Livelihood challenges of ‘double strike’ disasters: Evidence from rural communities of southwest coastal Bangladesh during the Covid-19 pandemic and cyclone Amphan

Abstract

The Covid-19 pandemic has adversely affected livelihoods in the Global South and exacerbated the effects of natural disaster. This study examines the ‘double strike’ of the Covid-19 lockdown and ‘super-cyclone’ Amphan on the 12 villages of Dakshin Bedkashi Union, southwest Bangladesh. It employs questionnaire surveys to assess the impact of the compound disaster on rural livelihoods over a period of five months, comparing pre-lockdown, post-lockdown, and post-Amphan phases. The results demonstrate the severe impact of the pandemic on income, occupation, and workdays, and consequently on livelihood resilience leading up to Amphan. Although effects vary moderately by livelihood category, village, and socio-spatial characteristics, overall rates of income reduction (58%) and occupation loss (77%) between the advent of the lockdown and the aftermath of Amphan are high. Thus, livelihood resilience to the double strike is generally poor. The study analyses the predictors of lost working days to illuminate the influences on livelihood resilience throughout the double strike. We conclude that the pandemic has presented novel challenges to this region, complicating vulnerability to more common cyclonic natural hazards. Finding that the degree of livelihood diversification is low in the sample, we recommend implementing diversification policies and strengthening local networks and community trust to better anticipate and combat the complex, varied impacts of double strike scenarios in future.

1. Introduction

The Covid-19 pandemic, together with intensifying climate change-related natural disaster, has heightened the potential for concurrent disasters to afflict communities (Shultz et al., 2020). Geographically and temporally overlapping hazards may have complex and cascading implications for disaster effects and response, compounding stresses, complicating effective management, exacerbating existing inequalities and vulnerabilities, and protracting recovery times (Salas et al., 2020; Cardil, and de-Miguel, S., 2020; Quigley et al., 2020; Shultz et al., 2020; Phillips et al., 2020; Mishra, Bruno & Zilberman, 2020). Rahman et al., 2020a, Rahman et al., 2020b forecast that future compound disasters are likely to have dire consequences for Bangladesh, jeopardising the economic advances the country has made in recent decades. This paper assesses the impact of the ‘double strike’ of the Covid-19 pandemic and ‘super-cyclone’ Amphan, one of the major natural disasters to coincide with the pandemic, on rural communities in southwestern Bangladesh. It quantifies their impact on livelihoods, conceptualising the successive events as one ‘double disaster’ with aggregating effects on household income and occupation. Furthermore, by analysing the relevance of spatial attributes and resource dependency to livelihood maintenance during the period of emergency, it suggests how rural livelihoods may be better fortified against future multi-hazards in this vulnerable region.

On 11th March 2020, the World Health Organization (WHO) declared the novel coronavirus outbreak a pandemic, urging governments to act swiftly against the rapidly spreading virus (Cucinotta and Vanelli, 2020). The first cases of Covid-19 in Bangladesh were reported on March 8th 2020 by the Institute of Epidemiology, Disease Control and Research (IEDCR). A national lockdown was enforced on March 26th 2020, with multiple extensions until the end of May 2020 (Anwar et al., 2020). The Covid-19 response raised several challenges for Bangladesh, a densely populated country, including concerns of social distancing, efficacy of testing, shortage of testing kits and centres, poor health infrastructure, and poor coordination between authorities (Anwar et al., 2020; Rahaman et al., 2021). Reduced economic activity greatly impacted livelihoods and food security, particularly among low-income informal workers (Zabir et al., 2021; Ruszczyk et al., 2020). 55 days into Bangladesh's lockdown, Cyclone Amphan struck Bangladesh and India, at around 4 pm local time on May 20th 2020. Developing rapidly from a depression to a super cyclonic storm within 24 h, it was designated as the first ‘super-cyclone’ (category 5 tropical storm) of the century to arise in the Bay of Bengal (Vinoj and Swain, 2020). Per government estimates, Amphan incurred a death toll of 129, and caused US$13.7 billion and US$130 million in damages in India and Bangladesh, respectively, making it the costliest storm recorded in the North Indian Ocean (Sud and Rajaram, 2020; GoB and UN, 2021). In Bangladesh, 10 million people were affected. Across nine highly impacted districts, 330,667 houses were damaged, of which 55,667 were destroyed (GoB and UN, 2021). Impacts on agriculture and aquaculture were also extensive (Parvez, 2020). Destruction of safe water and sanitation facilities led to disease outbreaks (Rafa et al., 2021).

