Estimation of transplanting and harvest dates of rice crops in the Philippines using Sentinel-1 data

Rice is a staple crop in the Philippines, thus, identifying the ideal window to carry out crop management activities is valuable for efficient monitoring and resource allocation. This study used Sentinel-1A and 1B Synthetic Aperture Radar (SAR) data to estimate the transplanting and harvesting dates of paddy rice under dry and wet seasons and varying climatic conditions. A total of 99 rice fields in three provinces with distinct climatic patterns were considered in this study. From Sentinel-1, we extracted the mean backscatter coefficients in VV, VH, and VH/VV polarizations for each field to generate time series curves with a temporal resolution of 6 days. To mitigate noise, locally weighted scatterplot smoothing (LOWESS) was applied. Periodogram analysis and the Breusch-Godfrey test were used to identify repetitive patterns and their statistical significance. Local extrema and corresponding dates suggest potential transplanting and harvesting dates. The identified dates were then compared with field data from farmer interviews. The root mean squared difference (RMSD) for transplanting ranged from 9 to 16 days and 14–29 days for dry and wet seasons, respectively. Harvest estimates followed similar trends with generally less scattered RMSD during the dry season (16–17.5 days) compared to the wet season values (8–22 days). Results show that VH and VV polarizations are promising for estimating transplanting and harvest dates during the dry season, whereas, VH/VV polarization were better during the wet season. The study emphasized the importance of SAR data for monitoring crop management strategies which are important for the agricultural sector. • Sentinel-1 data was utilized for determining transplanting and harvest dates of rice crops in different seasons. • The mean backscatter coefficients in VV, VH, and VH/VV polarizations were integrated to generate time series curves with a temporal resolution of six days. • The locally weighted scatterplot smoothing (LOWESS) method reduced noise in the data and enhanced the accuracy of transplanting and harvest dates estimation. • VH/VV polarization was found effective for estimating transplanting and harvest dates during the wet season.