Spatial variability of ENSO-based monsoon rainfall in the coastal region of southeastern India

The ocean-atmospheric interaction across the tropical Pacific Ocean is termed El Niño Southern Oscillation (ENSO), which influences the behavior of global precipitation. ENSO can be observed in three phases termed El Niño, La Niña, and Neutral. El Niño/La Niña denotes the rise/fall of sea surface temperature (SST) above/below the normal over the equatorial Pacific Ocean (Pandey et al. 2018). While in the Neutral phase, the SST remains nearly equal to normal. This temperature change during El Niño and La Niña creates pressure differences between the western Pacific Ocean and eastern Pacific Ocean (EPO) while changing the wind flow patterns. It is a global phenomenon that occurs every 3–5 years. Among several influential forces, El Niño and La Niña have been considered significant factors affecting rainfall behavior.

The impact of ENSO is worldwide, but the influence on the tropics is severe, especially in monsoon-affected countries such as India, Indonesia, and Australia (Kumar et al. 1999). It is acknowledged that Indian summer monsoon rainfall exhibits a 10% inter-annual variability compared to its long-period average (LPA) (Krishnamurthy & Kinter III 2003). This monsoon rainfall inter-annual variability is associated with droughts and floods impacting the agricultural, water resource, and economic sectors. About 40% of inter-annual variability is under the influence of ENSO events (Sikka 1980; Rasmusson & Carpenter 1983; Shukla & Paolino 1983). Along with El Niño, a strong Indian Ocean Dipole (IOD) also affects this inter-annual variability (Saji et al. 1999; Webster et al. 1999). The impact of any co-occurring ENSO event on the monsoon rainfall is modulated by strong IOD events (Karumuri et al. 2001; Ashok & Saji 2007). Additionally, the Atlantic Ocean is also recognized as a driver of the Indian monsoon rainfall (Kucharski et al. 2008; Pottapinjara et al. 2016; Yadav 2017). Among all of these, the relationship of Indian monsoon rainfall with ENSO demands more of our attention.

In India, El Niño events are often accompanied by low southwest monsoon rainfall (SMR), whereas La Niña events are typically accompanied by robust SMR (Subrahmanyam et al. 2013; Pandey et al. 2018). However, this relationship has weakened recently (Kumar et al. 1999) and has also strengthened at times (Song et al. 2018). Although severe droughts in India have consistently been associated with El Niño, not every El Niño event has led to drought in India (Kumar et al. 2006).

The influence of ENSO on rainfall can vary significantly across different regions of India. For instance, 8 out of 10 El Niño events led to lower-than-normal rainfall in Andhra Pradesh (Rao et al. 2011). Similar rainfall patterns were observed in Himachal Pradesh (Prasad et al. 2014) and Chhattisgarh state (Manikandan et al. 2016). Conversely, Gujarat experiences more rainfall during El Niño than during non-El Niño years (Patel et al. 2014). These regional differences in rainfall emphasize the need to consider local and regional distribution patterns rather than viewing India as a single homogenous entity (Shukla & Mooley 1986; Gregory 1989). Thus, this study is specifically focused on region-specific impacts.

This study focuses on the southeastern coastal region of India due to its high vulnerability to climate variability and extreme weather events. The unique geographic and climatic conditions of the region can lead to distinct rainfall patterns and responses to ENSO events compared to other parts of India. Thus, this research aims to analyze the spatial variability of ENSO events in the southeastern coastal region of India. Additionally, it conducts a comparative analysis of the impacts of ENSO events on the SMR in this region to address the need for region-specific analyses and improve the predictability of monsoon rainfall patterns.

The climate of the coastal region is usually tropical, wet, and warm. Odisha is vulnerable to cyclones, surges, and floods. Andhra Pradesh and Tamil Nadu coasts have been affected by flooding and erosion. People of the coastal regions are engaged in fishing, agriculture, shipping, and shrimp cultivation for their livelihood.

