The development of Internet of Things (IoT) technology has increasingly become an interesting topic for discussion by researchers and in the industrial world. Fortino et al. [10] define IoT as a concept that integrates “smart objects” with the internet so that it is possible for them to communicate with one another. The IoT research field has been widely carried out in various aspects, Niaraki [11] categorizes it into four categories: infrastructure applications, environmental applications, consumer applications, and community applications. Meanwhile, the agriculture, plantation, and aquaculture sectors are included in the industry category, which is part of infrastructure applications. In his paper, Niaraki describes that IoT research implemented in the industry (agriculture, plantation, and aquaculture) has experienced a rapid increase since 2019. IoT is part of the industrial revolution 4.0 where researchers and practitioners collaborate with machine learning, big data and cloud computing, and Artificial Intelligence/Machine Learning [12]. The IoT domain has expanded, and there is now the term Internet of Underwater Things, which is a concept that integrates underwater autonomous vehicles (AUVs) and underwater wireless sensor networks (UWSNs) supported by Internet services. In general, there are 3 main layers in IoUT, i.e., the application and service layer, the networking layer, and the physical layer [13]. In the healthcare sector, IoT could be implemented in patient monitoring, increasing the efficiency with real-time data and improving service performance through remote monitoring [14]. IoT can also be implemented in remote health monitoring, which is integrated into a wireless body area network (WBAN) for monitoring critical patients [15]. The IoT development is combined with machine learning (ML) as implemented by Bhoi et al. [16] for an environmental monitoring system with features of carbon dioxide (CO₂), humidity, temperature, and carbon monoxide (CO). In the field of agrotechnology, IoT plays a role in monitoring and increasing food production by integrating various sensors, actuators, and protocols as a tool of communication between the user and the system [17]. The main focus of IoT research in agrotechnology is on sensors used in IoT, IoT controllers, IoT communication standards, and IoT-based smart machinery [18]. While in the field of aquaculture, which is also related to this study in IoT, in general, there are two challenges: the first is common challenges that relate to how IoT devices interact with waters, while the second is specific challenges that relate to species and accessibility with water aquaculture objects [19]. In this study, IoT is the basic foundation of a system of communicating with users that has an impact on increasing fish productivity through an aquaculture monitoring system.

The production from aquaculture is increasing, so innovation and technology are needed to support market needs. One of the technological implementations is the IoT in the Aquaculture Monitoring Water System. Chiu et al. [20] integrate sensor devices and Wi-Fi modules to control water quality, which is combined with deep learning random forest modeling to predict fish production potential, with the result being a smart aquaculture monitoring device. The implementation of temperature and pH sensors in the water monitoring system is carried out to see the actual conditions that have an impact on fish growth, using an Arduino Mega 2560 and Wi-Fi Module ESP8266 [21]. IoT-based Smart Aquaculture System (ISAS) has been developed by Tsai et al. [22] to control water quality automatically and replenish water to keep the growth of fish and shrimp stable. ISAS can provide automatic feeding through fuzzy rules, and the result is a shrimp survival rate that has increased by 33.3%. An AI buoy system combined with IoT has been developed to measure real-time data on salinity, water temperature, dissolved oxygen, and velocity in a low-cost and easy-to-build system [23]. The implementation of IoT in aquaculture provides benefits for fish cultivators to get environmental conditions in real-time, remotely, and accurately [24]. Hawari et al. [25] developed an IoT monitoring water system device with a background of climate and weather changes locally in Malaysia, which have caused a decrease in fish production levels. A communication system built on cloud portals and mobile devices that provide accurate and real-time information.

A microcontroller is a component with a minimalist size that serves as a system controller [26]. Microcontrollers are an integral part of helping humans work; there are 250 billion microcontrollers in electronic devices that are used daily [27]. Davis innovates the microcontroller as a low-cost dual-polarity high-voltage controller, which provides a stable high-voltage output in both positive and negative modes [28]. In the field of aquaculture technological innovation, microcontrollers are widely used as part of system development [29]. In its development, the microcontroller not only integrates sensors and actuators but furthermore can be integrated with cloud data and machine learning or deep learning [30]. Saha et al. [31] describe in their paper a microcontroller that can be integrated with deep learning to become an intelligent system. Whereas in this study, the role of the microcontroller is to control data transfer and monitoring from sensor devices to an output that can be accessed by end users.

The architectural system used in this study is represented in Fig. 2 below, where there are three main layers: (1) the physical layer, which consists of sensors and microcontroller devices; (2) the network layer, in the form of a Wi-Fi module that connects the physical layer with internet services; and (3) the application layer, in the form of a mobile-based application that can be accessed by end users to monitor the current conditions of the observed environment.

