Acoustic approach as an alternative to human-based survey in bird biodiversity monitoring in agricultural meadows

silent for the most time (e.g., storks, herons, birds of prey), should be detected by a human- observer visually and undetected on the recorder, because they do not vocalise, or their vocali- sation is too low to be recorded. Additionally, some birds avoid the surrounding area of the point after the arrival of an observer [48], potentially leading to a decrease in the probability of detection of birds near to a survey point. Unlimited detection radius has been applied in many studies comparing the effectiveness of acoustic recorders and human-observers in bird biodiversity estimation. Most of them reported higher biodiversity detected by human-observers than by recorders [22,47], which is an expected result. In general, recorders have similar or lower detection range than human- Table 1. Results of three generalised estimating equations. All bird species B SE Wald χ2 df p (Intercept) 2.468 0.0602 1681.515 1 <0.001 Method [observer unlimited] 0.388 0.0266 212.675 1 <0.001 Method [observer 100 m] -0.400 0.0405 97.414 1 <0.001 Method [observer 50 m] -1.492 0.0717 433.160 1 <0.001 Survey [late] -0.023 0.0350 0.420 1 0.517 Observer ID [2] -0.284 0.0665 18.248 1 <0.001 Songbird species B SE Wald χ2 df p (Intercept) 2.154 0.0673 1024.928 1 <0.001 Method [observer unlimited] 0.252 0.026 94.254 1 <0.001 Method [observer 100 m] -0.349 0.0418 69.703 1 <0.001 Method [observer 50 m] -1.351 0.0729 343.587 1 <0.001 Survey [late] 0.049 0.0413 1.409 1 <0.235 Observer ID [2] -0.237 0.0727 10/633 1 <0.001

Farmland bird species B SE Wald χ2 df p (Intercept) 1.100 0.0648 288.471 1 <0.001 Method [observer unlimited] 0.444 0.0356 156.083 1 <0.001 Method [observer 100 m] 0.030 0.0414 0.519 1 0.471 Method [observer 50 m] -0.883 0.0656 181.295 1 <0.001 Survey [late] 0.145 0.0491 8.701 1 0.003 Observer ID [2] -0.182 0.0652 7.774 1 0.005 Meadow bird species B SE Wald χ2 df p (Intercept) 0.588 0.1126 27.311 1 <0.001 Method [observer unlimited] 0.360 0.0509 49.851 1 <0.001 Method [observer 100 m] -0.010 0.0530 0.034 1 0.854 Method [observer 50 m] -1.002 0.0963 108.374 1 <0.001 Survey [late] 0.011 0.0730 0.024 1 0.877 Observer ID [2] 0.043 0.1177 0.137 1 0.712 Models examining the effect of detection method (recorder, observer up to 50 m, observer up to 100 m, observer surveying birds in unlimited distance), time of survey (early or late) and observer (MJ or PS) on the number of detected bird species. Separate models were conducted for: (1) all bird species, (2) songbird species, (3) farmland bird species, and (4) meadow bird species. As a reference category we used: recorder for Detection method, early survey for Survey and first observer for Observer ID. https://doi.org/10.1371/journal.pone.0266557.t001 PLOS ONE Acoustic bird biodiversity monitoring in agricultural meadows PLOS ONE | https://doi.org/10.1371/journal.pone.0266557 April 8, 2022 8 / 16