Monthly rainfall data for analyzing monsoon rainfall variability in the coastal districts of Odisha, Andhra Pradesh, and Tamil Nadu from 1970 to 2020 was obtained from an India Water Resources Information System (WRIS) database. India WRIS, a data dissemination platform created by the Ministry of Jal Shakti, collects the daily gridded rainfall with a resolution of 0.25° latitude by 0.25° longitude from the Indian Meteorological Department (IMD) (Pai et al. 2014). IMD collects rainfall data from 6,995 rain gauge stations across India, and the data are interpolated using the inverse distance weighted method to generate the gridded dataset. The data for analyzing the effect of ENSO on SMR is based on the Oceanic Niño Index (ONI), developed by the National Oceanic and Atmospheric Administration (NOAA). The ONI is considered the benchmark for identifying the occurrence of El Niño and La Niña events in the tropical Pacific Ocean and was obtained from the Physical Sciences Laboratory using the HadISST1 dataset (Rayner et al. 2003). According to NOAA, this index is calculated as the running 3-month average of SST anomalies for the Niño 3.4 region (i.e., 5 °N–5 °S, 120°–170 °W) (Pandey et al. 2018). If the index is greater than 0.5, it is classified as an El Niño event, while a value below −0.5 indicates La Niña. Otherwise, it is considered a neutral year. Based on the ONI index, out of 51 years of rainfall data (1970–2020), there are 17 El Niño events, 20 La Niña events, and 14 Neutral events. The ENSO-associated years are shown in Table 2.

The variability of rainfall in the coastal districts is evident through the distinct coefficient of variation observed across each district. For example, the average coefficient of variation in the coastal districts is 0.57 for Tamil Nadu, 0.30 for Andhra Pradesh, and 0.20 for Odisha. Therefore, the analysis was first conducted for individual coastal districts, followed by a composite analysis at the state level. Following the evaluation of the anomalous SMR data, the study aims to identify whether the impact of the ENSO event on the SMR is random or a true effect. To achieve this, a z-test was conducted to assess the statistical significance of the differences between the LPA of SMR and the SMR averages during El Niño/La Niña years. The null hypothesis (H₀) posits no significant difference between the LPA of SMR and the SMR during El Niño/La Niña years, suggesting any observed differences are due to random chance. In contrast, the alternative hypothesis (H₁) asserts that there is a significant difference, indicating that El Niño/La Niña events have a substantial impact on SMR. If the null hypothesis is rejected, the study proceeds by examining long-term trends in the rainfall data. Identifying trends can aid in determining if there are systematic changes in SMR patterns over time that are independent of El Niño/La Niña events, offering deeper insights into SMR behavior.

This study then focuses on validating the established notion that El Niño is associated with below-normal rainfall and La Niña with above-normal rainfall by using a confusion matrix (Silva et al. 2022). A confusion matrix provides a framework for comparing actual monsoon rainfall during ENSO events with expected rainfall patterns. It classifies rainfall years into true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN) based on ENSO events.

The SMRA was used to classify rainfall as above normal or below normal. Positive SMRA values indicate above-normal rainfall, while negative values represent below-normal rainfall. This method captures variability, including minor deviations from the LPA, in a statistically meaningful way. This approach allows us to assess how even small changes can reflect broader climatic influences, such as El Niño or La Niña events. Using SMRA avoids arbitrary thresholds and ensures that the criteria for evaluating TP, FP, TN, and FN during El Niño and La Niña events are based on the inherent variability of rainfall patterns.

In this study, large-scale circulation significantly links seasonal anomalies of ocean basins with the SMR in India. The large-scale circulation that causes significant alteration in rainfall over southeastern India is due to the negative or positive SST across the equatorial Pacific Ocean.

The z-test was implemented to determine the statistical significance of the differences between the climatological average of monsoon rainfall and the rainfall average during El Niño/La Niña years. Initially, the climatological monsoon rainfall from 1970 to 2020 was evaluated, followed by the evaluation of average rainfall during ENSO years. The monsoon rainfall averages of 17 El Niño events and 20 La Niña events were compared with 51 years of rainfall climatology. Statistical significance was analyzed at both 90 and 95% confidence levels.

Table 3 presents the results of the z-test, with columns 4 and 5 indicating the statistical significance at 90 and 95% confidence levels, respectively. A ‘Yes’ in the table signifies that the difference between an ENSO condition and climatology is statistically significant. At both confidence levels, the difference in average monsoon rainfall between the ENSO event and the climatology was statistically significant in Tamil Nadu and Andhra Pradesh. However, this difference was insignificant in Odisha.

