As the third largest river of Myanmar, the Chindwin River has great importance as a water resource and transport artery. At 113,800 km2 the basin is comparable in size to the Elbe basin in Europe, although with higher rainfall and runoff. During the southwest monsoon high rainfall intensities with spatial and temporal variation causing severe floods are threatening the region. The study aims to analyze the hydrologic aspects of monsoon floods using statistical and frequency analysis. Flood responses vary due to the complex topography and rainfall distribution over the catchment. Time series of annual maximum floods shows no trend of the mean value. The deviation of annual maxima from the respective mean values, however, has increased significantly in recent decades. Flood quantiles are determined for return periods of 2 to 1,000 years using the data covering the period 1966 to 2011. Flood probability analysis shows that the upper and middle parts of the basin have particularly high flood risks. To analyze the change in flood values, the relative differences of flood quantiles in two time phases, 1966–1990 and 1991–2011, with respect to the entire observation period are compared. The expected floods of the latter period are the highest.

INTRODUCTION

Many regions of the world are experiencing an intensification of floods caused by changing land use and climate. Consequently, risks to human life and possessions are increasing, made worse by population growth, urbanization, settlement in flood plains and development (Huntington 2006; Marchi et al. 2010). Global warming has caused greater climatic volatility shown by changing precipitation patterns, increased frequency and intensity of extreme weather events including flooding and has led to a rise in global mean sea level (Parry et al. 2007; ADB 2009). Streamflow variability is not only highly dependent on anthropogenic activities, but also on seasons and climate (Daniel & Daniel 2006). With their dynamic nature, floods may develop at any space and time scales that conventional rainfall and discharge observation systems are not able to monitor (Marchi et al. 2010). Consequently the atmospheric and hydrological generating mechanisms of flash floods in many regions are poorly understood (Borga et al. 2011). Hence, understanding the hydro-meteorological processes of flash flooding is extremely important, from both scientific and societal perspectives.

Most Asian countries have suffered from flood disasters frequently. As stated by Dutta & Herath (2004), out of the total number of flood events in the world during the past 30 years, 40% occurred in Asia. The regional distribution shows that South Asia is the most affected region with 39%, followed by Southeast Asia with 30%, East Asia with 25% and with 6% the West Asia region is the least affected. Sharma (2012) observed that the flood events across Asia have increased threefold and sixfold between 2000 and 2009, in comparison with the events in the 1980s and 1970s, respectively. In humid tropical and subtropical climates, especially in the realms of the monsoon, river flooding is a recurrent natural phenomenon (Sanyal & Lu 2004). The southwest monsoon, the seasonal change of winds caused by the reversal of the land–ocean temperature gradient, brings high rainfall over large areas of Asia especially in the South and Southeast Asian countries. Hydrological processes related to land surface and water resources management are strongly affected by the regional characteristics (Kondoh et al. 2004). The Intergovernmental Panel on Climate Change (Parry et al. 2007) reported that the mean surface air temperature in Southeast Asia increased at the rate of 0.1–0.3 °C per decade between 1951 and 2000. Consequently, the frequencies of extreme weather events such as heavy precipitation and tropical cyclones have increased considerably in Southeast Asia along with an increase in the interannual variability of daily precipitation in the Asia summer monsoon. These climatic changes have brought massive flooding, landslides, and droughts in different regions and have caused extensive damage to property, assets and human life. Studies on the Bagmati watershed in Nepal by Sharma & Shakya (2005) and Dhital & Kayastha (2012) explained that the frequency and duration of monsoon floods have increased since 1991 while their magnitudes have already reached the statistical 100-year flood. Delgado et al. (2010) described that in the Mekong River, likelihood of extreme floods has increased during the last decades although the probability of an average flood decreased.

Over global land monsoon regions, monsoon rainfall intensity showed a downward trend during 1950–2004 (Zhou et al. 2008). However, a regional study by Yao et al. (2008) showed that frequent extreme precipitation events are found over South and Southeast Asia, with the exception of a narrow zone over the Indo-China peninsula along 100 °E. At low latitudes, there are both regional increases and decreases of rainfall over land areas, and increased rainfall intensity, particularly during the summer (southwest) monsoon, could increase the extent of flood-prone areas in temperate and tropical Asia (Houghton et al. 2001; Yao et al. 2008). Myanmar's location in the transition zone between the South Asian and East Asian monsoon systems results in a particularly complex spatial pattern of precipitation variability which is not very well understood (D'Arrigo et al. 2013). With a distinct nature, Myanmar's rainfall has no significant relationship with the contiguous area of India and Bangladesh, even though all of them are under a similar weather system (Sen Roy & Kaur 2000). According to the study of the Association of Southeast Asian Nations Disaster Risk Management Initiative (ASEAN DRMI 2010), a catastrophic 200-year flood (0.5% annual probability of exceedance) would have a major impact on ASEAN countries’ economies, which are already fragile. In a comparative analysis of social vulnerability to disaster risks, among ASEAN countries Myanmar is the second worst-affected country after Indonesia, with 3,480 killed per year from natural hazards.

To the knowledge of the authors, few works on the hydrologic characteristics of the large-scale basins in Myanmar have been reported and particularly the Chindwin catchment has received relatively little attention. Further motivation is the public consensus on an increase in flood damage and risk in recent decades. This paper is intended to assess the hydrologic aspects of monsoon floods along the Chindwin River in northern Myanmar. An attempt was also made to identify the flood probability of the mesoscale Chindwin River Basin in the tropical monsoon country. As floods are complex and dynamic processes characterized by spatial and temporal variations, in this paper special attention was paid to the understanding of runoff behavior and its changes which play a crucial role in operational flood assessment.

Regional situation and monsoon floods

Myanmar is located in the northwestern part of the Indo-China peninsula, between 9°32′N and 28°31′N latitudes (with most of the area between the Tropic of Cancer and the Equator) and between 92°10′E and 101°11′E longitudes. Based on topographic conditions, Myanmar is divided into three parts – the western ranges (Himalayan ranges that divide India and Myanmar), the central plains (Ayeyarwaddy delta and other river basins) and the eastern hilly regions (Shan Plateau). River basin characteristics in Myanmar are quite variable due to the physiographic differences (Ti & Facon 2004).

