Early-season mapping of winter canola using satellite optical and SAR data

Early and accurate mapping of crops is critical for monitoring crop growth and predicting crop yield for decision makers. However, previous efforts generally produce canola planting area maps at the flowering or late stage. Thus, early-season canola mapping is still a challenge. Here, we aim to propose a method that can be used for early-season winter canola identification and solve the problem on the use of optical images in cloudy and rainy seasons. We analyzed the spectral characteristics, vegetation indexes, backscatter coefficients, and texture features of these crops at the early stage based on Sentine1-2 and Sentinel-l images. Random forest(RF) and support vector machine(SVM) classification models were built to identify winter canola based on early-season images and field samples in 2017. Experimental results demonstrated that the proposed approach overcomes the lack of optical images, and hence provides an effective option to improve the accuracy of early-season mapping of winter canola.

I. Introduction

Early crop mapping is critical for guiding precise farmland management, as well as to local and national food security [1, 2]. It also facilitates early detection of famine and rapid response to potential risks (e.g., agricultural disasters, such as floods, droughts, and storm damage). The production of canola, which is one of the four major oil crops in the world [3], occupies a critical position in the national economy. The canola area should be estimated before winter to accurately obtain the frozen damaged area of winter canola, thereby identifying corresponding management and protection measures.