Assessing the Impact of Past Flood on Rice Production in Batticaloa District, Sri Lanka

Abstract

Flood risk to rice production has previously been examined in terms of river basins or administrative units, incorporating data about the flood year, inundated area, precipitation, elevation, and impacts. However, there is limited knowledge about this topic, as most flood impact studies have focused on loss and damage to people and the economy. It remains important to identify how flood risk to rice production can be better identified within a long-term, community-based, analytical framework. In addition, flood risk studies in Sri Lanka tend to focus on single-year flood events within an administrative boundary, making it difficult to fully comprehend risks to rice production. This paper aims to fill these gaps by investigating long-term flood risk levels on rice production. With this aim, we collected and analyzed information about rice production, geospatial data, and 15-year precipitation records. Temporal-spatial maps were generated using Google Earth Engine JavaScript coding, Google Earth Pro, and OpenStreetMap. In addition, focus group discussions with farmers and key informant interviews were conducted to verify the accuracy of online information. The collected data were analyzed using descriptive statistics, GIS, and linear regression analysis methods. Regarding rice production impacts, we found that floods in the years 2006–2007, 2010–2011, and 2014–2015 had significant impacts on rice production with 20.5%, 75.8%, and 16.6% reductions, respectively. Flood risk maps identified low-, medium-, and high-risk areas based on 15-year flood events, elevation, proximity to water bodies, and 15-year flood-induced damage to rice fields. High risk areas were further studied through field discussions and interviews, showing the connection between past floods and poor water governance practices in terms of dam management. Our linear regression analysis found a marginal negative correlation between total seasonal rainfall and rice production.

1. Introduction

Floods are among the most hazardous natural disasters and are a major constraint to rice production in many parts of the world [1,2]. While about 40 million hectares of lowland areas are used for rice cultivation [2], more than 100 million rice farmer’s livelihoods are impacted annually around the world by unfavorable submersion [3].

Past studies on flood impact on rice production are largely derived from case studies in a given river basin or administrative district. Using a climate model (MRI-AGCM3.2S), Shrestha et al. [4] assessed the future impact of large-scale flood incidences on rice crops within four major river basins in the Philippines, Vietnam, Cambodia, Indonesia, and Thailand. Their projection showed that flood damage areas from a 100-year flood were expected to increase by 13–55% across different river basin in the future. Another study [5] conducted a questionnaire survey to understand 20-year flood impacts on rice crops within an administrative district of Cambodia. It found that rice farmers in the district experienced an average loss of about 3.9 tons/ha due to floods [5]. Sri Lanka’s Batticaloa district experienced a 75% decline in rice production due to extreme weather events, including floods, from 2009 to 2010. In addition, approximately 143,560 acres of rice fields were destroyed by floods during the 2010–2011 season [6,7].

To the best of our knowledge, no case study has previously examined periodic changes to determine long-term flood risk. The accuracy of basin- or administrative-level flood risk predictions can be questioned, as some studies showed varied flood impacts in different communities within the same district, elevation, or river basin [8,9]. Considering these points, this study argues that community-based, long-term flood risk analyses might better understand localized flood risk levels, which, in turn, could help prioritize flood protection measures in the future. This paper, therefore, attempts to identify community-based flood risk areas for rice farmers before and after major flood years. It examines the frequency and intensity of seasonal and daily rainfalls alongside the rice production data for the period between 2007 and 2021. The following discussion first outlines the significance of the study area in addressing flood impacts and then provides an explanation of the data collection approach and methodology used to assess the impact of floods on rice cultivation. The ensuing main discussions highlight the study’s findings and significance.

2. Methodology

2.1. Study Location

Batticaloa district is situated in the eastern part of Sri Lanka, with geographical coordinates at 7°34′ N and 81°41′ E, covering a total land area of 2854 km² (Figure 1) [10,11]. It belongs to the dry zone because of its low annual rainfall. However, flood incidents seriously affect the area during the wet season due to intense monsoon rains. The district has a mean annual temperature of 27 °C and an average annual rainfall of 1756 mm [12]. In Sri Lanka, rice cultivation occurs in two major seasons: the maha season, which ranges from October to March, and the yala season, which spans from April to September [13]. This study focuses on the maha season, as 65% of farmlands in Batticaloa are rain-fed [14]. In this season, the north-east Monsoon brings to this region the highest amount of rainfall each year [15].

