefore allowing for consistency in data collection and repeatability of all steps involved in a project’s data collection, processing and storage. This is crucial to Natural State’s mission of facilitating nature restoration at scale by using the latest technology and methods to revolutionise impact monitoring for carbon, biodiversity and human well-being.

1.2 SOP SCOPE

This document details how this project will be implemented. All methodological steps are explained, and the principal team members overseeing the project are listed in case further clarification is required. It also further directs readers to where they can find additional information relevant to the project. This document is intended to be printed out and taken to the field for reference sake.

2. GLOSSARY

Acoustic monitor: a passive acoustic recording device.
Ambient noise: Regular (continuous or intermittent) non-target noises that may obscure the target sounds. Common Ambient noises include wind, running water, vehicles, people’s voices and construction.
Band filter: Subset frequencies to be recorded. For example, it may be practical to filter out all frequencies more than half the sampling rate.
Centroid: A point at the center of a corresponding sampling grid cell denoted by a geohash, a longitude coordinate, and a latitude. Here, the centroid of the CPP.
Deployment: The period of time a single remote sensor is active within the environment at a single, defined station as part of a survey.
Deployment group: Temporally distinct sets of geographically or thematically grouped and contemporaneous (e.g., seasonal deployment groups).
Dominant woody vegetation: The primary species that provides structure to the habitat within or around a sampling unit. Must stand taller than the height of the grass. The dominant woody vegetation might not be the most abundant type but is rather the one that consistently provides maximum vertical structure within the grid cell.
Features, Sampling: Physical fine-scale characteristics of the landscape that correlate with animal movement and space use such that detection within a used habitat patch is assumed to be greater on or near such features than away from them.
Frequency: The number of times per second that a sound pressure wave repeats itself. Frequency is measured in Hertz (Hz) or, more often, in kilohertz (kHz). Most bird calls are between 0 and 16 kHz.
Gain: In a receiving antenna, the gain describes how well the antenna converts radio waves arriving from a specified direction into electrical power. It is basically a measure of sensitivity; the higher the gain the more likely the sensor is to detect a signal (e.g., bird call) but the more background noise is generated.
Grid cell: A geohash or other rectangular area within which one or more sensors may be deployed as semi-independent sampling units.
Habitat type: Coarse, subjective classifications of the amount of woody cover in a heterogeneous savanna typically measured within a ~100m radius of the sampling location (see open grassland, open woodland, and closed woodland).
LLBN: The Lewa-Lolldaiga-Borana-Ngare Ndare landscape.
Project: A concerted, data-driven effort to robustly measure variation in Biodiversity, Carbon, or Human-wellbeing in response to one or more sources of heterogeneity in a designated landscape.
Recording cycle: The combined duration (usually measured in seconds) of one recording period and one sleep period. For example, if a recording period is 3570 seconds and a sleep period is 30 seconds, then the recording cycle is 3600 seconds.
Recording duration: A defined duration measured in seconds during which a device is actively recording during the recording period. The recording duration defines the duration of each individual audio file.
Recording periods: Defined periods of a 24-hour day during which the recorder operates recording cycles. The recording periods must be set in UTC.
Remote sensor: An electronic sensor with automated data collecting capabilities that operate independently of a human handler.
S123: Survey123, a field-data collection app from ESRI which NS uses for recording all field observations and survey metadata.
Sampling design: The set of field methods employed in a survey and the manner of their use.
Sampling protocol: Explicit survey methodology that describes the design, effort, duration, configuration, and operation of a survey.
Sampling rate: The sampling rate is how rapidly continuous sound is discretized during digitization. It is equivalent to the frames-per-second rate in video. The sample rate must be at least double the highest frequency you want to record. For example, in order to record sounds up to 20 kHz, a sample rate of at least 40 kHz is required.
Site, Sampling: A distinct, discrete spatial unit defined in at least two dimensions where sampling occurs.
Sleep duration: A defined duration measured in seconds during which a device does not record sounds. The sleep duration must be less than the recording period.
Station, Sampling: A point location where sampling occurs in space.
Trigger type: A method to program recorders to only record when sounds above certain amplitudes are detected.
Vegetation structure: Variation in height and woodiness of vegetation that extends above the grass.

