Historic drought in northern Kenya shifts diet and mobility away from traditional pastoralism as resource security and nutrition falter

Abstract

The Greater Horn of Africa suffered the worst drought in >40 years between late 2020 and early 2023. Using a prospective, interdisciplinary unbalanced panel study between 2019–2024 among Daasanach pastoralists in northern Kenya, we measured the effects of and recovery from this drought in relation to experiential and objective measures. We examined water and food insecurity, coping and dietary strategies, and nutritional status among 515 adults and 384 children. Livestock herds dropped precipitously with little recovery by 2024. Individuals’ consideration of moving due to water problems alongside experiential water and food insecurity scores increased substantially during the drought indicating severe distress but returned to pre-drought levels by 2024. The drought led to lower mobility, which continued post-drought for Daasanach and their remaining animals, indicating increased sedentarization. Milk intake declined drastically but rebounded by 2024. In contrast, meat intake declined and stayed lower a year post-drought, while consumption of previously stigmatized fish increased. As the drought progressed, nutritional status worsened. Children who consumed milk had significantly better nutritional status over time. Compared to men, women experienced greater levels of water and food insecurity and greater declines in adiposity at the drought’s peak. Lower socio-economic standing was associated with worse outcomes except nutritional status over time. Overall, experiential measures of water and food insecurity recovered faster than livestock and nutrition, while shifts in livelihood and dietary strategies persisted post-drought. Policymakers should use these findings to inform drought mitigation plans as diversification of diet and livelihoods may help ameliorate negative impacts.

1. Introduction

Extreme climatic events are increasing in frequency and negatively affecting people’s lived experiences, well-being, and health (Ebi & Bowen 2016, Rosinger 2023). Many climatic events affect human biology through increased physical, nutritional, and psychosocial stress and result in worsened measures of well-being (Lawton et al 2023, Reed et al 2022, Rosinger et al 2023). Yet, droughts are a unique extreme climatic event due to their slow onset. Several studies have demonstrated the detrimental impacts of drought for food insecurity and malnutrition across populations (Alulu et al 2024, Bauer & Mburu 2017, Cooper et al 2019a, Prall & Scelza 2023) as well as greater risk of long-term stunting if the drought is experienced in early life (Cooper et al 2019a, Hyland & Russ 2019).

Hot semi-arid environments are particularly susceptible to drought due to environmental limitations, including highly seasonal patterns of precipitation, limited water and vegetation, and high temperatures (Little 1989). However, lacking to date are high resolution data on how people’s lived experiences relate to resource security, livelihood, and change in coping strategies from prior to, during, and after droughts. Such data are necessary to understand the mechanisms through which extreme climatic events become embodied in poor health and malnutrition. Knowledge about these mechanisms and local processes is critical for policymakers to understand so that interventions, adaptations, and resilience plans can be designed to cope with extreme climatic events, particularly for marginalized groups (Pörtner et al 2022).

This research helps fill this gap. This paper leverages a unique natural laboratory by drawing on a longitudinal study in the Turkana basin of northern Kenya – a climate change hotspot (Muheki et al 2024, Passey et al 2010) – which began data collection in 2019 before the onset of a historic drought that occurred in the region between late 2020 to 2023.

1.1. Pastoralism, droughts, and resource insecurity

Classic anthropological research discusses pastoralism as a livelihood strategy that developed as an adaptive strategy to marginal and patchy environments and is interwoven with climatic changes (Galvin 2009, Mccabe 1990, Ndiema et al 2010). Pastoralists use several behavioral adaptations and coping strategies to navigate these complex environments (Bollig 2006b). These include mobility (often in prescribed seasonal patterns (Little 1989)), long-term cultural knowledge of the local ecosystem, herd size management through increased selling of livestock and formal exchange networks, and dietary use of livestock – increased slaughter and changes in dietary habits (Bollig 2006b, Galvin 2009).

As a result, over time pastoralists developed cultural institutions including sharing networks (Ford et al 2023) for mitigating against drought and water scarcity (Galvin 2009) to minimize vulnerability (Bollig 2006a). Michael Little (Little 1989) contended that droughts are a part of life for pastoralists, and that it is critical to understand their behavioral responses and the biological effects of droughts. Thus, as a livelihood strategy, pastoralism should be resilient against droughts. Yet, climate change is making rainfall more unpredictable and droughts longer (Cook et al 2018).

The experiences and perceptions of pastoralists as they navigate growing uncertainty with drought is critical to understand, since it may upend assumptions based on past decades of work (Pike & Williams 2006). Prior work has demonstrated that despite long-term experience inhabiting marginal environments without reliable access to clean water, pastoralists experience high levels of water insecurity (Pearson et al 2015, Roba et al 2025). Water, rainfall, and climate change are deeply interwoven with pastoralist emotional well-being (Cooper et al 2019b). Further, thinking about moving because of water problems (water availability, access, and quality) but not being able to move increases stress (Rosinger et al 2024). Even greater environmental uncertainty, increasing market integration, and national governmental policies pushing pastoralists towards sedentism may impact how pastoralists respond to the stress of dealing with livestock death and increasing food and water insecurity (Fratkin 2019, Galvin 2009, Rosinger et al 2024).

1.2. Effects of droughts on pastoralists’ well-being

Exposure to drought affects health and becomes embodied, or gets under the skin, in ways that are both observable and less visible (Ebi & Bowen 2016, Krieger 2005). Some of the more visible impacts of drought are the aforementioned nutritional status effects (Bauer & Mburu 2017). Experiential measures of food and water insecurity can help connect the visible and invisible effects of drought on well-being (Rosinger & Young 2020).

While the nutritional and economic effects of several droughts have been studied among pastoralist populations (Butt et al 2009, Fratkin & Roth 1990, Kirui et al 2022, Mccabe 1990, Straight et al 2022), little work to date has provided in-depth longitudinal examination of pastoralists’ lived experiences related to food and water insecurity during a drought alongside objective measures of nutritional status and corresponding changes in diet as well as coping strategies. Experiential measures of food and water insecurity are crucial for understanding effects of extreme climatic events by helping to connect the visible (e.g., malnutrition) and invisible (e.g., stress) effects of drought on health and well-being (Rosinger & Young 2020). Water insecurity is defined as lacking access to safe and sufficient water for a healthy, active, and culturally-appropriate lifestyle (Jepson et al 2017).

Recent work examining Himba pastoralists in Namibia demonstrated that drought years negatively affected dietary diversity, the availability of sour milk, and likely resulted in increased food insecurity (Prall & Scelza 2023). As pastoralists traditionally rely on milk from livestock for both calories and sources of hydration, livestock death from droughts can pose large challenges to food and water insecurity and act as a mechanism to worsened well-being.