Myanmar is the second biggest country in Southeast Asia, which is characterized by tropical rain forest and monsoon climates with a high and constant seasonal rainfall (Parry et al. 2007). Due to the diverse topographic conditions the climate varies across the country. As stated by Htway & Matsumoto (2011), the southwest monsoon advances in southern Myanmar and onset date is end of May, though the onset date may differ year to year. The monsoon periods with their areal average monthly rainfalls and temperatures (1960–2009) are shown in Figure 1. There is the southwest (summer) monsoon (June to September) with a cloudy, rainy, hot, humid summer and the northeast (winter) monsoon (December to April) with less clouds, scanty rainfall, lower humidity and moderate temperatures. Annual rainfall ranges from as high as 4,000–6,000 mm along the coastal reaches and in the western mountainous region to as low as 500–1,000 mm in the central dry zone. Two-thirds of the country lies in the tropics and one-third in the temperate zone. Lying within the tropics and the great Asiatic continent to the north and the wide expanse of the Indian Ocean to the south, Myanmar furnishes one of the best examples of a monsoon country. Extreme events in Myanmar such as heavy precipitation during the southwest monsoon vary across the country, depending on the monsoon intensity in the Bay of Bengal, while droughts are related to El Niño and El Niño Southern Oscillation on a global scale as well as to the regional monsoon trough and synoptic situations (Houghton et al. 2001; Myint et al. 2011).

Figure 1

Typical monsoon seasons of Myanmar.

Figure 1

Typical monsoon seasons of Myanmar.

The complex topography of this mountainous country, high rainfall intensities and the large number of glaciers mean that Myanmar is highly exposed to flood hazards. While the contrast between the Asian continent and the surrounding oceans drives the large-scale swing of the monsoon, the regional distribution of monsoon rain is governed, to a large part, by orography (Liu et al. 2005). According to the Department of Meteorology and Hydrology (DMH) Myanmar, flood occurrence in Myanmar can be generally recognized as 6% in June, 23% in July, 49% in August, 14% in September and 8% in October. Recorded data by the DMH reveal that several severe floods have occurred in major rivers in Myanmar during recent decades: for example, 2004 Ayeyarwaddy; 1991, 2002, 2011 Chindwin; 2002, 2011 Thanlwin; 1997, 2011 Sittaung; and 1995, 2011 Bago. In the public's perception, there seems to be a trend of frequent hydrological extreme events, leading to a high risk of flood hazards. Floods usually occur every year in one river system or another during the southwest monsoon. The number of recorded floods with significant impacts continues to rise, making floods one of the most costly natural hazards. A hazard-specific distribution and the impacts of the various disasters that occurred in Myanmar during 1970–2009 are shown in Figure 2. In contrast to the prevailing flood hazards, the present flood management system in Myanmar is not satisfactory for most situations. Moreover, data scarcity still hampers the application of distributed hydrological models for predicting streamflow over a range of spatial scales.

Figure 2

Disaster impacts in Myanmar during 1970–2009. Source: ASEAN DRMI (2010).

Figure 2

Disaster impacts in Myanmar during 1970–2009. Source: ASEAN DRMI (2010).

Study area

The study was conducted for the Chindwin River Basin with a catchment area of 113,800 km2 and a length of 985 km, in northern Myanmar as shown in Figure 3. It is the third largest river and one of the principal water resources of the country. It is comparable in size with the Elbe River, which has a catchment area of 148,268 km2 and a length of 1,091 km, which is the fourth largest river in Europe. The Chindwin's catchment is a mountainous forested terrain with the exception of its lowest southern part, which comprises a wide flood plain. The Chindwin River flows down through the Hukaung valley, Hkamti, Homalin, and Mawlaik to the southwest and changes its course at Kalewa in the southeast, crossing a number of vast plains and finally joins the Ayeyarwaddy River which is one of the major rivers in Asia (Figure 4). The Chindwin River is 350 m wide near Hkamti, and spreading over the meridian direction. Downstream of the defile, the river valley gets wider, with a width of 1,200 m near Monywa, and flow velocities slow down when flowing through the central flat terrain. The main tributaries are: the U Yu River just below the Homalin with a catchment area of 7,485 km2, the Yu River above Mawlaik with 6,423 km2 and the largest tributary, Myitthar River at Kalewa, with a 25,563 km2 catchment. The Chindwin with its major tributaries is the most convenient means of communication within the basin connecting it with the main economically developed areas of the country. The river basin occupies almost the entire northwestern part of Myanmar and is important to the development of the country.

Figure 3

Location of the Chindwin River Basin.

Figure 3

Location of the Chindwin River Basin.

Figure 4

Longitudinal profile of the Chindwin River from head to confluence.

Figure 4

Longitudinal profile of the Chindwin River from head to confluence.

Chikamori et al. (2012) reported that the Chindwin basin is mainly formed by tertiary continental sediments such as sandstones of different hardness, shale and limestone. There is also exposure of crystalline rocks (granite, granite-gneisses, diorites, etc.) stretching to the east of Chindwin and a range of hills to the north of Monywa station. Closed forest covers 50% of the basin area. The second dominant land type is degraded forest including shrubs which covers about 33%. Agricultural land, alluvial island cultivation and homestead gardening cover 15% of the basin area while shifting cultivation and swamp areas represent 2%. Though the social and economic conditions have changed and the population in the basin has increased in recent years, urbanization effects do not seem to have greater impacts on the basin hydrology compared with the geomorphic conditions. The important role of the basin in national socio-economic development is hampered by flood hazards due to climate conditions. Severe floods hit the Chindwin basin every year at one place or another due to high rainfall intensities during the southwest monsoon. Since 1965, flood occurrences in the Chindwin basin have been highest in July and August contributing 72% of the total number of floods in the basin.

MATERIALS AND METHOD

Data analyzed in this paper mainly consist of daily discharges and rainfall (1966 to 2011) at five gauging stations and relevant hydro-meteorological data of the basin. The gauging stations also refer to the outlets of the sub-basins of the Chindwin catchment along its main course. Geographic information system (GIS) was used to characterize catchment parameters. The hydro-meteorological data are mostly obtained from the DMH in Myanmar and from the numerous research articles and published reports which are cited in the reference list. A simple and straightforward approach was applied to characterize the rainfall and flood runoff patterns over the basin based on the analysis of observed hydro-meteorological records and catchment characteristics.

