A MIS developed for orchard management can be expected to address major tasks in commercial orchard operation, such as the management of irrigation
[8][9], pests (weeds, insects and disease)
[10][11], plant nutrition, manipulation of tree physiology and structure, e.g., through pruning and phytohormone applications
[12][13], and harvest, e.g., planning harvest labour, packing and marketing needs
[14]. Many farm MIS technical features and adoption drivers and barriers will be common across all agricultural production systems, with greater commonality expected within the plant production systems of perennial and annual horticultural production and broadacre cropping than with intensive and extensive animal production systems. For example, a report on the use of system analysis methods to define the design requirements of a farm MIS
[6], explanations for the lack of success in the adoption of a computer-based MIS in farming systems
[15], and a review of irrigation MISs
[16], all have relevance to tree-fruit production systems.
2. Adoption Barriers and Drivers
Despite their promise to increase the efficiency of farm operations, farm MIS adoption has been less than expected, with MIS technology described as ‘under-utilized’ in agricultural production
[1][19]. The major barriers to farm MIS adoption identified in previous studies
[1][19][20][21][22][23][24] include:
- (i)
-
‘Internal’ operational issues, including system ‘bugs’, feature incompleteness, poor user experience through poor wording of logic flow, the effort required for data entry, and poor system value as manifested in a low integration of system output into decision-making;
- (ii)
-
‘External’ operational issues, including poor system reactivity caused by limited internet connection, poor input data quality and poor lack of system integration and interoperability through failure to use standardised data formats;
- (iii)
-
System maintenance issues, including low adaptation rates, high costs and poor user support resources;
- (iv)
-
Lack of trust in system reliability and data security;
- (v)
-
Affordability.
Similar operational challenges were reported in the reviewed orchard MIS papers, including:
- (i)
-
‘Internal’ operational issues, including e-device or platform compatibility, scalability and performance/efficiency
[25][26][27];
- (ii)
-
‘External’ operational issues, including connectivity and adequacy of bandwidths
[27] and poor integration of third-party services and new technologies
[28];
- (iii)
-
System maintenance issues, such as software features and poor support services and training
[29];
- (iv)
-
Data privacy and trust
[25][26][27];
- (v)
-
Affordability and user willingness to pay for the services
[30][31][32][33].
The issues noted above are often inherent in system design, resulting from a mismatch of designer and developer focus and the requirements of the ultimate users of the system. As with any technology adoption, there exists a scale-up problem between the proof-of-principle stage and the offer of a user-friendly, cost-effective product to a large user base.
MIS adoption is also a function of the client base. For example, a relationship was noted between the ‘technical efficiency’ of Brazilian citrus farms and their adoption of a MIS, with MIS users producing greater output using the same level of inputs
[34]. The perceived benefits of orchard MIS use are related to increased efficiency in planning and the coordination and monitoring of production, particularly in larger-scale enterprises.
3. Orchard MIS Evaluation
An orchard MIS is a software artefact. While some quantitative performance metrics can be reported, e.g., the speed of data rendering and accuracy of forecasts based on models, the value of such a system is ‘in the eye of the beholder’, i.e., the user. User evaluation is, therefore, required. Many studies reporting evaluation use a survey tool. However, attention is required to the goal of the survey goal. Questions such as ‘which of the following features were most useful?’ may help guide developer efforts but do not provide information on the product’s usability and, thus, likely uptake. For example, a web-based agro-ecological monitoring system was evaluated through a user survey to check the satisfaction of registered users, with 95.8% of the users reported agreeing that “the pest density forecasting service helped to prevent pest outbreak”
[33]. The willingness of users to adopt (and pay) for the use of the system is a separate question.
To assess the likely uptake, questions can probe the relative performance of the orchard MIS to an existing management approach or probe the economic value achieved in the use of the MIS. For example, a quantitative approach was described for a user satisfaction survey on a decision support system for integrated pest management in an apple orchard, involving a comparison of the decision support recommendation to that of a domain expert
[35]. The decision support system agreed with the domain expert in all 100 cases of spray recommendations and in 95% of 283 cases of disease diagnosis. In another example, the Washington State University-Decision Aid System (WSU-DAS) was evaluated using a web-based user survey
[27]. It was reported that the tool had achieved a high market penetration in the area of integrated pest management (IPM) due to the provision of time-sensitive information, e.g., recommendations on the timing of insecticide application to achieve disruption on insect mating with fewer applications. Users were asked to estimate the economic value gained in the use of WSU-DAS from a reduction in the number of chemical sprays
[11]. Adoption of such questions in the evaluation of future orchard MIS developments is recommended.
4. Technological Features
Technological features, including development tools, were described in 31% of all papers (
Table 1 and
Table 2), with a useful review provided by Kaloxylos et al.
[36] on the architecture of an FMIS and the use of generic software modules in building farm-specialised systems. Decadal progression in technical features can be summarised in terms of (i) the use of personal computers from the 1990s, (ii) the use of WebGIS mapping and remote sensing from the 2000s, (iii) the use of mobile phones for data entry and LoRa-enabled sensors from the 2010s, and (iv) the use of IoT sensors, Lidar, UAVs and machine learning in image analysis in the 2020s. Twelve features were identified within the MISs (
Table 1 and
Table 2). Less common features, such as a search engine, location-based services (LBS), WebGIS-based mapping, notification and alert messaging systems, e.g., web/email/SMS, add to the farm usability of a MIS. These features can be expected to be added to commercial products intended for orchard management use, as opposed to the systems reported in publications, which tend to focus on the demonstration of proof-of-concept for the management of a particular task. Likewise, software features such as RESTful APIs, which allow for the progressive development of an application, are more important in a commercial product than a proof-of-concept project.
Table 1. Technological features utilised within orchard MISs: code key (for use with Table 2).
Table 2. Technological features utilised within orchard MISs: categorisation of research papers.
In concert with the evolution from static to dynamic and onto progressive web app delivery of the orchard MIS, there has been an evolution in the development tools used (Table 3). Of the publications reporting on an orchard MIS as a web application, the majority used HTML/CSS and JavaScript for the development of a front-end graphical user interface (GUI) (Table 3). PHP was popular for backend development (including RESTful APIs) along with Java, .Net and Folium/Django (Python). Most developers used MySQL for database management, followed by PostgreSQL/PostGIS, MS SQL Server and Oracle. Google and ESRI platforms were predominantly used for web mapping and WebGIS app development, with some use of other platforms, e.g., GeoServer/OpenLayers and OpenStreetMap.
Table 3. Development tools and technologies used in MIS-related publications.
The implementation of sensor and control networks on orchards has been constrained by communication capability. The most implemented orchard-wide management tool has been for irrigation management, involving soil moisture monitoring and irrigation valve control. This has typically involved proprietary radio communications. LoRa and related technologies such as LoRaWAN and SigFox have enabled low bandwidth applications, such as temperature monitoring across orchards, although line-of-sight requirements require the use of elevated aerials and repeater stations. To date, the coverage of high-bandwidth 4G and 5G networks across (Australian) orchards has been limited, given their location in low-population-density areas. The advent of lower-cost satellite-based communication systems such as StarLinkTM (Redmond, Washington, DC, USA), which operates in both the 2.4 GHz and 5 GHz frequency spectrums, opens possibilities for high-bandwidth orchard applications. Competitor systems are expected to expand during the next decade, potentially further lowering the cost. Another relatively recent development is the deployment of mesh-wifi across farms, based on 900 MHz, 2.4 GHz and 5 GHz frequencies in Australia. Which of these high-bandwidth applications will eventually dominate in-orchard use will reflect the reliability and cost of the technologies. It is clear, however, that their availability will lead to an increase in the use of control actuators and sensors, particularly cameras, across orchards operating through orchard MISs.
5. Management Aims
5.1. Plant Health
Publications in the ‘Plant health’ category commonly involved the monitoring of environmental parameters in the context of pest, disease and weed risks, e.g., the use of wireless sensor networks for near real-time environmental monitoring of environmental parameters in the context of disease risk, with data storage in a database system and visualisation through a web-based system
[32][70]. Other work involved the development of models for the management of pests and disease
[71][72] or weed invasion risks
[73]. These management tools were generally intended for use at a farm level; however, some papers reported on tools intended for regional use, for use by a catchment or government agency.
5.2. Irrigation
Publications in the ‘Irrigation’ category dealt with the development of MISs around water stress assessment and irrigation management. For example, a multi-modal sensor system for plant water stress assessment was developed in the context of an irrigation decision support system for an apple orchard
[7]. Several authors report on the design and development of a generic MIS for the irrigation management of tree crops
[8][26]. Other examples of models that could be used within a farm MIS include the CropSyst model for the irrigation management of pear orchards
[74], while apple-specific irrigation management systems were developed
[7][41] and implemented
[8][26]. Similarly, a system (‘Irriman’) was developed for use across an irrigation community based on input from a regional array of sensors
[9][75].
5.3. Nutrition
Publications in the ‘Nutrition’ category dealt with the in-field assessment of nutritional status and fertiliser use. For example, a mobile application was developed for the assessment of the nutritional status of apple trees
[76], a web-based fertiliser information system was developed for banana production
[44], and a web-based expert system was developed for the fertilisation of orange trees
[63].
