The discovery of the role of global teleconnections in the seasonal precipitation patterns across the world has led to a substantial improvement in forecasting skills during the last few decades. One of the most widely researched global teleconnections index is in the form of El Niño Southern oscillation (ENSO). ENSO is broadly defined as a disruption in ocean surface temperatures and atmospheric circulation over the Pacific Ocean leading to wide spread changes in precipitation regimes around the world. It has been noted that there is an increased frequency of warm episodes of ENSO, which affect the precipitation regime in the tropics and sub-tropics (Houghton et al., 2001). The relationship between warm phase of ENSO and below average summer monsoon precipitation over the Indian subcontinent has been investigated in several previous research studies (Sikka and Gadgil, 1980; Rasmusson and Carpenter, 1982; Shukla and Paolino, 1983; Ropelwski and Halpert, 1989; Thapliyal et al., 1998). However, India's immediate neighbour, Myanmar (formerly known as Burma) which is an agriculturally important country, occupying a vast landmass in the tropical belt of Southeast Asia has received comparatively scant climatological attention. It has a predominantly agricultural economy, which is heavily dependent on the amount and timing of summer monsoon precipitation. In general, not many studies exist on Myanmar's climatology. Most of the earlier literature that mention about Myanmar's general climatology includes Maung (1945) and Huke (1965), which, however, do not include a very detailed description of the climatology of this region. Recently, Sen Roy and Kaur (2000) conducted a very detailed study about the summer precipitation climatology of Myanmar using 33 years of (1947–1979) station level monthly data. This is perhaps the first such detailed climatological study specific to Myanmar. The results of the study showed the existence of five homogenous precipitation regions, namely, north Myanmar, west Myanmar, central Myanmar, east Myanmar and south Myanmar across the country (Figure 1). Their study also indicated approximately 10% below average precipitation during El Niño years for all the regions. They were somewhat handicapped for want of a longer series of data. However, the availability of high-resolution long-term gridded datasets such as the monthly level climate dataset created by CRU (East Anglia University) at a spatial resolution of 0.5° × 0.5° has made it possible to extend research studies for areas where long-term continuous data records are not so readily available.
In recent years, scientists have discovered the influence of a relatively newly found global teleconnection in the form of Pacific decadal oscillation (PDO) in modulating the impact of ENSO. More specifically, it has been indicated that the warm phase of PDO amplifies the effect of warm phase of ENSO on precipitation patterns. As is well known, PDO is a decadal level teleconnection with clear prolonged warm and cold phases, extending over a few decades. The impact of PDO is more visible in the North Pacific Ocean and North American continent. Studies have also indicated the complementary role of warm phase of PDO during El Niño years, resulting in a stronger ENSO signal over North America (Gershunov and Barnett, 1998). The main explanation proposed for this complementary role between ENSO and PDO is the relative sensitivity of the Aleutian low to the PDO phase and the ENSO phase. Similar results were also reported in the case of winter precipitation over Arizona by Goodrich (2004). The enhancing effect of warm phase of PDO on the ENSO signal has also been shown in the case of India (Krishnan and Sugi, 2003; Sen Roy et al., 2003). In these studies, it was found that warm (cold) PDO is associated with decrease (increase) of precipitation over India. Given the clear modulating effect of PDO on ENSO, the present study focuses on the role of different phases of PDO on the monsoon precipitation of Myanmar occurring during El Niño, La Niña and neutral years.
