پایش آب‌های سطحی ناشی از پایش سیلاب با استفاده از قطبش‌های مختلف و شاخص‌های طیفی آب مبتنی بر سنجش از دور در آبخیز کرخه

Introduction and Goal
Global statistics show that floods affect more than 40 % of the world's population and have resulted in significant loss of life and property. In Iran, more than 80 % of cities are at risk of flooding. Therefore, investing in preventive measures and modern technologies to combat floods is more necessary than ever. Preparedness and risk reduction before disasters could significantly reduce the cost of post-disaster recovery. Accurate mapping and long-term monitoring of floods play a key role in planning and prevention. Although Sentinel-1 synthetic aperture radar (SAR) images are useful for identifying flood areas in cloudy conditions, this method also has limitations. The sensitivity of radar data to dense vegetation and vertical structures may lead to under- or over-estimation of flooded areas. Besides, the Normalized Differential Water Index (NDWI), Modified Normalized Differential Water Index (MNDWI), and Automatic Water Extraction Index (AWEI) derived from optical data in the presence of clouds or plant shadows may have errors, leading to incomplete or erroneous identification of flood areas. Therefore, the results obtained from both methods should be interpreted with caution, and the combination of radar and optical data can reduce these limitations, but does not completely eliminate them. Accordingly, this study aims to analyze and compare radar and optical data using remote sensing indices related to the water spectrum to identify flooded areas in the Karkheh watershed located in Khuzestan province.

Materials and Methods
The Karkheh watershed witnessed a widespread and unpredictable flood in March and April 2019, which resulted in significant damage. In this study, three different methods were used to identify and analyze surface water caused by flooding: (1) To detect changes, data from three time periods were used, including pre-agriculture season (September 28 to October 25, 2018), post-agriculture season (February 28 to March 25, 2019), and during the flood (March 28 to April 25, 2019). (2) SAR images obtained from the Sentinel-1 satellite with VV and VH polarizations were processed using the Refined Lee filter and the Otsu threshold algorithm to identify flooded areas. (3) To more accurately extract surface water from optical images, spectral water indices including NDWI, MNDWI, and AWEI obtained from the Sentinel-2 satellite were used. It is worth noting that AWEI has been introduced in two versions: shadowless for open areas and shaded to reduce the effect of shadow in urban and mountainous areas. In this study, the shadowless version was used. Finally, by combining the results from radar and optical methods and analyzing multi-temporal variations, comprehensive maps of flood spread were produced and validated.

Results and Discussion
The results showed that the VH polarization with a value of 254 km2 performed slightly better in detecting flood areas than the VV polarization with a value of 252 km2. A combined method using both polarizations identified 239 km² of flooded regions by selecting only areas with low backscatter in both polarizations. Combining polarizations improved the accuracy and reliability of flood mapping, offering an advantage over optical data due to radar’s insensitivity to cloud cover and lighting conditions. Temporal analysis of surface water across three periods revealed that water coverage was high (about 1560 km²) before agricultural activities, decreased to 847 km² in late winter, and then increased to approximately 974 km² during the flood event. Despite this lower water extent compared to the pre-agriculture period, a widespread flood occurred due to intense, sudden rainfall and the limited capacity of land and infrastructure to manage the runoff. The MNDWI, with a value of 227 km2, identified the largest area covered by water, while the AWEI, with a value of 126 km2, was more conservative and identified only areas with a very high probability of surface water. By combining all three spectral indices, an area of 62 km2 was identified as flood-prone areas. The variations between the indices are due to the differences in spectral bands and sensitivity to water. Analysis of the results of the NDWI, MNDWI and AWEI showed a significant change in the permanent and temporary water areas during the flood, which increased from 98 km2 to 324 km2. The Otsu thresholding method was effectively applied to standardize and classify water areas across indices.
Conclusion and Suggestions
Overall, it is concluded that the combination of radar and optical data led to the identification of 58 square kilometers of flood areas, with a 39% overlap between the results of the two satellites; this combination significantly increased the accuracy and reliability of identifying flood-affected areas. Since, changes in water extent alone are not a sufficient indicator of flooding, and hydrological and management factors play a more critical role in flood occurrence. Therefore, the use of combined data along with analysis of environmental factors can be highly effective in early warning systems and disaster management planning. This method enables periodic flood mapping and plays a significant role in developing preventive measures, optimizing flood management, and supporting agricultural planning.