Abstract
This study assessed the performance of remote sensing and geographic information system techniques for analysing flood events in agricultural areas and estimating flooded areas in agricultural lands. The Berdan Plain in the southern part of Turkey is below sea level and prone to flood events. The flooded agricultural areas were estimated by using Sentinel-2 satellite data for the Berdan Irrigation Scheme from December 2019-January 2020. The flooded area in the old Aynaz marsh was estimated as 486.79 hectares on 9 January 2020, while the total flooded area in the Berdan Irrigation Project was 4,515.58 hectares. Flood volume was calculated as 2.61 hm3 and the highest water level was estimated to be −0.76 metres in the old Aynaz marsh.
HIGHLIGHTS
Flood inundation maps were produced by using Sentinel-2 images.
Modified normalised difference water indices were used for inundation area detection.
Dynamically monitoring the extent of the flood area during the event is possible with remote sensing.
Graphical Abstract
INTRODUCTION
Disaster management practices are activities intended to minimise the damage caused by disasters by preventing destruction of the environment and loss of life and property. Flood events are one of the most dangerous disasters and are experienced frequently in different residential areas of the world. Floods can occur due to natural causes or human activities. Some causes of flooding are heavy rainfall, storms, snow melting, vegetation issues, or dam failure. Flood control studies start by establishing monitoring stations for rainfall and water levels. The second step is to develop flood early warning systems. To avoid negative consequences, decision-making processes should also take into account assessments of the drainage systems efficiency, the impact of climate change and the location of settlements and irrigation areas. Sediment load in the rivers reduces the volume of the cross sections in time. This negative effect should also be considered for all areas vulnerable to floods.
Flood mapping is a useful tool for better understanding the flood regime, and contributes to reducing its negative effects (Dinh et al. 2019). Flood risk assessment and flood risk mapping are the technical activities carried out in pre-flood planning. Various studies on planning and monitoring flood events are available in the literature. Several studies used remote sensing (RS) and geographic information system (GIS) techniques to evaluate the effects of floods on agricultural production, human life, infrastructure, income status, property, and settlements.
As the need for water resources increases due to population growth and climate change, sustainable water resources management has increasing demands for new methodologies and, in parallel, new technologies such as RS and GIS. Using satellite data makes it much easier to determine the boundaries of large flood areas, eliminating the need for time-consuming and labour-intensive fieldwork. Many studies have been conducted to assess the monitoring of dam reservoirs (Akgül & Çetin 2018), investigation of flood effects (Zoka et al. 2018; Akgül & Çetin 2019; Çoşlu & Sönmez 2019; Gülbaz 2019), analysis of flood risk zones (Sufiyan & Zakariyab 2018), flood analysis (Özcan 2017), investigation of flood hazard (Baishya & Sahariah 2017), investigation of flooded areas (Şorman & Doğanoğlu 2001; Hazır et al. 2016; Vishnu et al. 2019; Goffi et al. 2020), mapping of flooded wetlands (Solovey 2020), monitoring of changes in water bodies (Abazaj 2020), and estimation of flood duration (Rättich et al. 2020). Phiri et al. (2020) presented a review of research on mapping land cover/use by using Sentinel-2 data to figure out the contribution of Sentinel-2 data to the monitoring and evaluation of land cover or use.
Akgül & Çetin (2018) investigated the surface areas of five dam reservoirs in the Seyhan and Ceyhan basins in Turkey with RS and GIS, and applied the modified normalised difference water index (MNDWI) using images from the Landsat 8 satellite to estimate the surface area of the dam reservoirs. Akgül & Çetin (2019) also used RS and GIS to investigate the effects of floods on agricultural areas in the Akarsu sub-basin in the Lower Seyhan Plain in Turkey, and showed that the MNDWI and Landsat 8 satellite images can be used to estimate inundation areas. Zoka et al. (2018) used RS techniques to monitor and assess flooded areas; they investigated the impact of a flood event on land use and land cover in a study area in Central Greece in May 2016; their study applied four water indices, one of which was MNDWI, and the indexes accurately classified pixels that indicated water. Similarly, Hazır et al. (2016) estimated flood inundation areas on the Amik Plain in Hatay, Turkey; they used the NDWI (normalised difference water index) to detect the flooded areas on the plain. Abazaj (2020) proposed a method for mapping water bodies, using Sentinel-2 images to monitor the water body variations in the Buna River area by detecting them with NDWI. Babagiray & Kalkan (2021) analysed the daily total precipitation data of 18 automatic meteorological observation stations in and around Turkey's Adana City between 24 December 2019 and 8 January 2020; Sentinel-1 data were used to analyse the flooded area; changes in soil moisture due to precipitation were also investigated.