Covid-19 precautions modulated the management of the disaster, particularly evacuation efforts. Among the protocols activated by the disaster was the HCTT Contingency Plan 2020 for Climate-Related Disasters in the COVID-19 Pandemic Context, prepared by the Government of Bangladesh in coordination with the UN (GoB and UN, 2021). Precautionary measures reduced cyclone shelter capacity. To compensate, authorities increased total shelters from 5000, the number reportedly used in the previous major cyclone, to over 12,000, enlisting facilities like mosques and schools as additional shelters (Syfullah, 2020). 2.2 million people were evacuated following social distancing measures; without these restrictions, shelter capacity would have been 5.1 million (bdnews24, 2020). The cyclone intensified the crippling effects of lockdown on vulnerable populations, particularly rural households (Ober, 2020). The predominantly rural population in cyclone-affected districts was already swollen as migrant labourers returned home from cities, lacking work during lockdown; the reduced urban income thus affected the capacity of village households to recover from cyclone damage (Savage and Farhad, 2020). Moreover, the cyclone likely heightened the psychological anxiety already raised by the pandemic (Islam et al., 2020) and linked to the resource-constrained context of much of the Bangladeshi populace (Shammi et al., 2020).

In light of their temporal overlap and their cumulative impacts, this study conceptualises the twin perils of the pandemic and the cyclone as a ‘double strike’ on rural households, with the view that this holistic perspective can best account for the complex, prolonged effects of overlapping disasters on vulnerable livelihoods. The study employs field data from the southwestern coastal region of Bangladesh, quantifying livelihood impacts over the period of emergency. To investigate the effect of the double strike on livelihood patterns, it explores: (1) the impact of the double strike on income and working days, (2) the changes in occupation it precipitated, and (3) the effect of spatial attributes and resource dependency on the severity of impacts.

2. Methods and materials

2.1. Conceptual approach: the pandemic, disaster, and livelihood nexus

Disease outbreaks in the Anthropocene are more numerous than before and likelier to reach a globalised distribution (Morand and Walther, 2020). The interconnectedness of today's world increases the risks posed by pandemics (Hirschfeld, 2020). The concomitant rise in the frequency and severity of climate-related hazards (Shultz et al., 2020) has yielded a spate of grievous natural disasters that have overlapped with the Covid-19 pandemic. These ‘double strikes’, seen as a portent for the future, have complicated disaster response and extended recovery, such as when mobility restrictions delayed disaster relief packages to Vanuatu following Cyclone Harold (McGarry, 2020). At the household level, they have prolonged the adverse health effects of the pandemic and impaired livelihood resilience, as in Honduras, where successive hurricanes Irma and Eta destroyed family businesses already shuttered for months under lockdown (Tucker and Pindado, 2020). Like single disasters, multi-hazards deepen existing inequalities (Phillips et al., 2020). Natural resource dependence and socio-political context exacerbate vulnerability; thus, the intersecting challenges double strikes present are felt most keenly by marginalised households. For example, poor, fishery-dependent Fijians identified ‘financial stress’ as the major anxiety experienced under the dual impact of Covid-19 and Cyclone Harold (Mangubhai et al., 2021).

A livelihood comprises a household and the capabilities and resources employed to meet its needs (Scoones, 1998), while a ‘sustainable livelihood’ comprises strategies and resources that can cope with economic or environmental duress, enduring through generations (Chambers and Conway, 1992). Livelihood strategies are thus critical to a community's vulnerability to disaster. Natural resource dependence is linked with greater sensitivity to environmental stress, making resource dependent livelihoods more susceptible to external disruptions (Kumar and Yashiro, 2013), especially events that damage key resources (Biswas and Mallick, 2020). Environmental threats narrow livelihood options, often exacerbating existing ramifications of resource exploitation (Rahman et al., 2020c). Recurrent shocks push communities into long term poverty traps (Carter et al., 2006). Therefore, improving livelihoods is the most sustainable route to bolstering disaster resilience (Le Dé et al., 2018). Diversification presents a prime strategy as “diverse rural livelihoods are less vulnerable than undiversified ones” (Ellis, 2000, 1). People are often forced to diversify their livelihoods after disaster as their occupations become insufficient to sustain survival and recovery (Eakin et al., 2012). Diversification as a deliberate adaptive strategy entails the expansion of a community's livelihood portfolio to enhance security against disruptions and multiply opportunities to raise the standard of living (Ellis, 1998).

In Bangladesh, climate change stressors have strongly influenced rural livelihood patterns in recent decades. Livelihood loss from disaster has increased resource dependence (Ahsan et al., 2017) and concentration of industry and services away from rural areas has encouraged reliance on natural resources (Etzold and Mallick, 2016). In climate-vulnerable riverine islands, poverty traps and socio-spatial attributes like poor access to services restrict livelihood options; however, some residents pursue diversification, typically in the form of labour migration (Sarker et al., 2020; Sarker et al., 2019). Diversification is shown to improve livelihood sustainability, enabling people to stay in place despite environmental risks (Mallick et al., 2020). Given the increasing incidence of compound disasters, it is crucial to understand their effects in specific place and dependency contexts. Barring loss of life, day-to-day livelihood burdens are the major consequence of natural disasters and pandemics alike. Therefore, we investigate the Covid-19 pandemic and Cyclone Amphan as a ‘double strike’ on rural livelihoods in Bangladesh.

