Determination of Olive Trees with Multi-sensor Data Fusion

Abstract

Global warming, which triggers climatic changes, has direct effects on the phenology of plants. For a sustainable agricultural production, continuous monitoring of crops and trees is critical to have updated information and producing effective agricultural plans. Remote sensing is an efficient option for this purpose and is a very popular technique. Olive is an essential agricultural product for the economy of Mediterranean countries such as Turkey. Determination of olive trees, which are expanded all around Aegean and}{Mediterranean regions of the country, is critical to assess the production capacity and the quality of products. In this study, combinations of time series of Sentinel-1 satellite images, Sentinel-2 satellite images and NDVI products obtained from Sentinel-2 satellite images are used to investigate the classification accuracy of olive trees. According to analysis results, a significant correlation with R2 = 0.67 found between NDVI and SAR data (sigma nought VH/VV in decibel scale). This result pointed out probable accuracy improvement in classification of fused data from different sensors. In the next step, supervised random forest classification was applied on the fused data combinations and results showed that Sentinel-1 – Sentinel-2, Sentinel-1 – NDVI and Sentinel-2 – NDVI combinations achieved the highest overall accuracy with 73 %, while standalone Sentinel-1 and Sentinel-2 image time series classification accuracies are 48 % and 68 % respectively.