Researchers Introduce AI Tool to Help Olive Farmers Predict Harvest Timing

Summary

The Predic 1 Operational Group suc­cess­fully devel­oped a plat­form to pre­dict olive har­vests up to an entire sea­son in advance with up to 90 per­cent accu­racy, using data min­ing method­olo­gies and machine learn­ing algo­rithms. The project, funded by European agri­cul­tural funds, aims to pro­vide a freely avail­able web-based appli­ca­tion to improve farm man­age­ment and resource opti­miza­tion in the olive sec­tor, with the poten­tial to enhance deci­sion-mak­ing and sus­tain­abil­ity in the indus­try.

Following more than three years of devel­op­ment, the results of the Predic 1 Operational Group’s work were pre­sented last month at a con­fer­ence in Mengíbar, Jaén.

The group’s remit was to deliver a plat­form capa­ble of pre­dict­ing olive har­vests an entire sea­son in advance, a goal they said they accom­plished with an accu­racy of up to 90 per­cent.

The work was car­ried out by a con­sor­tium com­pris­ing the University of Jaén, Cetemet, Citoliva, Cooperativas Agro-ali­men­ta­rias de Andalucía, a farm­ers’ union, and Nutesca, using tra­di­tional Picual olive groves in Jaén, Córdoba and Granada as test cases.

According to María Isabel Ramos, a pro­fes­sor at the University of Jaén’s Department of Cartographic, Geodetic and Photogrammetric Engineering and cor­re­spond­ing author of a 2022 study about the tech­nol­ogy, pre­dic­tive sys­tems are cru­cial to the future of the olive sec­tor.

“At the sci­en­tific level, crop har­vest pre­dic­tion is one of the most com­plex prob­lems within pre­ci­sion agri­cul­ture,” she said. ​“There are sev­eral stud­ies that make these pre­dic­tions based on the close rela­tion­ship between the emis­sion of pollen and fruit pro­duc­tion, oth­ers from aer­o­bi­o­log­i­cal, phe­no­log­i­cal and mete­o­ro­log­i­cal vari­ables, all with effi­cient and accept­able accu­ra­cies from July onwards.”

“We intend to advance this pre­dic­tion and be able to make opti­mal pre­dic­tions in the period before flow­er­ing… long before the farmer car­ries out their strate­gic plan­ning and eco­nomic invest­ment in the farm,” Ramos added.

The group used data min­ing method­olo­gies pre­vi­ously used in pre­dic­tive health­care projects to cre­ate regres­sion mod­els from mete­o­ro­log­i­cal data and his­tor­i­cal har­vest data from across the ini­tial tar­get area.

This was com­bined with cur­rent data from drones equipped with ther­mo­graphic sen­sors and mul­ti­spec­tral cam­eras, satel­lite imagery, phe­no­log­i­cal assess­ments, foliar and soil analy­ses and data col­lected from model farms.

The model uti­lizes machine learn­ing, the best-estab­lished field of arti­fi­cial intel­li­gence and one with a proven track record in agri­cul­ture, to pre­dict crop yields as accu­rately as pos­si­ble.

Using a sup­port vec­tor machine algo­rithm made it pos­si­ble to use mul­ti­ple ker­nels, namely the lin­ear and Gaussian ker­nels. This makes it eas­ier for the algo­rithm to adapt to the nature of the data, allow­ing infi­nite trans­for­ma­tions to be car­ried out.

The plat­form will be freely avail­able as a web-based appli­ca­tion sim­i­lar to SIGPAC, the Spanish government’s geo­graphic infor­ma­tion sys­tem for agri­cul­tural parcels.

Users can view an inter­ac­tive graph­i­cal rep­re­sen­ta­tion of the requested infor­ma­tion and export the data.

Francisco Ramón Feito Higueruela, chair of com­puter graph­ics and geo­mat­ics at the University of Jaén and tech­ni­cal coor­di­na­tor of the project, explained that as the num­ber of users increases and the results of future har­vests are fed back into the sys­tem, the accu­racy of pre­dic­tions will improve. More effi­cient mod­els tai­lored to each area will be pos­si­ble.

José Menar Pacheco of the Cooperativas Agro-ali­men­ta­rias de Andalucía high­lighted the impor­tance of his organization’s role in dis­sem­i­nat­ing the project results and knowl­edge to stake­hold­ers.

He hopes to ensure broad aware­ness and adop­tion of the pro­jec­t’s find­ings to improve his mem­bers’ farm man­age­ment and resource opti­miza­tion. Those mem­bers account for more than €11 mil­lion in annual turnover and over 70 per­cent of Andalusia’s total olive oil pro­duc­tion.

The project is financed through the European agri­cul­tural funds for rural devel­op­ment and the Andalusian regional gov­ern­ment as part of the call for regional oper­a­tional groups of the European Innovation Partnership in agri­cul­tural pro­duc­tiv­ity and sus­tain­abil­ity in the olive sec­tor.

“Within the Common Agricultural Policy, a series of new reforms are being imple­mented, includ­ing the fight against cli­mate change with these envi­ron­men­tal objec­tives, as well as the achieve­ment of a sus­tain­able and com­pet­i­tive agri­cul­tural sec­tor by sup­port­ing farm­ers, and all this with a strong com­mit­ment to the dig­i­tal­iza­tion of the olive sec­tor to achieve these objec­tives,” Ramos said.

She added, ​“The ful­fill­ment of these objec­tives depends on the appro­pri­ate deci­sion-mak­ing by each of the actors involved in the sec­tor. Therefore, pre­dic­tive sys­tems are a cru­cial tool in man­age­ment and deci­sion-mak­ing.”

• University of Jaén