As of 2023, the district had a population of 581,751, comprising 276,947 males [16]. The population has grown steadily since 2012. The average household size in 2023 was about four persons. About 71.3% resided in rural areas. In the 2022–2023 maha season, about 12.4% of the rural population was engaged in rice cultivation [16,17]. Batticaloa district had 51,213 farmers who were actively involved in rice cultivation during the 2022–2023 season [16].

In the Batticaloa District Flood Contingency Plan of 2022–2023, Batticaloa is mentioned as a key agricultural region of Sri Lanka, as it contributes approximately 5% of Sri Lanka’s rice output [12,18]. The Ministry of Agriculture and Plantation Industries reported that 67,414 hectares of rice fields in Batticaloa produced 190,167 metric tons (MT), or 2.8 MT per hectare, during the 2018–2019 maha season [19,20]. In the 2020–2021 maha season, 65,630 hectares yielded 220,640 MT, or about 3.4 MT per hectare [21].

The rice production process in the district is supported by institutionalized services. It begins with 23 nurseries which supply saplings and plant materials that are essential for initiating rice cultivation. There are 18 agro-equipment sale centers where farmers can procure tools and machinery for field preparation and maintenance [10]. To nurture crop growth, farmers procure fertilizers and agrochemicals from one of 43 fertilizer sales centers. Additionally, there are 25 seed rice sales centers, offering high-yielding seed varieties [10].

After harvest, farmers bring their harvested rice grains to one of 22 rice purchasing centers. Among them, 12 centers have rice storage facilities [10]. These grains are distributed to 33 agro-product markets or 31 purchasing centers for sale to customers. Farmers may receive financial assistance from one of 32 agricultural credit facilities for procuring seeds, fertilizers, and equipment, facilitating the entire agricultural cycle. Rice cultivation in the district predominantly relies on rain-fed conditions; only about 35% of its arable land is under year-round irrigation [10].

The flat terrain and fertile alluvial soil that are prevalent in the district have historically provided ideal conditions for rice cultivation. For this reason, this district gained substantial attention from Portuguese, Dutch, and British colonial regimes, which introduced various rice farming practices, including irrigation systems in limited areas. As a result, the district has a long history of being the nation’s agricultural hub [14,22].

In 2011, after mainstreaming climate change policies into various policy areas, the Ministry of Environment identified the rice sector in Batticaloa as being highly vulnerable to climate change due to more frequent and intensifying weather events, as well as to its low-lying coastal areas. Farmers in this region, who rely on rainfall, face significant risks from changes in precipitation patterns, particularly during the north-east Monsoon [23].

Figure 1. Location map of Batticaloa district highlighting rice fields covering 676.73 km² and DSD boundaries [24].

Figure 1. Location map of Batticaloa district highlighting rice fields covering 676.73 km² and DSD boundaries [24].

According to past records, Batticaloa experienced severe flood incidents in 1878, 1904, 1906, 1913, and 1957 [7]. A major flood event in 1957 resulted in the breakage of the Unnichai Dam, although no detailed information is available about agricultural damage and losses, partly due to the Civil War that lasted for more than 25 years and ended in 2009, devastating the study area. The floods in 2010 and 2011 were considered the most severe since 1957, causing economic losses of about LKR4.5 billion in Batticaloa alone [25]. About 143,560 acres of paddy cultivation were destroyed, affecting 44,179 farmers [7]. Even in the 2009–2010 maha season, rice production experienced a 75% decline due to exceptionally heavy rainfalls [6]. From 1974 to 2016, Batticaloa was the most flood-affected district in the country in terms of the number of people impacted [26].