2.2. Study area

The field survey was conducted in the Dakshin Bedkashi Union of Koyra Upazila, Khulna District, southwestern Bangladesh. The union comprises 12 villages: Boro Angtihara, Choto Angtihara, Choramukha, Dakshin Bedkashi, Gharilal, Golkhali, Holudbunia, Jorsing, Patakhali, Matiabhanga, Binapani and Meder Char (Fig. 1 ). It borders the world's largest mangrove forest, the Sundarbans, and is encircled by the rivers Rai and Kopothako. The total population is around 16,755 people, across 3881 households, within 110,452 acres of land (BBS, 2011). The area has a lack of urban amenities, formal employment opportunities, and primary healthcare facilities. It is poorly connected with transport systems. The literacy rate is lower than the national average of 75% (World Bank, 2021). The housing pattern and structure are traditional, exhibiting low resilience against cyclone or flood disasters (BBS, 2011). Due to frequent disasters, livelihood dynamics in this coastal area are continually shifting and susceptible to disruption. Notably, the land-use pattern has shifted in recent decades from rice cultivation to more labour-efficient shrimp-farming, to cope with precarity (Biswas and Mallick, 2020). Migration is also extensively practised as a coping strategy, usually in the form of seasonal labour migration (Mallick, 2019; Rahman and Gain, 2020). Following recent diversification, day labour, shrimp-farming, fishing, and mangrove resource extraction are the most common occupations in the villages (Biswas and Mallick, 2020).

Fig. 1.

Study villages in southwest Bangladesh.

2.3. Sampling and data collection

Dakshin Bedkashi Union was randomly selected as the field survey site among seven unions in Koyra Upazila severely affected by Cyclone Amphan. We applied a simple random sampling method to the union's 12 villages to select respondents for face to face interviews. The sample comprised 327 households, 8.42% of the union total. The study villages were visited prior to data collection to assess their conditions and ensure the relevance and robustness of the designed questionnaire. Data was collected from 7th August to 9th September 2020 with the assistance of trained data collectors (see, Fig. 2)). We collected a total of 327 structured interviews from the 12 villages (Table 1 ), using the Kobocollect online data collection tool. The survey was administered to household heads. In their absence, the next family member was interviewed. Each interview lasted 15 to 20 min.

Fig. 2.

Timeline of double strike events.

Table 1.

Total population and sample size of the study communities.

Village	Total Population	Total Households	Sample Size	Percent
Boro Angtihara	2360	520	20	6.12
Choto Angtihara	724	195	22	6.73
Binapani	1435	357	20	6.12
Choramukha	762	183	24	7.34
Dakshin Bedkashi	1497	351	38	11.62
Gharilal	1396	312	27	8.26
Golkhali	1895	386	20	6.12
Holudbunia	361	105	21	6.42
Jorsing	3345	803	62	18.96
Patakhali	844	199	22	6.73
Matiabhanga	1483	325	25	7.65
Meder Char	653	145	26	7.95
Total	16,755	3881	327	100

Alongside the household's contextual attributes, the survey concerned the changes and losses incurred from the double strike over the preceding months. The temporal variation of the surveyed period was 148 days, from the onset of the Covid-19 pandemic in Bangladesh (i.e. the first reported case) to the advent of the field study. The survey distinguished three phases: the period before the Covid-19 lockdown (Phase I), the period before Amphan, during lockdown (Phase II), and the final period after Amphan, until the field survey (Phase III). As the field survey was undertaken 80 days after Cyclone Amphan struck, and over two months after the end of lockdown, a key limitation is that it relies on respondents' recollection of personal experiences during a period of emergency and upheaval.

2.4. Data preparation and analysis

The data was analysed to examine the degree of income and occupation loss by village and occupation during the Covid-19 lockdown (Phase II) and the post-Amphan phase (Phase III), as compared to before lockdown (Phase I) (See Fig. 2). Further, we employed regression models to analyse several variables (see Table 2 ) as predictors for the loss of working days in each phase. We also analysed the factors influencing resilient livelihoods, i.e. no loss of working days despite the double strike. Spatial variables like distance to the market were calculated using GIS in the software QGis 3.10, applying the tool distance to the nearest hub. Elevation from mean sea level was calculated using the digital elevation model (DEM) of the region and the point sampling tool. Currency conversions from BDT to USD were calculated on 2nd June 2021.

Table 2.

Description of variables employed in regression analysis.

Variable	Unit of measurement	Description
Dependent variables:
•Number of working days lost in Phase II	Days	Total number of days people could not work due to the Covid-19 pandemic, before Amphan.
•Number of working-day lost in Phase III	Days	Total number of days people could not work after Amphan, while the pandemic persisted.
•No lost working days in either phase	Dummy	Binary variable expressing the working status of the household heads in both phases. If they were able to stay in their occupations in both phases, it was coded as 1 = Yes, otherwise, 0 = No.
Independent variables:
•Age (x1)	Years	Age of the household head.
•Education (x2)	Years	Years of schooling completed by the household head.
•Practice seasonal migration(x3)	Dummy	Binary variable denoting whether the household head has practiced seasonal migration to supplement their livelihood, coded as 1 = Yes and 0 = No.
•Multiple occupations (x4)	Dummy	Binary variable indicating whether the household is involved in more than one occupation or not, coded as 1 = Yes and 0 = No.
•Height above mean sea level (x5)	Metres	The elevation of the household residence from the mean sea level.
•Distance to market (x6)	Kilometres	Distance of the nearest market from the residence.
•Distance of embankment (Z1)	Kilometres	Distance at which the interviewed household heads placed the nearest riverbank.
•Occupations (Z2)	Categorical	The principal livelihood source of the household heads.

