A scoping review of the digital agricultural revolution and ecosystem services: implications for Canadian policy and research agendas

PA and smart farming technologies make production systems more precise by using advances in Global Navigation Satellite System technologies, big data, and automated robotic farm equipment to improve guidance, recording, and management (“reacting”) technologies (Balafoutis et al. 2017; Finger et al. 2019). Often PA also takes advantage of specialized software, data analytics, and modern information and communication technologies to inform decision-making (Finger et al. 2019). This requires sensors attached to individual animals, plant locations, other geographic locations on farms to gather data. The interweaving of these technologies creates an IoT involving feedback between sensors and equipment such as smart tractors or robotic milking machines to provide real-time automated adjustments that aim to improve agricultural objectives (Schrijver et al. 2016; Tzounis et al. 2017). In addition to use on large grain operations, in North America and Europe, PA techniques are widely applied to dairy production and may have led to improved cow health and reduced environmental footprints (Rotz et al. 2019b). While PA is typically associated with large-scale grain and animal operations, the broad definition of PA as site-specific, sensor-based monitoring and automated adjustment also applies to smaller-scale operations that can afford such technologies (Finger et al. 2019).

PA is associated with sensors, big data, and smart farming (Table 5). Finger et al. (2019) reviewed the quality of evidence in studies on PA and smart farming tools’ ability to improve nutrient management (Q1 and Q3), assist with climate regulation (Q1 and Q3), save water (Q1), reduce herbicide and pesticide (Q3). Regarding herbicides and pesticides, Clapp (2021) argued that DATs may be one reason we have seen increasing glyphosate use rather than the anticipated decrease with DAT adoption. PA and smart farming tools are associated with improved nutrient management (Q1 and Q3) on large farms (Balafoutis et al. 2017) and potentially on small farms (Kanter et al. 2019), potential contributions towards climate regulation through lower reliance on external inputs (Q1) and avoided fuel use (Q3) (Balafoutis et al. 2017; Kakamoukas et al. 2021), and reduction of the land footprint of agriculture (Q1–3) (Capmourteres et al. 2018). PA benefits to ES are most likely to be empirically measurable in large-scale farming operations (such as grain and oilseed) and biophysical monitoring of animals in red meat, poultry, pork, and aquaculture sectors. Yet, smart farming technologies deployed in smaller farms using agroecological approaches may also show measurable enhancements to ES by increasing farm level input efficiency and allowing monitoring of environmental indicators (Wittman et al. 2020). While PA and smart farming literature indicates potential environmental benefits, there are very few empirical studies showing causal inference on specific ES (see Finger et al. 2019). Many of the controlled studies focus on variable rate technology control of nutrient (mainly nitrogen) and water inputs but report only on agricultural productivity instead of environmental outcomes (see for example Khosla et al 2002). Targeted randomized controlled trials that examine specific PA and smart farming technologies while measuring environmental variables would strengthen PA and smart farming claims to benefit ES.

There is considerable overlap of big data literature examined in this review with sensors, PA, and AI and machine learning. Big data is cited as a way to improve nutrient management through more precise application of nutrients leading to reduction in input demands (Q1) and reduction of nutrient leaking (Q3). Similar dynamics are cited for reduced herbicide and pesticide use (Q3) (Balafoutis et al 2017; Weersink et al. 2018). Big data may also reduce GHG emissions (climate regulation) by using atmospheric and local emission monitoring to inform farm management (Q3) (Delgado et al. 2019). As well, big data can reduce the need for land (Q1) by identifying less profitable areas and allowing the cultivation of more biodiversity (Q2) in agricultural landscapes (Capmourteres et al. 2018). While big data drives the DAR by allowing rapid advances in fields such as genetics and genomics (Zaidi et al. 2020) and CEA (Charania and Li 2020), better models for linking big data sets in PA, precision conservation, and ES modelling are needed (Capmourteres et al. 2018; Khanna et al. 2018; Delgado et al. 2019). Additionally, research gaps include ways in which big data might drive positive management practices for reducing agricultural energy demands and increasing pollination, soil health, nitrogen fixation, local air quality, and pest management.

