1 Introduction

The boreal summer monsoon (June to September; JJAS), characterized by southerly-southwesterly winds and deep moisture content, usually brings a torrent of heavy rains across South Asia, particularly over Bangladesh. Frequent inundations during the summer monsoon are a distinctive feature of the climate of Bangladesh and also a major concern to the stakeholders and policymakers in the country, which is now among the fastest-growing economies (https://www.worldbank.org/en/country/bangladesh) in the world over the past decade. Two recent examples are the intense flooding that occurred during the summer of 2015 when several dozen people died in floods and landslides triggered by heavy rains, and the deadly floods and landslides in the Cox Bazar District during the summer of 2021 that damaged the shelters for nearly one million Rohingya refugees (the world's largest refugee settlement) (https://floodlist.com/asia). A slight variation in the summer monsoon rainfall can have colossal impacts on infrastructure, agriculture production, water supply management, hydropower generation, fisheries, industry, health, daily human lives, and livelihoods of several million people living in this low-lying coastal country. Given the exposure of almost the entire population (Coirolom et al. 2013) of Bangladesh to summer monsoon rainfall, identifying factors that modulate it, e.g., the El Niño-Southern Oscillation (ENSO; the largest mode of variability at interannual time-scales) and using that information to improve forecasts of Bangladesh monsoon variability on seasonal timescales is an issue of considerable urgency.

Many South Asian countries receive a substantial fraction of their annual rainfall during a span of 4 months from June to September (Webster and Yang 1992; Ehsan et al. 2021; Attada et al. 2022). As a first approximation, the boreal summer monsoon over the Indian landmass is often considered a proxy for the entire South Asian monsoon. A plethora of studies have sought a relationship between summer monsoon rainfall over South Asia and the remote and local sea-surface temperature (SST) anomalies in different oceanic basins (Rasmusson and Carpenter 1983; Shukla 1987; Annamalai et al. 2007; Kucharski et al. 2009; Roxy et al. 2015; Venugopal et al. 2018; Kirtman and Shukla 2000; Sharma et al. 2021). For instance, on the interannual timescale, positive SST anomalies in the central-eastern equatorial Pacific related to ENSO bring below-normal rainfall over India. In contrast, below-average SST anomalies associated with La Niña conditions are usually accompanied by above-normal precipitation amounts (Ju and Slingo 1995; Kripalani and Kulkarni 1999; Goswami and Xavier 2005). However, the strength of the relationship between summer monsoon rainfall in different parts of South Asia and ENSO undergoes alternate increases and decreases on decadal time scales (Kumar et al. 1999, 2006). In addition to ENSO, SST anomalies in the tropical Indian (Webster et al. 1999; Cherchi et al. 2007) and Atlantic (Borah et al. 2020; Yadav et al. 2018) Oceans, as well as other factors (Boos and Kuang 2010; Turner et al. 2012) also have a substantial impact on the summer monsoon rainfall over South Asia.

While considerable attention has been devoted to exploring the interactions between the summer monsoon rainfall across South Asia, particularly over India (the largest country in terms of area, population, and economy in South Asia) and SSTs as described above, little attention has been given to understanding the relationships between SSTs and summer monsoon rainfall over Bangladesh (Singh et al. 2000; Wahiduzzaman and Luo 2021), and its impacts (Banu et al. 2015; Cash et al. 2010). To this end, we address the need for a more rigorous investigation of the ENSO and BSMR relationship. It has been demonstrated that there is a weak positive correlation between BSMR and SSTs in the tropical Pacific, while there is an inverse relationship between BSMR and ISMR based on observations. The initialized seasonal forecast models are incapable of replicating these observed connections.

