A retrospective study to understand the impact of flooding on natural selection and physical growth dynamics

Abstract

Flooding ranks among the most devastating natural disasters, causing substantial economic disruption and significant loss of life. Its impact on human populations has both immediate and long-term consequences that can extend from birth through adulthood. Studies on the effects of flooding have primarily focused on preschool children, often within diverse populations characterized by heterogeneous genetic ancestry. Ultimately it has left a gap in understanding its long-term developmental effects, particularly during adolescence and within genetically cohesive populations. Notably, no prior study has examined its potential evolutionary significance through natural selection, and finally its impact on adolescent physical growth dynamics as well as the prevalence of stunting within endogamous population groups. To address this gap, the Mishing population from Assam has been selected, which exhibits a degree of genetic homogeneity due to their shared ancestry and endogamous practices. They are characterized by differential habitation patterns in relation to perennial flooding. This variation in exposure to flood-affected and non-affected areas within the same ethnic group provides a unique opportunity to conduct a comparative analysis, minimizing genetic variability while assessing the environmental impact on health and related parameters. The present study thus investigates the effects of flooding across the life course among the Mishing, with specific focus on the opportunities for natural selection and the long-term impacts on stunting and physical growth parameters viz final height, peak height velocity (PHV), and age at peak height velocity (APHV). Using stratified random sampling, two flood-affected and two non-affected Mishing villages were selected from four districts of Assam, covering 1687 households. All 309 post-menopausal women (186 from flood-affected and 123 from non-affected villages) were interviewed for their fertility records, achieving a 100% response rate. A modified index for natural selection (Sikdar’s index) was used to estimate selection pressure across life stages. Out of 3761 children and adolescents (aged 6-20 years) identified in these households, a cross-sectional study was conducted on 2970 individuals (1464 boys and 1506 girls). With a 78.9% response rate, the study assessed height growth variation using the Preece and Baines Growth Curve Model 1. Additionally, the prevalence of stunting was assessed among 2752 individuals (1353 boys and 1399 girls aged 6-19) using the WHO 2007 LMS approach. Results show a substantial selection pressure on the infants in flood-affected villages which accentuates the urgent need for targeted public health interventions and climate-resilient infrastructure. However, among surviving children and adolescents, flood exposure had a negligible impact on postnatal growth parameters like final height, PHV, APHV and stunting, possibly reflecting cultural adaptation, community resilience, and the hygiene hypothesis. Additionally, the findings underline the protective role of higher household income and better maternal education, both of which were significantly associated with reduced odds of stunting which suggest that in flood-prone regions interventions must go beyond emergency response and include enduring as well as development focused policies.

Introduction

In recent years, floods have emerged as one of the most frequent and devastating natural disasters, significantly impacting human populations at different level. Climate change has intensified the frequency and severity of extreme weather events, contributing to heavier rainfall and rising sea levels, which in turn have increased flood risk in both urban and rural areas^1,2. According to the Emergency Events Database³, floods affected over 99 million people between 2015 and 2022, disproportionately impacting low-income populations who often reside in flood-prone areas with inadequate infrastructure. Further global warming is expected to cause an increase in the frequency and intensity of floods in the next decades, potentially affecting up to 1.2% of the world’s population^4,5. The majority of flood-exposed individuals are located in South and East Asia, with China (395 million) and India (390 million) accounting for over one-third of global exposure⁶. Over the past couple of decades, India has experienced an increase in the frequency of floods, primarily due to climate change induced extreme precipitation events⁷. The advancement of spatial mapping techniques has made it possible to identify flood-prone areas and assess their associated vulnerabilities more accurately^8,9. Assam is one such vulnerable states in India so far as the effect of flood on human life is concerned and its recurring flood situation is linked by experts to the huge earthquake of 1950. According to data from National Flood Commission, thirty-one thousand square kilometers, or roughly 39.58 percent of the state’s total area, are under flood-affected areas¹⁰ giving us ample opportunity to understand different dimension of flood and its effects.

There are direct as well as indirect effects of flood which can be described in a simple schematic diagram (Fig. 1). While talking about direct impacts, floods can lead to drowning, traumatic injuries, electrocution, and other fatalities, especially in high-density areas with poor drainage¹¹. Flooding damages the public infrastructure like homes, roads, bridges, and interrupts critical services such as power, water supply, and conveyance¹².Sometimes floodwaters mix with sewage water and industrial waste and leads to the outbreaks of diarrhoeal and vector-borne diseases¹³. Floods often damage the infrastructure of hospitals as well as clinics preventing access to prenatal check-ups, delivery, and postnatal care¹⁴. Beyond immediate injuries and fatalities, floods have profound long-term effects on public health¹⁵.

Fig. 1.

Schematic diagram to illustrate the relationship of flood with other confounding factors.

Sometimes floods have been linked to the negative reproductive outcomes, particularly in low-resource and disaster-prone regions. Displacement, psychological distress, and limited access to antenatal care associated with flood have been shown to increase the risk of pregnancy complications¹⁶. Other findings also highlight an increased incidence of low birth weight (LBW) among neonates born to flood-affected mothers¹⁷. Floods also augment the risk of respiratory infections, and mental health disorders like post-traumatic stress disorder (PTSD)^18–20. Floods also have shown to adversely affect child health and development worldwide. For instance, a systematic review highlighted that floods contribute to undernutrition among under five children in low and middle-income countries, with stunting being the most frequently reported outcome^21–24.

Often floods act as selective pressure on human populations. From evolutionary point of view, Darwin’s theory of natural selection suggests that individuals with traits that enhance reproductive success and survival are more likely to pass on their genes to the next generation²⁵. In the context of recurrent hazards like floods, there are possibilities that the traits viz physiological resilience, immune robustness, or adaptive behaviour (e.g., migration patterns and subsistence approaches) may become advantageous²⁶. Over generations, such traits can become more prevalent within a population. For instance, historical exposure to malaria in certain regions has led to increased frequencies of protective genetic variants such as sickle-cell trait and other abnormal haemoglobins in different parts of the world²⁷. Similarly, in flood prone regions, traditional knowledge systems like housing, and other adaptive behaviour may serve as cultural and biological buffers against selection pressure²⁸.