2.2. Data Collection and Analysis

For this research, we collected and analyzed the data from 2006 to 2021 on rice production (gross extent, net harvested hectares, average yield per hectare, and total rice production), rainfall (seasonal, monthly, and daily), floods, and factors associated with flooding in Batticaloa district. We collected these data from the Department of Census and Statistics of Sri Lanka [27]. These helped us identify (1) rice production loss and damage, (2) rainfall fluctuations and intensity, (3) flood risk levels in the study area, (4) correlation between rice production and seasonal rainfall, and (5) flood determining factors (Figure 2).

As to the first objective of identifying rice production loss and damage, we quantified the volume of unharvested rice directly attributable to flood-related damage. Rice field damage was then identified by estimating the loss in rice production. The estimation was derived by multiplying the extent of the damaged area with the average yield per hectare [28].

Production loss = Rice field damage × Yield per hectare

(1)

Production loss % = Production loss/(Gross extent × Yield per hectare) × 100

(2)

The second task of identifying rainfall fluctuations and intensity were done by examining seasonal, monthly, and daily precipitation patterns and their influence on rice production losses from 2006 to 2021. Google Earth Engine (GEE) was utilized to analyze the data through JavaScript coding. Since the maha season spans from October to March, rice production for 2007, for example, was calculated based on the maha season data from October 2006 to March 2007. Rainfall frequency and intensity were then visualized in multiple years to assess trends and anomalies in rainfall events during the maha season, the crop maturing period, and the harvesting period. The coding procedure used to extract rainfall data is provided in Appendix A.

To identify community-level flood risks, our third task, three types of flood inundation and risk level maps were created. Geospatial tools and datasets were utilized to identify flooded areas and determine flood risk levels in the Batticaloa district. The region of interest (ROI) was defined by importing a boundary shapefile and centering the map for effective visualization. Subsequently, the Global Surface Water (GSW) dataset (JRC/GSW1_4/GlobalSurfaceWater) was loaded and clipped to the ROI in Google Earth Engine (GEE) to analyze historical surface water presence. The method of extracting global surface water is illustrated in Appendix B.

To visualize the flooded areas on a map, the occurrence band from the dataset was extracted to identify water bodies and their variability from 2006 to 2021. Pixels with high occurrence values (>80%) were classified as permanent water bodies, while intermediate values (20–80%) indicated areas prone to flood inundation. These flood-prone areas were isolated by applying thresholds and masking the corresponding pixels. The processed data were exported to Google Drive for further analysis, maintaining a 30-m spatial resolution. Accuracy issues may have arisen due to the similarities in spectral signatures between floodwater and other land cover types, particularly in densely vegetated or shadowed areas. To mitigate these limitations, the results were validated using ground-truth data and historical flood records based on government maps from riskinfo.lk [29]. The national Disaster Management Centre (DMC) previously identified flood-prone areas nation-wide by district [29]. This information was used as a reference for mapping the flooded areas (Figure 5a). The identified flood inundation area was then used to calculate the extent and percentage of flood-affected rice fields through spatial intersection and geometric analysis.

To illustrate flood risk levels in the map (Figure 5b), geographical datasets related to flood risk factors were collected, including flooded areas, elevation points from Google Earth Pro, and drainage and rice field data from OpenStreetMap. All datasets were projected to a common coordinate system (UTM Zone 44N, WGS 84). The flood risk factors were then analyzed by reclassifying the Digital Elevation Model (DEM) into low (4–6 m), medium (2–4 m), and high (0–2 m) elevation zones to assess susceptibility to flooding. Additionally, 500-m drainage buffer zones were created to represent varying flood impact distances. Identified flood inundation areas were intersected with rice field polygons to determine the extent of flooded rice fields.

