Spatial variability of chlorophyll and NDVI obtained by different sensors in an experimental coffee field

S.A.S. Silva¹, G.A.S. Ferraz¹*, V.C. Figueiredo², M.M.L. Volpato², M.L. Machado², V.A. Silva², C.S.M. Matos², L. Conti³ and G. Bambi³

¹Federal University of Lavras, School of Engineering, Department of Agricultural Engineering, Rotary Clover Professor Edmir Sá Santos, BR37200-900 Lavras, Brazil
²Agricultural Research Company of Minas Gerais, Av. José Cândido da Silveira 1647, Bairro União Belo Horizonte, BR31170-495 Belo Horizonte, Brazil
³University of Florence – UniFI, Department of Agriculture, Food, Environment and Forestry (DAGRI), Via San Bonaventura, 13, IT50145 Florence, Italy
* Correspondence: gabriel.ferraz@ufla.br

Abstract:

The objective of this research was to study the spatial variability of NDVI and chlorophyll sampled by different sensors, as well as to evaluate the correlation between them in a coffee field. The study was carried out on a coffee farm located in the municipality of Três Pontas, Minas Gerais. A sampling grid containing 30 points was created for the study area. Each sampling point was represented by one plant, which was georeferenced by a GNSS RTK. For each sample plant, NDVI and chlorophyll were obtained by the optical and active sensors GreenSeeker and ClorofiLOG, respectively. In addition, it was carried out a flight with an RPA equipped with a passive and multispectral sensor. Using the data obtained by active sensors, a geostatistical analysis was carried out to evaluate the spatial variability of NDVI and chlorophyll. The geostatistical analysis verified the existence of spatial dependence for the two attributes, and thus it was possible to generate spatialization maps through kriging. The images obtained by the passive sensor resulted in five multispectral orthomosaics, making it possible to calculate the NDVI, thus generating a spatialization map of this index. It was possible to observe in the generated maps, points that presented a certain similarity and for this purpose a correlation analysis was carried out for the values of each attribute, sampled directly in the maps, and in different sampling grids (30, 60, 90 and 120 points). By analyzing the Pearson coefficient (R) it was possible to quantify the level of correlation between the data obtained by the different sensors and through the t test it was possible to find significant correlations between them.

Key words:

active sensor, geostatistics, passive sensor, spatial distribution, spectral indices

Use of geostatistical analyses for wheat production areas throung the variables NDVI, surface temperature and yield

A.L. Abreu¹, G.A.S. Ferraz¹*, R. Morais¹, N.L. Bento¹, L. Conti³, G. Bambi³ and P.F.P. Ferraz¹

¹Federal University of Lavras - UFLA, School of Engineering, Department of Engineering (EENG/DEA), Aquenta Sol, P.O.Brox 3037, 37200-900 Lavras - MG, Brazil
²University of Florence – UniFI, Department of Agriculture, Food, Environment and Forestry (DAGRI), Via San Bonaventura, 13, 50145 Florence, Italy
*Correspondence:gabriel.ferraz@ufla.br

Abstract:

Geostatistics is a crucial tool for data analysis in the field of precision agriculture, allowing the characterization of spatial variability magnitude, optimizing profitability and yield in agricultural areas. In this context, the present study aimed to evaluate the spatial dependence of the variables yield, Normalized Difference Vegetation Index (NDVI), and surface temperature in winter wheat plants. This was achieved through fitting semivariograms with different statistical models and interpolating the study variables using Ordinary kriging. The experiment was conducted at Fazenda Santa Helena, located in the municipality of Lavras in the state of Minas Gerais, Brazil, with a 12-hectare winter wheat crop of the TBIO Calibre variety. Data were collected using a grid sampling method at different stages of wheat plant growth (tillering and elongation). The analyzed variables included yield, NDVI, and surface temperature. Statistical analyses were performed using the R software. Initially, the spatial dependence of the study variables was analyzed by fitting semivariograms using the Restricted Maximum Likelihood (REML) method and considering spherical, exponential, and gaussian models. The evaluation of errors was carried out through cross-validation, and subsequently, the data interpolation was performed using ordinary kriging with the best-fitted semivariogram model. The results demonstrated a proper fit of semivariograms for the study models, with the spherical model standing out for surface temperature variables (elongation and tillering), NDVI (tillering), and the exponential model for NDVI (elongation) and yield. Therefore, the use of geostatistics is emphasized as an important tool to assist in precision agriculture management in winter wheat crops.

Key words: