Data from: Early detection of human impacts using acoustic monitoring: an example with forest elephants

The impacts of human activities and climate change on animal populations often take considerable time before they are reflected in typical measures of population health such as population size, demography, and landscape use. Earlier detection of such impacts could enhance the effectiveness of conservation strategies, particularly for species with slow population growth. Passive acoustic monitoring is increasingly used to estimate occupancy and population size, but this tool can also monitor subtle shifts in behavior that might be early indicators of changing impacts. Here we use data from an acoustic grid, monitoring 1250 km2 of forest in the northern Republic of Congo, to study how forest elephants (Loxodonta cyclotis) assess the risk of poaching across a landscape that includes a national park as well as active and inactive logging concessions. By quantifying emerging patterns of behavior at the population level, arising from individual-based decisions, we gain an understanding of how elephants perceive their landscape along an axis of human disturbance. Forest elephants in relatively undisturbed forests are active nearly equally day and night. However, they become more nocturnal when exposed to a perceived risk such as poaching. We assessed elephant perception of risk by monitoring changes in the likelihood of nocturnal activity relative to differing levels of human activity. We show that logging is perceived to be a risk on short-time and small spatial scales but with little effect on animal density. However, risk avoidance persisted in areas with relatively easy access to poachers and in more open habitats where poaching has historically been concentrated. Increased nocturnal activity is a common response in many animals to human intrusion on the landscape. Provided a species is acoustically active, passive acoustic monitoring can measure changes in human impact at the early stages of such change, informing management priorities. Core data are 24-hour continuous sound files acquired on 50 digital sound recorders deployed in the northern Republic of Congo. Sound files are processed with a detection algorithm running in MatLab, which tags putative elephant rumble vocalizations. Output files from the detection process are reviewed to remove false positive detections. Data tables were compiled from the edited detector output files, which include the location, date, and time of each verified vocalization. Ecological metadata (season, habitat, study stratum, etc.) were then added to these files to construct analysis dataset tables. # Data from: Early detection of human impacts using acoustic monitoring: an example with forest elephants This README file was generated on 2024-07-09 by Peter H. Wrege ## GENERAL INFORMATION **Author Information** * Principal Investigator Contact Information * name: Peter H. Wrege * Institution: Cornell University * Email: [p.wrege@cornell.edu](mailto:p.wrege@cornell.edu) **Date of data collection:** 15Jan2017-27April2021 **Geographic location of data collection:** Nouabale-Ndoki National Park, Republic of Congo ## SHARING/ACCESS INFORMATION 1. **Licenses/restrictions placed on the data:** CC0 1.0 Universal (CC0 1.0) Public Domain 2. **Publications that cite or use the data:** Peter H. Wrege, Frelcia Bien-Dorvillon Bambi, Phial Jackel Ferdy Malonga, Onesi Jared Samba, Terry Brncic (2024). Early detection of human impacts using acoustic monitoring: an example with forest elephants. PLOS ONE 3. **Data is also available at:** The Registry of Open Data on Amazon Web Services “Sounds of Central African Landscapes”. This registry holds daily (~24hr) sound files recorded during the study, from each of the 50 monitoring sites. These sound files are 'soundscapes', recording all ambient sounds, from which forest elephant 'rumble' vocalizations were tagged using detection algorithms or by hand-browsing. 4. **Recommended citation for this dataset:** Wrege et al. (2024). Data from: Early detection of human impacts using acoustic monitoring: an example with forest elephants. Dryad Digital Repository. [https://doi:10.5061/dryad.x…](https://doi:10.5061/dryad.x…) ## DATA & FILE OVERVIEW 1\. File List: * (A) allEleObs_dep1_10.txt * (B) dielPctWeek_dep1_10.txt * (C) dielPctWeek_kabo_dep1_10.txt * (D) gunShots_dep1_10.txt 2\. Relationship between files: File (A) is a comprehensive listing of all acoustic signals (aggregated per hour), accepted as a 'rumble' vocalization. Files (B) & (C) are aggregations, by week, of file (A), with additional environmental and location variables added for statistical analysis. File (D) is independent of other files. 3\. Data tables are provided as tab-delimited text files. ### DATA-SPECIFIC INFORMATION COMMON TO MOST FILES 1\. Variable List: * site (txt) - one of 50 recorder locations * year (num) - year * month (num) - month * day (num) - day * week (num) - week of the year * fiscalYr (num) - study year, beginning 15Dec2017 * stratum (txt) - one of three study strata. values: * PNNN = national park * KaboE = active logging * Bonio = inactive logging * season (txt) - <60mm rain/month = dry * satHab3 (txt) - based on satellite image analysis: mixed forest (Fmixed), gilbertiodendron forest (Fmono), open * totDay (num) - number of calls recorded 0600-17:59 * totNite (num) - number of calls recorded 00:00-05:59 plus 18:00-23:59 * totCalls (num) - total calls in 24hr recording day * contDensity (num) - total calls in 24hr recording day (for use as continuous predictor) ### DATA-SPECIFIC INFORMATION FOR: allEleObs\_dep1\_10.txt 1\. Number of variables: 10 2\. Number of cases/rows: 980691 3\. Unique Variable List: * diel (txt) - 'day' = 06:00-17:59 hrs, 'nite' = 00:00-05:59, 18:00-11:59 hrs (i.e., other than day)\ * count (num) - number of individual rumbles in the given hour * useCode (txt) - code to restrict the use of observation in some types of date or time-sensitive data aggregates due to imprecise date information for a given sound file. Values: * n = do not use * all = all analyses * addZeroHr = some daily files that started 1hr later on a given date than desired. We examined the potential for bias because of this missing hour but decided no correction was necessary because the frequency of rumble calls was so sparse. * weekOK = sound files with uncertainty about the exact date, because of temporary malfunction of recorders, but accurate enough for aggregation on the week level (most analyses). 4\. Missing data codes: na = hour of call unknown (clock error), but date OK ### DATA-SPECIFIC INFORMATION FOR: dielPctWeek\_dep1\_10.txt 1\. Number of Variables: 14 2\. Number of cases/rows: 3422 3\. Unique Variable List: * contMonth (num) - the sequential month of the study, beginning with 12 = December 2017 (for plotting) * numInStrat (num) - number of recording sites within a stratum 4\. Missing data codes: none ### DATA-SPECIFIC INFORMATION FOR: dielPctWeek\_kabo\_dep1\_10.txt 1\. Number of Variables: 13 2\. Number of cases/rows: 610 3\. Unique Variable List: * exposHist5 (txt) - code for each recording site indicating the number of years since active logging within that sector of the logging concession. Values: * active = logging activity during that fiscal year * done1-done6 = logging ended 1-6 years previous * preExp = logging began in the following fiscal year 4\. Missing data codes: none ### DATA-SPECIFIC INFORMATION FOR: gunShots\_dep1\_10.txt 1\. Number of Variables: 8 2\. Number of cases/rows: 86 3\. Unique Variable List: * eventID (num) - unique identifier for each gunshot event, including all gunshots within 30 min of one another * sumShots (num) - number of individual discharges in the event 4\. Missing data codes: none