The 2011 monsoon season was exceptionally heavy, leading to extensive and long-lasting flooding in the Chao Phraya river basin. Flooding was exacerbated by rapid expansion of urban areas into flood plains and was the costliest natural disaster in the country's history, with direct damages estimated at US$ 45 billion. The present study examines the flood behavior in 2011 and flood impact from changing climate. Two generations of the global climate model (GCM), ensembles CMIP3 and CMIP5, are statistically downscaled through historical 20th century and future projections. The majority of GCMs overestimate the dry spell (in June and July) and underestimate the peak precipitation (in May and September). However, they can simulate the mean precipitation reasonably well. Use of the Multi Model Mean shows continuously increased precipitation from near-future to far-future, while the Multi Model Median shows increased precipitation only for the far-future. These findings in changing precipitation are assessed by flood simulation. With several adaptation measures, flood in the lower Chao Phraya river basin cannot be completely avoided. One of the best practices for a high flood risk community is to raise the house with open space in the first floor. This is promoted as one resilient approach in Thailand.

INTRODUCTION

The impact of global warming is likely to increase vulnerability to the potentially damaging impacts of climate change, especially increases in the frequency and harshness of weather events such as heavy rainfall, shifts in the rainy season, and an increase in the number of wet days. Therefore it would need rather long-term future climate projection to be able to clearly detect the change in future climate patterns. Several communities in the coastal areas are vulnerable to a range of hazards, especially coastal flooding. An increase in mean sea level and sea level extremes will mainly affect the terrestrial landscape, increasing the risk of inundation of low-lying coastal areas. Coastal areas in both developing and more industrialized economies face a range of risks related to climate change and variability. Potential risks include accelerated sea level rise, an increase in sea surface temperatures, intensification of tropical and extra tropical cyclones, extreme waves and storm surges, altered precipitation and runoff, and ocean acidification. The Intergovernmental Panel for Climate Change Fourth Assessment Report (IPCC 2007) points to a range of outcomes under different scenarios. It identifies a number of hotspots, including heavily urbanized areas situated in the low-lying deltas of Asia and Africa, as especially vulnerable to climate-related impacts. The number of major cities located near coastlines, rivers, and deltas provides an indication of the population and assets at risk. Thirteen of the world's 20 largest cities are located on the coast, and more than a third of the world's people live within 100 miles of a shoreline. Low-lying coastal areas represent 2% of the world's land area, but contain 13% of the urban population (McGranahan et al. 2007). Bangkok is one of the cities vulnerable to climate change.

Floods have had an impact on human society from time immemorial. Damage from flooding has been increasing, and floods are occurring more frequently. Flooding is a byproduct of persistent heavy rain, which is associated with the passage of a tropical cyclone or a monsoonal depression. However, flood disasters are increasing because of deforestation, poor land drainage, expanding agriculture and urbanization. All of these activities promote an increased rate of runoff and a greater flood peak. Hydrologic models provide a framework in which to conceptualize and investigate the relationship between climate change and floods. The use of hydrological models (hard computing) in climate change studies can range from the evaluation of annual and seasonal streamflow forecasting using simple water-balance models (Arnell 1992) to the evaluation of flood impact processes (Singh & Kumar 1997; Bronstert 2003). Recently, hydrologic modeling uncertainty has been examined for climate change impact study (Poulin et al. 2011; Steinschneider et al. 2015). In addition, the use of soft computing in the field of hydrological forecasting is a relatively new area of research. Soft computing technologies can produce a single, hybrid solution for the enhancement of operational river levels and flood forecasting systems. Moreover, it offers a more flexible, potentially self-adaptive approach to modeling flood processes, which by their nature are inherently complex, nonlinear and dynamic. Recent application of flood forecasting using the soft computing approach can be found (Wu et al. 2008; Chidthong et al. 2009; Taorumina & Chau 2015).

Since the Coupled Model Intercomparison Project (CMIP) was launched in 1995, coupled ocean–atmosphere general circulation models developed in dozens of research centers around the world have been compared and analyzed extensively. The program has improved our scientific understanding of the processes of Earth's climate system and of our simulation capabilities in this field. CMIP also plays an important social role by contributing to the Intergovernmental Panel on Climate Change (IPCC). CMIP phase three (CMIP3) provided the scientific base for the Fourth Assessment Report (AR4) of IPCC published in 2007. CMIP phase 5 (CMIP5) was initiated in 2008, and the CMIP5 data are now available for analysis and are expected to provide new insights on our climate for the Fifth Assessment Report (IPCC 2013). The latest generation of global climate models (GCMs), the framework of the fifth phase of the CMIP5, reflects 5–6 years of effort by multiple climate modeling groups around the world. Compared to CMIP3, CMIP5 models typically have finer resolution processes, incorporation of additional physics, and better-developed or well-integrated earth system components (Taylor et al. 2012). Emerging literature on CMIP5 (Meehl et al. 2009; Taylor et al. 2011a, 2011b, 2012) has reported improvements in simulating certain key processes. Kug et al. (2012) reported that the CMIP5 suite of models performs slightly better than the CMIP3 models in simulating two types of El-Nino events: Warm Pool El-Nino (a new type) and Cold Tongue El-Nino (the conventional El-Nino). A few studies have been performed on the intercomparison of the performance of CMIP3 and CMIP5 (Brands et al. 2013).

