Research on the optimization of smart agriculture based on machine learning: Taking Qingdao as an example

Abstract

Faced with the challenges of rural labor shortage, aging population, and food security pressure, machine learning supported intelligent agriculture has become a promising approach to promote sustainable development in rural areas. However, existing research often overlooks the locality of technology adoption and its impact on local economic development. This study takes the coastal city of Qingdao in China as a typical case, examining the constraints and optimization strategies of machine learning intelligent agriculture from a local development perspective to address this gap. Through literature review and semi-structured interviews, four core bottlenecks were identified: technological adaptation barriers, insufficient power infrastructure, shortage of human capital, and lack of collaborative mechanisms. This study proposes targeted and situation specific improvement paths that are consistent with local resource endowments and institutional conditions. The research results not only provide practical guidance for the upgrading of smart agriculture in Qingdao, but also contribute to the literature on digital agriculture and local economic development, and provide valuable insights for other coastal areas facing similar challenges.

Keywords

machine learning smart agriculture internet of Things optimization strategy Qingdao

Introduction

Smart agriculture is a new revolution. Under the background of agriculture and rural areas in the future, according to the current assumption, smart agriculture through machine learning will become the mainstream in the next five to 10 years, while machine learning will promote the automation and intelligence of agricultural production and improve the efficiency of agricultural production with the wide application of artificial intelligence (Mao et al., 2018). The core of machine learning is to enable computers to systematically learn and train by using large models and image recognition, sensors, and the Internet of Things, and to use intelligent methods of new algorithms, new modes, and new platforms, including artificial vector machine SVM, artificial neural network, random forest RF, and decision tree DT. The core concept is shown in Figure 1.

Figure 1.

Core concepts of machine learning.

In 2024, Qingdao’s rural population was only 2.267 million. With the hollowing out and aging of rural areas in Qingdao, the development of agriculture has been severely challenged. The total output value of agriculture, forestry, animal husbandry, and fishery in Qingdao is 97.38 billion yuan, with a per capita disposable income of 31,560 yuan for rural residents. The city’s total vegetable production is 7.039 million tons, total fruit production is 1.3053 million tons, and total meat, egg, and dairy production is 1.1867 million tons, showing an overall upward trend. Qingdao’s agriculture presents a situation of combining traditional agriculture with modern smart agriculture, and in recent years, traditional agricultural production methods have been challenged. Given the complexity and diversity of local agriculture in Qingdao, traditional agricultural production methods are affected by factors such as weather, resulting in low yields, weak technological progress, and a large amount of abandoned land. In order to increase the income of agricultural and rural farmers, help to quickly achieve rural revitalization, achieve high-quality development of agriculture and rural areas in Qingdao, promote rural industrial revitalization, and ensure agricultural informatization and intelligence, we aim to achieve this goal. At the same time, some rural residents give up agricultural production due to the high wages brought by working in cities, resulting in uneven distribution of wealth. At the same time, there are also some problems with machine learning for smart agriculture, such as poor model adaptability, farmers not knowing how to use it, and high IoT costs. Some rural residents abandoned agricultural production due to the high wages brought by working in the city. In order to solve the problems of reduced rural labor, poor model adaptability, farmers’ inability to use it, and high IOT costs, machine learning can significantly solve this problem. Through machine learning, smart agriculture can be improved and optimized to realize the revitalization of rural industries in the future. Starting from local development, this study examines how the technological efficiency of the differentiated rural economy in Qingdao depends on local infrastructure, social, and institutional conditions. This paper uses the methods of interviews and literature to analyze and make suggestions for the optimization and upgrading of smart agriculture in Qingdao.

Smart agriculture has become a powerful means to improve the efficiency of agricultural production, while providing favorable support for food security and high-quality agricultural and rural development. Through the research of water resources (Li et al, 2025), it shows that the current trade model cannot effectively alleviate the pressure of water and food security in water-deficient areas, and can reallocate water resources through international trade, so as to alleviate the water pressure in water deficient areas and strengthen food security. In other areas of food security, at the same time, relevant scholars have proposed an intelligent system based on open source technology, which uses sensing ground parameters, such as soil moisture, soil temperature, and environmental conditions, as well as weather forecast data from the Internet. Perception nodes, involving ground and environment perception (Goap et al., 2018), sometimes intelligent irrigation systems also appear under the typical characteristics of large evaporation and high precipitation variability (Nouri et al, 2023). Relying on the embedded system, other professional systems are used to maintain its stability, thereby reducing the complexity of coping with extreme situations of the system. In some cases, hardware perception benchmarks show that real-time prediction on ultra-low power devices is feasible. Using decision trees and random forest models can achieve an accuracy of up to 86%, while maintaining a low memory footprint and energy consumption (Li et al, 2026). Some scholars believe that intelligent agriculture using embedded systems should not pay more attention to irrigation systems (Lachgar et al., 2025). On the other hand, if there is no irrigation system in intelligent agriculture, resulting in the lack of timely supply of water resources, some scholars have added the use of sensors and algorithms to collect and analyze data in the research and development of embedded systems to optimize water resources management. The system adjusts irrigation levels to specific crop needs, thereby contributing to more sustainable water resource management. An ML algorithm (such as a linear regression algorithm) is used to model the relationship between environmental factors and crop water demand so that the required irrigation level can be accurately predicted according to the data collected by the sensor. Using an embedded system, combined with temperature, humidity, and water level sensors, an automatic and efficient system is developed to collect data in real time and process it for decision-making (Morchid et al., 2025). At present, especially in the intelligent irrigation system in the direction of the Internet of Things and UAV, it can be concluded that by integrating UAV imaging, sensor based monitoring, and advanced machine learning, this method provides a promising solution for active crop disease management and sustainable yield increase in sugarcane planting (Gorijavolu et al., 2026), which is particularly useful for the Internet of Things and is suitable for irrigation and precision agriculture Internet of Things systems, because these devices it will not hinder the work of machines and farmers and will reduce the possible physical damage to nodes deployed in the field. García et al. (2020) and Spigel et al. (2020) proposed a series of key tools for innovation and regional development in the study of entrepreneurial ecosystems, but ignored the vast rural areas and did not refine the subdivision ecology of rural areas, agriculture, and small and medium-sized enterprises. They suggested that local adaptation should be carried out in the future. In terms of human resources Jones and M K’s (2004) research on rural Wales found significant differences in regional policy and urban-rural duality. However, the key limitation lies in the lack of in-depth analysis of the influencing factors of these differences. The regional reasons only involve Wales, UK, and cannot be extended to other countries and regions. Marian (2021) proposed to achieve precise and customized production methods through technologies such as satellites, sensors, and the Internet of Things. The organizational model will also shift from a fragmented agricultural production mode to a factory and enterprise based production mode. The transformation from agricultural humanization to agricultural digitization and informatization. The key point is that its impact on developing countries is relatively low, and there is no discussion on the cost of technology implementation, post maintenance, and strict technological gaps.

