Identifying the local factors of resilience during cyclone Hudhud and Phailin on the east coast of India

Abstract

The vulnerabilities of coastal communities have increased in recent years, as more people have to live in hazard-prone areas due to population growth. The east coast of India is one such vulnerable area that faces the combined challenge of climate risks and poverty. This study identifies the environmental and socioeconomic factors contributing to the resilience building that helped the communities in the study area to cope with cyclone Phailin in 2013 and cyclone Hudhud in 2014. We used questionnaire surveys, GIS and satellite data, and econometric analysis to identify these features. The damage suffered by the households, and the time needed by them to return to the pre-cyclone situation were used as indicators of resilience. We found that factors such as high education, decision-making power of women, and presence of coastline vegetation of local species such as cashew and palmyra helped the communities to build resilience toward cyclones.

Electronic supplementary material

The online version of this article (doi:10.1007/s13280-019-01241-7) contains supplementary material, which is available to authorized users.

Introduction

Building resilience to climate extremes is necessary to improve the human wellbeing and to eradicate poverty. Natural disasters affect the poor disproportionately and cause more loss to their welfare than to that of the rich (Sakai et al. 2017). Studies on the relationships between poverty, disasters, and climate extremes indicate that climate change and its impacts accelerate poverty in the vulnerable developing countries (Skoufias 2003; Shepherd et al. 2013; Sawada 2017). Economic growth in South Asia averaged 6% per annum over the past 20 years and is expected to increase to 7% or higher in coming years.1 Despite such growth, this region accommodated 34% of the world’s poor, around 274 million people who subsisted on USD 1.90 a day in 2013.2 Nearly half of these people live in the low-lying coastal regions of South Asia.

Large population and poverty force more people to live in marginal, hazard-prone regions (such as low-lying urban areas, disaster-prone coastal areas, and flood plains) and put them under persistent risk from combined impacts of rapid-onset extreme events (e.g. cyclones or storm surges) and slow-onset processes of climate change such as sea-level rise (Stephenson et al. 2010; Woodruff et al. 2013). Population pressure causes degradation and loss of ecosystems, the first line of defence against natural disasters (IPCC 2012; UNISDR 2015). Increased vulnerability from climate change is reflected in the global economic loss caused by natural disasters, which rose from USD 50 thousand million in the 1980s to USD 250–300 thousand million per annum in 2010s (Benfield 2014; UNISDR 2015). These factors have reinforced an interest in studying the linkages between poverty and climate extremes to prescribe resilience-building policy options (IPCC 2012). Increasing the adaptive capacity of people decreases their exposure to natural disasters in vulnerable areas and makes them more resilient (Ross 2014).

The east coast of India witnesses frequent cyclonic disturbances (Sahoo and Bhaskaran 2016). Some of the most densely inhabited areas in India, with population density exceeding 400 persons per square kilometre, are located on the east coast probably due to the availability of water and fertile land e.g. delta plains of Mahanadi, Godavari, Krishna, and Cauvery (Kumar et al. 2005). The combination of climate risks, poverty, and a high population density calls for the need to develop resilience building in these communities. The government of India adopted the National Policy on Disaster Management in 2009 with a vision “to build a safe and disaster resilient country”. Accordingly, the cyclone-prone areas are provided with multi-purpose storm shelters, approach roads, saline embankments, early-warning systems, rescue and evacuation training, and disaster rapid action force to manage storm hazards effectively.3 These measures are essential to provide the emergency response needed when an extreme climatic event such as a cyclone starts and have been proved effective in some areas (Das 2019). Socioeconomic and environmental factors contributing to a community’s resilience to natural disasters, however, remain unacknowledged by the existing policy. Addressing these factors will make coastal communities more resilient and sustainable. This article identifies such features by studying one site in the state of Andhra Pradesh in India, which was hit by cyclone Hudhud in October 2014. The areas had the above-mentioned disaster management measures in place, when the cyclone hit.4 We identify the features which helped these communities to cope with the impacts of this disaster and contributed to their resilience. As the climatic hazards are projected to intensify further (IPCC 2012; Sawada 2017), the present study provides an important input for developing an appropriate climate adaptation policy for vulnerable coastal regions. The paper first explains its conceptualization of resilience, which is then followed by objectives, description of study area, analytical methods, results, discussion and the concluding remarks.

