Evaluation of coastal erosion by using Landsat data in Ca Mau Cape, Ngoc Hien District, Ca Mau Province

Coastal erosion as well as loss of aquaculture, cultivated and residential lands are of great concern in the Mekong Delta, especially in the Ca Mau Cape region. Historical map data and remote sensing images such as Landsat data combined with GPS field surveys are the optimal methods for assessing shoreline changes including coastal erosion and sedimentation. The 57-year coastline change during the period from 1965 to 2022 has been described through the processing of topographic map data published in 1965 and a series of Landsat-1 to Landsat-8 images collected in 1973, 1993, 2003, and 2022. The study found that the northeast of the study area in Cho Thu hamlet, Tam Giang Tay commune was the place where the largest erosion reached an annual average of 42.4 m/year. Due to the development program of afforestation, largest sediment reaches an annual average of 88.7 m/year in Ca Mau Cape National Park in Lach Vam hamlet, Dat Mui commune. However, during the 57-year period, coastal erosion has caused a loss of 1,714 hectares of land or 2.60% of the total area. Results from this work also suggest that the accuracy of the shoreline studies should be improved by using higher satellite data spatial resolution and the GPS system with dual frequencies for Real-time Kinematics (RTK) for coordinate positioning measurement.


Introduction
In nature, there is a continuous process of wearing of land called erosion.This erosion is transferred to different areas and natural ecosystems.Coastal erosion involves sediment removal from dunes caused by ocean currents and waves [1].Waves are one of the visible erosive agents.However, tides and fauna also play an important role in the erosion process.This type of erosion can also result in collapsed cliffs, damage to houses and beaches, and destruction of coastal facilities and mangroves.Besides the main cause of the continuous erosion of the coastline due to natural phenomena such as waves and ocean currents [2], some living organism contribute to the erosion of an ecosystem.Human factors include vegetation clearing, dredging, harbor development, land reclamation, and climate change [3].A shoreline is known as the interface between water and land [4].Many researchers have applied remote sensing image data to evaluate the quantitative loss of land and shoreline changes [5].Remote sensing technology and Geographic Information System (GIS) provide better morphological characteristics of the offshore environment and historical shoreline measurements [6].In the Southeast Asia region, Landsat image with a spatial resolution 30 m/pixel was used for visual interpretation of the coastline changes [7].Since 1972, Landsat satellite images and other satellite images are taken in the infrared wavelength range, land and water are clearly separated in this wavelength range.Remote sensing and image processing techniques offer alternative solutions to coastal erosion problems [8].
IOP Publishing doi:10.1088/1755-1315/1247/1/012009 2 Nowadays, with the integration of remote sensing technology and GIS, monitoring and calculating shoreline changes is done rather quickly and efficiently.In the past, shoreline assessment was conducted by field survey method with traditional tools and equipment, but now we can perform shoreline measurement with modern equipment such as a global positioning system GPS, or camera system and information sources from satellite images.The data source from satellite images will be input into a GIS for semi-automatic processing, analysis, and assessment of shoreline changes [9].Some previous works on Ca Mau Cape in the short period as 1995 -2010 due to historic data unavailable [10], additionally, other studies have not presented detailed quantitative shoreline changes at commune level [10].The results of statistic measurements were verified with field survey with GPS equipment and local land use planning map.This study is to determine the scale and trend of shoreline changes in order to provide useful information to aid planning strategy at commune level and rational exploitation of estuary and coastal areas; such this study has a meaningful task for both scientific and practical benefits.
Available data sources include the historical maps s such as the topographic map at scale 1:50,000 issued in 1965 under the authorization of the Department of Defense of the Unites State of America USA.The map shows the shorelines visible on near-infrared IR aerial photographs.Those IR images showed a sharp contrast between land and water in order to identify for identify and map the shoreline in the topographic maps.This is a valuable source for historic analysis of surface topography, land use and land cover, and human-related features.This study uses the topographic map at scale 1:50,000 for generating the shoreline in 1965.Landsat-1 was launched in 1972 with the sensor Multi-spectral scanner (MSS) within the visible near-infrared wavelengths and the imaging coverage at 185 km x 185 km.The Landsat-1 has spatial resolution at 60 m/pixel while the Landsat 5 Thematic Mapper (TM) has better spatial resolution at 30 m/pixel for the multi-spectral bands and 90 m/pixel for the thermal band.In this study, Landsat-1 data acquisition on 03 January 1973 was used for extracting the shoreline in 1973 and Landsat-5 data collection on 20 January 1993 was employed for delineating the shoreline in 1993.Landsat-7 Enhanced Thematic Mapper (ETM) images taken on 05 April 2003 and Landsat-8 acquired on 18 January 2002 with the Operational Land Imager (OLI) and the Thermal Infrared Sensor (TIRS) was used for identifying the shoreline in 2003 and 2022, respectively.Information from existing land use maps and GPS devices was used for the field survey to evaluate the accuracy of the results.
The distinctions of this research are the analysis of a long period (57 years) for the coastline changes, the quantitative comparison in a more detailed scale (commune level), and using GPS device for field survey to verify remote sensing data while most of the previous researches for the region lack the above.

