Introduction

Similar to many other industries, agriculture is experiencing dynamic and diverse developments in the digitalization of operations and operational infrastructure. Consequently, farmers in mid- and western Europe have access to an increasing number of digital applications and precision agriculture (PA) technologies. A broad range of technologies are available as decision-making support tools for practitioners and to facilitate site-specific and more efficient processes in both crop and livestock farming (Paustian & Theuvsen, 2017). PA in crop farming supports the management of spatial and temporal variability through the use of information and communication technologies (Blackmore et al., 2003). PA in livestock farming involves the use of technologies that allow continuous real-time monitoring of individual animals (Berckmans, 2017), increase on-farm production efficiency (Rowe et al., 2019), and bring forward automation of operational tasks. Thus, for both sectors, PA can help farmers optimize input management based on actual agronomic needs (Tey & Brindal, 2021). In addition, a wide range of “smart farming” technologies goes beyond PA by supporting management tasks and decisions that are not related to field locations or individual animals. Technologies such as smartphone applications, digital field recordings or ground sensors are digital and autonomous applications that support the processing of real-time data. This paper addresses representative technologies from all areas of digital, automatization and precision farming currently available to farmers. Thus, it goes beyond the consideration of only “classic” PA technologies to encompass complementary smart digital technologies for data collection, analysis, and decision making (Tey & Brindal, 2021).

The term digital transformation has been defined many times in the literature. In relation to companies or business sectors, it is often referred to either as evolution, or the creation of entire new business models (Berman, 2012). Both, digitization, and digitalization can be considered to be part of digital transformation, but the distinction between the terms should be made clear. Digitization is the starting point for, and also describes the conversion from, analog to digital processes. It is often referred to as the 3rd industrial revolution (Greenwood, 1997). Digitalization goes a step further, as it describes socio-technical processes and their impacts on human activities that result from the use of (interconnected) digital technologies (Nambisan et al., 2017). In this context, PA is mainly related to digital on-farm activities (e.g., sector control, yield mapping), while complementary digital technologies broaden the spectrum of applications, and serve to interconnect farmers with other farm systems or value chain stakeholders (Rolandi et al., 2021).

This paper examines the current situation regarding PA and digital technology adoption in a small-scale farming region in Europe. Rather than using the common scientific approach of testing preconceived hypotheses, an engineering method is applied here. Koen (1984) give a definition of the engineering method derived from several existing concepts and morphologies. Originally conceived as a methodological approach to generating solution designs from limited information, the main focus of research that employs this method is the use and potential of heuristics. Tey & Brindal (2021) have recently demonstrated the limitations of the use of binary models in a meta-analysis of adoption studies (see literature review). Thus, the methodological approach used in the research presented here relies instead on descriptive measurement to describe trends in the adoption of specific technologies. This approach allows identification of patterns of sequential steps that actually occur in-farm digitalization, in order to better understand the diffusion of PA and digital technologies in both crop and livestock farming. This includes a deeper understanding of which digital technologies currently serve as important entry points in the digitalization process. Another focus of this analysis is on the sequential adoption of multiple PA and digital technologies, building on the work of Schimmelpfennig & Ebel (2016) and Miller et al. (2019). Both actual joint usage of technologies on-farm, and potential future trends in technology diffusion at four sequential points of adoption (entry technology, technologies currently in use; planned short-term investment; planned mid-term investment) are considered.

Another unique contribution of this study is the explicit consideration of a small-scale farming region in Europe. While a majority of the literature investigates adoption of PA and digital technologies in larger-scale farming, this paper focuses on farm-level analysis in a small-scale agricultural context. Small-scale agriculture is a loosely used term and defined differently in different regions of the world (Bosc et al., 2013). It is often used in reference to agriculture in developing countries by smallholders (those with less than two hectares of farmland), and is commonly directly connected to family farming (Graeub et al., 2016). The FAO defines family farming as: “a means of organizing agricultural […] production which is managed and operated by a family and predominantly reliant on family capital and labor, […]” (FAO, 2013, p. 2). However, small-scale agriculture in Europe is not the same as in developing countries. In this paper, the consideration of Bavarian farms as “small-scale” agriculture is more due to structural composition (family farms, part-time operations, low degree of specialization) than to farm size. With an average size of around 35 ha of cropland, Bavarian farms are larger than those in many other regions in Europe, e.g., in specific regions in Poland, Slovenia, Romania, Greece, and Southern Italy (Eurostat, 2018). Still, the findings on the adoption of PA and digital technologies in Bavaria presented here can provide indications of how the digitalization process is progressing in other agricultural regions in Europe with similar structural characteristics.

Literature review

As digitalization in the agricultural sector has intensified, the number of studies investigating the adoption and use of digital and PA technologies in crop production and livestock farming has increased accordingly. Such studies focus both on farm characteristics and aspects of farmers’ behavior in adopting digital technologies (Klerkx et al., 2019). Due to the relatively long history of PA technologies, the number of studies of PA clearly outweighs those on more recently developed automatic and “smart” farming technologies (Tey & Brindal, 2021). The focus of the following literature review is limited to the influence of structural conditions on the adoption of technologies at both the country and farm levels.

