Search

Soybean crop plays an important role in world food production and food security, and agricultural production should be increased accordingly to meet the global food demand. Satellite remote sensing data is considered a promising proxy for monitoring and predicting yield. This research aimed to evaluate strategies for monitoring within-field soybean yield using Sentinel-2 visible, near-infrared and shortwave infrared (Vis/NIR/SWIR) spectral bands and partial least squares regression (PLSR) and support vector regression (SVR) methods. Soybean yield maps (over 500 ha) were recorded by a combine harvester with a yield monitor in 15 fields (3 farms) in Paraná State, southern Brazil. Sentinel-2 images (spectral bands and 8 vegetation indices) across a cropping season were correlated to soybean yield. Information pooled across the cropping season presented better results compared to single images, with best performance of Vis/NIR/SWIR spectral bands under PLSR and SVR. At the grain filling stage, field-, farm- and global-based models were evaluated and presented similar trends compared to leaf-based hyperspectral reflectance collected at the Brazilian National Soybean Research Center. SVR outperformed PLSR, with a strong correlation between observed and predicted yield. For within-field soybean yield mapping, field-based SVR models (developed individually for each field) presented the highest accuracies. The results obtained demonstrate the possibility of developing within-field yield prediction models using Sentinel-2 Vis/NIR/SWIR bands through machine learning methods. [ABSTRACT FROM AUTHOR]

The characterization of physical and chemical soils attributes is a pressing necessity for the agricultural land management optimization in many countries. Currently, soil analyses are performed by chemical treatments in a laboratory, generating environmental and time-consuming problems. Remote sensing techniques can be faster and cheaper than conventional methods, do not generate chemical residues and are non-destructive to the samples. The objective of the present work was to determine a remote sensing technique to estimate soil physical and chemical attributes in the regions of Paraguaçu Paulista and Rio Brilhante, in the States of São Paulo and Mato Grosso do Sul, respectively, Brazil, using reflectance data obtained by a sensor located in orbit. Fieldwork was performed to validate orbital data. A total of 110 soil samples were collected representing 43,000 h for the development of spectral models. Landsat-7 ETM+images were atmospherically corrected and transformed to reflectance. The soil samples observed in the field were located by GPS and evaluated in the orbital image. The method used consists in a detailed investigation of the spectral data, in which the spectral curve, the position of the data in a graphic dispersion, the colour compositions and the pixel cursor values are evaluated. Spectral models were determined to quantify soil attributes. Soil samples from a different area had their attribute contents determined by the models. The attributes studied were sand, silt, clay, pH (CaCl2), organic matter, phosphorus, potassium, calcium, magnesium, aluminium, hydrogen, cation exchange capacity (CEC), sum of cations (SC) base saturation (BS) and aluminium saturation (AS). The results showed weak correlations with some soil attributes, such as SC, K, Ca, Mg, Al and P. High correlations (reaching 0.86) were obtained with sand, clay, silt, OM, CEC and H. On the other hand, validation procedure indicated that the best attributes for quantification were clearly clay (0.61) and sand (0.5). Therefore, there is strong evidence that these attributes can be predicted in similar landscapes, using the multiple-linear equations developed in this study. This knowledge can be useful in many ways in agriculture, such as soil mapping, determining soil attributes, determining soil information in difficult access regions, and diminishing traditional soil analyses with environmental protection. [ABSTRACT FROM AUTHOR]