Can Digital Intelligence and Cyber-Physical-Social Systems Achieve Global Food Security and Sustainability?

Yanfen Wang , Mengzhen Kang ,,, Yali Liu , Juanjuan Li , Kai Xue , Xiujuan Wang ,Jianqing Du , Yonglin Tian ,,, Qinghua Ni , and Fei-Yue Wang ,,

Abstract—Plants sequester carbon through photosynthesis and provide primary productivity for the ecosystem.However, they also simultaneously consume water through transpiration, leading to a carbon-water balance relationship.Agricultural production can be regarded as a form of carbon sequestration behavior.From the perspective of the natural-social-economic complex ecosystem, excessive water usage in food production will aggravate regional water pressure for both domestic and industrial purposes.Hence, achieving a harmonious equilibrium between carbon and water resources during the food production process is a key scientific challenge for ensuring food security and sustainability.

I.INTRODUCTION

AGRICULTURAL production not only provides indispensable food for human survival but also maintains the stability of ecosystems [1], [2].For example, ecological agriculture is conducive to rural economy, protecting and improving the environment, so as to obtain higher economic, ecological, and social benefits [3], [4].Key elements, main processes,and critical mechanisms for regulating and controlling ecological systems are intrinsically entangled.In order to balance the needs of various stakeholders with a holistic view, decision support is desirable for farmers, researchers, governors, service providers, companies, etc.

In ecosystems, carbon and water are the fundamental sources of life and the core elements of ecological structures and functions.Essentially, plants fix carbon through photosynthesis and provide primary productivity for ecosystems.Meanwhile, through transpiration, they also consume water resources that could be used for human production and life,thus maintaining the natural balance between carbon and water.Agricultural production inevitably consumes a lot of water resources, thus improving water use efficiency in agriculture is essential to ensure food and water security in the world [5].In many parts of the world, more than 70% of fresh water is used for agriculture, which increasingly competes with domestic and industrial water.The agricultural water use efficiency varies greatly in different regions affected by climatic conditions.For example, the agricultural water use efficiency in the upper basin of the Yellow River in China is much lower than that in the middle and lower basins.

For governors, a key decision is to achieve optimal water allocation among agriculture and industry [6], [7].Significant challenges remain in developing national and international regulations to support more optimal production in both industrialized and developing countries.Therefore, it is necessary not only to adjust the industrial structure within a certain region according to the water use efficiency but also to follow the principle of coordinated development between regions from the holistic perspective [8], [9].Reasonable and feasible water resource trading standards are being formulated under the premise of ensuring the basic needs of food security,domestic and ecological water, and taking ecological compensation into account, to guide low and high water efficiency areas to carry water trading, to improve the overall water resource utilization efficiency, and achieve a win-win situation of ecological protection and economic development [10].

For growers, a critical decision is to select the appropriate crops that optimize both the planting structure and water use efficiency based on local climatic and hydrological conditions[11], [12].In addition, the rational use of water-saving technologies is adopted to improve water use efficiency, including water storage and soil conservation technology, dropper technology, pipeline water transport method, and field cover technology [13].Fertilization and field management are integrated to optimize crop growth to its fullest potential [14].

Proper decision-making is crucial for ensuring food security.According to the 1996 World Food Summit, food security has been defined as a situation when all people, at all times, have physical and economic access to sufficient safe and nutritious food that meets their dietary needs and food preferences for an active and healthy life.Given this definition, it is not superising that culture impacts food availability,access, utilization, and stability, which is in turn influenced by gender, family, and decision-making power [15].Poor decisions on resource use may harm the resource utilization efficiency, and ultimately food security.As defined, food security means not only food sufficiency (quantity), but also food safety (quality).Long-term food security calls for wisely-managed, resistant, reliability and robust agricultural systems.However, climate change, population growth, unbalanced supply and demand, world-wide pandemic, and even regional conflicts, altogether bring significant uncertainties to food security.Ensuring food security requires that food production and distribution function properly after disruptions.Factors that improve the ability of the global food system to respond and adapt to disturbances (i.e., resilience) could be considered in three dimensions, including socio-economic conditions,biophysical capacity, and production diversity.Specific measures should focus on increasing income generating capacity,infrastructure construction, migration, and the education level of farmers to strengthen the resilience to food security [16].