In order to analyse the influence of PDO and ENSO on the precipitation patterns in Myanmar, we have conducted our study at the regional level for the five homogenous summer monsoon precipitation regions of Myanmar, recently identified by Sen Roy and Kaur (2000). Analysis was also made for precipitation for the country as a whole. As in India, the summer monsoon season in Myanmar extends from June to September when most of the country is under the influence of the monsoon. This makes it possible to compare the results of the present study with those of previous studies conducted for the adjacent Indian subcontinent. The monthly average precipitation data of Myanmar was extracted from a high-resolution gridded dataset (0.5° latitude × 0.5° longitude) made available by the CRU, referred to as the CRU TS 2.0 dataset (available at: http://www.cru.uea.ac.uk/∼timm/grid/CRU_TS_2_1.html). These datasets are an updated version of a pervious gridded dataset created for the time period 1901–1996 (New et al., 2000; Mitchell and Jones, 2005). The precipitation data in this dataset have been quality controlled for gauge biases. The final datasets were created from actual station level data by using angular distance weighted (ADW) method. ADW method of interpolation consists mainly of a distance weighted function in which the closest stations carried a greater weight than stations located farther, with a total of eight stations being taken into consideration while interpolating each grid point (New et al., 2000). There are a total of 289 grid cells covering the entire country. The present study is limited to the time period 1951–2002, mainly due to questionable quality of the dataset before 1950 over this region found after detailed exploratory data analysis. Specifically, in the case of Myanamar, monthly precipitation data were available for 13 stations well spread across the study area, with considerable density of stations around Myanmar. Seven of these stations had data available from before 1951 onwards, whereas for the rest of the six stations monthly precipitation data were available from 1961 onwards. Given the distance weighted spatial interpolation technique used in the creation of the dataset, data from surrounding stations were also used in the creation of the dataset during years with lesser data availability. Similar data quality issues before 1930 were reported in case of India by Sen Roy and Balling (2005). Further detailed description of the dataset, including issues related to missing data maybe found in Braganza et al. (2004).
We used two different variables to characterize the oceanic and atmospheric components of ENSO for our study. There are several regions in the Pacific Ocean that are used for monitoring the strength and persistence of El Niño, which are referred to as niño regions. Chattopadhyay and Bhatla (2002), who studied the relationship between India's summer monsoon precipitation and sea surface temperatures (SSTs) anomalies over different niño regions, found the strongest inverse correlation with niño 3 region. Therefore, in the present study, we have used the widely referred niño 3 index for El Niño, which involves an area bounded by 5°N–5°S and 90°–150°W in the Pacific Ocean. The SST anomalies for the months of June to September (based on the 1950–1979 normal period identified by the Climate Prediction Center) were assembled from 1951 to 2002. The atmospheric circulation component of the ENSO index is represented by Southern oscillation index (SOI), defined as the standardized sea level pressure differences between Tahiti and Darwin. Both these variables (El Niño and SOI indices) were highly correlated, with negative values indicating El Niño years and positive values indicating La Niña years. Furthermore, the data for both SOI and El Niño indices were obtained from the widely used database made available by International Research Institute (IRI) for Climate Prediction, and there were no missing data.
As is well known, PDO is defined as the leading eigenvector of the mean monthly SSTs observed in the Pacific Ocean north of 20°N. It is a decadal level oscillation, with positive values indicating a warm phase, and negative values indicating a cold phase of the PDO. The warm phase of PDO is associated with above normal SSTs along the west coast of North America and below normal SSTs in central and western North Pacific around 45°N. Several studies have indicated the evidence of the warm phase of PDO extending between 1925–1946 and 1977 through the mid-1990s, with the cold phase of PDO extending between 1890–1924 and 1947–1976 (Mantua et al., 1997). There is also a general consensus about the beginning of a cold phase of PDO from 1998 onwards (Hare and Mantua, 2000; Schwing and Moore, 2000). In the present study standardized PDO index values, derived as the leading principal component of monthly SST anomalies in the northern Pacific Ocean, for June to September were assembled for the time period 1951–2002, with no missing values in this series. The PDO index data were acquired from the PDO index dataset developed by Joint Institute for the Study of the Atmosphere and Ocean (JISAO), University of Washington (Mantua et al., 1997; Zhang et al., 1997). The main data sources used by JISAO for calculating this index include the United Kingdom Meteorological Office (UKMO) historical SST dataset (HSSTD) for 1900–1981, Reynold's optimally interpolated SST from 1982 to 2001 and Optimally Interpolated Sea Surface Temperatures (OISST) variability version 2 beginning 2002. The UKMO HSSTD dataset, which is provided by the CRU of East Anglia, are available on 5° latitude × 5° longitude resolution. The OISST dataset, which has a resolution of 1° latitude × 1° longitude, were averaged into 5° × 5° gridded resolution to extend the HSSTD data through 1994–1996 period of record. Further detailed information regarding the creation of PDO index data is available in Mantua et al. (1997).