The number of users of PA in developed countries varies widely between countries (Lowenberg-DeBoer & Erickson, 2019), and regions (Pfeiffer et al., 2021; Schimmelpfennig & Ebel, 2016; Llewellyn & Ouzman, 2014). In the European context, the highest user rates of digital farming technology are currently seen in Germany, France, the Netherlands, and the United Kingdom (Maloku, 2020), with the level of use in crop farming even higher than, e.g., in forestry or viticulture (Kernecker et al., 2020). In their twenty-year review of adoption studies, Lowenberg-DeBoer & Erickson (2019) demonstrated that the use of PA technologies in crop farming is significantly higher in the US, Australia, and South America than in Europe. Depending on the technology, European studies have shown adoption rate percentages mostly in the single-digits or low double-digits over the past twenty years. For example, study results list adoption percentages of up to 20% for GNSS-guided lightbar systems (in UK, Pickthall et al. 2017), 12% for yield mapping (in Germany; Lawson et al., 2011), and 8 to 14% for variable-rate fertilizer applications (Lowenberg-DeBoer & Erickson, 2019). For many years, the study by Reichardt & Jürgens (2009), who surveyed the user shares of PA field technologies among visitors of several agricultural trade shows, was considered authoritative for Germany. The authors found relatively high rates of adoption of GPS-based soil sampling or yield mapping methods (up to 40% of those surveyed), with lower adoption rates for variable-rate site application technologies (10–20%). However, the sample consisted mainly of trade show visitors, who are likely to be highly interested in investing in new technologies.

In European livestock farming, the focus of digital technology deployment is currently on farm management information systems (FMIS), on barn cameras and on sensors for measuring the behavior of individual animals. Adoption studies conducted in the United States (Borchers & Bewley, 2015), Switzerland (Groher et al., 2020), and Germany (Pfeiffer et al., 2021) have shown that data capturing technologies (barn cameras, animal behavior monitoring sensors) already play an important and increasing role in these countries. Robotics has also made its way into animal production, with the adoption of automatic or autonomous feeding, milking, and cleaning technologies beginning in the 1990s. The number of dairy farms with automatic milking systems (AMS) has increased significantly worldwide over the last two decades; with the Netherlands, France, and the Scandinavian countries leading the way in Europe (De Koning, 2010). Adoption rates vary regionally, depending on prevailing regional production intensity and operational structures (Eastwood & Renwick, 2020; Gargiulo et al., 2018). For example, the adoption rates of AMS are already at 30% in Iceland and Sweden, and at more than 20% in the Benelux countries, but only at the single-digit level in the UK (Vik et al., 2019). In the case of AMS, farmers are opting for the use of such systems due to increased working time flexibility (Straete et al., 2017) and accompanying improvements in animal welfare through better monitoring of livestock operations (Vik. et al., 2019; Latvala & Pyykkönen, 2005). Considering all available adoption rate surveys, Lowenberg-DeBoer & Erickson (2019) emphasized the problems of comparability between these studies, due to different methods of sampling and data analysis and the use of different terminologies to refer to specific digital technologies.

A large body of literature has investigated the factors influencing the adoption of digital technologies and the application of PA at the farm level. Some reviews have sought to provide an overview of this research by classifying potential effects of digital transformation and summarizing the empirical evidence in numerous individual studies (Shang et al., 2021; Pierpaoli, 2013; Tey & Brindal 2012). The meta-study by Pierpaoli et al., (2013) summarized the main influences on adoption of PA technologies among farmers gathered from studies conducted both before and after adoption. The authors identified farm size, expected benefits of the technologies by the farmers, and individual farm manager characteristics as critical factors in adoption. Farm size - measured as both absolute production field size and ratio of (main) crop land to total farmland – has been identified in many studies as a crucial factor influencing the adoption of multiple PA technologies (Walton et al., 2008; Robertson et al., 2012; Pierpaoli et al., 2013; Tamirat et al., 2017). The fact that adoption rates increase significantly from small to larger farms has been supported by several additional studies (Annosi et al., 2019; Daberkow & McBride 1998; Kernecker et al., 2020; Paustian & Theuvsen 2017). The relatively higher level of income generated by larger farms lowers (perceived) investment risk in newer technologies (Barnes et al., 2019) and provides greater access to capital for acquiring PA technologies (Tamirat et al., 2017). Further, adoption rates have been shown to be higher on owner-occupied farms (Paustian & Theuvsen 2017). Other studies indicate that farms with higher proportions of rented land are quicker to adopt new technologies (Daberkow & McBride 1998), which could be explained by tenant farmers’ efforts to maintain good long-term relationships with landlords by the use of future-proof technologies. However, a large proportion of adoption studies have concluded that the higher the share of farmer-owned land the greater the incentive for and thus, the probability, of technology adoption (Tey & Brindal, 2012: Isgin et al., 2008). The same argumentation leads to the conclusion that the nature of legal form of the farm operation itself can play a role in the adoption rate of new technologies (Paustian & Theuvsen, 2017; Moreno & Sunding 2005). Cooperative members, for instance benefits from better access to investment capital and lower investment risk for technologies (Barnes et al., 2019).