However, the pace and impulsiveness of contemporary hazards like flood can outpace evolutionary responses, often leading to maladaptive outcomes such as increased mortality, reduced fertility as well as undernutrition^3,29. Therefore, hazards in one hand may act as evolutionary filters but on the other hand contemporary socioeconomic buffers like better healthcare, technology, and education sometimes moderate or override natural selection pressures in human populations³⁰. Thus, studying opportunity for natural selection due to differential mortality as a selective pressure at different stages of life not only enriches our understanding on human adaptability but also informs targeted interventions to reduce health inequalities. To date, no such study has specifically examined these parameters in relation to the evolutionary implications of floods. There is a notable gap in the literature concerning the intersection of flood exposure and natural selection. Further human growth parameters, such as height, body size, and the timing of developmental events are fundamental to evolution, as they are closely linked to reproductive success and species survival^31,32.

These growth patterns are highly sensitive to environmental stress and often reflect adaptive responses to ecological stressors. However, the impact of floods on children’s growth parameters such as the pubertal growth spurt as well as peak velocity has not been adequately investigated. Moreover, presently there is no data available on endogamous populations with a shared genetic background that has experienced differential exposure to flood events. Addressing this gap is crucial for understanding how both environmental and socio-cultural factors interact to influence developmental outcomes in the entire post-natal period. To address these gaps, Mishing community from Assam, Northeast India has been considered because of their differential habitation across flood-prone and non-flood-prone areas. It allows the present discourse to isolate the effects of flood minimising the confounding effect of genetic assortment as well as enhancing the validity of observed association.

Materials and methods

The population

The Mishing population is the second-largest tribal population in Assam, Northeast India and they exhibit a relatively high genetic interconnection due to the endogamous practices within such geographic confinement. As far as cultural homogeneity is concerned, anthropological and sociological studies from the colonial period often portrayed tribal groups in India as homogeneous entities³³. However, there are arguments that such representations tend to oversimplify the complex social fabrics of tribal communities. Basically, their main proposition is that tribal societies are shaped by internal multiplicities as well as historical, political, and cultural influences³⁴. The Mishing community traditionally inhabits low lying riverine areas of Brahmaputra River which is locally known as chars or chapori. As such they are disproportionately affected by the perennial as well as seasonal flooding. Further being a Scheduled Tribe population, they often experience economic and social marginalization, limited access to healthcare, education, and essential services. Most of them live in bamboo huts and rely on small-scale farming or fishing, making them more vulnerable to the destruction caused by floods. They have unique cultural adaptations like stilt houses to live with floods. However, these adaptations are being strained by increasing flood intensity and frequency due to climate change. According to the National Family Health Survey³⁵, tribal populations are marginalised people of India and they used to suffer from poorer health indicators, including higher infant mortality, malnutrition, and lower immunization coverage. Flooding further disrupts healthcare access in these regions, making them a critical case study for public health research. The Mishing people are also known as ‘river loving’ people. After many years of settlement in the river bank of Brahmaputra and its tributaries, the community has webbed deep relations with river ecosystem. The rationale behind focusing on the Mishing community lies in their high physical exposure to floods, structural and socioeconomic vulnerabilities, and the need for culturally grounded public health interventions. Studying them offers a microscopic lens on macro-level environmental injustice and disaster risk, especially in the context of climate-sensitive health outcomes.

Ethical consideration

The ethical considerations in this study centered on ensuring informed consent, data privacy, and the protection of vulnerable participants. While collecting and utilizing data related to fertility outcomes from elderly females, special care was taken to respect their autonomy and uphold their right to refuse participation at any stage. Sensitivity to potential emotional and social vulnerabilities was preserved throughout the process. In the case of children and adolescents, the study prioritized their well-being and autonomy, safeguarding the idea that consent was obtained in an age-appropriate manner, either directly or via legal guardians. The research process was guided by principles of fairness, justice, and confidentiality, with robust safeguards in place to protect the privacy and dignity of all participants. The feasibility and ethical aspects of the study were thoroughly reviewed by the Departmental Research Committee of Dibrugarh University. Following this, the study received formal approval from the Ethical Committee for Biomedical and Health Research Involving Human Participants, Dibrugarh University, Assam, India. All procedures involving human participants were conducted in accordance with the ethical principles outlined in the Declaration of Helsinki. Prior to participation, informed consent was obtained from all individuals, ensuring they were fully aware of the study’s purpose, procedures, and their right to withdraw at any time without consequence.

Minimum household number and sample size

In Assam, the Mishing population is distributed in 8 districts with a total population of around 7 lakhs individuals (Source: Mising Autonomous Council). To estimate the total number of households to be covered from a population of 700,000 individuals, I have used the average household size. In Assam, the average household size is approximately 4.5 members per household (Source: Ministry of Statistics and Program Implementation). So, for a population of 700,000 individuals the approximate total number of Mishing households will be

Following Yamane’s formula³⁶ we can estimate the minimum households to be considered to represent 1,55,555 total Mishing households in Assam. Thus, minimum household size (n) will be:

where, N = 155,555 (total number of households available for the study). e = 0.05 (margin of error, 5%).

Therefore, ≈ 400.

Following a stratified random sampling, a total of 4 villages were selected from 4 different districts of Assam having a minimum household number of 400 in each village (Fig. 2). Out of the 4 villages, 2 represented flood affected village and 2 non-affected village as per district disaster management authority. A total of 1687 households were selected. All the post-menopausal women from these households (186 from flood affected and 123 from non-flood affected villages) were interviewed showing cent percent response rate. Out of 3761 total children and adolescents available in these 1647 households, a cross-sectional study involving 2970 children and adolescents (1464 boys and 1506 girls from 6 to 20 years) was undertaken (78.9% response rate) to understand varying timing and tempo of height growth. The descriptive statistics of height growth has been derived in terms of absolute growth along with growth percent in the both the sexes within two varying environments. The difference in height growth has been measured in terms of Welch’s t-test. All the children and adolescents from 6 to 19 years (1353 boys and 1399 girls) were evaluated for stunting based on their height-for-age Z-score values as per WHO criteria.

Fig. 2.

Map of Assam showing selected study districts.

