Development of a Mango Harvest Management Information System

A review of literature on management information systems (MISs) for orchard management was undertaken. MIS development and use for this application was noted to lag that for broadacre cropping applications. Most published work and commercially available product focused to chemical and labour documentation, and to irrigation and pest/disease management, with harvest forecast poorly represented. Methodologies relevant to mango harvest forecast were reviewed, including use of heat units and in-field measurements of fruit dry matter content for forecast of fruit maturation time, and fruit count and fruit size for estimation of fruit load and a ‘harvest forecast engine’ was proposed that combines and codifies this information. Inputs to the system include 15-minute interval temperature data, human and machine vision-based estimates of flowering levels, fruit load and fruit size, and fruit dry matter content (DMC) measurements. Estimation of harvest timing was based on heat units and on target DMC levels, total yield estimation was based on fruit counts and fruit size forecast, and weekly harvest load was based on time distribution of flowering and fruit load estimates. The harvest forecast engine was then incorporated into an electronic MIS artefact. This artefact was developed using design science principles with adoption of an agile software development approach, with the iterative development informed by feedback from agronomic experts, growers and other players in the mango supply chain. The key components of the system are (a) data acquisition and management, (b) the harvest forecast engine and (c) data visualization using WebGIS, tables and charts/graphs. The fundamental design requirements of the system are accuracy of prediction and system responsiveness. The MIS is a progressive web application enforced by RESTful APIs that bridges multiple sub-systems for forecast of harvest timing and fruit load. Limitations of web map visualization were explored through empirical assessment of the spatial accuracy of the web imagery services, and recommendations were made in context of their use in a MIS. Further development of the system, such as time series overlays of satellite derived vegetation index data, is recommended based on evaluation received through interview of 14 stakeholders. 

The thesis is structured into nine chapters (with five published): (i) an introduction; (ii) a review of MIS use for tree fruit; (iii) development of a ‘harvest forecast engine’; (iv) methodology for software development; (v) the proposed software architecture of a mango harvest MIS; (vi) development and evaluation of a harvest forecast MIS; (vii) evaluation of the positional accuracy of web mapping in context of use in an orchard MIS and (viii) an application case study, involving harvest bin placement in orchard, based on use of machine vision fruit load, and (ix) conclusions and recommendations.