The role of the farm as the main source of income in willingness to adopt is somewhat unclear. Zheng et al., (2018) identified a higher willingness to adopt when the farm is the main source of household income, although off-farm income can offer additional funds for investments in digital technologies (Schimmelpfennig & Ebel, 2016). Likewise, investments are always associated with economic risks, which can have an inhibiting effect on decision-making - especially for part-time farmers and smaller farms. This is especially true if the advantage of the technology is not clear, or technologies are not yet perceived to be ready for the market (Kernecker et al., 2020). Consequently, there is growing concern that smaller farms will be left behind in digitalization (Rotz et al., 2019).

The degree, and type of specialization of agricultural production on individual farms can also have an impact on adoption rates of PA and digital technologies within a specific region. Specialization in certain crops or specific livestock are also potential influencers of adoption (Barnes et al., 2019), especially if the focus is on higher-value production in terms of high margin crops and products (Kernecker et al., 2020; Blackmore et al., 2003; Daberkow & McBride, 1998). Highly specialized farming systems do not require the same breadth of assistive technologies as highly diversified farming systems. Specialized farmers also benefit from longer and deeper knowledge of their production sector which can be conducive to the adoption of digital technologies (Asare et al., 2018; Paustian & Theuvsen, 2017).

Another driver of adoption is the use of complementary technologies. Combinations of multiple tools may be appealing to farmers when one technology builds on another (Isgin et al., 2008), or when technologies are in the same “technology bundle” (Griffin et al., 2017). Information from long-term farmer and machine trader panels is often used to assess adoption rates in the U.S. The world’s most long-term study format to investigate the use of digital farming technologies, the biennial U.S. Precision Agriculture Dealership Survey, asks crop input dealers about PA technologies and customer adoption of precision farming in both the current and the three upcoming cropping seasons (Erickson et al., 2017). Adoption curves for farms in the Kansas Farm Management Association (KFMA) were generated by asking them to specify both the year of adoption and time until abandonment of PA technologies (Griffin et al., 2017). Continuously gathered panel data also provides information on sequential adoption of multiple technologies. These data also help explain, e.g., farmer’s shift from the use of “embodied knowledge”-applications, such as automatic steering systems, to “information-intensive”-technologies, such as yield mapping or geo-referenced soil sampling applications. The panel information shows that these changes happen very slowly over the long term (Miller et al., 2019; Fernandez-Cornejo et al., 2001). Schimmelpfennig & Ebel, 2016) adopted a similar approach, using data from U.S. farmers to examine both the sequential use of several technologies, and the associated cost effects of specific combinations of technologies. Another recent paper goes a step further, and extracts PA adopter categories from USDA farm data, and compares their economic efficiency gains through the use of PA bundles (DeLay et al., 2020). Applied stochastic analysis shows that farms in the early adoption stages are more efficient than non-adaptors due to the use of PA, but are usually also larger in size. Looking at the European situation, Barnes et al., (2019) studied the uptake of automatic machine guidance, which also includes automatic steering, in five different European countries. The authors also considered farm size as significant indicator between adoption or non-adoption of this PA technology considered.

A significant number of adoption studies used binary indicators to measure farmers’ decision to adopt particular technologies, the tendency to combine various technologies, and the frequency, and use intensity of specific PA technologies (Tey & Brindal, 2012). Logit or probit regression analysis with dichotomous dependent variables were used in the first case, and tobit models for the latter. This methodological approach is widely used to test predictors for adoption but is not free of criticism. Weersink & Fulton (2020) argued against a binary understanding of adoption, proposing instead an examination of the adoption process that recognizes multiple stages during which economic, social, and attitudinal influences need to be weighted differently depending on the adoption stage. Montes de Oca Munguia & Llewellyn (2020) argued against a strong a focus on farmer and farm characteristics, and a lack of consideration of the performance of the technology itself in adoption studies. They suggested replacing binary measures of adoption with more interactive factors combining farmer preferences and technology characteristics. A recent meta-analysis of PA adoption highlighted how different study designs and statistical methods used in binary approaches hamper the comparability of study results (Tey & Brindal, 2021). An Australian researcher group went beyond binary measures to develop the ADOPT forecasting model that quantifies the speed and peak level of technology diffusion at the farm level (Kuehne et al., 2017). The ADOPT model is composed of multiple variables that consider both characteristics of the user population and of the technology itself. Many of those model variables reflect classic attributes of innovation diffusion theory (for instance, relative advantage of an innovation, compatibility with other applications, complexity, trialability and observability of the technology for the user population) (Rogers, 2003). It is, therefore, worth investigating the adoption rate of individual technologies in a certain population on the basis of these product characteristics.