Household income

The household income categories in Assam are delineated based on various government schemes and economic surveys. Government of Assam has utilised these classifications to identify the eligibility for specific programs and to analyse economic demographics. The same categories were used here at first to categorize a household into low Income Group, Middle Income Group and High-Income Group. Households with an annual income below ₹4 lakh (1$ = ₹86.27 on 08.04.2025) was categorised as Low-Income Group. This threshold is used to determine eligibility for the Antyodaya Anna Yojana (AAY) and Priority Household (PHH) cards under the National Food Security Act, 2013. Income between ₹4 Lakhs to ₹5 Lakhs were categorised as Middle-Income Group. This category is considered for various state initiatives, including the Atal Amrit Abhiyan, a cashless healthcare scheme covering critical diseases. Households with an annual income exceeding ₹5 lakh were considered High-Income Group. Typically, high-income group people did not qualify for income-based government assistance programs and was expected to afford services without subsidies. Among the investigated households, low-income household was found to be more common (81.97%), often under ₹1 lakh/year due to seasonal or subsistence livelihoods. Therefore, to assess the association between household income and stunting, logistic regression analysis was performed using monthly household income as a continuous variable, measured in ₹100/month increments.

Demographic variables

Data on reproductive history were gathered from all the post-menopausal women using a semi-structured interview schedule, followed by enumeration of pregnancies. Details including the number of pregnancies, embryonic losses, live births, infant and childhood mortalities, and age at death were documented via extended genealogies. To enhance reliability, the provided information was cross-checked with other elderly family members.

Opportunity for natural selection at different stages of life

Crow (1958)³⁷ segregated the index of the opportunity for natural selection () into two parts, due to selective mortality and due to differential fertility, where:

1

2

3

Here is the probability of survival from birth to maturity, is the mean number of births per survivors and is the variance of

In developing countries there is a variability of post-natal mortality in different stages of life taking a heavy death toll at infant stage. Differential mortality at different stages of life viz. early embryonic age, infant age, early childhood age and late childhood age has tremendous impact on demographic trends of a population. For example, a cohort of children born in a specific time period might show different mortality patterns, which can affect future generations in terms of workforce composition, dependency ratios, and even social stability. For instance, if infant mortality is high in a given period due to environmental hazards, the cohort size will be smaller, and the population may face labour shortages or increased pressure on social support systems later, which subsequently affect the opportunity for natural selection in a particular population. We don’t have any measure to calculate these evolutionary forces on different population groups of the world. Thus, I proposed an index () by disintegrating Crow’s index of natural selection to specifically assess the force of selection due to embryonic mortality, infant mortality (0–1 years), early childhood mortality (1–5 years) and late childhood mortality (5-15 years)³⁸.

Assuming, = proportion of early embryos that survive to birth.

= proportion of survivors from birth to infancy.

= proportion of survivors from infancy to early childhood.

= proportion of survivors from early childhood to reproductive age.

= proportion of survivors form early embryonic stage to infancy.

 = proportion of survivors from early embryonic stage to early childhood.

= proportion of survivors from early embryonic age to reproductive years.

= the mean number of offspring per embryo (i.e. when all ‘individuals’ are counted during the early embryonic stages).

= the mean number of offspring per surviving adult.

= the mean number of offspring per newborn (where all individuals are counted at birth).

Therefore, Index due to fertility

4

where (= ) is the variance in offspring number of the survivors and is the square of the mean number of offspring per surviving adult.

Now the variance from embryonic mortality is,

5

The variance from all mortality,

6

The variance from early embryonic to infant mortality.

7

The variance from early embryonic to early childhood mortality.

8

Thus, the variance from infant (birth -1 year) mortality becomes

The variance from early childhood (1–5 years) mortality will be

The variance from late childhood (5–15) mortality is

The selection potential due to embryonic mortality is the ratio of embryonic deaths to survivors,

The index due to infant mortality (birth-1 years) will be

The index due to early childhood mortality (1–5 years) is

The index due to late childhood mortality (5-reproductive years) becomes

The index of selection potential, considered from early embryonic stage to reproductive years thus composed of five components i.e. prenatal mortality (), infant mortality (), early childhood mortality (), late childhood mortality () and differential fertility of the survivors ().

Thus,

[Since, and ]

9

To identify the significance level of the differential mortality components of natural selection in the two groups, Z score has been applied following standard method³⁹ where and are the number of successes in sample 1 and 2, and are the respective sample sizes, and P is the pooled proportion

Growth models

In order to obtain information on timing and tempo of growth events, growth-velocity curves can be constructed and mathematical models can be used to estimate the biological and mathematical parameters, allowing for the investigation on the variability of growth velocities⁴⁰. The non-invasive biological parameters like the Age at Peak Height Velocity (APHV) (years) and the Peak Height Velocity (PHV) (cm/years) can be determined using the Preece and Baines 1 Mathematical Model⁴¹. This model is a member of a family of curves that closely resembles the human growth curve. Using cross-sectional data from children and adolescents living throughout the globe, a number of research have concentrated on simulating linear physical growth^42,43.

The following formula infers the five parameters of this nonlinear model:

Here, resembles the final height (cm), whereas and θ are the average height (cm) and age (years) for height on the decreasing slope of the Peak Height Velocity. Controlling height velocity (cm/years), S₀ and S₁ are prepubertal and pubertal rate constants. The normal distribution of the data was verified using the Kolmogorov-Smirnov test. The basic concept that has been used using this modelling process was calculating the best fitting equations to describe the real data based on the criterion of the least sum of squares. The SOLVER setup in Microsoft excel-365 for least sum of squares in linear curve fitting was used for the purpose. All the descriptive statistics of mean, standard deviation (SD), standard error (SE) was calculated as per standard method. Welch’s t-test was used to compare the two groups of participants from flood affected and non-affected villages. Statistical significance was set at p value less than 0.05.