Satellite data should be used to ensure high temporal and spatial resolution for flood mapping. The current study investigated flooded agricultural areas on Turkey's Berdan Plain between December 2019 and January 2020, using Sentinel-2 satellite data in the scope of flood mapping to support flood management activities by analysing the flooded agricultural parcels. The study area has been used for agricultural production although it is prone to flooding (see Figure 1).
MATERIAL AND METHODS
The study area is located between Berdan Dam and the Mediterranean coastal region between 36°44′00″ and 36°55′20″N latitudes, and 34°50′45″ and 35°18′25″E longitudes in the eastern Mediterranean Basin of Mersin City which is in the southern part of Turkey. Berdan Dam was built on the Berdan River in 1984. The main tributaries of the Berdan River are the Pamukluk and Kadıncık streams. The Berdan Dam is operated for irrigation, energy production, drinking water supply, and flood control purposes to enhance the region's economic growth. The area of the irrigation project is 24,940 hectares.
Limited to a narrow coastal area to the west of Mersin, the alluvium expands much further to the east, where the Berdan and Seyhan rivers accumulate large amounts of alluvial sediments (IECO 1966). The Berdan Plain has fertile agricultural areas and the project's water resources are the Berdan river and its tributaries. Sediment from high elevation lands from the north sides along the rivers boosts agricultural productivity. There are agricultural areas below sea level on the coastal plain. Berdan Dam's multi-purpose operation and the benefits of the project delivered in the region have made it possible to timely examine the problems encountered, and it has become an interesting project field for researchers. In the official reports, the variables of the floods’ occurrence in this study area and suggestions have been provided in the framework of assessing flood risk areas by analysing the flood occurrence problem on the Mersin, Tarsus, and Berdan rivers and tributaries (DSİ 2015).
The average annual temperature is 19.1 °C (DSİ 2021). The climate of the Berdan Plain is Mediterranean, so summers are hot and dry, and winters are mild and wet. The topography of the study area, with elevations between −2.8 metres and approximately 3 metres, was investigated in the scope of flood inundation conditions. The entire area below the floodplains is agricultural land, and there are greenhouses, fruit gardens, and crop fields in the inundation area.
Flood control measures have been carried out for long years to prevent flooding and to protect agricultural areas in the Berdan Plain region, which contains the Berdan Irrigation Project as the study area of this research. The maps of 1911 and 1929 indicate that there were inland lakes (Kipert 1911; Bertarelli et al. 1929) before the Aynaz marsh was formed and the Berdan Irrigation Project was initiated (Kipert 1911; Bertarelli et al. 1929). IECO prepared a feasibility report explaining the development stages of the Berdan Project in 1966. As described in this report, much of the irrigable land is below sea level. The Aynaz marsh area is prone to flooding due to insufficient drainage (IECO 1966). Efforts to drain the Aynaz marsh took place between 1958 and 1969, and the marsh was turned into agricultural fields (Akgül 2018). Akgül (2018), provided historical evidence of the geographical changes in the old Aynaz marsh area, and investigated inundation in the area from December 2016 to January 2017 by using synthetic aperture radar data from the Sentinel-1 satellite.
Sentinel-2 satellite data were used for investigating the flood inundation area in the old Aynaz marsh and its surroundings between December 2019 and January 2020. Sentinel-2 satellite images were examined to analyse the 2019–2020 winter floods, since the widest flood inundation area was observed between December 2019 and January 2020. The location of the flood inundation area in the Berdan Project is presented in Figure 1. The same location is illustrated in Figure 2, showing the location of the pumps and drainage canals in the study area on 9 January 2020.
The location of pumps and drainage canals in the inundated study area.
The Sentinel-2 satellites called Sentinel-2A and Sentinel-2B were launched on 23 June 2015 and on March 2017, respectively. The Sentinel-2 mission operates as part of the European Commission's Copernicus program (EOS 2020), and the specifications of its 13 spectral bands are provided in Table 1. The Sentinel-2 data is accessible via the European Space Agency (ESA 2015).