Method and data

Implementation of the farmer online survey

An online survey of Bavarian farmers provided the data used here to investigate the current level of use of digital technologies on Bavarian farms. Questionnaire structure, selected technologies and question forms were initially compiled based on questionnaires from farmer surveys of smart technology and PA adoption in the U.S., UK, and Denmark (Ericksen et al., 2017; DEFRA, 2020, Statistics 2018), and were subsequently adapted to fit the German context. An initial revision of the questionnaire concept was carried out with experts from agricultural research, extension, and government agencies. A pre-test employing a quoted sampling procedure was carried out in Germany at the end of 2019 (n = 591) (Gabriel et al., 2021). The extended pre-test was relevant for identifying the terminology for specific technologies most familiar to farmers, among other things. Studies in German-speaking countries have shown that terminology has an influence on the understanding of digitalization in agriculture (Reissig, 2020), and unclear formulations can impact farmers’ understanding and assessment of technologies (Schukat, 2019). The pre-test findings regarding comprehensibility and technical implementation were used to adjust the questionnaire for application in the main survey to be conducted in Bavaria. The final questionnaire comprised seven question groups: (1) socio-demographics and firmographics; (2) current use of and future plans to invest in 30 PA and digital technologies; (3) general barriers to and attitudes towards digitalization in farming; (4) perceived problems and perceived benefits of specific technologies; (5) issues of data security and data access; and (6) information sources and services. The structure of question groups and questions was adaptive, in that it was conditioned by previous answers, in order to ensure a compact and user-friendly questionnaire. Farmers who specialize in livestock farming, for example, were asked only about technologies relevant to this sector during the survey process.

The target population of 103,552 Bavarian applicants were accessed via the support and funding platform of the Bavarian State Ministry of Food, Agriculture and Forestry (StMELF), which is available to all farmers online. Applications for EU agricultural subsidies in Bavaria must be submitted via this platform during a fixed time period each year. The application period in 2020 took place from 12th March to 9th June (including a three-week extension period). The link to the survey was prominently displayed in the application portal, together with introductory information explaining the purpose of the questionnaire. Thus, every Bavarian farmer applying for EU subsidies in 2020 had the option to participate in the survey. Additional dissemination of the survey link via other media was omitted in order not to bias the probability of access. The survey system (LimeSurvey V3.22, Hamburg, Germany) ensured that only one response was submitted by each individual study participant using the same device.

A total of 3,739 participants started the questionnaire, of which 2,458 completed each of the seven consecutive groups of questions. The dropout rate was thus 34%. The median time needed to complete the questionnaire was 18 and a half minutes, a user-friendly time frame for an online survey (Revilla & Ochoa, 2017). The dataset of completed responses was checked for plausibility (answer consistency) and quality (total response time). A final set of 2,390 completed responses was used for frequency analysis with SPSS Statistics 26 (IBM, Armonk, NY, USA).

Conceptual framework of measuring multiple usage and sequential adoption patterns

A total list of 30 technologies was used in the survey. A subset of this list was shown to each survey participant. The subset was individually adapted based on their responses to initial questions identifying the crop production and animal categories in which they are currently active (Fig. 1). From this list, farmers indicated which technologies they were currently using. The first technology that was purchased and used on the farm is considered to be the entry technology. For those technologies that had not yet been adopted on the farm, farmers were asked whether they could see themselves investing in them within a year (short term) or within five years (mid-term). Based on this data, it was possible to determine multiple usage of technologies already exist or are likely to exist in the future at the farm level. While actual future investment decisions cannot be inferred from the statements, farmers’ indications of willingness to engage with certain technologies can be seen as potential readiness to adopt. Figure 1 provides an overview of the conceptual model of the user rate survey including the four sequential points of adoption. While the methodological approach follows the engineering method and there is no hypothesis testing, it is still important to validate the empirically measured frequencies to avoid the occurrence of data noise that might affect interpretation. Bootstrap methods for statistical validation of frequency probabilities (e.g., technology user onboarding rates) or differences in mean rates between user and non-user usage rates helped determine the 95% confidence intervals of the means.

Fig. 1
figure 1

Concept model of variables investigated

Full size image

Sample description and evaluation

The total population of Bavarian farmers who were eligible to apply for EU funding in 2020 had the opportunity to participate in the survey. Assessment of the representativeness of this “convenient random sample” takes place post hoc, through comparison of operational characteristics and socio-demographic attributes of the surveyed farm managers with available statistics for the population. Based on this comparison, the actual distribution of Bavarian farms surveyed is well represented in the sample in terms of legal form and types of land ownership (Table 1).