Resilience in disaster management

Resilience is the ability of a system to absorb changes and still persist (Holling 1973). After the seminal work of Holling, resilience has been studied in different contexts and is defined as the ‘ability of an entity (individuals, communities, organizations, states) to recover from the effects of exogenous shocks, such as natural hazards without compromising the long-term prospects of growth’ in disaster management (Adger et al. 2005; Briguglio et al. 2008; Kousky and Shabman 2015). It denotes “the ability to prepare and plan for, absorb, recover from, and more successfully adapt to adverse events” (NRC 2012, p. 1). Resilience is linked inversely to vulnerability and directly to the adaptive capacity of a system (Cutter et al. 2008; Ross 2014; Yoon et al. 2016). Though resilience is usually perceived as a long-run phenomenon, it also reflects the short-term immunity of a system immediately after a shock. The “ability of a system to maintain function when shocked or the ability to limit the magnitude of immediate loss of income after a disaster” is described as ‘static resilience’, whereas “hastening the speed of recovery or the ability to recover and reconstruct quickly after a shock” is called ‘dynamic resilience’ (Hallegate 2014; Chang et al. 2018). Static resilience reduces the initial damage, while dynamic resilience reduces the recovery time after a disaster.

Figure 1 explains the concepts of static and dynamic resilience for a representative household (Kousky and Shabman 2015). When a disaster strikes, the household’s financial condition declines. Based on its locational features, socioeconomic status, and the pre-disaster capacity-building/adaptation activities, the household may fall to the points of ‘A’ or ‘B’. A static-resilient household will reach point ‘A’, but a non-resilient household will suffer more damage and reach point ‘B’. Then, the household takes steps to recover after the disaster to reach the pre-disaster level pp*. Depending on its dynamic resilience, it may attain recovery 1 or 2, with four possible paths.

Fig. 1.

Pathways of recovery under conditions of resilience. Adapted from Kousky and Shabman (2015). The y-axis represents the household’s financial condition (income), the x-axis represents time, and PP* is the household’s pre-disaster financial situation

Path 1 depicts the most resilient scenario of ‘smaller impact, faster recovery’. It can be achieved with multiple adaptation measures (e.g. storm-resistant construction, early warning, crop insurance, disaster aid, disaster management training etc.). With limited measures, e.g. only post-disaster aid, the recovery may take Path 2, which improves recovery time but not the initial damage. If adaptations undertaken are too limited (e.g. only storm-resistant construction with no post-disaster aid), initial damage may be low, but recovery time would be longer with the Path 3. Path 4 shows the least resilient scenario with larger impact and slowest recovery. Therefore, the scope and extent of the adaptation measures define the losses and post-disaster recovery of households. This study examines the factors that contribute toward the static and dynamic resilience of affected households.

Factors determining the resilience to natural disasters

Natural hazards disrupt economic activity and destroy resources, infrastructure, and livelihoods (Das 2016). In poorer communities, this immediate loss of income and assets can push people into poverty (Hallegatte and Dumas 2009; UNISDR 2009). However, communities can bounce back with the appropriate protective measures and post-disaster support systems. Several factors help communities to cope better after a disaster. Income diversification through alternate livelihoods improves the resilience of the vulnerable groups such as women (Akter 2014; Giegerich 2014). Opportunities of employment in infrastructure-building projects can help communities overcome the impacts of disasters (Venn 2012). Migration is another means of resilience that people adopt post-disaster to continue their livelihood and regain lost income (Gray and Mueller 2012). In developing countries, such migration often proves to be long-term and detrimental to the local economy (Drabo and Mbaya 2015) as the skilled and educated workers tend to migrate more than the unskilled. This pattern of migration reduces the region’s capacity of resilience building. Godschalk et al. (1999) argue that measures like retrofitting, reconstruction, and insurance cover can be remodelled as resilience building for the future (Godschalk et al. 1999). Murty et al. (2006) show how the communities affected by the 2004 Indian Ocean tsunami were helped by post-disaster relief and aid from different sources of the state, non-profits, and global community. Extensive coverage by both conventional and social media helped widening the relief aid speeding up the recovery. These factors contributed to the resilience of the affected communities. Factors that negatively impact resilience building include features such as extreme poverty, dependence on private moneylenders, gender inequality, presence of marginalised communities without property rights, lack of communication and warning systems, absence of alternate livelihood opportunities, etc. (Gupta and Nair 2013).