Method and Materials
Remote sensing, GIS, and GPS data were integrated to detect shoreline change during the 57-year period in the Ca Mau Cape area is Ngoc Hien district, Ca Mau province.

The study area
Ca Mau Cape is a place with highly sacred meaning not only for Ca Mau people but also for any Vietnamese.The study area: Ca Mau Cape region is located in Ngoc Hien district, Ca Mau province(Figure 1).This is the southernmost land of the country, about 100 km from the center of Ca Mau city.Surrounded by the East Sea and the West Sea (the Gulf of Thailand), the study area is an ecotourism centre with resorts and sightseeing opportunities, as well as an object of research on mangrove ecosystems belonging to the worldwide biosphere reserve, recognized by UNESCO in May 2009.Ngoc Hien district has an area of (Figure 1).,These communes areas analyzed for shoreline changes in study each period.
Weather and climate such as wind pattern influence wave and sea currents that cause coastal erosion.The tide in Ca Mau Cape occurs twice a day.The average hourly wind speed in Ngọc Hien is subject to significant seasonal variation over the course of the year.The windier part of the year lasts for 4.4 months, from November 12 to March 24, with average wind speeds of more than 19.3 km per hour (Figure 2).The windiest month of the year in Ngọc Hien is January, with an average hourly wind speed of 24.0 km per hour.The calmer time of year lasts 7.6 months, from March 24 to November 12.The calmest month of the year in Ngọc Hien is May, with an average hourly wind speed of 14.6 km per hour.

Data used
Four historical maps such as the topographic map at scale 1:50,000 compiled in 1965 under the authorization of the Department of Defense of the USA were used.This map series has the datum of Everest 1930 that should be converted into the standard datum of World Geodetic System WGS84, the names, map series, and Id are listed in the Table 1.A shoreline with slope refers to the angle of the land where it meets the water.The slope of a shoreline can vary depending on the type of coastal landform present, such as a rocky cliff, sandy beach, or marshy wetland.The slope of a shoreline can also be affected by human activities such as dredging, land reclamation, and construction of seawalls.The bathymetric data from Navionics (www.navionics.com)were used to identify the slope and terrain along the shoreline as shown in Figure 3.This data has been derived from the digital elevation model (DEM) and vessels users and communities measurement over 30 years.Within 5 km, the terrain of the study area has a very low slope (0.02 degree) with the depth of water at 1.5 m depth (Figure 3) The tidal regime refers to the pattern of tides along a shoreline.Tides are the rise and fall of sea levels caused by the gravitational pull of the moon and sun.The tidal regime of a shoreline can vary depending on the location and geography of the area.Some areas may experience diurnal tides, which occur once a day, while others may experience semi-diurnal tides, which occur twice a day as displayed in Figure 4. Together, the slope of a shoreline and the tidal regime is used to identify the shoreline by using the tidal at the satellite acquisition time at about 10:00 am local time [5,11].The average slope is 0.02 degree based on the bathymetric data from Navionics, and tide water level is about at 1m based on the local tidal regime (Figure 4).The data can significantly impact coastal processes such as erosion and deposition.For example, a steep shoreline combined with strong tidal currents may result in more erosion of sediment, while a gently sloping shoreline with weaker tidal currents may lead to more deposition of sediment.Understanding the characteristics of a shoreline, including its slope and tidal regime, is important for managing and protecting coastal areas.

Data processing and analysis
Remote sensing, GIS, and GPS data integrated to detect shoreline change during the 57-year period is described in Figure 6.The software used for remote sensing data processing is IDL-ENVI Ver.5.3 and GIS analysis is Quantum GIS Ver.3.2.4.The GPS device is iPhone SE with 4G internet mobile networks and Apple Apps as Google Map for boat navigation with Compass and Measurement App for GPS data collection.