With the introduction of new technologies such as the Internet of Things, big data, and artificial intelligence, machine learning has been rapidly integrated into the current agricultural production work to quickly and accurately identify field management and production decisions in agricultural work. In Qingdao, agricultural and rural institutions strongly support the development of smart agriculture and strive to achieve a significant improvement in the overall efficiency in the future.

At present, advanced research progress has been made in the United States, Germany, South Korea, and Japan (Kamienski et al., 2019; Li et al., 2023; Washizu and Nakano, 2022). Although these studies provide valuable insights for advanced agricultural systems, they often ignore the limitations faced by developing regions, especially in terms of socio-economic conditions, infrastructure gaps, and human capital constraints. In 2024, China’s machine learning has achieved phased development, but still appears to be inadequate and unbalanced. At present, it mainly includes the high implementation cost of Internet of Things technology, the limited digital literacy of farmers, the lack of infrastructure support, and the poor adaptability and popularization of machine learning modes in different agricultural environments. These problems seriously restrict the large-scale adoption and effectiveness of the intelligent agriculture system. Please refer to Table 1 for specific research details.

Table 1.

Relevant reviews.

Author	Core view	Core findings	Key limitations
Kamienski C, Soininen J-P and Taumberger M et al.	Scalability is a major concern for IOT applications	It provides sufficient performance for swamp pilots, but requires specially designed configuration and redesign of some components to provide higher scalability with less computing resources	The integration of different technologies required for seamless work has not yet been fully realized.
Washizu A Nakano S	By improving the level of data visualization and introducing intelligent agricultural technology, the production efficiency of Japanese agriculture can be improved	Even if the level of data visualization is still very low, the agricultural efficiency of Japan can be improved by introducing intelligent agricultural technology. Intelligent agricultural technology automatically visualizes information and optimizes conditions without relying on farmers’ information literacy	Only the development of software did not include farmers’ educational background and level into the research.
Goap A, Sharma D and Shukla A K et al.	An intelligent farmland irrigation demand forecasting system based on open source technology is proposed	The perception of soil moisture, soil temperature, environmental conditions and other ground parameters, as well as the weather forecast data from the Internet, to predict the demand for farmland irrigation	Only remote sensing data is considered, real-time data is not considered
Li D, Nanseki T et al.	The practice of intelligent agriculture in Japan was introduced	Through real-time data collection, comprehensive and professional data mining, and specific and practical feedback, it has made contributions to improving the rice yield, production efficiency and profitability of large-scale farms	R&D system only
Li C, Sun J Wen X et al.	As an innovative water resources management strategy, virtual water trade can reallocate water resources through international trade, so as to alleviate the water pressure in water shortage areas and improve food security	It reveals the imbalance of virtual water trade mode, and shows that the current trade mode cannot effectively alleviate the pressure of water and food security in water shortage areas	Research from China’s macro level, lack of micro level research
Lachgar N, Saikouk H and Essabbar M et al.	Intelligent agriculture using embedded systems rather than focusing more on irrigation systems	How artificial intelligence and embedded systems change the way we use this valuable resource	Regional research is not done well
Morchid A, Elbasri A and Oughannou Z et al.	An innovative irrigation management method is proposed, which integrates real-time data and prediction model to improve irrigation efficiency	Improving food security and sustainable water management	Reliability unknown without combination of extreme weather (high temperature, rainstorm, drought)
García L, Parra L and Jimenez JM et al.	The possible consequences of global warming have led to consideration of water adaptation measures to ensure the supply of water for food production and consumption	Determine the parameters monitored in the irrigation system related to water quantity and quality, soil properties, and weather conditions	Insufficient research on remote areas

In order to make up for the deficiencies of these studies, this study takes Qingdao as an example and adopts the methods of literature and interviews to investigate multiple machine learning smart farms and technologies. The main contributions of this paper are as follows: first, through interviews with farm operators and retired teachers in Qingdao, the current situation of machine learning intelligent agriculture in Qingdao is analyzed; second, the factors affecting machine learning are analyzed, including the limitations of infrastructure, the shackles of talent shortage, and the bottleneck of technical barriers. The third is to put forward a series of countermeasures and suggestions through the investigation to improve the level of machine learning intelligent agriculture in Qingdao, improve the adaptability of the system, and improve the efficiency and sustainability of the system program. This countermeasure and suggestion can help other coastal areas in China to provide examples of policy designation and provide experience for agricultural practitioners in other coastal cities (Figure 2).

Figure 2.

Overall framework of machine learning-driven smart agriculture in Qingdao.

The current situation of smart agriculture in Qingdao

At present, Qingdao has built a land-sea linkage characteristic agriculture dominated by vegetables, fruits, Lao Shan tea, prawns, and sea cucumbers. Qingdao has three national agricultural high-tech zones. In terms of policy, Qingdao will invest 200 million yuan in smart agricultural machinery subsidies and digital smart agricultural base construction in 2023.

Internet of things technology and precision agriculture

The intelligent farm in Jimo District has realized the real-time perception of the Internet of Things sensor. Through the sensor, the monitoring of soil temperature and humidity and facility agricultural environment temperature and humidity is realized, automatic irrigation and intelligent temperature control are realized, and the Internet of Things technology is used to link the computer. At the same time, relying on a 5G high-speed network to achieve efficient real-time data transmission and control its low latency. By the end of 2024, Qingdao will have built a big data platform for agriculture and rural areas, covering 15 digital application directions such as origin traceability. Among them, Pingdu City has achieved differentiated irrigation through Internet of Things technology, and the effect of irrigation water saving is good. JiaoZhou City uses UAV precision agriculture to generate a farm map to achieve differentiated, regionalized, and variable precision fertilization.

High-tech agricultural demonstration project in the west coast new area

Dachang Town, West Coast New Area of Qingdao, has built about 3000 hectares of farmland intelligent platform, and built 500 hectares of unmanned farm to realize automatic irrigation of farmland and save water resources. Field management automation to ensure the stability of agriculture. At the same time, it is combined with the traditional ‘water and fertilizer integration’ to improve the utilization rate of fertilizer. Da-chang Town uses intelligent equipment to carry out precise operations to ensure the stability of wheat fields, and the degree of food security is significantly improved. Intelligent agricultural machinery is used to ensure accurate grain harvesting at harvest, so as to realize intelligent agricultural work before, during, and after production, reduce labor intensity, improve production efficiency, and increase crop yield.

Big data platform

Qingdao Municipal Bureau of Agriculture and Rural Affairs has built a big data platform for Qing-Nong Cloud Brain Smart Agriculture, which takes standardized management as a means to improve the construction of an agricultural big data center as the foundation, ensures the collection and analysis of current data, estimates the output of agricultural products in the next year, and sets a target suitable for Qingdao’s output. At the same time, the data of agricultural enterprises is collected in real time, a demonstration scenario is established, and a map of digital agriculture and rural areas is matched. As the first comprehensive data platform for agriculture and rural areas in Shandong Province, relying on the infrastructure and cloud computing capabilities of Qingdao’s ‘government cloud’, the city’s agriculture-related data can be collected, governed, managed, analyzed, and shared.