Role of coastal bio-shields in building resilience to extreme climate events

Nature-based solutions to address coastal disasters and build social ecological resilience are mainstream now.5 Coastal vegetation (e.g. mangroves and mixed-species native forests) along with natural infrastructure such as sand dunes provide protection and limit damages from disasters such as tsunamis and cyclones (Danielsen et al. 2005; Das and Vincent 2009; Das and Crépin 2013; Das and Sandhu 2014; NOAA 2017). Given the escalation of both frequency and intensity of natural calamities in coastal regions, there has been extensive discussion on the natural coastal bio-shields (FAO 2007). However, most of the studies examining the role of coastal vegetation during natural disasters are limited to mangroves (Iverson and Prasad 2007; Cochard et al. 2008; Feagin 2008; Das and Vincent 2009; Koch et al. 2009; Das and Crépin 2013). After the 2004 Indian Ocean tsunami, the Food and Agriculture Organization (FAO) and the United Nations Development Programme (UNDP) implemented coastal plantation schemes on the east coast of India (FAO 2007; Mukherjee et al. 2009). Coastal plantation schemes often use monocultures of casuarina (Casuarina equisefolis) trees in non-mangrove habitats. However, there is a little evidence to support casuarina’s effectiveness as a coastal bio-shield compared to the native plant species and natural ecosystems of mixed vegetation (Das and Sandhu 2014). Cashew (Anacardium occidentale), coconut (Cocos nucifera), and palmyra (Borassus flabellifer) are some of the common species in the local mixed forests in the study area. This paper addresses the research gap on the protective services of this vegetation by studying the services that this non-mangrove vegetation provided communities during cyclones Hudhud and Phailin.

Materials and methods

Objectives of study

This article seeks to identify the local factors contributing to both static and dynamic resilience building for the local communities in the chosen study area. Hence, it examines the following set of questions:

1. How did different kinds of households suffer and cope with the impacts of the cyclone?

2. Did households surrounded by thick vegetation cope better?

3. Did the non-mangrove vegetation along the coastline help households during the cyclones?

4. Which type of non-mangrove vegetation provided storm protection and helped in building resilience to cyclones?

Study area

The east coast of India lies along one of the core areas of cyclogenesis, the Bay of Bengal, that witnesses a significant number of cyclones almost regularly (Sahoo and Bhaskaran 2016). This coastline was hit by two severe cyclonic storms, Phailin over 12–13 October 2013 and Hudhud over 12–13 October 2014. Phailin made a landfall at 19.2°N, 84.9°E near Gopalpur town in Ganjam district of Odisha. Hudhud made a landfall at 17.7°N, 83.3°E near Vishakhapatnam city in Andhra Pradesh. Both the cyclones were severe with landfall wind velocities of 230 km h⁻¹ and 200 km h⁻¹ respectively. They caused heavy damage to the agricultural harvest of the season. The study site of this research is located near the landfall point of the cyclone Hudhud. It consists of 15 villages falling under 12 gram panchayats and six Mandals6 of Vishakhapatnam district. These villages were damaged by both the cyclones, but received maximum damage due to Hudhud being close to its landfall. Figure 2 shows the study area and the track of cyclones Hudhud and Phailin. The coastal forest belts observed in this area were patchy and mixed, formed mainly of palmyra, coconut, and casuarina.

Fig. 2.

Study area location and the paths of cyclones Hudhud and Phailin.

