Bridging expert judgement and simulated evidence: A cross-validated iot technology stack for African catfish (Clarias gariepinus) fingerling survivability

Résumé

Selecting an appropriate Internet of Things (IoT) technology stack comprising sensors, actuators, connectivity technologies, edge-computing devices, and cloud platforms is a critical step in developing aquaculture monitoring systems, yet technology selection is often unsystematic. This challenge is particularly important in resource-constrained environments, where cost, power availability, and connectivity reliability constrain system deployment and operation. This study aimed to identify, evaluate, and validate an appropriate IoT technology stack for monitoring African catfish (Clarias gariepinus Burchell, 1822) fingerling survivability. A mixed-methods sequential explanatory research design was adopted, integrating quantitative expert-survey data with simulation-based evidence. First, a structured questionnaire was administered to 48 IoT experts, of whom 87.5% reported direct implementation experience. The questionnaire demonstrated good internal consistency (Cronbach's α = 0.82). Survey data were analyzed using descriptive statistics, including frequencies, percentages, and weighted rankings, to identify preferred sensors, actuators, connectivity technologies, edge devices, cloud platforms, and implementation priorities. Second, six simulation and prototyping platforms were comparatively evaluated against the requirements of aquaculture IoT development. Third, an independent 300-observation quasi-experimental simulation was conducted to examine the effects of key water-quality parameters on fingerling survivability. Multiple linear regression was used to estimate standardized sensitivity indices, which were subsequently compared with expert rankings for cross-validation. Dissolved oxygen sensors (16.79%), ammonia/nitrite sensors (14.18%), and temperature and pH sensors (11.57% each) received the highest expert rankings. For actuation, aerators and pumps (34.91%) and automated water-exchange systems (29.25%) received the highest rankings. Raspberry Pi Pico microcontrollers were identified as suitable edge-computing devices, with Message Queuing Telemetry Transport (MQTT) recommended for device communication, Wi-Fi for relatively close-range deployments, and Global System for Mobile Communications/Long-Term Evolution (GSM/LTE) for remote sites. ThingSpeak and Blynk were identified as suitable cloud platforms for data management and visualization. The principal implementation challenges were security (21.88%), interoperability (20.83%), and power consumption (19.79%). The simulation produced sensitivity indices of 0.7875 for dissolved oxygen, 0.7620 for ammonia, and 0.5462 for temperature, corresponding to the parameters receiving the strongest expert endorsement. The convergence between expert rankings and simulation-derived sensitivity indices provides evidence of consistency between practitioner judgment and simulated system behavior. The study therefore proposes a low-cost IoT technology stack comprising dissolved-oxygen and ammonia sensors, aeration and water-exchange actuators, Raspberry Pi Pico edge devices, MQTT-based communication, and ThingSpeak/Blynk cloud platforms for resource-constrained aquaculture environments. Adoption of standardized communication protocols, secure data transmission using Transport Layer Security (TLS), and low-power hardware designs is recommended to address interoperability, security, and energy constraints. The methodological framework also provides a basis for combining expert assessment with simulation-based evidence when evaluating IoT technologies for precision aquaculture and related agricultural applications.