A Comparative Study on 2015 and 2023 Chennai Flooding: A Multifactorial Perspective

Abstract

Floods are highly destructive natural disasters. Climate change and urbanization greatly impact their severity and frequency. Understanding flood causes in urban areas is essential due to significant economic and social impacts. Hydrological data and satellite imagery are critical for assessing and managing flood effects. This study uses satellite images, climate anomalies, reservoir data, and cyclonic activity to examine the 2015 and 2023 floods in Chennai, Kanchipuram, and Thiruvallur districts, Tamil Nadu. Synthetic-aperture radar (SAR) satellite data were used to delineate flood extents, and this information was integrated with reservoir data to understand the hydrological dynamics of floods. The classification and regression tree (CART) model delineates flood zones in Chennai, Kanchipuram, and Thiruvallur during the flood years. The study region is highly susceptible to climatic events such as monsoons and cyclones, leading to recurrent flooding. The region’s reservoirs discharged floodwaters exceeding 35,000 cubic meters per second in 2015 and 15,000 cubic meters per second in 2023. Further, the study examines the roles of the Indian Ocean Dipole (IOD), which reached its peak values of 0.33 and 3.96 (positive IOD), and El Niño in causing floods here. The complex network of waterways and large reservoirs poses challenges for flood management. This research offers valuable insights for improving the region’s flood preparedness, response strategies, and overall disaster management.

1. Introduction

In recent years, the frequency and intensity of extreme weather events, particularly floods, have posed significant challenges to vulnerable regions across the globe. A flood, characterized as a natural calamity, involves the submergence of a location for a specific period, resulting in the destruction of property and loss of life [1,2]. Inland flooding occurs when significant precipitation over a short period of time causes flash floods, or when several heavy rainfall events increase flooding in inland locations. Coastal flooding, on the other hand, is caused by strong winds and high tides, which raise and fluctuate the mean sea level [3,4]. United Nations Office for Disaster Risk Reduction (UNDRR) [5] reported that in the two-decade span from 2000 to 2019, Emergency Event Database (EM-DAT) documented 7348 disaster events, resulting in the loss of around 1.2 million lives and impacting over 4.03 billion individuals. The yearly average consisted of 367 disaster events, with floods and storms comprising the majority at 44% and 28%, respectively.

Precise delineation of flooded areas is a crucial element in flood management. Nevertheless, the traditional method of forecasting flooded regions is time-consuming and fails to offer prompt information during disasters [6]. Additionally, conventional land observation techniques are both costly and time-intensive [7]. Satellite imagery plays a pivotal role in flood mapping, offering a bird’s-eye view of large geographic areas. Mapping floods helps forecast inundated areas with higher precision and offers the potential for accessing near real-time data [8]. Moreover, the fusion of remote sensing data with geographic information systems (GIS) has advanced significantly in accurately mapping inundated areas [9].

Flood mapping using remotely sensed data is an essential effort for planning flood mitigation measures and assessing flood impacts [10]. The high spatial resolution and frequent revisit times of modern satellites enable efficient monitoring of changes in land cover, water bodies, and flood extents. Different remote sensing methodologies, such as multispectral and synthetic aperture radar (SAR) imagery, can be used for flood mapping [11]. Flood analysis using satellite images has been explored using machine learning (ML) and deep learning (DL) methodologies. ML and DL models such as Random Forest (RF), Support Vector Machine (SVM), U-Net, DeepLabV3, and Mask RCNN have been employed for flood detection and mapping. These models have been implemented using various tools and platforms such as SNAP, ESRI ArcGIS Pro, and Python programming language [12]. Machine learning models based on satellite imagery have shown potential in detecting and forecasting disasters, aiding in disaster management and post-disaster activities [13]. The above methods can now be processed and analyzed using the Google Earth Engine (GEE) code editor platform without extensive computational resources or data downloads [14].

Integrating the satellite flood map data with reservoir information provides a holistic understanding of the hydrological dynamics during flood events. Reservoir details, including inflows, outflows, and storage capacities, are crucial indicators of the water management infrastructure’s response to extreme precipitation. Dams are crucial for water resource management, serving various purposes such as supplying water, generating hydroelectricity, managing floods, supporting ecology, and facilitating recreation [15]. In recent years, numerous dam failures worldwide, including incidents in Oroville, USA (2017) [16,17,18,19], have had significant and devastating impacts on downstream areas.

Furthermore, multiple studies have been conducted to understand the influence of climatic phenomena such as the Indian Ocean Dipole (IOD) and El Niño on weather patterns, precipitation levels, and the occurrence of cyclonic activities [20,21,22,23]. Understanding these climatic anomalies is essential for anticipating and mitigating the impacts of floods. Cyclones, originating from warm ocean waters, constitute a crucial element in causing floods. Their interaction with coastal areas can lead to storm surges, intense rainfall, and subsequent inundation. Past research has concentrated on examining the connections between Indian Summer Monsoon Rainfall (ISMR), El Niño Southern Oscillation (ENSO), and IOD, offering substantial evidence on how sea surface temperature (SST) conditions and circulation patterns across the Indo-Pacific region influence the variability of ISMR [20,24]. This is measured using the IOD, or Dipole index, which is derived by subtracting the average sea surface temperatures (SST) of the Western Indian Ocean from those of the Eastern Indian Ocean. This difference is used to determine the Dipole index values. This study aims to assess the effects of ENSO and IOD on the Northeast monsoon, a major cause of flooding in the study area.

Existing literature often analyses flood causes independently, including satellite images, reservoir dynamics, climatic anomalies (IOD/El Niño), and cyclonic activities. This study aims to examine the interconnections between these factors to offer a thorough comprehension of flood events. It specifically examines the 2015 and 2023 floods, identifying the interrelated factors that contributed to flooding in the study area. The objectives are to (a) examine how urban and environmental elements are interdependent, (b) compare the floods of 2015 and 2023, and (c) assess the necessity of integrating different datasets to improve flood management and climate resilience in urban environments. Integrating these diverse datasets may promise a more accurate assessment of the factors influencing flooding, contributing valuable insights to flood management strategies and climate resilience planning in the selected study area. The findings are expected to inform future flood preparedness strategies, taking into account the diverse climatic conditions and urban features that contribute to the vulnerability of this dynamic metropolitan area.

2. Study Area

The study area encompasses the geographically contiguous regions of Chennai, Kancheepuram, and Thiruvallur districts in Tamil Nadu, India, as shown in Figure 1. Chennai, the capital city of Tamil Nadu, serves as the focal point of this study, while the Kancheepuram and Thiruvallur districts flank it to the southwest and northwest, respectively. Geographically, the study area is positioned along the eastern coast of India, overlooking the Bay of Bengal. Chennai, a bustling metropolis, is renowned for its urban and industrial landscape, serving as the region’s major economic and cultural hub. Kancheepuram district, situated southwest of Chennai, is characterized by a mix of urban and rural areas and is known for its rich cultural heritage and historic temples. Thiruvallur district, located to the northwest, features a diverse landscape that includes urban centers, agricultural regions, and water bodies. The study area covers an area of 7949.6 sq. km.

In recent years, Chennai and its neighboring districts of Kancheepuram and Thiruvallur in southern India have witnessed significant flooding, illustrating the intricate interconnections between climatic conditions, hydrological dynamics, and urban resilience. The floods of 2015 and 2023 have notably emphasized the region’s susceptibility to extreme weather events. In 2015, it was reported that 347 lives were lost due to rain-related incidents, and economic losses amounted to approximately 1.03 billion USD [25]. According to a BBC article dated 8 December 2023, 14 people were killed in the 2023 floods, but economic loss estimates for this catastrophe are unavailable. The region’s hydrological dynamics are further complicated by its network of water bodies, including reservoirs such as Chembarambakkam and Poondi, which impact the area’s ability to manage excess rainfall and inflow. The urbanization process has profoundly transformed the physical environments of Chennai, Kancheepuram, and Thiruvallur, exacerbating the challenges these areas face. The study area is particularly interesting due to its susceptibility to climatic events, including monsoons and cyclones, which can lead to flooding. The hydrological dynamics in the study area are intricate due to the presence of reservoirs, water bodies, and densely populated urban zones. Kancheepuram district is traversed by the west-to-east flow of the Palar River, accompanied by various water bodies. Notably, Chembarambakkam Lake is a significant water reservoir in the district. Additionally, minor water bodies, such as Thenneri, Mathuranthagam, and Sriperambathur, contribute to the region’s hydrological complexity. In Thiruvallur district, major water bodies include Poondi and Puzhal lakes, further influencing the local hydrological system. The intricate network of waterways is complemented by the flow of the Koovam and Adyar rivers in Chennai, contributing to the overall hydrological dynamics of the study area. These water features introduce challenges and complexities, particularly concerning flood management and resilience in the face of climatic events. This study examines inland flooding from heavy rainfall and a severe cyclone, aiming to improve flood management strategies in urban areas with significant reservoir networks.

The spatial mapping and analysis of flood events in these districts provide valuable insights into the vulnerability and resilience of the region to extreme weather conditions, contributing to the broader understanding of flood management and climate resilience in coastal urban environments.

3. Materials and Methods

3.1. Materials

The materials used for the study include rainfall data, satellite images, and other data like tide height, reservoir details, IOD, and ENSO. Daily rainfall data from 19 stations provided by the Indian Meteorological Department (IMD) using a gridded 0.25° × 0.25° dataset were used (

Table 1. Comprehensive Details of Data Used for the Study.

S. No.	Data Used	Scale/Data Resolution	Year/Range	Source
1	IMD Gridded	0.25 × 0.25	2015 and 2023	IMD
2	RISAT-1A	33 m	12 November 2015	Bhoonidhi, NRSC
3	EOS-4	25 m	7 December 2023	Bhoonidhi, NRSC
4	Reservoirs	-	2015 and 2023 flood events	Chennai Metropolitan Water Supply & Sewerage Board
5	Tide data	-	2023	Survey of India
6	ENSO and IOD	-	2015 and 2023	NASA

). Flood extent mapping was extracted from RISAT-1A and EOS-04 satellites launched in 2012. RISAT-1A, a radar imaging satellite developed by the Indian Space Research Organization (ISRO), is part of the RISAT series designed for all-weather surveillance using SAR technology. EOS-04 is the fifth mission in the RISAT program, which was also developed by ISRO. Both satellite sensors have a SAR-C band as their primary instrument with a spatial resolution of 33 and 25 m (Table 1). The data in

Table 2. Rainfall data, including annual and seasonal rainfall during the flood event of 2015.

S. No.	Name	Annual Rainfall (mm)	Annual Rainy Days	Seasonal Rainfall (mm)	Seasonal Rainy Days	Rainy Days (Flood Event)	Rainfall (Flood Event)
1	Arambakam	2119.86	81	1516.87	39	28	1241.31
2	Periyapalayam	2342.65	77	1760.28	39	28	1421.85
3	Kanchipuram	2151.60	73	1680.51	34	28	1428.04
4	Karunguzhi	1911.99	65	1622.79	43	28	1348.33
5	Mahabalipuram	2410.06	70	1994.19	40	28	1644.72
6	Chembarambakkam	2425.14	78	1859.28	42	28	1547.46
7	Sengundram	2410.06	70	1994.19	40	28	1644.72
8	Amaipandalur	1465.05	65	1005.50	34	28	787.50
9	Kovalam	2253.35	66	1910.32	42	28	1636.73
10	Meenambakkam	2253.35	66	1910.32	42	28	1636.73
11	Ananthamangalam	1334.50	50	854.30	30	28	721.24
12	Poondi	2332.03	74	1765.14	36	28	1407.83
13	Puzhal Lake	2425.14	78	1859.28	42	28	1547.46
14	Pulicat lake	2311.10	81	1742.13	40	28	1417.16
15	Thiruttani	2053.77	79	1381.22	32	28	1093.63
16	Perambakkam	2332.03	74	1765.14	36	28	1407.83
17	Marakkanam	1908.92	66	1548.20	38	28	1395.50
18	Sholinger	1429.21	75	820.45	30	28	604.01
19	Ponnur	1334.50	50	854.30	30	28	721.24

are crucial for evaluating water quantities within reservoirs and the amounts released during flood events. This information was meticulously gathered from the Chennai Metropolitan Water Supply and Sewerage Board, ensuring its reliability and authority. Tidal data for the 2023 flood event were obtained from the Survey of India to examine tidal wave heights during cyclonic activity and extreme rainfall events. Additionally, a detailed study of ENSO and IOD activity was conducted using data from the NASA portal to understand the causative factors behind extreme rainfall and cyclonic activity. The integration of these diverse datasets enhances the robustness of the research, providing a comprehensive foundation for understanding the complex dynamics of the Chennai flood events.

3.2. Methodology

The RISAT-1A and EOS-4 images were pre-processed and speckle-filtered using the Refined Lee filter in the GEE code editor platform. The EOS-4 corrected images were geometrically co-registered with a RISAT-1A image for accurate alignment. The Classification and Regression Tree (CART) classifier extracted flood pixels from the SAR images. This machine learning method is known for its robustness in developing predictive models by constructing classification and regression trees based on input factors [26,27]. Depending on the type of dependent variable, the rule-based CART algorithm can operate as either a regression tree (RT) or a classification tree (CT) [28]. The dataset is divided into smaller subsets using a recursive technique according to the feature that provides the maximum information gain. This approach efficiently captures intricate, non-linear connections within the data [29]. To improve the model’s accuracy, CART examines variables multiple times, unlike typical regression approaches that only consider them once and might ignore crucial information. As a non-parametric approach, CART does not make any assumptions about the specific distribution of the variables under analysis. Unlike other machine learning algorithms like artificial neural networks (ANNs), which have a complex structure and are often viewed as black boxes, decision tree (DT) algorithms have the advantage of being presented as a white box, making it easy to interpret and understand the decision-making process [30].

Before conducting supervised classification, when learning a dependence from data, it is important to divide the data into the training and validation sets to avoid over-fitting. This research separates 70–30% of training and validation sets. The creation of a training set can be performed manually or automatically. While human-labeled samples provide a more accurate representation with reduced false positives and negatives, automated algorithms may suffer from local dependence [31]. In the GEE Code Editor environment, approximately 400 to 500 training datasets are collected from each SAR image to conduct supervised classification depending on the extent of flooding. The accuracy of the classification is calculated on the GEE platform.