Limitations of using Preece and Baines Model 1 in cross-sectional data

The use of cross-sectional data to estimate dynamic growth parameters such as Age at Peak Height Velocity (APHV) and Peak Height Velocity (PHV) introduces significant boundaries in the context of growth modelling using the Preece-Baines Model 1 (PB1). The PB1 model is a parametric growth curve model and it was designed to fit the longitudinal data where repeated measures of continuous variables over time allow for accurate tracking of individual growth trajectories⁴¹. However, cross-sectional studies provide only a “snapshot” of individual growth trajectories at different ages, rather than tracking the same individuals over time. This design mismatch can lead to several potential biases like ‘temporal misalignment’, since each age group in a cross-sectional study comprises different individuals, the natural inter-individual variability may be mistaken for intra-individual growth patterns, thus distorting the estimation of growth parameters⁴⁴. Similarly, there may be cohort effect, where environmental, nutritional, and socioeconomic factors that vary across cohorts can confound age-related changes in continuous variables. Sometimes derived growth parameters may reflect population level variation which may not define the true biological timing of maturation⁴⁵. PB1 fittings in cross-sectional data sometimes tend to over-smooth or under-smooth rapid changes in growth parameters especially during puberty. Therefore, there may be a tendency in smoothing error too. This may lead to inaccurate estimation of the timing and magnitude of growth spurts⁴⁶. Despite these limitations, cross-sectional studies often employ the Preece-Baines Model 1 to smooth growth curves^47,48.

Assessment of stunting and associated factors

Stunting has been defined as the height-for-age below minus two standard deviation (-2SD) from the median value of WHO Child Growth Standard⁴⁹. A diverse range of environmental factors are already found to be associated with stunting^50,51. To understand the distribution of stunting, both boys and girls within the age group of 6-19 years were evaluated as per the WHO 2007 height-for-age reference. Individual Z-score values were calculated following the LMS method⁵² and categorized to represent normal, moderate, and severe cases of malnourishment. Chi-square values were calculated to understand the significant difference in the prevalence of malnutrition in two environmental setups. A heatmap was created using Microsoft Excel 365 on Windows 10, utilizing conditional formatting to visualize the distribution of different forms of stunting (moderate as < -2SD and severe as < -3SD) across different age groups (6-19 years) in both boys and girls from flood-affected and non-affected regions. Forest plot showing adjusted odds ratios and 95% confidence intervals for socioeconomic and environmental predictors of stunting was generated through Python version 3.13.2 utilising the matplotlib along with seaborn. A multivariable logistic regression analysis was performed to examine the association between stunting (dependent variable: stunted vs. not stunted) and a set of socio-demographic and environmental predictors. The analysis included flood exposure, household income, maternal education levels, and child sex as independent variables. Flood exposure was coded as a binary variable (exposed = 1, not exposed = 0). Household income was treated as a continuous variable in increments of 100 Indian rupees. Maternal education was categorized into three levels viz no formal education (reference category), primary education, and secondary education. Sex of the individual was coded as male = 1 and female = 0. The logistic regression model estimated odds ratios (ORs) with 95% confidence intervals (CIs) and corresponding p-values. Model coefficients (β) and standard errors (SE) were also reported. Statistical significance was set at p < 0.05. All statistical analyses were conducted using SPSS version 25.

Results

The differential intensity of natural selection, based on fertility and mortality variations in the two discrete environmental backgrounds is presented in Table 1. This table presents the Sikdar’s index of natural selection () and its components across two environmental contexts i.e. flood-affected and non-affected populations. The indices are broken down into embryonic mortality (), infant mortality (), early childhood mortality (), late childhood mortality (), and fertility component (). The ratios and total selection index () are also shown, along with the relative contributions of each component (in%). The results reveal a marked difference in the intensity of natural selection at the infant stage between the two environments, with an value of 0.070 in the flood-affected environment compared to 0.41 in the non-affected environment. However, the overall intensity of natural selection () appears to be relatively similar across both the settings (in flood affected villages = 0.48, in the non-affected villages = 0.42). The data clearly indicate that infant mortality is the key differential factor driving higher natural selection intensity in flood-affected areas. While the fertility component dominates in both groups, ecological stress (flood exposure) appears to shift the burden of selection more heavily to the postnatal period especially infancy highlighting its impact on population fitness. The Z-score values of intergroup variability with regard to mortality components show significant differences in embryonic (Z = − 2.37) and infant mortality (4.07) between the two groups, while other components show no statistically significant difference.

Table 1.

Indices of selection potential among the Mishing population based on early embryonic and postnatal mortality (Based on ).

Environment											% of embryonic mortality	% of infant mortality	% of early childhood mortality	% of late childhood mortality	% of fertility component
Flood affected	0.08	0.07	0.02	0.02	0.23	0.08	0.02	0.02	0.28	0.48	17.4	15.8	4.98	4.56	57.05
Non-affected	0.09	0.04	0.02	0.02	0.2	0.05	0.02	0.03	0.24	0.42	21.6	9.73	4.03	6.17	57.00
Z score	–	–	–	–	–	–	–	–	–	–	-2.37*	4.07*	1.02	-1.60	0.02

*Significant at 5% level.

The mean and standard deviation of stature, along with absolute growth and growth percent values for the participants from flood-affected and non-affected children and adolescents (both boys and girls) have been presented in Tables 2 and 3 respectively. Table 2 compares the growth pattern of stature among the boys from flood-affected and non-affected areas between 6 to 20 years. While both groups show similar growth trends in early years, peak height growth occurs between 13 and 14 years of age, with the non-affected group exhibiting slightly higher growth spurts. By the age of 16 years, growth plateaus with minimum gain have been observed in both the groups. By age 20, the non-affected group averages 1.6 cm taller than the flood-affected group (165.04 cm vs. 163.44 cm). However, Welch’s t-values remain below significance thresholds throughout, indicating no statistically significant height differences across ages. The Table 3 presents the comparative account of stature among Mishing girls from 6 to 20 years in both the flood-affected and non-flood-affected setting. It also highlights noticeable growth differences between the two groups. However, none of the Welch’s t-values indicate statistically significant differences. While both groups show typical growth patterns peaking around ages 10 to11 years, girls from flood-affected areas exhibit slightly lower mean statures and an earlier tapering of growth, with negative or negligible growth beyond age 15. In contrast, girls from non-flood-affected areas continue to grow modestly until about 18 years.

Table 2.

Descriptive statistics of stature (absolute) among the Mishing boy of two different setting.