A common view, or belief, of future agriculture, is important for agricultural sustainability, which is a prerequisite of food security [17].A plain understanding of sustainability,shared in western and eastern culture, is to meet society’s food needs in the present without compromising the ability of future generations to meet their own needs.Sustainability is both a social-cultural and a technological challenge, requiring the combined efforts of researchers from the natural and social sciences and a range of stakeholders from government, business, and civil society [18].Concerns about sustainability in agricultural systems center on the need to develop technologies and practices that do not have adverse effects on environmental goods and services, are accessible to and effective for farmers, and lead to improvements in food productivity [19].Redesigning toward a resilient agricultural system [20] needs new approaches that not only integrate biological and ecological processes into food production but also make productive use of the knowledge and skills of farmers and people’s collective capacities to work together.In general, sustainable transitions in food systems could be facilitated by increasing efficiency (e.g., sustainable intensification), demand restraint(e.g., sustainable diets), and transforming food systems (e.g.,alternative food systems) to build resilient and efficient food systems [17], [21].

Evidently, there is a social dimension in food security and sustainability that covers the farmer’s behavior, the culture,the policy making, etc., beside the optimal use of natural resources.Natural-social-economic complex ecosystem is a coupling ecosystem composed of natural conditions, human society, and economic activities [22], [23].This multi-disciplinary subject calls for the frontier technologies in big data and artificial intelligence (AI) to achieve deep understanding of the system, to support decision making, and ultimately global food security and sustainability.Digital intelligence(DI) and cyber-physical-social systems (CPSS) are becoming the new research paradigms that have significantly shifted normal and conventional thinking and practices in many scientific fields, including ecological science and sustainability studies.This paper outlines our recent effort in using big data,artificial intelligence, digital twins, metaverses, and parallel intelligence to model, analyze, and manage the complex relationships and balance among plants, carbons, and waters in ecosystems.

The organization of this paper is as follows.In Section II,the intension of DI and CPSS are presented, as well as the organization mode of CPSS.Section III presents the roadmap of achieving the agricultural CPSS of various scales.A case study of decision making for water use efficiency is shown in Section IV.The agricultural foundation models are envisioned in Section V.The last two sections give perspectives and conclusion respectively.

II.DIGITAL INTELLIGENCE AND CYBER-PHYSICAL-SOCIAL SYSTEMS (CPSS)

A. Description, Prediction and Prescription Intelligence

With the development of technologies in the internet of things (IoTs), remote sensing (RS), AI, etc., smart agriculture has developed rapidly in the past few years [24].How data and information are utilized for a specific goal like sustainable development, and how to deal with systems that include social components, require new methodology that facilitate the achievement of desired outcomes.A parallel system has been proposed that is featured by artificial systems, computational experiments and parallel execution (ACP) (Fig.1),which is supported by DI of descriptive, predictive and prescriptive intelligence.

The software-defined artificial systems deal with the modelling of the crop, their natural and social environment, the management at various scales.The method includes crop models, environment models, and agent-based modelling such as those for human behavior.For crop model, it is recognized that the knowledge- and data-driven model supplement each other in the big-data age to make best use of the multi-source data and the well-developed process-based model in the past century [25].The temporal scale varies from hours to months.Similarly, the environment model includes those that predict the environmental conditions for a short period at a fine spatial scale or for a long period over a wide area over of time.The interaction among the crop, environment, and management are components of the system description.Corresponding models need to be chosen according to the temporal and spatial scales of interest.The artificial system helps to give a deep understanding of the real system.The description of artificial systems is named in general Description Intelligence.

Based on the artificial system, the computation experiment gives the possible optimal solutions that can guide the real system.Such experiments can be very costly in reality, in terms of time, labor or money.For example, to demonstrate the benefits of a proper cultivar and fertilization approach, it takes years and thousands of people to collect data and show the final results [26].Computational experiments conducted in cyberworlds are significantly faster than their real-world counterparts, even though running highly complex systems might be computationally expensive.With the visualization techniques, it helps to give a clear view of “What-If” question that is easy to understand for people of various backgrounds,for example, what is the ecological and economic gain if the cropping mode is updated in a region.Optimization techniques can offer support to give the needed actions toward a certain goal.The underlying intelligent methods are called Prediction Intelligence.

