Danang Bird Club Partners with Cornell Lab to Improve Merlin Bird ID in South East Asia

Danang Bird Club is now working with the Cornell Lab of Ornithology on an important initiative to help improve Merlin Bird ID in South East Asia. The project forms part of a wider effort to make Merlin’s Sound ID function work better across the whole Asian continent, where the app still has much more limited coverage than in North America and Europe.

For anyone who has used Merlin while birding in South East Asia, this will probably come as no surprise. The app can be remarkably good at recognising some familiar calls, but there are many species for which it either gives no suggestion at all or produces an uncertain result. This is particularly noticeable in forests, where many of the birds we encounter are more often heard than seen.

Merlin Sound ID is a machine-learning system, and like any such system it needs a large quantity of good-quality training data. Cornell's Merlin team uses sound recordings archived in the Macaulay Library, many of them submitted by birders through eBird. The recordings are converted into spectrograms and then annotated by experts, who identify and mark the individual bird sounds that the computer needs to learn. 

For many species, Merlin needs roughly 150 useful recordings before there is enough material to train a reliable model. Across Asia, Cornell is now working with local birding communities to collect and annotate recordings for species that are still under-represented. The current Asian project already covers hundreds of species, but there are many more for which either sufficient recordings exist but still need to be processed, or for which there are still insufficient recordings.

I am personally heading up the effort in the birding region of Indochina (Vietnam, Laos, Cambodia and Thailand) in my role as an annotator. My job is to review audio recordings submitted to eBird checklists and annotate the bird sounds they contain. This means looking at the recordings as spectrograms, listening carefully to the calls and songs, and identifying the relevant vocalisations so that they can be used as training material for Merlin. A recording may contain several bird species, insects, mammals, amphibians, reptiles, humans, vehicles, background noise or other sounds. Identifying exactly which songs and calls belongs to which species requires considerable experience. 

Annotating the spectrogram of an audio recording using the Merlin Vision software.

Good recordings are therefore extremely valuable. Merlin needs many more recordings for its data set. A recording does not have to be made with expensive equipment to be useful. Modern smartphones can produce perfectly usable recordings if the birder gets reasonably close to the bird without disturbing it, keeps still and quiet, and records for long enough to capture several examples of the vocalisation. Cornell currently recommends recordings of at least 30 seconds where possible, with clear examples of the target bird's vocalisation and as little interference as possible. Recordings attached to eBird checklists are archived in the Macaulay Library and can subsequently be used for research and to improve Merlin. Cornell specifically encourages birders to retain and upload their Merlin recordings rather than deleting them after an identification has been made. 

This is an opportunity for all DBC members to contribute directly to improving bird identification in Indochina. Whenever you are out birding, please make recordings as often as possible. This is especially useful for species that are difficult to see, species that are poorly represented by existing recordings, and species for which Merlin currently gives no useful result.

Several different recordings can be particularly valuable if they capture different individuals, locations or types of vocalisation. If several species can clearly be heard, the recording may potentially be useful for more than one species. Most importantly, don't worry if Merlin cannot identify your recording. That may actually make it especially interesting.

One particularly easy way to contribute is now built directly into the eBird mobile app. As of September 2026, birders can make a sound recording while entering an eBird checklist and upload the recording directly with the checklist, without needing a complicated workflow, or a laptop. The recording is then archived in the Macaulay Library along with the observation, making it available not only as evidence of the sighting but also as potential material for projects such as Merlin Sound ID. This makes it much easier for all of us to contribute useful audio while we are actually out birding. 

Vietnam has a large and diverse bird fauna, and improving Merlin here will require local recordings from local birders. The sounds of Vietnamese birds are something that international databases cannot obtain without the help of people actually going into the field and recording them. For DBC, this project is therefore more than just helping to improve an app. Every good recording submitted by a club member becomes part of a much larger scientific resource and may eventually help thousands of birders identify birds that they otherwise might never be able to see.