Age in years	Flood affected					Non flood affected					Welch’s t value
	N	Mean (cm)	SD	Absolute growth	Growth percent	N	Mean (cm)	SD	Absolute Growth	Growth percent
6	51	105.34	5.32			45	106.49	4.92			1.10
7	52	112.93	5.67	7.59	7.20	41	114.16	5.26	7.67	7.20	1.08
8	50	119.54	6.15	6.61	5.86	50	120.22	5.85	6.05	5.30	0.57
9	51	125.76	5.83	6.22	5.20	45	125.99	7.31	5.78	4.81	0.17
10	52	132.21	5.36	6.44	5.12	55	131.38	6.03	5.39	4.28	0.75
11	51	136.88	6.62	4.67	3.53	51	136.19	6.69	4.81	3.66	0.52
12	54	140.90	6.61	4.02	2.94	47	142.32	6.89	6.13	4.50	1.05
13	50	144.85	7.41	3.94	2.80	53	144.93	7.75	2.62	1.84	0.05
14	50	153.32	6.70	8.47	5.85	40	154.13	6.22	9.20	6.34	0.59
15	52	158.14	6.41	4.82	3.15	38	159.31	6.33	5.18	3.36	0.86
16	50	161.45	5.34	3.31	2.09	41	160.17	5.17	0.87	0.54	1.16
17	51	161.41	3.99	-0.04	-0.02	42	161.06	4.27	0.89	0.56	0.41
18	50	162.19	7.04	0.78	0.48	49	162.92	6.30	1.86	1.15	0.54
19	50	163.39	5.51	1.20	0.74	46	163.64	5.93	0.73	0.45	0.21
20	54	163.44	6.19	0.05	0.03	52	165.04	5.33	1.40	0.85	1.43

N Number of individuals, SD Standard deviation, cm centimeter.

Table 3.

Descriptive statistics of stature (absolute) among the Mishing girls of two different setting.

Age in years	Flood affected					Non flood affected					Welch’s t value
	N	Mean (cm)	SD	Absolute growth	Growth percent	N	Mean (cm)	SD	Absolute Growth	Growth percent
6	53	105.01	4.76			52	106.27	5.24			1.68
7	52	112.69	5.27	7.68	7.31	50	113.06	5.69	6.79	6.39	1.52
8	50	116.89	5.72	4.20	3.73	51	117.83	5.86	4.77	4.22	0.57
9	50	122.80	5.24	5.91	5.05	50	122.86	6.02	5.04	4.27	0.55
10	52	127.89	6.39	5.09	4.15	53	127.58	6.76	4.72	3.84	0.75
11	52	136.40	6.25	8.51	6.65	53	134.99	7.10	7.41	5.80	0.52
12	50	140.90	6.16	4.50	3.30	57	142.15	6.27	7.17	5.31	1.05
13	54	144.94	6.32	4.04	2.87	38	146.69	5.64	4.53	3.19	0.05
14	50	148.22	6.10	3.28	2.27	41	148.86	5.89	2.17	1.48	0.33
15	52	149.73	5.55	1.51	1.02	45	149.65	5.78	0.80	0.54	0.86
16	53	150.84	5.08	1.11	0.74	47	150.29	6.24	0.63	0.42	1.62
17	52	151.41	6.34	0.57	0.38	52	151.76	6.52	1.47	0.98	0.53
18	50	152.18	6.67	0.77	0.51	50	152.26	6.84	0.50	0.33	0.65
19	52	152.10	6.17	-0.08	-0.05	38	153.27	6.96	1.01	0.66	0.63
20	51	152.23	6.63	0.13	0.08	56	153.19	5.97	-0.08	-0.05	0.78

N Number of individuals, SD Standard deviation, cm centimeter.

The five mathematical parameters estimated by the Preece and Baines-1 Model have been shown in Table 4 and it demonstrate small residuals in both flood-affected and non-affected environments (among boys, 0.37 and 0.40, and in girls, 0.35 and 0.38). In both sexes, no significant difference (as per Welch’s-t test) was found in final height (h₁) or peak height velocity size (h₀) between both the environmental setups. As far as the biological parameters are concerned (APHV and PHV), they were found to be similar in both the environment and in both the sexes (APHV among girls = 12.1 and 11.8 in flood affected and non-affected, respectively, whereas APHV among boys = 13.7 and 13.8 in flood-affected and non-affected, respectively).Taking cognizance from the different biological parameters, the pictorial graph of Preece and Baines Model 1 fit to the distance and velocity curves of height among boys and girls of both setups (flood-affected vs. non-affected) is depicted in Fig. 3. Among the girls, the velocity curve shows delayed and reduced peak height velocity in flood-affected setting, while the distance curve indicates consistently lower final stature compared to non-affected peers. The same situation prevails in boys too, suggesting trivial impact of flood affected environment towards growth parameters.

Table 4.

Sexual and environmental variation in mathematical and biological parameters of height growth estimated through Preece-Baines Model 1.

PB model parameters	Boys				Girls
	Flood affected		Non affected		Flood affected		Non affected
	Mean	SE	Mean	SE	Mean	SE	Mean	SE
Mathematical parameters
s0	0.11	0.00	0.11	0.00	0.12	0	0.12	0
s1	0.86	0.02	0.87	0.05	0.87	0.02	0.93	0.05
θ	14.50	0.06	14.4	0.07	11.04	0.05	11.07	0.09
hθ	152.20	0.17	152.5	0.39	145.09	0.19	145.07	0.34
h1	163.60	0.22	163.80	0.34	156.98	0.14	156.62	0.25
RSE	0.37		0.40		0.35		0.38
Biological parameters
APHV (y)	13.8	0.57	13.8	0.72	10.2	0.39	9.96	0.47
PHV (cm/y)	3.74	0.38	3.95	0.45	3.46	0.82	3.70	0.55

Fig. 3.

Sexual and environmental variation in PB1 fit to distance and velocity curves of height.

The distribution of different forms of stunting among the Mishing boys and girls as per WHO 2007 criteria has been depicted in Tables 5 and 6, respectively. It is evident from Table 4 that there is no significant difference in the prevalence of stunting between the boys of two environmental setups, either in early childhood or late childhood. 29.3% of the flood-affected boys are found to be stunted, whereas 27.5% of the non-affected boys are found to be stunted. Table 5 shows the prevalence of stunting among the girls in the two environmental setups. The table also shows that there is no significant difference in the prevalence of stunting among the girls between the two environmental setups (34.1% in flood-affected and 33.5% in non-affected).