3. Analysis and results
In the present study, we calculated the area weighted average summer monsoon season precipitation occurring between June and September for the country as whole and five homogenous precipitation regions in Myanmar. The two indices for ENSO, namely, niño 3 SST and SOI, were found to be highly correlated with each other. Therefore, we conducted principal components analysis (PCA) on these two variables, for the period of June to September, to identify the leading vector representing most of the variance in the two variables. The results of PCA showed the first eigen vector explaining more than 93% of the variance among the two variables. The scores for the first eigen vector were used in the present study to represent ENSO index. The indices were taken into consideration for all three summer monsoon months of June to September to detect the overall relationship between the two above-mentioned teleconnection indices and the prevailing precipitation patterns. In order to delineate the role of different phases of ENSO on the summer monsoon precipitation during the cold or warm phase of PDO, we divided our analysis according to the different phases as listed in Table I. In the case of precipitation data, the study area was covered by 285 grid cells, with 61, 35, 48, 66, 75 cells covering the regions of northern Myanmar, western Myanmar, central Myanmar, eastern Myanmar and southern Myanmar regions, respectively. Finally, the dataset for the entire time period was divided into six separate matrices according to the phases of ENSO and the corresponding PDO (warm or cold) episode. The warm and cold phases for PDO were determined according to the widely used classification proposed by Mantua et al. (1997). Each matrix consisted of six columns containing average monsoon precipitation data for the country as a whole and five precipitation regions, and two columns each for the two teleconnection indices taken into consideration, namely, ENSO vector and PDO.
Table I. List of El Niño, La Niña and neutral years according to the cold and warm phases of PDO
In order to determine the role of ENSO and PDO on precipitation, we conducted multiple regression analysis with precipitation as the dependent variable, in addition to the ENSO vector and PDO index as independent variables for different phases of ENSO (La Niña, neutral, El Niño) during the warm and cold episodes of PDO, at the regional level as well as at the national level. During La Niña years, the response of precipitation to ENSO vector was predominantly negative during both warm and cold phases of PDO (Figure 2a). However, the role of ENSO vector was stronger on the regional level precipitation during the warm phase of PDO, except in case of northern Myanmar where the relationship reversed to a weak positive coefficient. The strength of the regression coefficients during La Niña years was strongest in the case of southern Myanmar at − 0.94 (warm phase of PDO), followed by western (−0.87) and central (−0.86) Myanmar regions. Furthermore, the results of the regression analysis were statistically significant at 90% or greater level for all regions and as well as at the national level during the warm phase of PDO, except for northern and western Myanmar. Next, we examined the relationship of regional and national level precipitation with ENSO during El Niño years (Figure 2c). The results of our analysis again showed a predominantly negative role of ENSO on prevailing precipitation patterns across all the regions and at the national level. Interestingly, the negative role of ENSO was stronger during the cold phase of PDO, except in the case of southern Myanmar where there was a slight decline in the strength of the coefficient. The strongest coefficients for El Niño years were observed in the case of northern (−0.614) and eastern (−0.425) Myanmar regions during the cold phase of PDO. Moreover, the regression coefficient during the warm phase of PDO in the case of Northern Myanmar was statistically significant at 98% significance level. Comparatively, the relationship between the ENSO vector and the prevailing precipitation patterns at both regional and national levels were stronger during the La Niña years than El Niño years.
Most studies examining the role of ENSO on regional climatic processes concentrate specifically on La Niña and El Niño years only (Juneng and Tangang, 2005). However, recently it has been observed that neutral phase of ENSO may also play a significant role on the regional level climate. Some of these studies, including Goodrich (2004), indicated a clear modulating effect of PDO on winter precipitation in Arizona, during neutral ENSO years. Therefore, in the present study, we also analysed the role of neutral years on the resulting precipitation in Myanmar (Figure 2b). There was an overall reversal in the role of ENSO on precipitation patterns during the neutral years from the cold phase to the warm phase of PDO for all regions, except southern Myanmar. During the warm phase of PDO, the relationship was negative in the case of all the regions as well as at the national level. On the other hand, during the cold phase of PDO, the relationships were mostly positive for all regions except in case of southern Myanmar. Moreover, in the case of southern Myanmar, the relationship between the ENSO vector and the prevailing precipitation pattern was weaker during the cold phase in comparison with the warm phase, similar to that observed during El Niño and La Niña years. The negative coefficients were observed for western Myanmar region (−0.87) followed by southern (−0.84) and central Myanmar (−0.68). Furthermore, the maximum strengthening in the observed coefficients between the two phases were observed for western Myanmar and at the national level. The positive coefficients observed during the cold phase of PDO were relatively weaker compared with the observed regression coefficients during the warm phase.