With more than 100,000 farms producing on an average of 35 ha farmland, Bavaria - the south-eastern federal state of Germany - contains almost one third of the farms in Germany (StMELF, 2020). A comparison of data on farm structure in Bavaria with that from other federal states in Germany shows the smaller structures in Bavaria in contrast to other states, especially those in the northern and eastern parts of Germany. In neighboring state Thuringia, for example, 3,500 (mostly corporate) farms operate an average of 220 hectares of farmland (Destatis, 2018). Another characteristic of smaller agricultural structures is a low degree of specialization, with many farms operating in multiple agricultural production sectors (e.g., crop farming, animal feed production / grazing, livestock farming). Energy production, forestry, services provided to other farmers or municipalities, processing and direct marketing of agricultural products, and agro-tourism provide additional revenues for 53% of Bavarian farms (Destatis, 2021). Finally, about two thirds of farmer households in Bavaria have at least one other source of income from salary or wage work outside the farm (Destatis, 2021).

Table 1 Survey sample description (n = 2,390)
Full size table

With regard to legal form and land ownership, Bavarian agriculture is still dominated by sole proprietorships (94%, Destatis, 2018), with 70% of farms having leasehold shares below 50% (data from 2013; StMELF, 2020). The sample contains a significantly higher share of full-time farmers (48%) than official statistics for Bavaria indicate (38%; StMELF, 2020). The type of operation plays a crucial role in farmers’ motivation to use specific technologies (Mittenzwei & Mann, 2017). Organic farm managers also participated in the survey proportionally more often than the actual share of organic farms in Bavaria (9%). The proportion of livestock farmers (cattle, pigs, poultry, laying hens) in the sample is 57% and significantly lower than in the official statistics (71%; Destatis, 2018). Regarding average size of the farms surveyed, there are significantly more larger farms in the sample than in the official statistics for Bavaria. Farm size was indirectly queried in the survey via the indication of land used for crop farming (average of 56 ha), feed production (40 ha) and herd size (46 cows on average for the dairy farms surveyed). Current Statistics for Bavaria indicate an average farm size of 35 ha production land and herd sizes of about 40 cows for dairy producers (StMELF, 2020).

Socio-demographic characteristics of the surveyed farm managers approximate the actual situation in the largely family-run farms in Bavaria (Table 1). The distributions of age classes, gender, and level of agricultural education in the sample deviate only slightly from the most recent data from the statistics on the Bavarian agricultural sector (Destatis, 2018). The sample also reflects the dominance of male operators on Bavarian farms at more than 90%, while the distribution of age categories of farm managers differed slightly from official statistics. In particular, the two highest age classes (50–59 years; 60 years and older) are much more strongly represented in the sample than they are in the population, while the age class of 40 to 49 years is proportionately underrepresented.

One way to evaluate the representativeness of the sample is to show that the farm structures and socio-demographic characteristics are similar to those in the official statistics. It should be noted, however, that random samples can be further biased in ways not captured in official statistics (e.g., methodological errors, unknown patterns in the overall population). One potential source of bias in the survey approach implemented here may be that a large number of applications were filled out by service providers (e.g., tax consultants, agricultural support organizations), and the farmers themselves did not have access to the survey link on the application platform. However, the official documentation of the application process shows that, in 2020, only 7.9% of the applications received for Bavarian farms were submitted by service providers (personal communication, Bavarian State Ministry of Food, Agriculture and Forestry). Furthermore, it can be expected that a higher number of younger farmers would voluntarily complete an online survey (Blasius & Brandt, 2010). However, a comparison of the distribution of age classes in the sample compared to that of the population from official statistics shows that especially older farmers (50–59 years, 60 years and older) are overrepresented in the sample. The fact that every agricultural enterprise in Bavaria is eligible for EU funding, and that applications are submitted exclusively online shows that an overriding bias of the sampling procedure due to its online nature is unlikely. Another indication of the strength of the sampling method lies in the fact that responses from all production areas and farm sizes in Bavaria were received. Farms under a certain minimum size were not excluded in advance, as is sometimes done in other surveys conducted by the private sector (e.g., DEFRA 2020; Rohleder et al., 2020).

Results

Current use of digital technologies

The Bavarian farmers surveyed were provided with a list of available PA and digital technologies for agricultural purposes and asked about their current use or non-use of each, and their intention to invest in each given technology in the short-term (within one year) and/or the mid-term (within the next five years) (Fig. 2). Based on a complete list of 30, the technologies offered to each respondent were tailored based on previous questions pertaining to the production sectors in which they actually practice. Regardless of production sector, all respondents were asked about their use of forecast models and apps (e.g., weather and pest pressure forecast models), and online communication and trading platforms. These two categories had the highest rates of reported adoption of all groups of technologies − 23% and 38%, respectively. High numbers of users of these, mostly free or inexpensive, tools suggests that easy access to digital technologies also increases the adoption rate. Of the respondents involved in crop farming, the most frequent rates of reported purchase or investment in technologies were as follows: digital field records (21%), automatic steering systems (17%), maps from satellite data (14%), and farm management information systems (13%). However, respondents also noted that, despite investments in these technologies, not all of them are actually in use on the farm. Section control (13%) and lightbar guidance systems (10%) are two additional PA technologies that showed relatively high adoption rates by crop farmers in the sample. Further, up to 11% of farmers surveyed expressed interest in investing within the next five years in those technologies. In general, digital technologies that have prevailed in farming so far are those that are particularly beneficial to quality of farm work and reduction of workload, rather than those that yield positive environmental effects.