Previously, projected changes of the climate by GCMs were derived mainly from the simulation of CMIP3 and more recently, by the latest generation of coupled models in CMIP5. However, most research concentrates on large domains such as the global monsoon region (Chen & Sun 2013), the western Pacific region (Grose et al. 2014), or the Asian monsoon region (Sperber et al. 2013). Regional assessments of models have been attempted for a limited number of regions: South America (Blázquez & Nuñez 2013), Europe (Cattiaux et al. 2013), and Australia (Mehrotra et al. 2014). Hydrologic projections require climate projections that are derived from GCMs, the only tools we have available to us to explore the response of the global climate system to scenarios of future greenhouse gas emissions. GCMs are run at relatively coarse resolutions with a grid spacing of a 100 km or more. The regional climate response will be affected by large topographic relief, strong precipitation gradients and continentality. Statistical downscaling is an approach often adopted to translate large-scale information to a local scale, where relationships between large-scale variables in the GCM and local-scale climate variables in the observed record are used to adjust GCM projections to better represent local conditions.

In this paper, we assess the CMIP3 and CMIP5 simulations on the local scale of Thailand and look into future change in precipitation for Bangkok due to its vulnerability to climate change. First, the flood history and flood behavior of the 2011 great flood are examined. Then, the future vulnerability and statistical downscaling from nine model pairs of CMIP3 and CMIP5 are explained. Subsequently, hydrological simulations (Mike11 and Mike21 models) with input climate change variables from the climate downscaling are described, the results of several adaptation measures are given, followed by a summary and short discussion.

FLOODING HISTORY AND FLOOD BEHAVIOR IN BANGKOK AND VICINITY

The Chao Phraya river basin is the largest basin in the country, covering an area of 159,000 km2 or about 35% of the total land area of the country. There are two main rivers bisecting the delta area: the Tha Chin River on the west and the Chao Phraya River (the main stream) on the east. The basin is formed by four large tributaries: the Ping, Wang, Yom, and Nan originate from the mountainous terrain in the northern part of the country. These four tributaries flow southward to join each other in Nakhon Sawan to become the Chao Phraya River. The river flows southward through a large alluvial plain to reach the sea at the Gulf of Thailand as illustrated in Figure 1. The basin area is flat at an average elevation of 1–2 m from mean sea level, with certain spots where the elevation is lowered down to sea level due to land subsidence. There are a number of canals crossing the whole basin. Bangkok straddles the Chao Phraya River, 33 km north of the Gulf of Thailand. Owing to the flatness of the area and close proximity to the seashore, the area annually faces the problems of floods from rivers from the north and inundation due to high tides from the sea.
Figure 1

Chao Phraya river basin.

Figure 1

Chao Phraya river basin.

Bangkok covers an area of 1,569 km2, and is located in the delta of the Chao Phraya river basin. As the capital and largest city it has about 7 million people, with 12 million people in Bangkok Metropolis, and contributes 43% to the country's Gross Domestic Product (GDP). Over the past three decades, severe flooding in Thailand has become increasingly common. Figure 2 shows the flood extents of five of the most severe floods in Thailand's recent history. Not shown in this figure are the floods of 2010 and 2011. The 2011 floods in the Chao Phraya basin are the worst floods ever recorded in the country, and the estimated US$45.7 billion in costs make it more expensive than Hurricane Katrina (World Bank 2011). Although efforts have been made to mitigate flood damage in the Chao Phraya river basin through several structural measures (construction of dams, reservoirs, dikes and pumping stations), flooding still causes much more impact as a result of deforestation, farmland expansion and urban development. The flood damage potential is increasing due to rapid urbanization and land development in downstream areas, particularly in Ayutthaya and its municipalities along the Chao Phraya River.
Figure 2

Past severe floods in the Chao Phraya river basin.

Figure 2

Past severe floods in the Chao Phraya river basin.

Bangkok is still under threat of flooding, especially the increasing flood risk due to climate change and the rapid urbanization in the floodplain. The frequency of devastating floods tends to be higher and there are increasing signs of loss of human lives and property. Flood devastation such as in Thailand in 2011 is not simply the result of extreme rainfall and poor reservoir management. It results from failure to prepare for recurrent floods (Ziegler et al. 2012).

The 2011 rainy season was influenced by both the northwest monsoon and tropical storms, commencing with Nock-Ten, which made landfall in Vietnam and became a tropical depression before moving to Thailand at the end of July (Nan province). In addition, there were four storms (Haima, Haitan, Nesat and Nalgae) that caused medium to heavy rainfall from June to October in the north and northeast of Thailand (see Figure 3). The continuous rainfall in the north accumulated nearly 1,675 mm of water, which is 42% more than the 30-year average value (see Figure 4). This caused significant accumulated run-off volume passing Nakhon Sawan province, more than 30,000 mcm compared to approximately 25,000 mcm and 27,000 mcm 1995 and 2006 floods, respectively, as given in Figure 5. In total, the floods damaged 18,291 km2 of farmland and 804 factories, and killed 813 people.
Figure 3

Storm tracks in 2011 (TMD).

Figure 3

Storm tracks in 2011 (TMD).

Figure 4

Cumulative rainfall in the northern part.

Figure 4

Cumulative rainfall in the northern part.

Figure 5

Run-off volume passing Nakhon Sawan.

Figure 5

Run-off volume passing Nakhon Sawan.

