1 Introduction

Sandy beaches may be affected by coastal erosion, decreasing their value for tourism and recreation.

In international context, the regional review of Asian Development Bank (2009) showed that there are four adaptation options on coastal erosion; mangrove conservation and plantation, strengthening existing dike and seawall, better design and standard for construction and provision of information and awareness. A comparative study and review of Krainara (2014); Walter (2018) showed that several countries tend to utilize adaptation options in terms of nature conservation/plantation and soft structure than hard structure. For instance, beach nourishment with some engineering structures in United States of America including conservation of coastal natural process in the Netherlands and France. Moreover, the option of revegetation of coastal forests and trees (as bioshields) combined with some types of soft and hard structures including ecosystem-based adaptation with proper institutional design/arrangement will be one of the adaptation options that are increasingly favored (also in sandy coast/ beach). On the other side, in term of impacts estimation, Addo et al. (2011) projected the expected impacts of sea level rise in three communities in the Dansoman coastal area of Accra, Ghana using the SimCLIM model based on global scenarios, the Commonwealth Scientific and Industrial Research Organization General Circulation Models (CSIRO_MK2_GS GCM), the modified Bruun rule, and aerial photographs taken in 2005. The results show that sea level will rise approximately 60.27 cm (SRES B2) to 79.71 cm (SRES A1FI) in 2100 compared to the period 1970–1990. The Dansoman coastline could recede about 202 m in 2100 relative to 1970–1990. Moreover, 650,000 people, 926 buildings, and a total area of about 0.80 km of land are vulnerable to permanent inundation in 2100.

Thailand has approximately 320,000 square kilometers of maritime zone, 2800 km of shoreline (including the Gulf of Thailand and the Andaman Sea), and 23 coastal provinces (Aquatic Resources Research Institute-Chulalongkorn University, 2011). Furthermore, it has a number of renowned and attractive beaches, such as Mae Phim beach in Rayong province, Patong beach in Phuket province, Pattaya beach in Chonburi province, SaiRee beach in Chumphon province, Railay beach in Krabi province, and Pak Meng beach in Trang Province. Consequently, if these valuable beaches are ruined by coastal erosion, it will affect tourism and the economic system in Thailand (National Research Council of Thailand, 2012; Tourism Authority of Thailand, 2013).

Adaptation is one of the options/strategies to confront the future effects of climate change (also sea-level rise [SLR]-induced beach erosion). It aims to build resilience in several sectors and communities without causing them other problems. In the case of adaptation to sea level rise and beach erosion, three possible options are (1) retreat: people migrating due to coastal erosion; (2) accommodation: impact avoidance by heightening buildings or planting flood/salinity tolerant crops; and (3) protection: seawall/dike construction/strengthening or beach nourishment (Mclean et al., 2001; Nicholls, 2003). Thailand has adaptation strategies for coastal erosion in several areas: Samut Prakarn province, Pak Phanang district, and HuaSai district of Nakhon Si Tammarat province and some provinces along the coast of the Andaman Sea (Boonma & Saelim, 2011; Ritphring et al., 2021; Roy et al., 2023). Nonetheless, the beach erosion situation in Thailand has still not been alleviated.

Previous studies in Thailand have focused mostly on erosion in coastal provinces (local scale) such as Surat Thani, Nakorn Si Thammarat, Krabi, and Phuket (Saengsupavanich et al., 2009; Snidvongs et al., 2008; Thampanya et al., 2006). However, few studies have been conducted on national-scale coastal erosion, and to the best of our knowledge, there is no national estimation of the sandy beach erosion caused by rising sea levels (Department of Mineral Resources, 2001, 2002). Thus, this paper aims to fill this information gap, using the coastal erosion model SimCLIM/CoastCLIM as a tool for estimation.

The main objectives of this study were (1) to forecast the rate of sandy beach erosion and (2) to estimate the impact of sand loss and forced migration due to global/regional SLR at the Thai national level for the period 1940–2100. Moreover, this research also attempts to present an adaptation option resulting from the economic evaluation approach: a cost–benefit analysis (CBA) (with/without adaptation strategies) under various climate change scenarios. The results will contribute to national-scale data on sandy beach erosion in terms of the rate and potential impacts of SLR for future Thai scenarios. The CBA results could aid various stakeholders and local communities as they are forced to adapt to beach erosion in Thailand.

This paper is classified and organized as follows: Sect. 1 introduces the objectives and study areas, Sect. 2 describes the methodology, and Sect. 3 presents the results. Section 4 discusses the findings, notes the limitations, and makes recommendations for further study. Section 5 presents the conclusions. Section 6 presents the limitations of study.

2 The study area

Regarding the study areas, three critical provinces and 27 sandy beaches were selected as the study areas from a total of 18 coastal provinces and 152 sandy beaches in Thailand (Department of Marine & Coastal Resource, 2013d). Rayong is located between 12 and 13° N and 101 and 102° E on the eastern coast of the Gulf of Thailand, while Nakhon Si Thammarat and Trang are located at 8–10° N, 99.15–100.05° E and 7–8° N, 99.20–99.90° E on the western coast of the Gulf and the Andaman Sea coast, respectively (2007b; Eastern Province Cluster Office of Strategy Management, 2007; Land Development Department, 2007a). The locations of the study areas are shown in Fig. 1. In regard to the 27 sandy beaches, 11 are located in Rayong: Pla, Phayun, Namrin, Suchada, Laem Charoen, Mae Ramphueng, Sai Kaew, Phe, Suan Son, Pak Khlong Klaeng, and Mae Phim. Nakhon Si Thammarat contributed 11 beaches: Kanom, Thung Sai, Sichon, Hin Ngam, Piti, Baan Roh, Pothong, Sai Kaeo, Tha Soong Bon, Ban Ko Fai, and Chan Chaeng. The remaining five beaches are located in Trang: Leam Makham, Hua Hin, Khlong Son, Pak Meng, and Chao Mai (Department of Marine & Coastal Resource, 2013d). These three provinces were listed in the 18 critical/vulnerable areas where erosion rates exceeded 5 m/yr, as mentioned by Kraipanon (2010).