Parallel execution links the artificial systems with the real ones.From the real to the virtual, various data are useful to ensure that the artificial system behaves like the real one [27],so that the following prediction is not simply a computer game.From the virtual to the real, the former gives guidance based on the computational experiments, such as the optimal allocation of water in crop lifetime.The results support the decision-makers by giving optimal result.When ecological and economic constraints are both considered in crop planning, decision makers can choose the weight or the rules in making computational experiments and the final candidate results with their own knowledge.The continuous loop between the real and virtual systems gives updated results under changing environments, giving online “If-Then”answers.This methodology fits both the real-time control of a complex system with numerous agents, and the management of a long-term ecological system.Prescription Intelligence supports the parallel execution of real and virtual systems.The description, prediction and prescription as a whole form the DI.

The real and artificial agricultural systems compose the CPSS, which incorporate a social dimension compared to the cyber-physical system [28], the latter being featured by a tight integration among computation, communication, and control in system operations and interactions with their task environment.CPSS integrates social system and physical system through intelligent human-machine interaction in cyber-space[29], [30].CPSS in energy, smart vehicles, intelligent manufacturing and smart cities, etc.have been proposed for better management and control of real systems [31].Among them,an agricultural CPSS [32] has been raised that takes into account both environmental conditions and price information for crop planning, the decision on the timing, area and type of crop planting.

B. DAOs for CPSS Operation

The interaction between the real and artificial systems is based on the sensing of the real world.While the IoTs or RS solve the sensing of the physical world, the internet of mind(IoM) synergizes various intelligent agents in human society by connecting machines, as well as human and machine intelligence [31].The goal of building IoM is to complete the collaborative intelligence of the whole society.For agriculture,IoM can serve for the acquisition, representation, transmission, association, and utilization of knowledge as to sustainability from both regional ancient traditions and frontiers in smart agriculture.However, a large number of farms are located in developing countries [20], and out of the 570 million farms worldwide, 84% are smallholder farms covering less than 2 hectares [33].Thus centralized data and knowledge gathering is not efficient for smallholder farmers.Moreover, the information systems for smallholders need to be lightweight, low-cost, and intelligent [34].

Blockchain offers a complete, novel, and self-consistent architecture for establishing decentralized autonomous societies and ecosystems in a bottom-up fashion.The characteristics of blockchain technology can be summarized as “TRUE”(trustable, reliable, usable, efficient/effective).Decentralized autonomous organizations and operations (DAOs) provides a novel approach to CPSS operation (Fig.2).Under this architecture, it is not necessary to share the raw data, laden with private information, but rather the processed and secure results.Smart contracts, which could be flexibly predefined as“If-Then” types of scenario-response rules, can be composed.For example, if a question is answered, a certain amount of credit is given to the provider, which can be used to incentivize the inheritance of traditional knowledge from locally experienced people.For decision making, community autonomy needs various un-predefined “What-If” types of scenariodeduction rules that can be executed automatically.For example, what can be the alternative crop if the dry season dominates in the following months.

III.THE ROADMAP OF AGRICULTURAL CPSS

Like fractals, ecological systems operate on various scales,making the scale a fundamental consideration in modeling.Roughly speaking, the spatial scale ranges from a monotype field consisting of individual plants to regional and ultimately global levels (Fig.3).Key stakeholders, factors and challenges are related to the scale, as are the underlying infrastructures, technologies, data, models and services.

Fig.3.Decisions, stakeholders and organizations for different scales of agricultural systems.

A. Crop Level

At the crop level, the data of interest mainly include those of the plants themselves and their environment.Infrastructures designed to capture plant phenotype include hyper-spectral cameras, multi-spectral cameras, infrared cameras, lidar, etc.Using deep-learning models, non-destructive phenotypic data can be obtained, including plant height, leaf area index,ground cover, leaf count, flowering and fruit count, diseases,etc.[35] (Fig.4).Besides, there are sensors for obtaining the internal physiological state of crops such as the real-time sapflow, which is very useful to learn about crop water use.Recently, plant portable systems have been developed that can simultaneously monitor morphological traits, physiological processes, and plant-environment interactions [36].For the environment, it is very common to monitor air temperature and humidity, light conditions, soil temperature and humidity,etc., using IoT techniques, making remote real-time monitoring practical.