2 Data and Methodology

2.1 Datasets (Observations and Models)

The observed sea surface temperature (SSTs) is from NOAA Extended Reconstructed Sea Surface Temperature (ERSST), version 5 dataset (Huang et al. 2017), which is also used to define the warm and cold episodes of historical ENSO events during 1981 to 2021. The precipitation data is from Bangladesh ENACTS (ENACTS-BMD: the Enhancing National Climate Services for Bangladesh Meteorological Department: BMD) datasets for 1981–2021. The ENACTS initiative led by the International Research Institute for Climate and Society (IRI), Columbia University, in collaboration with National Meteorological Services has been targeted for the provision of high-quality climate data in different countries. An open-source, R-based software, developed and maintained (for data organization, quality control, combining station data with other products) at IRI namely Climate Data Tool (CDT) is used to construct ENACTS in different countries (Dinku et al. 2022). EANCTS-BMD is a product of such cooperation between IRI and BMD. This gridded high-resolution high-quality dataset (0.05° × 0.05°) covers the whole of Bangladesh and was constructed by merging station and satellite products (Acharya et al. 2020). In order to investigate the relationship between BSMR and summer monsoon over India we also utilized the rainfall data, created and managed by the Indian Meteorological Department (Pai et al. 2014). The data were directly downloaded from their website (https://www.imdpune.gov.in/). We also utilize other precipitation products, including Climatologies at High resolution for the Earth's Land Surface Areas (CHELSA) 1981 to 2018 (Karger et al. 2017), and Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) 1981 to 2021 (Funk et al. 2015). Observed lower-level (850 hPa) horizontal (U850), meridional (V850) winds, and vertically integrated moisture divergence (VimD) used in this research is obtained from the fifth generation of ECMWF atmospheric reanalysis dataset (Hersbach et al. 2020).

Seasonal forecast data from two models ‘CanSIPS-IC3 (Lin et al. 2020) (20-members: 1981 to 2020: 40-years)’, and ‘European Center for Medium-Range Weather Forecast (ECMWF) seasonal forecast system 5 SEAS5 (Johnson et al. 2019) (25 members: 1981 to 2021: 41-years)’ is utilized in this study. SEAS5 have different ensemble sizes in hindcasts and forecast (25 members in hindcast and 51 in forecasts). However, we use an ensemble size common to the hindcast period. Recently, there have also been studies (Ehsan et al. 2020a, 2020b) on the prediction of boreal summer monsoon rainfall using these seasonal forecasts models. The ensemble mean data for SST, precipitation, and winds (zonal and meridional components) at 850 hPa at Lead-1 month forecast from these models are obtained from the Data Library of the IRI at the following link: (https://iridl.ldeo.columbia.edu/). For details about individual model configurations and forecasting performance, readers are encouraged to consult the reference papers cited above.

2.2 Methods

The climatological based (WMO 2017) period considered in this study is from 1991 to 2020. The anomalies for observation and models (Ensemble mean considered for each start and target (always Lead-1 month) month and season) are calculated by subtracting the 1991 to 2020 climatology. Unless stated otherwise, all the precipitation time series (observed and model) represent the spatially averaged values over the Bangladesh country mask, while the Indian Summer Monsoon Rainfall is spatially averaged over 75°–85oE, 18°–30oN. Tests employed for statistical significance are t-test for correlation and composites (Wilks 2006) at alpha = 0.05.

2.3 Identification of ENSO Events

Warm and cold ENSO events during June to September 1981–2021 are identified based on a JJAS exceedance of SST anomalies in the Niño3.4 region (5°S-5°N, 190°E–240°E) relative to a 1991–2020 climatology. Warm/El Niño years are defined when the Niño3.4 index exceeds + 0.5 °C. Similarly, cold/ La Niña years are defined when the Niño3.4 index must be − 0.5 °C or less. We selected 8 warm and 11 cold events based on this criterion, while the rest are considered ENSO-neutral years. A summary of warm and cold ENSO events for JJAS during 1981–2021 is presented in Table 2.

