Monitoramento da cultura da soja em agrossistema com espectrorradiômetro orbital de resolução moderada

Abstract

The techniques of remote sensing that use vegetation indices (VI's) can help in the study of spatial and temporal patterns of soybean (Glycine max L. Merr.), and associated to the edaphoclimatic attributes, cano also enable the economical use of inputs in order to provide higher grain yield and less impact on agrosystems. This study aimed to evaluate the spectral and temporal relationships of the MODIS sensor with normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) with grain yield, relief, texture and soil organic matter (MOS), during the phenological cycle of soybean in Campo Verde (MT) in the 2012/2013 harvest, identifying the best phenological stages to generate predictive models on soil attributes variability and productivity prediction. The EVI / NDVI of the MODIS orbital sensor products (MOD13Q1 and MYD13Q1) and the Savitzky-Golay (SG) filtering for noise correction (anomalous values) present in time series of these VI’s were used. Pearson (r) correlation was made considering P ≤ 0.05 between the variables using two methods of SG filtering in the time series of the phenological cycle. The SG filtering corrected the interference present in the series, and improved the correlation of the EVI with the physical attributes of the soil. The coefficients of determination (R²) of EVI in the R1 stage (56 days after germination) with SOM, clay, silt and sand were, R² = 0.77; 0.75; 0.74; 0.75, respectively. With NDVI in the phenological stage R2 (64 days after germination) it was obtained R² = 0.44 with the productivity. The SG filtering was a necessary tool to study the noises present in the time series of the VI's. From the MODIS image and the information of the phenological stage with greater spectral correlation with the soil attributes, it is possible to perform predictive modeling.