Source: mapsofindia.com

We probed all the aforementioned research questions for this area, and additionally, supplemented the results related to the question (d) with some more analysis using data from a Phailin-affected area, which had widespread thick plantations of cashew, casuarina, and other local species along the coast. As the study area lacked diversity in the coastal vegetation, this extra analysis was undertaken to provide additional insight into the role of coastal vegetation types in resilience building. “Additional analysis to examine coastal vegetation effect on resilience to cyclone” section provides the details of this extra analysis.

Data

This paper uses a survey data generated through a structured household survey. The household survey was conducted during June–July 2015, 7 months after Hudhud. It collected the information on damages suffered by the households, time needed to come back to normalcy, and the households’ demographic, economic and locational features (e.g. density of vegetation surrounding the households and the vegetation pattern of the closest coastline). The species type and approximate width (vertical to coast) and length (parallel to coast) of the vegetation patch along the nearest coast were recorded from households as measuring it through satellite images was beyond budget. All households had such knowledge as they lived very close to the coast. The survey instrument used to collect these data is shown under electronic supplementary information.

Methodology

The study uses a combination of methods such as questionnaire survey and econometric analysis that include both univariate and multivariate regression analysis. We followed a multi-stage random sampling to conduct the questionnaire survey. We used high impact from Hudhud and proximity to coast as the first round of identification strategy and selected five coastal mandals (Bheemunipatnam, Visakhapatnam rural, Visakhapatnam urban, Padagantyada and Achutapuram)7 from north and south direction of Visakhapatnam city, the landfall point. Next, we listed all villages of the mandals having 500 or more households and then, selected fifteen villages, seven from each direction and one from Visakhapatnam urban mandal. All such villages were closest to the sea from the list. We took only one coastal village from Visakhapatnam urban due to its large number of households. Highly populated villages were chosen to have wide variation in occupations within the sample. We surveyed 900 households to ensure that the sample represents that occupational diversity.8 The number of households to be surveyed from each village was then decided proportionately. Systematic random sampling was done within the villages. Questionnaire survey was carried out for these systematically selected households. All the selected villages were from the high-impact zone to neutralise the effect of wind velocity while identifying the role of socioeconomic factors on damage occurrences and resilience.

We used STATA software for data cleaning, descriptive analysis, and estimation of multivariate regression models to derive the results. The regression models are estimated at the household level and use multiple control variables along with mandal dummies that control for the aggregative macro features of the mandals. Huber-White robust standard errors are used for test of significance.

Results

Summary findings (Supplementary Information Tables S1 and S2) from the survey show the sample to be predominantly illiterate (76%) with only 9% with no more than primary education, and 13% with no more than secondary education. All the selected households were Hindu, 98% of which belonged to Scheduled and backward castes that are historically disadvantaged in terms of social status and economic security. Nearly, 37% of the households practiced subsidiary occupations, casual labour being the most common.

Table 1 shows the socioeconomic distribution of sample households and self-reported losses (in INR) suffered during Hudhud. The households were either fishers, fish-vendors, salt producers, farmers, or manual labourers. More than 50% of these households practiced either fishing or fish vending; 26% were manual labourers, and 11% engaged in salt production. Farmers were few in the sample, probably due to the scarcity of cultivable land. About 66% of the households were poor with the median annual household income between INR 50 000 (USD 781) and INR 150 000 (USD 2344). When households were grouped as per their occupations, the annual income of all the groups averaged under INR 100 000 (USD 1562) (Column 3). Fishers earned the highest, followed by farmers, drivers, and salt producers, respectively.

Table 1.

