Multispectral reflectance and geostatistic methods to estimate leaf nitrogen content and cotton yield.

ABSTRACT: Satellite images and geostatistics are useful tools to assess the nutritional status of plants, and thus, understanding the variability of cotton yield in farmers´ fields, within the precision agriculture onfarm experiments. The objective of this study was the spatial identification of the NDVI index and spatial distribution of nitrogen contents and cotton yield through different phenological stages of the crop using geostatistical methods in Goiás state, Brazil. The experiment was carried out in a commercial field with 47.4 ha, in 80x80m georeferenced grid with 74 plots. Yield monitor data and multispectral satellite images at 56 m spatial resolution were collected in a rainfed cotton field in two dates to monitor the plant vigor, as well as leaf samples were collected for laboratory analysis. Satellite images of AWiFS sensor were acquired on two fenological states, during the first flowering and fruiting cotton stages, respectively, corresponding to 70 and 120DAE (days after emergence). Measures of canopy reflectance, plant height and leaf nitrogen content were determined and cotton yield was obtained by mechanical harvest in August, 2012. Data were analyzed using descriptive statistics, correlation and geostatistical analyses by building and setting semivariograms and ordinary kriging interpolation. Best correlation was found between NDVI and cotton yield at 120DAE. At first flowering at 70DAE, the NDVI and cotton yield showed strong spatial dependence, while for 120DAE there was no dependence, probably due to the enlargement of vegetated coverage. There were similarities in the bottom left of the area with high values of NDVI 70DAE, as well as the highest values of cotton yield due to excellent plant vigor in the cotton flowering stage. By using geostatistics methods with remote sensing data retrieved by satellite images of medium resolution, it was possible a spatial identification of differences in plant development and also to predict cotton yield