Rainfall data are analyzed to identify whether a flood occurs due to monsoonal rainfall or unexpected excessive rainfall induced by cyclonic activity. This includes an examination of yearly, seasonal (Northeast Monsoon—NEM), flood event rainfall patterns, and rainy days. Although monsoonal activity is the primary factor that drives rainfall, it is also susceptible to being influenced by other climatic circumstances, such as cyclones or the IOD/El Niño event. The purpose of this study is to determine which climatic variables influence rainfall. Further, surface runoff into reservoirs occurs during occasions of heavy rainfall. Therefore, the study uses reservoir data, including inflow and outflow rates, to understand flood events. The methodology for this study is shown in Figure 2.

4. Results

4.1. Rainfall Analysis

In the study area, rain gauge stations are distributed throughout and around the region (Figure 3). In 2015, the annual rainfall was approximately 2000 mm across all stations, with around 70 rainy days. During the NEM season, the rainfall was 1570 mm over 37 days, as shown in Table 2. In contrast, the annual rainfall in 2023 was about 1544 mm, with 72 rainy days. During the NEM season, the rainfall amounted to 728.95 mm over 29 days, as detailed in

Table 3. Rainfall data, including annual and seasonal rainfall during the flood event of 2023.

S. No	Station Name	Annual Rainfall (mm)	Annual Rainy Days	Seasonal Rainfall (mm)	Seasonal Rainy Days	Rainy Days during Flood Event	Rainfall in the Flood Event
1	Arambakam	1483.58	94	777.10	34	1	125.67
2	Periyapalayam	1604.14	73	960.38	29	1	211.35
3	Kanchipuram	1529.79	76	530.22	25	1	78.14
4	Karunguzhi	1363.41	65	622.66	31	1	96.20
5	Mahabalipuram	1762.05	69	922.79	31	1	172.59
6	Chembarambakkam	2034.81	70	1046.79	29	1	243.97
7	Sengundram	1762.05	69	922.79	31	1	172.59
8	Amaipandalur	1133.47	69	396.93	25	1	51.07
9	Kovalam	1392.40	61	733.63	26	1	145.87
10	Meenambakkam	1392.40	61	733.63	26	1	145.87
11	Ananthamangalam	1497.21	64	617.29	29	1	23.72
12	Poondi	1717.65	85	745.82	31	1	134
13	Puzhal Lake	2034.81	70	1046.79	29	1	243.97
14	Pulicat lake	1568.56	75	922.26	30	1	180.82
15	Thiruttani	1368.59	86	471.03	24	1	86.36
16	Perambakkam	1717.65	85	745.82	31	1	134
17	Marakkanam	1340.50	69	660.38	33	1	51.05
18	Sholinger	1147.24	76	376.54	24	1	41.29
19	Ponnur	1497.21	64	617.29	29	1	23.72

. Most stations recorded over 1000 mm of rainfall, except for Ananthamangalam, Sholinger, and Ponnur stations. Sengundram and Mahabalipuram stations received a maximum rainfall of 1644.72 mm during the flood event between 9 November 2015 to 6 December 2015. The average rainfall in 2015 during the flood event was 880.47 mm.

The situation in 2023 was different from 2015. In 2023, a single day’s rainfall triggered the flood, whereas in 2015, the flood was caused by continuous rainfall events over 28 days across three spells. In 2023, most stations in the study area received heavy rainfall within a single day, except for Amaipandalur, Ananthamangalam, Marakkanam, Sholinger, and Ponnur stations. The total rainfall recorded in the 2023 flood event was 2362.27 mm. In both events, the stations that received the lowest rainfall were inland within the study area. This difference in rainfall patterns explains why the study area received more rainfall during the 2015 NEM season than in 2023.

4.2. Spatial Mapping of Flood

This study maps the spatial extent of the flood events during 2015 and 2023 in Chennai, Thiruvallur, and Kancheepuram districts, providing a comprehensive analysis of the inundation patterns.

The inundation pattern observed during the 2015 flood event, as depicted in Figure 4, reveals distinct characteristics in the affected regions. Particularly in the upper part of the Thiruvallur district, a pronounced and extensive inundation is evident, surpassing the severity observed in other areas. This heightened inundation suggests that specific geographical factors or local topography contributed to a more substantial accumulation of floodwaters in this region.

In contrast, the floodwaters show a more dispersed distribution in Kanchipuram district, spread across various parts. Unlike the concentrated inundation observed in Thiruvallur, the floodwaters in Kancheepuram appear to be distributed more evenly, suggesting a different spatial dynamic in this district. Furthermore, a noteworthy observation is the tendency of floodwaters to stagnate in close proximity to water bodies or rivers. The CART ML model has an accuracy of 0.96, indicating a robust fit of the classification model to the image.