By early October the two reservoirs stored approximately 10 billion m3, which is an amount equivalent to two-thirds of the total flood volume. Figure 6 shows the inflow and release from the two major upstream reservoirs (Bhumibol and Sirikit dams). Several inflow peaks occurred after the passage of the five storms mentioned before. The overbank flow occurred during the middle of September at several communities downstream from Nakhon Sawan. The water level in both reservoirs rose sharply and touched the maximum pool level during the first week of October. The sudden released discharge from Bhumibol dam (more than 100 mcm/day) and the Sirikit dam (more than 50 mcm/day) caused tremendous impact on the downstream communities, especially the industrial estates. Within the first 3 weeks of October, seven major industrial estates were submerged 2–3 m during high flood level (see Figure 7). This caused interruption to the supply chain of car parts regionally and worldwide, e.g. electronic components and hard disk drives. The Ministry of Industry estimated damage in various industrial estates to be approximately THB (Thai Baht) 513.9 billion. Similarly, the Ministry of Industry estimates the losses from reduced production at THB 493 billion (World Bank 2012).
Figure 6

Inflow and release of (a) Bhumibol and (b) Sirikit dams.

Figure 6

Inflow and release of (a) Bhumibol and (b) Sirikit dams.

Figure 7

Flood inundated seven industrial estates.

Figure 7

Flood inundated seven industrial estates.

ASSESSMENT OF PRECIPITATION CHANGE BY CMIP3 AND CMIP5 MODELS

One plausible consequence of global warming is acceleration of the hydrological cycle, which is simply the balance among global evapotranspiration, rainfall, surface runoff, and storage (Ziegler et al. 2003). Acceleration may increase the frequency and/or intensity of extreme events, which occur annually throughout monsoon Asia. Thailand is affected by two tropical weather patterns during its ‘rainy’ season: the southwestern monsoon that originates in the Indian Ocean and tropical storms that originate in the Pacific. The monsoons are the primary driver of rainfall during the rainy season, which lasts from June to December, and flooding often occurs during the regular heavy monsoon rains. Tropical storms are less common in Thailand, but combined with monsoon rains, they helped to cause the two largest floods in recent years.

In the present study, we investigated the capability of CMIP3 and CMIP5 models in simulating precipitation over land through historical 20th century skills and future projections. Because the GCM selection was not the primary objective of this study, therefore, it was done by: (1) the availability of the variables (precipitation), scenarios (CMIP3:B1 and A2; CMIP5: RCP4.5 and RCP8.5), and based line and target projection years (1980–1999 and 2010–2099); and (2) guidelines from previous studies for the southeast Asia region (Kumar et al. 2014; McSweeney et al. 2015). Finally, we used nine climate model pairs from CMIP3 and CMIP5, downscaling for Bangkok (see Table 1). The horizontal resolution of the CMIP5 models is generally improved for the majority of the models.

Table 1

List of IPCC CMIP3 and CMIP5 GCMs used in this study

CMIP3 Resolution CMIP5 Resolution Center 
CNRM-CM3 128 × 64 CNRM-CM5 256 × 128 Centre National de Recherches Meteorologiques, France 
CSIRO-Mk3.0 192 × 96 CSIRO-Mk3.6 192 × 96 CSIRO, Australia 
GFDL-CM2.0 144 × 90 GFDL-CM3 144 × 90 Geophysical Fluid Dynamics Laboratory, NOAA 
GFDL-CM2.1 144 × 90 GFDL-ESM2M 144 × 90 Geophysical Fluid Dynamics Laboratory, NOAA 
GISS-ER 72 × 46 GISS-E2-H 144 × 90 Goddard Institute for Space Studies, USA 
INM-CM3.0 72 × 45 INM-CM4 180 × 120 Institute of Numerical Mathematics, Russia 
IPSL-CM4 96 × 72 IPSL-CM5A-LR 96 × 96 Institut Pierre Simon Laplace, France 
MIROC3.2 (medres) 128 × 64 MIROC5 256 × 128 CCSR/NIES/FRCGC, Japan 
MRI-CGCM2.3.2 192 × 96 MRI-CGCM3 320 × 160 Meteorological Research Institute, Japan 
CMIP3 Resolution CMIP5 Resolution Center 
CNRM-CM3 128 × 64 CNRM-CM5 256 × 128 Centre National de Recherches Meteorologiques, France 
CSIRO-Mk3.0 192 × 96 CSIRO-Mk3.6 192 × 96 CSIRO, Australia 
GFDL-CM2.0 144 × 90 GFDL-CM3 144 × 90 Geophysical Fluid Dynamics Laboratory, NOAA 
GFDL-CM2.1 144 × 90 GFDL-ESM2M 144 × 90 Geophysical Fluid Dynamics Laboratory, NOAA 
GISS-ER 72 × 46 GISS-E2-H 144 × 90 Goddard Institute for Space Studies, USA 
INM-CM3.0 72 × 45 INM-CM4 180 × 120 Institute of Numerical Mathematics, Russia 
IPSL-CM4 96 × 72 IPSL-CM5A-LR 96 × 96 Institut Pierre Simon Laplace, France 
MIROC3.2 (medres) 128 × 64 MIROC5 256 × 128 CCSR/NIES/FRCGC, Japan 
MRI-CGCM2.3.2 192 × 96 MRI-CGCM3 320 × 160 Meteorological Research Institute, Japan 