To assess the factors influencing loss of working days in Phase II and III, we employ two Poisson regression models. This model is suitable as the dependent variables (number of lost days) are measured in a manner where variance does not exceed the conditional mean, and there are no negative and excess zero values. To identify the predictors of resilience among households who indicated no workday loss, we use a logit regression model, as this dependent variable is dichotomous. The three models employ the same set of regressors (see Table 2). We checked the severity of multicollinearity among the regressors and found that the mean Variance Inflation Factor (VIF) of the models are all lower than 1, indicating no extreme collinearity. To reduce the heteroscedasticity among the villages, we employed cluster analysis using the statistical software STATA14. In our analysis, we investigate the interaction effect between the household occupation and distance from the embankment on the outcome variables, supposing a relation between the two. It was assumed that households nearer to the river were likelier to depend on its resources through fishing, extracting mangrove resources, etc. The equations of the Poisson and logit regression models are presented below. In the Poisson model, y _t indicates the number of lost working days. In the logit model, y _t indicates whether an occupation persisted during the double strike. ψ ₀ is the intercept, ln (p _i /1-p _i) is the log odds ratio, and ε _i is the error term. X _ji , Z ₁ and Z ₂ represent the regressors, and b ₁ b ₂ is an interaction term.

Poisson regression model:

$$E\left(y_t|X_t\right)=\mathit{exp}\left({\sum }_{j=1}^m{\psi }_{\mathit{ji}}{X}_{\mathit{ji}}+Z_1{\psi }_7{b}_1+Z_2{\psi }_8{b}_2+Z_{1i}{Z}_{2i}{\psi }_9{b}_1{b}_2\right)$$ (1)

Logit regression model:

$$L_i=l_n\left[\frac{P_i}{1-P_i}\right]={\psi }_0+\sum _{i=1}^m{\psi }_{\mathit{ji}}{X}_{\mathit{ji}}+Z_{1i}{\psi }_7{b}_1+Z_{2i}{\psi }_8{b}_2+Z_{1i}{Z}_{2i}{\psi }_9{b}_1{b}_2+{\epsilon }_i$$ (2)

Where,

J = 1, 2, 3, ……m; i = 1, 2, 3, …..n

3. Results

3.1. Household profile

3.1.1. Demographic information

The total population of the study area is 16,755, with our sample representing around 8.5% of inhabitants. The average household size is 4.26 persons. Approximately 34.2% of the population are children (≤14 years) and 5.2% are elderly (≥64 years), with the remaining 60.6% of active working age (BBS, 2011). The gender distribution is 55% male and 45% female. Most respondents come from Muslim households (72.17%), while 27.83% are Hindu. Respondents range from 17 to 80 years of age, with a mean age of 42. They have completed five years of education on average, indicating most have only primary education. 79.51% reported that their households did not practice migration prior to the Covid-19 pandemic, whereas 20% did. 70.95% claimed they did not aspire to migrate while 29.05% did, suggesting most residents prefer alternative strategies that bolster the resilience of their current homes. Just 25.9% of the respondents were engaged in multiple occupations, suggesting low levels of resilience capacity.

3.1.2. Socio-spatial pattern

The socio-spatial characteristics include elevation of the homesteads and the distance from each household to the nearest markets, education facilities, health facilities, embankment, and mangrove area (see Table 3 ). The mean elevation of the households in the study area is 2.05 m above mean sea level. The mean distance to the nearest embankment was 0.352 km, and the maximum distance was 1.24 km. The mean distance to the mangroves was 1.8 km, and the maximum distance was 3.83 km. Mangroves were located to the south, west and east of the study area. Education facilities were the most accessible local services, at a mean distance of 0.68 km. Meanwhile, the nearest market and health facilities were on average 1.23 km and 2.24 km away, respectively.

Table 3.

Socio-spatial characteristics of the homesteads.

Variable	Obs	Mean	Std. Dev.	Min	Max
Elevation (m)	327	2.058	1.594	0	4
Distance from embankment (km)	327	0.352	0.335	0.026	1.242
Nearest market (km)	327	1.227	0.702	0.053	2.416
Nearest educational facility (km)	327	0.683	0.347	0.161	1.537
Nearest health facility (km)	327	2.242	0.834	0.08	3.955
Distance to mangroves (km)	327	1.812	1.104	0.18	3.833

3.2. Effects on income and working days

3.2.1. Income-expenditure

The relationship between the income and expenditure of the respondents contributes to their resilience capacity against the double strike. In Phase I, prior to the Covid-19 lockdown, the mean monthly income of the respondents was 9082 BDT (approximately US$107), with mean expenditure being 73.5% of the total income. Following lockdown, the mean income was initially reduced by 53% to 4285 BDT (US$50) in Phase II. In Phase III, after Cyclone Amphan, it was further reduced to 3896 BDT (US$45), a reduction of 58% from the base income reported in Phase I. This represents a high economic impact for the communities.