Fig. 1
figure 1

Location of the study areas and estimated retreat of coastline (m) in 2100 of the 27 sandy beaches in Rayong (a), Nakhon Si Thammarat (b) and Trang (c) (in the counterclockwise direction)—critical beaches showing in purple and red colors (Data: revised from Department of Marine & Coastal Resource, 2013e, 2013f)

Full size image

In Thailand, critical areas are defined by the coastal erosion rate. Coastal areas with severe erosion rates—more than 5 m/yr—are classed as critical areas. In addition, risky areas have moderate erosion rates between 1 and 5 m/yr (Duriyapong & Nakhapakorn, 2011; Department of Marine and Coastal Resource 2013 h; Prukpitikul et al., 2014). From the recent data (2003–2011), Rayong, Nakhon Si Thammarat, and Trang all had high impacts from erosion at 21.75, 53.46, and 43.7 km of eroded shorelines, respectively (Department of Marine & Coastal Resource, 2011). Furthermore, the three provinces are renowned for beach tourism, with several beautiful and famous beaches such as Mae Ramphueng, Sai Kaew, and Mae Phim in Rayong Province; Hin Ngam, Sai Kaeo, and Kanom in Nakhon Si Thammarat Province; and Pak Meng and Chao Mai in Trang Province (National Research Council of Thailand, 2012; Tourism Authority of Thailand, 2013).

3 Methodology

3.1 The SimCLIM/CoastCLIM model

The input parameters for SimCLIM/CoastCLIM analysis (Table 1) consisted of six site parameters and two storm parameters. Site parameters were shoreline response time (τ), closure distance (l), depth of material exchange (d), dune height (h), residual movement (RM), and vertical land movement (VLM). Storm parameters were storm surge cut mean (SSCM) and storm surge cut standard deviation (SSCSD).

  • Shoreline response time refers to the responsiveness of the coastal system to SLR in a given year and influences the annual change in the shoreline.

  • Closure distance is the distance offshore at which the process of sediment exchange ceases and the sediment is lost.

  • Depth of material exchange is the water depth at the closure distance.

  • Dune height is the frontal dune/berm/beach height.

  • Storm parameters represent random storm characteristics, including storminess (frequency and intensity). These factors determine the erosion potential of the shoreline due to storms in terms of the means and standard deviations of impacts (meters of erosion). Users can select and add values in storm surge cut mean (SSCM) and storm surge cut standard deviation (SSCSD) flexibly, as representative of the mean and standard deviation of erosion potential in any given year.

Table 1 Input Parameters of the 3 study areas for SimCLIM/ CoastCLIM Analysis
Full size table

For the SimCLIM/CoastCLIM model, actual storm erosion was assumed to be 10% of the value selected in the potential one. For analysis with CoastCLIM, all parameters mentioned in Eqs. (1) and (2) were used as input parameters, including another two parameters: residual movement and vertical land movement. Residual movement is the long-term variation in shoreline position (erosion and accretion), which influences trends of sediment supply and transport. Vertical land movement is the change in relative sea level that excludes climate change–related components (e.g., land subsidence or uplift) (CLIMsystems, 2013; Warrick, 1998).

Closure distance and depth of material exchange were obtained from observation data from the Aquatic Resources Research Institute and Chulalongkorn University (2006). Dune height and residual movement were obtained from observation data from the Department of Mineral Resources (2001, 2002). Vertical land movement data were collected from SLR with VLM for cities’ data from CLIMsystems (1995) and observation data from Trisirisatayawong et al. (2011). The VLM values from CLIMsystems data were generated from direct continuous observations of the Global Positioning System—GPS (the SONEL program)—and from trend analysis of tidal observations (the PSMSL program). Due to the lack of observations and secondary data, the default/initial values (1940 values) of the model were applied to shoreline response time and storm parameters.

Sand loss was calculated using Eq. (3), while forced migration was calculated in terms of the area of sand loss multiplied by the average population density per segment. The segment length data were collected from observations and secondary data from the Department of Mineral Resources (2001, 2002), Nakhon Si Thammarat Provincial Governor’s Office (2009), Department of Marine and Coastal Resource (2013a, 2013b, 2013c, 2013d, 2013e), Rayong Provincial Governor’s Office (2014), and Trang Provincial Governor’s Office (2014). The erosion factor of a sandy beach (assuming the value from 1940 remained constant over time) was taken as 1 according to Vafeidis et al. (2004) and Hinkel et al. (2013). The populations of the three study areas were obtained from the Bureau of Registration Administration, Department of Provincial Administration (1994). The areas of the three study regions were obtained from Earthpower (2002), the Nakhon Si Thammarat Provincial Governor’s Office (2009), the Rayong Provincial Governor’s Office (2014), and the Trang Provincial Governor’s Office (2014).

This study utilized the coastal impact model (CoastCLIM) of the Simulator of Climate Change Risks and Adaptation Initiatives model (SimCLIM 2013 version 3.3) to forecast sandy beach erosion due to sea-level rise. CoastCLIM operates following the modified Bruun Rule and focuses on change in the equilibrium shoreline position of a beach-and-dune system due to variation in sea level. The equilibrium shoreline position is readjusted and reestablished landward and eroded when the sea level rises, as shown in Eq. (1). The Bruun Rule was modified by adding the time lag of the shoreline response and the variation in the occurrence of severe stormy seasons, as shown in Eq. (2). The modified Bruun Rule attempts to overcome its two main drawbacks, the inability to estimate the change of yearly actual shoreline position and the failure to consider storm parameters (Warrick, 1998).

$$ {\text{C}}_{{\text{eq}}} = {\text{ z l }}/ \, \left( {{\text{h }} + {\text{ d}}} \right),{\text{ and}} $$
(1)
$$ {\text{dC}}/{\text{dt}} = \, \left( {{\text{C}}_{{\text{eq}}} --{\text{ C}}} \right) \, /\tau + {\text{ S}} $$
(2)

where Ceq is the equilibrium change in shoreline position, z is the rise in sea level, l is the closure distance, h is the dune/berm height at the site, d is the depth of material exchange at the closure distance (l/(d + h) thus gives slope), t is the time (yr), C is the shoreline position relative to t = 0, τ is the shoreline response time, and S is a stochastically generated storm erosion factor.