Across East Africa, several drought monitoring programs have been designed as early warning systems (Mugabe et al 2019). In Kenya, the National Drought Management Authority (NDMA) tracks several metrics including, rainfall anomalies, livestock body conditions, vegetation index, food and water conditions, and nutrition monthly (Authority 2022). These warning systems are used by policymakers to address drought effects by intervening through food aid, water development projects, disease outbreak response, cash transfers, and restocking programs.

1.3. Greater Horn of Africa 2020–2023 drought

The Greater Horn of Africa experienced the worst drought in >40 years between November 2020 and March 2023 (Department 2024, World Health Organization 2023). In northern Kenya, the drought began in late 2020 with moderate drought in 2021, severe drought in 2022, and peaked in extreme drought between December 2022 and February 2023 resulting in the highest “emergency” rating by the NDMA (Authority 2023). Recovery began in March-May 2023 when the long rains reached average precipitation for the first time in three years and were followed by a full year of above average rains through 2024. The fourth and fifth consecutive failed rainy seasons (taking into account their dual rainy season) resulted in the deaths of hundreds of thousands of livestock and acute food insecurity for an estimated 5.4 million people in the region (World Health Organization 2023). It is assumed that extreme climatic events affect resource insecurities, but measured lived experiences alongside objective nutrition data and responses to these events are scant in the literature as they have been mostly limited to measuring food insecurity (Rosinger et al 2023). For example, in Kenya, there has been systematic effort to assess how experiential food insecurity and nutritional outcomes respond to drought (Bhavnani et al 2023, Kipkorir et al 2024), but experiences of water insecurity using a validated scale are not yet collected as part of national drought monitoring.

Here we advance the field by drawing on rich panel survey, dietary, and anthropometric data from Daasanach semi-nomadic pastoralists in northern Kenya between 2019–2024 alongside climatic and animal GPS mobility data. We test the effects of and recovery from this historic drought on indicators of well-being, coping strategies, and nutritional outcomes of adults and children. We further test whether the drought had a greater negative effect on women than men as existing gender inequities often are exacerbated in patriarchal societies when resources become scarce (Balehey et al 2018) and how perceived socioeconomic status (SES) played a role. This case study helps document how lived experiences and resource insecurities in already under-resourced settings are affected by extreme drought and what the recovery process looks like. In this way, this case study helps shed light on the impacts of extreme weather events so that policies can be implemented to avert such effects on local communities.

2. Methods

2.1. Field site

This research took place in northern Kenya among Daasanach semi-nomadic pastoralists, who live on the eastern side of Lake Turkana near the Ethiopian border. Approximately 19,000 Daasanach live in and around Illeret, Marsabit County in 26 villages (Kenya National Bureau of Statistics (KNBS) 2019). A further ^~40,000 Daasanach live on the Ethiopian side of the border. Daasanach living in Kenya have been more reliant on pastoral forms of subsistence than those in the Omo River Basin of southern Ethiopia (Almagor 1978) where they have greater access to market and agricultural goods; however, access to market goods has increased in recent years (Swanson et al 2023a). See supplemental materials for additional details about Daasanach (Appendix).

Even during non-drought years in this hot, semi-arid environment, water and food insecurity and dehydration are major sources of concern for Daasanach (Bethancourt et al 2022, McGrosky et al 2025). These resource insecurities are worsened by increased climate variability in the region (MoALF 2017). The study site (Marsabit County) has been rated as the worst within Kenya in terms of social vulnerability index due to low access to healthcare, sanitation, education, and other services, yet it rated as the lowest risk in terms of epidemiological vulnerability index due to the lack of population density (Macharia et al 2020).

Data for this paper come from the Daasanach Human Biology Project, which is an ongoing panel study measuring water and food insecurity, mobility, stress, energy expenditure, and nutrition among Daasanach. It began in 2019 and aims to understand changes in these factors over time by conducting annual surveys (Bethancourt et al 2022, Rosinger et al 2025). Household surveys inquired about household-level and individual-level questions related to water, food, sociodemographics, mobility, nutrition, and health with up to two household heads (male and female) with up to two children interviewed/measured (Bethancourt et al 2022, Rosinger et al 2025). The surveys were followed by anthropometric, biomarker, and environmental measurements, including ambient weather conditions and water quality analyses.

The study aimed to carry out annual study waves beginning in June-July 2019; however, survey data collection was halted in 2020 and 2021 due to COVID-19 and the shutdown of international travel. Thus, the second, third, and fourth study waves were conducted in June-July of 2022, 2023, and 2024, respectively. In 2019, we enrolled adults in seven communities at different distances to the main market town of Illeret to capture variation in access to market goods and water sources. Three communities were in and around Illeret, three communities were 4–9 km away from Illeret, and one community was 20 km away (Figure S1). The initial 2019 survey used a random sampling design of every third household, which resulted in sampling 242 adults and 132 children (of all ages) in 132 households (Bethancourt et al 2021). In 2022, we attempted to re-contact the 2019 sample and re-enroll them; we enrolled neighbors of the original household with similar characteristics if they were absent during data collection to refresh the sample. In 2023, we again attempted to re-enroll all original participants from 2019 and any new participants from 2022; we also expanded to an 8^th village that provided a mix of characteristics of the other communities (Rosinger et al 2025). In 2024, we attempted to re-enroll all participants from prior waves and any replacements along with allowing for new participants to enroll to reflect changes in community characteristics.

Prior work to date examining the effects of this drought among Daasanach have used cross-sectional waves of the data ending in 2022 (Roba et al 2025, Rosinger et al 2024). Thus, this represents the first full longitudinal analysis of the drought’s effects from before to after the drought ended.

Total observations and analytic sample sizes between 2019–2024 for adults were between n=960–978 observations (depending on the survey variable) for 514–515 household heads aged 18–83 on water and food insecurity and survey data; n=882 observations on 487 household heads with anthropometric data (restricting to those aged 18+ and excluding pregnant women). For children, we restricted analyses to those aged 2–9 years to capture the most vulnerable group to nutritional shocks. Height and weight were collected on n=510 observations on 384 children; and slightly fewer mid-upper arm circumference data (n=489 observations on 371 children). Thus, this resulted in an unbalanced panel with replacements to refresh the panel. As we used refreshment enrollment to maintain sample size and village characteristics, 51 adults were measured in all four waves, 89 were measured in three of four waves, 132 were measured in two waves, and 217 were measured in one wave. Loss-to-follow-up after the 2019 wave (defined as not participating again) was 25.5% for adults and 43.1% for children (See Figure S2 for sample sizes, enrollment, and additional details on attrition for the full study).