The trend analysis at a 5% significance level reveals that SMR in Tamil Nadu and Andhra Pradesh shows an insignificant decreasing trend, as indicated by the MK statistic values in Table 4. Conversely, Andhra Pradesh and Odisha show an insignificant increasing trend.

During La Niña years, the distribution of rainfall shifts toward more rainfall in both Tamil Nadu and Andhra Pradesh, except Odisha. In Tamil Nadu, 15% of the rainfall events lie above the range of 1 sigma, while in Andhra Pradesh, this figure is 20%. Meanwhile, in Odisha, rainfall distribution has shifted toward less rainfall, with 27% of rainfall events observed below the range of 1 sigma.

During El Niño years, most of the coastal districts in Tamil Nadu and Andhra Pradesh experienced negative anomalous rainfall (Figure 8(a)). Tamil Nadu and Andhra Pradesh received an average rainfall anomaly of −0.40. Heading toward Odisha, the effect of El Niño seems to be diminished or weakened with anomalous rainfall in the range of −0.20 to 0.40. Odisha receives an average positive anomalous rainfall of 0.15.

During La Niña years, coastal districts in Tamil Nadu and Andhra Pradesh experienced a positive deviation in the range of 0–25. No coastal district experiences a deviation of rainfall beyond the range of 25. Tamil Nadu and Andhra Pradesh receive more rainfall with an average positive deviation of 12, and 9% and less rainfall in Odisha with a 4% deviation from normal rainfall. During neutral years, the percentage deviation of monsoon rainfall is in the range of −3 to 16.

As per the IMD press release on the long-range forecast for the SMR in 2023, the rainfall is categorized into five ranges: deficit, below normal, normal, above normal, and excess, as given in Table 5. The SMR during El Niño falls into the category of deficit rainfall in Tamil Nadu and Andhra Pradesh, with 83 and 89% of LPA, respectively, whereas Odisha receives normal rainfall with 104% of LPA. During La Niña years, the average SMR in Tamil Nadu lies in the category of excessive rainfall with 111% of LPA, Andhra Pradesh falls into the above-normal rainfall category with 108% of LPA, and Odisha remains in the normal rainfall category with 96% of LPA. In neutral years, all three states receive normal rainfall, with 103, 102, and 101% of LPA in Tamil Nadu, Andhra Pradesh, and Odisha, respectively.

The results of the identified complex matrix are shown in Table 6. The higher number of TP events in Tamil Nadu and Andhra Pradesh indicates the high probability of receiving below-normal rainfall during El Niño. Also, FN events in Tamil Nadu and Andhra Pradesh are few, resulting in a low FNR. This low rate indicates that the probability of experiencing above-normal rainfall conditions during El Niño is low. In contrast, Odisha has fewer TP events than FN events. However, the difference between the number of TP and FN events is not much; hence, the effect of the events is not significant.

Both Tamil Nadu and Andhra Pradesh have 14-13 FP and 20-21 TN, respectively. The higher number of TN suggests that both states recorded expected above-normal rainfall during non-El Niño events. However, there were instances when both states received below-normal rainfall despite the expectation of above-normal rainfall. In contrast, Odisha shows a different scenario. The higher number of FP and lower number of TN indicates the high probability of below-normal rainfall during non-El Niño years that did not align with expectations.

During La Niña events, TP events are more frequent than FN events in Tamil Nadu and Andhra Pradesh, indicating a low FNR and a low probability of below-normal rainfall. Interestingly, the TN events are higher than FPs, indicating the above-normal rainfall events are less during non-La Niña years. In contrast, the condition is reversed with more FP events compared to TN in Odisha. The higher number of FP events in Odisha suggests above-normal rainfall during non-La Niña years, indicating the possibility of the influence of local climatic factors other than ENSO.

The recall rate, which indicates the probability of above-normal rainfall, is highest in Andhra Pradesh during La Niña among all three states. Also, the recall rate for the El Niño event is more than the La Niña event.

Whereas in Odisha, El Niño years led to most of the SMR above the −1 sigma range of normal levels (Figure 10(c)), with some events recording rainfall in the 3-sigma range (Table 7). Also, the El Niño events associated with below-normal rainfall are less than those with above-normal rainfall, especially after 1991.