Table 5.

Prevalence of stunting (as per Z-score) among the boys of two environmental setup.

Age in years	Flood affected							Non-affected							X2 values for stunting
	Total boys	HFAZ <  − 2SD		HFAZ <  − 3SD		Total stunt		Total boys	HFAZ <  − 2SD		HFAZ <  − 3SD		Total stunt
		N	%	N	%	N	%		N	%	N	%	N	%	Total	Severe
6	51	11	21.6	12	23.5	23	45.1	45	10	22.2	8	17.8	18	40.0
7	52	12	23.1	9	17.3	21	40.4	41	9	22.0	4	9.8	13	31.7
8	50	11	22.0	2	4.0	13	26.0	50	9	18.0	2	4.0	11	22.0
9	51	19	37.3	2	3.9	21	41.2	45	14	31.1	6	13.3	20	44.4
10	52	5	9.6	0	0.0	5	9.6	55	7	12.7	0	0.0	7	12.7
(6–10)	256	58	22.6	25	9.8	83	32.4	236	49	20.8	20	8.5	69	29.2	0.58,df = 1	0.02,df = 1
11	51	5	9.8	1	2.0	6	11.8	47	10	21.3	0	0.0	10	21.3
12	54	13	24.1	1	1.9	14	25.9	47	8	17.0	1	2.1	9	19.1
13	50	10	20.0	2	4.0	12	24.0	53	13	24.5	1	1.9	14	26.4
14	50	10	20.0	1	2.0	11	22.0	40	6	15.0	1	2.5	7	17.5
15	52	14	26.9	0	0.0	14	26.9	38	4	10.5	1	2.6	5	13.2
(11–15)	257	52	20.2	5	1.9	57	22.2	225	41	18.2	4	1.8	45	20.0	0.34,df = 1	0.00,df = 1
16	50	17	34.0	0	0.0	17	34.0	41	13	31.7	2	4.9	15	36.6
17	51	12	23.5	1	2.0	13	25.5	42	11	26.2	1	2.4	12	28.6
18	50	15	30.0	7	14.0	22	44.0	49	14	28.6	5	10.2	19	38.8
19	50	14	28.0	3	6.0	17	34.0	46	13	28.3	3	6.5	16	34.8
(16–19)	201	58	28.9	11	5.5	69	34.3	178	51	28.7	11	6.2	62	34.8	0.01,df = 1	0.08,df = 1
Total	714	168	23.5	41	5.7	209	29.3	639	141	22.1	35	5.5	176	27.5	0.49,df = 1	0.00,df = 1

*Significant at 5% level, HFAZ = Height-for-Age Z score.

Table 6.

Prevalence of stunting (as per Z-score) among the girls of two environmental setup.

Age in years	Flood affected							Non-affected							X2 values for stunting
	Total girls	HFAZ <  − 2SD		HFAZ <  − 3SD		Total stunt		Total girls	HFAZ <  − 2SD		HFAZ <  − 3SD		Total stunt
		N	%	N	%	N	%		N	%	N	%	N	%	Total	Severe
6	53	22	41.5	5	9.4	27	50.9	52	17	32.7	4	7.7	21	40.4
7	52	17	32.7	2	3.8	19	36.5	50	18	36.0	1	2.0	19	38.0
8	50	13	26.0	5	10.0	18	36.0	51	12	23.5	4	7.8	16	31.4
9	50	7	14.0	5	10.0	12	24.0	50	8	16.0	6	12.0	14	28.0
10	52	26	50.0	0	0.0	26	50.0	53	23	43.4	4	7.5	27	50.9
(6–10)	257	85	33.1	17	6.6	102	39.7	256	78	30.5	19	7.4	97	37.9	0.17,df = 1	0.29,df = 1
11	52	9	17.3	2	3.8	11	21.2	53	14	26.4	4	7.5	18	34.0
12	50	14	28.0	0	0.0	14	28.0	57	14	24.6	0	0.0	14	24.6
13	54	16	29.6	3	5.6	19	35.2	38	7	18.4	1	2.6	8	21.1
14	50	20	40.0	0	0.0	20	40.0	41	16	39.0	0	0.0	16	39.0
15	52	17	32.7	2	3.8	19	36.5	45	15	33.3	2	4.4	17	37.8
(11–15)	258	76	29.4	7	2.7	63	24.4	234	66	28.2	7	2.9	73	31.2	2.82,df = 1	0.08,df = 1
16	53	15	28.3	1	1.9	16	30.2	47	16	34.0	1	2.1	17	36.2
17	52	12	23.1	4	7.7	16	30.8	52	11	21.2	5	9.6	16	30.8
18	50	11	22.0	4	8.0	15	30.0	50	11	22.0	4	8.0	15	30.0
19	52	11	21.2	3	5.8	14	26.9	38	8	21.1	1	2.6	9	23.7
(16–19)	207	49	23.7	12	5.8	81	39.1	187	46	24.6	11	5.9	57	30.5	3.23,df = 1	0.48,df = 1
Total	722	210	29.1	36	5.0	246	34.1	677	190	28.1	37	5.5	227	33.5	0.05,df = 1	0.25,df = 1

*Significant at 5% level, HFAZ = Height-for-Age Z score.

The heatmap (Fig. 4) shows that pre-adolescents boys and girls are having highest rates of moderate stunting irrespective of the environmental stressors like flooding. Boys from 9 years of age and girls from 10 years of age are having high prevalence of moderate stunting. The red colour (under < -2SD = moderate stunting) is clearly visible in both flood-affected and non-affected villages. This phenomenon suggests that there are factors beyond flood exposure that are influencing the nutritional status of this age group of children.

Fig. 4.

Heat map showing colour gradient and intensity of stunting in different age groups in both the setting.