3. PROJECT OVERVIEW

3.1 PROJECT AIMS

The passive acoustic monitoring of biodiversity project aims to:

Non-invasively collect observations of bird species across habitat types and management regimes.
Non-invasively collect audio recordings of other vocal species such as invertebrates and mammals across habitat types or gradients and management regimes.
Non-invasively collect soundscape recordings across habitat types or gradients and management regimes.
Collect audio recordings of domestic animals such as cattle to measure their activity across habitat types and management regimes.

3.2 PROJECT BACKGROUND

Natural State’s mission is to catalyse large-scale restoration globally by revolutionizing impact monitoring, developing new financial mechanisms and supporting local changemakers. As part of this, NATURAL STATE aims to develop and optimise impact monitoring systems to reduce costs and provide independently verifiable results on interventions.

Passive monitoring of biological communities is a staple of long-term and large-scale ecological monitoring. Passive monitoring, deploying remote sensors continuously or intermittently to gather information about the landscape, tends to be less invasive and less resource-intensive than most active monitoring approaches (e.g., animal tracking, direct counts) and thus offers much lower costs to scale. The emergence of highly accurate machine-learning classifiers for passively collected data has amplified the scalability and effectiveness of passive monitoring.

Bioacoustics offer especially pronounced benefits because they can be used to monitor such a wide range of taxa from insects to whales, because they are applicable in nearly every biome, and because animal vocalizations have evolved to project through space meaning that signals can be detected at greater distances than many other passive methods. Unlike visual cues, bioacoustics are not impeded by dense vegetation. For these reasons, bioacoustics are among the fastest developing and most promising methods for large-scale ecological monitoring.

3.3 STUDY AREA

Lewa-Lolldaiga-Borana-Ngare Ndare

The Lewa-Lolldaiga-Borana-Ngare Ndare (LLBN) study area is a savanna rangeland landscape in the central highlands of Kenya that extends north from the lower slopes of Mt. Kenya, straddling the boundary between Meru and Laikipia counties. The four main properties are located at 0.11 = 0.34° latitude and 37.07-37.53° longitude. The study site-centered ecosystem (SCE) exhibits three main “arms” with on extending south around the western edge of Mt. Kenya, one extending east along the southern side of the B9 highway to the northeastern edge of Meru County, and the longest and largest arm extending 130 km northwest to Maralal with some isolated islands of similar ecosystem to the north and west. The entire SCE is located between -0.3° and 2.3° latitude, and between 36.3° and 38.1° longitude.

Rainfall is highly variable, but is typically between 400 and 600 mm annually. During droughts, total annual rainfall can drop below 200 mm. Rainfall generally follows a north-south gradient with more rainfall at higher elevations in the south. The landscape sits at 1400-2370 m with higher elevations in Ngare-Ndare forest and south-central Lolldaiga and the lowest elevations in Lewa and northeast Borana. Soil types are highly variable throughout the landscape but consist primarily of Luvisols (Haplic and Vertic) and Dystric Regosols. Vegetation communities are predominantly Acacia-Commiphora bushlands and thickets with montane forests at higher elevations (Dinerstein et al. 2017).

Currently, passive acoustic monitoring surveys are conducted across Lewa and Borana Conservancies as well as Lolldaiga ranch.

Lewa Wildlife Conservancy is mostly owned by The Nature Conservancy but has a number of small, privately owned enclaves. The reserve is almost exclusively used for photographic tourism but does have some livestock grazing, especially along the community road and in a section nicknamed ‘Bosnia’ where the community have grazing rights. Over the last 30 years, the conservancy has increasingly adopted elephant and large-mammal exclosures as a major management practice aimed at increasing tree cover on the conservancy.