Monthly precipitation data have been extracted from nine climate model pairs from CMIP3 and CMIP5. We have examined mean climatology for the historical period (1980–1999), for the near-future period (2010–2039), for the mid-future period (2040–2059) and the far-future period (2080–2099) with only one initial condition ensemble for each, ‘run1′ from CMIP3 and ‘r1i1p1′ from CMIP5. In this study, we considered the last two decades of the 20th century and extracted from the ‘20c3m’ and ‘historical’ experiments from CMIP3 and CMIP5, respectively. Future monthly precipitation data have been taken from comparable greenhouse warming scenarios, SRES B1 and A2 from CMIP3, and RCP4.5 and RCP8.5 from CMIP5 models. The IPCC AR4 scenario SRES B1 has been reported to best match the RCP4.5 temperature and total anthropogenic RF projections (Rogelj et al. 2012; Stocker et al. 2013). All these models were regridded to a 0.5 ° (720 longitude × 278 latitude) and were interpolated (by bi-linear interpolation) to the Bangkok station (with latitude 13.73 N and longitude 100.56 E) as shown in Figure 8. Daily observed precipitation data at Bangkok station during the period 1980–1999 were obtained from the Thailand Meteorological Department (TMD). The multi-model ensemble mean and median were constructed with equal weights.
Figure 8

Climate model grid boxes (dots) and the Bangkok station (red star). Please refer to the online version of this paper to see this figure in colour.

Figure 8

Climate model grid boxes (dots) and the Bangkok station (red star). Please refer to the online version of this paper to see this figure in colour.

The annual cycle of the mean precipitation for nine pairs of CMIP3–CMIP5 models and mean observed data (in bar) are shown in Figure 9. It was found that most CMIP5 models give a precipitation peak in September (agreeing with the observed data), while CMIP3 models give a precipitation peak in August. Two models of CMIP3 (MRI-CGCM2.3.2, INM-CM3.0) and five models of CMIP5 (Miroc-CGCM3, GISS-E2-H, CSIRO-Mk3.6, GFDL-ESM2M, GFDL-CM3) give double peaks of mean precipitation similar to the observed data. The lower peak in May indicates the southwest monsoon season beginning, while the higher peak in September indicates the influence of the southwest monsoon and the Inter Tropical Convergence Zone (ITCZ). The lower precipitation (dry spell) in June and July is caused by a rapidly northwards movement of the ITCZ across southern China. The majority of CMIP3 and CMIP5 models overestimate the dry spell and underestimate the peak precipitation.
Figure 9

Mean monthly precipitation of nine pairs of CMIP3 and CMIP5 models for the 20th century (1980–1999). (a) CMIP3 and (b) CMIP5. Please refer to the online version of this paper to see this figure in colour.

Figure 9

Mean monthly precipitation of nine pairs of CMIP3 and CMIP5 models for the 20th century (1980–1999). (a) CMIP3 and (b) CMIP5. Please refer to the online version of this paper to see this figure in colour.

Two statistical downscaling methods were used in this study. They are linear scaling (LS; Lenderink et al. 2007) and distribution mapping (DM; Teutschbein & Seibert 2012). The LS operates with monthly correction values based on the differences between observed and GCM during the reference period; corrected GCM will perfectly agree in their monthly mean values with observations. The adjusted monthly precipitation for the reference and future period is obtained using Equations (1) and (2), 
formula
1
 
formula
2
where P =precipitation, m = the monthly interval, obs = observed, ref = GCM 1980–1999, fut = GCM 2010–2099, and μ = the mean value.
The DM corrects the distribution shape of the monthly precipitation based on cumulative distribution functions (CDFs), and these were constructed for both the observed and the GCM (1980–1999) for all months. Thereafter, the value of GCM precipitation of month m was searched on the empirical cumulative distribution function (ECDFs) of the GCM together with its corresponding cumulative probability. Then, the value of precipitation of the same cumulative probability was located on the ECDFs of observations. Finally, the monthly precipitation for reference and future periods are obtained by Equations (3) and (4) in terms of the Gamma CDF and its inverse . 
formula
3
 
formula
4
where α = the shape parameter of gamma distribution, β =the scale parameter of gamma distribution, F = the cumulative distribution function (CDF), F−1 = the inverse of CDF, and γ = the gamma distribution.

We downscaled precipitation for each of the nine model pairs from both generations of models (CMIP3 and CMIP5) and computed with respect to Bangkok station observed data. Then, we also evaluated the models' agreement on projecting mean climatology over the last two decades of the 20th century (1980–1999). Specifically, we generated changes in projection precipitation for all model pairs from the CMIP3 and CMIP5 models with three target periods, the near-future (2010–2039), the mid-future (2040–2069) and the far-future (2070–2099).

The similarity between the observed and model-simulated field can be described by the Taylor diagram. The reference data set is plotted along the abscissa. The model data set is plotted in the first or second quadrant depending upon whether the correlation coefficient is positive or negative, respectively. The azimuthal position of model data is given by the arccosine of the correlation coefficient between the reference and model data set. The radial distances of reference and model data points from the origin are proportional to their standard deviations. The centered RMS (Root Mean Square) error is proportional to the distance between the points representing reference and model data sets. The closer a model point is to the reference data point, the lower its centered RMS error; it implies that the model is performing relatively well. High correlation between reference and model data signifies model-simulated seasonal cycles are reasonably phased. In this study, we use observations (TMD data) as a baseline for measuring historical performance, while the ensemble medians are used as a baseline for multi-model performance measuring in the future. As a measure of central tendency, Wilks (2011) suggested the use of median rather than the mean to be robust to outliers.

Figure 10 shows the Taylor diagrams for historical rainfall data over Bangkok for LS and DM, respectively. It can be seen that DM gives better results compared to LS. The models show both generations performed reasonably well in capturing the amplitude and phasing of past mean annual precipitation over Bangkok. The correlation coefficient over Bangkok from all models lies between 0.6 and 0.8, implying most of the models simulate the mean rainfall reasonably well. In addition, both model generations have approximately the same standard deviation as the observed, implying similar spatial variability. However, they are different in RMS error.
Figure 10

Taylor diagram of the 20th century annual mean precipitation climatology (1980–1999) for CMIP3 and CMIP 5 models. (a) LS and (b) DM.

