Sentinel 1A SAR Backscattering Signature of Maize and Cotton Crops

Agricultural crops are the important food source for most of the terrestrial species. As a result, the monitoring of crops and their growth stages and discrimination between the crops are considered most important. Monitoring the crop phenology and observing their conditions within the growing season allows the farmers to modify the cultivation practice as needed.

Crop discrimination is a critical first step which allows to discriminate between crop types, mapping the cropped area and estimating the yield at an early stage. For these purposes, the use of remote sensing is essential. Optical remote sensing for crop monitoring has increased over the past several years and become one of the major civilian operational applications. However, several images acquired at specific times during the crop growth cycle are required to reach a suitable accuracy. This temporal constraint limits the use of optical data for operational applications because cloud cover may prevent or delay image acquisitions in many places.

Space-borne Synthetic Aperture Radar (SAR) imagery is able to observe the Earth’s surface independently of such conditions as cloud cover and guarantees a temporal frequency of images throughout the growing period (Boerner et al., 1987).

In the case of SAR polarimetry, the sensitivity of microwave polarization to crop structure (size, shape, and orientation of leaves, stalks, and fruits), dielectric properties (related to the water content), and the physical properties of the underlying soil (roughness and moisture) varies as a function of crop type, growth stage, and crop condition (Haldar et al., 2012 and Iyyappan et al., 2014).

As a fact, different crop types, or the same type at different growth stages, produce different polarimetric signatures, which can be identified in the acquired images. The objective of this paper is to exploit the potential of polarimetric Sentinel 1A SAR data for monitoring and extraction of backscattering signature at different phenological stages of Maize and Cotton crop in Perambalur district of Tamil Nadu.

Sentinel-1A, carrying a C-band radar system with VV (Vertical-Vertical) and VH (Vertical-Horizontal) polarization obtained by interferometric wide (IW) swath mode (1) of High Resolution (HR) and processed at Level-1 ground range (GRD), is used for this study.

In order to have full coverage during the crop growing period of Maize and Cotton in Perambalur district of Tamil Nadu, the Sentinel-1A data were downloaded from 8th August 2015 to 23rd January 2016 at a 12-day interval from the Sentinel Scientific Data Hub – Copernicus website (https://scihub.copernicus.eu/dhus/). The satellite pass is not available between 1st September 2015 and 19th October 2015.

Pre-processing of the satellite images was performed to suppress unwanted distortion or to enhance the image features important for further processing. The pre-processing operations like Orbital correction, Subsetting, Border Noise Removal (Carlo, 2015), Radiometric calibration (Rosich and Meadows, 2004; Laur et al., 2004; and Lavalle and Wright, 2009), Geometric correction (Range Doppler Terrain Method), Co-registration, Layer stacking, Multi-temporal speckle filtering (Quegan et al., 2000), and Linear to dB conversion were done using the Sentinel-1 toolbox.

The ground survey was conducted during the months of October, November, and December 2015 in the cotton and maize growing areas for extraction of backscattering signatures. The survey was planned in such a manner to collect information at different stages of the crop viz., vegetative, tasseling/boll formation, and harvesting.

A total of sixty-nine points were surveyed for maize (39 points) and cotton (30 points), and they were geocoded using the GPS Latitude and Longitude values using the Sentinel-1 toolbox. Further, the temporal backscattering signature values were derived using the geocoded points for VV and VH polarization separately for further analysis.

The radar backscattering coefficient (σ0) is a measure of crop biomass, plant height, water content, underlying soil, and crop phenology. Data collected during the cropping period were processed and analyzed using the training pixels to derive the temporal backscattering coefficient (σ0) for Cotton and Maize. The minimum, maximum, and mean temporal backscattering signature for Vertical-Vertical (VV) and Vertical-Horizontal (VH) polarized SAR data are given in Tables 1 and 2.