Fig.4.Phenotyping system of crop.

The description of crop growth and its interaction with the environment have been greatly studied.In the AI age, with easily accessible plant and environmental data, purely datadriven model has become possible that predicts the yield from the environmental data [37], given enough historical training data.The need to crop data for training has partly pushed the appearance of the knowledge-data driven model [25].The knowledge includes the known cultivar characteristics, or more comprehensively, the process-based models [38] that describe crop eco-physiological process and yield formation based on a set of knowledge and hypotheses of crops.Such models have the advantage of generating numerous kinds of crop phenotypic data speedily that are close to real crops,which contribute greatly to the initial training of the model.The analysis and prediction of crop phenotypic behavior for the target environment are helpful for breeders to select the best cultivar candidates efficiently, supporting the intelligent breeding, or breeding 4.0 [39].

The development and sharing of crop models call for decentralized science (DeSci) [40].Most crop models center on several main cereal and horticultural crops, developed by advantageous institutes.Research has been conducted in a centralized way, even for open-sourced crop models.It is not easy for less-sponsored institutes or start-ups in smart agriculture to benefit from already published models.The decentralized science brings a novel approach to funding, data owning and sharing, potentially fostering a unique research ecosystem among world modelers.

B. Farm Level

At the farm level, a key dimension for farm-holder is the economic benefit.High yield and high income are dominating goals.Unmanned aerial vehicles (UAVs) -born cameras are useful in obtaining farm-level information.Additionally, the product prices are of high importance.The price information is usually open to the public, accessible from Internet.For example, the biggest whole-scale market in Beijing shows the daily highest and lowest price on website.Weekly trends and the comparison with the last year are also shown.The infrastructures are mainly mobile phones which are widely accessible in the countryside.

Because of the long crop growth period, a high price at planting does not necessarily mean a high income at harvest because of price variation.Farmers often prejudge prices empirically, which can greatly deviate from the reality due to the Cobweb phenomenon, such as the unexpected price of ginger and garlic of year 2020 in China.Price is a complex social-economic index, dependent on the supply, the weather,public preference, etc.Price forecast models have been developed based on demand-supply information, or on historical data using machine learning algorithms.Temporal scale can be chosen for effective prediction, such as weekly and monthly average prices [41].Farm- or cooperative-level decision-making can be supported by parallel crop planning [42]by taking into accounted the predicted product prices, ecological constraints, profit model, etc.This work could be otherwise demanding if it is done manually when dozens of types of crop are to be planted on dozens of pieces of field, to meet a minimal market demand.A successful match between supply and demand can support contract agriculture [43] which is meaningful for economic sustainability.

For farm managers, another decision linked to sustainability is about the fertilizer and herbicide application.Digital intelligence can support with knowledge graphs [44], such as disease and insect recognition, diagnosis.The techniques of phenotyping can serve in smart agriculture (Agriculture 4.0)for precise locating of weeds and pulling with machinery [45],saving labors and avoiding the use of herbicides.However,small and cheap agricultural robots are still scarce that fit small-scale farms.

DAOs can serve as the digital cooperative of small-holders,helping to reduce cost and increase income jointly.Individual farms often lack strong bargaining power in the market.The origin of agricultural cooperatives was to enhance this power through collective purchasing and sales.In reality, although cooperatives have achieved the desired effect in some countries, some are not partly because of poor organization.Management rules can be defined as smart contracts in DAOs to be transparent and fair.Besides, agricultural knowledge from the experienced farmer can be monetized and imparted in this digital society.For the market side, DAOs also serve for the consumer cooperative comprised of individuals who insist on ecological products.Such preferences and needs are essential for guiding the production.

C. Regional Level

At the region level, information of priority includes the area and yield of staple crops.Remote sensing images from satellite provides data source to identify the type, area, yield, nutrition level of crops and the entropy condition of the soil, etc.[46].The historical RS images can serve to learn the rotation experience in the region [47], giving a data-based model of crop planning.The decision for policymakers and governors include the planning on divisions for the development goal of the region, for example, the expansion of planting area for certain types of crop in the coming years, or the best match of natural resource condition.Similar to farm-level, the crop planning models can be applied to have an optimized solution for local economy and ecology.Decisions include the policy to encourage or subside farmers toward the dedicated goal.The traceability feature of DAOs can enhance the reliability of subsidy distribution, as it serves in agricultural insurance [48].

