Future of Edge AI in biodiversity monitoring

TL;DR

It is argued that edge computing offers opportunities for responsive biodiversity management, but realising this potential requires increased collaboration between ecologists, engineers, and data scientists to align model development and system design with ecological questions, field constraints, and ethical considerations.

Abstract

1. Many ecological decisions are slowed by the gap between collecting and analysing biodiversity data. Edge computing moves processing closer to the sensor, with edge artificial intelligence (AI) enabling on-device inference, reducing reliance on data transfer and continuous connectivity. In principle, this shifts biodiversity monitoring from passive logging towards autonomous, responsive sensing systems. In practice, however, adoption remains fragmented, with key architectural trade-offs, performance constraints, and implementation challenges rarely reported systematically. 2. Here, we analyse 82 studies published between 2017 and 2025 that implement edge computing for biodiversity monitoring across acoustic, vision, tracking, and multi-modal systems. We synthesise hardware platforms, AI model optimisation, and wireless communication to critically assess how design choices shape ecological inference, deployment longevity, and operational feasibility. 3. Publications increased from 3 in 2017 to 19 in 2025. We identify four system types: (I) TinyML, low-power microcontrollers (MCUs) for single-taxon or rare-event detection; (II) Edge AI, single-board computers (SBCs) for multi-species classification and real-time alerts; (III) Distributed edge AI; and (IV) Cloud AI for retrospective processing pipelines. Each system type represents context-dependent trade-offs among power consumption, computational capability, and communication requirements. 4. Our analysis reveals the evolution of edge computing systems from proof-of-concept to robust, scalable tools. We argue that edge computing offers opportunities for responsive biodiversity management, but realising this potential requires increased collaboration between ecologists, engineers, and data scientists to align model development and system design with ecological questions, field constraints, and ethical considerations.

Figures (6)

• Figure 1: Conceptual overview of edge computing systems in biodiversity monitoring. Edge computing integrates four key components to enable autonomous, in-situ data processing and decision-making. (1) Biodiversity data are captured through various sensing modalities including vision-based camera systems, acoustic recorders, tracking technologies, environmental sensors, and environmental DNA. (2) AI inference occurs on resource-constrained hardware platforms, where data undergo preprocessing before being analysed by classical ML algorithms or DL models. Model optimisation techniques, including quantisation and graph optimisation, combined with specialised inference libraries, enable efficient on-device computation despite limited memory capacity, processor cores, and clock frequency. (3) Wireless networks support selective data transmission from sensor nodes to network gateways via cellular, LoRa, or WiFi before final storage in remote infrastructure. (4) Ecological insights derived from edge computing systems include wildlife monitoring and human-wildlife conflict management.
• Figure 2: Different categories of hardware chips (A) Microcontroller units (MCUs) integrate all core elements of a computer system, which are a central processing unit (CPU), a volatile memory (i.e. RAM), and programmable input and output (I/O) peripherals into one single chip. (B) Microprocessor units (MPUs) integrate only the CPU on a chip. They are often mounted on a printed circuit board (PCB) and interact with other computing elements (i.e., RAM, I/O Control, SSD Storage) via conductive tracks. (C) Systems-on-chip (SoCs) integrate multiple CPUs along with additional processing chips (i.e. graphic processing unit (GPU), digital signal processing (DSP), radio communication modules, and MEMS sensors (e.g. accelerator, pressure, microphones) in one compact chip.
• Figure 3: Hardware Trade-Offs in edge computing. Platform selection balances three primary considerations: AI model complexity, energy consumption, and device cost. Triangle sizes illustrate these trade-offs. MCUs (small yellow) typically feature single core, around 100 kB memory, and operate at frequencies near 100 MHz, while MPUs and SoCs (medium orange and large brown) offer multiple CPUs cores, gigabytes of memory, and operate at frequencies above 1 or 2 GHz, with SoCs incorporating specialised processors (GPU, NPU, TPU). Quantitative metrics for specific hardware platforms are provided in Appendix (\ref{['appendix:hardware_table']}).
• Figure 4: Conceptual overview of edge AI systems types for biodiversity monitoring.I TinyML systems (or Edge AI on MCUs) run quantised models on microcontrollers (MCUs) for specific events or species detections, transmitting AI inference results via low-power networks such as LoRaWAN. Common for acoustic and movement applications. II Edge AI systems (or Edge AI on SBCs) execute computationally intensive models on single-board computers (SBCs) for real-time multi-species classifications, using wide range of connectivity. Common in vision-based and acoustic applications. III Distributed Edge AI coordinates multiple sensing nodes with a centralised gateway, allowing spatially distributed data capture while aggregating or coordinating inference locally. Well suited to landscape-scale monitoring. IV Cloud AI streams raw data to remote servers, trading transmission energy and inference latency for model sophistication. Common when no sufficiently accurate pre-trained models are available, or for retrospective long-term analysis.
• Figure 5: Publications Analysis Overview Studies using edge computing systems for biodiversity monitoring are summarised by (A) year of publication, (B) processor characteristics and data modalities, (C) country, and (D) ecological and technical categories including monitored taxa, data types, hardware platforms, AI models optimisation techniques, and supported connectivity. Results based on data extracted from 82 articles (See Supplementary Materials \ref{['sec:data_availability']}).