Figure 10

Taylor diagram of the 20th century annual mean precipitation climatology (1980–1999) for CMIP3 and CMIP 5 models. (a) LS and (b) DM.

We then applied DM for future precipitation. Figure 11 shows Taylor diagrams of annual mean precipitation for the target future periods of CMIP3 and CMIP5 models. The images on the left are the results from CMIP3(B1) and CMIP5(RCP4.5), and the images on the right are the results from CMIP3(A2) and CMIP5(RCP8.5). The correlation coefficient for the target future periods (Figure 12) does not change significantly from the historical data. Both CMIP3 and CMIP5 models still simulate the timing of rainfall reasonably well. However, more spatial variability and more RMS error are found for all future periods compared to the historical period. In addition, the results of CMIP3(A2) and CMIP5(RCP8.5) show higher spatial variability and higher RMS error than the results of CMIP(B1) and CMIP5(RCP4.5), especially for the far-future period. Therefore, past model performance does not guarantee future results.
Figure 11

Taylor diagram of the annual mean precipitation climatology in future climate for CMIP3(B1) and CMIP5(RCP4.5) (left hand pictures) and CMIP3(A2) and CMIP5(RCP8.5) (right hand pictures). (a) Near-future (2010–2039), (b) mid-future (2040–2069) and (c) far-future (2070–2099).

Figure 11

Taylor diagram of the annual mean precipitation climatology in future climate for CMIP3(B1) and CMIP5(RCP4.5) (left hand pictures) and CMIP3(A2) and CMIP5(RCP8.5) (right hand pictures). (a) Near-future (2010–2039), (b) mid-future (2040–2069) and (c) far-future (2070–2099).

Figure 12

Mean precipitation in the wet season (MJJASO). (a) Near-future (2010–2039), (b) mid-future (2040–2069) and (c) far-future (2070–2099).

Figure 12

Mean precipitation in the wet season (MJJASO). (a) Near-future (2010–2039), (b) mid-future (2040–2069) and (c) far-future (2070–2099).

The future rainfall over Bangkok was analyzed for the period 2010–2099 for all scenarios in this study (B1, A2, RCP4.5, RCP8.5). The delta change factor approach was used to indicate the difference between future and reference day climate. The delta change is compiled for the seasonal (MJJASO) scale in Figure 11. The mean precipitation during the wet season (MJJASO) of CMIP3 and CMIP5 models and observed data (1980–1999) are shown for all target future periods. The model color is similar to Figure 9. Most GCMs show higher mean precipitation than the observed data except for CSIRO-Mk3.0, IPSL-CM4 (CMIP3) and CSIRO-Mk3.6, and IPSL-CM5A-LR(CMIP5). Results show higher variability for a higher emission scenario similar to Figure 12. We observed that all CMIP3 and CMIP5 models give a significant increase in precipitation for the far-future period, but not so much difference between the near-future and mid-future periods. We also cannot see any improvement in CMIP5 models over CMIP3 models.

Various studies have shown that a Multi Model Mean yields better prediction and compares more favorably to observations than any single model when compared over multiple variables (Knutti et al. 2010). The Multi Model Mean tends to be an improvement over individual models, because the bias in one model is canceled out by another. It is very interesting to see the model performance between the use of Multi Model Median and Multi Model Mean for future projection (see Figure 13). The Multi Model Median and Multi Model Mean in the historical period do not give significant differences. We also found that most GCMs of CMIP3 and CMIP5 give higher mean than median values in the historical period (figure not shown). Use of the Multi Model Mean shows continuously increased rainfall from the near-future to the far-future in the order of 2–12%. The Multi Model Median shows increased rainfall only for the far future. Both show a potential increase in future precipitation, implying more flood vulnerability. Similar trends are found for CMIP3(A2) and CMIP5 (RCP8.5) models, but with different scales.
Figure 13

Delta change factor of both model generations for future mean precipitation at Bangkok. (a) Multi Model Median, (b) Multi Model Mean.

Figure 13

Delta change factor of both model generations for future mean precipitation at Bangkok. (a) Multi Model Median, (b) Multi Model Mean.

FLOOD SIMULATION CONSIDERING CLIMATE CHANGE IMPACT

Regarding the concerns about the future flooding in Bangkok and its vicinity, several studies, for example, IPCC (2007), Nicholls et al. (2008), World Bank (2010) and Kundzewicz et al. (2014) indicate that Bangkok may have to face major flooding in the future. Significant drivers are increasing rainfall and upstream discharge, land use change, land subsidence and an increase in sea level. Any or all of these significant factors would result in Bangkok and its vicinity experiencing greater flood risk and vulnerability.

This study covers areas of Bangkok Metropolitan Administration and surrounding provinces (SamutSakhon, SamutPrakan, NakhonPathom, and Nonthaburi). These areas are located in the Chao Phraya-Tha Chin river Basin (about 31,885 km2). The downstream boundary of the computational domain was set up along the coastline from Tha Chin river to the Bangpakong river (80 km long) and 50 km offshore. The MIKE family (Mike11, Mike 21), similar to World Bank (2009), was used in this study. The computational grid size is 200 × 200 m (see Figure 14). The above two models were integrated into a single model by using the MIKE FLOOD module.
Figure 14

Model configuration.