3 Results

3.1 Annual Cycle of Bangladesh Rainfall in ENACTS-BMD

The monthly climatology of rainfall over Bangladesh are presented using newly constructed high-resolution ENACTS-BMD data (See Data Section) during 1991–2020 (30 years period). Bangladesh is an extremely wet country in South Asia, receiving a substantial amount of rainfall (~ 2200 mm of rainfall per year or even more) across the country in a given year. However, the rainfall magnitude and distribution across time and space (East–West and North–South rainfall gradient) is extremely heterogeneous in Bangladesh. First, intense rainfall usually occurs during the boreal summer months. Second, the northern and southern territories in the eastern side of the country receive even more rainfall as compared to western and other regions, before (March to May), during (June to September) and after (October and November) the monsoon. December to February is considered the dry season (rainfall is less than 1 mm/day) across the country (see supplementary Fig. S1).

We also analyze the annual cycle of the monthly precipitation fraction of the annual total across the whole country (Fig. 1). The monthly precipitation data for Bangladesh have been extracted from the ENACTS-BMD data set. The annual cycle of Bangladesh rainfall is dominated by the intense summer monsoon rains that happen from June to September. However, as noted in the monthly rainfall climatology (Supplementary Fig. S1), pre- and post-monsoon months also provide nearly 20–25% to the total annual rainfall. In contrast, contribution during the winter months is almost negligible. The rainfall in Bangladesh peaks during boreal summer and contributes about 70–80% of the annual rainfall during JJAS. Therefore, most of the intense flooding in Bangladesh is typically related to active and consistent rainfall during boreal summer, which causes damages to lives and property during these times of the year.

Fig. 1
figure 1

Observed monthly rainfall over Bangladesh. Monthly precipitation as a percentage of annual total over Bangladesh landmass (country average) based on the 1991–2020 period, using ENACTS-BMD (see text) gridded data from Bangladesh Meteorological Department. The red dot indicates the average monthly precipitation fraction

Full size image

3.2 BSMR-ENSO Relationship in Observations

In the present study, the 4-month wet season average (June to September) of the Bangladesh country averaged precipitation is used as a reference time series. This time series is normalized (divided by the standard deviation for 1991–2020) and is denoted as the Bangladesh Summer Monsoon Rainfall (BSMR) Index shown in Fig. 2a. We examined the relationship between the BSMR time series and global sea-surface temperature anomalies during JJAS. Figure 2b shows the simultaneous correlation of BSMR with the JJAS SST anomalies across the globe. A weak positive correlation coefficient is found in the central-eastern equatorial Pacific region. To explain further, the Niño-3.4 index, which is the SST averaged for the domain of 5°S–5°N, 190°E–240°E, is constructed. Table 1 illustrates the correlation and 95% confidence interval between the two-time series from 1981–2021, keying Niño-3.4 index for individual months (January to December), and concurrent to BSMR. The association of BSMR with ENSO in the starting month is near-zero and increases gradually as the ENSO state approaches the beginning of the summer monsoon season. Weak positive correlations are found between BSMR and Niño-3.4 SST occurring nearly simultaneously during summer months (June = 0.14, July = 0.21, August = 0.22, September = 0.18, JJAS = 0.21). Furthermore, the correlation with BSMR is virtually unchanged if Niño-3.4 is replaced by the Niño-3 (r = 0.23), moderately positive in case of Niño-1 + 2 (r = 0.27) and very weak in case of Niño-4 (r = 0.07), which shows that the far eastern and central-eastern Pacific SSTs are more impactful as compared to western Pacific SSTs. In further analysis, however, we considered Niño-3.4 index for representing the ENSO state. To investigate this issue further concerning data sensitivity, we employed an ensemble of different observation datasets (see data for details), and prepared a similar correlation table (see supplementary table S1) between BSMR from other precipitation datasets and the Niño-3.4 index. For the purposes of this analysis, datasets were required to span at least 35 years and also of high-quality, high spatial resolution product like ENACTS-BMD. Based on the analysis conducted, it can be inferred that the positive association between BSMR-ENSO is not limited to the ENACTS-BMD data alone, but is also evident in various other datasets of similar nature, as indicated by the positive correlations observed across them. Figure 2c, shows the regression of the 850 hPa winds anomalies against the BSMR. From the analysis shown in Fig. 2c, it is evident that there are anomalous westerlies present along the foothills of the Himalayas, and an anomalous cyclonic circulation at the lower level over Bangladesh.