We also mapped the standardized regression coefficients for PDO in relation to El Niño, neutral and La Niña years (Figure 3). Standardized regression coefficients are measured on the same scale, which have a mean of zero and a standard deviation of one. It is specifically useful in determining the relative role of different independent variables on the dependant variable, specifically when the units of measurement for the different variables under study have different units of measurement. Therefore, the standardized regression coefficients were used in the present study to enable comparison of the role of the different predictor variables on the resulting precipitation patterns. There was almost complete reversal in the relationship between the regional precipitation and the PDO index during the warm and cold periods of PDO for La Niña, neutral and El Niño years. During La Niña years, the role of PDO index on the prevailing precipitation patterns was predominantly positive at both regional and national levels (Figure 3a). The observed role of PDO index was negative during the cold phase of PDO in eastern Myanmar (−0.106), as well as during the warm phase of PDO in southern Myanmar (−0.035). The strongest positive relationship was observed during the warm phase of PDO over eastern (0.974), central (0.52) and northern (0.389) Myanmar. In the case of eastern Myanmar, the result was also statistically significant at 98% confidence level. However, during the El Niño years, there was a clear reversal in the role of PDO during the cold phase, in the form of a negative relationship, to the warm phase at both regional and national levels, except in the case of eastern Myanmar (Figure 3c). In general, the positive relationship with PDO was stronger during the warm phase of PDO, with the highest observed coefficients in western Myanmar (0.403), which was followed by northern Myanmar (0.318) and central Myanmar (0.27). The negative relationship between PDO and prevailing precipitation patterns during El Niño years was universally observed across all the regions as well as at the national level. It is noteworthy that in general the coefficients were much lower during the cold phase of PDO relative to that observed in the case of warm phase of PDO. Overall, in southern Myanmar the role of PDO index was weakest for both phases of PDO. Finally, the spatial patterns of the role of PDO on region level precipitation during neutral ENSO and warm PDO years were similar to that observed during the El Niño years and warm phase of PDO (Figure 3b). However, the role of PDO was stronger specifically in the eastern (0.549) and southern (0.717) Myanmar regions, as well as at the national level at 0.45 in comparison with that observed during the El Niño years. Furthermore, during the cold phase of PDO, both central and eastern Myanmar experienced positive signals, whereas it was reversed in the case of the remaining three regions.
Additionally, the R2 values, which are mainly used to determine the strength of a particular regression model, were examined to identify the variations in strength of the relationships in the form of variance explained by the selected independent variables for regional level and country level precipitation (Table II). Overall, the highest variance was explained during the years associated with warm PDO period superimposed on La Niña years. For instance, in the case of eastern and southern Myanmar, the R2 values were 0.936 and 0.896. The variance explained at the national level was also the highest among all the years at 0.579. On the other hand, the lowest variance explained were observed during the El Niño years superimposed with warm phase of PDO. In general, southern Myanmar experienced relatively higher levels of explained variance during the entire study period. On the other hand, the lowest levels of R2 values were observed during El Niño years superimposed with cold phase of PDO. At the regional level, the highest average R2 values were observed in the case of southern Myanmar, with northern Myanmar showing the lowest average R2 values.