In 2024 we conducted six focus groups (three with men and three with women) in two communities about the drought and changes in livelihood strategies. While ongoing work is analyzing the qualitative results thematically from those FGDs (Roba et al Under Review), here we use them to provide additional contextual information about the impact of the drought to situate the quantitative survey results. Finally, we used participant observation and stakeholder discussions from fieldwork conducted in 2025 and 2026 to provide additional context for quantitative results.

The research was approved by Pennsylvania State University’s Institutional Review Board and the Kenyan Medical Research Institute. Permission was also obtained from the Director of Health in the county government of Marsabit, Kenya and from community leaders in all the communities sampled. All study procedures were explained through Daasanach translators and a local community health volunteer. All participants provided written (via signature if literate or fingerprint otherwise) and verbal informed consent, while parents provided consent for their children and children provided assent.

Data availability statement:

As these data come from an Indigenous population, requests for individual-level, non-GPS, de-identified data can be made from the corresponding author with institutional IRB permission detailing research question to minimize the risk to our study participants and ensure their privacy.

2.2. Measures

2.2.1. Indicators of pastoral lifestyle

Livestock:

During each wave of the survey, we asked households about all livestock the household owned, including the number of cows, goats and sheep (combined in 2019 and 2022, and separately thereafter), camels, donkeys, and other livestock. We report on individual species separately since there are different uses for animals and resilience to drought. This allows us to test how individual species were affected by the drought and how this may be tied into differences in mobility in the instance of donkeys or milk availability since Daasanach drink goat milk more frequently and goats are often kept closer to houses whereas cattle are kept at pasture.

Mobility:

To assess mobility and capture migration, we asked households the number of times they moved for more than a week at a time in the prior year as described elsewhere (Rosinger et al 2024). These trips did not include short trips to the market in Ethiopia to buy food or visits to other communities, which typically last 1–4 days.

Daily range of select livestock:

We attached passive GPS receivers (CatLog Gen2 by Mr. Lee Science Line) to nylon collars which were carried by livestock (cows, n=17; goats, n=23) during field seasons between October 2020 and July 2023 (See supplemental materials for additional details). These were set to record animal locations at 10-minute intervals and were recovered during the same or subsequent field seasons when the collared animal and its owner could be located. Total recorded periods ranged from 19 to 632 days (Figure S3). Due to high levels of retreat and mortality among cattle herds during the drought, several collars were not recovered, including all those deployed during and after the February 2022 season.

2.2.2. Experiential measures of well-being

Water Insecurity:

Each adult household head was asked the 12-item Household Water InSecurity Experiences (HWISE) scale separately. The HWISE scale has been validated in low- and middle-income countries (Young et al 2019), and was translated and implemented among Daasanach (Bethancourt et al 2022, Ford et al 2023). The HWISE items describe the frequency of twelve different water-related experiences (e.g., worrying about one’s water situation, inability to wash hands) that occurred over the past four weeks. Score of 0, 1, 2, or 3 are given to answers of never (0 times), rarely (1–2 times), sometimes (3–10 times), or often/always (11+ times), with a range of 0–36. The Cronbach’s alpha between 2019–2024 was 0.85 indicating good reliability.

Mobility ideation:

To assess mobility ideation or rumination about moving due to water problems, we asked adults: “how frequently did you think about moving dwellings in the past month due to the water situation?” The development of this question is described further elsewhere (Stoler et al 2026) and the response categories were the same as the HWISE scale. This question helps capture sentiments driven by shifts in climatic factors (Rosinger et al 2024, Stoler et al 2026).

Food Insecurity:

The 9-item Household Food Insecurity Access Scale (HFIAS) (Bethancourt et al 2022, Coates et al 2007) was similarly asked of each adult separately. This widely used scale has also been used among Daasanach as described elsewhere (Bethancourt et al 2022). It addresses the frequency in the prior four weeks with which any household member experienced nine different dimensions of food insecurity, including anxiety and uncertainty of food access, inadequacy of food quality, insufficient food intake, and hunger. Like the HWISE scale, scores of 0, 1, 2, or 3 were given to answers of never, rarely, sometimes, or often/always, respectively, for a range of 0–27. The Cronbach’s alpha between 2019–2024 was 0.80 indicating good reliability.

2.2.3. Dietary intake

Milk intake frequency:

In the survey, we asked individuals (adults and children) how frequently they consumed fresh plain and sour milk per week with the response options 0 times, 1–3 times, 4–6 times, once per day, 2 times per day, 3 times per day, or 4 or more times per day.

Meat and fish intake:

For adults, we asked how many days in the prior week they had eaten meat and, separately, fish with the variables ranging 0–7 days. We then created a dichotomous variable of any fish intake in the prior week.

2.2.4. Nutritional Status

All participants had their heights measured on a Seca stadiometer and weights measured on a Tanita bioelectrical impedance analysis (BIA) scale, which were both placed on a hard, flat surface without shoes, hats, and in light clothing. To assess adiposity, body mass index (BMI) was calculated (kg/m²), and the percent total body fat was used from the BIA scale. Finally, participants had triceps, biceps, subscapular, and suprailiac skinfolds measured using Lange skinfold clippers and these four skinfolds were summed (SUM4) following methods by Lohman (Lohman et al 1988).

For children aged 2–9, we used WHO age- and sex-specific reference curves to calculate z-scores for weight-for-age (WAZ), height-for-age (HAZ), BMI-for-age (BMI-Z), and classified stunting as HAZ<−2 and wasting as WAZ<−2 (World Health Organization 2006). While Daasanach children generally fall below the 50th percentile for WAZ and weight-for-height, their measures track with WHO centile curves versus age (Swanson et al 2023b). The specific values (z-scores) reported here may differ if Daasanach-specific growth curves were employed but the patterns would remain the same as WHO-reference values reliably indicate if a child’s nutritional status is changing over time. Since nutritional supplementation in Illeret was based on malnutrition as defined by the WHO curves, we thus report those here to allow for a comparison to Kenya wide malnutrition reporting and global references. For children, we also measured mid-upper-arm circumference (MUAC) rounding to the nearest 0.1 cm.

2.2.5. Primary Predictor: Timeline of drought

We analyze trends in time by using time fixed effects for the year data were collected and test whether measures changed pre-, during, and post-drought. We use 2019 data as the pre-drought reference year. However, we note that the first six months of 2019 also received low rainfall, but it did not classify as a drought, whereas the first six months of 2020 provide an average, non-drought year (Figure 1; Figure S4). The onset of the drought began with the missed short rains in November-December of 2020 and lasted until April 2023. The 2022 survey wave occurred during the height of the drought, the 2023 wave occurred at the end of the drought, and the 2024 wave occurred one year following the end of the drought, which included above-average rains (Figure 1).