Fig. 2
figure 2

Observed BSMR variability, and teleconnections. a The time series of the spatially averaged JJAS (BSMR) rainfall over Bangladesh. b The simultaneous correlation between BSMR and global SSTs anomalies. c The 850-mb winds from ERA5 regressed on BSMR (ENACTS). The anomalous westerlies along the foot hill of Himalayas (northwest Indian region and Nepal) are evident. In (b), and (c) the black dotted points/arrows show correlation/regression coefficients that is statistically significant at the 5% level. The analysis period is from 1981–2021

Full size image
Table 1 Correlation and 95% confidence interval (CI), based on 1981–2021 (41-years), between Bangladesh Summer Monsoon Rainfall (BSMR) and Niño-3.4 SST index during monthly (Jan-Dec), and four-month (JJAS: Jun to Sep) periods prior to and simultaneous with the BSMR
Full size table

Bangladesh's JJAS rainfall differs from one year to another year. Figure 3 shows countrywide BSMR relative to the recent long-term mean (1991–2020). The El Niño and La Niña years, denoted by red and blue stars respectively (refer to method section), have been listed in Table 2. Over the period 1981–2021, Bangladesh's summer monsoon precipitation is consistent from one El Niño event to another: above-average JJAS precipitation was observed during 7 of the 8 El Niño events, including the top two extreme summer monsoon seasons on record, during the strong El Niños of 1987, and 2015. However, slightly below-average BSMR happened during the weak warm event of 2009. Considering the cold ENSO episodes (Fig. 3: blue stars), above-average JJAS Bangladesh precipitation was observed during 7 of the 11 La Niña events. Furthermore, these results are in accordance with prior research indicating an increase in summer monsoon rainfall over Bangladesh during the warm phase of ENSO (e.g., Chowdhury 2003; Wahiduzzaman and Luo 2021). This also demonstrates that the association between BSMR and ENSO is asymmetric.

Fig. 3
figure 3

BSMR, relationship with ENSO. Observed BSMR relative to the long-term mean (1991–2020). Red and blue asterisk denotes warm and cold ENSO events. El Niño years tend to be wetter than average (7 out of 8), La Niña showed mixed signal. Two El Niño years that are nearly 3 decades apart (1987 and 2015) produced extreme BSMR (~ 2 × \(\sigma\): standard deviation). ENACTS-BMD precipitation data is used to estimate BSMR. See text for details about ENACTS-BMD data

Full size image
Table 2 List of El Niño/warm (08 Events) and La Niña/cold (11 events) ENSO events happening in JJAS during 1981 to 2021. The JJAS Niño 3.4 index anomaly is shown in parentheses. The ERSSTv5 data and 1991 to 2020 normal period is used to compute anomalies
Full size table

Although these findings suggest a weak and statistically insignificant positive association between BSMR and ENSO, further investigation is necessary to determine whether this association influences the representation of seasonal monsoon circulation. Table 2 presents the Niño-3.4 index value in JJAS for the warm and cold ENSO years during the period. A total of eight (8) warm and eleven (11) cold events are outlined in Table 2, along with their SST anomaly values. Composite anomalies for sea surface temperature, 850 hPa winds, precipitation, and vertically integrated moisture divergence are shown in Fig. 4. White stippling indicates the statistically significant (5% level) values. A prominent feature of warm (cold) ENSO composites is positive (negative) SST anomalies pattern in the central-equatorial Pacific. It is evident from Fig. 4a and 4b, that the magnitude of SST anomalies is higher for warm composites than for cold composites.

