Time-series Sentinel-1A SAR remote sensing of rice planting methods in Ebonyi State, Nigeria

Abstract

Timely and reliable agricultural data are essential to monitoring the implementation of rice intervention programs and policies aimed at sustainably reducing hunger and poverty. The difficulties associated with traditional data-gathering methods have made using satellite imagery for data acquisition imperative due to its coverage of large agricultural areas, high repetitive coverage, and cost-effectiveness. However, the use of satellite images for rice research in Nigeria is hampered by the peculiarities associated with methods of rice crop establishment, where manual transplanting and direct seeding methods are practiced in single rice plots due to financial constraints. To examine the effect of different planting methods on sensor signals, we analysed the temporal variation of Sentinel-1 synthetic aperture radar (SAR) data. Also, the relationship between SAR backscatter coefficients and field-measured rice plant parameters (wet biomass and plant height) as a function of planting methods was determined in contrast to similar studies where only the temporal variations in the planting methods were investigated. Multi-temporal, dual-polarised C-band Sentinel-1A backscatter data at 20 m spatial resolution and field measurements were simultaneously acquired every 12 days over the study area during the rainy season, from August to December 2019, in Ebonyi state, Nigeria. Average backscatter values were derived per sampling rice plot, with the temporal profiles of rice SAR backscatter plotted to show variability due to planting methods. Simple regression analysis determined the relationships between measured rice parameters and VH (Vertical transmit, Horizontal receive) and VV (Vertical transmit, Vertical receive) polarisation backscatter coefficients. Results revealed variations in the temporal profile and responses of transplanted and direct-seeded rice fields at different crop growth stages. Regression analysis showed that VH and VV backscatter values for transplanted rice were significantly related to wet rice biomass (R2 = 0.6; 0.51) and rice plant height (R2 = 0.5; 0.59). Conversely, wet biomass (R2 = 0.24; 0.19) and plant height (R2 = 0.18; 0.2) were reported for broadcasted rice for VH and VV polarisations, respectively. These findings show that significant differences exist in the SAR backscatter responses of these crop establishment methods, affecting the retrieved data's accuracy and reliability. Therefore, a cautious approach is advocated in using SAR data from small rice plots with multiple planting methods.