One in six of the Bavarian livestock farmers surveyed currently had digital farm-management information systems in use. Barn cameras are currently used by 17%, and 12% already use sensors for behavior monitoring (e.g., pedometers for activity recording). Of the dairy farms surveyed, 15% currently use automatic milking systems (AMS), while automatic (robotic) slat cleaners and feeding robots are each used by 7% of respondents. These low adoption rates are not surprising, as low average herd size among the farmers surveyed indicate that such automation technologies are not yet cost-effective for many of these farms. Still, about 10% of the livestock farmers surveyed reported having plans to invest in cost-intensive robotic barn technologies. Considering all 30 of the digital applications and PA technologies surveyed, their current use on Bavarian farms is quite limited, although respondents stated a strong intention to invest in most of the technologies queried. The overall rate of technology adoption (proportion of farmers using at least one of the technologies in the survey) was 62% for the total sample.

Fig. 2
figure 2

Current use of and planned investment in 30 PA and digital technologies

Full size image

Sequential adoption patterns and joint usage of technologies

The data from the farmer survey covers different points on the timeline of technology adoption. This includes information on the entry technology (the first technology purchased and used on a farm), the current use of additional digital technologies, and two potential future adoption times (based on the intention to invest within one year and/or within five years) (Table 2).

Table 2 Selection of PA and digital technologies and rates as entry technology
Full size table

Forecast models and apps are considered to be easily accessible digital technologies. These models and apps provide basic agricultural weather forecasts, as well as more specific recommendations for optimal timing of pesticide application. Such applications are also used very frequently, as they are also the entry technology for more than half of the users. In crop farming, digital field records, lightbar systems, and automatic steering systems were the most frequently cited entry technologies. Similar to automatic steering systems and lightbar systems, the focus of digital field records is on facilitating farm work without overburdening the farmer with complex information-intensive technologies (Miller et al., 2019). Many livestock farmers in the sample reported making the first step towards digitalization with barn cameras and AMS, while farm-management information systems such as herd management software and animal behavior monitoring sensors were not reported as common points of entry. For more than half of the AMS users in the sample, this very cost-intensive digital technology was the first introduced on the farm. Studies from Denmark and Holland identify important thresholds with regard to time after investment and number of cows (45) above which AMS become more profitable than conventional milking systems (Hansen et al., 2019; Floridi et al., 2013). The average number of 48 cows per farm in the present sample seems to indicate the relatively high acceptance of this technology among Bavarian farmers.

In for a penny, in for a pound - sequential adoption based on selected entry technologies

Overall figures for the 30 digital and PA technologies examined in this study show that the adoption rates of digital technologies are low on Bavarian farms in comparison to those in large-scale agricultural regions worldwide (Lowenberg-DeBoer & Erickson, 2019). In particular, the adoption of PA and digital technologies is a slow and long-term process, and the tendency to stick with the status quo in terms of technical equipment is strong among farmers (Miller et al., 2019). Experience with an entry technology can support farmers in getting used to digitalization and increase the likelihood that they will move from being entry-level to more advanced adopters (Schimmelpfennig & Ebel, 2016). Based on a selection of technologies that were often cited as entry technologies in the sample, the level of digitization due to the addition of further on-farm technologies was calculated (Table 3).

Joint usages were identified at both the single and multiple levels for selected entry technologies showing noticeable differences. For example, ‘none’-combinations (the sole use of the entry technology) varied from 6% (for AMS) to 50% (for forecast models). Forecast models and apps are also most often referred to at the ‘single’-category, describing the presence of a second technology on the farm. Digital field records, automatic steering systems, and AMS have high percentages of ‘multiple (4+)’- combinations, with more than one third of all users of these technologies in the sample reported having four or more other technologies in use on their farms. Dairy farmers who use the cost-intensive AMS technology on their farms usually also use additional technologies that are compatible with automated milking systems, such as sensors or farm-management software. Two PA technologies that are highly similar to one another - automatic steering systems and lightbar system - both show very high levels of combinations with additional technologies when first deployed. Thus, starting from the rather easy-to-use (“knowledge-embedded”) PA technologies, farmers are expanding their use of digital technologies.

Table 3 Selected entry technologies with number (n) and share (in percent) of respondents currently using them, number of additional technologies used, and adoption rate of selected technology
Full size table

Pull effects – current and potential usage of further technologies

In addition to questions concerning current use of technologies, Bavarian farmers were also asked about planning investments in technologies within one year and within the next five years. This makes it possible to make forecasts about the frequencies of additional technologies in use at two points in the future. Table 4 shows proportions of users and non-users for a selection of frequently adopted PA and digital technologies in the subsample of crop farmers. For each of these technologies, user rates for an additional technology were compared to those of non-users. This was done for three points in time: at the time of the survey, and two specified future points in time - (one year) and (five years). Taking the first line as an example, half of the users of digital field records also have FMIS for crop farming in use (user rate: 0.50). In contrast, a rate of only 0.03 of non-users of digital field records use FMIS. Regarding the planned investment in FMIS for crop farming as a second technology within the next year, there are no differences in the rates between the users and non-users of digital field records (both 0.03). Yet, there is a statistically significant difference between the number of non-users of digital field records (user rate of.10) and the number of current users of digital field records (user rate of 0.05) who plan to purchase FMIS within the next five years. In this specific case, this can be explained by the fact that many FMIS for crop farming now include some functions commonly offered by digital field record products.