5.4. Plant Development
Publications in the ‘Plant development’ category dealt with a selection of germplasm/cultivars, planting issues, prediction of growth, flowering or fruiting, and thinning requirements. Several publications did not target orchard management per se but rather the management tasks in the other areas of the value chain, e.g., bayberry breeding
[45]. Examples of models that could be used within an orchard MIS include a method for yield estimation of apple trees
[77], a carbon balance model for use in guiding apple tree thinning
[13], a model for apple and pear growth
[42][78], apple and orange yield estimations
[14][43][74], an agent-based decision support system for mango flower initiation
[79] and a tool for the forward estimation of mango harvest timing
[80]. The most comprehensive management system was that of Jianwei et al.
[81], who applied research on plant growth modelling within a framework to enable management decisions on, e.g., pruning, irrigation, fertilisation, yield prediction and cultivation.
Publications addressing harvest management were categorised as related to the estimation of fruit maturity for the timing of harvest, estimation of on-tree fruit load or the management activity during the harvest. These reports document the development of sensors and methods, with a number of studies also exploring the presentation of data to users. In particular, methods based on machine vision are rapidly developing for a direct assessment of tree-fruit attributes, underpinning the forecast of crop timing and loads. Indeed, a number of machine vision-based commercial systems have recently become available for in-orchard use, as reviewed by Anderson et al.
[18].
The timing of the ‘decision to pick’ requires information on fruit maturation. This can be estimated based on the heat unit requirements between flowering and fruit harvest maturity, e.g., Sousa et al.
[82] and Amaral et al.
[83], given the measurements of orchard temperatures. Depending on the commodity, fruit maturity can also be assessed based on the level of storage reserves, fruit skin colour, flesh colour and/or fruit shape. For example, the classification of melon ripeness/maturity from field imagery of the fruit surface was reported to support the estimation of optimum harvest timing
[14]. The development of portable, non-invasive assessment technologies has enabled the inclusion of internal attribute specifications into farm management systems. For example, a hand-held device for the assessment of the ratio of absorbances at two wavelengths was used to assess flesh pigment levels in fruit
[84], and the use of hand-held near-infrared spectrometry has been recommended for the estimation of fruit storage reserve levels as a measure of fruit maturity
[85], e.g., of olive fruit
[86]. An online MIS was developed for the optimisation of olive orchard harvesting orders based on monitoring networks for environmental data and physio-chemical fruit analysis
[87]. In other examples, a flower initiation model was adopted into a decision support system for mango production in a greenhouse environment
[79], and a web-based MIS for mango harvest time forecasting, based on non-invasive in-field measurements of fruit dry matter, was implemented
[57]. A MIS has also been developed that provides a financial assessment of the value of the use of sensor technologies in harvest timing
[88].
The estimation of harvest loads, i.e., the fruit number per tree and the fruit size, can be based on manual assessments. Wulfsohn et al. (2018) report on sampling strategies that are appropriate for use in these applications, with commercial use exemplified by Pronofrut (San Fernando, Chile) (
https://pronofrut.cl/en) (accessed on 18 December 2023) in which a yield estimation support service is provided for the bud, flower and fruit counts and sizing using multistage systematic sampling designs and manual measurements. In another example, the ‘AKFruitData’ software was used to acquire manual count data for the estimation of apple yields
[37].
A number of researchers have explored the use of machine vision in fruit load assessment. For example, a deep learning model for an Android smartphone app, “KiwiDetector”, was developed for the yield estimation of kiwifruit
[89], a smartphone camera application was used in a client-server architecture for the estimation of apple yields
[77], and the orange fruit count was estimated from UAV collected imagery, with results broadcast to an online map
[90]. The use of depth cameras or LiDAR has enabled the estimation of size profiles of fruit on trees, with a coupling to a fruit growth model that allows a forecast of the fruit size distribution at harvest, as reviewed by Neupane et al.
[91]. Moving beyond the provision of yield data per se to use in management, a system for estimation of the optimal number of harvest containers and their field placement for efficient logistics was developed based on apple yield mapping from orchard video images
[43].
A number of systems also exist for recording fruit yields and quality at harvest. For example, a tree-fruit yield map was generated using labour time data collected from ‘labor monitoring devices’
[66], which was later developed into a cloud-based harvest management system for monitoring harvest labour
[31]. Several cloud-based harvest information systems have been developed, e.g., involving real-time harvest data and the generation of yield maps for hand-harvested cherry, blueberry and apple fruits
[49] or the recording of the in-orchard location of harvest field bins, enabling a postharvest association of fruit to a set of trees for the construction of a yield map
[38]. In another example, a system was developed for the quality assessment of cherry fruit at harvest based on the use of a mobile device to upload pictures of harvested fruit, with a cloud image processing pipeline providing a report on the quality of the fruit
[46].