Despite the significant reduction in household incomes, monthly expenditure was not similarly reduced. By Phase III, expenditure remained at 67% of the average Phase I income, although present incomes had dropped, on average, by half. This implies that most households sought alternative strategies to recoup the economic loss, like access to credits and savings. In Phase II, 44.3% of respondents reported that they drew on savings, and 38.4% took credits. In Phase III, 57.8% turned to their savings and 41.3% to credits.

By Phase III, income had fallen greatly for all occupations compared to the slightly reduced rate of expenditure (Fig. 3 ). Among the occupation categories, mangrove dependent livelihoods were most affected with an average income reduction of 8526 BDT (US$100). This corresponds with the high rate of occupation loss, which saw 94% of mangrove dependent livelihoods experience loss between Phase I and Phase III. Trading and other occupations also experienced a high income reduction of 5540 BDT (US$65) on average, linked to the rate of occupation loss (89%). Day labour, the category with the highest rate of occupation loss, experienced an average income reduction of 5170 BDT (US$61).

Fig. 3.

Changes in monthly income and expenditure between Phase I and Phase III.

3.2.2. Working days

Most respondents experienced a loss in working days in Phase II and Phase III, however, these losses showed notable variation by village and occupation category. In Phase II, the most affected occupation was day labour, with a mean loss of 45 days overall. At the village level, day labourers lost a maximum average of 55 working days, with labourers in the villages of Binapani, Holudbunia, Boro Angtihara and Meder Char losing at least 50 days on average throughout the period (Fig. 4 ). This was followed by mangrove dependent livelihoods, with a mean loss of 37 working days. The villages of Choto Angtihara, Binapani, Choramukha, Patakhali, Mathia Vanga and Boro Angtihara were particularly affected, all losing 55 days on average. The third most affected occupation category was small business and others, losing an average of 35 working days, followed by fishing, with 29 days. The least affected occupations were shrimp-farming, with a mean loss of 4 working days, and farming, with 11 lost days.

Fig. 4.

Number of days of no work across the villages and livelihood categories during Phase II.

Overall, the average number of working days lost in Phase III showed a slight increase from Phase II. There was greater variation by village, suggesting the locational impact of Amphan compared to the broader effects of the lockdown alone. The more affected livelihood categories remained day labour and mangrove dependent livelihoods, with a mean value of 45 and 38 days of lost work, respectively. Fishing also remained highly affected with an increase to 32 lost workdays. Notably, the number of lost working days for mangrove dependent activities increased greatly following Amphan in the villages of Choto Angtihara, Choramukha, and Patakhali, which faced more than 60 lost days in Phase III (Fig. 5 ). Traders and others lost an average of 34 days of work, but in two villages (Choto Angtihara and Choramukha) the number increased sharply from Phase II, reaching 90 lost workdays, the highest documented in the survey. Although shrimp-farming remained comparatively less affected, Amphan appeared to have a greater influence on it than the pandemic, as the average loss doubled to 8 days in Phase III. The average workdays lost by farmers also increased to 15.

Fig. 5.

Number of days of no work across the villages and livelihood categories during Phase III.

3.3. Effects on occupation and livelihood activities

3.3.1. Occupation loss

The livelihood activities most commonly performed as a main occupation in the study area before the double strike were day labour (33.7%) and shrimp-farming (20.5%), followed by resource extraction from the mangroves (15.2%), trading and other occupations (11.3%), farming (10.7%), and finally, fishing (9.4%) (see Table 4 ). All respondents reported having an occupation in Phase I prior to the pandemic, however, by Phase II, 70.64% reported that they had lost their occupation. Following Cyclone Amphan, this increased to 77.06%.

Table 4.

Distribution of respondents by occupation across the double strike phases.

	Phase I (%)	Phase II (%)	Phase III (%)
No occupation	0	70.64	77.06
Day labourer	32.72	0.92	0.31
Fishing	9.48	1.53	0.31
Mangrove dependent	15.29	0.92	0.92
Small business and others	11.31	2.14	1.22
Shrimp-farming	20.49	17.74	15.29
Farming	10.70	6.12	4.89
Total	100.00	100.00	100.00

After the Covid-19 lockdown, the proportion of people engaged in day labour decreased from 32.7% to 0.92%, representing the greatest loss of occupation for any livelihood activity across either phase. Mangrove-dependent activities, fishing, and small trading also experienced a high rate of occupation loss in Phase II, while all other occupations experienced smaller reductions. The greatest losses were incurred in this period as nearly two thirds of the respondents lost their main occupation between Phase I and Phase II. In Phase II, all occupations barring mangrove dependent activities experienced further, albeit minor, losses, raising the total proportion of respondents with no occupation from 70.64% to 77.06%. Overall, the occupation most likely to be lost under the double strike was day labour, followed by mangrove dependent activities, fishing, and small business. The occupations comparatively less likely to be lost were shrimp-farming and farming.