Countries send their satellites to track the status of the nation’s food supply, which is important to decide the food policy.In China, for example, the “hyper-sensing” (high spectral, spatial, and temporal) satellites have been launched for monitoring, modelling and mapping of crops [49].As to the environment, global temperature, which is of interest even to a common individual, has drawn considerable attention [49].Such information influences people’s mind, for politicians to consider nation’s behavior in the world, for researchers to consider breeding for the future, and for citizens to consider their daily activities.The balanced planting strategy, that is,choosing the crop cultivar that best fits the natural hydrothermal condition, seems to be a low-cost action.This can be made possible by monitoring the hydrothermal condition the corresponding staple crop, and matching the target region with a similar condition.

Fig.5.Foundation models and services for food systems, including the levels of crop, field and region.

D. The Circular Causality

The common feature of ecological systems of all scales is that the attitude and opinion of stakeholders take effect on the natural systems, while the performance of real world influence on people’s thinking about what should be the future (the circular causality).The attitude (“believing”) is reflected in the decision making expressible in cyberworld with prescription intelligence.In the Internet, examples have already shown how the opinions of individuals can be guided, expressed and influence the real world.In ecological agriculture, the power of “believing” is shown in some alliances that appeal for the organic food, which supports small organic farms.With the digital intelligence of prediction, the opinion gets a computational base of “becoming” where the vision for the future can be shown to policymakers, the public and other stakeholders with figures and data, based on the description of situation(being).DAOs potentially provide a technically solution to the participatory management of complex system where conflict between development and conservation exists.

The circular causality also exist in the outcomes among system of different scales.As mentioned above, in agricultural systems, the components vary from crop to a regional and even global scale.An analog is system biology, which is the computational and mathematical analysis and modeling of complex biological systems, including the interactions and behavior of the components of biological entities (molecules,cells, organs, and organisms).The behavior of a higher level system is usually an emergent property of the sub-system,with an underlying intelligence that self-regulates the system’s behavior.Such property applies in system agriculture [50].For example, at higher planting density, greater mortality takes place in wheat tillers to get the best seed survival.It applies also in social-economic phenomena visible in product price variation.

For the description of natural properties at different scales,the linkage of models of different scales can be assured with the key input/output result.For example, at the individual plant level, the model deals with organogenesis, organ expansion, and organization into a three-dimensional (3D) structure,as in the GreenLab model [27].Although desirable, it is not necessary to have 3D visualization of all crops in a farm where millions of individual plants are inside.Instead, the representative structures are sufficient to visualize the effect of fertilizer application.Linkage can be achieved by making sure that both models give the same average yield [51].

IV.AGRICULTURAL FOUNDATION MODELS FOR FOOD SYSTEMS

Although distinguished, the borderlines between systems of different scales are not strict, and there are common technical needs as summarized in Fig.5.The integrated sky and ground monitoring technology combine RS, UAVs, cameras, and sensors, creating an efficient infrastructure to collect various types of data including monitored environmental data, crop growth data, agricultural domain knowledge and rules, farmers’ information, market prices, etc.The data types encompass text, imagery, audio, and video.To illustrate, in the field of plant phenotyping, the predominant data source focuses on the utilization of images collected from a variety of cameras.The crop models can also be used as data resources by simulating various scenarios.The agricultural data from various sources offers key information for decision-making.

Although the general big models, as represented by Chat-GPT, have been developed and are being developed rapidly,agricultural decision support is still not widely applied.Big foundation models can handle massive datasets efficiently,which is crucial for tasks requiring extensive data analysis,such as language translation and sentiment analysis.They can generalize well across various tasks and domains.These models can be fine-tuned for specific tasks, saving time and resources in developing custom models from scratch.They can achieve impressive results with minimal task-specific fine-tuning.This versatility makes them valuable in a wide array of applications.Despite their impressive capabilities, it's important to acknowledge that big foundation models present substantial challenges related to resource demands, ethical considerations, and the real-world implementation of these technologies.