In Fig. 2 above, several sensors, i.e., a turbidity sensor, a pH sensor, and a temperature sensor, read the environmental conditions of catfish pond water. The reading data from the sensors is received by the microcontroller device and then sent to the data cloud via the Wi-Fi module as an accessibility connection that connects physical layer devices with internet services. Real-time data is uploaded to the cloud system and then accessed by end users via an Android-based application. End users can see the current condition of catfish pond water in the form of temperature conditions, level of turbidity, and level of pH (acidity or alkalinity). In Fig. 3, can be seen in detail the system flow of the Firebase cloud data as a real-time data storage medium from sensors that can be accessed by end users. IoT devices via Wi-Fi modules, communicate with cloud systems via the Message Queuing Telemetry Transport (MQTT) protocol. The MQTT protocol is one of the communication protocols that are commonly used for the IoT and uses standard ISO subscription publishing-based messages that work in the Transmission Control Protocol/Internet Protocol (TCP/IP), which does not have a special address either between the servers or with other connected devices. The MQTT protocol was specifically designed for machine-to-machine communication [32].

MQTT works as a broker or server that has the function of filtering incoming messages and distributing them to clients who are interested in receiving them. Each client can be a publisher (message sender) or subscriber (recipient of messages). In this study, IoT devices work as clients that publish data from sensors that are received and sent to brokers. The MQTT protocol forwards real-time data to Firebase via a cloud gateway. Mobile applications for end users are developed based on Android, which requires the Rest API as a link between cloud Firebase data and end users. Rest stands for Representational State Transfer and is a type of software architecture. Rest represents a set of constraints that must be followed when developing applications on the internet. Application Programming Interface, or API, is a set of standards and protocols that integrate users with applications so that users can access and use the application software functions. Through the Rest API, developed applications can have well-organized interfaces and standardized procedures that are easy to represent. This also facilitates communication between devices and users.

Fig. 11 shows a flowchart of a system built with several processes. The first stage begins with reading the data from the sensor in the fish pond. The data read is the value of temperature, the value of pH, and the value of turbidity. The microcontroller receives data from the sensor device, and then the Wi-Fi module connects the data to be stored on Firebase via the MQTT protocol. Next, the MQTT protocol distributes data into Firebase as real-time data that can be accessed directly by end users through Android-based applications.

Data collection was carried out by testing IoT devices in catfish ponds with a time range of 06.00 a.m. to 08.00 p.m. and a break every 2 h. The data that appears on the mobile app is real-time data from sensor readings accessed via Firebase. The dimensions of the fish pond are 5 m long, 3 m wide, and 1.5 m deep. Fish pond environmental temperatures range from 25°C to 30°C with occasional light rain. Data collection was carried out in as many as 8 categories of data collection time, as shown in Table 2 below.

Fig. 12 is a configuration of data collection where the sensor interacts directly with the fish pond. Turbidity sensors, pH sensors, and temperature sensors are connected directly to the microcontroller via a connector so that the data can be read by the microcontroller. The microcontroller is connected to the Wi-Fi module device, and data can be uploaded to Firebase.

Fig. 13 is the actual condition of the catfish ponds, where there are several fish ponds. However, this study only focused on 1 fish pond, the orange line is the data collection area.

To test the accuracy of the data generated by the sensor in an IoT device, a test is needed by comparing the output data from the sensor with data based on measurements of turbidity, pH, and temperature measuring instruments. This study used a temperature measuring pen to measure temperature, a portable turbidity meter to measure turbidity, and a digital pH meter to measure the pH level, as shown in Table 3 below.

Table 4 shows the measurement results of IoT devices and measurement tools, which consist of 8 specific time categories and are measured in Celsius (C). The highest measurement results for both IoT devices and measurement tools are at noon. When compared between the results of these two devices, the gap is not too high, while the highest measurement gap is in the time range at 06.00 am, where there is a 1.1°C difference between the results of measurements using measurement tools (24.2°C) and the results of IoT device measurements (25.3°C). As for the gap measurements at other specific times are less than 1°C, and if it is averaged, the gap between IoT devices and measurement tools is 0.4°C.

Visually, in Fig. 18, it can be seen that the gap between the IoT device (blue line) and the measurement results (red line) is not that significant. Between 10.00 a.m. and 12.00 p.m., both devices detected an increase in temperature and a decrease in temperature between 02.00 p.m. to 08.00 p.m.