Data review and analysis

Through a GIS process, catchment characteristics were extracted from the global digital elevation model (DEM) available from the USGS (EROS) Hydro 1 K Asia database as well as from ground survey data and maps. EsriArcGIS 10 software was used to analyze spatial parameters of the watershed. First, raw DEM data were corrected to create a depressionless DEM by filling the sinks. Then watershed characterization was done by using the ArcHydrol Tools function in ArcGIS which processes and analyzes the corrected DEM data to characterize topography, measure basin parameters, identify surface drainage, subdivide watersheds, and quantify the drainage network. Base flow index (BFI) program was used for identification of annual base flow contributions of the total annual runoff. Among several hydrograph separation methods, the program implements a deterministic procedure developed by the British Institute of Hydrology and the method combines a local minimums approach with a recession slope test (Wahl & Wahl 1995). This program was included in the Developments in Water Science Series, volume 48, edited by Tallaksen & van Lanen (2004). Long-term variability of annual maximum floods in the Chindwin River was studied using flash flood magnitude index (FFMI), suggested by Kale (2003), which is the standard deviation of the logarithms of annual peak discharge, for defining the interannual variability of monsoon floods. A literature review on different aspects of water resources in Myanmar (Ti & Facon 2004; ADB 2009; ASEAN DRMI 2010; Chikamori et al. 2012) helps in understanding the study area and provides some preliminary conclusions. Particular attention was paid to information on flood hydrology. Statistical and regional analysis was applied to the observed streamflow and rainfall data at the five gauging stations of the Chindwin River. Mean, standard deviation and coefficient of variation were mainly used to understand the general hydrology of the watershed. Statistical analysis was done using SPSS 19 software and Microsoft Excel 2010.

Time series analysis

Time series analysis involves applying a linear regression to detect the trends of annual maxima (AM) series and their deviation trends. The significance of a linear trend was assessed using a linear regression function in SPSS 19. The issue of a linear trend is whether the slope value is significantly different from zero (i.e. no trend). The slope is the average rate of change over the years being examined. For the test of significance, the P-value was determined by referring to a t-distribution. The test statistic was calculated dividing the estimated slope coefficient by the estimated standard error. The null hypothesis of no trend is rejected if the p value is smaller than the significance level. In this study, a trend was considered to be significant at 5% significance level. If the p value is less than or equal to 0.05, there is a significant trend. If not, there is not enough evidence of a meaningful trend at this significance level.

Frequency analysis

Frequency analysis was applied to evaluate the probability of flood occurrences and possible trends during the observation period. Having listed a series of annual maximum floods, they were then ranked in descending order. The empirical recurrence interval (T) and probability (P) of the data could be computed using a plotting position and fitted by Log-Pearson Type III (LP-3) distribution. LP-3 distribution is mostly recommended in flood frequency analysis and is used for design purposes. It is determined by three parameters: mean, standard deviation and the skewness coefficient. Using the AM of the entire study period (1966 to 2011), the hypothetical floods were determined for the return periods of 2 to 1,000 years and then the flood risks were defined in terms of probability for each station along the main river. In the next step, the time series was split into two time phases: TP-1 (1966 to 1990) and TP-2 (1991 to 2011). The moving average method was used to determine at which year time series were split best into two periods. The moving average method smoothes the fluctuations of AM series. The year at which an obvious change occurs in the moving average series was taken as the trend change year. With this, the time series are also divided into two partial series of similar length. To detect the changes in flood quantiles, the relative differences of flood quantiles in two time phases with respect to the entire observation period were compared.

Based on the flood frequencies, the index flood method could be applied to derive the regional frequency curve (Riggs 1982). The basic premise of this method is that a combination of streamflow records obtained at a number of gauging stations will produce a more reliable, not a longer, record, and thus will increase the reliability of frequency analysis within a region (Jain & Singh 2003). A dimensionless frequency curve representing the ratio of the flood of any frequency to an index flood, which is the mean annual flood, was generated.

RESULTS AND DISCUSSION

Hydrology

Orographic effects, as a natural barrier to the southwest monsoon in the meridian direction, cause much greater precipitation on the west side of the mountains and a rain shadow on the east side. Further, from the source of the Chindwin to its mouth the amount of rainfall also decreases because of the western disturbance and tropical cyclones in the Bay of Bengal. Assessment of the long-term average annual rainfall demonstrates that the spatial variation of rainfall was also influenced by topography. As a monsoon-dominated catchment, riverine floods and flash floods are most common to the Chindwin River when intense rainfall persists at the headwaters of the basin seasonally and annually. Generally the northern part of the Chindwin basin receives an average of 3,800 mm per year, while the lower (southern) part of the basin gets only 760 mm. Nearly 90% of the rainfall in the northern part and 75% in the southern part of the basin falls between June and October. In a flood warning context, floods are expected when critical values are exceeded. The DMH has defined the threshold discharge for a given cross-section of the Chindwin River. Threshold runoffs are 13,500 m3 s−1 at Hkamti, 14,200 m3 s−1 at Homalin, 16,200 m3 s−1 at Mawlaik, 16,800 m3 s−1 at Kalewa and 19,000 m3 s−1 at Monywa.

 Figure 5 shows the annual rainfall comparison at different stations and typical rainfall pattern and streamflow of the upper Chindwin basin. It can be seen that intense rain falls almost every day during the southwest monsoon season. Lag or response time of rainfall to flow is about 1 to 3 days depending on the rainfall distribution. While single peak floods are characteristic during the southwest monsoon season especially in the lower basin, sometimes multiple peak events occur due to rain on successive days mostly in the upper Chindwin basin. They are imposed on the annual flood hydrograph, which is unique for any given year but similar in shape between different years. Annual cycle and flood season are stable and defined by monsoon precipitation that arrives approximately at the same time of the year. The high fluctuation of streamflow patterns in the Chindwin catchment is influenced by the extreme variation of rainfall in the region.

Figure 5

Annual rainfall and typical rainfall pattern of the Chindwin River. (a) Annual rainfall and (b) typical rainfall and streamflow pattern of the Chindwin River.