3.3.2. Occupation shift

Occupation shift refers to a change in the primary livelihood activity between phases rather than an outright loss (see Table 5 ). While most respondents experienced either loss or no change in their occupation following the Covid-19 lockdown, a minority of respondents (2.14%) reported a shift in occupation in Phase II. This was sustained in Phase III, where otherwise some further occupation loss occurred. However, at 2.14% of the sample, the rate of occupation shift was not representative.

Table 5.

Shift in occupation across phases.

	Phase II		Phase III
	Freq.	Percent	Freq.	Percent
No change in occupation since Phase I	89	27.22	68	20.80
Changed occupation	7	2.14	7	2.14
No occupation	231	70.64	252	77.06
Total	327	100	327	100

3.4. Predictors of loss of working days

The regression models employed in this study analyse the socio-demographic and socio-spatial characteristics of the households and their changing livelihood situation to determine the predictors of working day loss under the Covid-Amphan double strike. In particular, these models examine the interaction between the locational variable distance from embankment/river, and different occupations, as this relates to resource dependence. As the most represented occupation, day labour is used as the reference category. Table 6 displays the key findings of the regression analysis.

Table 6.

Summary of regression models.

Predictors	Poisson regression model-1 Number of working days lost in Phase-II (COVID)	Poisson regression model-2 Number of working days lost in Phase-III (COVID + Amphan)	Logistic regression model Never lost working days in either phase (Yes = 1, No = 0) (26.3%)
Age of the household head	0.000291	0.00164	−0.0272
	(0.00210)	(0.00292)	(0.0214)
Years of schooling	0.0231⁎	0.0196	−0.125⁎⁎⁎
	(0.0125)	(0.0134)	(0.0323)
Seasonally migrate? (Yes = 1, No = 0)	0.0742	−0.0772	−0.683⁎⁎
	(0.0802)	(0.0833)	(0.269)
Multiple occupations? (Yes = 1, No = 0)	0.223⁎⁎⁎	0.0232	−1.124⁎⁎
	(0.0692)	(0.0526)	(0.545)
Height above mean sea level (m)	−0.00416	−0.00552	0.326⁎
	(0.0212)	(0.0199)	(0.166)
Distance from nearest market (km)	0.0938	0.103⁎	−0.212
	(0.0695)	(0.0592)	(0.455)
Distance from embankment (km)	−0.00320	0.114	0.529
	(0.0857)	(0.134)	(1.409)
Occupation: Day labour (reference category)			–
Occupation: Fishing	−0.205⁎⁎⁎	0.0799
	(0.0758)	(0.149)
Occupation: Mangrove dependent	0.0510	0.147	0.508
	(0.101)	(0.117)	(1.134)
Occupation: Small business and others	−0.588⁎⁎⁎	−0.544⁎⁎	4.149⁎⁎⁎
	(0.120)	(0.231)	(1.242)
Occupation: Shrimp-farming	−1.934⁎⁎⁎	−1.015⁎⁎⁎	4.056⁎⁎⁎
	(0.415)	(0.378)	(1.136)
Occupation: Farming	−1.626⁎⁎	−1.617⁎⁎⁎	4.384⁎⁎⁎
	(0.661)	(0.541)	(1.638)
Occupation: Day labor ⁎⁎Distance from embankment (km) [reference category]
Occupation: Fishing ⁎⁎Distance from embankment (km)	1.357⁎⁎⁎	−0.0409
	(0.406)	(0.649)
Occupation: Mangrove dependent ⁎⁎Distance from embankment (km) [reference category]	−0.0162	−0.101	−3.061⁎
	(0.164)	(0.265)	(1.777)
Occupation: Small business ⁎⁎Distance from embankment (km) [reference category]	0.674⁎⁎⁎	0.0502	−13.03⁎⁎
	(0.238)	(0.486)	(5.209)
Occupation: Shrimp-farming ⁎⁎Distance from embankment (km) [reference category]	−2.565⁎	−3.299⁎⁎⁎	2.769
	(1.496)	(1.225)	(2.133)
Occupation: Farming ⁎⁎Distance from embankment (km) [reference category]	1.067	1.109⁎⁎	−2.310
	(0.754)	(0.554)	(1.587)
Constant	3.544⁎⁎⁎	3.509⁎⁎⁎	−1.634
	(0.259)	(0.236)	(1.865)
Observations	327	327	296

Robust standard errors in parentheses

^⁎⁎⁎p < 0.01.

^⁎⁎p < 0.05.