Fig. 4
figure 4

ENSO Composites and Regional Climate. a Observed (a, b) SST (shaded in ºC), c, d 850 hPa winds (UV850: vectors in m/s) and precipitation (in inset: shaded in mm), and vertically integrated moisture divergence (VimD: shaded in kg/m2). The stippling white dots show values that are statistically significant at the 5% level. The country of Bangladesh is surrounded by a black square box

Full size image

As a component of our analysis of the impact of warm and cold SST composites across the South Asian region as a whole, and particularly over Bangladesh, we investigated the lower-level circulation. It is probable that discrepancies in the monsoonal circulation can result in dissimilarities in precipitation patterns during the South Asian summer monsoon. The warm ENSO composites have resulted in a weakening of the monsoon circulation (as illustrated by vectors in Fig. 4c) and a decrease in moisture content (as indicated by shading in Fig. 4e) over the Indian landmass. Conversely, the cold ENSO composites indicate a strengthening of the lower-level summer monsoon circulation (illustrated in Fig. 4d by vectors) and an increase in moisture content (as displayed in Fig. 4f by shading) over India. However, it is worth noting that in Bangladesh, there is a noticeable upsurge in moisture as well as precipitation during warm ENSO composites (as shown in the inset of Fig. 4c), which is significantly greater than during cold ENSO composites (as displayed in the inset of Fig. 4d). These findings are consistent with previous studies that have demonstrated a weakening of the walker circulation during the summer monsoon, associated with El Niño, which typically results in a decrease in summer monsoon rainfall over the Indian landmass (e.g., Kumar et al. 2006). Moreover, Cash et al. (2015) conducted an investigation on the regional characteristics of the South Asian summer monsoon using various reanalysis products and model simulations. Their study revealed that the intensified summer monsoon rainfall observed in Bangladesh and northeast India is associated with anomalous anticyclonic circulation at 850 mb and westerly anomalies along the foothills of the Tibetan Plateau. Conversely, other regions displayed cyclonic circulation and easterly anomalies linked to enhanced rainfall. It is important to note that the choice of reanalysis products influenced their findings. Our findings are partly consistent with the conclusions of Cash et al. (2015) as we also detected anomalous westerly flow along the foothills of the Himalayas. Nevertheless, this atypical circulation induced a cyclonic circulation pattern instead of the anticyclonic circulation over Bangladesh and the surrounding regions. It should be emphasized that a lower-level anticyclone typically suppresses rainfall by inhibiting cloud formation and restricting the transport of moisture, as demonstrated in the composite maps of vertically integrated moisture divergence (Fig. 4e and 4f). The anomalous vertically integrated moisture divergence associated with the warm (Fig. 4e) ENSO composite pattern is substantially different from those associated with the cold (Fig. 4f) ENSO composite pattern. A significant negative VimD value can be observed during warm composites, which indicates a large moisture convergence that could intensify precipitation over Bangladesh (also in Myanmar and northeast India) and result in flooding (Fig. 4e). As a result of the above analysis, it can be inferred that rainfall over Bangladesh during the summer monsoon is determined by large-scale shifts in the center and intensity of the seasonal mean monsoon circulation.

The potential physical mechanism can be outlined as follows:

  1. 1)

    During a warm ENSO event (El Niño), the walker circulation tends to be weakened.

  2. 2)

    Resulting in a corresponding weakening of the mean summer monsoon circulation.

  3. 3)

    This weakening is evident in the diminished intensity of easterly winds along the Himalayan foothills (anomalous westerly anomalies are observed along the foothills of the Himalayas).

  4. 4)

    A lower-level anomalous cyclonic circulation is established over Bangladesh and adjoining areas, facilitating the convergence of moisture within the boundary layer, which in turn, leads to intensified rainfall over Bangladesh and the surrounding regions during warm ENSO (El Niño).