Figure 5

Annual rainfall and typical rainfall pattern of the Chindwin River. (a) Annual rainfall and (b) typical rainfall and streamflow pattern of the Chindwin River.

Generally monsoon floods display temporal patterns characterized by long period fluctuations and non-random behavior in terms of discrete periods of low and high floods (Kale 1999). The high rises of water levels and discharges are often noticed in July and August, the mid-season of the monsoon when soils are saturated with water concurrently and infiltration during intense rainfall is less. Nevertheless, early and late monsoon floods should also be considered as common events. According to past flood events since 1966, the highest river level reached 4 m above the danger level at the upper Chindwin basin contributing discharges of 19,613 m3 s−1 and 19,400 m3 s−1 at Hkamti in 1991 and 1997, 26,773 m3 s−1 and 26,443 m3 s−1 at Mawlaik in 1976 and 2002, and 26,220 m3 s−1 at Kalewa (lower basin) in 2002. The severe floods, which were over critical levels on 22 July 2004 (17,673 m3 s−1) at Hkamti, 15 July 1997 (19,470 m3 s−1) at Homalin, 01 September 1999 at Mawlaik (24,093 m3 s−1), 10 July 2008 (23,720 m3 s−1) at Kalewa and 19 August 2002 (23,957 m3 s−1) at Monywa, lead to the assumption that extreme events have occurred more often in the last two decades. Duration of floods above the danger levels varied from 9 to 18 days. The danger level corresponds to the crest height of the levees. Historical flood records indicated that the middle part of the basin also has a high potential flood hazard although rainfall is not as high as in the most upstream catchment. During the southwest monsoon, flood flows at the middle station (Mawlaik) are moderately correlated with those of the upstream station, Hkamti (R2 = 0.5), but strongly correlated with the lower station, Monywa (R2 = 0.8). The latter situation is likely due to the similar geomorphic condition and rainfall distribution over these areas.

According to Kondoh et al. (2004) Myanmar belongs to region B2 of Monsoon Asia where the climate is characterized by distinct wet and dry seasons with a large water deficit in the dry season. Maximum mean annual temperature at the Monywa station (lower part of the basin) is 29 °C whereas the minimum mean annual temperature amounts to 25 °C. Average annual evaporation is high with about 1,400 mm/a in the southern part of the basin to 1,000 mm/a in the north. This causes an average basin loss of about 45% of total rainfall. From daily streamflow records, annual base flow contributions were determined and expressed in BFI, which is the ratio of base flow to total flow volume for a given year. The average specific discharge at five stations ranges between 0.20 m3 s−1 km−2 and 0.53 m3 s−1 km−2. Basin area and slope, river length and slope are determined using GIS based on the DEM of 1 km2 resolution. All parameters, together with basin properties, are shown in Table 1.

Table 1

Catchment characteristics of the Chindwin Basin

ParameterHkamtiHomalinMawlaikKalewaMonywa
Location N-26° 00′ E-95° 42′ N-24° 52′ E-94° 55′ N-23° 38′ E-94° 25′ N-23° 12′ E-94° 18′ N-22° 06′ E-95° 08′ 
Basin area (km227,210 49,137 69,057 99,072 113,814 
River length (km) 347 546 660 741 985 
River slope 0.0013 0.0002 0.0004 0.0002 0.0003 
Basin slope 0.104 0.122 0.118 0.120 0.114 
Mean annual rainfall (mm) 3,830 2,287 1,738 1,685 764 
Mean annual max flow (m3 s−114,387 16,243 19,542 20,054 20,603 
Specific discharge (m3 s−1 km−20.53 0.33 0.28 0.20 0.18 
BFI 0.61 0.69 0.73 0.74 0.77 
ParameterHkamtiHomalinMawlaikKalewaMonywa
Location N-26° 00′ E-95° 42′ N-24° 52′ E-94° 55′ N-23° 38′ E-94° 25′ N-23° 12′ E-94° 18′ N-22° 06′ E-95° 08′ 
Basin area (km227,210 49,137 69,057 99,072 113,814 
River length (km) 347 546 660 741 985 
River slope 0.0013 0.0002 0.0004 0.0002 0.0003 
Basin slope 0.104 0.122 0.118 0.120 0.114 
Mean annual rainfall (mm) 3,830 2,287 1,738 1,685 764 
Mean annual max flow (m3 s−114,387 16,243 19,542 20,054 20,603 
Specific discharge (m3 s−1 km−20.53 0.33 0.28 0.20 0.18 
BFI 0.61 0.69 0.73 0.74 0.77 

 Figure 6(a) shows the FFMI against the ratios of observed highest discharge and average annual maximum discharge at different stations. FFMI values range between 0.08 and 0.1 showing that the interannual variability in flood peaks was not very different from station to station. Figure 6(b) shows the box plot of AM standardized by the basin area. The annual maximum flood relative to the basin area is highest at Hkamti followed by Homalin, Mawlaik, Kalewa and Monywa. Flood generation decreases with the increase in the catchment area. The standardized AMs are more variable over time at Hkamti and Mawlaik, with higher standard deviations than at other stations.

Figure 6

Variability of annual floods: (a) FFMI against the ratio of observed largest floods to mean annual maximum discharges; (b) distribution of standardized AM.

Figure 6

Variability of annual floods: (a) FFMI against the ratio of observed largest floods to mean annual maximum discharges; (b) distribution of standardized AM.

Time series and frequency analysis of floods

The trends in AM and variability of extreme floods were evaluated. Temporal trends of AM series and their deviations were detected at five stations along the Chindwin River using a linear regression. Annual maximum discharges (Qmax) and deviation trends (1967–2011) at two selected gauging stations are shown in Figure 7. Table 2 shows the slope factors and p-values as the significance test for a linear trend.