Intelligent aquaculture

As the world’s first 100,000-ton intelligent fishery breeding ship of Qingdao Guo Xin Group, Guo Xin No.1 promotes the marine fishery breeding from blue sea to deep blue by means of ship-borne warehouse breeding mode, intelligent temperature control mode, and industrialization mode. On the ship, an intelligent control center at the ship end and a shore-based intelligent cloud platform for ship-shore integration are constructed. The 2108 measuring points of the whole ship carry out centralized control and real-time monitoring of water, oxygen, light, feed, and fish in the breeding cabin. The cleaning robot of the breeding warehouse adopted by Guotin No.1 maintains the sanitary condition of the breeding warehouse, with an annual output of more than 3,000 tons of large yellow croaker, etc., which realizes the improvement of the agricultural quality of Qingdao facilities and breaks through a series of constraints of offshore breeding. Realize the innovation of subversive agricultural production mode and promote the transformation and upgrading of other industries. At the same time, ensure the safety of fishery products and build a ‘smart marine granary’.

Interview design

The interview time is from July 10, 2025, to July 17, 2025. This interview survey mainly focuses on seven farms in Qingdao. With the consent of the interviewees, the farm manager is interviewed on the spot or by telephone according to the interview outline. Unlike survey methods that focus on data, using interview results as data can enable machine learning smart agriculture to be deeply integrated with local economic characteristics and industrial development. All interview content will be organized into interview details to reveal current economic development. The scope of investigation involves Dachang Town, West Coast New Area, Pingdu City, Nancun Town, Duanbolan Town, Jin-Kou Town, Daxin Town of Jimo District, and Jihongtan Street of Chengyang. The southern part of Qingdao is a tourist area, while the northern and western parts are agricultural production areas. The survey scope selected in this article is the agricultural scale in Qingdao, where large areas of land resources are used for agricultural production. Among them, the West Coast New Area is a representative city of agriculture in Qingdao. Qingdao is also a model of temperate maritime climate and one of the developed cities in eastern China. It has similarities with other cities in terms of economy and technology. The location of each farm is shown in Figure 3, and the name and the main business are shown in Table 2.

Figure 3.

The location distribution of smart farms in Qingdao.

Table 2.

Interview farm-related information.

Name of farm	Primary business	Interviewee identity	Interview format	Interview information
Qingdao Tongfuqindeng Cooperative	Wheats	Cooperative manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation
Xiang Kun Agricultural Science and Technology in Jimo District of Qingdao	Flower planting and sales	Manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation
Qingdao Shan-Mei Ecological Agriculture Industrial Park	Laying hens breeding	Manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation
Qingdao XinXiangyuan Ecological Agriculture Co., Ltd.	Laying hens breeding	Manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation
Ding-Xin Oasis Vineyard, Jimo District, Qingdao	Grapes	Section chief of an office	On-site interview and telephone interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application of machine learning models and scenarios, infrastructure, and talent development.
Dongguang Minghui Intelligent Ecological Park of Chengyang District, District of Qingdao	Tomato	Retired teacher of the agricultural university	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application of machine learning models and scenarios, infrastructure, and talent development.
Pingdu Nancun Convergence Farm	Broiler breeding	Manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation
Huiquan Dairy Farm, Guhe Street, Laixi of Qingdao	Dairy cows breeding and dairy products sales	Manager	On-site interview	The degree of perfection of the Internet of Things system, the understanding of machine learning, the application aspects and scenarios of machine learning models, and the infrastructure situation

Interview process

With the consent of the interviewees, the author collected and sorted out the interview contents and screened them after the interview. A total of 10 interview questionnaires were issued, and 8 valid interview records were obtained. In order to ensure the privacy of the interviewees, written according to their alphabetical surname.

The degree of perfection of the internet of things system

The author: Are you satisfied with the current Internet of Things system and related facilities?

Teacher C of Dong Guang Ming Hui Farm in Chengyang District believes that ‘although we have the Internet of Things of China Unicom’s 5G big data here to transmit data to the big screen, in the extreme high temperature situation of 2024, the high temperature in the greenhouse cannot pass through the natural conditions of their own day and night temperature difference, we still lost some tomato seedlings, which is a huge loss for the farm. Therefore, machine learning for temperature control still needs to continue improving’.

Qingdao West Coast New Area, Tong Fu Qin Deng Cooperative B manager said: ‘We have invested hundreds of thousands of yuan in this set of Internet of Things technology. This cost covers only the system purchase. Inside the farmland, we have sensors set up every 10 m set up a, for monitor land temperature and humidity. Some of the sensors are imports, more expensive. The wireless network setup, Internet of Things, and AI program I need to cooperate with a computer company in Qingdao Shinan District, which maintains the program every 6 months for 10 thousand yuan, which is not cheap’. The implementation of this Internet of Things (IoT) technology required an initial capital outlay of several hundred thousand RMB, a figure that accounted strictly for system procurement. Regarding field deployment, sensors were installed at intervals of 10 m to facilitate the monitoring of soil temperature and moisture; notably, certain high-precision sensors were imported, resulting in escalated hardware costs. Furthermore, the configuration of the wireless network infrastructure, along with the development of IoT and AI algorithms, necessitated collaboration with a computer technology firm located in the Shinan District of Qingdao. This partnership entails a biannual maintenance fee of 10,000 RMB per session, representing a substantial recurring operational expenditure.

R manager of Jimo Shanmei Agricultural Ecological Park: ‘Despite the substantial upfront investment and high cost of the Internet of Things, it is inevitable that high returns cannot be obtained; it must be done. In today’s high-tech era of laying hens that rely on traditional methods of production cannot achieve scale expansion and output value increase. Notwithstanding the substantial initial capital expenditure and the inherently high costs associated with IoT infrastructure—expenses that are arguably unavoidable—I remain highly satisfied with the initiative. Such investment is an imperative prerequisite for realizing superior economic returns, given the nature of high-tech integration. In the contemporary agricultural landscape, reliance on conventional production methodologies within the layer hen industry precludes the realization of scalability and the enhancement of production value’.

J manager of Xinxiangyuan Agricultural Science and Technology in Jimo District: ‘The initial investment is indeed large. In any case, the Internet of Things system and sensors can realize temperature monitoring at night to prevent laying hens from dying due to high temperatures, which is very good. It is very labor-saving. Workers cannot be stationed on the farm 24 hours a day. Only one person is on duty, and the Internet of Things system needs to be arranged at night to realize the perfect cooperation between people and the system, so as to ensure the operation of the farm’.

Jimo Dingxin Oasis Vineyard D director: ‘A company has strong financial resources. It doesn’t matter that the Internet of Things and sensors are expensive, and the cost of program maintenance is high. The company can afford it. The Internet of Things system can also reduce labor costs. The farm only needs to arrange some personnel to do security work, and the rest is left to the central control system, very satisfied with the system’.

