METHOD OF IDENTIFYING AGROECOLOGICAL GROUPS USING REMOTE INFORMATION Russian patent published in 2024 - IPC G01N33/24 G01V99/00 

FIELD: agriculture.

SUBSTANCE: method of identifying agroecological groups using remote information is characterized by the fact that multispectral satellite survey of medium resolution of the mapped territory is carried out, with a pixel side of 10–30 m, creating a digital model of the relief of the territory with a similar resolution, combining the obtained information into time series of data, wherein the time series of data covers at least three years or not less than the duration of the crop rotation cycle for the mapped territory, calculating the morphometric characteristics of the relief based on the digital relief model with a resolution of at least 30 m in 1 pixel, using two-dimensional spectral analysis, calculating the fractal dimension of the relief and determining the changes of said dimension depending on the magnitudes and forms of the relief using the inverse spectral transformation, creating digital models for additional morphometric characteristics of the relief, overlaying the mask of agricultural lands on the datasets of remote information and morphometric characteristics of the relief, wherein the mask is applied so that only agricultural lands to be mapped are included in the analysis, on the remote data prepared in this way, performing the procedure for selecting agroecological differences and checking their stability, wherein the physical meaning of the obtained differences is determined using the field agroecological descriptions, when selecting the agroecological differences, spatial stationary clusters are successively identified, determining the physical meaning of the obtained clusters with the agroecological properties of elementary soil contours being identified, determining the degree of stationarity of the obtained elementary soil contours, selecting soil differences and constructing maps of agro-ecological properties and soil characteristics, wherein the task of selecting stationary clusters in sets of source data of multispectral satellite survey is solved by applying hierarchical factor analysis, carrying out step-by-step procedures for reducing the dimension, to identify the relationship between clusters and agroecological characteristics of soils, machine learning procedures are used using a set of field agroecological descriptions as a training sample, wherein to obtain a set of field agroecological descriptions, a field observation network is planned, at that, in order to achieve statistical reliability of semantic interpretation of obtained clusters, from 5 to 10 descriptions of agroecological characteristics of soils are carried out for each selected cluster, determining the degree of stationarity of the obtained maps of elementary soil contours and identifying soil differences by using discriminant analysis of the second row of prepared remote information covering a period which is 3–5 years from the first.

EFFECT: invention makes it possible to specify and accelerate works on agroecological and soil mapping, and also allows to provide the possibility of retrospective analysis of agroecological conditions of agricultural lands.

5 cl, 12 dwg