Figure 6. Description of data processing and analysis
Landsat data required radiometric correction, as we use data with different acquisition dates, the spectral characteristics from the two scenes should be corrected for the next steps of processing.Geometric correction has been also applied for Landsat-1 as this data is too old and high-level data processing is not available from the USGS data service.Mosaicking two scenes of Landsat is required because the study area lies between the different scenes (row 136 / path 54 and row 126 / path 54) as seen in Figure 5.
Atmospheric correction was conducted for the Landsat data used in this study , and then the Normalized Difference Water Index (NDWI) was applied as it is widely used in surface water monitoring from satellite data.The index uses two spectral channels, Green (green) and Near Infrared (near infrared) of remote sensing data.
The general formula are NDWI = (NIR -SWIR) / (NIR + SWIR) NDWI (Landsat 8) = (B3 -B5) / (B3 + B5) NDWI (Landsat 4 to 7) = (B2 -B4) / (B2 + B4) The NDWI values correspond to the following ranges: 0,2 -1.0: Water surface, 0.0 -0.2: Flooding, humidity, -0,3 -0.0: Moderate drought, non-aqueous surfaces, -1.0 --0.3:Drought, non-aqueous surfaces The 2022 land use planning map of Ngoc Hien district was found on the website of the Office of Natural Resources and Environmental Management of the Ca Mau province.This data is in raster standard format as JPEG file with the grids of VN2000 map projection.This map was then geo-rectified and transformed into WGS84 with UTM map projection zone 48 North by using 11 ground control points with a polynomial level 2 algorithm.This map could be matched with the data set of Landsat 8 and Google image data by using QGIS Version 3.14 software.The 2022 land use planning map was used for visual interpretation and comparison with the detected shorelines (Figure 7).The field survey was conducted from December 9, 2022 to December 16, 2022.The oil-engine boat has been used for field surveying.The positions of checkpoints were measured along the shorelines and riverbanks using GPS with Google Maps navigation.As it is impossible to stand on the shoreline for positioning measurement, to solve this problem, we use offset mode positioning measurement by using azimuth angle from North direction (Compass App) and distance (Measurement App) the position and the target point as explained in Figure 8. Accuracy assessment: Root mean squared error (RMSE) is the square root of the mean of the square of all of the errors of shoreline extraction from the satellite data as Landsat-8 18 Jan.2022 with the GPS field survey data [12].It is the measure of how well an extraction shoreline fits the checkpoints that have been collected from GPS device during the field survey.
The Root square error RSE in the North and East directions of 83 checkpoints are also visually described as the formula Eq.1 The formula Eq. 2 for calculating RMSE for the North and East directions is: where, Satellite i = The point data extracted from satellite data for the ith observation.Actual j = The actual (GPS data) value for the jth observation N = The total number of observations.

Results and discussion
The results of data processing and analysis show significant erosion along the coastline during the 57year period.The total area measured for 1965 was 65,803.26hectares and reduced to 64,089.29 hectares in 2022.In other words, 1,714 hectares of land was lost due to coastal erosion during the period, and the rate of erosion is estimated at 42.4 m/year in Cho Thu hamlet, Tam Giang Tay commune.However, the largest sediment loss reaches an annual average of 88.7 m/year in Ca Mau Cape National Park in Lach Vam hamlet, Dat Mui commune.

Accuracy assessment
The results of field survey is 83 checkpoints of positioning collected with the GPS device.These checkpoints have adequate distribution (Figure 10).The Root mean square error RMSE between the extracted shoreline from Landsat-8 data dated 18 Jan.2022 and the GPS field survey data is 12.04 m in North and 9.36 m in the East with the confidence level (95.0%) at 1.50 and 1.43, respectively (Table 2).
For more detailed descriptive statistics data of Root square error (RSE), the errors are shown in Figure 11 as most of the errors are located in the center of the chart with an area within 20 m (less than the one-pixel size of the Landsat-8 image.There are some points that are less than 50 m in the North and 35 m in the East direction due to the GPS signal error and boat movement as well as other environmental limitations. Statistical results, analyzed over many years by remote sensing images and GIS technology, indicate that coastal erosion in Ngoc Hien District occurs in the rainy season, after heavy rain at the time of low tide.The new way to use simple GPS available in the smartphone (iPhone or Android) can be applied for field surveys to evaluate the accuracy of shoreline detected from the satellite data for coastal environmental change monitoring.
For further research, the accuracy mapping for the shoreline should be improved by using higher satellite data spatial resolution and the GPS system with dual frequencies for Real-time Kinematics (RTK) for coordinate positioning measurement.The GPS system used in this research is approximate 10m error while RTK system has less than 0.1m error in both horizontal and vertical measurement.

Figure 1 .
Figure 1.Map of the study area and the Mekong Delta with seven communes in Ngoc Hien district, Ca Mau province (Source: OpenStreetMap).

Figure 2 .
Figure 2. The average of mean hourly wind speeds (dark gray line) in Ngoc Hien district (Source Weather Spark)

Figure 7 .
Figure 7.The 2022 land use planning map 2022 geo-rectified and transformed into WGS84 with UTM map projection zone 48 North (Source: MONRE Ca Mau province)

Figure 8 .
Figure 8. GPS data collection for checkpoint positioning of the shoreline in offset mode measurement (Source: the author)

Figure 11 .
Figure 11.RSE of field survey of 83 checkpoints and their distribution in North and East directions

Table 1 .
Description of data used