S manager of Nancun Hui He Farm in Pingdu City: ‘The interactive paradigm within the IoT architecture exhibits significant flexibility; in instances of program anomalies, immediate technical assistance is available from developers located in the Shinan District of Qingdao. Should the integration between the IoT framework and the physical operational environment prove suboptimal, a strategic replacement of the system infrastructure remains a viable contingency. Currently, maintenance protocols are being executed with a high degree of efficacy; however, a prevailing technical challenge pertains to the occasional insufficiency in sensor sensitivity’.

K manager of Xiangkun Agricultural Science and Technology in Jimo District: ‘The manifestation of systemic dysfunction within the IoT infrastructure presents a severe operational impediment, characterized notably by a lack of accessible technical support from agricultural technologists. Given that my immediate professional network lacks domain-specific expertise in this niche, I am necessitated to seek external assistance to address such technical contingencies’.

The F manager of Huiquan Dairy Farm in Guhe Street, Laixi City, said: ‘A significant technical limitation is the system’s lack of cross-browser compatibility, which imposes considerable usability constraints. Furthermore, the financial burden associated with essential training protocols for the IoT infrastructure is occasionally prohibitive. Nevertheless, at the current operational stage, the interaction mechanism between the user and the system retains a satisfactory degree of flexibility’.

Through this interview, it can be found that teacher C believes that there is more room for improvement in the system, which shows that the innovation degree of the Internet of Things system is not good. Other farm managers generally believe that the Internet of Things system is good, but the maintenance cost is expensive, and the technical barriers of network cloud service still exist.

Understanding of machine learning

The author: ‘Do you know the algorithm used in the program when you are finished? Do you want to receive training in the future?’.

Teacher C in the Dong Guang Ming Hui Farm of Chengyang District: ‘Only heard of support vector machine SVM and artificial neural network ANN’.

B manager: It is acknowledged that the Random Forest (RF) algorithm is employed. Consultations with software developers were conducted during the system debugging process, and the artificial intelligence (AI) programs compiled by these developers are available. Given the advanced age, the comprehension capacity may fail to match that of younger generations if relevant training is required in the future.

J manager: Proficiency in information technology is not attained, yet it suffices that the individual’s son and nephew possess relevant expertise, as they will serve as the core driving force for the future development of the farm. Given the advanced age, there is no willingness to learn such technologies even if mastery proves unachievable. Last year, the son was admitted to a university, majoring in animal husbandry. He exhibits high proficiency in operating mobile devices, far exceeding the capabilities of the elderly. Despite the son’s instructional efforts, only basic functions—making phone calls, sending text messages, and chatting via WeChat—are operable; mastery of other digital operations remains unattainable, with no corresponding desire to pursue such learning.

D Director: ‘The company has special operators, and training can also be done here.'

S Manager: ‘A basic understanding of machine learning has been attained; however, no specific attention has been devoted to the relevant algorithms. In terms of professional training, external programmers are retained to deliver relevant instructional sessions’.

K manager: ‘This pertains to the domain of programmers; they have not delved deeply into research on it. Although they frequently participate in training sessions, their time is constrained’.

F manager: ‘Only one training session hosted by the Bureau of Agriculture and Rural Affairs was attended. The relevant algorithms and models were not comprehended, and the corresponding content proved challenging to understand’.

In summary, at present, people lack understanding and cognition of machine learning, and training is not in place. At present, the age of Qingdao farm managers is too old, and many people do not understand information technology. Even if they understand, they cannot rely on their own cognition to complete the actual use of machine learning.

Machine learning model application aspects and scenarios

The author: ‘Your farm applies machine learning technology to those aspects and scenarios?’.

Teacher C: ‘The technologies for crop pest identification and early warning, as well as soil fertility detection, have been implemented on my farm. Such applications are primarily adopted for farms located in plain regions’.

Manager B: ‘Key operational domains include pest identification, soil fertility assessment, precision fertilization, intelligent irrigation, data analytics, and agrometeorological forecasting. Given the geographical constraints of Qingdao, facility deployment is restricted to plain areas due to superior drainage capacity and lower hydrological risk compared to low-lying regions. The elevated terrain mitigates the threat of monsoon-induced flooding, thereby reducing potential asset depreciation. Furthermore, the integration of IoT sensors and machine learning algorithms is optimized for stable topographies, as lowland environments introduce significant technical barriers to infrastructure setup and software development’.

J Manager: ‘An integrated agricultural production data and decision support platform has been developed, incorporating dipteran pest (flies and mosquitoes) identification functionalities. The platform is deployed exclusively in plain regions, deliberately excluding mountainous terrain to ensure operational stability’.

R manager: ‘A mosquito surveillance platform has been deployed, integrated within a broader framework encompassing agricultural product traceability systems. The operational infrastructure is situated in the peri-urban plains, a strategic location chosen to optimize spatial efficiency and logistical accessibility’.

Chief D: ‘The operational scope encompasses soil fertility diagnostics, smart irrigation systems, and agricultural product traceability mechanisms. These interventions are deployed across a contiguous 1300-hectare expanse of flat arable land’.

Manager S: ‘Animal husbandry only needs agricultural products traceability, data analysis, and decision-making. The farm is also on the flat land near the village, because I am located in Pingdu City, with many hilly areas, and I have not heard of anyone breeding there’.

K manager: ‘Flower products need pest warning, soil nutrition monitoring, and precision fertilization. Intelligent irrigation is essential. It is close to the urban area of Jimo and belongs to suburban smart agriculture. It is better to keep the greenhouse away from the mountains and sloping fields, because the flowers need quantitative irrigation. It only takes time to set it up, and no other operation is required’.

F Manager: ‘Intelligent robots and traceability of agricultural products are required for dairy farming, and the staff is responsible for dosing’.

In summary, the intelligent farm of machine learning in Qingdao only exists in the plain area. At present, there are crop pest identification and early warning, soil fertility detection, data analysis, and decision-making platforms in Qingdao farms. Other projects exist in various farms to varying degrees. There are almost no farms in the mountainous areas of Qingdao, and most of the farms are in the plain area, the edge of the township, or near the main traffic arteries.

Infrastructure situation

The author: ‘Do you think that rural infrastructure in Qingdao can support Smart Agriculture in the form of machine learning?’.

Teacher C: ‘Further advancements remain contingent upon infrastructural readiness; without robust foundational support, subsequent initiatives risk remaining theoretically infeasible’.

B manager: ‘Notable progress in infrastructure enhancement is observed within the West Coast New Area, positioning it as a favorable locale for smart agriculture implementation’.

J Manager: ‘Ensuring power grid reliability is imperative; even transient power interruptions can trigger cascading operational failures, resulting in substantial economic depreciation and data integrity compromise’.

R manager: ‘Who knows how to arrange improvement measures in the future? Look at the future development’.

