Abstract

With the increase of the world’s population, while urbanization is increasing, agricultural lands are decreasing. Therefore, monitoring of up-to-date agricultural lands is important for agricultural product estimation. The study investigates suitability of Sentinel-2 data for the phenological stage analysis and yield estimation of sunflower plant. To this aim, fieldworks was conducted and sunflower parcels were identified in Zile district of Tokat province, Turkey which has dense sunflower production. In this study, ten Vegetation Indices (VIs) were performed by using multi-temporal Sentinel-2 data obtained during the growth stages of sunflower plant and yield estimation was obtained. As a result, the indices obtained on 30 June, at the stage of inflorescence emergence, provided coefficient of determination (R²) higher than 0.67 and The Root Mean Square Error (RMSE) lower than 13 kg/da. Among the VIs, the best forecast obtained by NDVI (R² = 0.74 and RMSE = 10.80 kg/da) approximately three months before the harvest of sunflower.

摘要

随着世界人口的增加，在城市化进程加快的同时，农业用地也在减少。因此，对最新农业用地的监测对于农产品估算很重要。该研究调查了 Sentinel-2 数据对向日葵植物物候阶段分析和产量估算的适用性。为此，在土耳其托卡特省 Zile 地区进行了实地考察，并确定了向日葵地块，该地区的向日葵产量很高。在这项研究中，利用向日葵植物生长阶段获得的多时相 Sentinel-2 数据进行了 10 个植被指数 (VI)，并获得了产量估计值。因此，在 6 月 30 日，在花序出现阶段获得的指数提供了决定系数（R ²) 高于 0.67 且均方根误差 (RMSE) 低于 13 kg/da。在 VI 中，NDVI 获得的最佳预测值（R ²  = 0.74 和 RMSE = 10.80 kg/da）大约在向日葵收获前三个月。