Next, the vector datasets for elevation, drainage buffers, flood inundation, and inundated rice fields were converted to a raster format. Here, a weighted overlay analysis assigned weights to the following four factors based on their contribution to flood risk: (1) flood inundation was given the highest weight (40%) because it directly represents the spatial extent and frequency of past flooding events, which is the most significant indicator of future flood risk; (2) Elevation was prioritized next (30%), as low-lying areas are inherently more prone to flood accumulation and water retention; (3) Drainage proximity was considered to be moderately influential (20%), as areas closer to drainage paths or rivers are more likely to experience flood overflow or backflow during heavy rainfall events; (4) Rice lands was assigned the lowest weight (10%), as it is more an exposure indicator than a driver. This analysis led to the classification of low, medium, and high-risk zones in the district. As a reference point, a related study by Acted [30] was reviewed, which identified the product of hazard, exposure, and vulnerability using the World Risk Index methodology.

Additionally, this study reviewed some methods used by previous studies to ensure the accuracy of the applied maps. For example, Desalegn and Mulu [31] combined GIS, a 30 m × 30 m Digital Elevation Model (DEM), rainfall data, and soil data to delineate flood areas by calculating water surface elevations at various cross-sections. These elevations were imported into ArcGIS 10.3 to generate floodplain maps for different return periods (e.g., 5, 10, 25 years). Bukari et al. [32] used the Pairwise Comparison Method (PCM), a component of the Analytical Hierarchy Process (AHP), to analyze flood impacts by assigning weights to different flood-related criteria using digital terrain data, rainfall, flow direction, and flow accumulation.

A 15-year community-based flood risk level map (Figure 5d) was visualized using color-coded symbology. Grama Nildhari Divisions (GNDs) represent community-level boundaries, i.e., the smallest administrative units in Sri Lanka. Following GND boundaries and using information about flood risk level percentages, this study incorporated information about past 15-year flood events, elevation, proximity to water bodies, and 15-year flood-induced damage to rice fields. As a reference point, this study examined the flood impact risk map of 2011 (Figure 5c) made by the Disaster Management Centre of Sri Lanka. We obtained it from the Disaster Management Centre, Batticaloa. This government map was prepared based on the information of the Disaster Risk Information Platform and desinventar, using data about flooded areas and the number of impacted people. Regarding this government map, we had a key informant interview with an assistant director of the Disaster Management Centre in charge of Batticaloa district regarding the accuracy of flood risk estimates. This comparative method allowed flooding risk classification, ensuring that vulnerable GNDs were accurately identified. The process was supported by multiple validation sources, including focus group discussions, field observations, and key informant interviews conducted in 2024.

In the final stage of this research, this study conducted a linear regression analysis to explore the correlation between seasonal rainfall and rice production. The model incorporated key statistical indicators such as R-squared, adjusted R-squared, and Root Mean Square Error (RMSE) to assess the strength and significance of this relationship and how rainfall amount fluctuations affected rice yields over time. Rainfall is crucial for rice cultivation, but it is not the sole determinant of productivity. For instance, excessive or poorly timed rainfall can have adverse effects, such as flooding, which may lead to significant crop damage due to heavy and prolonged rains [33].

The linear regression equation used to model this relationship is expressed as:

Rice Production = β₀ + β₁ × Rainfall + ϵ

(3)

Here, rice production represents the predicted yield, and seasonal rainfall is the total rainfall recorded during the maha season. The term β₀ (intercept) reflects the expected rice production when rainfall is zero, while β₁ (slope) indicates the change in rice production for a one-unit increase in rainfall. The error term ϵ accounts for variations in rice production that cannot be attributed to rainfall alone.

The coefficients β₀ and β₁ are estimated using the least squares method, which minimizes the sum of squared differences between observed and predicted rice production. Their interpretation is as follows:

• β₀ (Intercept) represents the predicted rice production in the absence of rainfall. While this value may not hold practical significance (since rice cultivation requires water), it provides a baseline for the regression model;
• β₁ (Slope) indicates the effect of rainfall on rice production. For example, if β₁ = 0.5, it suggests that every additional millimeter of rainfall increases rice production by 0.5 units (e.g., tons).