Table 2

Linear trend statistics of two time series for different stations

AMDeviation of AM from the means
StationSlopeP-valueSlopeP-value
Hkamti −21 0.6 55 0.001 
Homalin −28 0.4 30 0.150 
Malwaik −42 0.5 68 0.050 
Kalewa −28 0.6 25 0.320 
Monywa −10 0.8 45 0.068 
AMDeviation of AM from the means
StationSlopeP-valueSlopeP-value
Hkamti −21 0.6 55 0.001 
Homalin −28 0.4 30 0.150 
Malwaik −42 0.5 68 0.050 
Kalewa −28 0.6 25 0.320 
Monywa −10 0.8 45 0.068 
Figure 7

Linear trends of annual Qmax at (a) Hkamti and (b) Mawlaik and deviation of annual Qmax from mean value at (c) Hkamti and (d) Mawlaik.

Figure 7

Linear trends of annual Qmax at (a) Hkamti and (b) Mawlaik and deviation of annual Qmax from mean value at (c) Hkamti and (d) Mawlaik.

AM series at all stations show slightly decreasing trends with negative slopes. But trends were not significant as p-values are much greater than 0.05. Thus, the AM series of the Chindwin River can be regarded as stable for the observation period. The statistical mean value was not changing with time and there was no significant trend. However, the deviation of annual peaks from their mean (regardless of positive or negative) showed increasing trends with positive slopes at all stations. Deviation trends at Hkamti and Mawlaik stations were highly significant at the 5% level with p-values of 0.001 and 0.05, respectively. At Monywa station, the deviation of AM series showed an increasing trend with a marginal significance (p = 0.068). At Homalin and Kalewa stations, there was not enough evidence for meaningful trends at the 5% significant level.

The trend to a higher deviation and thus variation of annual maximum floods is in accordance with the hypotheses of climate change, making the region experience more extreme events. For example, ADB (2009) reported that Southeast Asia is one of the world's most vulnerable regions to the impact of climate change in terms of frequency and intensity of extreme weather events which are projected to increase. Furthermore, frequent extreme precipitation over Southeast Asia is predicted in the next few decades due to climate change and that is definitely going to worsen the flooding situation in the region (Houghton et al. 2001; Yao et al. 2008; Turral et al. 2011). Statistically, the time series are thus not homogeneous.

In many cases, large rivers with the highest annual variability may have potential impacts from dams because of substantial control over downstream hydrology (Graf 2006). However, flow changes by dam building were not evident in the Chindwin watershed. According to the collected data from the Myanmar Irrigation Department, there is no dam across the main stream yet and only three dams have been implemented in recent years on the Chindwin's tributaries. Locations of the dams are on the Neyinsara River (23°31′ N and 94°06′ E), on the Manipur River (22°58′ N and 93°58′ E), and on the Myitthar River (22°00′ N and 94°02′ E). But none of these dams is finished yet and regulated controls are not capable of exerting substantial influence on downstream hydrology. Although their impacts could not be fully assessed, the damming was not a prime influencing factor for the variability of floods whose magnitudes have increased in the last two decades. Furthermore, as stated by Bruijnzeel (2004), it is difficult to evaluate the effects of land use change on flood peaks in large rivers because such changes are rarely fast and consistent with the exception of high population pressure and thus often compounded by climatic variability. High variability of streamflow could be expected as an effect of an interannual structure in regional climate as well as changes in monsoon intensity. Regarding other monsoon-dominated rivers in Southeast Asia, Delgado et al. (2010) came to a similar conclusion that the variation of extreme floods in the Mekong River was validated with the precipitation data, which suggests climatic causes for the increase in variability.

Frequency analysis was then applied to annual maximum discharges of five stations during 1966 to 2011. Although not all stations covered the same data length, all streamflow data used span over 40 years of records. Mean, standard deviation, skewness coefficients and coefficients of variation (CV) of the annual maximum series of the different stations are shown in Table 3. Probable floods at each gauging station were computed by using the LP-3 distribution for different return periods of 2 to 1,000 years. The expected hypothetical floods were compared with the highest observed flows in the past 40 years as shown in Figure 8. Comparing the expected floods with the highest observed floods, suggests that these correspond in the upper mountainous catchment (Hkamti and Homalin) to about 100-year events. In the central flat terrain with medium elevations (Mawlaik) the maximum observed flood has a statistical return interval of about 50 years, while in the lower part of the dry zone area (Kalewa and Monywa) it is about 15 years. It is generally assumed that the reliability of statistical analyses increases with the length of time series or the number of data. An important precondition for this, however, is that the set of data is homogeneous, drawn from a single data population. As found above, this condition is not fulfilled for the flood data of the Chindwin River. There was a rise in standard deviation particularly in the last two decades caused by outer (most extreme) influences (Table 3). The most likely effects on this high flood variability are the changes in rainfall intensity and pattern in the region, as discussed above. The standard deviations of AM series for the data from the last two decades are 2.4, 1.5, 1.5, 1.3 and 1.4 times greater than that of the period 1966–1990 at Hkamti, Homalin, Mawlaik, Kalewa and Monywa, respectively.

Table 3

Statistical parameters of annual maximum series

Mean (m3 s−1)Standard deviation (m3 s−1)Skew coefficientCV
StationEntire period1966–19901991–2011Entire period1966–19901991–2011Entire period1966–19901991–2011Entire period
Hkamti 14,387 14,582 14,165 2,481 1,404 3,343 −0.31 −0.24 −0.80 0.17 
Homalin 16,243 16,782 15,653 2,715 2,128 3,189 −0.64 0.32 −0.66 0.17 
Mawlaik 19,542 20,017 19,113 3,931 3,001 4,650 −0.31 −0.05 −0.20 0.20 
Kalewa 20,054 20,504 19,518 4,175 3,663 4,750 - − 0.36 −0.40 −0.21 0.21 
Monywa 20,603 20,804 20,362 4,103 3,480 4,820 −0.02 0.16 0.07 0.20 
Mean (m3 s−1)Standard deviation (m3 s−1)Skew coefficientCV
StationEntire period1966–19901991–2011Entire period1966–19901991–2011Entire period1966–19901991–2011Entire period
Hkamti 14,387 14,582 14,165 2,481 1,404 3,343 −0.31 −0.24 −0.80 0.17 
Homalin 16,243 16,782 15,653 2,715 2,128 3,189 −0.64 0.32 −0.66 0.17 
Mawlaik 19,542 20,017 19,113 3,931 3,001 4,650 −0.31 −0.05 −0.20 0.20 
Kalewa 20,054 20,504 19,518 4,175 3,663 4,750 - − 0.36 −0.40 −0.21 0.21 
Monywa 20,603 20,804 20,362 4,103 3,480 4,820 −0.02 0.16 0.07 0.20 
Figure 8

Observed maximum floods and expected flows with different return periods.