Table 4 Share (in percent) of sampled farmers who currently use a given technology in crop farming, rates of users and non-users of that technology who currently use a specified additional (2nd ) digital technology, and rates of users and non-users who plan to acquire the specified (2nd ) technology in the future
Full size table

It turns out that a handful of PA and digital technologies form the most common joint usage combinations. The decisive factor here is which user group serves as the starting point. There is a higher user rate of using automatic steering system by users of section control (0.67) than vice versa (0.51). A comparison of the current usage of a second technology between non-users and users of a given technology reveals statistically significant differences in adoption rates. This again indicates that the actual use of PA and digital technologies on-farm encourages the adoption of other technologies. Farmers who have implemented PA technologies such as automatic steering systems and lightbar systems, often use them in combination with section control or digital field records (0.40 - rates for each). It is notable that the most frequent rates of joint usage occur between technologies that some authors have classified as “embodied technologies”, or those that can be used without further application skills (Miller et al., 2019; Lambert et al., 2004). Similar studies have also shown the dominance of these technologies over information-intensive PA technologies, such as variable-rate application technologies (Griffin et al., 2017; Schimmelpfennig & Ebel, 2016; Barnes et al., 2019). Variable-rate fertilizer - nitrogen technology is currently used in particular with maps of satellite data in the sample (0.19). The use of satellite maps is one main requirement for the use of variable-rate applications. The investment plans of survey respondents suggest that the use of variable-rate technologies is going to increase in the next few years among these farmers (up to 0.37).

With regard to the future planning of all used technologies considered, it is apparent that increased user rates for satellite maps and section control technologies as further technology can be expected at least in the medium term (within 5 years). For non-users of some individual technologies, the comparison with users shows that digital farm records, in particular, are being more often considered as a possible future use. This is stated, for example, by the non-users of FMIS, automatic steering systems and section control. Based on the survey responses, adoption of PA technology automatic steering systems by current lightbar system users will increase by nearly 30% in the next five years. Conversely, only 4% of automatic steering adopters are considering investments in lightbar systems. This indicates a shift towards automatic steering systems. However, it must be emphasized that the any predictions from the data are based purely on the point of view of the farmers at the time of the survey.

Joint usage of more than one digital technologies in livestock farming provides a similar picture (Table 5). In that the central role of automatic milking systems (AMS) for dairy farm management is readily apparent. In many cases, AMS adopters also use behavior monitoring sensors (0.80), barn cameras (0.58), and farm management information systems (0.57). Regarding additional barn robots (slat cleaner, feed pushing), high rates of planned investment (around 0.80) are expected in the next five years. Combinations of farm management information systems for livestock farming are also used with digital crop farming technologies (farm management information systems – crop farming) on farms operating in multiple production sectors. Similar to crop farming technologies, it is evident across all livestock farming technologies used shown in Table 5 that they also adopt other automated and digital technologies more frequently than non-users. Automated technologies such as robotic slat cleaners and robotic feed pushers are used most frequently by users of AMS and behavioral monitoring sensors (for both rates around 0.30). In mid-term planning, such automated technologies are also attracting increased attention from these two user groups. Joint use of FMIS, sensors and barn cameras for control and decision making is evident in many livestock farms in the sample.

Table 5 Share (in percent) of sampled farmers who currently use a given technology in livestock farming, rates of users and non-users of that technology who currently use a specified additional (2nd ) digital technology, and rates of users and non-users who plan to acquire the specified (2nd ) technology in the future
Full size table

Conclusions

The heterogeneous and small-scale structure of Bavarian agriculture is characterized by a majority of small and family-managed farms, a diversity of production sectors, and different types of farm organization (e.g., part-time operators). This heterogeneity makes it especially difficult to evaluate the adoption of digital and PA technologies, and to forecast future developments. The aim of this research was first, to assess the current situation with regard to entry technologies, and joint usage of multiple digital technologies in both crop farming and animal husbandry in Bavaria. In addition, survey results allow us to draw conclusions about potential developments and trends in sequential steps of digitalization. Through the administration of an online farmer survey, reliable data on the use of digital technologies on farms in this small-scale agricultural region in Europe was collected for the first time. In total, farmers were questioned about 30 digital technologies, including PA, automation technologies and digital “smart farming” applications. Adoption rates for all technologies sampled lag clearly behind those in many other large-scale agricultural regions worldwide (Lowenberg-DeBoer & Erickson, 2019).

