Method for generating a treatment plan suitable for developing a cultivated area without chemical pesticides, method for treating the cultivated area, and associated learning model and system

A system for monitoring a cultivated area and generating a treatment plan adapted for the development of the cultivated area including a first natural formulation adapted to promote growth and stimulate a plant variety defense system, the system comprising a computer device and at least one measuring device, in which:

• the measuring device is adapted to measure at least one environmental parameter of the cultivated area,
• the computer device is configured to:
• load values of parameters of the cultivated area, said parameters comprising at least one plant variety, at least one varietal resistance index associated with the plant variety and a development index of the plant variety,
• load at least one value of an environmental parameter, for a predetermined past period of time, from the cultivated area,
• load at least one value of a forecast environmental parameter, for a predetermined future period of time, of the cultivated area based on geographical coordinates of said cultivated area, the at least one value of the environmental parameter comprising a precipitation level,
• generating a treatment recommendation of the cultivated area according to the at least one value of the environmental parameter, the at least one value of the forecast environmental parameter and the values of the parameters of the cultivated area, said recommendation including:
• a first treatment of the cultivated area at a first instant of the predetermined future period of time, said first treatment including at least one root and/or foliar application of a second natural formulation comprising an agent that elicits defense reactions;
• said first treatment being carried out at least twice at least 5 days apart;
• a second treatment of the cultivated area, at a second instant of the predetermined future period of time, said second treatment including the application of a third natural formulation comprising the antigerminative, the third natural formulation comprising the antigerminative being applied less than two days before an occurrence of a forecast precipitation episode if the forecast precipitation level is greater than a predetermined threshold,
• a third treatment of the cultivated area at a third instant of the predetermined future period of time, said third treatment including the application of a fourth natural formulation comprising a curative agent, the fourth natural formulation comprising the curative agent being applied less than two days after an end of the forecast precipitation episode.

A method for treating a cultivated area for a predetermined period of time, said method including:

• a first treatment of the cultivated area at a first instant of the predetermined period of time, said first treatment including at least one root and/or foliar application of a second natural formulation comprising an agent that elicits defense reactions;
• said first treatment being carried out at least twice at least 5 days apart;
• a second treatment of the cultivated area, at a second instant of the predetermined period of time, said second treatment including the application of a third natural formulation comprising an antigerminative, the third natural formulation comprising the antigerminative being applied less than two days before an occurrence of a forecast precipitation episode if the forecast precipitation level is greater than a predetermined threshold,
• a third treatment of the cultivated area at a third instant of the predetermined period of time, said third treatment including application of a fourth natural formulation comprising a curative agent, the fourth natural formulation comprising the curative agent being applied less than two days after an end of the forecast precipitation episode.

A machine learning model trained to predict a treatment plan adapted for the conditions of development of a cultivated area including a first natural formulation adapted to promote growth and stimulate the defense system of a plant variety, the trained machine learning model being obtained according to the following steps:

• acquiring, at each of acquisition instants of a plurality of successive acquisition instants, values of reference parameters of a plurality of reference cultivated areas, the reference parameters comprising, for each reference cultivated area, at least:
• values of parameters of the cultivated area, the parameters of the cultivated area including a plant variety, a varietal resistance index associated with the plant variety, a development index of the plant variety,
• a value of an environmental parameter,
• processing data including at least one other natural formulation selected from a second natural formulation adapted to stimulate the defense system of the plant variety, and/or a third natural formulation adapted to prevent the development of a pest, and
• data on the effectiveness of the treatment with the at least one other natural formulation,
• calculating, at each acquisition instant, for each of the plurality of reference cultivated areas, mathematical quantities, from the values of the parameters of the cultivated area, the value of the associated environmental parameter, acquired at acquisition instants included in a predetermined period of time preceding the current acquisition instant,
• creating, for each of the plurality of reference cultivated areas, a mathematical reference function from the parameter values of the cultivated area, the value of the environmental parameter and the mathematical quantities,
• associating, at each acquisition instant, for each of the plurality of reference cultivated areas, the mathematical reference function with the efficiency data of the treatment of the reference cultivated area, so as to obtain variables of the machine learning model, and
• training the machine learning model to predict a treatment plan including at least one other natural formulation, based on a machine learning algorithm and machine learning model variables.