Distribution of households based on primary occupation in the study area A

Occupation categories	Percentage of households engaged	Average annual income (INR)	Total self-reported loss due to Hudhud (INR)	Loss as percentage of annual income
Fish vending	12.12	58 144	44 935	0.77
Fishing	40.38	82 155	105 136	1.28
Salt	11.12	68 450	26 697	0.39
Cultivation	3.23	76 931	117 069	1.52
Driver	1.56	68 714	61 083	0.89
Labour (manual)	26.36	57 080	36 319	0.64
Labour (mechanical)	1.22	47 400	26 500	0.56
Unemployed (household heads were either homemakers or unemployed)	1.33	59 667	57 917	0.97
Others	2.67	49 783	27 045	0.54

Damage caused by Hudhud

All occupational groups suffered losses during Hudhud. However, farmers and fishers suffered the worst—their loss exceeded their average annual income (Columns 4 and 5). Salt producers and manual labourers suffered the least. Around 64% of the sampled households were headed by males, and 36% by females. The loss suffered by female-headed households was lesser than their annual income, whereas the loss suffered by male-headed households exceeded their annual income (Fig. 3). Among the types of loss suffered, farmers reported loss of land quality due to saline inundation, and loss of harvest (for crops of rice, brinjal, chilies, onion, drumsticks, coconut, papaya, and banana) due to strong winds and saline inundation.

Fig. 3.

Annual income and loss suffered during Hudhud by male- and female-headed households (in INR)

People also suffered different types of house damage. In the sample, nearly 38% of families had mud-thatched houses, 33.5% had concrete houses, and 28.5% had semi-concrete houses. Among these, 71% of the mud-thatched, 26% of the semi-concrete, and 0.1% of the concrete houses were fully damaged. Table 2 explains the types of damage and shows the number of houses in each category. While 81% of concrete houses had no damage, such houses were just 3% in the semi-concrete and 1% in the mud-thatched categories.

Table 2.

Type of damage suffered by different types of houses

Type of houses	Total number in the sample	Degree of damage suffered
		Fully damaged (the roof needs complete repair and walls need partial repair)	Partially damaged (only roof needs to be repaired, not walls)	Washed away (the house needs reconstruction, as both roof and walls are completely damaged)	Not damaged
Concrete houses	306	3	55	0	248
Semi-concrete houses	260	70	178	4	8
Mud-thatched houses	335	237	73	21	4

Fishers suffered both direct and indirect losses. Direct losses included boats, nets, engines and other fishing materials, and loss of fishing days. The indirect losses were the lower catch rates (reduced sizes and quantities) after the cyclone. Figure 4 shows the pre- and post-Hudhud catch per trip of sampled fishers. Prior to Hudhud, only 13 fishers caught 10 kg or less per trip. After the storm, 141 fishers reported catching that amount per trip. Hudhud caused an increase in the number of fishers reporting lower catch sizes and a decrease in the number of fishers with higher catch sizes.

Fig. 4.

Inshore fish catch per trip before and after Hudhud

Other than the agricultural, house, and fishery-related losses, 82% of the households reported to have lost work days because of the cyclone, varying from less than 10 days to more than 2 months. Figure 5 shows the number of people with their respective number of lost work days. Most people had to sit idle for nearly a month after Hudhud.

Fig. 5.

Post-Hudhud loss of work days

Factors determining resilience to cyclone

In the study area, people were asked how long they would take to recover completely from the loss from Hudhud. Nearly, 86% of households expected to recover within 2 years; while 5% felt that they would take between 3 and 6 years for full recovery. Using this self-reported recovery period as an indicator of resilience, the study probed the household characteristics and environmental features contributing to the quicker post-cyclone recovery. We estimated a multivariate linear regression using self-reported recovery period as dependent variable and a set of explanatory variables related to the household and their surrounding (Table 3). We consider two types of vegetation in the regression model: tree density surrounding the house and tree cover on the coastline closest to the household. Only 14% of the households had tree cover along the closest coastline in the form of mixed forest. Rest of the households faced open coast. Our query was whether this mixed vegetation along the coast provided protection to households compared to an open coastline devoid of vegetation.

Table 3.