Sea-level rise and coastline retreat were projected by employing the SimCLIM Model on the basis of two forcing scenarios (RCP2.6 and RCP8.5) for the period 1940–2100. The preselected presumptions were a high climate sensitivity and an averaging of the 24 GCMs. The output scenarios comprise VLM. RCP 2.6 and RCP 8.5 stand for the best and worst case scenarios for analysis in this paper. RCP 2.6 (as the best-case scenario) represents the ‘peak and decline’ pathway of radiative forcing and GHG atmospheric concentration, which peak at approximately 3 Wm−2 and 475–490 ppm CO2-eq in 2050 and decline to 2.6 Wm−2 in 2100. RCP 8.5 (as the worst-case scenario) shows a ‘rising’ pathway of the two parameters, which leads to approximately 8.5 Wm−2 and 1,313–1,370 ppm CO2-eq in 2100 (Intergovernmental Panel on Climate Change, 2014; International Institute for Applied Systems Analysis, 2015; Moss et al., 2010; Yin et al., 2013). These scenarios could be applied as extreme situations (in terms of high and low extreme future climates) for various climate-related analyses. Countries, including Thailand, can use the ‘extreme scenarios’ as input for climate modeling, atmospheric chemistry modeling, and thread and impact analysis for future climate-related planning (International Institute for Applied Systems Analysis, 2009). Nevertheless, it has to be noted that the sea-level rise rate in the 1940–1995 period was fixed to be constant referred to input-historical rate whilst the RCP scenarios and VLM were applied to the rate at 1995 onward. In regards to the model fundamental philosophy and algorithm, the 1940–1995 periods was determined as “warming-up period” for the model estimation in purpose of model calibration as well as errors detection and correction (Warrick, 1998).

Nevertheless, SimCLIM/CoastCLIM was unable to estimate sand loss (SL) and forced migration (FM). Thus, SL was calculated from the following equation, revised from Hinkel et al. (2013), and FM was calculated in terms of area of sand loss multiplied by the average population density per segment.

$$ {\text{SL}} = {\text{s }}*{\text{ C }}*{\text{ E}}_{\text{f}} $$
(3)

where SL is sand loss, s is segment length; C is the erosion rate, and Ef is the erosion factor. Ef stands for the factor used for estimating the proportion of s that is composed of sandy beaches and could therefore be inferred to represent sand supply.

In this paper, the root mean square error (RMSE) and the mean absolute error (MAE) were used together as statistical metrics for the model evaluation and validation process. The combination of these metrics provided a more complete picture for the assessment of model prediction errors. While RMSE is properly used to describe a normal distribution of errors, MAE is suitable for uniformly distributed errors. However, both these metrics have been beneficially used for model performance measurements in the areas of meteorological, climatic, and environmental data analysis (Akpinar et al., 2013; Chai & Draxler, 2014; Chen et al., 2012). The RMSE and MAE were calculated using the following equations:

$$ RMSE = \sqrt {{\frac{1}{n}\sum_{i = 1}^n {e_i^2 } }}, \,\rm{and} $$
(4)
$$ MAE = \frac{1}{n}\sum_{i = 1}^n {\left| {e_i } \right|} $$
(5)

where ei is the model estimation error of n samples (ei, i = 1,2,…,n), equal to the difference between the observed value (oi) and the estimated or predicted value (pi). Moreover, sensitivity analysis was introduced to assess the uncertainty and variation of the two main input parameters (SLR and storminess), which influenced the results (see details in the Discussion section).

3.2 The CBA method

The secondary data used for CBA were collected from several organizations. Two field studies were conducted for data gathering during (1) March 30th to April 2nd, 2016, and (2) June 16th to June 18th, 2016. The cost of adaptation under various scenarios was collected from Prince of Songkla University (2008), Department of Van Rijn (2010), Marine and Coastal Resource (2013), along with field study data. Benefits of each adaptation option/measure which consist of decrease in sand loss, decrease in forced migration, and increase in tourism revenue were calculated using the following equations:

$$ \begin{gathered} DSL = SLA \times LP \hfill \\ \qquad [z*R*E_f ] \times LP \hfill \\ \end{gathered} $$
(6)

where: DSL = Decrease in sand loss (baht).SLA = Sand loss area (km2). LP = Land price as Market price in 2014 (baht).z = The segment length (km). R = Coastal erosion rate (m per year). Ef = Erosion factor.

Remarks: Market prices of land in 2014 were used for calculations since the research was conducted in year 2014.

$$ \begin{gathered} DFPM = NOM \times MCPM \hfill \\ \quad \quad \quad [SLA \times (PETS/AOB)] \times [TRC + HRC + LIC] \hfill \\ \end{gathered} $$
(7)

where: DFPM = Decrease in forced people migration (baht).N OM = Number of migrants (people). MCPM = Migration cost per migrant (baht per person). PE TS = Population employed in (beach) tourism sector (people). AOB = Area of beach (km2). TRC = Transportation cost (baht per person).H RC = House/infrastructure rebuilding cost (baht per person). LIC = Loss of income per capita as opportunity cost of migrating (baht per person).

$$ \begin{gathered} {\text{ITR}} = {\text{NOTA}} \times {\text{EPT}} \hfill \\ {\text{NOTA }} \times \left[ {{\text{FDE }} + {\text{ ACE }} + {\text{ TRE }} + {\text{ REE }} + {\text{ SOE}}} \right] \hfill \\ \end{gathered} $$
(8)

where: ITR = Increase in tourism revenue (baht). NOTA = Number of tourist arrivals (thousand people). EPT = Expenditure per tourist (baht per person). FED = Food and drink expenditure (baht per person). ACE = Accommodation expenditure (baht per person).TRE = Transportation expenditure (baht per person).REE = Recreation expenditure (baht per person).SOE = Souvenir expenditure (baht per person).