Figure 1.

Monthly precipitation in research site using Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) data and a local weather station between 2010–2024 in Illeret, Kenya. Notes: Gray area indicates drought period. Data at Illeret station only available until end of 2022.

2.2.5. Covariates

In trend analyses we adjust for participant age, sex, perceived socioeconomic status (SES) via the MacArthur ladder which asks individuals to rank themselves from the top rung (1) to the bottom (10) (Giatti et al 2012), household size as the number of family members who live under the same roof as additional individuals require extra resources, and community of residence to adjust for time-invariant unobserved characteristics of those communities and address potential omitted variable bias (e.g. proximity to town and Lake, water source access) following analyses of panel data (Godoy et al 2010). Depending on the outcome assessed and pathways, we additionally adjusted for BMI, water insecurity, food insecurity, and number of times moved.

2.2.6. Statistical analyses

All analyses were conducted using Stata V19 (College Station, TX). To test for changes in continuous and dichotomous outcomes, we used the econometric approach of panel random effects linear and logistic regression models with the individual as the group variable and random effects to account for repeated measurements and time fixed effects to model the drought, respectively (see supplemental materials for additional details) (Wooldridge 2015). For ordinal outcomes (mobility ideation and milk intake), we used panel ordered logit regressions exponentiated to provide odds ratios. Robust standard errors were clustered at the community level to account for clustering of observations within communities and any potential heteroskedasticity. Random effects allow inclusion of time invariant variables in the model, such as sex, for an examination of how drought effects varied for men and women. For adult models we adjusted for the covariates listed in section 2.2.5. For adults, apart from livestock and mobility models which provided household level data, we tested interactions between year and sex retaining the interaction terms if they were statistically significant. We also present the sex and perceived SES coefficients to show how these factors were related to changes over the drought. To test for linear and quadratic time trends, we used post-estimation contrasts. Predicted values of outcomes were visualized using the marginal standardization method.

Calculation of daily ranges of livestock was performed using the R statistical computing platform (Team 2016) and the packages moveHMM (Michelot et al 2016) and splancs (Bivand et al 2017). Specifically, the GPS units record longitude and latitude (WGS 84) at regular intervals. First, location point data from each collar were subset to daytime hours (06:00 – 20:00) and then aggregated by day. Livestock daily ranges were derived by creating a polygon from a convex hull around each day’s GPS coordinates and then calculating area (in km²) of the polygon. We used this output to plot the LOESS curve of the daily range.

For children, we adjusted models for age, sex, household size, and community residence. We conducted a post-hoc model in which we tested how milk intake (any in the past week compared to none) was associated with nutritional outcomes for children (14 children had missing milk intake data) and plotted trends stratified by milk intake.

To test the sensitivity of the results and potential attrition bias/changes in the sample characteristics due to the refreshment samples that were not observed throughout the full study period, we re-estimated analyses restricting models of adults to individuals who were present in the baseline 2019 wave and at least two additional times in the 2022–2024 survey waves. This results in the exclusion of the eighth community that was added to the study in 2023. As we have limited repeated observations data on children (as many aged out), this sensitivity analysis was not possible for children’s models.

3. Results

Sample characteristics and demographics of Daasanach adults and children aged 2–9 are presented by survey wave and overall for all observations in Table 1.

Table 1:

Descriptive characteristics of Daasanach adult and children by year between 2019–2024

Adults	2019 (n)	Mean	SD or n	2022 (n)	Mean	SD or n	2023 (n)	Mean	SD or n	2024 (n)	Mean	SD or n	Total (n)	Mean	SD or n
Age (years)	232	39.7	14.8	182	42.2	15.9	248	40.6	15.6	312	39.3	15.0	974	40.2	15.3
Male (%)	232	45.3 %	105	182	42.9 %	78	248	37.1 %	92	312	37.5 %	117	974	40.2 %	392
Water insecurity
(HWISE) score	232	20.5	6.6	182	22.7	5.2	248	25.3	6.1	312	20.1	5.4	974	22.0	6.2
No-to-low WI	3	1.3 %	3	0	0%	0	1	0.4 %	1	1	0.3 %	1	5	0.5 %	5
Low WI	8	3.4 %	8	5	2.7 %	5	9	3.6 %	9	24	7.7 %	24	46	4.7 %	46
Moderate WI	141	60.8 %	141	103	56.6 %	103	63	25.4 %	63	205	65.7 %	205	512	52.6 %	512
High WI	80	34.5 %	80	74	40.7 %	74	175	70.6 %	175	82	26.3 %	82	411	42.2 %	411
Food insecurity (HFIAS)
score	232	17.6	4.9	181	17.6	4.0	248	21.2	3.5	312	16.6	3.5	973	18.2	4.3
Severely FI	225	97.0 %	225	180	99.4 %	180	248	100 %	248	306	98.1 %	306	959	98.6 %	959
Household size	232	7.4	2.7	182	6.9	2.7	248	5.9	2.9	312	6.3	2.6	974	6.6	2.8
Times moved	232	4.3	6.1	180	3.3	3.7	240	2.2	3.7	308	2.1	3.5	960	2.9	4.4
Meat	232	1.3	1.4	182	1.0	1.1	248	1.1	1.1	312	1.0	0.9	974	1.1	1.1
Fish	232	0.8	1.6	182	1.2	1.7	248	1.2	1.6	312	1.7	1.9	974	1.3	1.8
Weight (kg)	219	51.5	9.4	164	50.8	9.6	220	50.8	8.4	282	51.4	8.9	885	51.2	9.0
Height (cm)	219	167.9	7.8	164	167.0	8.2	220	167.6	8.2	282	167.8	8.2	885	167.6	8.1
BMI	219	18.3	3.2	164	18.2	2.9	220	18.0	2.3	282	18.2	2.8	885	18.2	2.8
SUM of 4 skinfolds (mm)	219	30.6	19.7	162	32.3	17.4	217	30.0	14.4	281	34.5	16.7	879	32.0	17.2
Body fat percent (BIA)	213	16.2	8.6	159	16.3	8.6	214	17.2	8.2	279	19.4	8.3	865	17.5	8.5
Children	2019 (n)	Mean	SD	2022 (n)	Mean	SD	2023 (n)	Mean	SD	2024 (n)	Mean	SD	Overall (n)	Mean	SD
Age (years)	72	7.1	1.4	103	6.4	1.8	147	6.3	1.8	188	5.9	1.9	510	6.3	1.8
Male (%)	72	37.5 %	0.5	103	48.5 %	0.5	147	48.3 %	0.5	188	48.4 %	0.5	510	46.9 %	0.5
Weight (kg)	72	18.5	3.9	103	16.9	4.0	147	16.7	3.8	188	16.1	4.2	510	16.8	4.1
Height (cm)	72	118.0	10.5	103	113.7	13.8	147	113.4	12.9	188	109.9	13.8	510	112.8	13.4
WAZ	72	−1.64	1.06	103	−1.88	1.01	147	−1.92	1.06	188	−1.76	1.06	510	−1.81	1.05
HAZ	72	−0.65	1.35	103	−0.79	1.31	147	−0.73	1.28	188	−0.86	1.27	510	−0.78	1.29
BMI-Z	72	−1.87	0.83	103	−2.14	0.97	147	−2.26	1.02	188	−1.87	1.02	510	−2.04	1.00
MUAC (cm)	71	15.3	1.3	95	14.7	1.2	144	14.5	1.2	181	14.6	1.3	491	14.7	1.3
Stunted (<-2 HAZ)	72	11.1 %	0.3	103	19.4 %	0.4	147	17.7 %	0.4	188	17.0 %	0.4	510	16.9 %	0.4
Wasted (% <-2 WAZ)	72	36.1 %	0.5	103	43.7 %	0.5	147	40.1 %	0.5	188	35.1 %	0.5	510	38.4 %	0.5