Chief D: ‘This may not be able to support some places. According to the geographical terms, the suburban area of the farm infrastructure is very good, but in the north of Jimo District, it is not enough’.

K Manager: ‘The data of Qingdao Agricultural and Rural Bureau is not connected to my side. As for infrastructure, my side is close to Jimo City, and it is OK’.

F manager: ‘It is very remote here, and the infrastructure is not very good. Sometimes the network will be paralyzed when there is a power failure, which is very troublesome. It can’t be solved at this time’.

In Qingdao rural infrastructure, the suburban areas of the farm problem is small, in the depths of the countryside, or in other areas of the farm. The infrastructure is not well constructed, especially in the northern part of Qingdao City, namely, Pingdu City and Laixi City. Due to the constraints of geographical conditions, the degree of infrastructure construction is slightly insufficient compared with other regions of Qingdao City. Rural and agricultural areas in the developed areas of southern Qingdao have better power supply.

Talent aspects

The author: ‘Do you farm workers have a bachelor’s degree?’

Teacher C: ‘Yes, our farm has a master’s degree graduate from the School of Resources and Environment of Qingdao Agricultural University, and there is also an undergraduate graduate working here’.

D Director: ‘Our company also has some information technology personnel is a bachelor’s degree, but the farm security personnel do not’.

Other farms have no undergraduate degree, which is not within the scope of this interview.

Through field research, it is found that machine learning is in urgent need of a large number of talents. Talents are the source of innovation for development. The shortage of talent leads to the development of Internet of Things programs and barriers to innovation and development.

The dilemma faced

Barriers to technology adaptation

Technological adaptability is not only an algorithm limitation, but also a geographical challenge formed by the terrain division of the coastal plain in the region. It is learned from the survey that at present, smart agriculture only stays on the big screen, and only realizes the control of facilities through the real-time transmission of data to the central control center of the farm. First, the machine learning algorithm needs a large amount of data for training, debugging a large number of models, and its application in temperature and humidity treatment. Second, due to different seasonal temperatures, the ability to accurately control humidity in the peak rainfall season is insufficient. When waterlogging occurs in the facility’s agricultural sites, the facility’s agriculture is damaged, and the recovery time is long, resulting in a disruption of the normal growth cycle of agricultural products. The existing algorithms are only for family farms in the plains of Qingdao, mainly for high-standard farmland construction and environmental development. Due to the topography and geomorphology of Qingdao, the northern and southeastern regions are mountainous areas, and the southern and southwestern regions are coastal areas. This survey found that the existing machine learning models are mainly focused on flower crops and fruit products, and other agricultural products have corresponding models. Some smart farms have unstable adaptability in certain environments. For example, random forest RF and CNN models perform well in a single scene, but their effects have changed significantly in special scenes.

The price of IOT equipment is on the high side (Simo et al, 2022) is these equipment worth investing in for farmers? At present, the IOT sensor has not formed a labor-intensive industry, and the network technology cannot be broken through in a short time, and the more complex IOT programs are deployed with AI models when needed (Muhammed et al., 2024). Due to the strict confidentiality of foreign AI technologies and the blockade of China, the training level and sensitivity of China’s AI models have great room for improvement. In recent years, due to the low degree of innovation of Qingdao’s information technology companies, they can’t design, change, and train AI models with high flexibility in a short time, so the network technical barriers need to be broken through within a certain time. The wider communication range and high data transmission function are most likely to dominate in the future, which is also the goal of smart agriculture in the future. Since the deployment of these technologies in rural areas is a major challenge, it is an opportunity of low installation costs (small units) and the inability to obtain high-speed times. In the cloud service of Internet of Things, the system faces a lot of data processing work (such as soil sensor data, temperature, and humidity control). How to use network technology to quickly process and optimize data processing is an urgent problem to be solved. In terms of technical complexity, the network technology of the Internet of Things usually needs a large number of information technology disciplines and a series of interdisciplinary and interdisciplinary support. How to reduce the network technical barriers in the future, so that farmers and universities, research institutions, and information technology enterprises ‘tripartite linkage’.

Shackles of power infrastructure

Qingdao’s agricultural and rural infrastructure is insufficient, the power supply is small, and the power supply is unstable, which seriously restricts the power quality of smart farms’ Internet of Things and sensor facilities. The deployment and operation of smart agricultural machinery to a large extent need to have a stable power supply and constant voltage, covering a small range of farms, resulting in backward infrastructure, which affects the popularization of smart agriculture in rural areas. Popularization should not lead to the fact that the actual agricultural production cannot reach the expected goal.

Shortage of human capital

The shortage of human capital reflects the social embeddedness of technological knowledge. In Qingdao, farmers’ education level is low, especially the lack of growth of ‘new farmers’ and the shortage of professional and technical personnel (Wei, 2024), and farmers cannot control the heavy system. Secondly, silver age farmers lack the operation of the smartphone system, and cannot give full play to the advantages of smart agriculture. At the same time, the inability to understand and actively receive training on machine learning has led to the slow development of smart agriculture. Machine learning requires a lot of professional knowledge. In the information and networking era, some silver age farmers do not trust the new production technology due to the deep-rooted small-scale peasant thought and family education concept in rural areas, as well as the traditional concept of agricultural production for a long time. They did not receive higher education, lacked basic knowledge of information technology, and were not empowered with the changes of the times in the later period. They believed that machine learning was the so-called learning and use of agricultural machinery. When their children learned information technology, they did not consult their children in time, and missed the opportunity to learn information technology knowledge.

Lack of high-level talents, lack of application experience in training, and low willingness to work in agriculture. First, the mechanism of talent introduction and retention is not mature. Rural areas generally lack the competitiveness to match the high-level talent market. The process and standard of talent introduction are not clear and scientific enough, which often leads to a mismatch between the introduced talents and the local demand. The follow-up service and support system for high-level talents in rural areas is not perfect, which is difficult to provide them with continuous career growth and life security, thus increasing the brain drain rate. Second, a large proportion of publicly funded agricultural students have no rural life experience, and a master of Agronomy and agriculture has no experience in planting crops. Even if Qingdao introduces highly educated and high-quality intelligent agricultural talents, the traditional concepts and cultural atmosphere in rural areas may restrict the innovative spirit and career development of high-level talent, resulting in the inability of high-level talents to integrate into the rural environment and carry out research and improvement work. Third, the existing training system is often not designed according to the actual needs and characteristics of the countryside (Xu and Song 2024). Fourth, agricultural colleges and universities in Qingdao have not started the training program of machine learning, including agronomy, plant protection, and the interdisciplinary integrated development of agricultural and forestry economic management. Colleges and universities have educational barriers in cultivating talents. Students of agricultural and forestry economic management who have not learned agricultural knowledge do not understand the normal growth cycle of crops. The subjective willingness of agricultural colleges and universities to return to rural areas for farming or entrepreneurship is still weak. This is not only due to historical reasons, traditional ideas, and intergenerational thinking, but also due to the comprehensive factors such as the special rural environment, social status, economic treatment, and development potential. The talent resource group generally lacks identity for farmers and is unwilling to engage in or discriminate against agricultural labor. ‘Learn what not to learn from agriculture, learn from agriculture and do not work in agriculture, do not love agriculture, love agriculture and do not specialize in agriculture’ is a true reflection of widespread existence (Xu and Zhou 2024).