Figure 14

Model configuration.

The ground surface elevation in the study area was derived from the benchmarks surveyed by the Department of Mineral Resources and the Royal Thai Survey Department. The ground survey elevation outside the study area is derived from the topographic map of 1:50,000 scale. The bathymetry was derived from an eco-sounding survey by the Hydrographic Department. The channel cross-section was surveyed by the Royal Irrigation Department from the upstream (Nakhon Sawan province) to the downstream at the river mouth, a distance of 375 km. The Tha Chin river cross-section was input from the Wat Sing District, Chainat Province, to the downstream station at the Tha Chin River mouth, a distance of 319 km. Several tributary channels in the Chao Phraya-Tha Chin basin were also included in the river network.

The discharge data of the Chao Phraya River (at the Chao Phraya Dam) and the discharge release from the Rama IV Dam of the Pa Sak River were used as the upstream boundary conditions. The tidal level was input as the downstream boundary for the coastal area. The measured and forecasted precipitation were also input at each gauging station for rainfall–runoff computation in 15 sub-basins by the NAM model. The NAM model for each sub-basin was calibrated with the observed discharge from nearby gauging stations, and applying the catchment area proportion. However, all parameters were adjusted again in the HD model calibration. The model was calibrated and verified by the flood events of the years 2002 and 1995, respectively. Good agreements are found between the model results and observed discharges at several stations, as seen in Figure 15.
Figure 15

Model calibration and validation. (a) Model calibration in 1995, (b) model validation in 2002.

Figure 15

Model calibration and validation. (a) Model calibration in 1995, (b) model validation in 2002.

Warming of the global climate system will have a multitude of impacts on the monsoon-driven climate of Bangkok. Based on analysis by 2050 the local mean temperature will rise by 1.9 and 1.2 °C, and the basin mean precipitation will rise by (2.3%) corresponding to CMIP3 (B1, A1FI, respectively). Figure 16 shows only the maximum water level for the future flood scenario A1FI 2050 case (IPCC 2000). It was found that many areas may be inundated for 1–2 months (blue shade area), beginning with the upstream provinces, Ayutthaya, the whole area of PathumThani Province, most area of Nonthaburi Province and some areas of Thonburi and eastern Bangkok. With the flood protection system (a polder dike and pumping system), most of the areas east of Bangkok will be protected except some areas where the crest elevations of dikes are not high enough. For the western area of Bangkok, the crest elevations of dikes are not high enough to protect against flood and sea level rise, especially in the west and south of the area. The inundation area on both banks of the Chao Phraya River will expand where the crest elevations of dikes are not high enough. However, the water level will be higher than the crest elevations of dikes along the river banks, but the duration will be during the high tide period, so the flood waters will not flow into the inner area of Bangkok. Furthermore, the internal drainage system can drain the overflow water into the rivers and the Gulf of Thailand, resulting in less inundation area in the east and the city core of Bangkok. The initial cost of damages was estimated at about THB 150,000 million. More importantly, there will be major difficulties in the administration and management of the given measures, which may not be able to be applied suitably and efficiently owing to the fact that future floods will be too complicated and too severe.
Figure 16

Flood map in 2050.

Figure 16

Flood map in 2050.

Immediately after the 2011 great flood, several adaptation measures were proposed (see Figure 17). Both the hard and soft-engineering approaches of four case studies were examined by using a similar Mike11 model. The first case study is the base case of the 2011 flood (do nothing). The inundation area (for the lower Chao Phraya River) was found to be approximately 8.8 million Rais (14,080 km2). The use of 2 million Rais (3,200 km2) as a retention basin in the second case study can reduce the inundation area by 18%, or 7.2 million Rais (11,520 km2). The third case study uses the east floodway, of 1,000 cms capacity. This can reduce the inundation area by 35%, or 5.7 million Rais (9,120 km2). The fourth case study uses both floodways (east and west, of 1,000 cms capacity each). This can reduce the inundation area by 42% from the 2011 flood event. Therefore, floods cannot be completely avoided in the central part. Considering a flood-resilient approach, Integrated Flood Management in the Chao Phraya basin has to be adopted for a sustainable future. One of the best practices is to raise the houses, with an open space in the first floor as shown in Figure 18.
Figure 17

Adaptation measures.

Figure 17

Adaptation measures.

Figure 18

Raised floor houses with open space.

Figure 18

Raised floor houses with open space.

SUMMARY AND DISCUSSION

The 2011 monsoon season was exceptionally heavy and led to extensive and long-lasting flooding in the Chao Phraya river basin. Flooding was exacerbated by the rapid expansion of urban areas into flood plains, and was the costliest natural disaster in the country's history, with direct damages estimated at US$45 billion.