WAZ: weight-for-age Z-score; HAZ: height-for-age Z-score; BMI-Z: BMI-for-age Z-score; MUAC: middle-upper arm circumference (cm); No-to-low WI: 0–2 HWISE score; Low WI: 3–11 HWISE score; Moderate WI: 12–23 HWISE score; High WI: 24–36 HWISE score; Severely FI: defined per Coates 2007. Total column refers to all observations from 515 unique adults and 384 unique children.

3.1. Livestock numbers plummeted during the drought with slow recovery

Prior to the drought in 2019, households had 24.6 (SE=3.0) sheep and goats, which declined steeply in 2022 and slightly further in 2023 to an average of 8.0 (SE=2.4) (Figure 2a; Table S1). In 2024, a year post-drought, there was a partial recovery as the mean number of sheep and goats per household increased to 13.9 (SE=2.3), but was still significantly lower than 2019 (P=0.018). A similar pattern was observed with cattle as the average household had 6.2 (SE=0.9) cows, which declined to 1.5 (SE=0.3) in 2022 (P<0.001) and bottomed out at 1.0 (SE=0.6) in 2023 (P=0.001). A slight non-significant increase occurred in 2024 to 1.5 (SE=0.4) as cattle are more expensive and take more capital to build back stock (Figure 2b; Table S1). Arid adapted livestock showed a more modest decline. The mean number of donkeys stayed fairly stable from 2019–2024 at close to 1 (SE=0.2) per household except for a marginal decline in 2023 (B=−0.52, SE=0.32; P=0.10) (Table S1). The number of camels began low (mean 0.11 [SE=0.1]) and increased non-significantly to 0.44 (SE=0.07; P=0.085) in 2022 as the result of a local government intervention which purchased camels during the drought as well as raiding and stayed above 2019 levels in 2023 and 2024 as the drought subsided (Table S1). Across species, each step lower on the MacArthur ladder (corresponding to lower perceived SES) was significantly associated with fewer livestock over the drought (Table S1).

Figure 2.

Predicted mean and 95% confidence intervals of the number of household A) goats and sheep, and B) cows; C) frequency of mobility ideation due to water problems, D) number of moves in the prior year among Daasanach household heads, 2019–2024; and LOESS of daily ranges of GPS monitored Daasanach livestock in square kilometers of E) cows, and F) goats. Note: Note: reference lines indicate drought onset in November 2020 and drought cessation in March 2023. Note logarithmic scale on y-axis for panels e and f.

3.2. Drought increased mobility ideation, yet actual mobility declined

We asked household heads how frequently they thought of moving due to water problems. The shift in responses indicates heightened mobility ideation during (OR=3.0 [95%CI:1.62–6.68]; P<0.001) and at the end of the drought (OR=4.24 [2.28–7.89]; P<0.001) compared to pre-drought, which reverted back to pre-drought frequencies one-year post-drought (Table S2; Figure 2c). In contrast to the pattern of ideation which is experiential and expresses dissatisfaction with the environment, the reported number of moves in the prior year declined from 3.9 (SE=0.6) in 2019 to 2.2 (SE=0.3; P=0.035) by 2023 and remained marginally lower in 2024 (2.3; SE=0.3; P=0.057) post-drought (Figure 2d; Table S2). This indicates lower mobility, likely tied to lower number of overall livestock, which have been consolidated and necessitate fewer people moving with the herd as well. Further, adults with lower perceived SES moved less frequently over the drought (Table S2).

Figures 2e-2f illustrate distinctions between daily ranges during the drought for GPS-collared cows and goats. In late 2020, the average range for both livestock species was around 1 km². Cows, which are grass dependent, showed a marked uptick in mobility as range conditions deteriorated. This was followed by a striking decline in late 2021 preceding widespread cattle mortality and the cessation of cattle GPS data collection. In contrast, monitored goats illustrated a stepwise decline in range over the period of observation–average ranges in early 2023 being around half of that in late 2021.

3.3. Water and food insecurity tracked the drought

From 2019 to 2024, water insecurity (WI) worsened significantly from pre-drought through the end of the drought in 2023. The WI score increased from already high levels of 20 by 2.6 (SE=1.1; P=0.019) points in 2022 and 6.6 (SE=1.7; P<0.001) points in 2023. In 2024, the mean WI score improved and was no longer significantly above 2019 levels (Figure 3a; Table S3). Additionally, there was a gendered effect of the drought on WI as women experienced significantly higher WI levels compared to men by the end of the drought in 2023 (2.4; SE=0.69; P=0.001) and post-drought (1.4; SE=0.74; P=0.064) (Figure 3a). The distribution of WI shifted to the right, worsening from 2019 to 2022 and 2023 with the percent of those experiencing high WI (≥24 HWISE score) increasing from 33.8% to 40.9% to 71.0%, respectively; and then improved in 2024, as the distribution shifted back and the percent with high WI declined to 27% (Figure 3b).

Figure 3.

Predicted mean and 95% confidence intervals of A) household water insecurity and B) distribution of household water insecurity experiences scores, and C) food insecurity and D) distribution of food insecurity scores stratified by sex among Daasanach household heads, 2019–2024. Note: reference lines in (A) and (B) indicate drought onset in November 2020 and drought cessation in March 2023; and in (C) indicate low, moderate, and high levels of water insecurity.