  5. 5)

    However, the relationship between BSMR and ENSO is asymmetric.

3.3 ISMR-ENSO and BSMR-ISMR Relationship

Previous analyses have indicated that warm ENSO composites lead to a rise in precipitation over Bangladesh, but the Indian landmass experiences below-normal rainfall, pointing towards an inverse correlation between the summer monsoonal rainfall of Bangladesh and India. To further emphasize this point, a joint plot depicting the relationship between BSMR and ISMR in the observational data is utilized. Figure 5 shows the scatter plot and individual distributions of the BSMR and ISMR along with the curve (regression line) that fits the data. BSMR is negatively correlated (r = − 0.47 statistically significant at the 5% level) with the ISMR during the 41-year period. Consequently, there appears to be a significant heterogeneity within the adjacent region, with varying responses to events in the Pacific related to ENSO. Furthermore, these results are consistent with earlier studies showing that rainfall recorded at individual station sites in Bangladesh and rainfall in northeastern states of India were not correlated with monsoon rainfall in the main Indian landmass (Kripalani et al. 1996; Kumar et al. 2006). For instance, Kumar et al. (2006) investigated the changes in the Indian monsoon rainfall response to El Niño events during JJAS. They found that warm ENSO events are linked with drought conditions in India, whereas Bangladesh appears to experience an excessive amount of rainfall during this period (refer to Fig. 2b of Kumar et al. 2006).

Fig. 5
figure 5

BSMR, ISMR relationship. Joint plot (combination of scatter plot and histograms) between BSMR and ISMR for the period 1981–2021. The shaded area indicates the 95% confidence interval (estimated using a bootstrap of 10,000 samples)

Full size image

One may also be interested in recalling the correlation between Indian summer monsoon rainfall (spatially average over 75°–85°E, 18°–30°N) and ENSO as well as circulation. The correlation map constructed between the normalized time series of ISMR and global SST anomalies reconfirms their general out-of-phase relationship (Supplementary Fig. S2a). The correlation coefficient between ISMR and Niño-3.4 is negative and statistically significant, suggesting that a good portion (~ 30%) of the total interannual variance of ISMR is associated with ENSO during boreal summer. Supplementary Fig. S2b shows the regression of the 850 hPa winds against the ISMR. The enhanced ISMR is characterized by the anomalous easterlies, cyclonic circulation over western Indian region, which leads to moisture convergence over the region.

3.4 Teleconnections in Seasonal Forecasting Models

Till now, our findings show the following;

  1. A)

    BSMR and ENSO (Niño-3.4 Index) are weakly positively correlated.

  2. B)

    BSMR and ISMR are negatively correlated.

To investigate these in seasonal forecasting models, correlation coefficients between BSMR and global sea surface temperatures have been computed in two models CanSIPS-IC3 (Fig. 6a) and SEAS5 (Fig. 6c) of ECMWF (see data section for details). As opposed to the observed association (Fig. 2b), the BSMR and SSTs in the Pacific region in two models indicates a strong negative relationship, which suggests that models cannot reproduce the observed teleconnections (even the sign is reversed). Additionally, the 850 hPa circulation plots in two models (Fig. 6b and d) show the anomalous circulation pattern different from that of the observed. The models inaccurately depict the relationship between BSMR and circulation at the 850 hPa level.