Figure 8

Observed maximum floods and expected flows with different return periods.

Taken strictly, the frequency analysis of the entire observation period was not admissible. Therefore, the AM series should be analyzed in different time phases. A moving average with seven spans was used to check the trend change at all stations. The year when the AM series were split is the year in which a trend of flood variability changes. Moving average series for five stations are given in Figure 9. Although the trend change years for all stations are not identical, with less fluctuation, the moving average series for all stations are generally smooth till 1990 and the higher fluctuations occurred in the later periods. In this study, detecting exact change points using possible statistical tests is not the main focus. Instead, the authors would like to point out that the deluge of extreme floods frequently occurred after 1990, especially concentrated in the last 20 years of the time series.

Figure 9

Moving average series of annual maxima at five stations.

Figure 9

Moving average series of annual maxima at five stations.

In Monsoon Asia, substantial regional features are associated with the changes in precipitation amounts and duration, and in southeast China, a sharp increase in extreme precipitation (>50 mm per day) occurred in 1993 (Yao et al. 2008). The change point for annual maximum flood series of the Wijiang River in South China was found in 1991 with an increase in mean of AM by 45% due to increased rainstorms (Chen et al. 2013). In Southeast Asia, extreme weather events associated with El Niño were more frequent and intense in the past 20 years (Parry et al. 2007). Interannual rainfall and temperature variability in Myanmar is affected by the ENSO (El Niño Southern Oscillation) patterns (Lwin 2006; Baroang 2013). With the effects of warmer temperature, the increased water vapor will possibly result in an increase in precipitation amount and intensity (Houghton et al. 2001; Wang et al. 2008). Thus extreme floods in the Chindwin catchment were found not only in strong La Niña but also in strong El Niño years, and the extreme floods associated with moderate to strong El Niño years have been observed since 1991.

A sharp increase in the maximum flood level (4.3 m above danger level) at Hkamti station occurred in 1991 and the occurrences of extreme floods (>2 m above the danger levels) are more frequent at Hkamti, Mawlaik and Kalewa stations from 1991 onwards. These regional and local situations show that higher interannual variations of streamflows in the Chindwin catchment are expected in the past 20 years. Through this reasoning, AM series are analyzed with two time phases: TP1 (1966–1990) and TP2 (1991–2011). Overall, 1990 was considered the trend change year for the Chindwin watershed because the deviations of AM series at all stations are also getting larger around this year. Modification of the rating curve for the Monywa station was carried out in 2009 and any changes of monitoring method which affect the rating curves could not be observed for all stations. Sometimes human activities are more important in the regional hydrologic regime (Yang et al. 2004); however, the variation of extreme floods in the Chindwin catchment after 1990 is probably due to climatic causes especially precipitation since the major human-induced impacts such as damming and land use changes were not evident in the catchment.

The two respective frequency curves, along with the plotting positions, for Hkamti and Mawlaik stations are shown in Figure 10. Parameters are given in Table 3. As expected, the frequency curves based on 1991–2011 records, with the exception of Homalin, provided higher floods. Hypothetical floods of the two frequency curves were then compared for the return periods. The increase of flood magnitudes in TP-2 with respect to TP-1 was expressed as a percentage as shown in Figure 11. Beyond the 5-year return period, expected floods are 2.5% to 26% higher for the data of TP-2. Only at Homalin station no significant difference was found.

Figure 10

Comparison of frequency curves with the most significant differences: (a) Hkamti; (b) Mawlaik.

Figure 10

Comparison of frequency curves with the most significant differences: (a) Hkamti; (b) Mawlaik.

Figure 11

Percentage increase of flood quantiles in comparison with TP-1 and TP-2 series.

Figure 11

Percentage increase of flood quantiles in comparison with TP-1 and TP-2 series.

To analyze the change in flood peak series, relative differences of flood quantiles in two periods were compared with respect to that of the entire period and results are shown in Figure 12. In the latter period (1991–2011), flood values are increasing 3% to 15% at different stations while the flood values of the former period (1966–1990) are decreasing up to 13% with the exception of Homalin station. In both comparisons of relative differences in flood values of two series with respect to the entire series, the highest difference (either high or low values) was found at the Hkamti station followed by Mawlaik, Monywa and Kalewa, respectively. This significance shows that the change in flood quantiles of the Chindwin River is decreasing from upstream to downstream.

Figure 12

Comparison of flood quantiles: (a) TP-1 and entire period; (b) TP-2 and entire period.

Figure 12

Comparison of flood quantiles: (a) TP-1 and entire period; (b) TP-2 and entire period.

The index-flood method was also applied here as a tool for regionalization of the basin using five station data. Dawdy & Gupta (1995) assess this method as unrealistic with its assumption of simple scaling, that the coefficient of variation is not changing with the increasing catchment area. Normally, CV decreases with the increasing catchment scale due to damping effects (Kuzuha et al. 2009). In the Chindwin basin, mean annual flood rises with the increase of catchment areas and CV is changing with increasing catchment as shown in Table 3. This agrees with the former argument. Flood indices (FI) were calculated as the ratios of expected floods with different return periods (Qe) to the mean of observed annual maximum series (Table 4). The regional frequency curve is shown in Figure 13. In averaging the entire basin, the expected floods may range between 0.6 and 1.5 times the mean annual maximum discharge. Average FI for all return periods at each station lie within 1.2 to 1.3, from upstream to downstream. The average value of the highest expected floods would not be much different from station to station. It is also noticeable that the expected floods, even with higher return periods, are not largely different from the average annual maximum floods of the basin.

Figure 13

Flood regionalization of the Chindwin River Basin.

Figure 13

Flood regionalization of the Chindwin River Basin.