Figure 5 comprehensively shows the inundation patterns during the 2023 flood event, revealing distinctive features in Thiruvallur, Chennai, and Kancheepuram districts. This visual representation provides nuanced insights into the spatial dynamics of flooding, emphasizing several key observations. Thiruvallur district shows significant inundation in the northern part, similar to the 2015 flood event. Notably, inland areas and those near water bodies bear a substantial impact, suggesting a consistent vulnerability in this region across flood events. Chennai exhibits a dual inundation pattern, affecting coastal areas and certain parts of the inland regions. This may indicate diverse factors contributing to flooding, including rainfall intensity, urbanization effects, and proximity to water bodies. In contrast to the other districts, Kancheepuram experiences notably heavy inundation, encompassing various parts of the district. The widespread nature of flooding implies a complex interplay of factors contributing to the severity of inundation in this region. Across all three districts, areas near river zones are consistently more inundated. This emphasizes the crucial role of natural watercourses in influencing the spatial patterns of flooding, as riverbanks are prone to heightened inundation during such events. Chennai and Kancheepuram district’s coastal regions bear the brunt of inundation during the 2023 flood event. Coastal vulnerability is evident, highlighting the unique challenges urban areas near the coastline face. The accuracy of the model for this particular image stands at 0.91. This signifies a noteworthy level of precision in image classification.

4.3. Integrated Reservoir Study for Flood Events

The reservoir details play a significant role in gauging the capacity of the water bodies to absorb excess rainfall and manage the inflow from adjacent areas. A comprehensive examination of reservoir data from the Chennai Metropolitan Water Supply and Sewerage Board was conducted during the investigation of the flood event. This investigation included the assessment of the quantity of water present within the reservoirs at the onset of the flood event and the amount of water released during the event. The amount of water obtained and discharged from the reservoirs played a significant role in impacting downstream regions, contributing to the comprehensive flooding scenarios observed in 2015 and 2023, as depicted in Figure 6 and Figure 7 below.

Figure 6 above depicts the inflow and outflow pattern of each reservoir during the 2015 flood event. The initial phase of the Northeast monsoon, spanning from 8 November 2015 to 15 November 2015, witnessed substantial rainfall throughout the region. During this period, most of the reservoirs in the study area commenced filling. Eventually, they reached their full capacity, driven by the persistent and heavy inflow from precipitation in the surrounding areas. The subsequent spell of rainfall, occurring from 20 November 2015 to 24 November 2015, unfolded seamlessly from the preceding spell, exerting continued pressure on the reservoirs. This continuous rainfall led to substantial discharge of floodwaters from the reservoirs. These meteorological events were attributed to a low-pressure system in the Bay of Bengal. The third spell, from 30 November 2015 to 2 December 2015, proved more severe than the preceding two spells during the Northeast monsoon season. This spell resulted in all reservoirs reaching their maximum storage capacity, prompting the sudden release of floodwaters. The cumulative impact of sudden and prior discharges exacerbated the flooding scenario in the study area. Across all three spells, Poondi Reservoir consistently ranked first in receiving and discharging floodwaters, followed by Chembarambakkam.

Commencing from the onset of December 2023, the study area witnessed a consistent downpour, initiating a gradual but persistent rise in water inflow into the reservoirs. This incremental surge reached its zenith on 4 December 2023, precipitating a substantial floodwater discharge from the reservoirs. Notably, the Poondi reservoir took the lead in this discharge, closely followed by the Chembarambakkam reservoir. On the pivotal date of 4 December 2023, all the regional reservoirs concurrently unleashed a formidable quantity of floodwater. In contrast, the Kammam Kottai Thervoy Kandigal reservoir discharged the minimum volume of floodwater during this event. The extensive discharge from the reservoirs inundated the study area, submerging it in floodwaters for the ensuing days, as shown in Figure 7.

This sustained influx of water was attributed to the extraordinary and intense precipitation induced by “Cyclone Michaung” in the Bay of Bengal. The cyclonic activity, characterized by extreme heavy rainfall, played a decisive role in the relentless inflow into the reservoirs, ultimately culminating in the significant discharge on and after 4 December 2023. The aftermath of this inundation underscored the critical role of meteorological events in shaping the hydrological landscape of the study area during the mentioned period.