3. Results and Discussion

3.1. Rice Production Loss and Rice Field Damage in the Maha Season

Our first objective of assessing loss and damage revealed that 2010–2011 and 2016–2017 seasons experienced the highest losses, with reductions of 75.8% and 41.3% from the previous years, respectively (

Table 1. Rice production loss in metric ton at Batticaloa District in the maha Season from 2006 to 2021.

Year	Gross Extent (Ha)	Net Extent Harvested (Ha)	Net Average Yield Per Hectare (MT)	Total Production (MT)	Yield	Damage (Ha)	Production Loss (MT)	Production Loss %
2006–2007	18,469	14,685	2.6	37,476	2.0	3784	9657	20.5
2007–2008	19,134	17,030	3.2	55,020	2.9	2104	6798	11.0
2008–2009	45,905	45,854	3.2	147,278	3.2	51	164	0.1
2009–2010	54,855	54,329	3.6	193,274	3.5	526	1871	1.0
2010–2011	59,520	14,424	2.1	30,539	0.5	45,096	95,479	75.8
2011–2012	59,470	52,115	3.3	171,715	2.9	7355	24,234	12.4
2012–2013	66,276	62,747	1.8	115,630	1.7	3529	-	5.3
2013–2014	62,204	62,204	2.7	166,010	2.7	0	0	0.0
2014–2015	61,014	50,866	2.4	120,570	2.0	10,148	24,054	16.6
2015–2016	61,988	56,401	2.7	150,434	2.4	5587	14,902	9.0
2016–2017	62,836	36,866	2.2	81,878	1.3	25,970	57,678	41.3
2017–2018	64,796	64,796	2.7	177,288	2.7	0	0	0.0
2018–2019	67,214	67,214	2.8	190,167	2.8	0	0	0.0
2019–2020	66,996	61,244	3.3	204,489	3.1	5752	19,205	8.6
2020–2021	67,074	65,630	3.4	220,640	3.3	1444	4855	2.2

). This meant a loss of 95,479 MT in 2010–2011 and 57,678 MT in 2016–2017. The total rice field damages in these two seasons were 45,096 hectares and 25,970 hectares, respectively. Per hectare rice productivity declined to 0.5 MT in 2010–2011 and 1.3 MT in 2016–2017. Over the 15-year period, the average rice productivity in Batticaloa was calculated at 2.5 MT per hectare, compared to the national average of 3.6 MT per hectare [27].

Additionally, the 2006–2007, 2007–2008, 2011–2012, 2014–2015, 2015–2016, and 2019–2020 seasons experienced moderate losses in rice production, with reductions of 20.5%, 11%, 12.4%, 16.6%, 9%, and 8.6% from the previous years, respectively. The total rice field damage amounted to 3784 hectares, 2104 hectares, 7355 hectares, 10,148 hectares, 5587 hectares, and 5752 hectares, respectively.

3.2. Patterns in Rainfall, Floods and Rice Production Losses

Seasonal and daily rainfall patterns and their influence on floods and rice production losses in Batticaloa were examined using data from the cloud-based platform Google Earth Engine (GEE) (Figure 3 and Figure 4). In terms of seasonal rainfall, highly intense rainfall seasons were found in 2010–2011, 2012–2013, and 2014–2015. The rice production losses of these seasons amounted to 75.8%, 5.3%, and 16.6%, respectively. Conversely, the 2016–2017 season witnessed low rainfall, leading to a 41.3% loss due to drought conditions [34,35,36].

The daily rainfall data analysis results confirmed that the three-month period (November to January) in the study area was critical in understanding rice production losses (Figure 3). In 2006–2007, heavy rainfalls (about 100 mm) were observed during the harvest period (last week of January 2007). In the 2007–2008, 2011–2012, 2015–2016, and 2019–2020 seasons, the study area experienced frequent and significant rainfalls from November to December.