Number of tourist arrivals (thousand people) was calculated using the equation below (revised from Hamilton et al., 2004):

$$ {\text{ln Ai}} = {5}.{97 } + { 2}.0{{5 }}\times 10^{ - {7}} {\text{Area}}_{\text{i}} + \, 0.{{22 T}}_{\text{i}} - { 7}.{{91}}\times 10^{ - {3}} {\text{T}}_{\text{i}}^{2} + { 7}.{\text{15}}\times0^{5} {\text{Coast}}_{\text{i}} + \, 0.{8}0{\text{ lnY}}_{\text{i}} $$
(9)

where: ln Ai = Number of tourist arrivals (thousand people).A rea = Land area of sandy beach (km2). T = Annual average temperature (°).C oast = Length of coastline (km). Y = Per capita income. i  = Indexes destination beach.

The three scenarios mentioned in Table 2 for the three study areas were analyzed with the economic evaluation approach (cost–benefit analysis: CBA), including the Do-Nothing Scenario (Scenario 0; Baseline Scenario), by using the four steps of CBA (revised from Tubpun, 1998).

Table 2 Scenarios analyzed with the Economic Evaluation approach
Full size table

The four steps of CBA to analyze the costs and benefits of each option (revised from Tubpun, 1998) consist of the following: (1) Study and identify the boundary and objective of each option; (2) identify and measure the cost and benefit of each option in each year; (3) evaluate the costs and benefits in monetary terms; and (4) compare the costs and benefits in terms of net present value (NPV), benefit–cost ratio (B/C ratio), and internal rate of return (IRR).

The total yearly benefits for each adaptation option/measure were summations of decrease in sand loss, decrease in forced people migration and increase in tourism revenue. The calculation was based on Eqs. (6–9) which were basely estimated by difference of erosion-rate avoidance per year (assumed to be constant) for each option/measure. Furthermore, the fundamental theoretic idea behind the benefit and cost calculations was referred to below equation (Costa et al., 2009).

$$ {\text{Tb}} = {\text{Ai }} - {\text{ Ac}} $$
(10)

where: Tb = Total theoretic benefit of adaptation option to sea-level rise. Ai = Avoided impacts. = IBAU – IADP. I = Sum of sea flood, river flood, salinity intrusion, land loss and people migration costs. IBAU = Costs under business-as-usual scenario. IADP = Costs under adaptation scenario. Ac = Cost of adaptation. = AADP–M.M = Cost of maintenance option under business-as-usual scenario. AADP = Cost of adaptation under adaptation scenario.

The fundamental assumptions of the CBA in this study were determined as follows:

  1. (1)

    The number of tourists increases with population and income, and tourists prefer holidays at a temperature of 25 °C. There is no beach tourism if the warmest day of the month is below 15 °C; 65% of tourists spend their holidays in coastal areas (sandy beaches in a strip 1 km length); 25% of their expenditure is profit (Hinkel et al., 2013).

  2. (2)

    Other variables in Equation (8) were assumed to be constant over the time of analysis (temperature, length of coastline, and per capita income) in order to investigate the effects of beach erosion on beach-related tourism (through change in sandy beach area).

  3. (3)

    The boundary of analysis is the subdistrict (of each study area), and the main objectives of each adaptation option are to alleviate beach erosion problems and not to affect tourism activities (through the beach landscape).

  4. (4)

    The project lifetime is 22 years—initiated in 2014. It includes 1 year (2014) for preparation, 1 year (2015) for construction, and 20 years (2016–2035) for operation. In case of beach nourishment without integration of other adaptation options/measures, sand will be refilled every 5 years (Department of Marine & Coastal Resource, 2013a, 2013b, 2013c, 2013d, 2013e).

4 Results

4.1 Historical trends

Empirical studies with regard to sea level change and coastal erosion are represented by the observation data of the Hydrology Section Engineering Bureau, Marine Department () and the Department of Marine and Coastal Resource (2008; 2009; 2013 g,h). The observed SLRs of Rayong (1978–1995), Nakhon Si Thammarat (1986–1995), and Trang (1968–1992) were −0.37 to 0.12, −0.28 to 0.55, and 0.11 to 0.39 cm/yr, respectively. The observed coastal erosions of Rayong (1952–1995), Nakhon Si Thammarat (1952–1995), and Trang (1974–1995) were 1.40–5, 1.74–8, and 2.50–3.50 m/yr. The average values of SimCLIM/CoastCLIM estimation in the 1940–1995 period for the three provinces were 0.12, 0.14, and 0.13 cm/yr for SLR and 3.61, 5.33, and 5.27 m/yr for erosion, respectively. For the same period of observation data, the estimated values of SLR and beach retreat were 0.13–0.14, 0.14–0.15, and 0.14 cm/yr and 0.63–7.65, 1.50–1.48, and 2.95–7.33 m/yr, respectively. Moreover, the estimated values of sand loss for the three provinces in the period 1940–1995 were 0.03, 0.04, and 0.02 km2/yr, while around 18, 15, and 3 people/yr were forced to migrate (Table 3).

Table 3 Historical Trends of Sea-Level Rise and Retreat of Coastline of the 3 Study Areas (1952–1995)
Full size table

To validate/calibrate the model for future projections of SLR and sandy beach erosion, RMSE and MAE were introduced using Eqs. (4) and (5) along with the historical data mentioned above. RMSE and MAE represent the difference between the actual observed values and the estimated values; they also describe the accuracy of the model prediction. In this paper, the RMSE and MAE of sea-level rise were 0.17 to 0.33 and 0.14 to 0.30 cm/yr, while the values for sandy beach erosion were 1.06 to 1.94 and 0.75 to 1.90 m/yr. In the ideal case, these two values should be closer to 0, indicating a higher accuracy of model prediction. Thus, the accuracy of SimCLIM/CoastCLIM prediction was satisfactory and reliable compared to the high accuracy level (0.5–2 of the referred unit) mentioned by Marghany (2013) and Murdukhayeva et al. (2013).

