Investigating olive trees by monitoring phenological stages using multi-modal satellite sensor data

Abstract

Olive tree is crucial due to its social, cultural and economical values to the Mediterranean society. The demand for olive oil and other products produced from olives is constantly increasing with the global trend of healthy lifestyle. To meet this demand, olive trees are grown in different regions as long as climatic conditions allow. However, global warming and the environmental disasters adversely affect olive trees. As a result, during the past 20 years, several studies on olive trees have been carried out utilising a variety of remote sensing techniques and platforms for diverse agronomic aims, including tree identification, fertilisation, irrigation etc. However, researches on olive trees in Türkiye is limited and to the best knowledge, no studies on olive trees have been performed using remote sensing techniques on regional scale in Türkiye. Therefore, the aim of this research is to map olive trees of Izmir Province, one of the main olive growing regions, for the first time using Sentinel-1 and Sentinel-2 data. The phenology and productivity of olive trees of Izmir are investigated as well, using CLMS HR-VPP products on a regional scale for the first time. The effectiveness of multitemporal Sentinel-1 and -2 data for olive tree classification is evaluated in Chapter 2. Datasets consist of May and October composites of each year between 2017 and 2021 are used in order to capture the different phenological stages of olive trees. Random forest classfier is implemented to datasets. Classification map of Izmir province is produced and number of classes are reduced to two classes "olive" and "non-olive" to create binary olive tree map. Results are validated using both K-fold cross validation and confusion matrix methods. The results show that Sentinel-1 data can map olive trees with up to 70%, while Sentinel-2 data succed up to 90%. However, the combined usage of Sentinel-1 and Sentinel-2 datasets show the best classification accuracy for olive trees. This is a clear example of the advantage of mapping tree crops using radar and optical data fusion. In addition, this is the first olive tree distribution map of Izmir Province, and it's essential for creating a successful management plan. Using data from Sentinel-1 and Sentinel-2, this research assessed the possibilities of mapping olive trees regionally. The phenology and productivity of olive trees of Izmir Province are investigated in Chapter 3. Copernicus Land Monitoring Sevices's High-Resolution Vegetation Phenology and Productivity (HR-VPP) metrics are used. Interannual variability of phenology metrics (SOSD, EOSD) of olive trees are evaluated and the effects of climatic conditions such as precipitation and surface soil mositure as well as elevation to the productivity of olive trees are investigated. Results demonstrate that precipitation have a significant effect on EOSD. Additionally, a strong correlation between total productivity and elevation and climatic conditions, such as precipitation and surface soil moisture, is observed. The strong corrrelation between productivity of olive trees and TUIK crop yield data (R2 = 0.77) verifies the efficiency of CLMS HR-VPP metrics to be used on a regional and global scale. This is one of the first attempts to utilise the power and potential of WEkEO and GEE cloud infrastructures using Sentinel-1 and Sentinel-2 data as well as CLMS HR-VPP products. Moreover, this is the first regional satellite-based phenology and productivity research on olive trees in Türkiye. Field trips to the districts of Bergama and Bayındır are explained in Chapter 4, to look into the reasons of the overestimation and underestimation issues that are discussed in the Chapter 2 and Chapter 3. As olive forests in hillsides are mixed with different types of trees and the other background vegetation around olive trees, olive trees are more likely to be misclassified in hilly regions. In conclusion, this research encourages the use of the cloud infrastructure capabilities of GEE in order to develop a scalable methodology for studies on tree crops at regional to global scales. If this method is expanded to other parts of Türkiye, it will allow for the first country-scale research on olive trees.