Food codes as an AI-based framework to investigate bird foraging ecology and interspecific interactions

IF 1.9 4区生物学 Q2 BIOLOGY

Biosystems Pub Date : 2025-09-12 DOI:10.1016/j.biosystems.2025.105591

Almo Farina , Luca Biancardi , Giovanni Ancillotti

{"title":"Food codes as an AI-based framework to investigate bird foraging ecology and interspecific interactions","authors":"Almo Farina , Luca Biancardi , Giovanni Ancillotti","doi":"10.1016/j.biosystems.2025.105591","DOIUrl":null,"url":null,"abstract":"<div><div>Food codes activated by provisioning food through a garden bird feeder have proven to be an effective tool for investigating bird behavior and bird–human interactions. The experimental setup included a garden feeder with carefully controlled variables, such as its location and structure, the type of food provided, and its temporal distribution.</div><div>A total of 2.8 million image frames of birds at feeder were captured between November 2023 and May 2024 using a time-lapse camera system. Of these, 1,232,456 frames were classified through supervised image processing using the Squeezebrains SDK (<span><span>https://fabervision.com</span><svg><path></path></svg></span>), an AI-based image analysis tool specifically developed for wildlife monitoring. This approach enabled detailed insights into feeding preferences and interspecific interactions.</div><div>Nine bird species were identified: Great tit (<em>Parus major</em>) (65.59 % of all visits), blue tit (<em>Cyanistes caeruleus</em>) (13.62 %), house sparrow (<em>Passer domesticus</em>) (8.28 %), red-billed leiothrix (<em>Leiothrix lutea</em>) (5.90 %), siskin (<em>Carduelis spinus</em>) (2.84 %), chaffinch (<em>Fringilla coelebs</em>) (1.69 %), dunnock (<em>Prunella modularis</em>) (1.00 %), European robin (<em>Erithacus rubecula</em>) (0.59 %), and common blackbird (<em>Turdus merula</em>) (0.45 %). These species, belonging to different genera, visit the feeder at different times. The cluster analysis has categorized the bird species into three distinct groups based on their temporal preferences, leaving the chaffinch as an unclustered species: Group 1 (Blue tit, great tit, house sparrow, red-billed leiothrix), group 2 (Dunnock, siskin), group 3 (Common blackbird, European robin).</div><div>The potential interspecific competition, estimated by the number of frames shared concurrently by two species, was highest for the blue tit, great tit, and red-billed leiothrix, respectively. The chaffinch, siskin, common blackbird, European robin, and dunnock exhibited the highest scores for intraspecific, non-shared frames. Severe weather conditions appear to increase the number of visits to feeders.</div><div>Overall, food codes represent a robust approach, providing valuable insights into bird community dynamics while offering perspectives relevant to ecological management.</div></div>","PeriodicalId":50730,"journal":{"name":"Biosystems","volume":"257 ","pages":"Article 105591"},"PeriodicalIF":1.9000,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Biosystems","FirstCategoryId":"99","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0303264725002011","RegionNum":4,"RegionCategory":"生物学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"BIOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Food codes activated by provisioning food through a garden bird feeder have proven to be an effective tool for investigating bird behavior and bird–human interactions. The experimental setup included a garden feeder with carefully controlled variables, such as its location and structure, the type of food provided, and its temporal distribution.

A total of 2.8 million image frames of birds at feeder were captured between November 2023 and May 2024 using a time-lapse camera system. Of these, 1,232,456 frames were classified through supervised image processing using the Squeezebrains SDK (https://fabervision.com), an AI-based image analysis tool specifically developed for wildlife monitoring. This approach enabled detailed insights into feeding preferences and interspecific interactions.

Nine bird species were identified: Great tit (Parus major) (65.59 % of all visits), blue tit (Cyanistes caeruleus) (13.62 %), house sparrow (Passer domesticus) (8.28 %), red-billed leiothrix (Leiothrix lutea) (5.90 %), siskin (Carduelis spinus) (2.84 %), chaffinch (Fringilla coelebs) (1.69 %), dunnock (Prunella modularis) (1.00 %), European robin (Erithacus rubecula) (0.59 %), and common blackbird (Turdus merula) (0.45 %). These species, belonging to different genera, visit the feeder at different times. The cluster analysis has categorized the bird species into three distinct groups based on their temporal preferences, leaving the chaffinch as an unclustered species: Group 1 (Blue tit, great tit, house sparrow, red-billed leiothrix), group 2 (Dunnock, siskin), group 3 (Common blackbird, European robin).

The potential interspecific competition, estimated by the number of frames shared concurrently by two species, was highest for the blue tit, great tit, and red-billed leiothrix, respectively. The chaffinch, siskin, common blackbird, European robin, and dunnock exhibited the highest scores for intraspecific, non-shared frames. Severe weather conditions appear to increase the number of visits to feeders.

Overall, food codes represent a robust approach, providing valuable insights into bird community dynamics while offering perspectives relevant to ecological management.

查看原文本刊更多论文

基于人工智能的鸟类觅食生态学和种间相互作用研究框架——食物编码。

通过花园喂鸟器提供食物激活的食物代码已被证明是研究鸟类行为和鸟与人互动的有效工具。实验设置包括一个花园喂食器，精心控制变量，如其位置和结构，提供的食物类型和时间分布。在2023年11月至2024年5月期间，使用延时相机系统共拍摄了280万帧喂食鸟的图像。其中，1,232,456帧通过使用Squeezebrains SDK （https://fabervision.com）进行监督图像处理进行分类，这是一种专门为野生动物监测开发的基于人工智能的图像分析工具。这种方法使我们能够详细了解进食偏好和种间相互作用。共鉴定出9种鸟类：大山雀（Parus major）（65.59%）、蓝山雀（Cyanistes caeruleus）（13.62%）、家雀（Passer domesticus）（8.28%）、红嘴鸟（leiothrix lutea）（5.90%）、金雀（Carduelis spinus）（2.84%）、凤头燕雀（Fringilla coelebs）（1.69%）、褐叶雀（Prunella modularis）（1.00%）、欧洲知更鸟（Erithacus rubecula）（0.59%）和黑鹂（Turdus merula）（0.45%）。这些种类属于不同的属，在不同的时间访问喂食器。聚类分析根据鸟类的时间偏好将其划分为三个不同的类群，而苍头燕雀则是一个未聚类的物种：第一类（蓝山雀、大山雀、家雀、红嘴鸟），第二类（雀、黄雀），第三类：普通黑鹂、欧洲知更鸟。潜在的种间竞争（由两个物种同时共享的帧数估计）分别以蓝山雀、大山雀和红嘴长嘴鸟最高。苍头燕雀、黄雀、黑鹂、欧洲知更鸟和雀在种内非共享框架上得分最高。恶劣的天气条件似乎增加了前往喂食器的次数。总的来说，食物法典代表了一种强有力的方法，为鸟类群落动态提供了有价值的见解，同时提供了与生态管理相关的观点。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Biosystems 生物-生物学

CiteScore

3.70

自引率

18.80%

发文量

129

审稿时长

34 days

期刊介绍： BioSystems encourages experimental, computational, and theoretical articles that link biology, evolutionary thinking, and the information processing sciences. The link areas form a circle that encompasses the fundamental nature of biological information processing, computational modeling of complex biological systems, evolutionary models of computation, the application of biological principles to the design of novel computing systems, and the use of biomolecular materials to synthesize artificial systems that capture essential principles of natural biological information processing.