The analysis and determination of banana fields backscatter values using sar and optical satellite images

Monitoring of the agricultural fields is important for creating the policies and the decision makers. Because sustainable and effective use of agricultural land is of great importance in the development of countries. Rapidly increasing population, food security has become important for the societies. Therefore, a correct determination and detection of the agricultural fields will help to minimize the problems and setting up a good agricultural planning policy by the governments. To monitor the agricultural fields, remote sensing technology is useful to achieve results in a short time with low cost. Optical and radar sensors have been used frequently for agricultural studies. Optical sensors have a rich spectral resolution but have limitation in terms of data acquisition time and weather conditions. On the other hand, radar sensor have advantages with their capability to acquire image data in all weather conditions and 7/24 hours since they are passive sensors, do not need solar radiation to derive the images and sensors are sensitive to the object geometry, and soil moisture which is important for the agricultural fields. In our study, the banana fields in Antalya-Gazipaşa are investigated with the use of Sentinel-1 SAR and also Sentinel-2 optical datasets for comparison. The aim is separation of the banana fields from the forest area. The banana fields are open areas and also greenhouses. As agricultural cycle, the banana trees volume decreases by February till March, this is also time period of maintenance of the banana greenhouses. After this period, the volume of the banana trees increases, which are the indicators for their separation from the other trees. Radar images are useful to identify these changes as we have used. First, the temporal change in backscatter values of banana and forest areas with time was observed on Sentinel-1 SAR images. The dates when banana trees had minimum and maximum volumetric density were determined. Two different methods were applied on the SAR dataset: the two images difference and the difference in mean values of the enhancemed images. After that, Sentinel-2 optical images were applied the Enhanced Vegetation Index (EVI) to classify the study areas. The EVI made it easy to separate greenhouses covering a certain part of the study area and banana areas grown in the open field, as it is more sensitive to the structural variations of the vegetation. Minimum and maximum pixel values of banana areas were determined on Sentinel-1 SAR and Sentinel-2 optical images with the help of histogram. In evaluating the area accuracy of the banana fields, the fields obtained manually from Google Earth were used. In the Yakacık region, 223.93 ha with the two images difference, 184.83 ha with the difference of average values of the enhanced images, and 143.11 ha with the Enhanced Vegetation Index were determined. In the Zeytinada region, 330.20 with the two images difference, 204.98 ha with the difference of average values of the enhanced images, and 155.00 ha with the Enhanced Vegetation Index were determined. Finally, in the evaluation of the results obtained by different methods, the areas obtained manually were calculated as 163.27 ha for the Yakacık region and 197.89 ha for the Zeytinada region.