In the 2007–2008 season, Batticaloa experienced continuous rainfall throughout the entire period. Between the second and third weeks of December, rainfall ranged from 40 to 80 mm over three consecutive days, followed by a decrease to below 30 mm. This pattern was repeated with another three-day period of 40 to 80 mm, a subsequent drop below 30 mm, and finally, another three-day period with rainfall between 40 and 80 mm. In 2011–2012, the last week of November experienced five consecutive days of rainfall ranging from 40 to 50 mm. A similar pattern occurred from mid to late December, with another five days of continuous rainfall of 40 to 50 mm, followed by a brief reduction to below 15 mm. Rainfall then increased again, exceeding 40 mm in the first week of January 2012. In 2015–2016, the area experienced consecutive days of 40 to 50 mm of rainfall in the second week of November, followed by similar rainfall in the third week of November, as well as in the second and last weeks of December. In 2019–2020, the area experienced continuous rainfall ranging from 40 to 80 mm from the last week of November to the first week of December, followed by a drop to 0 mm. Rainfall then measured 40 to 50 mm for three days in the second week of December, dropped to 0 mm again, and then increased to 40 to 50 mm for three days in the third week of December.

This observed rainfall pattern contributed to the occurrence of floods (Figure 4). Hemachandra et al. [37] highlighted that increased heavy rainfall during specific periods can lead to waterlogging and potential flooding in agricultural areas.

Figure 3. Total seasonal rainfall distribution in millimeters (Google Earth Engine, 2023) [38].

Figure 3. Total seasonal rainfall distribution in millimeters (Google Earth Engine, 2023) [38].

Figure 4. Daily rainfall during the period of November to January in (a) 2006–2007, (b) 2007–2008, (c) 2011–2012, (d) 2015–2016, and (e) 2019–2020 [38].

Figure 4. Daily rainfall during the period of November to January in (a) 2006–2007, (b) 2007–2008, (c) 2011–2012, (d) 2015–2016, and (e) 2019–2020 [38].

3.3. Flood Exposure

Regarding the third objective of creating flood risk level maps, we used the Global Surface Water (GSW) datasets, RiskInfo data, Google Earth elevation data, and rice field and river data from OpenStreetMap. Using GIS, we then assessed flood exposure in the study area based on a flood inundation map developed for the 2006–2021 period (Figure 5a), along with a flood risk level map (Figure 5b).

Figure 5a shows color-coded areas for existing surface water area (blue), flooded area (red), and rice field (green). DSD boundaries are indicated to identify flooded areas within these administrative areas. It shows that Koralai Pattu South DSD had 3184.5 hectares of rice fields inundated over a 15-year period. In particular, the Maduru Oya River Basin, including Vakaneri Tank—where Punnani Anicut is located—experienced the most intense flood impact within the division. This is where 78 minor tanks are located [39]. Eravur Pattu DSD experienced the second-largest flooded area, with 3000 hectares of rice fields. The area surrounding Mavadiodai Anicut and the Mundeni Aru/River was flooded due to water discharge from Rukam Tank during the monsoon season. The Mundeni Aru/River Basin contains six major and medium-sized reservoirs: Tampitiya, Weligahakandiya, Borapola, Rugam, Kitul Wewa/Lake, and Rambaken Oya/River Reservoir [40]. The Rugam Reservoir gently slopes toward the sea through Valachchenai and Batticaloa Lagoons.

Figure 5b provides insights into risk levels for the entire district with DSD boundaries. Drawing on the study by Acted [30], which calculated the risk using the World Risk Index, where risk is determined as the product of hazard, exposure, and vulnerability, risk levels were color-coded with purple as low, blue as moderate, and red as high risk. The high, moderate, and low risk areas consisted of 36%, 22%, and 42%, respectively, of the total area. This map highlights the localized distribution of high flood damage across the district. High-risk areas are generally located near residential zones, particularly around Arukarkudah and Panichchankeni (south of Upaar Lagoon), where three rivers—Makarachchi Aru, Bodigoda Aru, and Kirimechchi Odai—flow into Upaar Lagoon. Other high-risk areas include Mavadivembu and Sittandy (south of Valachchenai Lagoon), where Miyangolla Ela, Mundeni Aru, and Lavani Aru converge. Eravur Town is another high-risk area within the Batticaloa Lagoon watershed. Muthalaikuda, Palugamam, and Mandur are located along the Namakada Aru. Thumpankeli is located near Batticaloa Lagoon.