4.4. Study of Cyclone Activities and Tidal Wave Heights during the Flood Event

As the annual cyclone season unfolded on 8 November 2015, a low-pressure area consolidated, evolving into a depression that gradually intensified into a deep depression. This atmospheric disturbance crossed the Tamil Nadu coast near Puducherry the following day. Land interaction and heightened vertical wind shear led to the system’s weakening, eventually transforming it into a well-marked low-pressure area over north Tamil Nadu by 10 November 2015. Although the precipitation from the preceding low-pressure system ceased by 24 November 2015, a new system emerged on 29 November 2015, resulting in additional rainfall. The IMD, recognizing the potential impact, forecasted heavy rainfall over Tamil Nadu throughout the week. The NASA Earth Observatory meteorological map captured the cloud movement on 1–2 December 2015 (Video S1), when the city of Chennai and neighboring districts Kanchipuram and Thiruvallur experienced a deluge, receiving the highest rainfall in a 24 h period than any recorded day since 1901. Video S1 illustrates the temporal progression of the phenomenon, as shown in the attached GIF file.

The Michaung cyclone started in late November 2023 as a low-pressure area originating in the Gulf of Thailand traversed into the Bay of Bengal. By 1 December 2023, IMD identified this disturbance as having evolved into a depression in the South Andaman Sea, forecasting its trajectory to the northwest while gaining strength in the Bay of Bengal. On 2 December 2023, the system further intensified, reaching the status of a deep depression positioned approximately 440 kilometers (270 miles) east–southeast of Puducherry. Subsequently, it escalated into a cyclonic storm, officially named Michaung by Myanmar. On 4 December 2023, cyclone Michaung attained its peak intensity as it neared the coast of Tamil Nadu, boasting winds with speeds of 110 kilometers per hour (68 mph). Gale-force winds, ranging between 80 and 90 kmph and gusting up to 100 kmph, prevailed over the affected region, gradually diminishing thereafter; details are shown in

Table 4. Cyclone Michaung—Chronological Timeline, Geographical Position, and Intensity Category of the Cyclonic Event.

Date/Time (IST)	Position (Lat/Long)	Maximum Sustained Surface Wind Speed (Kmph)	Category
04.12.23/0530	13.2/81.2	80–90 kmph gusting to 100 kmph	Cyclonic Storm
04.12.23/1130	13.7/80.9	85–95 kmph gusting to 105 kmph	Severe Cyclonic Storm
04.12.23/1730	14.2/80.6	90–100 kmph gusting to 110 kmph	Severe Cyclonic Storm
04.12.23/2330	14.7/80.5	95–105 kmph gusting to 115 kmph	Severe Cyclonic Storm
05.12.23/0530	15.4/80.4	90–100 kmph gusting to 110 kmph	Severe Cyclonic Storm
05.12.23/1730	16.3/80.6	70–80 kmph gusting to 90 kmph	Cyclonic Storm
06.12.23/0530	17.1/81.1	50–60 kmph gusting to 70 kmph	Deep Depression
06.12.23/1730	17.8/81.8	35–45 kmph gusting to 55 kmph	Depression

and Figure 8 [32].

The flood event in 2023 was the result of the torrential rainfall received on 4 December 2023 despite scanty rainfall on 3 December 2023 and moderate rainfall on 5th December, which followed a sporadic pattern. The rainfall on 4 December 2023 is classified as extremely heavy across all stations as per IMD classification, while on 3rd December 2023, it falls in the light to moderate category. On 5 December 2023, in most places, it falls in the heavy and very heavy categories [33].

In the wake of the cyclone, the Chennai area grappled with a storm surge ranging from 0.50 to 1.30 m, as shown in

Table 5. Storm surge height on Chennai coast during Cyclone Michaung (Source: Survey of India).

Date	Time (h:min)	Height (m)
1/12/2023	04:45	0.57
	10:22	1.09
	16:16	0.50
	22:56	1.30
2/12/2023	05:28	0.61
	11:03	1.04
	16:56	0.56
	23:39	1.25
3/12/2023	06:15	0.64
	11:52	1.00
	17:39	0.63
4/12/2023	00:22	1.20
	07:07	0.67
	12:56	0.97
	18:28	0.71
5/12/2023	01:11	1.15
	08:09	0.68
	14:16	0.97
	19:35	0.78
6/12/2023	02:07	1.11
	09:18	0.67
	15:45	1.01
	21:24	0.81
7/12/2023	03:11	1.08
	10:15	0.63
	16:52	1.08
	22:48	0.78

. This surge poses a significant challenge, necessitating vigilant measures to mitigate potential impacts on the coastal region.

4.5. Effects of ENSO/IOD on Northeast Monsoon

Meteorologists in India and abroad attribute the intensified rainfall to an exceptionally potent northeast monsoon. During the winter (October to December), prevailing winds typically traverse from the northeast to the southwest across the country, exerting a drying influence on many regions, particularly inland. However, these north-easterly winds intersect with the warm waters of the Bay of Bengal, facilitating substantial evaporation of moisture from the sea. This moisture-laden air is deposited over southern and eastern India, resulting in a significant portion of coastal eastern India’s annual rainfall—approximately 50 to 60 percent—during the winter (northeast) monsoon. The meteorological conditions in 2015 witnessed an amplification of this established pattern, attributed to the concurrent influence of record-warm seas and the far-reaching effects of El Niño. The IOD values for 2015 range from 0.23 to 0.33, with a peak on 1 November 2015, at a value of 1.41. Figure 9 below shows positive IOD throughout the year 2015. But in 2023, IOD had positive values after the month of June. The values range from 1.3 to 3.96, as shown in Figure 10.