In this study, we examined the flood behavior and the lessons learned, then we looked into future flood risk and flood vulnerability from extreme precipitation. The comparison across two generations of the GCM ensembles CMIP3 and CMIP5 were made through historical 20th century and future projections. The majority of CMIP5 models show double peaks of precipitation (in May and September) similar to the observed data. However, both models overestimate the dry spell and underestimate the peak precipitation. The precipitation downscaling was performed using two statistical downscaling methods, LS and the DM. The DM downscaling is found to be better than the LS. Overall, our results suggest that the performance of CMIP5 models cannot be readily distinguished from CMIP3 models, although there are clear signals of improvements over Bangkok. Both model generations perform reasonably well in capturing the amplitude and phasing of past mean annual precipitation. The correlation coefficient for all GCMs lies between 0.6 and 0.8, implying most of the models simulate the mean rainfall reasonably well. Even though both model generations display approximately the same standard deviation as the observed, more spatial variability and more RMS errors are found for all future periods. Therefore, past model performance does not guarantee future results. The forecasted precipitation changes, in the rainy season were examined through the Multi Model Mean and Multi Model Median of nine GCMs. The use of the Multi Model Mean shows continuously increased rainfall from the near-future to the far-future, in the order of 2–12%. The Multi Model Median shows increased rainfall only for the far future. Both show a potential increase in future precipitation, implying more flood vulnerability. The complex terrain and land-sea contrast at Bangkok may contribute to those findings. Further spatial analysis for the whole of Thailand will be done in the near future. In addition, more GCMs in CMIP3 and CMIP5 are needed to confirm our present results.

We then applied flood simulation models (Mike 11 and Mike 21) to examine the 2011 flood behavior, including adaptation measures for a changing climate. It is found that several areas in the lower Chao Phraya river basin may be inundated for 1–2 months. The initial cost of damages was estimated at about THB 150,000 million. Four adaptation measures were investigated. The first case study is the base case of the 2011 flood (do nothing). The inundation area is found to be approximately 14,080 km2. The use of 3,200 km2 as a retention basin in the second case study can reduce the inundation area by 18% or 11,520 km2. The third case study uses the east floodway, of 1,000 cms capacity. This can reduce the inundation area by 35%, or 9,120 km2. The fourth case study uses both floodways (east and west, of 1,000 cms capacity each). This can reduce the inundation area by 42% from the 2011 flood event. However, even with this, Bangkok is still under threat of flooding. The frequency of devastating floods tends to be higher and there are increasing signs of loss of human lives and property. In summary, flooding in the Chao Phraya river basin cannot be completely avoided, Therefore, the best practice for high flood risk communities is to raise the houses and have an open space in the first floor, and this should be used as one flood resilient approach.

ACKNOWLEDGEMENTS

The author is grateful for the valuable comments of anonymous reviewers, who contributed to an improvement of this paper. Thanks to the World Bank for sponsoring the project in 2009. Thanks also to the Thai Meteorological Department and the Royal Irrigation Department for providing the observed rainfall and discharge data. The climate model data sets were obtained from the PCMDI archive.