Cornell has provided a list of species that currently have fewer than 150 recordings in the Macaulay Library — in other words, species where additional good recordings could make a real difference to Merlin. The Vietnam list includes some birds that DBC members regularly encounter, such as Great Iora (43) and Green-billed Malkoha (78). Some species have far fewer recordings: Grey-faced Tit-Babbler has only 24, Indochinese Cuckooshrike 24, White-faced Plover just 15 and Vietnamese Greenfinch only 10. These numbers show where our efforts can be especially useful: if you encounter one of these species in Vietnam, please make a recording and add it to your eBird checklist. Even a single good recording brings that species one step closer to the 150-recording target, and if DBC members collectively concentrate on these under-recorded species, we can make a meaningful contribution to improving Merlin for Vietnam.

List of priority target species for Vietnam with fewer than 150 recordings:

1. White-nest SwiftletAerodramus fuciphagus
   45
2. Chinese Pond HeronArdeola bacchus
   42
3. Racket-tailed TreepieCrypsirina temia
   147
4. Green-billed MalkohaPhaenicophaeus tristis
   78
5. Medium EgretArdea intermedia
   39
6. Cinnamon BitternBotaurus cinnamomeus
   127
7. Blue-winged LeafbirdChloropsis moluccensis
   112
8. Dark-sided FlycatcherMuscicapa sibirica
   45
9. Burmese ShrikeLanius collurioides
   70
10. Great IoraAegithina lafresnayei
    43
11. Red-necked StintCalidris ruficollis
    113
12. Green-eared BarbetPsilopogon faiostrictus
    132
13. Tibetan Sand PloverAnarhynchus atrifrons
    140
14. Amur StonechatSaxicola stejnegeri
    56
15. Laced WoodpeckerPicus vittatus
    77
16. Flavescent BulbulPycnonotus flavescens
    77
17. Long-toed StintCalidris subminuta
    85
18. Black-and-red BroadbillCymbirhynchus macrorhynchos
    73
19. Grey-eyed BulbulIole propinqua
    112
20. Common FlamebackDinopium javanense
    122
21. Blue-winged MinlaActinodura cyanouroptera
    144
22. Chestnut-vented NuthatchSitta nagaensis
    149
23. Terek SandpiperXenus cinereus
    119
24. Mountain FulvettaAlcippe peracensis
    78
25. Necklaced BarbetPsilopogon auricularis
    41
26. Indochinese BarbetPsilopogon annamensis
    43
27. Grey-faced Tit-BabblerMixornis kelleyi
    24
28. Large WoodshrikeTephrodornis virgatus
    137
29. Bar-bellied PittaHydrornis elliotii
    77
30. Grey-cheeked WarblerPhylloscopus poliogenys
    112
31. White-faced PloverAnarhynchus dealbatus
    15
32. Indochinese CuckooshrikeLalage polioptera
    24
33. Curlew SandpiperCalidris ferruginea
    105
34. Indochinese RollerCoracias affinis
    149
35. Black-headed SibiaHeterophasia desgodinsi
    101
36. Vinous-breasted MynaAcridotheres leucocephalus
    51
37. Grey-headed LapwingVanellus cinereus
    136
38. Puff-throated BulbulAlophoixus pallidus
    144
39. Tiger ShrikeLanius tigrinus
    42
40. Violet CuckooChrysococcyx xanthorhynchus
    115
41. Asian Golden WeaverPloceus hypoxanthus
    80
42. Ruby-cheeked SunbirdChalcoparia singalensis
    96
43. Barred Cuckoo-DoveMacropygia unchall
    124
44. Painted StorkMycteria leucocephala
    60
45. Vietnamese GreenfinchChloris monguilloti
    10
46. Mountain Imperial-PigeonDucula badia
    138
47. Pacific SwiftApus pacificus
    86
48. Sulphur-breasted WarblerPhylloscopus ricketti
    73
49. Great KnotCalidris tenuirostris
    64
50. Broad-billed SandpiperCalidris falcinellus
    52

Maurizio is the founder of Danang Bird Club, a non-profit organisation focused on citizen science, education and advocacy for the conservation of Vietnamese avifauna. He is the author of Birdwatching in Central Vietnam; a guide to the region’s rich avian diversity. He is currently an annotator for Merlin Sound ID Asia, helping to train the app we all use to identify more Vietnamese bird sounds. With his tour company Avis, he offers science-informed wildlife experiences throughout Central Vietnam’s most iconic landscapes.