Shortage of a coordination mechanism

More serious cases will lead to high input, high management, and no income of agricultural products. At present, with the development of science and technology, cooperation with other institutions is another innovative choice to promote scientific and technological progress. At present, there is no substantial progress in the mechanism of industry-university research cooperation between universities, scientific research institutions and agricultural enterprises, or even farmers in Qingdao. The actual adoption rate of scientific research achievements of university scientific research institutions is low. In order to maximize benefits, agricultural enterprises use the minimum R&D funds to obtain the maximum scientific research output. The effect is far less than that of universities and agricultural scientific research departments in Qingdao. It is impossible to solve the problems of machine learning in actual smart agriculture in a ‘grounded’ way. At the same time, the models introduced from abroad, even if they are suitable for the local geographical characteristics, may not align with the agricultural conditions of Qingdao, or require improvements after introduction. Furthermore, it is challenging to integrate multiple models in a synergistic manner, and their compatibility with the smart agricultural monitoring system is insufficient.

Conclusions and perfect measures

Conclusion

The future optimization level of machine learning smart agriculture in Qingdao is very high. This study found and proposed that in the context of machine learning, the problems faced by machine learning in Qingdao’s smart agriculture were identified through research on different identities of ordinary workers, as well as their understanding in real work. The main conclusions of this article are as follows:

There are many obstacles to technological adaptability, and the cost of IoT facilities is expensive. The system can be further improved, and the level of algorithm innovation can continue to be enhanced. It is necessary to train better models to adapt to the diversity of Qingdao’s climate and enhance their universality in rural areas of Qingdao.

The level of infrastructure improvement in Qingdao is not satisfactory, mainly due to restrictions on its development in terms of power supply. The popularity of new energy supply facilities has certain development space compared to other large cities, highlighting the need for stable power infrastructure support for machine learning smart agriculture.

The prosperity of talent leads to the prosperity of rural areas. The growth space for agricultural technology talents and new farmers is strictly limited, and communication is limited. Silver-aged farmers have poor experience in using smartphones, and technological progress is slow. The training plan for agricultural universities in Qingdao has not been closely integrated with smart agriculture production. There is less communication between agricultural subject teachers and farmers, and the atmosphere for discussing smart agriculture among rural villagers is relatively low.

The lack of collaborative mechanisms sometimes leads to difficulties in research and development, making it impossible to solve practical problems of machine learning in smart agriculture in Qingdao on the ground. Excessive pursuit of maximizing benefits cannot achieve ideal results, and model integration is also necessary.

Optimizing measures

Improve the construction of power infrastructure to ensure the stable operation of machine learning equipment

Improving power infrastructure is crucial for supporting the stable operation of smart agriculture and further promoting local economic development. The quality of the rural power supply in Qingdao is not high. Based on this problem, first, we can combine urban and rural areas to jointly match high-skilled electricians, and let employees with sufficient experience in power maintenance participate in the maintenance of rural power supply equipment. At the same time, the rural power supply and distribution network will be upgraded and accurately planned. The villages with high load and high power consumption required for machine learning smart agriculture will be regularly inspected and overhauled. The hidden dangers of power consumption will be found in time. Transformers and distribution rooms with aging lines and a lack of maintenance will be upgraded to meet the power requirements of machine learning smart agriculture. The second is to conduct real-time monitoring of power facilities and timely understand the status of power supply equipment. When equipment problems occur, farmers, employees of power companies, and village cadres can use a ‘three-way linkage’ to eliminate faults in time, reducing the time from fault detection to fault elimination. The employees of power companies should not only monitor the power consumption data in real time, but also make early warning notifications, and feed back and notify the data that may affect the stable operation of equipment in machine learning intelligent agriculture. The third is to establish infrastructure construction, such as photovoltaic power generation and wind power generation, reduce farmers’ investment in electricity through clean energy, and increase farmers’ income. Fourth, establish standby energy storage facilities for electric energy, that is, equip with diesel generators or other traditional standby power sources, form a multiple power supply guarantee system, and quickly switch to ensure a continuous power supply in case of unexpected power failure.

Develop machine learning intelligent agriculture according to local conditions to ensure the integrity of technology

Developing localized machine learning models based on local conditions can better integrate smart agriculture with local resource endowments, promoting local based economic growth. Improve the localization model of machine learning in Qingdao, focus on model optimization technology (Li et al, 2023), cloud computing and edge computing, and develop edge computing technology to reduce the dependence on remote servers and improve the response speed of agricultural on-site decision-making. At the same time, combined with the ‘Qingnong yunnao’ big data platform of Qingdao Municipal Bureau of Agriculture and rural development, do a good job in relevant statistical functions, and then optimize the local model, so as to integrate the model with practice for innovation, so as to achieve ‘agricultural products + model’ double insurance. The second is to ensure that agricultural products are adaptive to the models of Jiaozhou cabbage, Majiagou celery, and other characteristic agricultural products under the machine learning intelligent agricultural production mode. On this basis, different vegetables and fruits are classified, and the data sets are trained in combination with local characteristics. The models are strictly screened, the results are tracked regularly, and the large models are adjusted regularly to prevent crop yield reduction due to model adaptation problems. The third is to represent the will and thoughts of farmers and make the model have ‘farmers’ consciousness’, that is, to design the model scientifically and reasonably according to farmers’ planting experience, and improve the interpretability.

In view of the lack of result orientation in some regions, the effect of machine learning in smart agriculture is poor after the implementation. First, we can take measures according to local conditions, choose the traditional agricultural production mode, moderate wisdom, moderate scale, moderate machine learning, and ensure that smart agriculture can play a more efficient, higher output, and faster speed under the machine learning mode. In areas that fully understand the economic strength of farmers in Qingdao and fully assess the development of machine learning intelligent agriculture, facilities and sites operated voluntarily by farmers can be retained in areas with poor implementation effect. Second, farms with better results can be appropriately expanded. According to the actual needs of different farmers and agricultural enterprises, a personalized intelligent decision-making system can be customized to ensure that the technical scheme truly conforms to the actual local agricultural production. The successful experience should be popularized, and farmers with experience and a college degree or above should be encouraged to develop machine learning intelligent agriculture and carry out intensive management under their own economic conditions. Third, silver age farmers who are unwilling to engage in machine learning smart agriculture are encouraged to voluntarily transfer land to provide land resources support for the implementation of smart agriculture.

