Mehdaoui, R; Anane, M (2020). Exploitation of the red-edge bands of Sentinel 2 to improve the estimation of durum wheat yield in Grombalia region (Northeastern Tunisia). INTERNATIONAL JOURNAL OF REMOTE SENSING, 41(23), 8984-9006.

Worldwide, cereals are of great interest for national food security. The recently released free Sentinel 2 imagery is promising for cereal production systems management, both at field and regional scale. The present study aims to estimate and map the durum wheat yield in Grombalia region using the Sentinel 2 red-edge bands. First, the spatial distribution of wheat in Grombalia region was mapped applying the maximum likelihood classification of Sentinel 2 images spread out between October, 2017, and June, 2018. Then the wheat yield of eleven sampled fields was estimated based on these images through empirical regressions models. The independent variable of the regressions was the yield measured in-situ by objective method. The dependent variables were different types of vegetation indices derived from Sentinel 2 red, red-edge and near-infrared bands. The vegetation index of the best model was chosen to estimate and map the wheat yield over the entire study area. The maximum likelihood method classifies accurately the wheat in Grombalia region with wheat Kappa index (K) of 0.95, wheat user accuracy of 0.95 and wheat producer accuracy of 0.99. The area of wheat cultivated in Grombalia during the campaign 2017/2018 is about 1200 ha. On the other hand, the coefficient of determination (R-2) of the tested empirical regressions models for yield estimation is between 0.55 and 0.73 and the Root Mean Square Error (RMSE) varies from 3.80 to 4.90 Qx ha(-1). The best model is the one that employsB(7)andB(5)bands. According to this model the wheat yield of the entire study area ranges from 0.90 to 53.00 Qx ha(-1)with a mean of 21.00 Qx ha(-1)and a total production of 23,452 Qx. This mean is very close to the official mean recorded during the campaign 2017/2018 in Nabeul governorate, which is 22.10 Qx ha(-1).