Ordinary least squares estimates of regression coefficients (Dependent variable: Years needed by households to recover the loss from Hudhud)

Explanatory variables	Model 1	Model 2
Total income of household (in thousands)	0.01 (1.03)	0.11 (1.02)
Female-headed household	− 0.24*** (3.23)	− 0.30*** (3.52)
Primary-educated household head	0.01 (0.06)	0.00 (0.02)
Secondary or higher-educated household head	− 0.29*** (3.12)	− 0.30 ** (2.82)
Mud-thatched house	0.07 (0.86)	0.09 (0.94)
Concrete house	− 0.19** (2.13)	− 0.23** (2.13)
Tree cover around house (dense)	0.01 (0.05)	0.03 (0.12)
Tree cover around house (scattered)	− 0.09 (1.13)	− 0.09 (1.07)
Whether the nearest coastline has tree cover	0.34 (0.36)	0.002*** (7.18)
Length (spread parallel to coast) of coastal vegetation patch (in km)	0.003*** (2.30)	0.0003 (0.93)
Width (spread vertical to coast) of coastal vegetation patch (in km)	− 0.01*** (2.94)	0.003*** (11.17)
Length of open coast (in km)	0.002*** (8.81)	− 0.01*** (7.58)
Width of open coast (in km)	0.0002 (0.64)	0.32 (0.66)
Bheemunipatnam mandal	1.74*** (10.54)	1.75*** (7.78)
Padagantyada mandal	0.93*** (4.24)	0.94*** (4.24)
Achutapuram mandal	0.678* (1.83)	0.67*** (2.44)
Visakhapatnam urban mandal	0.43 (1.44)	0.41* (1.85)
FemaleXconcrete_house	X	0.16 (1.21)
FemaleXsecondary_edu	X	0.06 (0.26)
FemaleXconcrete_houseXsecondary_edu	X	− 0.01 (0.03)
Constant	− 0.19 (0.65)	− 0.18 (0.52)
Number of observations	688	688
F(20. 667)	15.83, p < 0.001	47.05, p < 0.001
R squared	0.268	0.29
Root MSE	0.831	0.832

Figures in parentheses with estimated coefficients are student t values derived with robust standard errors. The explanatory variables are all dummies except the ones in km or in thousands

‘X’ means the variable is not used in estimation of the model

***p < 0.001, **p < 0.01, *p < 0.05

Three types of households were seen to have short recovery time post-Hudhud (Model 1 shown in Column 2). Those were female-headed households, households having secondary or higher-educated heads, and households with concrete houses. Next, we interacted these households, surmising that they may not be independent and higher-educated women may be living in concrete houses, and estimated Model 2. The results are shown in Column 3. All interaction terms were found insignificant while supporting the results of Model 1.

Among environmental factors, we observed an association between width of coastal vegetation and shorter recovery times. Households behind wider coastal vegetation suffered fewer damages from Hudhud and thus needed lesser time to recover. The vegetation around a house was seen to have an insignificant effect on the recovery period. However, it was seen to have a negative and significant effect when a similar equation is estimated taking the binary variable of ‘whether the household suffered fully damaged houses’ as a dependent variable (see Table S3).

Additional analysis to examine coastal vegetation effect on resilience to cyclone

As the Hudhud affected study area only had sparse, mixed vegetation, we did an additional analysis using the house damage data of some coastal villages of Odisha, which had widespread coastal forests and were severely damaged during Phailin. Secondary data from multiple sources has been used in this analysis. The house damage data were collected from the emergency department of Odisha government. Satellite images were used for generating forest cover and GIS files for village and coastline location. This data were procured in an earlier project of one of the authors of this article and had remained unused. The forest cover map was combined with the GIS polygons on village locations and the coastline (Fig. 6). The width of different forest types between the villages and the nearest coastline was measured using ArcView ArcGIS software 9.0. Tangential wind model9 and parameters of Indian Meteorological Department10 were used to measure the cyclone impact on villages in terms of potential wind velocity. All these data were then used to estimate multivariate regressions models in STATA.

Fig. 6.

Coastal vegetation of study of Phailin-affected region in Odisha

This analysis tests and compares the quality of protection offered by the mixed forests of casuarina, cashew, palmyra and other local species during Phailin. Table 4 shows the ordinary least square coefficients of the multivariate linear regressions. Different regressions are estimated taking different types of house damage11 as dependent variable and the same set of explanatory variables. Village-level information on house quality for the year of 2013, when Phailin occurred, was not available. Thus, we considered explanatory variables such as share of households belonging to the underprivileged castes and communities like scheduled caste and scheduled tribes, education, and the skill-level of workers to approximate the house quality. The results show that wide vegetation between village and coast protected houses from being severely or completely damaged during Phailin, and that plantations of cashew, not of casuarina, provided this service. Other vegetation in the area showed insignificant effects, though width of mixed vegetation showed a negative sign.

