TY - JOUR

T1 - ОПЫТ ПРИМЕНЕНИЯ МЕТОДОВ МАТЕМАТИЧЕСКОЙ СТАТИСТИКИ ДЛЯ ОЦЕНКИ СОСТОЯНИЯ СЕЛЬСКОХОЗЯЙСТВЕННЫХ РАСТЕНИЙ

AU - Bure, V. M.

AU - Petrushin, A. F.

AU - Mitrofanov, E. P.

AU - Mitrofanova, O. A.

AU - Denisov, V.

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Solving problems related to the assessment of the status of agricultural plants during the growing season, allows us to effectively use fertilizers, obtain favorable yields, improve the quality characteristics of plants, as well as the ecological condition of the field. To solve such problems of precision farming, the use of various methods of mathematical statistics is becoming an increasingly promising direction. The aim of our work was to assess the state of agricultural plants using an approach based on the combined use of kriging and binary regression methods, as well as the determination of nitrogen planting using the NDVI (Normalized Difference Vegetation Index) index. The studies were carried out at the site of an experimental agricultural field located on the territory of the branch of the Agrophysical Institute (Menkovo, Leningrad region) in 2015. With the help of aerial photographs taken from the automatized unmanned aerial vehicle complex Geoscan-401 (Geoscan Group of Companies, Russia), a set of NDVI (Normalized Difference Vegetation Index) vegetation index values was obtained at arbitrary points of the plot. A number of ground-based measurements were also conducted on the studied area of the field. The proposed approach to assessing the state of agricultural plants consisted in the joint use of two methods of mathematical statistics: ordinary kriging and logistic regression. A preliminary variogram analysis was carried out, and a variogram model was constructed. After this, the kriging method was used to calculate a series of predicted values of the parameter being studied. At the next stage, the threshold value of the parameter for the study area was established, and also a dummy variable was entered, taking the value 1 if the parameter value exceeded the threshold, and 0 otherwise. Then a logit model was built, in which one of the factors was a series of estimates of the parameter of interest, obtained using the ordinary kriging method. The input _data for building logit models were as follows: N(x i ) is the NDVI value at the location x i , i = 1.78; variable T = 1, if N(x i ) 0.46, otherwise T = 0; the variables X and Y are the coordinates of the observations, are considered as explanatory variables; N pred (x i ) is parameter values, predicted using the kriging method at the observed points. All calculations were performed using the R programming language. As a result of the experiment, three logit models were built with the dependent variable T: in the first model, the explanatory variables X and Y; in the second model — X, Y and N pred ; in the third model N pred . Testing showed that when adding the N pred variable, the logit model works better (2 times less than the erroneous determination of the level of the parameter under study). The results obtained suggest that adding in the binary regression factors a set of values predicted by the kriging method can significantly improve the accuracy of calculations.

AB - Solving problems related to the assessment of the status of agricultural plants during the growing season, allows us to effectively use fertilizers, obtain favorable yields, improve the quality characteristics of plants, as well as the ecological condition of the field. To solve such problems of precision farming, the use of various methods of mathematical statistics is becoming an increasingly promising direction. The aim of our work was to assess the state of agricultural plants using an approach based on the combined use of kriging and binary regression methods, as well as the determination of nitrogen planting using the NDVI (Normalized Difference Vegetation Index) index. The studies were carried out at the site of an experimental agricultural field located on the territory of the branch of the Agrophysical Institute (Menkovo, Leningrad region) in 2015. With the help of aerial photographs taken from the automatized unmanned aerial vehicle complex Geoscan-401 (Geoscan Group of Companies, Russia), a set of NDVI (Normalized Difference Vegetation Index) vegetation index values was obtained at arbitrary points of the plot. A number of ground-based measurements were also conducted on the studied area of the field. The proposed approach to assessing the state of agricultural plants consisted in the joint use of two methods of mathematical statistics: ordinary kriging and logistic regression. A preliminary variogram analysis was carried out, and a variogram model was constructed. After this, the kriging method was used to calculate a series of predicted values of the parameter being studied. At the next stage, the threshold value of the parameter for the study area was established, and also a dummy variable was entered, taking the value 1 if the parameter value exceeded the threshold, and 0 otherwise. Then a logit model was built, in which one of the factors was a series of estimates of the parameter of interest, obtained using the ordinary kriging method. The input _data for building logit models were as follows: N(x i ) is the NDVI value at the location x i , i = 1.78; variable T = 1, if N(x i ) 0.46, otherwise T = 0; the variables X and Y are the coordinates of the observations, are considered as explanatory variables; N pred (x i ) is parameter values, predicted using the kriging method at the observed points. All calculations were performed using the R programming language. As a result of the experiment, three logit models were built with the dependent variable T: in the first model, the explanatory variables X and Y; in the second model — X, Y and N pred ; in the third model N pred . Testing showed that when adding the N pred variable, the logit model works better (2 times less than the erroneous determination of the level of the parameter under study). The results obtained suggest that adding in the binary regression factors a set of values predicted by the kriging method can significantly improve the accuracy of calculations.

KW - Binary regression

KW - Kriging

KW - Language R

KW - NDVI

KW - Normalized Difference Vegetation Index

KW - Plant status

KW - Binary regression

KW - Kriging

KW - Language R

KW - NDVI

KW - Normalized Difference Vegetation Index

KW - Plant status

UR - http://www.scopus.com/inward/record.url?scp=85065130341&partnerID=8YFLogxK

UR - http://www.mendeley.com/research/experience-mathematical-statistics-methods-assessment-agricultural-plants-status

U2 - 10.15389/agrobiology.2019.1.84rus

DO - 10.15389/agrobiology.2019.1.84rus

M3 - статья

AN - SCOPUS:85065130341

VL - 54

SP - 84

EP - 90

JO - СЕЛЬСКОХОЗЯЙСТВЕННАЯ БИОЛОГИЯ. СЕРИЯ: БИОЛОГИЯ РАСТЕНИЙ. СЕРИЯ: БИОЛОГИЯ ЖИВОТНЫХ

JF - СЕЛЬСКОХОЗЯЙСТВЕННАЯ БИОЛОГИЯ. СЕРИЯ: БИОЛОГИЯ РАСТЕНИЙ. СЕРИЯ: БИОЛОГИЯ ЖИВОТНЫХ

SN - 0131-6397

IS - 1

ER -