5 Future projections

The estimated values for future scenarios resulting from analysis with the SimCLIM/CoastCLIM model are exhibited separately in two parts: (1) sea-level rise and sandy beach erosion and (2) sand loss and forced migration.

5.1 Sea-level rise and sandy beach erosion

In the best-case scenario (RCP2.6), with a peak-and-decline pathway of radiative forcing and GHG atmospheric concentration, global temperature will rise 1.79 °C by 2100 compared to 1995 levels (SimCLIM/CoastCLIM data). This will result in an increase in sea level (accompanied by local VLM, as shown in Table 1) of the three study areas. The estimated values of the 106-year sea-level rise are 0.13–83.48 cm (compared to 1995 levels), as shown in Fig. 2. The future scenarios show that sea level will rise 18.01, 38.70, 61.66, and 83.48 cm in 2025, 2050, 2075, and 2100, respectively (Table 4). At this rate of SLR (accompanied by the stochastic storminess shown in Table 1), the current shoreline (sandy beach) in long-term estimation (the 1995–2100 period) will be eroded, resulting in landward migration. The estimated values of the three study areas range from 1.09 to 675.58 m (Fig. 2). The sandy beach erosion situation (provincial average values) in future scenarios tend to become worse, and the shoreline is estimated to retreat 107.30, 216.41, 324.91, and 436.11 m by 2025, 2050, 2075, and 2100, respectively.

Fig. 2
figure 2

Estimated sea-level rise (cm/yr) (a) and retreat of coastline (m/yr) (b) for the best-case scenario (RCP2.6) of the study areas during the 1995–2100 period (Note: RA, NA and TR stand for Rayong, Nakhon Si Thammarat and Trang)

Full size image
Table 4 Estimated Sea-Level Rise and Retreat of Coastline of the 3 Study Areas (1995–2100)
Full size table

In the worst-case scenario (RCP8.5), with continuous increase of the two parameters, global temperature will increase by 5.4 °C by 2100 compared to the 1995 level (the model data). Sea level also shows an increase in the three study areas during the 106-year period, but with higher magnitude. Figure 3 shows that the estimated value of 106-year SLR could reach 124.38 cm by 2100 (compared to 1995 levels). The future scenarios suggest that the sea level will rise 19.15, 43.25, 78.70, and 124.38 cm by 2025, 2050, 2075, and 2100, respectively (Table 4).

Fig. 3
figure 3

Estimated sea-level rise (cm/yr) (a) and retreat of coastline (m/yr) (b) for the worst-case scenario (RCP8.5) of the study areas during the 1995–2100 period

Full size image

At this rate of SLR, the sandy beach of the three study areas in the long-term estimation will be highly eroded at 747.05 m by 2100 (Fig. 3). The sandy beach erosion situation (provincial-average value) in future scenarios tends to become worse, and the shoreline shows retreat of 108.27, 224.10, 353.13, and 507.90 m by 2025, 2050, 2075, and 2100, respectively.

Regardless of the scenario, Rayong is the least risky province in terms of sandy beach erosion, at 16.72% below the average value of the three study areas, while Trang is the most affected, at 17.62% above the average value in terms of the change in sandy beach position by 2100. According to Figs. 2 and 3, six sandy beaches have the lowest risk of erosion—Sai Kaew, Thung Sai, Sichon, Hin Ngam, Piti, and Chao Mai—while Mae Phim, Ban Ko Fai, Chan Chaeng, and Pak Meng are the most highly affected. Furthermore, Ban Ko Fai, Chan Chaeng and Pak Meng are defined as critical areas with the highest erosion, exceeding 900 m by 2100, or equivalent to 5–6 m/yr (Fig. 1).

5.2 Sand loss and forced migration

Sand loss and forced migration are the major impacts from sandy beach erosion. In regard to rapid short-term erosion from stochastic storminess, loss of coastal land and population could follow in an ad-hoc manner. Thus, the researcher applied a constant value of beach length and population density in the base year (or the nearest year) as the baseline to calculate changes in these two impacts.

In the best-case scenario (RCP2.6), with 675.58 m of erosion of sandy beach positions, the value of change in the land area of the three study areas over the period of 1995–2100 varied from 0.01 to −5.20 km2 (Fig. 4). This led to a change in the coastal population for the same period of between 1 and 2431 people. A simulation of the future impacts showed that the sandy coastal area (provincial average value) will be reduced by 0.46, 0.92, 1.38, and 1.85 km2 by 2025, 2050, 2075, and 2100, respectively (Table 5). The estimated values of the reduction in coastal population were 200, 402, 591, and 778 people over the same period.

Fig. 4
figure 4

Estimated sand loss (km2/yr) (a) and forced people migration (people/yr) (b) for the best-case scenario (RCP2.6) of the study areas during the 1940–2100 period

Full size image
Table 5 Estimated Sand Loss and Forced People Migration of 3 Study Areas (1995–2100)
Full size table

In the worst-case scenario (RCP8.5), with 747.05 m erosion of sandy beaches, the value of change in the land area by 2100 (Fig. 5) could reach 5.61 km2 compared to 1995 levels. In the 161-year period, the estimated value of reduction in the coastal population of the three study areas was higher, reaching 2,753 people by 2100. On the provincial scale, the sandy coastal area will be reduced (average values) by 0.46, 0.95, 1.50, and 2.15 km2 by 2025, 2050, 2075, and 2100, respectively. This will reduce human settlement by 202, 412, 628, and 873 people over the same period.

Fig. 5
figure 5

Estimated sand loss (km2/yr) (a) and forced people migration (people/yr) (b) for the worst-case scenario (RCP8.5) of the study areas during the 1940–2100 period

Full size image

In 2100, for both the RCP2.6 and the RCP8.5 scenario, Nakhon Si Thammarat was the least affected province, at 11.53% below the average values of land variation, while Trang showed the highest impact at 22.22% above average values (Table 5). Trang had the least impact on the coastal population, at 67.59% below the average values, while Rayong was the most affected province at 73.04% above the average values in terms of reduction in population by 2100 (Table 5). For the beach-scale analysis (Figs. 4 and 5), Sai Kaew, Pothong, and Leam Makham were the lowest-risk beaches, while Mae Ramphueng, Kanom, and Pak Meng were three beaches with the highest erosion. Furthermore, people living near Pla, Pothong, and Leam Makham had the lowest probability of migration, while several households in Phayun, Sichon, and Pak Meng faced the highest risk of displacement.