Food insecurity (FI) showed slightly different patterns than WI as the predicted mean HFIAS score did not change significantly in 2022 from pre-drought levels – though in 2019, 98.4% of households were already classified as severely food insecure. Further, in 2022 food aid delivery ameliorated the situation. However, in 2023 when food aid was delayed due to rains and flooding, FI worsened significantly by 4.8 points (SE=1.0; P<0.001). In 2024, food security improved, returning to pre-drought levels (Figure 3c; Table S3). Again, there was a significant interaction with gender, where women experienced significantly higher food insecurity than men in 2022 (B=1.47; SE=0.52; P=0.006). Each step lower in perceived SES was associated with 0.6 points (SE=0.17; P<0.001) higher FI over the drought (Table S3). We observed different shifts in the distribution of FI over the drought as there appeared to be a wider distribution of FI scores in 2022 – potentially highlighting uneven impacts of the drought on households or differential access to food aid, which worsened in 2023, and then uniformly improved in 2024 (Figure 3d).

3.4. Milk and meat intake declined during the drought, while previously stigmatized fish intake increased

The drought produced large-scale dietary changes. First, the frequency and odds of drinking milk in the prior week for children declined by 82% (OR=0.18; [0.06–0.50]; P=0.001)] in 2022 compared to 2019, began to recover some in 2023, and then recovered to pre-drought levels in 2024 (Figure 4a; Table S4). There was a similar trend among adults as the odds of milk consumption were 79% (OR=0.21; [0.09–0.50]; P<0.001)] and 76% (OR=0.24; [0.10–0.55]; P=0.001)] lower in 2022 and 2023 and then recovered in 2024 (Figure 4b, Table S4).

Figure 4.

Milk (plain and sour) consumption among Daasanach A) children and B) adults between 2019–2024; and predicted number of days and 95% confidence intervals eating C) meat and D) fish in the prior week among Daasanach adults 2019–2024. Note: reference lines indicate drought onset in November 2020 and drought cessation in March 2023.

Among adults there was a quadratic trend (Chi-2=17.6; P<0.001) of declining days eating meat in the prior week between 2019 and 2024, which overall dropped by 0.22 days. In contrast, there was a linear trend of increasing fish intake between 2019–2024 (Chi-2=44; P<0.001; Figure 4c; Table S5). At the peak of the drought in 2022, adults ate fish on 0.5 (SE=0.17; P=0.003) more days than in 2019. This remained slightly elevated in 2023 but was no longer significantly higher than 2019, but then further increased in 2024 (B=0.60; SE=0.18; P=0.001) as the stigma surrounding eating fish continued to decline and the community increased commercial fishing to access the market economy (Figure 4d; Table S5). Adults with lower perceived SES consumed meat (B=−0.10; SE=0.02; P<0.001) and fish (B=−0.11; SE=0.05; P=0.011) on fewer days over the drought (Table S5). The proportion of adults eating any fish in the prior week increased from 38.8% in 2019 to 62.9% in 2022 and remained elevated in 2023 and 2024 (Figure S5).

3.5. Nutritional status declined during the drought, while adiposity partially rebounded post-drought

Children aged 2–9 years demonstrated worsening nutritional status during the drought with a marginal rebound one-year post-drought. Specifically, a quadratic trend (P<0.001) emerged between 2019–2024 in weight-for-age z-score (WAZ) as it declined significantly in 2022 (−0.25 z-score; SE=0.11; P=0.016) and further in 2023 (−0.30 z-score; SE=0.12 P=0.011) compared to 2019 and slightly rebounded in 2024 (−0.20; SE=0.12; P=0.096) (Figure 5a; Table S6). There was a significant linear trend in declining height-for-age (HAZ) (Chi-2=14.1; P<0.001) between 2019–2024 as HAZ declined by −0.18 Z-score (SE=0.09; P=0.044) in 2022, slightly rebounded in 2023, and then declined non-significantly in 2024 (Figure 5b; Table S6). Like WAZ, children’s BMI-Z and middle-upper arm circumference (MUAC) both demonstrated quadratic trends (P<0.001) as they declined marginally in 2022, worsened in 2023, and partially rebounded in 2024 (Figure 5c-d; Table S6). Odds of stunting and wasting were both significantly higher in 2022 compared to 2019; but while elevated in later years were not statistically different (Table S6).

Figure 5:

Predicted mean and 95% confidence intervals of Daasanach children aged 2–9 years A) weight-for-age Z score, B) height-for-age Z score, C) BMI-for-age Z score, and D) mid-upper arm circumference (MUAC); and E) weight, F) body mass index, G) sum of 4 skinfolds, and H) BIA percent body fat by sex among non-pregnant Daasanach adults aged 18 years and older between 2019–2024. Note: reference lines indicate drought onset in November 2020 and drought cessation in March 2023.

A post-hoc analysis found that children who drank any milk in the prior week did better nutritionally, as they had Z-scores 0.31 (SE=0.09; P<0.001) higher in WAZ, 0.29 (SE=0.12; P=0.014) higher in HAZ, 0.16 (SE=0.076; P=0.03) higher in BMI-Z, and had MUAC 0.18 cm (SE=0.08; P=0.019) greater over time (Figure S6).

Adult nutritional status also demonstrated quadratic trends between 2019–2024 across outcomes. Mean weights did not change significantly for men or women (Figure 5e; Table S7). However, BMI decreased in 2023 (B=−0.21 kg/m²; SE=0.08; P=0.005) with no significant interaction by sex, and slightly rebounded in 2024 (Figure 5f; Table S7). Similarly, there was a significant decline in sum of 4 skinfolds by the end of the drought in 2023 (B=−3.8 mm; SE=2.0; P=0.058) with a significant sex interaction, demonstrating a more pronounced decline for women’s adiposity than men, which rebounded for both in 2024 (Figure 5g; Table S7). Similar patterns were observed for percent body fat, as it declined by 1.0% (SE=0.32; p=0.002) by 2023 and then rebounded above 2019 levels in 2024 (B=1.6%, SE=0.45; P<0.001) (Figure 5h; Table S7). There was no association between perceived SES and adult nutritional status (Table S7).

3.6. Sensitivity analysis

Adult models were re-estimated restricting to participants (n=131 adults, 431 observations) present in 2019 and at least two additional times over the study period (3.2 observations per participant). Results were consistent with the primary results (Tables S8-S13); however, the interaction between year of drought and sex for food insecurity was no longer significant (Table S10), and in 2024, meat intake was now significantly lower while fish intake was no longer significantly higher (Table S11).