Lolldaiga ranch is principally a livestock ranch but is also used by free ranging wildlife. Historically, the ranch was intensively grazed but recently new management has implemented a regenerative grazing regime based on intensive grazing by a few large, fast-moving herds. The grazing regime is intended to stimulate grass productivity and prevent selective grazing by livestock herds.

Borana Conservancy is a dual use wildlife and livestock ranch that also includes privately owned enclaves. In addition to photographic tourism, the conservancy produces cattle and has a partnership with a local community that extends limiting grazing rights to community herders. Borana has a few large-mammal exclosures, especially along the river that runs between Lewa and Borana.

The methods described here are broadly applicable to any savanna landscape.

3.4 PROJECT TIMELINE

In 2022, NS conducted a pilot study to collect data that could be manually annotated to creating training datsets for a machine-learning classifier. This work preceded this protocol.
In Q2-Q3 2023, NS deployed acoustic monitors as part of the Rangeland Baseline Survey to sample bird communities across varying management regimes within the three properties. Each 500m X 500m sample site consisted of three camera stations and one acoustic station. This work also preceded this protocol.
In Q3 2023, NS conducted a baseline survey as part of before-after-control-impact survey with 36 camera and acoustic stations evenly distributed across three treatment types.
In Q3 2023, NS introduced and S123 form for capturing AudioMoth metadata in the field.
In Q3 2023, NS developed the ability to identify 99 bird species in the LLBN landscape using the opensource neural network algorithm, BirdNET, with species-specific confidence thresholds identified for all 99 species.
In Q1 2024, NS began collecting AudioMoth data as part of the Carbon-Biodiversity Survey in the LLBN landscape.
In Q1 2024, NS began Version 2 for calibrating BirdNET confidence scores.
In Q2 2024, NS began baseline sampling at the Kupona experimental farm.

4. SAMPLING PREPERATION

The equipment mentioned in the list below needs to be gathered, checked and packed before sampling begins. If any sensors need to be configured prior to commencing sampling this will be documented below the equipment list.

NB: Ensure that landownders and managers have been notified of the planned fieldwork in advance.

4.1 EQUIPMENT LIST

AudioMoths. Devices must be programmed with the latest firmware. Each AudioMoth must be tagged with a QR code for scanning into S123.
Eneloop Pro Rechargeable batteries.
Micro SD cards (SanDisk Extreme or SanDisk Ultra 64 GB ). Each micro SD card must be tagged with a QR code for scanning into S123.
Casings and straps for securing AudioMoths.
Long (2m) poles for holding AudioMoths.
Handheld GPS with spare Eneloop Pro batteries and preloaded with coordinates of prospective sites.
Two fully charged mobile devices with cameras and preloaded with the S123 app and Acoustic Survey form.
A mallet for hammering poles into the ground.
Spare AudioMoths, batteries and SD cards.
Water and a medical kit in the vehicle.

4.2 DEVICE CONFIGURATION

AudioMoth Configuration Settings

Use the configuration app (https://www.openacousticdevices.info/applications), and enter the settings below to configure the device. This needs to be done from a computer at the office, before going into the field.

Setting	Selection
Recording Periods	00:00 – 08:00 UTC AND 12:00 – 21:00 UTC
Sample Rate (kHz)	48 kHz
Gain	Medium – Low (second from the left)
Sleep Duration (s)	30
Recording duration (s)	3570
Band filter	None
Trigger type	None

Example recording settings Example of proper recording settings.

Example recording schedule Example of proper recording schedule for EAT.

5. SAMPLING PROCEDURES

Unless otherwise noted in a survey design document, all passive acoustic monitoring stations should be sampled for 30 days with devices checked after two weeks. However, any deployment of at least three-weeks (21-days) will be considered a full sample and any deployment of at least two weeks (14 days) will be considered adequate sampling. Deployments of between 10 and 14 days will be considered adequate sampling at the discretion of the field team lead and the PI. Any deployment lasting fewer than 10 days will be considered incomplete.