AI and machine learning literature often also cites sensors, big data, robotics and automation, and UAV and remote sensing. AI and machine learning was linked to improvement in nutrient management (Q1 and Q3) as big data, UAV, robotics, and AI and machine learning are used for efficient spot application of nutrients, herbicides, and pesticides (Q3). In addition, UAV, big data, and AI and machine learning were also linked to monitoring that would allow early identification of pests and disease for treatment and limitation of spread in the surrounding ecosystem (Q3) (Gossen and McDonald 2020). In combination with genetic engineering and big data, AI and machine learning accelerates breeding of climate resilient plants (climate regulation) thus enhancing agriculture and ecosystems’ abilities to respond climate variability (Q2) (Harfouche et al. 2019). While AI and machine learning include several applications (such as targeted pest reduction) that have apparent ES benefits (reduced pesticide use), there is a paucity of empirical research linking AI and machine learning to most ES (see Table 5). This sector would benefit from targeted research projects examining how AI and machine learning are used in specific agricultural cases that lead to ES benefits.

CEA refers to a continuum of technologies that include greenhouses, container farms, aquaponics operations that integrate vegetable and fish production, and vertical farms. These are all variations of farming that use sensors, timers, lighting, and other technologies to create smart management systems using AI and automation. Academic research on CEA has proliferated³ and there is a rapidly growing business sector with start-ups such as Ostara⁴ and Freight Farms⁵ emphasizing high-tech, low environmental impact farming (Charania and Li 2020). Yet, the literature on CEA and ES often relies on hypothetical LCA models (see for example Forchino et al. 2017; Benis et al. 2017, and Wildeman 2020). Empirical studies that directly compare field agriculture to CEA in terms of ES benefits (see for example Newell et al. 2021 on hydroponic fodder) are rare. We found only four records that mention CEA directly addressing ES, but numerous CEA publications (and business media) that assume environmental benefits from CEA.

Regardless of the specifics of the CEA application, most studies of CEA find them to have hypothetically more efficient nutrient management (Q1 and Q3), herbicide and pesticide use (Q3), and water management (Q1 and Q3) when water cycling is used. There is a hypothetical argument regarding climate regulation (Q2 and Q3) based on potential reduction of food miles, but only if CEA are geographically distributed. Yet, Benis et al. (2017) showed in a LCA that depending on surrounding climatic factors some CEA operations may increase GHG emissions relative to field agriculture. There is also a hypothetical argument about CEA creating a lower land footprint (Q1 and Q2) that may lead to land sparing for conservation based on higher production efficiency per square meter and averted agricultural land conversion. This argument does not consider high energy demands (light, heating, and cooling) leading to increased GHG emissions and upstream impacts of energy generation on land, biodiversity, and other ES (Benis et al. 2017; Newell et al. 2021; Beacham et al. 2019). Additionally, much of the adoption of CEA may not be averting existing agricultural land use or new conversion. For example, Freight Farms emphasizes that many of its adopters have little agricultural experience and choose CEA in extreme conditions (e.g., Alaska and dense urban areas). The literature widely recognizes that these hypothetical ES benefits are highly context dependent and require renewable energy to lower the impacts of CEA high energy demand. The assumption that CEA will replace a substantial amount of any agricultural sector and benefit ES, therefore, remains speculative and requires more empirical research.

As shown in Table 5, robotics and automation are highly associated with AI and machine learning, sensors, and big data. In field agriculture, robotics use in relation to ES emphasizes autonomous field equipment that can precisely apply nutrients (nutrient management) (Q1 and Q3) and herbicides and pesticides (Q3) or may completely eliminate the use of herbicides and pesticides (Q3) through the use of lasers for weed control (Duckett et al. 2018; Gossen and McDonald 2020). In addition, robotics contribute to better water management (Q1 and Q3) through optimization of irrigation systems, lower energy demand (Q1), and GHG emissions (climate regulation) (Q3) by electrifying farm vehicles and implements, and increase soil health (Q2) as the autonomous field units cause less compaction than larger tractors and conventional equipment (Duckett et al. 2018). When robotics are deployed as part of CEA, they may contribute to land-use optimization (Q1) and disease reduction (Q2 and Q3) by reducing interface with the surrounding environment (Charania and Li 2020). While research on field applications of robotics and automation shows benefits to ES, the purported benefits of these technologies in CEA suffers from the same caveats of the overall CEA literature—lack of empirical research verifying supposed ES benefits of creating CEA rather than pursuing field agriculture.

Agricultural genetics (and genomics) reduce variability in agricultural production by targeting genetic modifications that enhance desirable traits that, for example, increase production potential, lower pest damage, and increase storage times. This field’s most numerous overlaps with other technologies are with robotics and automation and cellular agriculture. The association with cellular agriculture is likely a result of the novel foods literature examining cellular agriculture and advances in genetic engineering in the same publications. Genetics is highly associated with the management DAT cluster, but is also the second highest contributor to the novel foods cluster. While agricultural genetics includes traditional agriculture selection and crop breeding practices, we focused our search on new genetic engineering approaches. Following NASEM (2016), we use the term “genetic engineering” to refer to transgenic techniques (recombinant DNA), somatic cell transfer, genome editing (CRISPR)/Cas9 nuclease system), RNA-based approaches (e.g., dsRNA, mRNA, RNAi), synthetic chromosomes, and target epigenetic modification. In plants, such modifications have typically targeted pesticide tolerance but are now used to increase yield, nutritional content, stress tolerance, medicine production, and pest resistance (NASEM 2016). Animal modifications often target objectives such as increased growth, increased production (e.g., dairy and wool), and disease resistance (Forabosco et al. 2013; Long 2014; Lievens et al. 2015). Advances in agricultural genetics have been accelerated by DAR technologies such as AI, robotics, and big data (Harfouche et al. 2019).

Agricultural genetics includes some of the most scientifically supported and promising benefits to ES. These benefits include resource efficiencies such as with nutrient-use (nutrient management) (Q1 and Q3) (NASEM 2016; Brookes and Barfoot 2018) and water-use efficiency (water management) (Q1 and Q3). GE may lead to reduced land footprint (Q1) due to increased nutritional quality of forage and (theoretically) avoided agricultural land expansion (Brookes and Barfoot 2017; NASEM 2016). Waste reduction (Q1 and Q3) within agriculture may be reduced through increased storage times, thus leading to reduced need for inputs and harmful impacts on the environment (NASEM 2016). GE may benefit ecosystems by increasing biodiversity (Q2) in plants engineered for climate variability, drought tolerance, salinity tolerance, and disease/pest resistance in plants (e.g., species restoration of blight-resistant Castanea dentata) (NASEM 2016) and pollinators (pollination) (Q1 and Q2) (Greenlight Biosciences 2021). Furthermore, GE enables climate regulation through reduced methane production (Q2 and Q3) from better-quality forage (NASEM 2016), enhanced soil carbon sequestration, and avoided tractor fuel use (NASEM 2016). Other ES-related benefits include enhanced photosynthesis (biomass) (Q1) (NASEM 2016; Pellegrino et al. 2018), enhanced nitrogen fixation (Q2) (NASEM 2016), disease resistance (Q2 and Q3) (Zaidi et al 2020), pest resistance (Q2 and Q3) (Mezzetti et al. 2020), and reduced herbicide and pesticide use (Q3) (Zhang et al. 2017; Rodrígues and Petrick 2020).

Types of sensors include optical (remote observations of farmlands), mechanical (measuring soil compaction), location-based (recording latitude and longitude), dielectric soil moisture, electrochemical (measuring pH and soil nutrients), and airflow sensors. While sensors are common in most of the DATs, sensors were most often mentioned in cases of field-based nutrient management (Q1 and Q3) (see for example Li et al. 2016) in AI and automation applications (Talaviya et al. 2020) and monitoring emissions for climate regulation (Q2) (see Balafoutis et al. 2017). Continued development of micro and nanosensors that are more sensitive and smaller but have stronger connectivity than existing sensors creates big data for agriculture and drives most of the technologies mentioned above (Fraceto et al. 2016).

Supply chain management can benefit in multiple ways from digital technology, the most well-known of which are software management systems that manage, track, and help control shipping. More novel technology such as blockchain offers ways of tracking goods that theoretically offer both agricultural producers as well as consumers opportunities for increased transparency, better global market and direct access, alternative financing arrangements, and the chance to add value to food products (Ge et al. 2017; Antonucci et al. 2019; Astill et al. 2019). Blockchain is a distributed digital ledger that is immutable. When made public, the blockchain shows transactions over time that allow traceability and a great level of transparency and efficiency for market participants (Astill et al. 2019). Blockchain is currently being used to provide transparency and efficiency in agri-food supply chains including coffee, tuna, beef, beer, milk, and pasta (Antonucci et al. 2019). In the case of tuna, blockchain is being used to verify sustainability claims about fishing practices (biodiversity) (Q1) (Antonucci et al. 2019). Increases in efficiency and transparency have the potential to reduce GHG emissions (climate regulation) (Q3) and waste (Q3) (Astill et al. 2019) associated with complex food systems. Yet, there is inadequate empirical data supporting any of the above ES benefits in relation to blockchain. Moreover, there may be negative environmental impacts of current “proof-of-work” blockchain mining systems that consume large amounts of energy—blockchain technology may transition to renewable energy sources or to less energy intensive “proof-of-stake” frameworks to reduce energy demands, emissions, and related ecosystem impacts (Giungato et al. 2017; Zhang et al. 2020). In summary, research on the benefits of applying blockchain to agricultural interactions with ES and overall sustainability requires more empirical work.

ICT that increase human connections, mobile technology use, cloud computing, and supply chain management options compromise most records in this category. This type of connectivity has a necessary role as it enables many of the DATs but impacts on ES are not well-established.

Advances in ICT are changing food system dynamics. In 2021, there remains a digital divide in the form of both broadband internet and mobile telecommunications access between rural and urban areas in many countries that requires investment in infrastructure (Ramsetty and Adams 2020; Roese 2021). The Canadian Radio-television and Telecommunications Commission (CRTC) declared broadband internet an essential service in 2016—with minimum standards of 50 megabits per second download and 10 megabits per second upload (Ruimy 2018). The government launched a CAD$1.75 billion fund to connect 98% of Canadians to high-speed internet by 2026, and 100% by 2030 (CBC News 2020). By 2020, 45.6% of Canadian rural communities had access to broadband at high speed and 97.4% of these communities had access to Long Term Evolution (LTE) mobile services with 4G wireless access. Broadband internet and mobile technology may benefit all scales of producers by allowing more advanced farm management systems to link local sensors and machines into larger networks. Moreover, this connectivity may enhance the ability of small and medium producers to source inputs, access information and financial services, and sell directly to retail consumers and wholesalers (Quandt et al. 2020; Krell et al 2021). There is a need for more empirical evidence on how connectivity such as mobile services impact pricing, production strategies, and decisions regarding ES (Baumüller 2018).

While expanded ICT connectivity and social media access have increased producers’ ability to organize alternative, regional, and ecologically focused agricultural movements (Stevens et al. 2016), the relevance of online connectivity made especially clear during COVID-19 (Weersink et al. 2021). The COVID-19 crisis saw increasing demand for online food platforms (Change and Meyerhoefer 2020; Thilmany et al. 2021) and increasing relevance of DATs to food system functioning (Fairbairn and Guthman 2020). Having an internet presence and the ability to interact online to receive orders and communicate with customers allowed local producers overwhelmed by increased direct demand (Coppolino 2021) to respond and reorient perishable products in wholesale operations to direct to consumer sales (Thilmany et al. 2021). While there was limited time to scale-up production, digital connectivity increased the flexibility and resilience of producers and the entire food system during the pandemic (Coppolino 2021; Stoll et al. 2021).

While there is not sufficient empirical evidence to support links between these connectivity DATs and ES, theoretically they may (i) reduce food waste (Q3), (ii) optimize transport to lower emissions (climate regulation) (Q3), and (iii) provide opportunities to transparently link environmentally friendly producers with companies and individual consumers thus creating market-based incentives (and possible certification strategies) for agricultural practices that reduce input ES demand and waste (Q1–3). These interactions seem logical, but more empirical research is needed to show measurable benefits to ecosystems and ES.

Cellular agriculture uses tissue engineering or fermentation-based processes to produce protein-containing products and tissue-based foods such as meat, fish, and dairy. The primary distinction between tissue engineering and fermentation processes is that tissue engineering uses cells or cell lines from living animals and fermentation does not use such materials. In both cases, protein products mimicking animal muscle and products (e.g., milk) are produced from either cell cultures or bioreactors. Digital technologies allow for precise identification of cellular data needed to produce new products. The cost of producing such products is falling on an exponential curve, suggesting that price parity with animal production is rapidly approaching. Experts suggest that high-volume production using serum-free media is rapidly approaching feasibility (Post et al. 2020).

How these novel foods may impact ES is less clear. The main argument in the literature is that conventional livestock production has a disproportionately large impact on ES due to the GHG emissions caused by clearing forests for pastures, raising grain to feed livestock, and methane emission from ruminant animals (Willett et al. 2019). The energy necessary to produce cultured meat at scale, and how that compares with conventional meat production, is still unclear in scientific literature as LCAs tend to be highly speculative, use incompatible models, compare cultured protein only with ground beef, and use conflicting model assumptions (Mattick et al. 2015; Lynch and Pierrehumbert 2019; van Eenennaam 2019; Glaros et al. 2021). Furthermore, the technology for cellular agriculture is evolving so quickly that LCAs may quickly become out of date. Cultured dairy substitutes are moving rapidly to scale and are already sold in some regional and national markets (e.g., Perfect Day cultured nondairy ice cream).⁷ Yet, as of 2021, no LCA details potential environmental impacts of switching to cultured dairy substitutes versus conventional dairy products, while it is widely assumed that this alternative pathway would significantly reduce the number of emissions and other environmental impacts inherent to dairy production (GRAIN and IATP 2018). As mentioned above, even the climate regulation and energy demand benefits require assuming that low-carbon energy sources are readily available (Alexander et al. 2017; Lynch and Pierrehumbert 2019).