Table 4.

Estimated ordinary least squares regressions coefficients (Dependant variable: Number of houses damaged in a village during Phailin)

Explanatory variables	Severely damaged houses	Completely damaged houses	Washed away houses	Partially damaged houses
Adult males in a village (proxy for number of households)	0.04*** (4.39)	0.05*** (4.06)	0.01** (2.19)	0.25*** (6.42)
Share of underprivileged households	3.62 (0.49)	1.88 (0.22)	− 1.74 (0.88)	13.79 (0.60)
Share of literate people in total population	− 32.67 (1.17)	− 45.79 (1.43)	− 13.1* (1.86)	− 24.04 (0.37)
Share of farmers in total population	114.89* (1.93)	122.43* (1.90)	7.54 (0.77)	− 33.38 (0.43)
Share of other full-time workers in total population	84.56 (1.48)	115.44 (1.44)	30.88 (1.25)	− 35.76 (0.29)
Share of workers in household industries in total population	− 64.53 (0.72)	− 64.65 (0.65)	− 0.12 (0.01)	485.52 (1.37)
Share of marginal workers in total population	8.64 (0.33)	14.92 (0.44)	6.28 (0.69)	− 53.06 (0.91)
Approximate wind velocity at the village	0.15 (0.58)	0.16 (0.58)	0.01 (0.36)	0.91 (1.44)
Distance of village from coast	− 0.11 (0.15)	− 0.50 (0.59)	0.39** (2.29)	2.09 (0.91)
Width of casuarina between village and coast	12.66 (1.59)	14.99 (1.42)	2.33 (0.78)	44.61* (1.71)
Width of mixed plantation (of fruit bearing trees like coconut, mango etc.) between village and coast	− 3.25 (0.40)	− 5.51 (0.57)	− 2.26 (1.14)	9.94 (0.46)
Width of cashew plantation between village and coast	− 13.59 ** (2.03)	− 14.85* (1.85)	− 1.26 (0.76)	− 12.29 (0.48)
Width of kewda (Pandanus odoratissimus) plants between village and coast	13.51 (1.35)	12.62 (1.10)	− 0.89 (0.36)	27.67 (0.58)
Ganjam block dummy	3.13 (0.23)	4.34 (0.28)	1.21 (0.42)	143.06*** (3.2)
Khallikote block dummy	− 11.36 (0.53)	− 9.70 (0.40)	1.67 (0.38)	43.17 (0.62)
Rangeilunda block dummy	− 51.27*** (4.38)	− 57.07 *** (4.31)	− 5.80** (2.71)	− 26.93 (0.59)
Constant	− 13.07 (0.28)	− 8.92 (0.16)	4.15 (0.40)	− 213.65 (1.59)
Number of observations	288	288	288	288
F value F (16, 70)	3.61 (p < 0.001)	3.78 (p < 0.001)	1.83 (p < 0.001)	8.18 (p < 0.001)
R squared	0.33	0.32	0.23	0.59
Root MSE	40.36	48.18	11.22	121.05

Figures in parenthesis are student—t values derived with robust clustered standard errors, clustering done at the level of blocks

***p < 0.001, **p < 0.01, *p < 0.1

Discussion

The study analyses household survey data from Hudhud-affected areas in Andhra Pradesh to find out the factors helping households cope better with the cyclones and adding to their resilience. The household survey observed four types of households, which were more resilient— female-headed households, those with secondary school or higher-educated heads of family, those with concrete houses, and the ones located behind thick coastal vegetation. The self-reported recovery times of these households are shorter. High-income households of fishers and farmers suffered more loss and took longer to recover. About 95% of these households in the study area had either no or only primary education. The study surmised that education provided people with more livelihood options and better-informed insights on risk management, and thus contributed to their resilience. Choice of occupation was an influential factor in the household resilience. Similarly, the female-headed households reduced their vulnerability by not engaging in high-risk occupations. To reduce the vulnerability in fishing and agriculture, households can seek insurance and diversify their livelihoods. Earlier research on the socioeconomic factors of east coast’s resilience to natural disasters was mainly focused on macro units (Dasgupta and Shaw 2015). The uniqueness of the present study is in addressing the household-level resilience.