5.3 The adaptation option

Regarding the SimCLIM/CoastCLIM results, Mae Phim (in Kram subdistrict of Rayong province), Ban Ko Fai (in Khanap Nak subdistrict of Nakorn Si Thammarat province) and Pak Meng (in Mai Fat subdistrict of Trang province) were selected as the study areas for beach-scale analysis of adaptation options using the CBA method. The researcher gathered various kinds of data from field studies during March to June 2016, such as current adaptation options, erosion situation and rate, tourism activities, and transportation and livelihoods of native people, including some data for the CBA. These data affirmed that these beaches were better suited than others as the study area for this analysis. In Rayong province, Mae Phim beach was selected over another potential choice (Mae Ram Phueng beach) since it has more evidence of erosion and some adaptation options, including higher tourism activities (Fig. 6). Moreover, in Nakhon Si Thammarat province, Ban Ko Fai was selected over Chancheng beach due to its high erosion rate and various evidence of erosion although it has no tourism activities in the present time (Fig. 7). Lastly, in Trang province, Pak Meng beach was selected because it has high tourism activities and more evidences of erosion than Chao Mai beach (Fig. 8).

Fig. 6
figure 6

Tourism activities (a) and evidences of erosion/adaptation options (b) in Mae Phim beach

Full size image
Fig. 7
figure 7

Evidences of erosion (a) and adaptation options (b) in Ban Ko Fai beach

Full size image
Fig. 8
figure 8

Evidences of erosion/adaptation options (a) and tourism activities (b) in Pak Meng beach

Full size image

The values shown in the below tables were calculated using two discount rates: social rate of time preference (SRTP) and social opportunity cost rate (SOCR) (for details, see Tubpun, 1998). In practical ways of application, SRTP and SOCR are represented by the interest rate of government bonds and the minimum loan rate (MLR) after tax and inflation rate deduction in 2014. The values applied in this research were 3% and 6% for government bonds and minimum loan rate (MLR), respectively (Bank of Thailand, 2014, 2016). Moreover, the erosion rate in years 1 and 2 refer to the rate in scenario 0 (Do-Nothing Scenario). Benefits are the different values of each adaptation option in scenario 0.

Table 6 shows that on Mae Phim beach (Kram subdistrict, Rayong province), scenario 1—current adaptation option with stone sea wall—and scenario 3—potential adaptation option with artificial reef—should probably be selected for implementation. The stone sea wall option has the highest NPV (518.31 and 354.88 million baht for 3 and 6% discount rates, respectively), and the artificial reef option has highest benefits per unit of costs (7.10 and 5.47 for 3 and 6% discount rates). In the area of Ban Ko Fai beach (Khanap Nak subdistrict, Nakorn Si Thammarat province), scenario 3—potential adaptation option: hard-structural option (artificial reef)—could be promoted because it has the highest NPV (470.27 and 325.98 million baht), B/C ratio (7.22 and 5.57), and IRR (48.52%). However, scenario 2 with the beach nourishment option should be avoided: It not only has negative values of NPV (−113.32 and −88.45 million baht) but also has a B/C ratio less than 1 (around 0.5), representing that the cost of the option is more than its benefit. In the last study area, Pak Meng beach (Mai Fat subdistrict, Trang province), the wave attenuation dome accompanied by beach nourishment option (scenario 1) is proper for the initiation of adaptation activities. The high values of NPV (around 109.93–185.74 million baht) and the high ratio of benefits to costs (2.00–2.59) are the key factors for selection/decision. Besides, the negative values of NPV (−232.95) and IRR (−55.35%) affirm that the option of beach nourishment alone (scenario 2) without integrating other measures/options is not suitable for a high–erosion rate area such as Pak Meng beach (or Ban Ko Fai beach).

Table 6 Summary of NPV, B/C Ratio and IRR in various scenarios
Full size table

6 Discussion

The main outcome of the study using the SimCLIM/CoastCLIM model was a relevant contribution to knowledge on sea-level rise and its effect on sandy beach erosion. Storm alteration was also considered.

In terms of impacts, the loss of sandy beach area and population migration were calculated with simple equations and assumed to be linear functions in terms of impacts from sandy beach erosion. The economic loss due to these two impacts was not considered; this may be overestimated, particularly in terms of migration numbers. Further studies should utilize more sophisticated formulations, using only the population statistics of the affected sectors (e.g., tourism and beach-related activities), and investigate the interaction between socioeconomic development and local factors (e.g., freshwater inflow). Nevertheless, in the overall analytical concept, the study attempted to conduct a new scientific achievement in terms of distinguishing the factors relevant to sandy beach erosion. The SimCLIM/CoastCLIM model can be potentially applied in this aspect. Users can customize the input factors (e.g., sea-level rise, and storminess) that affect the beach system, and investigate the result of each contribution factor separately. However, only a limited range of influential factors were used in the model; human-induced factors and natural erosion were not included. In addition, the main assumptions of the model are about the way to generate a climate scenario for a climate variable under a certain emission scenarios from multiple GCMs and potential storm erosion. In term of multiple GCMs, combining results from multiple models is based on the fundamental assumption that errors tend to cancel if the models are independent, and thus uncertainty should decrease as the number of models increases. In regards to storm erosion, for the SimCLIM/CoastCLIM model, actual storm erosion was assumed to be 10% of the value selected in the potential one (CLIMsystems, 2013, 2014). Thus, future works or analysis should endeavor to focus on the others especially the human-induced factors or human activities.