Sentinel-2 satellite band specifications (ESA 2015)
Band Number . | Band name . | Central Wavelength (nm) . | Bandwidth (nm) . | Resolution (m) . |
---|---|---|---|---|
1 | Coastal Aerosol | 443 | 20 | 60 |
2 | Blue | 490 | 65 | 10 |
3 | Green | 560 | 35 | 10 |
4 | Red | 665 | 30 | 10 |
5 | Vegetation Red Zone | 705 | 15 | 20 |
6 | Vegetation Red Zone | 740 | 15 | 20 |
7 | Vegetation Red Zone | 783 | 20 | 20 |
8 | Near Infrared-NIR | 842 | 115 | 10 |
8b | Narrow Close Infrared-NIR | 865 | 20 | 20 |
9 | Water vapor | 945 | 20 | 60 |
10 | Medium Infrared–SWIR–Cirrus | 1,375 | 30 | 60 |
11 | Medium Infrared–SWIR | 1,610 | 90 | 20 |
12 | Medium Infrared–SWIR | 2,190 | 180 | 20 |
Band Number . | Band name . | Central Wavelength (nm) . | Bandwidth (nm) . | Resolution (m) . |
---|---|---|---|---|
1 | Coastal Aerosol | 443 | 20 | 60 |
2 | Blue | 490 | 65 | 10 |
3 | Green | 560 | 35 | 10 |
4 | Red | 665 | 30 | 10 |
5 | Vegetation Red Zone | 705 | 15 | 20 |
6 | Vegetation Red Zone | 740 | 15 | 20 |
7 | Vegetation Red Zone | 783 | 20 | 20 |
8 | Near Infrared-NIR | 842 | 115 | 10 |
8b | Narrow Close Infrared-NIR | 865 | 20 | 20 |
9 | Water vapor | 945 | 20 | 60 |
10 | Medium Infrared–SWIR–Cirrus | 1,375 | 30 | 60 |
11 | Medium Infrared–SWIR | 1,610 | 90 | 20 |
12 | Medium Infrared–SWIR | 2,190 | 180 | 20 |
Sentinel-2 data have been used in various research studies of earth sciences and atmospheric sciences in recent years, such as monitoring and evaluation studies on land use or land cover (Szostak et al. 2017; Miranda et al. 2018; Nezhad et al. 2019), flood mapping (Dinh et al. 2019; Rättich et al. 2020), wetland monitoring (Solovey 2020), and agricultural activities (Zhang et al. 2019; Kobayashi et al. 2020). The Sentinel-2 satellite that was used in the study has several advantages, including a 10-day revisit time at the equator and a 5-day revisit time at mid-latitudes, as well as a spatial resolution of 10 m.
Although they have the disadvantage of being affected by cloudiness, optical imaging satellites are preferred for flood studies such as this one to radar satellites, which are not affected by cloudiness but make it difficult to analyse and access data.
Using free access satellite data is a low-cost method of analysing flooded areas, while ground-based measurements and field surveys have high costs and are labour-intensive.
Sentinel-2 data are pre-processed by ENVI 5.3 software. Atmospheric corrections were performed using the QUAC (QUick Atmospheric Correction) module, while radiometric corrections were calculated according to Canty (2014).
CHIRPS (Climate Hazards Group InfraRed Precipitation with Station data satellite precipitation data) was used to verify rainfall intensity in the Berdan Plain. CHIRPS is a year quasi-global rainfall data set more than 35 years record in the public domain (https://www.chc.ucsb.edu/data/chirps).
CHIRPS-based data have been used in various research studies in different countries around the world in recent years (Paredes Trejo et al. 2016; Ayoub et al. 2020). Aksu & Akgül (2020) compared CHIRPS-based precipitation data throughout Turkey with measurements of precipitation data at meteorological stations to assess the usability of CHIRPS estimates and proved them usable for hydro-climatological studies.
The ground truth rainfall data for the study area were obtained from the https://rp5.md/Weather_archive_in_Mersin website. Precipitation data from weather station site number 17340 between 21 December 2019 and 10 January 2020 were used to estimate the size of the inundated agricultural irrigation areas.
RESULTS AND DISCUSSIONS
The Berdan River flood capacity was designed as 600 m3/s for the region between Berdan Dam and D-400 Highway Bridge while it is 1,200 m3/s between D-400 Highway Bridge and where the river reaches the Mediterranean Sea. However, the flood capacity has decreased over the years due to sediment load from the Kusun stream system (DSİ 2015).