The measurement results of IoT devices and measurement tools for pH levels show a gap less than or equal to 0.3, as shown in Table 5. The highest gap at noon was 0.3, where IoT devices showed a pH value of 7.0, and measurement tools showed a pH value of 7.3. The highest accuracy in data collection was at 04.00 PM and 08.00 PM where both devices show a pH value of 7.0.

Fig. 19 is a graphic of the pH measurement result where at noon there was an increase in the pH level based on the measurement tool result. This was because at that hour it was time for fish feeding, so it increased the pH level in the pond. While the measurement results of IoT devices at 12.00 p.m. and 02.00 p.m. did not change, and had a gap of 0.3 lower than the results of the measurement tool.

Turbidity level measurements were carried out using the Nephelometric Turbidity Unit (NTU). In Table 6, it can be seen that the turbidity measurement results through the IoT device produce the measurement values below through the measuring instrument. The highest gap between the two devices occurred between 08.00 a.m. and 10.00 a.m. with a value of 2.8 NTU. In aggregate, the average gap between IoT Devices and Measurement Tools is 2.2 NTU.

The graphical visualization is shown in Fig. 20 where the blue line shows the data generated by the IoT device and the red line shows the data generated by the measurement tool. In both devices there was an increase in turbidity at 12.00 p.m. and 02.00 p.m, this was because at that time it was the fish feeding schedule that increased the turbidity level in fish ponds.

The error percentage calculation is performed by finding the absolute difference between the expected value (νE) which is the value resulting from the measurement tool result and the actual value observed (νA) which is the result of IoT devices, Eq. (1) is the percentage error formulation

(1)∂=νA−νEνE×100%

Based on the results of calculating the error percentage which refers to the measurement results of IoT devices and measurement tools in Tables 3–5, the resulting average error percentage is represented in Table 7.

Table 7 above results in an average temperature error percentage of 1.75%; this indicates that the accuracy value of the temperature data generated by the IoT device is 98.25% (which is taken from 100% minus 1.75%). The pH error percentage shows 1.94%, which indicates an accuracy percentage of 98.06%. Meanwhile, the turbidity error percentage shows 9.78% or an accurate value of 90.22%. Of the three error percentage results above, the overall average error between temperature error percentage, pH error percentage, and turbidity error results in a percentage of 4.49%. From the percentage error, the level of accuracy can be measured with the Eq. (2):

(2)A=100%−∂¯

where A is the percentage of accuracy with a 100% formulation minus ∂¯. ∂¯ is the average error percentage of 4.49%. So the results of formulation produce an accuracy percentage of 95.51%.

Cost analysis is carried out by conducting a comparative analysis using a labor-based. The elements that are calculated in the IoT device consist of (a) the use of electric power in units of watts and (b) the prediction of maintenance costs in one month. The electricity cost charged is standard in Indonesia at Indonesian Rupiah (IDR) 1,352 kilowatt-hours (kWh). Based on the conditions of the IoT device, the maximum consumption of electric power is 20 watts with full-day or 24-h usage. The consumption of electric power is 20 watts, so it is necessary to convert the unit cost of electricity into watts per hour (Wh). So the parameters used in the calculation is IDR 1.352 Wh. The formula used to calculate the daily electricity load is maximum power consumption multiplied by the number of hours of use, multiplied by the cost of electricity (IDR 1.352 Wh). In this case, 20 watts × 24 h × IDR 1.352 Wh results in a daily cost of IDR 648.96 for device usage. If calculated for a month (30 days), it generates IDR 19,468.8 for costs charged to IoT devices. However, using an IoT device requires additional maintenance costs; in this case, the maintenance charge for the device is IDR 50,000 per month. Meanwhile, for the use of daily labor, the standard cost in Indonesia is IDR 40,000 per day with a work duration of 1 to 2 h. If calculated for one month (30 days), it will generate a cost of IDR 1,200,000. Table 8 shows a comparison of the monthly costs incurred between the IoT device and the daily labor-based one.

In Table 8 above, it can be seen that there is a significant difference between aquaculture monitoring using IoT devices and labor-based, wherein monthly usage for IoT devices, taking into account maintenance costs, results in a cost of IDR 69,468.8. While labor-based, the total monthly fee is IDR 1,200,000. In terms of monthly cost effectiveness, IoT device-based aquaculture has a very high advantage when compared to labor-based. This can be calculated by the difference between labor-based and IoT devices which results in IDR 1,130,531. This means that the effective cost of IDR 69,468.8 (IoT device) is IDR 1,130,531 (labor-based), which is less than IDR 1,200,000. In terms of percentage, the effective value formula is divided by the original value (labor-based = IDR 1,200,000) and multiplied by 100% to produce 94.21%. So, the effective cost of IoT devices is 94.21% when compared to labor-based costs.