Practical solutions to real-life problems, such as “What shall I plant next year?”, are scarce [52].The challenges include the huge diversities of natural environmental conditions, the big cost of obtaining and analyzing RS with the earth verification,the diversity of farm scales and economic levels, etc.It is well known within the research community that the quality of data/feature determines the upper limit of machine learning.Meanwhile, models and algorithms are the only means to approximate that limit.Crop models have been developed for the main staple and horticultural crops, but for the minor crops, the models are negligible.Thus there is still huge a digital gap for the service of farmers all over the world.There are well-intentioned efforts seeking effective solutions to promote sustainability and achieve the United Nation Sustainable Development Goal of “No Hunger”.Platforms with frontier technologies are desirable that feed the earth with human wisdom.

Scenario engineering (SE) [53] has been proposed based on the fact that feature engineering can increase the value of existing data and improve the performance of deep learning.SE is defined as an integrated reflection of the scenarios and activities within a certain temporal and spatial range, where all actionable AI are encouraged to complete the design, certification, and verification.For example, it can be a digital farm,where the duty of digital farmer is to make sustainable management according to the prediction of climate, yield, cost,ecological and economic profit.The description of farm-level operation is needed according to the digital natural and socialeconomic environment.It is also desirable for the training of professional farmers in schools, where the real facilities are expensive to build.The cyber-space scenarios are needed to answer specific agricultural challenges as well, which otherwise cost at least months to fulfill offline.The goal of SE is to realize 6S for smart development and sustainability of intelligent systems: safety in the physical world, security in the cyber world, sustainability in the ecological world, sensitivity to individual needs, service for all, and smartness in all [53].For each goal of “6S” in SE, the corresponding indexes must be designed in line with the specific applications.

The digital system around crops finally leads to farm operation system [54] together with intelligent machineries.The system encompasses the entire product chain, from the initial planning and the production process to the final connection to the market.The ongoing unmanned farm has shown this trend,where the sowing, fertilizing, and harvesting are conducted by machineries that are carefully scheduled.It can also be shown in highly automated plant factories, where the pipelines are arranged until the package of clean vegetables.

The domain knowledge, rules, and models plays a crucial role in the development of the agricultural foundation model.The development of crop foundation models calls for the global cooperation of AI suppliers, breeders, researchers and growers.It concerns not only the gathering of existing knowledge in literature but also the living knowledge in the human mind by integrating parallel learning theory [55] and federated intelligence [56] in DAOs.

V.CASE ANALYSIS ON WATER USAGE

As an important basic natural resources, water resources are essential to supplying services and maintaining ecosystem stability [57].However, since the Industrial Revolution, the contradiction between socio-economic development and ecological protection around water resources has become more and more prominent under the rapid development of society and the economy [7].The rapid socio-economic development cannot be separated from the massive exploitation and utilization of water resources, the accompanying water shortage, and ecological degradation, which in turn restricts the development of society and the economy [58].Therefore, optimizing the allocation of water resources in the Natural-Social-Economic complex ecosystem is crucial to comprehensively achieving sustainable development.

Scientifically evaluating the coordination and development state of water resources, social economy and ecological environment system is particularly important, and research on the coupling model between systems is increasing [59]–[61].The ecological water diversion project in the Heihe River Basin is a successful case of river basin ecosystem restoration in China, which indicates that the sustainable development in river basins needs to consider the basin as a whole and coordinate the relationship between the upstream, midstream and downstream areas [62].The establishment of coupling models in the Heihe River Basin, including an eco-hydrological model to characterize the natural system and a socio-economic model to characterize the socio-economic system, is realized through the land-use model and water resources model to simulate the interaction between natural and economic systems [63].

