Proceedings Article | 11 October 2018
KEYWORDS: Synthetic aperture radar, Agriculture, Data acquisition, Vegetation, Sensors, Satellite imaging, Satellites, Polarization, Remote sensing, Spatial resolution
The two gratest challenges of remote sensing in agricultural monitoring is to provide reliable information on yield and crop variability that is temporally consistent to efficiently analyze the crop in its development cycle, and spatially scalable, with sufficient spatial resolution in adition with large aquiring area which allows analysis inside farm and its regional context.
With the technological advancement of orbital sensors mainly in relation to the availability of free data (optics and SAR) provide by Sentinel’s satélite in the Copernicus program, they allow monitoring of agricultural areas every 5 to 12 days, also associated with the processing of data acquired in cloud based plataforms, making the more efficient management of agricultural production. Therefore, the use of SAR data associated with optical data allows the extraction of information on crop development and its productive capacity.
Considering this context, the objective of the study was to analyze the temporal development of the backsacattering obtained by the Sentinel-1 satellites, comparing it with vegetation indices associated with biophysical parameters obtained by Sentinel-2 images along the phenological development of the Maize second crop 2017 and soybean crop 2018, and the possibility of integration of the sensors for the agricultural monitoring.
The product used for the SAR data acquired by Sentinel-1 was collected in the Interferometric Wide Swath Mode (IW) with VV + VH polarization. The level of processing used was the Ground Range Detect using the following steps: 1. Thermal noise removal 2. Radiometric calibration 3. Terrain correction using SRTM 30. The final terrain corrected values are converted to decibels via log scaling (10 * log10 (x )) and quantized to 16-bits. The images acquired by the Sentinel 2 satellite were used at the L1-C TOA processing level, generating three vegetation indexes associated with NDVI vegetative vigor, NDWI water content and PSRI senescence. The online processing platform Google Earth Engine was used for data processing.
We analyzed all Sentinel 1 and 2 acquisitions for maize in the period from 01-03-2017 to 08-01-2017, and for soybean the period from October 10, 2017 to March 1, 2018. To validate the data, field control points were used in areas of the same seeding date with weighing of productive mass. The data were first and correlated for maize 2017 and soybean 2018, characterizing the phenological behavior of each crop for maize and soya. Backscatter data were compared using vegetation indices and correlated using Pearson's correlation test and simple linear regression.
The results obtained showed phenologycal characterization for the two crops with the high correlation between VV polarization and NDVI for corn, obtaining values of r² = 0.853 and for soybean r² = 0.678, both with p-value <0.005 . These results evidenced the integration capacity of the two technologies in the monitoring of agricultural crops, with emphasis in spite of the lower values of NDVI for the soybean crop where it is more difficult to acquire optical images due to the intense amount of associated precipitation.