Editorial Notes

NS will transition to 64 GB microSD cards in 2024.
NS will trial external microphones for better sound quality and wind reduction with v1.2 AudioMoths.
NS will trial sampling rates of 196 kHz and 384 kHz for bat detection.

5.1 ACOUSTIC MONITOR DEPLOYMENT

The field team should arrive at the grid centroid or other station coordinates and note their surroundings. This includes discussing the habitat type and dominant woody vegetation and noting any features that occur within 75 m of the centroid as well as any sources of ambient noise that that can be heard within the centroid.

Acoustic Monitor Placement

Each AudioMoth is assumed to have a detection radius for most bird species of 100 m. Therefore, AudioMoths should be placed at least 100m from any edges such as reserve or exclosure boundaries. AudioMoths should always be placed at least 250 m apart to minimize the overlap in sampling areas.

When an AudioMoth is deployed at a site that is 4 ha or smaller (i.e., site boundary is less than 100m from the centroid), the AudioMoth should always be deployed at or very close to the site centroid. When an AudioMoth is deployed at a site greater than 4 ha, the field team should identify the point within the site that is at least 100 m from the site boundary and which maximizes the diversity of vegetation structure within the 100 m radius.

AudioMoths should generally be placed in vegetation at a height of 1.5 - 2 m. AudioMoths should not be placed in depressions or against cliff faces, but on the edge at the top of a cliff is suitable. They should also not be placed near ambient noise. This means avoid placing them near running water, busy roads, or construction. If the wind typically blows from a particular direction, AudioMoths should be placed on the leeward (i.e., downwind) side of trees and bushes.

The deployment location of an AudioMoth is always defined as a station, even when it is located within a sampling site.

Acoustic Monitor Setup

Be careful not to jostle the AudioMoth too much since the batteries could come loose, and you will lose your programming If you are deploying in a place where there is public traffic, it is advisable to attach a notice to the tree saying this is a recording device for a scientific study. Please include a contact number, so they can seek further information.

Once a tree is selected or a pole is driven into the ground, select an AudioMoth and scan the barcode into S123. Choose and SD card and scan its barcode before inserting into the AudioMoth. Insert a full complement of 3 fully charged Eneloop Pro batteries. Set the mode from USB/Off to Custom. On the side where you inserted the micro SD card, a green light should start flashing. If it is flashing both green and red, there is an error, and you will have to go back to your computer to reprogram. Close the AudioMoth container around the device, amking sure that the clip is properly fastened.

AudioMoths should be firmly attached to the tree or pole with their velcro strap such that the device is 1.5-2 m above the ground and the microphone hole is angled parallel to the slope of the terrain. When placing AudioMoths along tourist roads, emphasis should be put on using trees.

Record details of the AudioMoth deployment in S123. Ensure the location is recorded within 2 m of the AudioMoth.

Speak clearly near the AudioMoth to record the date, time, and grid ID (preferably use the geohash ID and not the label).

After setting up an AudioMoth, confirm all steps in the AudioMoth deployment checklist have been followed:

Confirm the AudioMoth has fully charged batteries (hint: Check the battery level on the configure AudioMoth app).
Confirm the AudioMoth has an empty SD card.
Confirm the AudioMoth has been programmed to sample for 17 hrs per day at a sampling rate of 48 khz and medium-low gain.
Confirm AudioMoth sampling periods were set using the correct timezone.
Confirm the recording duration is less than 3600.
Confirm the AudioMoth and SD card barcodes were scanned in S123.
Confirm all appropriate fields in Survey123 have been filled.
Confirm all features the AudioMoth is covering have been recorded.
Confirm the AudioMoth is at the correct height and parallel to the ground.
Confirm the AudioMoth is secure.
Confirm the AudioMoth is on.
Confirm someone spoke near the AudioMoth to record the date, time, and grid ID (preferably use the geohash ID and not the label).
Confirm all equipment and other items have been collected.

When returning to the office, one team member is responsible for checking the deployments in S123 to ensure that the locations were captured correctly.

Report any challenges or issues to the PI or project officer.

Acoustic Monitor Check

Check the deployment coordinates on S123 prior to departing the office and ensure the exact station coordinates are loaded into a handheld GPS (do not rely on the coordinates of the prospective sites from the survey design). Ensure that all equipment including spare AudioMoths and a full set of replacement batteries and SD cards are in the vehicle.

Approach the AudioMoth, undo to velcro to remove it from the tree of pole on which it was deployed and open the plastic casing. Check to see if there is a green light flashing on the side where you inserted the micro SD card. If it is flashing both green and red, there is an error, and you will have to go back to your computer to reprogram. If there are no lights shining then the AudioMoth has run out of battery power. Scan the AudioMoth into S123. As a team, determine what, if any steps need to be conducted. Switch the AudioMoth off. Replace batteries and SD card and scan the new SD card. Place the old SD card in a clearly marked bag. Record details in S123 and redeploy the AudioMoth.

Set the AudioMoth to to its normal operating setting by switching the mode from USB/Off to Custom. Speak clearly near the AudioMoth to record the date, time, and grid ID (preferably use the geohash ID and not the label).

After checking an AudioMoth, confirm all steps in the AudioMoth check checklist have been followed:

Confirm batteries were replaced and the AudioMoth has proper charge.
Confirm the SD card was replaced and the new card was scanned.
Confirm all metadata has been captured in S123.
Confirm the AudioMoth is at the correct height and parallel to the ground.
Confirm the AudioMoth is secure.
Confirm someone spoke near the AudioMoth to record the date, time, and grid ID (preferably use the geohash ID and not the label).
Confirm all equipment and other items have been collected.

Acoustic Monitor Retrieval

Undo velcro to remove the AudioMoth from the tree. Turn the AudioMoth off and scan the Audiomoth into S123. Capture all metadata in S123 and check the site for any equipment or other items that may be left behind.

6. POST PROCESSING

This section details all steps that need to be followed after returning to the office from the field to ensure that samples are properly processed and stored and that data is uploaded and saved to the correct location.

6.1 SAMPLE PROCESSING AND STORAGE

No sample processing is required.

6.2 DATA ENTRY AND UPLOADS

All metadata must be captured in the S123 app. If data cannot be entered into S123 in the field, then it must be entered when returning to the office and locations edited manually.

Within 3 days of retrieving SD cards from the field, all data must be uploaded to the ‘ns-ii-field-data-acoustic-landing’ blob container within the ‘nsiifielddataacoustic’ storage account under the ‘natural-state-field-data-sub’ subscription using the Microsoft Azure bulk uploader.

7.1 DESIGN DOCUMENTS

7.2 OTHER RELEVANT SOPS

SOP_RBP_C2Bio_v0.4.0
S123 Data upload SOP [in development]
BirdNET Validation SOP
BirdNET Calibration SOP

7.3 DATA ELEMENTS

8. REVISION AND VERSION HISTORY AND DESCRIPTION

v0.1.0 - Initial operating procedure following the RBS survey and adoption of S123 for capturing metadata. v0.2.0 - Sampling protocols were updated to include a second quiet period from midnight to 3:00 AM local time and standard sampling rates were increased from 32 kHz to 48 kHz. Clarification was also provided on adequate and optimal sampling durations. v0.3.0 - Updating sampling schedule to include a quiet period from midnight till 3am. Added screengrabs of audiomoth configuration app.

9. SIGNATURES OF CONFIRMATION

Principal Investigator: ______________ Date: ___________

Director of Impact Insights: ____________ Date: ___________

10. APPENDICES

None currently available