Identification of sowing window and sown area of maize and sorghum in rice fallows using multi-source satellite remote sensing

Sunil Kumar Medida (Department of Agronomy, Acharya N.G. Ranga Agricultural University, Lam, Andhra Pradesh, Guntur)
Prasuna Rani Podila (Department of Agronomy, Acharya N.G. Ranga Agricultural University, Lam, Andhra Pradesh, Guntur)
Geetha Sireesha (Department of Agronomy, Acharya N.G. Ranga Agricultural University, Lam, Andhra Pradesh, Guntur)
K. Anny Mrudhula (Department of Agronomy, Acharya N.G. Ranga Agricultural University, Lam, Andhra Pradesh, Guntur)
Ponnurangam Gramani Ganesan (NPP Post Doctoral Fellow, JET Propulsion Laboratory (NASA))

SAINS TANAH - Journal of Soil Science and Agroclimatology
Vol 22, No 2 (2025): December

Identifying the start of sowing in rice fallows is challenging due to its typical low land agro-ecosystem. Tracking the spatio-temporal shifts that take place during the transition from a wet to dry ecosystem, identifying crops, assessing their extent, and identifying optimal planting periods are vital information for researchers and planners. This study aimed at determining the crop sown area and sowing window of maize and sorghum crops planted in rice fallows during the Rabi 2020-2021 season in the Krishna Western delta of Guntur district, Andhra Pradesh. Optical cloud-free satellite images of Landsat-8 and Sentinel-2 were downloaded and using band ratios NDVI and NDWI was derived. A Threshold based algorithm was developed to detect the crop sowing window. The total area sown was determined using the SVM algorithm. The threshold-based algorithm is well-suited for identifying the sowing windows. The sowing window in the second fortnight of January had the largest area for both crops compared to other sowing windows. The detected sowing windows exhibited a deviation of up to two satellite acquisition intervals. The estimated area using SVM algorithm for maize and sorghum was 29,518 ha and 65,417 ha, respectively. The threshold-based algorithm overestimated the maize and sorghum crops as compared to SVM. This study established the superior performance of the Support Vector Machine (SVM) algorithm for crop classification. Statistical validation confirmed that the SVM model achieved significantly higher accuracy in distinguishing both maize and sorghum from other land covers compared to the threshold-based algorithm, which exhibited a greater tendency for misclassification.