Table 4

FI at different stations

HkamtiHomalinMawlaikKalewaMonywa
Return period (yr)Qe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIAverage FI
7,759 0.54 10,162 0.63 9,844 0.50 9,929 0.50 11,568 0.56 0.55 
14,436 1.00 16,438 1.01 19,640 1.01 20,126 1.00 20,130 0.98 1.00 
16,543 1.15 18,545 1.14 23,088 1.18 23,819 1.19 23,703 1.15 1.16 
10 17,520 1.22 19,591 1.21 24,789 1.27 25,659 1.28 25,687 1.25 1.24 
20 18,256 1.27 20,423 1.26 26,129 1.34 27,116 1.35 27,393 1.33 1.31 
50 18,988 1.32 21,310 1.31 27,530 1.41 28,651 1.43 29,381 1.43 1.38 
100 19,425 1.35 21,876 1.35 28,408 1.45 29,618 1.48 30,754 1.50 1.42 
200 19,783 1.38 22,371 1.38 29,159 1.49 30,450 1.52 32,042 1.56 1.46 
500 20,175 1.40 22,949 1.41 30,017 1.54 31,406 1.57 33,654 1.64 1.51 
1,000 20,411 1.42 23,328 1.44 30,555 1.56 32,019 1.60 34,807 1.69 1.54 
Average FI  1.20  1.21  1.29  1.27  1.31  
HkamtiHomalinMawlaikKalewaMonywa
Return period (yr)Qe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIQe (m3 s−1)FIAverage FI
7,759 0.54 10,162 0.63 9,844 0.50 9,929 0.50 11,568 0.56 0.55 
14,436 1.00 16,438 1.01 19,640 1.01 20,126 1.00 20,130 0.98 1.00 
16,543 1.15 18,545 1.14 23,088 1.18 23,819 1.19 23,703 1.15 1.16 
10 17,520 1.22 19,591 1.21 24,789 1.27 25,659 1.28 25,687 1.25 1.24 
20 18,256 1.27 20,423 1.26 26,129 1.34 27,116 1.35 27,393 1.33 1.31 
50 18,988 1.32 21,310 1.31 27,530 1.41 28,651 1.43 29,381 1.43 1.38 
100 19,425 1.35 21,876 1.35 28,408 1.45 29,618 1.48 30,754 1.50 1.42 
200 19,783 1.38 22,371 1.38 29,159 1.49 30,450 1.52 32,042 1.56 1.46 
500 20,175 1.40 22,949 1.41 30,017 1.54 31,406 1.57 33,654 1.64 1.51 
1,000 20,411 1.42 23,328 1.44 30,555 1.56 32,019 1.60 34,807 1.69 1.54 
Average FI  1.20  1.21  1.29  1.27  1.31  

CONCLUSIONS

Severe floods with high rainfall during the southwest monsoon are the most serious natural disasters in Myanmar. Flood risk management, however, is not yet well-developed. Flood characteristics and trends at five gauging stations along the Chindwin River are evaluated. Under the regional climate the Chindwin is a perennial river with high seasonal variation in discharge and rainfall, since up to 90% of rainfall occurs during the southwest monsoon. In the past flood levels have reached up to 4 meters over the bank level at Hkamti in 1991 and 1997, at Mawlaik and Kalewa in 2002, resulting in relatively large inundations. Duration of most floods generally lasts over 10 days and such long-lasting floods are frequent occurrences in the basin seasonally and annually.

Statistical analysis shows that there are no trends of the mean values of annual maximum discharges but strong trends in standard deviation and probability curves along the main stream. Particularly at the upper station (Hkamti) and the middle station (Malwaik), over the last two decades the standard deviations of AM have significantly increased 2.4 and 1.5 times, respectively, compared with the period 1966–1990. The study pointed out the remarkable increase in flood occurrence as well as the increasing trend of deviation of annual maximum floods with time since 1990. Frequency analysis of the entire observation period indicates that the river has already experienced a 100-year flood in the upper part, 50-year flood in the middle and a 15-year flood in the lower part. The paper also concludes that flood risk is high in the upper and middle regions of the basin. To detect the changes in flood quantiles, two distinct phases were studied: 1966 to 1990 with small deviation of annual maximum discharge and 1991 to 2011 with increasing trend of high floods. As the basin is not well industrialized and urbanized, except in Monywa, any substantial changes in land management were not evident in the catchment. However, the cause for the increasing variability in annual peaks is probably the interannual structure of regional climate and changes in monsoon intensity. It is recommended that interannual variation of streamflow should be validated with the detailed spatial precipitation data at the regional level. Attention should also be paid to human activities which alter regional hydrologic regimes, affecting long-term changes in streamflow at both seasonal and regional scales.

As a result, better understanding of monsoon flood characteristics of the Chindwin River Basin would contribute to improved flood hazard management because an effective mitigation plan could never be realized without a proper understanding and assessment of the regional characteristics. The study also gives motivation for further analysis of flood extents in this poorly gauged basin with finer spatial scale. It also suggests that flood management could benefit by paying more attention to the area with relatively high probability floods with the application of reliable forecasting methods coupling with inundation assessments.

ACKNOWLEDGEMENTS

The study was funded by the Deutscher Akademischer Austausch Dienst (German Academic Exchange Service). The hydrometric data used in this analysis were provided by the DMH, Myanmar.