Similarly, most of the La Niña years were associated with above-normal rainfall in Tamil Nadu and Andhra Pradesh. There are events when rainfall exceeds the +1 sigma of normal. Also, the number of events recorded below-normal rainfall is low in both Tamil Nadu and Andhra Pradesh. None of the rainfall events in Tamil Nadu is below −1 sigma except in 2017, and in Andhra Pradesh, there are only two such events. Hence, the probability of rainfall not lying below −1 sigma of normal is 95% in Tamil Nadu and 90% in Andhra Pradesh. Whereas in Odisha, below-normal rainfall events are more common, although there are events when rainfall is recorded beyond the range of +1 sigma.

The sigma range in Table 7 defines a lower ratio of n/N beyond the range of in the case of El Niño (La Niña), better is the recall of the event. Where n is the number of monsoon rainfall events beyond the range of for El Niño (La Niña). N is the total number of El Niño (La Niña) events from 1970 to 2020. Similarly lower the ratio of n/N beyond the range of in the case of El Niño (La Niña), the more the FNR. Where n is the number of monsoon rainfall events beyond the range of for El Niño (La Niña). The acceptable range of n/N is 0–0.30. The ratio of n/N close to 0 gives a good recall rate and FNR.

During El Niño, the number of events beyond the range of in Tamil Nadu is 0 out of 17 El Niño events. So, the ratio of n/N (0/17) is 0. Therefore, the recall rate for El Niño events in Tamil Nadu is good. The ratio of n/N in Andhra Pradesh is 0.17, and in Odisha, it is 0.40. Hence, the recall rate of El Niño events in Tamil Nadu and Andhra Pradesh is good as the n/N ratio lies in the acceptable range. In the case of La Niña, the recall rate in Tamil Nadu and Andhra Pradesh is adequate, with an acceptable n/N ratio of 0.50 and 0.25, respectively.

Similarly, the FNR of El Niño and La Niña events is significant in Odisha with a ratio of 0.29 and 0.20, respectively. Whereas for Tamil Nadu and Andhra Pradesh, FNR is not significant as the n/N ratio is 0.40 and 0.50, respectively, during El Niño and is 0.35 and 0.50, respectively, during La Niña.

The analysis revealed that El Niño is typically associated with below-normal rainfall in Tamil Nadu and Andhra Pradesh, with a probability of 82% (Table 6), consistent with previous studies (Rao et al. 2011; Gowtham et al. 2019; Jyothi et al. 2023). Similarly, La Niña events tend to result in above-normal rainfall in these states, with probabilities of 65–70% (Table 6), aligning with the findings of previous studies (Subrahmanyam et al. 2013; Pandey et al. 2018). However, the relationship between ENSO events and SMR was found to be insignificant in Odisha (p > 0.10) (Table 3), suggesting a prominent role in regional climatic dynamics. This weak correlation between ENSO and rainfall in Odisha is consistent with previous studies (Singh et al. 2002; Manikandan et al. 2017), which emphasize the influence of local weather systems, including monsoon depressions and cyclones, over ENSO-driven variability.

Contrary to expectations, a few El Niño years, such as 1991, 2006, 2015, and 2018, exhibited above-normal rainfall, particularly in Andhra Pradesh (Figure 6). These anomalies can be attributed to the co-occurrence of positive IOD events, which are known to counteract the typical dry conditions associated with El Niño. Previous studies (Karumuri et al. 2001; Ratna et al. 2021) also highlight the role of the IOD in mitigating the effects of El Niño on Indian monsoon rainfall. For instance, instead of strong El Niño years such as 1991 and 2015, Andhra Pradesh recorded above-normal rainfall due to the co-occurrence of a positive IOD (Bhardwaj & Singh 2021). Similarly, Odisha experienced above-normal rainfall during several El Niño years, including 1977, 1986, 1991, 1994, 1997, 2006, 2015, and 2018, as illustrated in Figure 6. Interestingly, even in the absence of co-occurring positive IOD events, Odisha recorded above-normal rainfall during these El Niño years, highlighting the complex interaction between large-scale oceanic patterns and local weather variations. In contrast, Tamil Nadu largely followed the expected pattern of reduced rainfall during these periods, underscoring the region-specific nature of these climate dynamics.