In addition to the IOD conditions observed in 2015, the influence of El Niño on the Northeast monsoon in India becomes evident. Figure 11a depicts a noteworthy rise in the global land and sea temperatures, reaching up to 1.06 °C. This warming trend is intricately linked to the movement of warm winds from the Pacific Ocean, traversing across the west side of East Asian countries and concluding on the western side of the African continent. The co-occurrence of intense IOD conditions along the east coast of India, coupled with the influence of El Niño, creates a conducive environment for extreme precipitation, ultimately contributing to floods.

A parallel scenario unfolds in 2023, marked by the recurrence of a similar weather pattern. However, this time, the region experienced the added impact of a severe cyclone named Michaung. Figure 11b illustrates a temperature rise, reaching 1.44° C. The manifestation of such climatic conditions underscores the complex interplay of various meteorological phenomena, each contributing to the overall climatic dynamics. Further insights into the Sea Surface Temperature (SST) during November and December 2023 are presented in Video S2, depicted. The SST visualization reveals a gradual temperature increase from 28 to 30 degrees Celsius towards the end of November and the beginning of December 2023, accompanied by an anomaly rise from 0.5° to 1°. This temperature trend on the Indian Ocean surface indicates the intricate processes governing oceanic conditions during this period. The confluence of IOD, El Niño, and the impact of Cyclone Michaung in 2023 accentuates the complexity of climatic interactions, ultimately leading to heightened temperatures and extreme precipitation.

5. Discussion

From the above results, it is evident that extremely heavy precipitation was the primary cause of flooding in both 2015 and 2023. Stations that received higher amounts of rainfall were located near water bodies such as lakes and rivers, suddenly causing tanks to fill with surplus floodwater. Several scholarly works, including those by [34,35,36,37,38], have collectively noted that factors such as encroachments on waterways and water bodies, along with insufficient stormwater drainage systems, may increase the susceptibility of Chennai to flooding during heavy rainfall.

During the 2015 flood event, monsoonal patterns manifested in three distinct spells with intermittent intervals, a phenomenon extensively documented by both [39,40]. The prolonged and intense rainfall associated with the NEM triggered the release of floodwaters from the reservoirs in the study area, which struggled to cope with the relentless downpour. Rainfall began in the first week of November 2015, with the most extreme rainfall on 9 November 2015, resulting in widespread flooding. This peak rainfall resulted in considerable inflows into all reservoirs, including Poondi Lake, by 12 November 2015. However, the outflow of floodwaters from Poondi and Chembarambakkam lakes was postponed till 17 November 2015. The DSC Report 2015 states that the rainfall from 9 to 10 November was more than 500 mm, which caused reservoirs to fill up to their maximum capacity and caused flooding in both urban and agricultural areas. As the reservoirs reached their capacity, the surplus water was discharged. Additional rainfall and the release of excess water from lakes on 1–2 December increased flooding in already impacted areas, aggravating the 2015 flood disaster.

Consequently, runoff inundated settlements that were concurrently experiencing substantial rainfall. This outcome, as reported by [41], aligns with the observations of [42], who highlighted urban density as a contributing factor to the flood in 2015. The authors asserted that the high urban density hinders floodwater infiltration into the ground. The subsequent spells of rainfall, particularly the severe third spell, resulted in the sudden release of floodwaters. Poondi and Chembarambakkam reservoirs consistently played a primary role in receiving and discharging floodwaters, emphasizing their critical role in the hydrological dynamics of the region.

In contrast, the scenario in 2023 diverged. The onset of the extreme cyclone, “Michaung”, instigated heavy rainfall, causing certain reservoirs to attain full capacity and discharge floodwaters significantly during the intense downpour. This altered dynamic transformed the typical flooding scenario into a more severe condition, characterized by extensive inundation and water stagnation persisting for several days. The increased water inflow into the reservoirs caused a large discharge of flooding on 4 December 2023. The discharge was mostly caused by the Poondi Reservoir, influenced by the heavy precipitation from Cyclone Michaung in the Bay of Bengal. The continuous precipitation and surplus water from the reservoirs inundated the drainage systems, resulting in the accumulation of floodwaters in urban areas lacking adequate drainage. The cyclone played a decisive role in this sustained influx of water, underscoring the interconnectedness of meteorological events and hydrological consequences.

The analysis of reservoir data from the Chennai Metropolitan Water Supply and Sewerage Board provided crucial insights into the dynamics of the 2015 and 2023 flood events. The reservoirs were pivotal in storing excess rainfall and managing inflows from adjacent areas. Examining reservoir quantities, including the amount of water present and released during the flood events, illustrated their significant impact on downstream regions, contributing to the extensive flooding observed in these years.