In comparison with evidence-base data, relative sea-level change results of SimCLIM/CoastCLIM in Rayong and Trang (0.30 and 0.31 cm/yr) are likely consistent with observation data of Hydrology Sect. Engineering Bureau Marine Department (2014a, 2014b, 2014c) in the 1968–2014 period but seem to be different in case of Nakhon Si Thammarat (0.32 cm/yr). The underlying reasons are: (1) vertical land movement inclusion in the model analysis but not for the tidal stations and (2) the limitation (of the model) in translating/downscaling SLR scenarios (in GCMs) from global to local scale (CLIMsystems, 2013). The estimated value of retreat of coastline rate in Trang (5.37 m/yr) is quite similar to observed value of Department of Marine and Coastal Resources (2008; 2013 g,h) in the 59-year period. Nevertheless, retreat of coastline values of Rayong and Nakhon Si Thammarat (3.62 and 5.51 m/yr) are slightly underestimated comparison to the observation data. This possibly caused by: (1) inability of the model to estimate ‘total erosion’ and (2) limited data of depth of material exchange of Thailand in base year (CLIMsystems, 2013). The sand loss value of Trang (0.03 km2/yr) and Nakhon Si Thammarat (0.04 km2/yr) reinforces the conclusion in the study of Thampanya et al. (2006) which reported that estimated net change of coastal area in the 30-year period (1967–1998) should be accounted for 0.01–0.32 and 0.03–0.06 km2/yr for the east and west coast of Southern Thailand. Moreover, the estimated value of sand loss in Rayong (0.032 km2/yr) is also quite consistent with observation data of Department of Marine and Coastal Resource (2009) which ranges about 0.01–0.02 km2/yr in the 2002–2008 period. Unfortunately, there is no direct measurement/research on change in human settlement caused by sandy beach erosion in Thailand. Few studies (e.g., Saengsupavanich et al., 2009) have been attempted to measure the impacts but only in term of infrastructure loss (e.g., house, road).

With regard to the sensitivity analysis, the contributions of the major factors in the model to future shoreline retreat and loss of coastal land were analyzed. Residual movement (RM) showed the highest contribution, ranging from 70.02 to 75.44%, while 6.76–8.02% and 20.31–23.26% were attributed to storm and sea-level rise parameters, respectively. RM data for long-term change of coastline position (multi-century) should also be added to SimCLIM/CoastCLIM. Unfortunately, the availability of these data in Thailand on a local scale (appropriate for use as input) is limited, and the longest period of observation is only 35 years (1967–2003). Table 1 shows the RM values of Rayong, Nakhon Si Thammarat, and Trang at −1.5 to −3.5, −2 to −5.3, and −1 to −5 m/yr over that period (Department of Mineral Resources, 2001, 2002). The large values of RM may contribute to increased future changes in shoreline at around 70%, as previously mentioned. Thus, further studies should be aware of this issue and seek longer periods for this value to reduce the large influence of RM. Furthermore, the uncertainty and variation of the two main factors, SLR and storms, in the model were assessed using a sensitivity analysis approach. Based on the empirical data from the Hydrology Section Engineering Bureau, Marine Department () during 1995–2014, the variation of sea-level rise was about 4%. However, there were no observation data or studies concerning the issue of storms in Thailand. Thus, the researcher applied the same value of uncertainty for the storm parameter. After analysis, the results showed that when the storm parameter varied within ± 4%, shoreline retreat changed by approximately + 0.17 to + 0.21 and −0.17 to −0.21%, respectively. In addition, the same value of variation in sea-level rise influenced beach erosion to change by + 0.11 to + 0.18 and −0.11 to −0.18%. The RMSE and MAE values for sandy beach erosion in the period 1952–2010 were 2.58 to 2.59 and 1.52 to 1.53 m/yr with ± 4% uncertainty for both SLR and storm parameters. These values compare well with those mentioned previously by Marghany (2013) and Murdukhayeva et al. (2013).

In this study, the Bruun Rule was applied in SimCLIM/CoastCLIM to estimate the change of shoreline/erosion in sandy beaches due to SLR (as mentioned in Sect. 2). The Bruun Rule has two important limitations: Firstly, the Bruun Rule omits consideration of alongshore sediment transport, which influences the sediment budget and the erosion/accretion rate (Hinkel et al., 2013). There is also a lack of total/complete erosion analysis due to all factors. The Bruun Rule merely considers a portion of the factors that affect a beach-and-dune system: only sea-level rise and storm characteristics. Other factors, such as the variation of sediment budgets due to coastal protections on rivers (e.g., dams), types of coastal vegetation (as shoreline protection), and land use in coastal areas, are not included (Warrick, 1998). Moreover, the model also has uncertainties in projections/estimations/simulations for future scenarios due to the uncertainty in translating GCMs and RCPs to the local scale (CLIMsystems, 2013). A further limitation was the scarcity and unavailability of input parameters for SimCLIM/CoastCLIM analysis in the base year (1940). Several input parameters required expert advice and proved difficult to acquire, particularly shoreline response time (τ), storm parameters (SSCM and SSCSD), closure distance (l), and depth of material exchange (d) (Warrick, 1998). Both τ and storm parameters required historical data of storm frequency and intensity. Several studies estimated the value of τ as ranging from 3 to 15 years (Addo et al., 2011; Leatherman, 1984), while the values of SSCM and SSCSD were estimated at 4.5–10, and 1.57–5 m of erosion (Addo et al., 2011; Department of Mineral Resources, 2002). Furthermore, difficulties were encountered in the estimation of l and d due to lack of available data and in several empirical formulations, as mentioned by Ranasinghe and Stive (2009). Several studies have estimated the values of l and d as varying between 4.19 and 10 and between 595 and 1000 m, respectively (Addo et al., 2011; Batten et al., 2007; Farrel, 2007).