The effectiveness of palmyra-dominant mixed forests as coastal bio-shields during Hudhud was established in the study area, but the forest covers were less diverse and sparse and data on them was collected from households. To overcome this limitation, a supplementary analysis was conducted using secondary data for some Phailin-affected villages with wide and diverse coastal forests. This analysis proved the resilience-building role of cashew plantations. Both palmyra and cashew-mixed forests represent the local flora and provide several ecosystem services to coastal communities, including protection from cyclones. This is probably the first scientific evidence on storm-protection services of these plant species.

Conclusion

The concept of ecosystem-based protection from natural disasters has been examined and established in the past decade for areas very close to the present study sites. Two studies (Badola and Hussain 2005; Das and Vincent 2009) highlighted the protective services of mangrove forests during a super-cyclone of 1999 in Kendrapada district of Odisha. A similar study assessed the protective services of casuarina plantations, mangroves, and mixed-species forests following the same 1999 cyclone in Odisha and proved that the best protection was provided by mangroves, followed by the mixed forests (Das and Sandhu 2014). The present study confirms the protective role of local forests dominated by palmyra and cashew trees during the cyclones Hudhud and Phailin. These findings call for the ecosystem-based resilience-building efforts, which provide more sustainable, cost-effective, and ecologically sound alternative to conventional coastal engineering (Tammerman et al. 2013).

Hence, this study advises coastal managers and planners to develop and strengthen coastal bio-shields, but with caution against monocultures. Extensive casuarina plantations established along the Odisha coastline as storm protection were ineffective in preventing damage during the 1999 super-cyclone. The cyclone uprooted most of the trees near the coast and caused significant damage inland (Das and Sandhu 2014). Similarly, the extensive casuarina belts in Phailin-affected areas did not prevent damage to houses. In addition, casuarina causes harms to the local biodiversity (Feagin et al. 2009) and damages the nesting sites of olive ridley turtles (Lepidochelys olivacea). To prevent this damage, the state government of Tamil Nadu in India ordered the removal of casuarina plantations up to the high-tide line (Chaudhari et al. 2009). Therefore, the future attempts to develop coastal bio-shields should adopt local flora rather than casuarina monocultures. Ecosystem restoration should consider the biodiversity of these coastal habitats along with the intensity and frequency of disaster befalling this area.

Coastal communities, regardless of their financial status, can invest in movable assets that can be relocated to safe zones in times of need. Households in Hudhud-affected areas were given five options of managing the effects of a cyclone and were asked to choose the best way (SI Fig. S1). An overwhelming number of respondents favoured concrete houses, followed by cyclone shelters, and early-warning systems. The central government of India has announced the Pradhan Mantri Fasal Bima Yojana12 to insure households against the crop loss from cyclones. However, only one household in the study area favoured crop insurance, indicating the lack of awareness regarding insurance in these communities. Steps should be taken to raise the capacities of coastal communities to access opportunities and benefits of insurance policies and schemes, smart asset investment, and financial saving plans so that they are better-equipped to cope with natural disasters. Insurance cover for other assets such as fishing instruments should also be considered.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Biographies

Saudamini Das

works as NABARD Chair Professor at the Institute of Economic Growth of Delhi. She specializes in Ecosystem services valuation, Coastal vulnerability and coastal protection from natural disaster, climate change adaptation, and impact evaluation.

Nisha Maria DSouza

used to work as small grant officer with Mangroves for Future program in International Union for Nature Conservation (IUCN) India country office, New Delhi. Presently, she is Director, EcoNiche Consulting.

Author contribution

SD conceptualized the research, SD and ND conducted the research and wrote the paper.