Additionally, the causative factors behind extreme rainfall and cyclonic activity of 2015 and 2023 flood events reveal the nuanced influence of climatic factors, including El Niño and Cyclone Michaung, on the reservoir dynamics. An in-depth study of ENSO and IOD activity was conducted using data sourced from the NASA portal. ENSO is the Earth’s most dynamic and reliably predictable short-term climatic phenomenon [43]. This irregular but recurring behavior involves fluctuations in winds and SST over the tropical eastern Pacific Ocean. The two prominent extremes within ENSO are El Niño, characterized by a warm period when the SST anomaly in the western Pacific Ocean reaches or exceeds +0.5 °C, and La Niña, a cool period occurring when the SST anomaly falls below or equals −0.5 °C. In parallel, researchers have identified a distinct climate anomaly over the Indian Ocean in the last two decades, called the Indian Ocean Dipole (IOD). This critical air–sea coupled mode is characterized by contrasting SST anomalies in the western and eastern tropical Indian Ocean. Typically evolving in spring (May or June), peaking in the fall (October–November), and concluding in early winter (December), a positive (negative) IOD epoch manifests as cooler (warmer) than normal water in the eastern tropical Indian Ocean and warmer (cooler) than normal water in the western tropical Indian Ocean (http://www.bom.gov.au/climate/iod (10 April 2024). The 2015 flood, with its record-breaking rainfall, also bore the imprint of El Nino and IOD of about 1.41, which influenced the overall climatic patterns. The 2023 flood, on the other hand, showcased the impact of Cyclone Michaung as a result of IOD (3.92), emphasizing the significance of coastal storms in exacerbating flooding events. These episodes, spanning a 2-to-7-year cycle, exhibit considerable strength variability and wield substantial influence over global climatic patterns [44,45]. The temporal progression of these events and the corresponding meteorological conditions highlight the complex interplay of factors contributing to flooding in the Chennai region.

The storm surge associated with Cyclone Michaung further complicated the flood scenario, as detailed in Table 3. The study area faced a considerable challenge from a storm surge ranging from 0.50 to 1.30 m. Due to the tidal activity, the floodwaters could not drain into the Bay of Bengal. The floodwaters can only be drained into the sea once the cyclone has passed the coast and sea levels have returned to normal. This issue only arises shortly after the cyclone’s landfall. Correlating the frequency of cyclones and the occurrence of floods [46] are planning and vigilant measures required to mitigate the potential impact on the coastal region, which underline the importance of integrating meteorological data into flood management strategies.

The discussion shows that the study area experienced flooding due to continuous and sudden rainfall with cyclonic events, leading to tank overflows and urban inundation. Future flood preparedness and management strategies in the study region should consider these complex interactions and the potential impact of climatic anomalies on hydrological patterns.

6. Conclusions

In conclusion, the analysis of Chennai’s 2015 and 2023 floods highlights how cyclones, climate change, and extreme precipitation impact flood risk maps. These factors, along with urban encroachments and inadequate drainage, significantly heighten flood risk in cities, revealing the complex interplay that shapes flood scenarios. In 2015, the flood event characterized by prolonged monsoonal rains resulted in exceptional rainfall that exceeded 2000 mm at most stations. During this time, the Poondi and Chembarambakkam reservoirs were crucial in managing and discharging floodwaters, surpassing 35,000 cu. secs. In contrast, the 2023 flood event, caused by Cyclone Michaung, followed a different path. The discharge of floodwaters during intense rainfall exceeded 15,000 cu. secs due to the cyclone, which caused reservoirs to reach maximum capacity. This altered dynamic transformed the flooding scenario into a more severe condition marked by extensive inundation and prolonged water stagnation.

The comparison between the two events reveals the interplay of several climatic factors, with El Niño contributing to the 2015 flood and Cyclone Michaung playing a decisive role in 2023. The storm surge associated with Cyclone Michaung added complexity to the flood scenario, emphasizing the importance of integrating meteorological data into flood management strategies. This meteorological phenomenon exemplifies the intricacies of tropical cyclone development, underscoring the dynamic nature of atmospheric systems and their potential impact on coastal regions. The timely monitoring and analysis conducted by meteorological agencies play a critical role in informing and preparing communities for the evolving conditions associated with such weather events.

The conclusion highlights the importance of a comprehensive understanding of climatic events and reservoir dynamics in flood preparedness and management strategies for the study area. The study highlights the need for adaptive strategies for cyclones and climate change impacts on hydrological patterns. To improve flood risk maps, it is crucial to integrate hourly rainfall data, tidal activity, urban population information, and high-resolution SAR data. These measures will enhance cities’ resilience to future floods and support more effective mitigation strategies.