Results show that Bavarian farmers cannot be described as exceedingly digitalized. Apart from smart technologies such as communication and trading platforms or forecast apps, adoption rates are rarely in the double digits. In the livestock sector, where there are more full-time farms, the adoption rates of barn cameras, farm management information systems and behavioral monitoring sensors do not exceed 17%. Some digital and PA technologies, such as barn robotics, section control, variable-rate applications, and maps from satellite data show potential adoption rates of 15–20% within the next five years.

Digital field records and automatic steering systems as well as automatic milking systems in dairy farming were shown to be typical and frequent entry technologies which are rarely used exclusively, but rather, often combined with additional technologies. In many cases, the potential for medium-term acquisitions by non-users of individual technologies is significantly lower. This suggests that the use of individual digital technologies is an accelerator to invest in other technologies as well. Current or future combined adoption of matching technologies seems understandable. The frequent combined use of automatic steering, lightbar systems or section control, as simple PA technologies, also shows that farmers primarily prefer easy-to-use technologies. Crop farmers tend to stay at their current technology level (“embodied knowledge technologies”) suggesting a low probability that they will adopt information-intensive technologies, as has already been shown in previous studies (Miller et al., 2019; Erickson et al., 2017). Complexity describes the degree to which an innovative technology is perceived as difficult to understand and to use (Rogers, 2003). The complexity of “information-intensive” technologies, such as variable-rate applications, NIR-sensors and soil sensor systems are not (yet) as accepted by Bavarian farmers. Thus, the level of on-farm digitalization is still a question of farm size and type of operation. More than the half of farms in the sample are operated by part-time farmers. Not operating a farm as the main occupation means less motivation, access to capital, and time to invest in specific technologies (Mittenzwei & Mann, 2017). Detailed information on the financial possibilities of Bavarian part-time farmers was not included in this survey but should be considered in future research.

However, the incentive for farms in a small-scale context to digitalize is greater when the relative advantage of a specific technology becomes apparent. Thus, adoption rates of individual technologies can rise sharply in the future if a quick return on investment of a digital and PA technology can be achieved or the necessity of a technical change is reinforced externally, e.g., by changing regulatory frameworks for agricultural production. In Bavaria, for example, forms of cow tethering are still legal, although the political debate about it is intensifying due to growing social pressure. Most automated barn technologies are not designed for tethering systems. A legal prohibition of this housing system could potentially promote the spread of automation systems. The subsample of livestock farmers in this survey shows currently high combination rates of AMS, behavior monitoring sensors, and barn cameras. Barn robotics for feeding and cleaning are digital technologies in which livestock farmers are demonstrating a growing active interest. In crop farming, farmers’ interest in resource-saving technologies will increase if, for example, the use of pesticides or fertilizers is further limited due to legal requirements. Technologies such as variable rate applications, automated equipment and field robotics are all technologies that can help farmers meet societal demands to lower chemical inputs.

It is true that global agriculture is very advanced in terms of digitalization and can certainly compete with other sectors of the economy. However, it must also be considered that the degree of digitalization of farms, which is sometimes communicated as being very high also in other large-scale regions in the world, is driven by individual applications. Also, the terminology used for each technology in adoption studies and publication of results is also critical (Reissig, 2020). If one calculates the number of farms that use any digital technologies at all, the results are quite consistent with a recent industry survey in Germany (Rohleder et al., 2020) which arrived at the following conclusion: “8 out of 10 farmers use digital technologies”. However, if both categories “Communication and trading platforms” and “Forecast models” are excluded, the degree of digitalization on the farms surveyed here decreases by more than 10% points. Therefore, the results of the current user rates and the sequential digitalization presented in this paper show that the adoption of digital technologies in agricultural practice depends heavily on a specific application. It should be noted that the role of PA technologies does not stand out directly from other digital and smart farming technologies. Rather, it is individual product attributes that have been the success factor in market diffusion to date. Among the most used technologies in the farmer sample, the current focus is clearly on user-friendly automation solutions that reduce the workload.

A heuristic approach was deployed to identify patterns of adoption trends in small-scale agriculture at the farm level, identifying the current equipment and technology trends are also of great importance at the political level. For Bavaria in particular, there is currently a specific funding program to support the digitalization of agriculture so that family farms can also benefit from the opportunities offered by digital technologies. The Bavarian program to foster digital agriculture, which was launched by the Bavarian State Ministry of Food, Agriculture and Forestry in October 2018, provides financial support to farmers for investing in specific digital technologies. The funding program is intended to give innovative and sustainable technologies, some of which are still associated with a high investment risk, an initial boost to accelerate their implementation in agricultural practice. As the funding program covers specific areas of PA and digital technologies, it is important to have up-to-date information about the current adoption rates of these technologies. This makes it possible to target funding towards technologies that address the most relevant issues in the societal discussion on good farming practice (e.g., animal welfare, reduction of pesticide use), and improve the conservation of resources and the environment (e.g., through NIR sensor systems, variable-rate applications, field robotics).