Figure 5c illustrates flood risk zones by Grama Niladhari Divisions (GNDs). This map was developed by the Disaster Management Centre in 2019 using the 2011 flood data that are available on the Disaster Risk Information Platform, which was developed by the Disaster Management Centre (DMC) with funding and technical support from the Global Facility for Disaster Reduction and Recovery (GFDRR) and the World Bank [41]. This map is based on information about inundated areas and the number of people affected. With permission from the Centre, we placed this map here to show different risk interpretations from ours.

Figure 5d illustrates the spatial distribution of flood risk levels at the GND level, whereas Figure 5b highlights DSD risk level, and Figure 5c highlights GND risk levels. Risk levels were color-coded using the same colors for risk levels as in Figure 5c. Among them, 80 GNDs were identified as being highly vulnerable and 219 as moderately vulnerable. Out of the 80 GNDs, 6 in Koralaipattu North DSD are located near Upaar Lagoon and Bodigoda Aru/River.

Water governance practices, including flood management, are complicated within Koralaipattu North. There are several large- and medium-size tanks, including Anaisuddakadu, Mathurankerny, Kirimichchai Odai, Tharavai and Meiyankallu. The operation and maintenance of these facilities fall under the Provincial Irrigation Department (PID). Several major and medium-size tanks, including Kaddumurivu Tank, Rugam, Welikandiya, Kithulwewa, and Vahaneri, are placed under the jurisdiction of the Central Irrigation Department (CID). There are another 30 minor tanks in this DSD that are managed by the Department of Agrarian Development (DAD) [39]. Another DSDs have 13 GNDs that are located near Valachchenai Lagoon, Lavani Aru/River, Maduru Oya/River, Miyangolla Ela, Mundeni Aru/River, and Vakaneri Wewa/Lake. These watersheds are connected to Koralaipattu South and Eravur Pattu DSDs, where the operation of 151 minor tanks is controlled by different administrative authorities [39].

In our field work and interviews, it was found that this jurisdictional mess has created coordination problems. For example, a past flood event affected 61 GNDs located near Batticaloa Lagoon partly due to sediment accumulation in tanks that raised water levels and led to water overflow. Additionally, during the 2024–2025 flood event, the areas enclosed by riverbanks, Batticaloa Lagoon, and Valachchenai Lagoon were affected due to the opening of gates at major tanks (Navagiri, Unichchai, Rugam, and Kithulwewa) that were managed by the Central Irrigation Department. The Department opened the gates when water levels rose due to intense rainfall without prior notification to the District Disaster Management Centre or the District Secretariat of Batticaloa.

Figure 5. (a) Flood inundation map, (b) Flood risk level map, (c) Government flood risk zone map by Grama Niladhari Divisions [42], (d) Flood risk level map at the Grama Niladhari Division Level.

Figure 5. (a) Flood inundation map, (b) Flood risk level map, (c) Government flood risk zone map by Grama Niladhari Divisions [42], (d) Flood risk level map at the Grama Niladhari Division Level.

3.4. Correlation Between Total Seasonal Rainfall and Rice Production

In the final section of the analysis, a linear regression was conducted to examine the relationship between Total Seasonal Rainfall (TSRF) and Total Production (TP). The result showed an 8.7% variability in total production (R-squared = 0.0870), indicating a weak correlation between TSRF and TP. The adjusted R-squared value was 0.0167, suggesting that the model’s explanatory power is limited. The Root Mean Square Error (RMSE) was 0.84379, indicating the average prediction error. The coefficient for TRF was −0.0006726, with a standard error of 0.0006044 (

Table 2. Total Seasonal Rainfall (TSRF) and Total Production (PRO).