Similarly, during La Niña, there are few instances of below-normal rainfall (Figure 7), potentially due to shifts in temperature dynamics, changes in atmospheric and oceanic dynamics over the Indian Ocean, as well as reductions in depressions over the Bay of Bengal (Aneesh & Sijikumar 2018).

The study also shows that Odisha experiences significantly higher rainfall compared to Tamil Nadu and Andhra Pradesh (Figure 3), which can be attributed to its exposure to moisture-laden winds from the Bay of Bengal and the region's frequent cyclonic activity. Ramachandran (1967) found that monsoon rainfall tends to increase with latitude in the eastern half of the Indian peninsula, including Odisha. However, this relationship is driven by geographical and climatic factors rather than latitude alone. For instance, Tamil Nadu receives comparatively lower rainfall, particularly during the southwest monsoon, because it lies in the rain shadow of the Western Ghats, limiting rainfall in this state during the critical monsoon period. In contrast, Andhra Pradesh and Odisha, located on the windward side of these mountains, experience more direct exposure to moisture from the Bay of Bengal and are thus more susceptible to heavy rains, especially during cyclonic events.

In contrast to the more predictable ENSO-driven patterns observed in Tamil Nadu and Andhra Pradesh, Odisha presents a different scenario where the influence of ENSO is insignificant. Nageswararao et al. (2019) also noted this insignificant impact of ENSO on Odisha's monsoon, with only a 21% probability of below-normal rainfall during El Niño events (Manikandan et al. 2017). This insignificant result in Odisha is likely due to its proximity to the Bay of Bengal. As per the report of IMD (2002), Odisha is frequently impacted by monsoon depressions, storms, and cyclones. These storms and cyclones move in a westerly direction, crossing the Odisha coast and contributing heavy rainfall to this region, especially during El Niño years. Unlike other states that typically experience reduced rainfall during El Niño, Odisha often benefits from these depressions, particularly during July and August when they are more prominent in El Niño years (Singh et al. 2002).

The monsoon rainfall over Odisha is complex due to its interactions between large-scale atmospheric flows and local-scale orographic features, such as the Eastern Ghats, along with the dominance of the low-pressure systems that develop over the Bay of Bengal (Murty et al. 1986; Dube et al. 2000). The other atmospheric factors contribute to Odisha's rainfall, such as the stronger cyclonic flow at 850 hPa, convergent moisture flux, and the southward shift of the monsoon trough (Swain et al. 2019). As per the report (Jena & Kishore 2021) Odisha's experience with extreme cyclonic conditions during El Niño, such as the Super Cyclone in 1999, and tropical cyclones like Phailin, Hudhud (2014–2016), and Fani (2019), further underscores the complexity of ENSO's influence. Therefore, the variability in Odisha's rainfall patterns (Figure 8) during ENSO events can be attributed to these regional climatic dynamics, which diminish the direct impact of ENSO and should be studied separately to understand its dynamics better.

This study explored the spatial variability of the impact of El Niño and La Niña on the coastal districts of southeastern India, including Tamil Nadu, Andhra Pradesh, and Odisha. The composite analysis revealed that the impact of El Niño is more prominent than La Niña's impact on the eastern coast of southern India. The distribution of rainfall in Tamil Nadu and Andhra Pradesh is observed to be strongly affected by equatorial Pacific Ocean weather. While both Tamil Nadu and Andhra Pradesh are prone to the impact of El Niño and La Niña, La Niña events have a more dominant impact on monsoon rainfall in Andhra Pradesh than Tamil Nadu. Whereas El Niño has a stronger effect on the SMR in Tamil Nadu compared to the other two states.

Results have also shown that strong El Niño does not always mean below-normal rainfall. Also, every El Niño or La Niña event distinctly impacts different states.

The impact of El Niño and La Niña in Odisha was found to be insignificant. Facts reflect Odisha received more than normal levels of rainfall during El Niño and, interestingly, below the normal rainfall levels during La Niña. This may be because of some other factors that have a weighted influence than ENSO on Odisha's SMR.

It is well established and understood that El Niño is usually associated with below-normal rainfall and sometimes brings drought-like conditions in India. Similarly, La Niña has been associated with above-normal rainfall. However, this study indicates that El Niño will not experience excessive rainfall or flood-like conditions, and La Niña will not experience drought-like conditions.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.