Furthermore, the “partial analysis” of sandy beach erosion (only sea-level rise and storm patterns) conducted in the study possibly neglected the origin of problem. Major effects from human interventions were not accounted and analyzed. Although the CBA exhibited the cost-effectiveness of different adaptation options/measures but it didn’t show the “whole picture” of beach erosion and solutions. Process-based models could alternatively solve the problems. Shoreline development modelling package (UNIBEST-CL+) which can estimate erosion using bathymetry data accompanied with main anthropogenic interventions factors (effect of harbour jetties and river dams) as well as satellite images analysis should be deployed. Google Earth Engine (GEE) platform could be conducted as a mean of efficient global scale shoreline detection. In addition, a sand nourishment impact model (Ntool) together with current situation of waves (Delft3D-WAVE) should be implemented for detailed simulations and analysis of repeated sand nourishments over time as well as erosion and SLR impacts on the shoreline position (Giardino et al., 2018; Luijendijk et al., 2018; Stronkhorst et al., 2018).

The NPV, B/C ratio, and IRR values at a 6% discount rate for the three beaches were recalculated with sensitivity analysis in terms of ± 20% variations in cost and benefit. On Mae Phim beach, with variation in + 20% of costs and −20% of benefits (the worst case), scenarios 1 and 3 also showed the potential for implementation due to their high NPV (244.74 million baht) and B/C ratio (3.65). Moreover, on Ban Ko Fai beach, scenario 3 with the reef is still also recommended due to its high value of NPV (232.23 million baht), B/C ratio (3.71), and IRR (32.24%) in the worst case. On the other hand, this leads to a negative value of NPV in scenarios 1 (−73.13 million baht) and 2 (−140.77 million baht). Furthermore, on Pak Meng beach, scenario 1 should still also be promoted due to its high values of NPV (43.80 million baht) and B/C ratio (1.33). Besides, scenarios 2 and 3 tend to be avoided due to negative values of NPV: −268.88 and −9.01 million baht, respectively.

Since a sandy beach is the transition zone between land and sea and a dynamic system that changes over time (a complex-adaptive system), the introduction of hard structural options for adaptation (a jetty, groin, breakwater, or seawall) could affect the natural balance in the system of sediment transport. In the sandy beach system, there is a natural balancing process of sediment transport during monsoon and normal seasons. Erosion occurs in the monsoon season (with severe winds and waves), and accretion occurs in the normal season (with low levels of winds and waves). The eroded beach areas could be restored in a few years by this natural process and returned to dynamic equilibrium, as shown on Samila beach, Songkhla province (Faculty of Economics, Prince of Songkhla University, 2011). However, hard structural options could interrupt this process and change the directions of wave and alongshore sediment transport. These structures cause deposition of sediment on the updrift side and lead to erosion on the near (downdrift) side (Faculty of Economics, Prince Edward Island Department of Environment, Labour & Justice, 2011; Prince of Songkhla University, 2011; Ritphring et al., 2021; Roy et al., 2023). There is evidence of hard structure problems along the coast of the Gulf of Thailand. Hard-structural adaptation options and/or strategies in sandy beach areas should be applied with high awareness of the impacts mentioned before and with consideration of the natural system and the process of sediment transport. Moreover, the structures also block the inland migration of coastal wetlands, leading to loss of wetland areas and ecosystems.

7 Conclusions

This paper simulated the possible impacts of sea-level rise in terms of sandy beach erosion. These impacts included loss of land area and the number of people forced to migrate. Several input parameters were added into the SimCLIM/CoastCLIM program to generate variations in sea level and shoreline, while the two major impacts were calculated using fundamental equations.

The results show that the historical trend of sea-level rise was between 0.12 and 0.14 cm/yr, and sandy beach erosion was 3.61–5.33 m/yr from 1940 to 1995. During the same period, the historical sand loss was between 0.02 and 0.04 km2/yr, and 3–18 people/yr were forced to migrate. Future projections indicate that the sea level will rise by 122.33, 124.38, and 121.26 cm in Rayong, Nakhon Si Thammarat, and Trang, respectively by 2100 compared with the 1950 level. Sandy beaches will be eroded by between 348.31 and 507.90 m. Sand loss values varied from 1.49 to 2.15 km2. Moreover, 172–873 people will be forced to migrate because of erosion. Many sandy beaches will be affected by this rate of erosion, including Mae Phim, Pla, Phayun, Namrin, Sichon, Ban Ko Fai, Kanom, Pak Meng, Leam Makham, and Hua Hin. Moreover, high–population density areas, including industrial provinces like Rayong, will be more affected in terms of change in coastal population, despite having lower rates of erosion and sand loss. The interaction between economic activities and population migration due to sand loss must be considered, e.g. land-use regulation/zoning in areas at risk of flood or erosion and/or substantial financial compensated cost for existing property owners.

In the aspect of adaptation options, the CBA method was applied. In summary, scenario 2 is not proper for application as an adaptation option on a sandy beach with a high amount of erosion (Ban Ko Fai and Pak Meng beaches), while other options are likely suitable, such as a sea wall, a wave attenuation dome with beach nourishment, or an artificial reef. On Mae Phim beach, the option of beach nourishment alone (scenario 2) is likely to be applied due to its lower rate of erosion. However, a stone sea wall would be most proper, as it would have the highest NPV (518.31 and 354.88 million baht). Results from sensitivity analysis with variations in cost and benefit of about 20% do not change the results. Nevertheless, scenarios 2 and 3 tend to be avoided due to the previously mentioned negative values of NPV.

8 Limitations of study

The scarcity of data also hindered the CBA’s performance. The limited number of data available for evaluating benefits such as long-term erosion rate in the study areas affected the reliability of the results (for sand loss). Moreover, the erosion rate was assumed to be constant over time due to the lack of data. This could affect the amount of sand loss and forced migration, meaning they are not the best estimates. The uncertainty in future income growth and tourist behavior could affect the results. Nonetheless, the “partial analysis” of sandy beach erosion (sea-level rise and storm patterns) conducted in the study possibly abandoned the human -interventions effects. Although the CBA exhibited the cost-effectiveness of various adaptation alternatives but it did not express the “holistic framework” of SLR-impact analysis under similar coastal systems. Further studies should apply more complex and related models that can analyze the impacts of beach erosion or sea-level rise on tourism flow directly. Moreover, the erosion rate of a specific beach on a local scale can be estimated by applying MEPBAY (Model for Equilibrium Platform of BAY beaches).