4. Discussion

Overall, this paper aimed to assess the effects of and recovery from the 2020–2023 historic drought in northern Kenya. This paper demonstrates strong negative effects on well-being, dietary intake, mobility, and nutritional status. Importantly, it highlights differential recovery rates as experiential water and food insecurity largely tracked the drought, whereas the negative effects on livestock, mobility, diets, and nutritional status persisted after the drought ended while diversification of livelihoods and diet increased. Despite the widespread nature of the drought, women, those with lower perceived SES, and children without milk access fared the worst, indicating greater vulnerability.

Livestock are interwoven into the fabric of pastoralist life. Everything from social status and wealth to diet is dependent on livestock. The widespread collapse of herds is catastrophic for these communities (Galvin 2009, M’Mbogori et al 2022). Semi-nomadic pastoralists have adapted a livelihood strategy of moving in search of pasture and water with their herds for millennia (Ndiema et al 2010). Daily ranges of livestock can vary widely (^~0–15 km²) depending on the type of animal, the condition and distribution of forage, and movement between foraging areas. We found that the daily range of animals declined during the drought though due to the total loss of cattle with GPS, we could not assess how cattle movement changed post-drought. During the drought, a camel restocking program infused drought-tolerant camels into the study area. While the number of camels has not grown significantly after this effort, our participant observations in 2025–2026 found that families are continuing to hold and breed these camels. Restocking programs serve as important interventions which can help maintain pastoralism after livestock death.

During the drought, Daasanach had higher frequency of water-related mobility ideation yet mobility declined by half. These results demonstrate how droughts can create a dissonance between wanting to move and being able to move which may increase distress. This builds upon prior work which found increased mobility ideation and water insecurity among Daasanach at the height of the drought and greater chronic stress as a result (Roba et al 2025, Rosinger et al 2024). Participants indicated many community members moved to or stayed near the main town of Illeret during the drought. In work reported elsewhere, Daasanach indicated that top reasons to move or think about moving were due to environmental and climate change issues related to searching for pasture, lack of water, flooding, or being closer to food aid (Rosinger et al 2024). Participants were not systematically asked why they did not move as much in the present study, but many participants indicated their remaining herds were small enough to be managed by fewer individuals. Further, livestock like donkeys assist with moving and the loss of animals may have created a negative feedback loop of reducing ease of mobility. These results are similar to another study that found that drought and cattle-raiding led some pastoralists to abandon herding (Schumann et al 2025). Further, it is possible that Daasanach realized that with a widespread drought the ‘grass isnť greener’ (or even existent) in other places that have served as drought refugia in the past. Future research should aim to disentangle the role of drought in changing mobility ideation, aspirations, and actual mobility to uncover barriers.

Drought worldwide has been found to increase mental health problems and stressors, often because of the economic, resource, and nutritional consequences and associated uncertainty (Sugg et al 2020). While water issues have been noted in prior research dealing with drought (Ogutu et al 2026, Straight et al 2025), this paper is the first longitudinal study to our knowledge to track how people’s experiences with water problems changed over a sustained drought using the validated HWISE water insecurity scale. The HWISE scores found among Daasanach at the end of the drought (mean=25.3; with >70% of households highly water insecure) are higher than any reported in the literature demonstrating the severity of the drought (Mir Mohamad Tabar et al 2026, Stoler et al 2021). We found a pattern of heightened mobility ideation due to water problems which increased successively during the drought and then abated with the return of the rains. This result aligns with our resource security results, which demonstrated a striking increase from already high levels of water and food insecurity prior to the drought to significantly higher levels by the end of the drought. Food aid delivery in 2022 at the peak of the drought likely kept food insecurity stable. However, in 2023, when this food aid had not been delivered for more than three months due to logistical challenges of flooding and the drought was declared over by local aid agencies, food insecurity increased significantly. Our results align with rising food insecurity patterns in Kenya over this time (Kipkorir et al 2024).

Our results demonstrate that during the drought women experienced heightened levels of WI and FI compared to men as well as worsened nutritional status. Daasanach society has clear gender divisions of labor with women primarily in charge of water, food, and gathering firewood. At mealtime, men are served first, then children, and last women. Thus, women are more likely to receive less food within a family during shortfalls and due to their roles as household managers are more aware of the nutritional and water shortfalls experienced by the household. Thus, during droughts, women may experience increased stress related to these household managerial obligations.

Further, we observed a substantial decrease in milk intake frequency for children and adults. This is especially devastating as milk is traditionally one of the largest caloric contributions for pastoralists (Galvin 1992, Little 1989). A year post-drought, milk intake increased from drought-levels but was still lower than pre-drought as the number of livestock had not recovered. This reinforces our observations and prior literature that recovery takes a substantial time among pastoralist groups (Bollig 2006b). We also observed that meat intake became less frequent and fish intake became more popular. Initially, as animals began to die, households processed hides and butchered carcasses for consumption. However, as herd losses accelerated, often occurring in a single day, the emotional toll became overwhelming. Informants explained that people eventually stopped processing the carcasses entirely, leaving animals where they fell. In 2022, our research team observed dead animals covering the landscape often with twigs and plastic in their stomachs illustrating the severity of the lack of pasture. Livestock was also not sold during this time as prior work suggests occurs during droughts as a coping strategy (Bollig 2006b) due to the widespread deaths of livestock and lack of water and pasture to sustain new herds.

Focus group discussions with Daasanach in 2024 substantiate the quantitative results. Adults reported that they learned to fish during the drought and were using fishing both as a dietary food source but also as a livelihood strategy to sell and provide income. At the beginning of our study in 2019, when we asked Daasanach about fishing, they told us that only the poorest community members who did not have livestock, referred to as “Dies” (Sagawa 2021), would fish. This long-held stigma (Sagawa 2021) was gone by 2024 (Roba et al Under Review). While fishing and eating fish was more common in the communities closer to the lake, even those with livestock in communities farther from the lake said that fishing was a good way to diversify their livelihoods and complement it. During field work in 2025–2026, these observations were substantiated as the stigma about fishing is gone and fishing activity has increased (Figure S7). Even community members in the farthest and most traditional community with livestock viewed fishing favorably and said some of them relocate to the lakeside for a period of a week or more to fish and would get neighbors to watch their livestock while doing so. Archeological evidence has demonstrated that changes in aridity and subsequent Lake Turkana levels affected subsistence strategies of pastoralism, fishing, and foraging over long periods (Ndiema et al 2010). While social and cultural issues affect how pastoralists consider implementing livelihood diversification (McCabe et al 2014, Sagawa 2021), our research demonstrates how droughts can affect subsistence patterns, lead to diversification of strategies, and change stigma and cultural norms over a relatively short period of time.