The temporal variation of precipitation in Mersin is shown in Figure 4. CHIRPS-based precipitation data for 7 January 2020 in the Berdan Plain show the spatial variations over the study area (Figure 5). The figure demonstrates that the precipitation product indicates significant rainfall in the study area. After three small peaks in precipitation amounts (Figure 4), a much higher peak denotes floods due to heavy rainfall on 27 December 2019, 1 January 2020, 9 January 2020 and 11 January 2020 (Figure 6). As presented in Figure 6, the agricultural areas were flooded during this period.
Daily precipitation data observed at weather station 17340 during the study period.
Daily precipitation data observed at weather station 17340 during the study period.
Map of CHIRPS-based daily precipitation for 7 January 2020 on the Berdan Plain.
Sentinel-2 images of flood spread areas on the Berdan Plain on (a) 27 December 2019, (b) 1 January 2020, (c) 9 January 2020, and (d) 11 January 2020.
Sentinel-2 images of flood spread areas on the Berdan Plain on (a) 27 December 2019, (b) 1 January 2020, (c) 9 January 2020, and (d) 11 January 2020.
Sentinel-2 satellite data between December 2019 and January 2020 were evaluated for the case study. The dataset for 9 January 2020 has the widest flood inundation area and cloud-free conditions and was preferred for the analysis. The records show that the flooded agricultural areas were even larger on 16 January 2020. Digital cadastral data were used to determine that the number of parcels affected by the flood on 9 January 2020, in whole or in part, was 2491 and that 577 of the parcels were in the old Aynaz flood area. The inundated area in the old Aynaz region was determined as 486.79 hectares, while the total flood spreading area was 4,515.58 hectares.
A digital elevation model of the old Aynaz marsh produced from the existing maps by Akgül (2018) and elevation-area-volume charts calculated by Akgül (2018) were used to calculate the flood volume for the relevant time in the study area. Parcel boundaries in flood-prone areas were determined in the GIS environment, whether they were affected or not. The agricultural parcels in the flood area were estimated by overlapping digital parcel boundaries with the flood map. The analysis results indicate that the flooded area total was 486.79 hectares and the flood volume was 2.61 hm3.
The highest floodwater surface level in the old Aynaz marsh was determined as −0.76 metres. The flood area and the flooded parcels in the study area are shown in Figure 7. The reports and photographs of the technical survey after the flood event carried by DSİ staff on 8 January 2020 in Berdan Plain were obtained from DSİ archives (Figure 8).
The flooded agricultural areas on the Berdan Plain on 8 January 2020.
The agricultural land area affected by the flood event in Berdan Plain between December 2019 and January 2020 was determined in this study by applying MNDWI to Sentinel-2 data.
Agricultural areas were also affected by floods between the Berdan and Seyhan rivers. It is important that the responsible units regularly carry out maintenance of the drainage channels.
CONCLUSION
Remote sensing methods used in monitoring and analysis of natural disasters such as earthquakes, landslides, and forest fires have been also frequently used in flood management studies because they are fast, economical, and practical.
Emergency assessment to mitigate the loss is of great importance in flood events. It has been demonstrated that the use of Sentinel-2 satellite data acquisition capabilities is more useful than ground measurements and observations since it facilitates decision makers’ flood management activities to provide life and property safety in flood events. However, to minimise loss, a study is recommended of the feasibility of using visual and audio tools to warn inhabitants, especially people engaged in agricultural areas under flood risk as well as the parcels’ owners.
This study determined that 4,515.58 hectares in the study area were affected by the flood. The precipitation measured by the meteorological observation station in the last 17 days before the flood was calculated as 365.1 millimetres in total. The consequences of this study may lead to the formation of infrastructure for an early warning system for flood management.
Vegetation growth in the drainage channels should be prevented, and maintenance service should be regularly performed and this study shows it can be easily monitored.
Rapid damage assessment after flood events is crucial for prompt rehabilitation of first responders and damaged infrastructure in coordination with other flood management activities. The use of satellite data makes it significantly easier to determine the boundaries of large flooded areas, whereas fieldwork requires great effort to be performed in a limited time.
Detecting flood-affected areas during the flood event guides decision-making institutions and organizations in eradicating the flood's consequences. remote sensing methods should be used for this because it is not possible to produce a map of the flood inundation area using terrestrial methods in difficult terrain and conditions and in a limited time. The current study's use of Sentinel-2 satellite images demonstrates that it is possible to dynamically monitor the extent of the flood area and generate solutions while the flood is ongoing.
The study's findings show it is possible to determine which parcels were affected by flood events and how much agricultural insurance will cost. The outputs and information from flood maps based on from satellite data will aid in estimating losses in agricultural products in flooded areas and assessing institutional financial support required.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.