Fig. 6
figure 6

BSMR variability, and teleconnections in seasonal forecast models (CanSIPS-IC3 and ECMWF-SEAS5). a and c The simultaneous correlation between BSMR and global SSTs anomalies. In (a and c), the black dotted points show statistically significant correlation coefficient at the 5% level. (b and d) The 850-mb winds regressed on BSMR

Full size image

This inability warrants exploring the correlation between ISMR and global sea surface temperatures anomalies in models. We prepared similar correlation maps by replacing BSMR with ISMR, and results are depicted in supplementary Fig. 3S. In contrast to BSMR-ENSO, the ISMR-ENSO correlation maps are of correct sign in models, although stronger than that of the observed. This stronger association result from utilizing the ensemble-mean data instead of ensemble members (Ehsan 2020; Ehsan et al. 2021). Nevertheless, both models were able to capture the correct sign and structure of the correlation between ISMR and SSTs in the Pacific, and between ISMR and 850 hPa circulation respectively. Finally, we check the observed inverse association between BSMR and ISMR in models. Supplementary Fig. 4S, shows the scatter plot between BSMR and ISMR from models. CanSIPS-IC3 model shows moderate positive association (Supplementary Fig. 4Sa) between BSMR and ISMR as opposed to observation, however, ECMWF-SEAS5 shows almost no association (Supplementary Fig. 4Sb). However, the models fail to replicate the observed inverse correlation between BSMR and ISMR.

4 Discussion and Summary

The summer monsoon over Bangladesh during 2015 resulted in a net countrywide precipitation excess nearly equivalent to the \(2 \times \sigma\) (Standard Deviations), with some parts of the country experiencing even higher values, while India experiences a strong rainfall deficit. Understanding how ENSO impacts the boreal summer monsoon rainfall over Bangladesh, including its variability and modulating factors, has important implications for seasonal predictions. The warm ENSO events have a profound impact on the climate of Bangladesh. In particular, very strong events (e.g., the warm events of 1987, 1997, and 2015) force a significant shift in the precipitation distribution to wet conditions over Bangladesh, as opposed to the precipitation over the neighboring country, India, that experienced deficit in their countrywide summer monsoon rainfall. Our findings could have several practical considerations and suggestions.

  • To analyze the regional/local precipitation variability over South Asia, particularly boreal summer monsoon rains that cause significant flooding and catastrophe across the region, high-resolution and high-quality climate data like ENACTS-BMD is encouraged.

  • The relationship between Bangladesh summer monsoon rainfall and ENSO is asymmetric.

  • Seasonal forecasting models struggle in showing the observed BSMR-ENSO and circulation relationship, as well as local precipitation heterogeneity. Yet, useful summer rainfall predictions can be potentially achievable by using different statistical bias correction methods (Acharya et al. 2021), which could potentially provide useful predictions of summer monsoonal rains. This is the topic of our future research.

Limitations of the present study include the relatively small number of observed cases of warm ENSO. The small sample size (41-years: 1981 to 2021), particularly for observed warm events (8 events), effectively prevents the determination of the robust statistical distribution of BSMR based on warm episodes alone, as the BSMR-ENSO relationship is relatively weak. Data from initialized coupled model ensemble simulations can increase the sample size to obtain more robust precipitation distributions. Still, the outcomes are also limited to the extent that they cannot provide realistic responses to SST forcing. In particular, the BSMR-ENSO, BSMR-circulation, and inverse BSMR-ISMR connection is different in models as compared to the observed relationship and pattern. One possible explanation for the models' inability to accurately depict the signal over Bangladesh is the intricate physiography of the country. Bangladesh's unique climate setting, which differs from its neighboring countries such as India and Myanmar, is due to its complex topography. The small area and distinct location of Bangladesh pose a significant challenge for its accurate representation (e.g., Kumar et al. 2006) in seasonal prediction models. The experiments conducted in Kumar et al. 2006, as shown in their Fig. 4b, depict a reduction in summer monsoon rainfall over the Indian landmass, but fail to reproduce the observed rise in rainfall over Bangladesh. The issue of topography is further exacerbated in the southeastern regions of Bangladesh, where the sharp elevation cannot be resolved by current versions of Global Climate Models (GCMs) that have 1-degree resolution (Fahad et al. 2021). Additionally, apart from these topographic issues, the physics of the models could also add an extra layer of complexity and further contribute to making accurate forecasts of summer monsoon rainfall over Bangladesh even more challenging.