The mean backscattering values for cotton crop during the entire cropping period range from -10.58 dB to -6.28 dB and from -20.59 dB to -14.53 dB for VV and VH polarization, respectively. For maize crop, it ranges from -11.08 dB to -7.07 dB and from -19.85 dB to -14.14 dB for VV and VH polarized data, respectively.

When comparing the mean σ0 of VH with VV polarization of cotton crop, it is found that VV backscattering is lesser by 9.66 dB to 13.76 dB at different stages of crop growth. Similarly, for maize crop, backscattering is lesser by 5.82 dB to 9.37 dB. This shows that VV polarization is more sensitive in acquiring surface variation than VH, and this is in line with the work reported by Aubert et al. (2011).

In both cotton and maize signatures, it is found that the mean backscatter value during the land preparation (σ0D1) to the initial crop growth period (σ0D3) shows lesser scattering in both VV and VH when compared to the further crop development stages (σ0D4 to σ0D9). This might be due to soil moisture variation (Xavier Blaes and Pierre Defourny, 2003) or sowing, which made the soil surface smoother (Karjalainen et al., 2004).

Fig 2. Mean Temporal Backscattering Signature of Maize Crop

The mean backscatter value for cotton crop during the harvest (σ0D10) to post-harvest stage (σ0D12) has decreased to a maximum of 1.54 dB in VV and 1.27 dB in VH when compared to the developed crop stage of σ0D9. A similar trend was found in the maize signature, and values decreased to a maximum of 1.99 dB in VV and 1.84 dB in VH polarization.

The decrease in backscatter value may probably be due to the maturity of the crop, which lowers the water content of vegetation (Lillesand and Kiefer, 1994), or is related to vegetation biomass (Skriver et al., 1999), and/or reduced volumetric scattering due to maturity (drying and fall of lower leaves) (Panigrahy and Mishra, 2003).

In both VV and VH, the mean backscatter value for cotton crop has increased by around 4.0 dB during the vegetative, flowering, and boll formation stages (σ0D4 to σ0D9) when compared to the initial establishment period (σ0D3) (Fig 1). A similar trend was found in the maize signature, showing an increase in mean backscatter value ranging from 3 to 5 dB during the vegetative, flowering, and cob formation stages (σ0D4 to σ0D9) (Fig 2). The increase in mean backscattering may be caused due to the increase in vegetation biomass, as reported by Xavier Blaes and Pierre Defourny (2003).

While comparing the mean backscattering values of cotton and maize at the peak crop growth stage (σ0D4 to σ0D9), an increase in value of 1.0 dB to 2 dB was observed, clearly indicating that the crop geometry, volumetric backscattering from the crop canopy, and biomass lead to this increase (Fig 1 and Fig 2). Several research works are in line with these results (Haldar et al., 2014; Shang et al., 2009; and Xavier Blaes and Pierre Defourny, 2003).

While considering the entire crop period of cotton, the minimum backscattering value of -12.90 dB at σ0D3 and -20.05 dB at σ0D1 is recorded during the land preparation stage for VV and VH polarization, respectively. Similarly, the maximum backscattering value of -2.67 dB at σ0D5 and -13.01 dB at σ0D8 is recorded during the vegetative to boll formation stage for VV and VH polarization, respectively.

In the case of maize, the minimum backscattering value of -14.30 dB at σ0D12 and -22.87 dB at σ0D2 is recorded during the land preparation stage for VV and VH polarization, respectively. The maximum backscattering value of 1.86 dB at σ0D5 and -11.01 dB at σ0D8 is recorded during the vegetative to cob formation stage for VV and VH polarization, respectively.

From the above results, it is clear that the maize crop has recorded the highest backscattering value during the peak growth stage, which might be due to higher biomass content in maize when compared to cotton crop. This is in line with the work done by Soria-Ruiz et al. (2001) and Macelloni et al. (2001).

It is evident from the present study, that the multi-temporal Sentinel 1-A SAR sensor can be well used for the discrimination of cotton and maize crops because of its high temporal resolution which captures the complete phenology of the crops during the cropping period.