REFERENCES

REFERENCES
ADB
2009
The Economics of Climate Change in Southeast Asia: A Regional Review
.
Asian Development Bank
,
Manila
.
ASEAN Disaster Risk Management Initiative
2010
Synthesis Report on Ten ASEAN Countries Disaster Risks Assessment
.
World Bank, Association of Southeast Asian Nations (ASEAN), United Nations Office for Disaster Risk Reduction – Regional Office for Asia and Pacific (UNISDR AP), Global Facility for Disaster Reduction and Recovery (GFDRR)
.
Baroang
K.
2013
Myanmar Bio-Physical Characterization: Summary Findings and Issues to Explore. Background paper No.1. Report for USAID/Burma
.
Borga
M.
Anagnostou
E. N.
Blosch
G.
Creutin
J. D.
2011
Flash flood forecasting, warning and risk management: the HYDRATE project
.
Env. Sci. Pol.
14
(
7
),
834
844
.
Chen
X.
Zhang
L.
Xu
C. Y.
Zhang
J.
Ye
C.
2013
Hydrological design of nonstationary flood extremes and durations in Wijiang River, South China: changing properties, causes, and impacts
.
Mathematical Problems in Engineering
Article ID 527461. 10 pp. doi:10.1155/2013/527461
.
Chikamori
H.
Heng
L.
Daniell
T.
(eds)
2012
Catalogue of Rivers for Southeast Asia and the Pacific
Volume VI.
UNESCO-IHP Regional Steering Committee for Southeast Asia and the Pacific
,
Indonesia
.
D'Arrigo
R.
Palmer
J.
Ummenhofer
C.
Kyaw
N. N.
Krusci
P.
2013
Myanmar monsoon drought variability inferred by tree rings over the past 300 years: linkages to ENSO
. In:
El Nino-Southern Oscillation
(
Braconnot
P.
Brierley
C.
Harrison
S. P.
von Gunten
L.
Kiefer
T.
, eds).
PAGES news 21 (2), 50–51. PAGES (Past Global Changes). Bern, Switzerland
.
Daniel
T. M.
Daniel
K. A.
2006
Human impacts, complexity, variability and non-homogeneity: four dilemmas for the water resources modeler
. In
Climate Variability and Change: Hydrological Impacts. Proceedings of the Fifth FRIEND World Conference
,
IAHS Publ. 308
, pp.
10
15
.
Dawdy
D. R.
Gupta
V. K.
1995
Multiscaling and skew separation in regional floods
.
Water Resour. Res.
31
(
11
),
2761
2767
.
Delgado
J. M.
Apel
H.
Merz
B.
2010
Flood trends and variability in the Mekong river
.
Hydrol. Earth Syst. Sci.
14
,
407
418
.
Dhital
Y. P.
Kayastha
R. B.
2012
Frequency analysis, causes and impacts of flooding in the Bagmati River Basin, Nepal
.
J. Flood Risk Manage.
6
(
3
),
253
260
.
doi:10.1111/jfr3.12013.
Dutta
D.
Herath
S.
2004
Trends of floods in Asia and flood risk management with integrated river basin approach
.
Proceedings of the Second International Conference of Asian-Pacific Hydrology and Water Resources Association, Singapore, 1
, pp.
55
63
.
Houghton
J. T.
Ding
Y.
Griggs
D. J.
Noguer
M.
van der Linden
P. J.
Dai
X.
Maskell
K.
Hohnson
C. A.
(eds)
2001
Climate Change 2001: The Scientific Basis
.
Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
.
Htway
O.
Matsumoto
J.
2011
Climatological onset dates of summer monsoon over Myanmar
.
Int. J. Climatol.
31
,
382
293
.
Jain
S. K.
Singh
V. P.
(eds)
2003
Water resources systems planning and management
.
Dev. Water Sci.
51
,
3
858
.
Kale
V. S.
1999
Long-period fluctuations in monsoon floods in the Deccan Peninsula, India
.
J. Geol. Soc. India
53
,
5
15
.
Kondoh
A.
Harto
A. B.
Eleonora
R.
Kojiri
T.
2004
Hydrological regions in monsoon Asia
.
Hydrol. Process.
18
,
3147
3158
.
Kuzuha
Y.
Tomosugi
K.
Kishii
T.
Komatsu
Y.
2009
Coefficient of variation of annual flood peaks: variability of flood and rainfall intensity
.
Hydrol. Process.
23
,
546
558
.
Liu
W. T.
Xie
X.
Tang
W.
2005
Monsoon, orography, and human influence on Asian rainfall
.
Proceedings of the First International Symposium in Cloud-prone & Rainy Areas Remote Sensing (CARRS)
,
Chinese University of Hong Kong
.
Lwin
T.
2006
The impact of El Nino and La Nina events on the climate of Myanmar
.
12th Pacific Congress (PACON) on Marine Science and Technology in Asia
.
Myint
U. T.
Thaw
S. H.
Nyein
Y. Y.
2011
Overview of droughts in Myanmar
. In:
Drought in Asia Monsoon Regions
(
Shaw
R.
Nguyen
H.
, eds).
Community, Environment and Disaster Risk Management Series, Volume 8
.
Emerald Group Publishing
,
UK
, pp.
87
95
.
Parry
M.
Canziani
O.
Palutikof
J.
Linden
P.
Hanson
C.
(eds)
2007
Climate Change 2007: Impacts, adaptation and vulnerability
.
Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change
.
Cambridge University Press
,
Cambridge
.
Riggs
H. C.
1982
Regional Analysis of Streamflow Characteristics
. Techniques of Water Resources Investigation of the United States Geological Survey.
United States Government Printing Office
,
Washington, DC
.
Sen Roy
N.
Kaur
S.
2000
Climatology of monsoon rains of Myanmar (Burma)
.
Int. J. Climatol.
20
,
913
928
.
Sharma
D.
2012
Situation analysis of flood disaster in south and Southeast Asia: a need of integrated approach
.
Int. J. Sci., Environ. Technol.
1
(
3
),
167
173
.
Sharma
R. J.
Shakya
N. M.
2005
Hydrological changes and its impacts on water resources of Bagmati watershed, Nepal
.
J. Hydrol.
327
(
3–4
),
315
322
.
Tallaksen
L. M.
van Lanen
H. A. J.
(eds)
2004
Hydrological Drought: Processes and Estimation Methods for Streamflow and Groundwater
.
Developments in Water Science Series
, vol.
48
,
Elsevier Science BV
.
Ti
L. H.
Facon
T.
2004
From Vision to Action: A Synthesis of Experiences in Least-Developed Countries in Southeast Asia. FAO-ESCAP Pilot Project on National Water Vision Phase II
.
RAP Publication 32
,
Bangkok
.
Turral
H.
Burke
J.
Faurès
J.
2011
Climate Change, Water and Food Security. FAO Water Reports 36
,
Rome
.
Wahl
K. L.
Wahl
T. L.
1995
Determining the Flow of Comal springs at New Braunfels, Texas
. In
Proceedings of Texas Water ‘95, American Society of Civil Engineers
,
San Antonio, Texas
, pp.
77
86
.
Yang
D.
Ye
B.
Shiklomanov
A.
2004
Discharge characteristics and changes over the Ob River watershed in Siberia
.
J. Hydrometeorol.
5
(
4
),
595
610
.