Improve the talent mechanism

Improving the talent mechanism will provide human capital support for smart agriculture and enhance the long-term vitality of local economic development. First, the Qingdao Municipal Bureau of Agriculture and rural areas regularly organizes and invites agricultural experts to give simple guidance to farmers who voluntarily operate smart agriculture. Regularly return visits to farmers under the guidance of experts, and select excellent training experts for training so that farmers can understand the simple operation to deal with the possible risks on the farm. On the other hand, older farmers who are willing to engage in machine learning can ask their children about computers and program applications, smartphones, networks, and other related knowledge to make up for the lack of knowledge. They have the basic knowledge of a network for operating machine learning smart agriculture, as well as the basic knowledge of operating smartphones and machine learning principles. After the farmers have the basic knowledge, they can eliminate the knowledge barriers of communication with agricultural scientific and technological personnel. At the same time, study groups or interest groups are set up locally to encourage older farmers to help each other and make common progress through regular exchange meetings and experience-sharing meetings. The fourth is to make short video tutorials and graphical materials, and use animation and real-time cases to intuitively display the technical principles and reduce the threshold of understanding. Encourage students to apply their knowledge to actual agricultural production, and reward or commend farmers who have achieved remarkable results, forming a positive cycle of ‘learning for application’.

Second, master and doctoral supervisors in Colleges and universities can understand the current situation of machine learning smart agriculture according to their own social relations, and promote the integration of production and education according to the existing social conditions and in combination with the training program of colleges and universities so that master and doctoral students can fully understand the practical problems of machine learning smart agriculture after graduation, and use their theoretical knowledge, academic ability and understanding of society to carry out scientific research work. Colleges and universities set up special doctoral scholarships and scientific research projects to encourage outstanding doctoral students to carry out research in rural areas and solve practical problems in rural development (Xu and Song 2024).

The third is to use the ‘Science and Technology Institute’ platform to establish a linkage mechanism with local farmers, and communicate every 6 months to 1 year. Computer science graduates and postgraduates can use the Science and Technology Institute to communicate with farmers, and learn programming in combination with the reality of machine learning and intelligent agriculture, so as to lay the foundation for the establishment of a model suitable for Qingdao.

Complement synergy mechanism

Completing the collaborative mechanism will accelerate the transformation of smart agriculture achievements and effectively drive local economic development. In view of the current high price of Internet of Things facilities, it is impossible for small and medium-sized farmers to complete the investment. In view of this, first, we can provide national research subsidies for smart agricultural IOT sensors, and give it to enterprises, universities in Qingdao, and scientific research institutions funding for IOT program research and development to ensure the rapid release of research and development funding. Second, in the research and development of sensors, we will focus on the new mechanism of agricultural sensing based on optical, electrochemical, electromagnetic, ultrasonic, image and other methods, research and develop core components such as sensitive devices, photoelectric conversion, weak signal processing, and develop a batch of high-precision agricultural sensors to break the monopoly of foreign technology products (Zhao 2021). At the same time, we can establish scientific research cooperation with local colleges and universities, develop small linkage sensors to adapt to the harsh environment, monitor the internal operation of smart agricultural facilities in real time, and allow private information technology enterprises to participate in the research and development of the Internet of Things and sensors, forming a positive and competitive scientific research environment. The third is to carry out commercial bidding for the high manufacturing cost of sensitive materials (Shi et al., 2023). The R&D unit cooperates with 3D printing companies to produce sensitive materials by direct printing to improve the manufacturing efficiency of sensors. It can also use the material AI material platform to screen manufacturing materials, shorten the R&D time, achieve the purpose of reducing manufacturing costs, and reduce farmers’ investment (see Figure 4).

Figure 4.

Challenges faced and measures for improvement.

Discussions

These findings suggest that machine learning optimization in Qingdao’s smart agriculture is not a technical upgrade but a place-based developmental process. The plain-area concentration of smart farms shows technological efficacy is geographically contingent; older farmers’ reliance on kinship networks reveals it is socially embedded. Without governance aligning design with local needs, machine learning cannot generate sustained rural income growth, repositioning it as a driver of local economic development rather than an autonomous fix. This study combines smart agriculture with the current hot spots of machine learning to predict possible problems in the future and solutions, through the experience of developed countries in smart agriculture such as the United States, Germany, and South Korea (Hector et al, 2024; Michels et al, 2020; Schukat and Heise 2021; Sundaram et al., 2020), combined with the advantages of Qingdao (Li 2022; Wan et al, 2022), interviews were conducted with farm managers and retired agricultural university teachers to understand their real ideas, and refined research was conducted on them. Firstly, it is suggested that the future research direction should be adjusted to AI big data agriculture, combined with machine learning. At present, AI technology is constantly maturing. In the next 10 years, studying AI + models of machine learning in Smart Agriculture can promote the high-quality development of agriculture. The second is whether the security of data affects the agricultural production of machine learning. In the future, there is no need to share the original data, and whether the privacy of the data is protected.

The optimization of machine learning in intelligent agriculture is not just a process of technological upgrading, but a form of place-based developmental intervention that restructures local agricultural economic capacities, based on the combination of machine learning and local economy. Adjustments are made according to the annual data of the local area, and an annual adjustment plan for machine learning is formulated according to the actual situation. In certain circumstances, the plan is adjusted or re-audited according to the actual situation of the monthly data. The fourth is to focus on long-term risk challenges and how to respond to high-risk investments in machine learning smart agriculture to ensure the healthy and sound development of machine learning smart agriculture. Although the current agricultural income is at a loss, machine learning in Smart Agriculture is a trend. In order to meet the future urbanization process and create a precedent for urban agriculture, future research and development will gradually follow-up and promote the development of agriculture in a more efficient and intelligent direction.

Footnotes

Acknowledgments

The completion of this study cannot be separated from the support and assistance of all parties. We would like to express our sincerest gratitude. Firstly, I would like to express my sincere gratitude Journal editors and reviewers valuable suggestions for the revision and improvement of the paper.

ORCID iD

Yilin Wang

Consent to participate

I hereby declare that all the farm staff and retired university teachers listed in this article have verbally agreed and voluntarily participated in my interviews. They have the right to interrupt their participation at any time during the interviews. The author himself completed the interview work while respecting the wishes of others Consent for publication.

Consent for publication

Author [Yilin Wang] agree that the journal [Local Economy] will accept the accepted version of the paper, which is original and has no copyright disputes. I also agree that the publishing unit may make necessary editing and modifications to the work. I hereby agree!.

Author contributions

Yilin Wang: Research design, interview implementation, topic selection and writing of the paper, and paper revision and review.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: The author declares that there are no conflicts of interest (including but not limited to commercial interests, personal interests, and academic competition) that may affect the objectivity of the research during the process of this study and publication of the paper. There is no undisclosed association between the author and the interviewed farm, and the research is authentic and reliable, without any third-party intervention.

Statement

This paper [Research on the optimization of Smart Agriculture based on machine learning: Taking Qingdao as an example] was independently selected, written, and revised by Yilin Wang. Throughout the entire process, there was no involvement of third-party personnel, enterprises, or government organizations. If other authors are found to be impersonating, they can inform the relevant organizations and the author’s email (qdwyl1996@163.com) or the author themselves. The author will handle this issue in accordance with the law and regulations.