^⁎p < 0.1

In Model-1, the dependent variable is the number of working days lost in Phase II, between the advent of the Covid-19 lockdown, and Amphan. The model finds that respondents with more than one occupation lost more working days. People with more years of education also lost more days. Compared to day labourers, fishers, traders, shrimp-farmers, and farmers were less likely to lose working days; that is, directly resource-dependent occupations lost fewer working days compared to independent occupations. Likewise, although distance from embankment does not have a significant influence on lost working days overall, fishermen living close to the embankment lost fewer working days than day labourers also living near the embankment. Similarly, a small-business trader residing near the embankment was likely to lose fewer working days than similarly located day labourers. In contrast, shrimp-farmers living close to the embankment lost more working days than a similarly located day labourer. Model-1 shows that the age of the respondent, residence elevation, distance from the nearest market, and seasonal migration practice did not have a significant influence on workday loss during Phase II.

The dependent variable of Model-2 is the number of working days lost in Phase III, under the double strike of both Covid-19 and Amphan. In other words, this is the number of working days that those who experienced losses required to bounce back to their present occupation. Respondents who lost their occupation after Amphan, as opposed to during Phase II, were likelier to be living on or near the embankment. Distance from the market was significant during Phase III, with respondents closer to the market losing fewer workdays. As with Phase II, shrimp-farmers, farmers and traders were likely to lose fewer workdays than day labourers. Shrimp-farmers living close to the embankment lost fewer working days than day labourers living close to the embankment. In contrast, farmers located near the embankment lost more workdays than similarly located day labourers. All other variables such as age, education level, seasonal migration practice, multiple occupations, and elevation had no significant influence on working day loss in Phase III. However, residences closer to the embankment have a higher elevation above mean sea level compared to those inland (r = −0.19, p < 0.05); living farther from the embankment, with a lower elevation, was associated with requiring more days to bounce back after Amphan. Overall, results indicate that exposure level varies across the livelihood categories.

Model 3 analyses the predictors for always having work opportunities despite the pandemic and cyclone, i.e. a resilient livelihood. It examines respondents who did not lose working days in either phase. Respondents had a greater chance of losing working days if they had more years of education. This is because those with secondary education were more likely to be employed in formal sectors that lost official workdays during lockdown. People who practice seasonal migration were also more likely to lose workdays, as they could not venture beyond their village due to travel restrictions. Overall, having multiple occupations was associated with workday loss. Meanwhile, those residing at a higher elevation had a lower chance of losing workdays. On the whole, shrimp-farmers, traders, and farmers were less likely to lose workdays despite the double strike, as compared to day labourers. They are more highly represented among people who lost no workdays. However, for those living further inland, a mangrove dependent or trading occupation meant a greater chance of losing workdays as compared to day labourers.

Overall, livelihoods related to fishing, mangrove dependence, day labour and (depending on distance from the embankment) trading were more susceptible to losses under double strike conditions, whereas shrimp-farming and agriculture were less vulnerable. Here it is important to note that the trading category encompasses the minority of people involved in hatchery, crabs, vegetable selling, honey collection, etc. Thus, many of these livelihoods are dependent on other major occupations like fishing, agriculture and collecting mangrove resources.

4. Discussion

Climate-related dangers to livelihoods, particularly in disaster-prone, economically vulnerable regions, are compounded by pandemic conditions (Steenbergen et al., 2020). The Covid-19 pandemic disproportionately burdens marginal livelihoods in the Global South, interacting with pre-existing scarcities to imperil vulnerable livelihoods (Phillips et al., 2020; Mishra, Bruno & Zilberman, 2020). The high rates of livelihood loss and income reduction demonstrate an overall low resilience to the compound disaster of Covid-19 and Cyclone Amphan among the rural livelihoods of southwestern coastal Bangladesh. Approximately seven out of ten respondents experienced occupation loss, and average incomes shrunk by over half. Although the region is accustomed to weather-related calamities, the pandemic has posed new challenges to livelihood strategies while complicating disaster management during the cyclone season. Importantly, the co-occurrence of the pandemic and the historic super-cyclone Amphan has brought into sharp relief the ‘complex temporalities’ of disasters (Ruwanpura, 2008): the initial months of lockdown rendered rural livelihoods ever more precarious, shaping their vulnerability to Amphan. By the time Amphan struck, the resilience capacity of the locals in Dakshin Bedkashi Union had been severely weakened, precipitating the further injuries caused by the disaster.

Although both occupation loss and income loss rose in the aftermath of Amphan (Phase III), particularly for aquaculture and agriculture livelihoods, our analysis found that the far greater proportion of the losses followed the Covid-19 lockdown. Under the unlikely pandemic conditions, people found themselves without recourse. Furthermore, resource dependent livelihoods appeared to be slightly less affected in Phase II than independent ones, suggesting that whilst the pandemic brought fewer disruptions to the natural environment, the service sector was highly impacted. Effects of lockdown on day labour and trading mirrored the high levels of disruption to the informal service sector across the country (Mujeri and Mujeri, 2021). Across both phases, proximity to the embankment influenced livelihood vulnerability. Households closer to the river, likelier to depend on extraction, generally experienced fewer losses; again, this may owe to the relative stability of natural resources under lockdown. Embankments and polders also remain an important buffer in vulnerable areas (Nath et al., 2019). As they are built on higher ground than homesteads inland, this affirms the positive association between elevation and resilience as well. Again, proximity to the market corresponded with fewer lost workdays in Phase III, demonstrating the significance of socio-spatial connectedness. Major response and relief activities after Amphan took place at markets; therefore, people living nearby had prime access to governmental and NGO intervention programmes to help them recover their livelihoods more quickly. This, along with the interdependence of the local informal economy, gestures to the important role social connectedness can play for resilience in the study area.