The logistic regression analysis (Table 7) shows that the coefficient for flood exposure is positive but comparatively small (β = 0.12). The odds ratio (OR) is 1.13, which suggests 13% higher odds of stunting for individuals exposed to floods compared to those not exposed, but the p-value is 0.19 (greater than 0.05), indicating that this result is not statistically significant at the conventional 5% significance level. Household income has a significant negative effect on stunting. For each additional increase of ₹100 (Rupees) in household income, the odds of stunting decrease by 5% (OR = 0.95). The p-value is 0.01, which is statistically significant. No formal education serves as the baseline against which other educational levels are compared. Having a mother with primary education is associated with 18% lower odds of stunting compared to having a mother with no formal education. The p-value of 0.04 suggests that this effect is statistically significant. Primary education provides some benefits in improving child health, but the effect is smaller than secondary education. Having a mother with secondary education is associated with 30% lower odds of stunting, which is statistically significant (p-value < 0.01). This result suggests that maternal education has a strong protective effect against stunting in children. There is no significant difference between boys and girls regarding stunting in this sample, as the coefficient is very close to zero (β = 0.01) and the p-value is 0.85. Therefore, sex does not appear to be a factor influencing stunting in these age groups. The forest plot (Fig. 5) visually represents the estimated odds ratios (ORs) and their 95% confidence intervals (CIs) for key predictors of stunting.

Table 7.

Logistic regression showing probable compounding factors for stunting.

Variable	Coefficient (β)	Standard error (SE)	Odds ratio (OR)	95% confidence interval (CI)	p-value
Flood wxposure	0.12	0.09	1.13	(0.96, 1.32)	0.19
Household income (per 100 Rupees)	− 0.05	0.02	0.95	(0.91, 0.98)	0.01
Maternal Education
No formal education	Reference		1
Primary education	− 0.2	0.1	0.82	(0.68, 0.99)	0.04
Secondary education	− 0.35	0.12	0.7	(0.58, 0.84)	< 0.01
Sex	0.01	0.05	1.01	(0.91, 1.13)	0.85

Fig. 5.

Forest plot showing adjusted Odds Ratios and 95% Confidence Intervals for Socioeconomic and Environmental Predictors of Stunting (Ages 6-19).

Discussion

Floods remain a significant threat to human health and well-being, particularly in vulnerable and densely populated regions and the effects are always more noticeable among the marginalized population groups of the world. With the intensification of climate change, the frequency and severity of flooding events have increased, leading to widespread displacement, contamination of water sources, disruption of healthcare services, and heightened risk of waterborne diseases. In the aftermath of floods, communities often face long-term challenges such as food insecurity, loss of livelihoods, and inadequate sanitation, which exacerbate malnutrition and increase susceptibility to infections, especially among children and pregnant women. The impact of flooding among the pregnant mothers and birth outcome is found to be substantial^53–55. However, there is a lack of studies examining the impact of floods on reproductive loss from an evolutionary perspective. The present study highlights the impact of flooding on selection pressure, primarily through increased infant mortality, alongside notable contributions from embryonic mortality.

To reduce embryonic and infant mortality in flood-prone regions, targeted public health interventions are essential. These include ensuring access to antenatal and postnatal care, nutritional supplementation, clean drinking water, sanitation, and safe childbirth facilities. Mobile health units, vaccination programs, and maternal health awareness campaigns can help bridge service gaps during disasters. Integrating disaster preparedness into routine healthcare and strengthening local health infrastructure will enhance resilience and improve survival outcomes for mothers and infants during and after flood events. Over time, this can influence reproductive fitness, favouring traits linked to resilience. Such natural selection processes may shape population level adaptations in immune response, metabolism, and reproductive timing across generations. Studies shows that gestational flood exposure is shown to be associated with increased pregnancy loss⁵⁶ which is definitely an important contributor of embryonic mortality component. Although data on the temporal trend of high infant mortality associated with flood exposure in Northeast India is limited, comparable research from similar regions in Bangladesh’s Ganges-Brahmaputra-Meghna River basin found that living in flood-prone areas was associated with 5.3 additional risk of infant deaths per 1000 births over a period of 30 years. This risk became more vulnerable during rainy season compounding with 7.9 additional deaths per 1000 live births⁵⁷. Another significant contributing factor to elevated infant mortality was the lack of adequate sanitation and hygiene^58–60. The present study focuses on pregnancy enumeration based on retrospective recall from women who have completed their reproductive years. This approach provides a valuable insight into the cumulative burden of repeated flood exposure on pregnancy loss over the full reproductive period. Examining the role of pre-reproductive mortality to population dynamics in flood-prone areas offers important insights into evolutionary processes. However, capturing data on the basis of recall method has limitations for understanding the experiences during the gestational period. There may be memory bias which can affect the reliability of the reported outcomes. Longitudinal follow up studies in future covering the pregnancies in real time would provide more precise and nuanced understanding of gestational health outcomes along with associated vulnerabilities.

Children who survive flood events may face an increased risk of food insecurity that can have long-term consequences on their physical growth. However, empirical evidence on these long-term effects remains limited. The idea that growth is a means by which people adjust to their immediate physical environment has been around since long, despite the fact that these advantages have been quantified only to a limited extent⁶¹. The present study is the first to investigate the influence of a flood-prone environment on growth dynamics, specifically height, using the Preece-Baines growth model. Although the impact of living in a flood-prone environment on growth trajectories appears to be minimal, the differences were statistically insignificant across key parameters such as Age at Peak Height Velocity (APHV), Peak Height Velocity (PHV), and final height (h₁). Similar results came out in a study comparing the growth parameters in final height, peak height, APHV, or PHV of children and adolescents between low and moderate altitude in Colombia⁴³. The lack of growth variations between regions (altitude) was explained by the reduction in the disparities between different populations, the improvement of living conditions in Colombia, and lifestyles, including food habits and physical activity levels. It may be noted that the Mishing community has developed both ex-ante and ex-post strategies to cope up with the perennial floods. Their preventive measures include building elevated homes i.e. stilt houses which is locally called as chang ghors. They also use hybrid housing with durable materials along community preparedness planning. Most of the households are forearmed with locally made boats.