Ethics committee review statement

This study is based on China’s national conditions, without ethical review institutions. The research process strictly follows the relevant ethics, laws, and ethical norms of China’s national conditions. All research involving humans/animals has obtained corresponding informed consent/ethical permission, ensuring the legitimate rights and interests of the research subjects.

References

García

Parra

Jimenez

, et al. (2020) IoT-based smart irrigation systems: an overview on the recent trends on sensors and IoT systems for irrigation in precision agriculture. Sensors 20(4): 1042. https://doi.org/10.3390/s20041042

Goap

Sharma

Shukla

, et al. (2018) An IoT-based smart irrigation management system using machine learning and open source technologies. Computers and Electronics in Agriculture 155: 41–49. https://doi.org/10.1016/j.compag.2018.09.040

Gorijavolu

Sharma

Rao

(2026) Smart irrigation system and early plant disease detection using IoT and novel non-linear growing self-organizing map based artificial neural network. Scientific Reports 16(1): 9488. https://doi.org/10.1038/s41598-025-33323-w

Hector

Menendez Jameson

, et al. (2024) Applying precision rangeland grazing management systems in western South Dakota. Journal of Animal Science 102(3): 192–193. https://doi.org/10.1093/jas/skae234.226

Jones M K (2004) Rural labour markets: the welsh example. Local Economy 19(3): 226–248. https://doi.org/10.1080/0269094042000233619

Kamienski

Soininen

J-P

Taumberger

, et al. (2019) Smart water management platform: iot-based precision irrigation for agriculture. Sensors 19(2): 276. https://doi.org/10.3390/s19020276

Lachgar

Saikouk

Essabbar

, et al. (2025) PRISMA-guided systematic review on the adoption of artificial intelligence and embedded systems for smart irrigation. Pure and Applied Geophysics 182(6): 2533–2582. https://doi.org/10.1007/s00024-025-03707-0

(2022) Research on the Development Evaluation of Modern Agriculture Demonstration Area on the West Coast of Qingdao. Qingdao: Qingdao Agricultural University. Master’s Dissertation.

Liu

Cao

, et al. (2026) Artificial intelligence of things for smart agriculture with optimal embedded edge intelligence. Computers and Electrical Engineering 133: 109857. https://doi.org/10.1016/j.compeleceng.2026.109857

10.

Sun

Wen

, et al. (2025) Evaluating agricultural resource pressure and food security in China and “Belt and Road” partner countries with virtual water trade. Sustainability 17(4): 1599. https://doi.org/10.3390/su17041599

11.

Xiang

, et al. (2023a) Foundation models in smart agriculture: basics, opportunities, and challenges. Computers and Electronics in Agriculture 222: 109. https://doi.org/10.1016/j.compag.2024.109032

12.

Nanseki

Chomei

, et al. (2023b) A review of smart agriculture and production practices in Japanese large-scale rice farming. Journal of the Science of Food and Agriculture 103(4): 1609–1620. https://doi.org/10.1002/jsfa.12204

13.

Mao

Wang

Cheng

(2018) Application of artificial intelligence technology in modern agricultural production. Chinese Journal of Agricultural Network Information(5): 14–18.

14.

Marian

(2021) A vision of the farming sector’s future: what is in there for farmers in the time of the second machine age? Local Economy 35(8): 717–722. https://doi.org/10.1177/02690942211010151

15.

Michels

Fecke

Feil

J-H

, et al. (2020) Smartphone adoption and use in agriculture: empirical evidence from Germany. Precision Agriculture 21(2): 403–425. Available at: https://doi.org/10.1007/s11119-019-09675-5

16.

Morchid

Elbasri

Oughannou

, et al. (2025) An innovative smart irrigation using embedded and regression-based machine learning technologies for improving water security and sustainability. Access IEEE 13: 100731–100751. https://doi.org/10.1109/ACCESS.2025.3577911

17.

Muhammed

Ahvar

, et al. (2024) Artificial intelligence of things (AIoT) for smart agriculture: a review of architectures, technologies and solutions. Journal of Network and Computer Applications 228: 103806. Available at: https://doi.org/10.1016/j.jnca.2024.103905

18.

Nouri

Homaee

Pereira

, et al. (2023) Water management dilemma in the agricultural sector of Iran: a review focusing on water governance. Agricultural Water Management 288: 108480. https://doi.org/10.1016/j.agwat.2023.108480

19.

Schukat

Heise

(2021) Smart products in livestock farming—An empirical study on the attitudes of German farmers. Animals 11(4): 1055. https://doi.org/10.3390/ani11041055

20.

Shi

Shen

, et al. (2023) Research progress on chipless RF cross-domain sensing for smart agriculture. Chinese Journal of ACTA Agricultural Engineering 39(07): 10–23.

21.

Simo

Dzitac

Badea

, et al. (2022) Smart agriculture: iot-based greenhouse monitoring system. International Journal of Computers, Communications & Control 17(6): 1–13. Available at: https://doi.org/10.15837/ijccc.2022.6.5039

22.

Spigel

Mason

(2020) A manifesto for researching entrepreneurial ecosystems. Local Economy 35(5): 483–492. https://doi.org/10.1177/0269094220959052

23.

Sundaram

JPS

Zhao

(2020) A survey on LoRa networking: research problems, current solutions, and open issues. IEEE Communications Surveys & Tutorials 22(1): 371–388. https://doi.org/10.1109/COMST.2019.2949598

24.

Wan

Zhang

Liu

, et al. (2022) Exploration on the development of smart agriculture in the Chengyang district of qingdao. Chinese journal of Agricultural engineering technology 42(15): 20–22. https://doi.org/10.16815/j.cnki.11-5436/s.2022.15.005

25.

Washizu

Nakano

(2022) Exploring the characteristics of smart agricultural development in Japan: analysis using a smart agricultural kaizen level technology map. Computers and Electronics in Agriculture 198: 107001. https://doi.org/10.1016/j.compag.2022.107001

26.

Wei

(2024) The Road of China’s Agricultural Modernization Under the Background of Big Country and Small Peasant Smelter - Also on the Internal Logic and Realization Path of Cultivating New Productivity in Agriculture. Chinese Journal of People’s Forum · Academic Frontier (10), 47–54. Available at: https://doi.org/10.16619/j.cnki.rmltxsqy.2024.10.005

27.

Song

(2024) Strategic research on accelerating the empowerment of high-level talents from the perspective of rural revitalization. Chinese Journal of Agricultural economy,(08):129-130.

28.

Zhou

(2024) The path optimization of agricultural talent resource development from the perspective of rural revitalization. Chinese journal of Agricultural economy(05): 127–128.

29.

Zhao

(2021) The development status and future prospects of smart agriculture. Chinese Journal of South China Agricultural University 42(06): 1–7.