The high livelihood-related losses under the double strike underscore the urgency of improving resilience. Few households (2.14%) reported shifting occupations in the face of pandemic hardships, reflecting the low level of diversification previously observed in climate-vulnerable riverine communities (Sarker et al., 2020). Notably, the mode of diversification often employed by these households, labour migration to urban centres (Sarker et al., 2020; Mallick et al., 2020), was directly impeded by lockdown travel restrictions. Thus, having multiple occupations was correlated with workday loss in Phase II, despite its positive association with resilience in the literature. Livelihood diversification nevertheless presents a potent, underutilised strategy that can transform resilience capacity in this region (Biswas and Mallick, 2020). For resource dependent livelihoods, the danger of food insecurity in particular makes diversification a key strategy to mitigate the impact of disaster (Uddin, 2012). We emphasise that, as a broad, many-pronged strategy by definition, diversification is essential for strengthening livelihoods under double strikes like the Covid-Amphan disaster, which induce a range of sometimes contrasting and unpredictable effects.

In our study, shrimp-farming and agriculture emerged as the most resilient livelihood options overall. Shrimp-farming, while profitable, adversely impacts long term vulnerability (Nath et al., 2019). However, investment in sustainable agriculture presents an avenue for diversification. One possible strategy is floating agriculture, an indigenous method that can alleviate burdens in the long periods of waterlogging that follow natural disasters (Chowdhury and Moore, 2017). As a nature-based solution, it is a potentially highly fruitful route to strengthening resilience in a resource-dependent region. Mangrove afforestation is another strategy that has seen success in nearby regions. Mangroves nurture livelihood sources in the form of non-timber forest products (NTFPs) and provide a natural defence against weather hazards, with careful management proven to improve their disaster-resistance (Nandy and Ahammad, 2012). Still, nature based solutions alone are not enough under compound disaster pressures. Community cooperation is as valuable as diversification, especially as technical solutions to scarcity often remain untapped due to poor coordination (Pukinskis, 2015). In the case of fishing, for example, community-based fishery management schemes proposed by Islam et al. (2011) might improve livelihood outcomes.

Finally, while the economic effects of Covid-19 have begun to be documented, the cultural and socio-political dynamics of rural communities under the pandemic merit investigation, including the role of communications and media. Following Cyclone Aila in 2009, diversification and community risk sharing were core recovery strategies. Social capitals were found to contribute significantly to the means of diversification, yet socio-political issues of access presented the main hurdle to recovery (Masud-All-Kamal, 2013). Therefore, we posit that the issues of relief coordination, equity, and both institutional and informal management under the successive Covid-Amphan disasters are a key arena for further research, essential to improving the outlook of future double strike scenarios.

5. Conclusion

As a region exceptionally vulnerable to climate change, southwestern Bangladesh is at high risk of experiencing future compound disasters. Poverty and persisting weaknesses in governance and health infrastructure heighten these risks. It is critical to understand the vulnerabilities of local communities and fortify them against future risks. Our study provides a snapshot of the livelihood effects the Covid-Amphan ‘double strike’ has had on rural communities in southwestern Bangladesh. It reveals that the overall rates of income reduction (53%) and occupation loss (76%) between the advent of lockdown and the aftermath of Amphan are high, although effects vary moderately by occupation, village, and socio-spatial characteristics. It demonstrates that livelihood resilience to the double strike is generally poor. Particularly noteworthy is how losses concentrate in the pre-cyclone period, and how the effects diverge from those of weather-related disasters alone. The Covid-19 pandemic has impacted a broader swathe of the population over an extended time period; tertiary sectors bear as severe an impact as directly resource-reliant ones. The pandemic challenges existing disaster management strategies and exacerbates the vulnerabilities threatening livelihood resilience against extreme weather events in poor, rural communities. In the face of these multi-dimensional disasters, we posit that firstly, livelihood diversification, and secondly, strengthened community networks, are critical to boosting the low resilience levels identified. Diversification broadens the range of strategies to which communities under duress may turn, strengthening them against the wide-ranging, complex effects of double strikes. Strong social networks and coordination are essential to efficiently combatting multi-faceted threats. Thus, we believe that policies for diversification should be prioritised in the region, together with efforts to bolster community trust and positive social relations.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Many thanks equally go to the respondents who provided vital information during the empirical surveys despite Covid-19 imposed social distant rules. We appreciate the constructive insights offered by the anonymous reviewers. This publication is not intended to reflect the opinions of these institutions or individuals. Any errors remain the responsibility of the authors. B. Mallick has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 846129.