Source	SS	df	MS		Number of obs	=15
					F (1, 13)	=1.24
Model	0.881673533	1	0.881673533		Prob > F	=0.2859
Residual	9.25584399	13	0.71198		R-squared	=0.0870
Total	10.1375175	14	0.724108395		RootMSE	=0.84379
Variables	Coef.	Std. Err.	t	p > |t|	95% Conf.	Interval
tsrf	−0.0006726	0.0006044	−1.11	0.286	−0.0019783	0.0006332
cons	3.396023	0.9431542	3.60	0.003	1.358462	5.433583

). These results mean that the 15-year analysis diluted these results along with the influence of additional factors such as pest outbreaks, soil salinity, irrigation practices, and farming methods, as they may play a significant role in determining yield outcomes. However, analysis of individual years indicated that the significance of correlations varied, suggesting that seasonal rainfall variations are more relevant for understanding the impacts of rainfall on rice production.

This study then examined seasonal correlations. The analysis indicated that intensive rainfall events in 2010–2011, 2012–2013, and 2014–2015 seasons were statistically significant in terms of increased flood occurrences and corresponding reductions in rice production (Figure 6). Additionally, frequent and intensive rainfall events during the maturation period (November-December) in 2007–2008, 2011–2012, 2015–2016, and 2019–2020 exhibited a significant negative impact on rice production, suggesting a strong correlation between excessive precipitation during the maturation period and yield reduction.

Additionally, this study found that rice production in the 2006–2007 season experienced significant losses due partly to excessive rainfall during the harvesting period. However, the model accounts for only a limited proportion of the variability in production (R² = X), suggesting that other factors, such as soil conditions, pest infestations, or management practices, may also have influenced the observed outcomes. This implies that the 2006–2007 season may not be an anomaly but rather part of a broader pattern influenced by multiple interacting variables.

Although considering all possible factors is beyond the scope of this paper, after our field work and target group discussion, we argue that the following factors plausible: (1) rapid urbanization has increased impermeable surfaces such as roads and buildings, disrupting natural drainage systems and intensifying surface runoff. (2) Agricultural areas near riverbanks and lagoon became particularly vulnerable, as long-term heavy rainfall had accumulated sediments on the bottom of rivers and lagoon, raising water levels and causing river overflows [43,44].

4. Conclusions

This paper examined community-level, 15-year flood impacts on rice production, highlighting the complex interplay among seasonal rainfall, flood loss/damage, rice production, and productivity within Batticaloa district, Sri Lanka. While previous flood related studies examined flood impacts on crop production within a river basin/district or administrative unit, our long-term flood impact research revealed more nuanced flood outcomes on rice fields at a community level. Compared to the government flood vulnerability map, which was based on the 2011 flood impact on people, our maps showed 15-year flood risk levels on rice fields after analyzing elevation, flood events, proximity to water bodies, and damage to rice fields. This result was then validated by our field work and interviews.

The 15-year, community-based flood risk map provided new and more localized insights into the spatial distribution of flood vulnerability. This type of map could help government agencies in developing countries with limited resources to better identify high-risk areas, allowing them to prioritize resource allocation and design targeted disaster mitigation and climate adaptation measures. In Batticaloa district, 80 GNDs were identified as being highly vulnerable and 219 as moderately vulnerable. Incorporating insights gained from local interviews also shed important light on our understanding of flood risks. Flood risk, this study contends, means more than geographical characteristics. Flood incidents can occur partly due to the poor management of existing protection infrastructure. Whereas many tanks are located within the district, which function as flood water storage at some points, poor inter-jurisdictional communication and coordination have led to the intensification of flooded areas near anicuts and other types of dams.

The findings of the regression analysis underscored the significant relationship between heavy seasonal rainfall events—especially during the maturation and harvesting periods—and rice production decline in the high-risk area. The correlation between Total Seasonal Rainfall (TSRF) and Total Production (TP) with an R-squared value of 0.0870 indicated that only a small portion of production variability could be explained by seasonal rainfall alone. These seasonal and community-level patterns suggest that high-intensity and frequency rainfall events during the maturation and harvesting periods are more influential on rice production in this high-risk area. These insights are significant for informing targeted adaptation strategies that consider not just rainfall variability but also land use changes and localized flood risks.