While global agricultural water accounts for about 70% of the total water consumption, the complex external environment limits the optimization of agricultural water use.Considering the spatial heterogeneous characteristics of water-agricultural-ecology nexus system, uncertainty of the crop simulation model is used to improve the robustness of simulation model of agricultural water use and provide a reference for regional managers in measuring distributed agricultural ecological benefits as well as making tradeoffs among economic,ecological benefits and risks [64].Based on the framework of agricultural water-saving reallocation, water resources transfer and trading are established to maximize water resource benefits [65].Studies have pointed out that agricultural water transfer to the second industry (533×104 m3) and tertiary industry (235×104 m3) in the Fenhe region could bring an incremental benefit of about 8.66-8.94 billion CNY [66].Therefore, it is possible to optimize the allocation of water resources among different industries to maximize the benefit of water resources.Our recent study also found that strict water management regulations promote industrial transformation and sustainable development in Inner Mongolia, with the decreasing proportion of environmentally harmful industries such as coal and steel, and the increasing proportion of tertiary industries (especially tourism), effectively reducing the dependence of economic development on water resources and alleviating water stress in drylands.Through the water resources regulations, the internal momentum for industrial transformation is stimulated, ensuring a balance between economic development and environmental protection and promoting sustainable development in China’s northern drylands [9].However, there is still no systematic understanding of how to coordinate the distribution of water resources among ecology,industry and domestic.

System integration is critical for understanding the interconnections of the Natural-Social-Economic complex ecosystem and putting forward sustainable solutions [67].For a given region, with a definite total amount of water resources, taking into account ecosystem health, food security, economic development, and human security, there would be a tradeoff between agricultural water use and other industries once water allocations for ecological and domestic needs are ascertained.It involves crop yield, water use efficiency, industrial structure adjustment, water-saving technology, so the optimal allocation of water resources should be considered from a perspective of system coupling (Fig.6).As for the whole river basin, given the service functions in different basins, water resources allocation is carried out from the holistic perspective, which contains water resources cooperation and trading,ecological compensation, industrial transformation, technical assistance, and green development on the premise of ensuring food security and human basic needs.Operationally, largescale multi-system coupling models and digital presentations are needed to accurately guide the optimal allocation of water resources and promote sustainable development.

Fig.6.Framework of water coupling models for natural-social-economic complex ecosystem.

VI.FUTURE AGRICULTURE WITH DIGITAL, ROBOTIC AND BIOLOGICAL FARMERS

Fig.7.Three types of farmers in three worlds.

Taking the agricultural production and grain management in the Yellow River Region as an example, we envisage three types of humans to address the complex management of regional ecosystems, as depicted in Fig.7: digital farmers,accounting for about 80% of the future workforce; robotic farmers, comprising approximately 15%; and biological farmers like us, making up only 5%.Organizations in future Agriculture 5.0 [68] would operate under the following modes.

1) Under autonomous mode (AM), biological farmers play a crucial role in monitoring the overall farming processes.They oversee the activities to ensure the smooth functioning of the farmland and make informed decisions with the support of digital farmers.Digital farmers utilize sophisticated algorithms and artificial intelligence to process information from various sensors and devices installed across the farmland.This enables them to detect potential issues, optimize resource allocation, and provide valuable insights to biological farmers,allowing them to make data-driven decisions and maximize productivity.

2) Under parallel mode (PM), biological farmers at remote sites or in the cloud need to supervise and work with digital and robotic humans in the fields or at the edges.Robotic farmers, equipped with specialized agricultural machinery and automation capabilities, execute tasks such as planting, irrigation, and harvesting with precision and efficiency under the guidance and supervision of biological farmers.This collaboration allows for simultaneous work on different areas of the farm, significantly reducing the time required for completing tasks.

3) Under expert mode or emergency mode (EM), depending on the situation, biological farmers or experts must be rapidly deployed to the working sites or fields to address tough emergencies.

The seamless integration of biological farmers, robotic farmers, and digital farmers under the AM, PM, and EM modes has the potential to revolutionized agricultural practices.They would make small tasks, big models, and deep intelligence an ecological routine for intelligent agriculture.Through efficient collaboration, utilization of advanced technologies, and real-time data-driven decision-making, the overall efficiency of farming operations has skyrocketed, resulting in an astounding 960-fold increase in productivity, because the efficiency of biological farmers will be elevated 20 times,and each of them is equipped with 3 robotic farmers and 16 digital farmers, resulting in a staggering efficiency improvement of 960 times in agricultural operations.This transformative approach is paving the way for a new era of intelligent and highly efficient farming practices.

VII.CONCLUSION

This paper presents the approaches to support global food security and sustainability with digital intelligence for description, prediction, and prescription.The key elements toward agricultural CPSS on different scales with DAOs are identified.A case analysis on efficient water usage is done.Perspectives are given to future agriculture with digital, robotic, and biological farmers.