Document Type : Full length article
M.Sc. Student in Remote Sensing and GIS, University of Tehran, Tehran, Iran
Associate Professor Department of Remote Sensing and GIS, University of Tehran, Tehran, Iran
The Earth's ecosystems play an important role in regional and global climate. Most natural vegetation covers change through the year- as they are influenced by the seasons. In vegetation studies, different types of remote sensing images such as optics and synthetic aperture radar (SAR) have been used in different scales from leaf area to global scale. These images provide data that is difficult to access through other methods such as field surveys. Remote sensing sensors capture images from the Earth surfaces with an adequate spatial and temporal resolution for the environmental studies. In remote sensing approaches, the study of the phenological cycle (the study of plant life cycles and the way is affected by weather) is mainly based on changes in reflectance values in different spectral bands of optic sensors or vegetation indices (VI), such as normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and soil-adjusted vegetation index (SAVI). Spectral indices have been widely used to monitor the seasonal cycle of vegetation photosynthesis over the past decades. Many of these studies have reported promising results. SAR systems can capture images in all weather conditions and overcome the limitation of optic sensors in cloudy weather. Increased access to SAR images broadens the image application in vegetation studies. SAR sensors operate at a microwave range of electromagnetic spectrum and are able to penetrate more in vegetation canopy. In this study, the efficiency of Landsat 8, Sentinel-2, and Sentinel-1 images in monitoring vegetation phonological cycle have been verified. For that, three regions with different vegetation types including mangrove forests, woodland, and Shadegan date palms in Iran have been studied.
All available Landsat 8, Sentinel-2, and Sentinel-1images in 2017 have been acquired. The Landsat 8 and Sentinel-2 images have been pre-processed. NDVI, EVI, and SAVI have been calculated from corresponding optical bands. Field survey was not possible at the study areas, therefore, sample points have been chosen by the help of high-resolution Google Erath images. For that, archive Google Earth images with the acquisition date close to the acquisition date of images have been used to confirm the presence the vegetation covers at the specific location. The NDVI, EVI, and SAVI values have been extracted in the location of sample points. The Sentinel-1 images have been processed; speckle effect has been minimized and radiometric terrain corrections have been implemented by means of digital elevation model (DEM). Median filter with 5*5 window size has been applied. Median Filter has been chosen, because it is not affected by very high or low DN values and it is one of the efficient filter in minimizing speckle effect in SAR images. Then, digital number (DN) values have been converted into backscattering coefficient in dB. Backscattering responses in VV (vertical-vertical) and VH (vertical-horizontal) polarimetric bands have been extracted at the same sample location. The extracted NDVI, EVI, SAVI, VH, and VV values on different days of the year (DOY) have been separately analyzed for each study site.
Results and discussion
In woodlands, EVI and SAVI indices in comparison to NDVI are more compatible with natural phenological cycles. However, optical images were not available for the whole year, therefore the changes of optical indices cannot be surveyed completely over the year. The changes of backscattering values follow the natural trend of vegetation cover, however, optical indices match better with the natural cycle. The growth cycle of woodlands is affected by temperature and rainfall variation; therefore, it will change in different years.
The changes in spectral curves in date palms show that spectral indices present the initial steps of the growth cycles better than the final steps. Few optical images were available because there was cloud cover in this area. Spectral indices do not follow the last stages of natural phenological cycles. In this stage, backscattering values increase due to the increased volume scattering of the trees, therefore radar images are more efficient in presenting the last part of phenological cycles of date palms in comparison to optical images. Spectral indices are sensitive to the greenness of the leaves; in this stage, no substantial changes in vegetation greenness occur, therefor the spectral indices do not change accordingly.
Mangrove forests have specific phenological cycles and are affected more by environmental conditions. Both spectral indices and backscattering values follow the natural trend of this kind of vegetation.
VH backscattering values are more compatible with spectral indices in comparison to VV backscattering values. Spectral indices and VH backscattering values follow the natural seasonal changes of vegetation especially in deciduous vegetation such as woodlands. This matches with previous studies. The highest values of backscattering are observed in the time that vegetation cover reaches the highest amount of biomass. EVI and SAVI trends are more similar to the backscattering values trend in comparison to NDVI values. This study only considers images captured during one year (2017). Vegetation cover is influenced by seasonal, gradual, and sudden changes, therefore monitoring of vegetation in a longer period and shorter revisit time will provide complete monitoring of vegetation growth cycles.
Backscattering values in the cross-polarized VH (in comparison to the VV band) band show more sensitivity to vegetation changes over the year and are therefore more suitable for monitoring the annual growth cycle of plants. Among the optics indices, EVI and SAVI have shown more acceptable results since their variations are more consistent with the natural phenological cycle. In an aquatic ecosystem where mangrove forest grows, SAR responses show promising results as they can better represent the phenological cycle in comparison to spectral reflectance or vegetation indices. The results of this study show that backscattering responses at C-band follow the natural vegetation’s phenological cycle and can be used in vegetation monitoring in these three ecosystems. The results of this study can be further used to identify vegetation phenological stages in similar ecosystems. Additional studies are necessary to generalize these results to other areas.