Apart from relying on raised shelters in post flood period, the Mishing people try to diversify the crop from long-duration to short-duration crops. They also could develop certain informal flood warning system which help them to predict flood. Sometimes these strategies combine indigenous knowledge with modern interpositions, enhancing resilience and reducing vulnerability. Their approach exemplifies adaptive living in a flood-prone environment, balancing tradition and innovation to safeguard health, housing, and livelihoods. These resilience plans employed by the Mishing people can offer valuable lessons for other flood-prone communities around the world. Integrating indigenous knowledge with modern scientific techniques has already proved to be more sustainable and adaptive solutions to flooding⁶². This move can be an important model for communities worldwide that are facing similar challenges due to climate change and natural disasters. In the context of post-childhood development, particularly among children who have survived early-life exposure to flood-related environments, the hygiene hypothesis offers another captivating descriptive outline⁶³. The hygiene hypothesis postulates that exposure to a broader range of microbial agents and environmental stressors during the early phase of life may play a role in conditioning immunological resilience in later stage of life. Children raised in flood-prone environments often experience insanitary and unhygienic conditions along with, recurrent exposure to pathogens. While these factors contribute to preeminent infant morbidity and mortality, survivors from this background may undergo a form of natural immunological selection and adaptation. Some studies have suggested a paradoxical defensive effect whereby low-level exposure to certain environmental pathogens may enhance immune system development⁶⁴.

However, the hygiene hypothesis might only apply in controlled or moderate exposure settings, or where cultural practices (i.e. breastfeeding, food hygiene) buffer against the worst impacts. Therefore, in flood-prone regions, public health strategies should aim to mitigate extreme exposures while allowing for healthy microbial interactions, safe outdoor activities, sometimes probiotic nutrition and regulated exposure to diverse environments in child-rearing practices.

Flood exposure was found to have weak and statistically insignificant effect on stunting, suggesting that other factors play a more significant role. Household income was found to be a significant predictor, with higher income associated with a reduced risk of stunting. Maternal education shows a strong, dose-dependent relationship, with secondary education providing the greatest protection against stunting. No significant sex difference in the likelihood of stunting was observed suggesting higher household income to be associated with lower odds of stunting. This outcome supports the idea that economic resources can improve nutrition and health outcomes among the children of flood affected zones. Socioeconomic disparities influence both exposure to floods and access to post-flood healthcare which support the studies that spatial-economic analysis aids in interpreting the social determinants of health⁶⁵. Sometimes extended families and community solidarity can also buffer impacts. For instance, kinship-based food sharing can protect the youngest. There are studies which suggest that stunting is a social and community phenomenon, and not a fixed result of geography or genetics. It reflects the emotional, economic, and political wellbeing of a society^66,67. Comparable prevalence of stunting in the both the settings corroborate this understanding. There are studies that explores how social status differences within communities can lead to variations in insulin-like growth factor-1 (IGF-1) levels, subsequently affecting height development. Children who survive early adversity may exhibit catch-up growth or biological adaptations in nutrient utilization and growth regulation. A longitudinal study examined children who were in utero during the 2004 Indian Ocean tsunami. Initially, these children exhibited reduced linear growth compared to non-exposed peers. However, over time, most demonstrated substantial catch-up growth in height. The study highlighted that successful reconstruction efforts and alleviation of maternal stress were crucial for this recovery. Notably, children whose mothers reported high levels of post-traumatic stress remained vulnerable to persistent growth deficits⁶⁸.However, geographical inequalities, in terms of infrastructure, services, and environmental quality may translate into disparate health outcomes during and after floods. Several recent studies in India have highlighted the importance of geographical heterogeneity in designing targeted public health interventions^69–73, a finding that may be applicable to other countries as well.

Conclusion

The present findings indicate that while floods are significantly associated with increased rates of infant mortality, they appear to have limited direct influence on post-infancy developmental parameters such as final height, height velocity, peak height velocity, and the prevalence of stunting. Stunting being a long-term development indicator, seems to be more strongly influenced by socioeconomic determinants more importantly household income and maternal education. Floods seems to have contributed towards infant mortality through multiple acute pathways. These includes the disruption of healthcare services, heightened exposure to infectious diseases, nutritional stress, and unhygienic environmental conditions. These disruptions are particularly detrimental during infancy due to the vulnerability. Therefore, to mitigate the adverse outcomes public health interventions must prioritize the protection of maternal and child health during and immediately after flooding. This includes ensuring access to clean water, nutritional supplements, deploying mobile health clinics and sometimes maintaining continuity of antenatal and postnatal care services even in displaced settings. Low household income seems to limit access to adequate nutrition, healthcare, and safe living conditions which are critical for optimal child growth. Maternal education, in particular, plays a pivotal role by shaping health-seeking behaviours, caregiving practices, and dietary knowledge. Children from households with educated mothers and stable economic conditions exhibit more favourable growth outcomes, regardless of flood exposure. These findings have practical implications highlighting that in flood prone areas interventions must go beyond immediate responses and should include long-term, development-focused strategies. Post-flood recovery programs should also incorporate nutrition rehabilitation, growth monitoring, and education campaigns targeted at mothers. Conditional income support programs and school-based nutrition programs can also be added to buffer the long-lasting socioeconomic vulnerabilities that are responsible for poor child development. In conclusion, the dual burden of critical environmental stressors and persistent socioeconomic shortcoming necessitates a combined approach to child welfare. While immediate responses should aim to reduce immediate mortality risks during floods, long-term policies must address the underlying social determinants of health to improve developmental outcomes among children in disaster-prone settings.

Limitation of the study

This cross-sectional study is limited by the absence of full life-cycle assessment that restricts causal inference and limits insights into long-term health trajectories. This is not a part of follow-up or longitudinal study that can give more accountable outcome. Cultural continuity could not be thoroughly examined owing to the lack of ethnographic follow-up. Additionally, reliance on self-reported data, particularly maternal recall, might have introduced potential recall bias. Environmental exposure was also assessed retrospectively rather than through real-time monitoring. The study also did not include genetic or epigenetic data, constraining the scope for biological interpretation. Future research should incorporate longitudinal, life-course, and molecular approaches to better understand bio-cultural adaptation and health in hazard-prone populations.

Author contributions

The author was solely responsible for the study design, data collection, statistical analyses, and manuscript preparation. The Anthropological Survey of India did not provide funding or contribute to the preparation of the manuscript. However, the author gratefully acknowledges the support and encouragement received from Prof. B.V. Sharma, Director, Anthropological Survey of India. Special thanks are also extended to Prof. Farida Ahmed Das of Dibrugarh University for her guidance and continuous support during the data collection phase.

Data availability

The data and materials supporting the results or analyses presented in the paper will be available upon reasonable request at msikdar@hotmail.com.

Competing interests

The authors declare no competing interests.