Given the state of the literature and LCAs that purport to detail cellular agriculture’s environmental impacts, the ES benefits of cellular agriculture are not empirically well supported. There is a need for more systematic and comparable models for LCA on cellular agriculture and the products that are supposed to lose market share. In addition, the field suffers from some of the modelling flaws mentioned above in that consumers are assumed to make a replacement choice between conventional meats/by-products and these novel foods. Yet, people may simply continue consuming conventional meats/by-products while adding these novel foods to their diets. More social and environmental research is required for cellular agriculture.

Technologies are enabled and constrained by the sociopolitical and financial systems in which they are embedded. Mosby et al. (2020) pointed out that farmers may not necessarily use DATs to enhance ES unless they receive financial rewards to do so—that simply being aware of inefficiencies on the farm is not enough to provide motivation for good management. This point was elaborated by Pannell (2017) who reviewed the economic logic of PA tools that allow for the precise use of inputs. Pannell noted that while there are optimal levels of nitrogen application that maximize yield while minimizing waste, these “payoff curves” (defined as the relationship between inputs applied and yields) are quite flat. This suggests that farmers pay little in the way of penalties if they over or under apply inputs such as fertilizers. Thus, the potential benefits of DATs may not provide sufficient economic payoffs to farmers to justify the expense of investing in these technologies. Higgins et al. (2017) showed that investing in new DATs when harvest yield risk is unknown and new technologies’ data formats are not interoperable can also hinder agriculture technology adoption. These strains of research on farmer behaviour suggest that economic incentives and disincentives are needed to ensure technology is deployed to create positive outcomes for the environment. The adoption of DATs that benefit ES requires a long-term process of scholars, stakeholders, and decision-makers engaging in co-development of effective solutions that address diverse needs (Bennett et al. 2021).

Virtually all the technologies described as part of the DAR depend on large volumes of data that must be integrated together to produce the insights farmers need to reduce their environmental footprint. In this way, the DAR is no different than similar technological transformations that have unfolded in IT, transportation, or medicine. Data-related challenges are identified in at least three key areas. First, data that come from diverse sources must be interoperable. This requires protocols that allow remote sensing, market, soil, weather, and harvest data to be standardized and harmonized so that algorithms, operations, and apps can access and use the data (Wolfert et al. 2017). Second, there is an urgent need for data governance that addresses privacy protection, data ownership, and compensation for farmers when companies use their field data to improve the performance of private sector technologies (Jakku et al. 2019). Rotz et al. (2019b) pointed out that farmers are justifiably concerned when the data they generate is sold back to them in the form of decision-support tools. Third, as agricultural digitalization occurs these systems also become susceptible to cybersecurity vulnerabilities (Barreto and Amaral 2018; Nikander et al. 2020). Organized, high profile cyber-attacks by hacker groups such as Revil on agricultural businesses such as JBS S.A. (the largest beef producer in the world) and Agromart in Canada are raising alarms about the lack of cybersecurity throughout the agricultural supply chain.⁸ Such issues must be addressed if DATs are to be adopted to enhance ES.

While novel foods and food production strategies may offer environmentally friendly ways of restructuring food systems, it is well documented that many novel technologies may not be readily accepted by communities—whether it be a subtle shift in colour or texture or a move to GE foods (Kuhnlein and Receveur 1996; Finucane and Holup 2005). In many ways, these challenges are somewhat intractable and different in nature than other technological, governmental, or economics of scale issues.

Labour and training gaps stand in the way of deploying DATs in a way that helps enhance ES. Demographic dynamics of the agricultural sector make investing in modern technologies challenging. The average age of farmers in North America is rising (in Canada it is 55 years old) and 70% of farmers do not have succession plans (Beaulieu 2014). There is an urgent need to assess agricultural labor shortages and train digital agriculture workers in Canada (Stackhouse et al. 2019). Federal reports argue that promoting environmental sustainability is integral to the identity of the Canadian agri-food sector (Advisory Council on Economic Growth 2017; Senate of Canada 2019). Taken together, these reports suggest that there is a skilled labour shortage in the agri-food sector in Canada and a need to invest in sustainability and technology training in post-secondary agricultural programs.