Several findings in this paper support the Marsabit County Climate Change Action Plan (2023) in which they note the adaptation action needs of supporting development of water infrastructure and livelihood diversification. By fishing and finding alternate sources of water, Daasanach demonstrate resilience and use of coping strategies to reduce the effects of drought. Efforts to increase fishing in a sustainable manner as a diversified livelihood strategy in the region is now a policy priority to help prevent overfishing (UNESCO 2025).

Droughts affect multiple biological systems (Ebi & Bowen 2016), often working through nutritional status. Our prior work demonstrated that in 2022 at the peak of the drought, high water insecurity and chronic stress were associated with inflammation as measured through mildly elevated systemic C-reactive protein (Roba et al 2025). One weekly panel study with a small number of households in southern Ethiopia and northern Kenya found that vegetation conditions affected available forage and resultant food insecurity during the Horn of Africa drought in 2021–2022 (Alulu et al 2024). A systematic review and meta-analysis found that drought exposure was associated with 46% higher odds of wasting and underweight prevalence among children (Lieber et al 2022). Panel data from 2009–2013, which covered an earlier drought in Marsabit county, found that each standard deviation in reduced precipitation was associated with 0.52 standard deviation lower MUAC for children (Bauer & Mburu 2017). At the county level, food insecurity was strongly correlated with stunting prevalence over time (Kipkorir et al 2024).

At the sub-county level in which Daasanach live, tracking by NDMA of MUAC found 30% of children under 5 were at risk for severe acute malnutrition in early 2023 (Authority 2023). Our study contributes to this literature and provides additional depth and substantiation as we found that WAZ, HAZ, BMI-Z and MUAC for children were significantly lower than before the drought and worsened by 2023. Yet, children who consumed milk at least once in the prior week had significantly better nutritional indicators than children who did not drink milk – nearly a third of a standard deviation. This pastoral foodway is critical in mitigating the negative effects of drought. The loss of livestock herds thus has multiple negative impacts on pastoralists from psychosocial stress to worsened diets and nutritional outcomes. For adults, who already have low adiposity levels, mean BMI, which was below the underweight threshold of 18.5 kg/m² for all years, dropped by 0.2 kg/m² by the end of the drought with rebounds in adiposity a year post-drought. The nutritional effects were more severe on women than men and worse for children than adults as they had not recovered from the nutritional shortfall at one-year post-drought.

Policymakers should prioritize food aid interventions that maintain milk access for children during drought. Interventions that address pinch points, like lack of water sources in pasture areas, may help maintain pastoral lifeways but must do so in ways that are compatible with nomadic and semi-nomadic pastoralist livelihoods to avoid unintended consequences such as sedentarization (Fratkin 2019).

4.1. Limitations

In terms of long-term trends, our study is limited in that we only have one year of pre-drought data and data collection was halted due to COVID-19 in 2020. The drought effects presented may be conservative as the comparative pre-drought year had low rainfall whereas the first half of 2020 had greater rainfall (Figure 1; Figure S4), thus if we had data from June-July 2020 immediately preceding the drought the observed drought effects may be more pronounced. As the drought was widespread in the region, there was not a non-drought-exposed part of the Daasanach population in Kenya to compare results.

Panel studies are subject to attrition, particularly ones with semi-nomadic populations. This ongoing study uses replacements as a refreshment strategy for the panel, which has pros and cons. It allows for evolving changes in study communities and helps mitigate attrition bias by enrolling individuals with similar characteristics to those more likely to be gone (e.g., younger adults and men who may be more mobile) (Deng et al 2013), but it reduces the ability to make inferences about individuals lost to follow-up who may differ from those who stayed. Reasons for temporary or permanent attrition included moving permanently, being away outside the study communities (in the fora), or being deceased. Our sensitivity analyses attempting to address attrition bias provided consistent results with the primary analyses. We had limited repeat measurements on children, thus the interpretations of findings for children are limited to broad community level changes over time and not individual level changes.

While we tracked the total number of specific livestock animals that households owned, we did not collect data on potentially milking and reproductive animals, which would be useful to immediate food resources and herd recovery through reproduction compared to other means (purchase, aid, theft). Further, prior literature and pastoralist cultural norms suggests that households may underreport actual livestock holdings, meaning livestock figures presented likely represent conservative estimates of herd sizes, though this may not be systematic where some might do so more during and in the immediate aftermath of drought when hoping for assistance. However, as we did not conduct objective counts of each households’ livestock holdings due to practical difficulties (livestock out to pasture during the day), we cannot be certain as to whether all households underreported holdings. Nevertheless, our participant observations over the course of the study along with our GPS-collared animals which we embedded with different community herders’ flocks increased the representativeness of those findings and provide substantiation of the overall decline and pattern of livestock numbers. Finally, all cattle that were GPS-collared died in 2022 during the drought, thus our findings while mimicking the overall patterns of the community are limited in terms of generalizing their mobility after the drought.

Dietary data is subject to recall bias and there may be gendered aspects to this in some societies. While food frequency approaches can miss items, they are better than 24-hour dietary recalls for the detection of episodically consumed foods because of greater day-to-day variability and have a lower participant burden and a proven record of construct and predictive validity (Willett & Hu 2007). Finally, an alternate statistical approach to the econometric one would be to conduct a mixed effects model nesting observations within time and community with varying effects. When models are re-estimated using that nested structure, results are highly consistent and more significant (results not shown). We thus, presented the more conservative estimates herein.

5. Conclusion

Overall, this case study helps demonstrate how lived experiences and resource insecurities in already under-resourced settings are affected by extreme drought and the uneven recovery process in outcomes that unfolds afterward. Livestock numbers plummeted and with them Daasanach became less mobile, consumed less milk, and had worsened nutritional status – particularly women and children – which had not recovered one-year post-drought. However, during the drought, Daasanach exhibited resilience by increasing their reliance on fishing to supplement their diet and income. As has occurred with prior droughts and other pastoralist populations (Fratkin 2019), droughts can hasten sedentarization as movement declines and people are drawn to local towns that have greater educational and food aid opportunities. Working with communities like the Daasanach sheds light on the local impacts of extreme climatic events. Drought monitoring surveillance systems should incorporate experiential water insecurity measures alongside food insecurity as the two phenomena differ. Simultaneous measure of experiential resource insecurity and questions about mobility ideation can be useful to understand subjective well-being and psychosocial distress. Coupling these measures with objective nutritional and health markers can further uncover the full range of drought effects. Policymakers should use these findings to inform drought mitigation plans as diversification of diet and livelihoods may help ameliorate negative impacts.

Highlights.

• Northern Kenya experienced a historic drought between 2020–2023

• During the drought mobility declined while thinking about moving increased

• Drought monitoring should assess experiential water insecurity like food insecurity

• Post livestock death, diversification with fishing increased, lowering prior stigma

• Children with milk access did better nutritionally than peers without milk