Table II. R2 values calculated from multiple regression analysis for years associated with different phases of ENSO and PDO
PDO cold/ La Niña
PDO cold/ neutral
PDO cold/ El Niño
PDO warm/ La Niña
PDO warm/ neutral
PDO warm/ El Niño
Overall, it is evident that for the monsoon precipitation of Myanmar, PDO has a modulating effect on the influence of ENSO for the country as a whole. Variance is accounted for to a greater extent for warm PDO–La Niña, warm PDO–Neutral and cold PDO–El Niño years. At the regional level, the signal is less strong. It is of interest to note that during an earlier study (Sen Roy et al., 2003), it was found that regression coefficients showing the relationship between July–August precipitation in India and ENSO were negative for northeast India, which is contiguous with Myanmar. In the case of Myanmar, under similar ENSO conditions, the relationship is negative as with northeast India. Similar results maybe noted with warm PDO and El Niño years. This appears to indicate that the descending leg of the Walker circulation cell perhaps affects Myanmar and the adjacent areas of India similarly during these conditions. Incidentally, it may be noted that the nature of association between Myanmar's precipitation and different combinations of El Niño and PDO undergoes noticeable change in sign and magnitude for different regions of Myanmar. The signal is thus not so clear at the regional level.
Finally, we calculated the absolute differences in average annual precipitation over different regions for El Niño, neutral and La Niña years separately for warm and cold episodes of PDO. Figure 4 shows the absolute difference in precipitation by subtracting the average annual precipitation occurring during warm episodes of PDO from that occurring during cold episodes of PDO for La Niña, neutral and El Niño years. The differences in percentage are represented in parentheses for each region as well as at the national level. As is evident from Figure 4, there was excess precipitation during La Niña, neutral and El Niño years occurring during the cold phase of PDO, with the exception of northern Myanmar. In general, the highest differences for all years were observed in southern Myanmar followed by central Myanmar. In the case of northern Myanmar, there was slightly excessive precipitation in El Niño years coinciding with warm PDO episodes. At the national level, the average annual precipitation for the country as a whole was greater during the cold period of PDO for all phases of ENSO, being 15, 18 and 12.31 mm during La Niña, neutral and El Niño years, respectively. The maximum difference in precipitation was observed during neutral and La Niña years between the different phases of PDO, both at the national and regional scale.
In order to further find the validity of the variations in responses of regional level precipitation, we conducted discriminant analysis on the national level precipitation during the warm and cold episodes of PDO along with different phases of ENSO. The Wilk's lambda, which is a measure of the amount of variance not explained by the differences in the group means, was 0.218 that was statistically significant at 0.00 confidence level. The overall discriminative ability of the two above-mentioned indices was 63.5%. More specifically, the years were identified accurately for neutral years superimposed with the cold phase of PDO. The reclassification ability was also high in case of La Niña years superimposed with the warm phase of PDO (80%) and El Niño years superimposed with the cold phase of PDO. In conformity with the results from the regression analysis, in general the results from the discriminant analysis were more accurate during El Niño years superimposed with the cold phase of PDO (72%), and La Niña years superimposed with the warm phase of PDO (80%).
In the present study, we have examined the combined effect of cold or warm episodes of PDO and different phases of ENSO on the monsoon precipitation over Myanmar for the country as a whole, as well as for five homogenous precipitation regions. The main findings of our study are listed below:
The role of ENSO vector on the regional level precipitation was stronger during La Niña years for both warm and cold phases of PDO.
The overall role of ENSO vector on precipitation patterns in Myanmar was negative with stronger negative coefficients observed during the warm phase of PDO.
The role of PDO, on the other hand, was predominantly positive during its warm phase for the different ENSO years, whereas it was negative during the cold phase of PDO.
The coefficients associated with both PDO and ENSO vectors were overall stronger during the warm phase of PDO.
Finally, the difference in average precipitation during different ENSO years throughout Myanmar showed greater amount of precipitation during the cold phase of PDO compared with the warm phase of PDO. This is in conformity with the findings of a previous study on Myanmar by Sen Roy and Kaur (2000).
At the regional level, southern Myanmar showed the strongest response to the different phases of the teleconnection indices.
In view of the above results, it is evident that there is a clear modulation of the ENSO signal depending on the cold or warm PDO episodes. Therefore, in order to make more effective monsoon forecasts for this region based on ENSO, the warm or cool episodes of PDO should also be taken into consideration.