REFERENCES

REFERENCES
Brands
S.
Herrera
S.
Fernández
J.
Gutiérrez
J. M.
2013
How well do CMIP5 earth system models simulate present climate conditions in Europe and Africa
?
Climate Dynamics
41
,
803
817
.
Chen
H. P.
Sun
J. Q.
2013
How large precipitation changes over global monsoon regions by CMIP5 models
.
Atmospheric and Oceanic Science Letter
6
,
306
311
.
Chidthong
Y.
Tanaka
H.
Supharatid
S.
2009
Developing a hybrid multi-model for peak flood forecasting
.
Journal of Hydrological Processes
23
,
1725
1738
.
Grose
M. R.
Brown
J. N.
Narsey
S.
Brown
J. R.
Murphy
B. F.
Langlais
C.
Gupta
A. S.
Moise
A. F.
Irving
D. B.
2014
Assessment of the CMIP5 global climate model simulations of the western tropical Pacific climate system and comparison to CMIP3
.
International Journal of Climatology
34
,
3382
3399
.
IPCC
2000
Emissions Scenarios
. In:
Special Report of the Intergovernmental Panel on Climate Change
(
Nakicenovic
N.
Swat
R.
, eds).
Cambridge University Press
,
Cambridge
,
UK and New York
, USA, p.
570
.
IPCC
2007
Climate Change 2007 The Physical Science Basis
.
Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change
(
Solomon
S.
Qin
D.
Manning
M.
Chen
Z.
Marquis
M.
Averyt
K. B.
Tignor
M.
Miller
H. L.
, eds),
Cambridge University Press
,
Cambridge
,
UK
and
New York
,
USA
, p.
996
.
IPCC
2013
Climate Change 2013: The Physical Science Basis
.
Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change
(
Stocker
T. F.
Qin
D.
Plattner
G.-K.
Tignor
M.
Allen
S. K.
Boschung
J.
Nauels
A.
Xia
Y.
Bex
V.
Midgley
P. M.
, eds).
Cambridge University Press
,
Cambridge
,
UK
and
New York
,
USA
, p.
1535
.
Knutti
R.
Abramowitz
G.
Collins
M.
Eyring
V.
Gleckler
P. J.
Hewitson
B. M.
Mearns
L. L.
2010
Good Practice Guidance Paper on Assessing and Combining Multi Model Climate Projections
.
Meeting Report of the Intergovernmental Panel on Climate Change Expert Meeting on Assessing and Combining Multi Model Climate Projections
.
IPCC Working Group I Technical Support Unit, University of Bern
,
Bern
.
Kug
J. S.
Ham
Y. G.
Lee
J. Y.
Jin
F. F.
2012
Improved simulation of two types of El Niño in CMIP5 models
.
Environmental Research Letters
7
,
034002
.
Kundzewicz
Z. W.
Kanae
S.
Seneviratne
S. I.
Handmer
J.
Nicholls
N.
Peduzzi
P.
Mechler
R.
Bouwer
L. M.
Arnell
N.
Mach
K.
Muir-Wood
R.
Robert Brakenridge
G.
Kron
W.
Benito
G.
Honda
Y.
Takahashi
K.
Sherstyukov
B.
2014
Flood risk and climate change: global and regional perspectives
.
Hydrological Sciences Journal
59
,
1
28
.
Lenderink
G.
Buishand
A.
Deursen
W. V.
2007
Estimates of future discharges of the river Rhine using two scenario methodologies: direct versus delta approach
.
Hydrology and Earth System Sciences
11
,
1145
1159
.
McSweeney
C. F.
Jones
R. G.
Lee
R. W.
Rowell
D. P.
2015
Selecting CMIP5 GCMs for downscaling over multiple regions
.
Climate Dynamics
44
,
3237
3260
.
Meehl
G. A.
Goddard
L.
Murphy
J.
Stouffer
R. J.
Boer
G.
Danabasoglu
G.
Dixon
K.
Giorgetta
M. A.
Greene
A. M.
Hawkins
E.
Hegerl
G.
Karoly
D.
Keenlyside
N.
Kimoto
M.
Kirtman
B.
Navarra
A.
Pulwarty
R.
Smith
D.
Stammer
D.
Stockdale
T.
2009
Decadal prediction: can it be skillful
? B
ulletin of the American Meteorological Society
90
,
1467
1485
.
Mehrotra
R.
Sharma
A.
Bari
M.
Tuteja
N.
Amirthanathan
G.
2014
An assessment of CMIP5 multi-model decadal hindcasts over Australia from a hydrological viewpoint
.
Journal of Hydrology
519
,
2932
2951
.
Nicholls
R. J.
Hanson
S.
Herweijer
C.
Patmore
N.
Hallegatte
S.
Corfee-Morlot
J.
Chateau
J.
Muir-Wood
R.
2008
Ranking port cities with high exposure and vulnerability to climate extremes-exposure estimates
,
Environmental Working Paper No. 1
,
Organisation for Economic Co-operation and Development (OECD)
,
Paris
,
France
.
Poulin
A.
Brissette
F.
Leconte
R.
Arsenault
R.
Malo
J. S.
2011
Uncertainty of hydrological modelling in climate change impact studies in a Canadian, snow-dominated river basin
.
Journal of Hydrology
409
,
626
636
.
Sperber
K. R.
Annamalai
H.
Kang
I.-S.
Kitoh
A.
Moise
A.
Turner
A.
Wang
B.
Zhou
T.
2013
The Asian summer monsoon: an intercomparison of CMIP5 vs. CMIP3 simulations of the late 20th century
.
Climate Dynamics
41
,
2711
2744
.
Steinschneider
S.
Wi
S.
Brown
C.
2015
The integrated effects of climate and hydrologic uncertainty on future flood risk assessments
.
Hydrological Processes
29
(
12
),
2823
2839
.
Stocker
T. F.
Qin
D.
Plattner
G. K.
Tignor
M.
Allen
S. K.
Boschung
J.
2013
Climate Change 2013. The Physical Science Basis. Working Group I Contribution to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change-Abstract for decision-makers, Groupe d'experts intergouvernemental sur l'evolution du climat/Intergovernmental Panel on Climate Change-IPCC, C/O World Meteorological Organization, 7bis Avenue de la Paix, CP 2300 CH-1211 Geneva 2, Switzerland
.
Taylor
K. E.
Stouffer
R. J.
Meehl
G. A.
2011a
A summary of the CMIP5 experiment design (Program for Climate Model Diagnosis and Intercomparison (PCMDI) 2011); http://cmip-pcmdi.llnl.gov/cmip5/docs/Taylor_CMIP5_design.pdf (accessed 30 November 2014).
Taylor
K. E.
Balaji
V.
Hankin
S.
Juckes
M.
Lawrence
B.
Pascoe
S.
2011b
CMIP5 Data Reference Syntax (DRS) and Controlled Vocabularies (Program for Climate Model Diagnosis and Intercomparison (PCMDI); http://cmippcmdi.llnl.gov/cmip5/docs/cmip5_data_reference_syntax.pdf (accessed 30 November 2014).
Taylor
K. E.
Stouffer
R. J.
Meehl
G. A.
2012
An overview of CMIP5 and the experimental design
.
Bulletin of the American Meteorological Society
93
,
485
498
.
Wilks
D. S.
2011
Statistical methods in the atmospheric sciences.
Academic Press, New York, NY
, p.
740
.
World Bank
2009
Climate Change Impact and Adaptation Study for Bangkok Metropolitan
.
Main Report
,
World Bank
,
Washington, DC
, p.
85
.
World Bank
2010
Climate Risks and Adaptation in Asian Coastal Megacities
.
A Synthesis Report
,
World Bank
,
Washington, DC
, p.
120
.
World Bank
2011
The World Bank supports Thailand's post-floods recovery effort
.
World Bank
,
Washington, DC
; .
World Bank
2012
Thai Flood 2011 Rapid Assessment for Resilient Recovery and Reconstruction Planning
.
World Bank
,
Washington, DC
, p.
377
.
Wu
C. L.
Chau
K. W.
Li
Y. S.
2008
River stage prediction based on a distributed support vector regression
.
Journal of Hydrology
358
,
96
111
.
Ziegler
A. D.
Sheffield
J.
Maurer
E. P.
Nijssen
B.
Wood
E. F.
Lettenmaier
D. P.
2003
Detection of intensification of continental-scale hydrological cycles: temporal scale of evaluation
.
Journal of Climate
16
,
535
547
.
Ziegler
A. D.
Lim
H. S.
Jachowski
N. R.
Wasson
R. J.
2012
Reduce urban flood vulnerability
.
Nature
481
,
145
.