Relationships between land tenure insecurity, agrobiodiversity, and dietary diversity of women of reproductive age: Evidence from Acholi and Teso subregions of Uganda

Abstract Land tenure security is central to food security of rural agricultural‐dependent communities, but there is limited evidence linking the state of agrobiodiversity to perception of land tenure security and access to and quality of food eaten. This study explores this relationship using data captured from 1,279 households in Acholi and Teso subregions of Uganda, and the relationships are established using a study sample of 1,227 women of reproductive age (WRA). Sixteen percent of respondents perceived themselves to be land tenure insecure. Although approximately 275 species were reported available for food, household access to a variety of plant and animal species is limited to <10 species by 69% of the study population. Dietary diversity was also low, with 53% of women meeting minimum diet diversity. Evidence from estimation of a generalized Poisson regression reveals that dietary diversity of WRA is consistently, positively correlated with species diversity available for food and negative with land tenure insecurity. A unit increase in species diversity led to 18% increase in dietary diversity of WRAs. Land tenure insecurity was likely to reduce dietary diversity of WRAs by 26% (p < .05). Interventions with an aim to increase species diversity can deliver positive dividends for food and nutrition security. Land policy reforms and interventions that strengthen land tenure security for both men and women are likely to contribute positively to dietary diversity leading to improved food and nutrition security of vulnerable communities in rural areas.


| INTRODUCTION
One of the greatest global challenges is to secure sufficient and healthy food for all and to do so in a sustainable manner (Burchi, Fanzo, & Frison, 2011). Under global mandate of sustainable development goals, world leaders and proponents of development agreed on agenda 2030 that chiefly aims at achieving sustainable development in all dimensions namely social, economic, and environmental dimensions. Key among the 17 goals is eradication of poverty in all its forms and dimensions and ending hunger and malnutrition through efforts that promote food security and sustainable agriculture to ensure access to safe, nutritious, and sufficient food for all but more so for the world's most vulnerable (United Nations [UN], 2015). The UN targets expect to ensure that all vulnerable men and women have equal rights to economic resources as well as access to basic services, ownership and control over land and other forms of inheritance, natural resources, and technologies. Additionally, it is also recognized that achieving a food secure world requires maintenance of genetic diversity of seed, cultivated plants, farmed and domesticated animals, and their related wild species that form a great part of agricultural biodiversity. Agricultural biodiversity also known as agrobiodiversity is defined as "the variety and variability of plants, animals, and microorganisms at genetic, species, and ecosystem level." Although most efforts for biodiversity conservation have traditionally aimed at supporting protected areas, the indicated link to food and nutrition security creates the need to investigate participatory models of biodiversity management in agricultural ecosystems that embrace biodiversity for farmers' food and livelihoods (United Nations University [UNU], 2003).
Land is the key foundation for agrobiodiversity. A farmer's perception or sense of security in relation to their land can have positive or negative effect on short-term decisions and long-term investments especially in terms of conservation practices and crop and animal choices that later determine nutrition outcomes for subsistence households, where the primary objective for production is household consumption. It influences the extent to which farmers are prepared to invest in improvements in production and land management (International Fund For Agricultural Development, 2015). It is also suggested that tenure security shapes social relations and contributes to social stability-or rather, situations of tenure insecurity contribute to social instability and conflict. Without expounding on the nature of land ownership, land size has a negative correlation with dietary diversity (Hossain, Jimi, & Islam, 2016). The certainty with which dwellers on land have on the stability of their future survival on it without threats from external and/or internal parties is thus pertinent to household livelihood and both agrobiodiversity and dietary diversity. Any factor that undermines ones' confidence of land tenure stability is a precursor for land impermanence syndrome. Land tenure security implies that the farmers or people using and living on a certain piece of land are certain of its ownership status, therefore land impermanence syndrome/land tenure insecurity involves farmer apprehension or uncertainty about the future ownership status on land and leads to disinvestment in an agricultural operation as well as erosion of producer confidence (Parry & Skaggs, 2014). Thus, land impermanence syndrome can undermine planning and encourage speculation according to Parry and Skaggs (2014). In addition, secure access to sufficient land is an important means of achieving food security in poor agrarian land-scarce societies, and strong tenure security for landowners stimulates investment and efficiency of land use (Holden & Ghebru, 2016).
Despite all the above evidence linking, land tenure security to agrobiodiversity conservation and utilization and food and nutrition security, these kinds of linkages have not been studied extensively in developing countries and especially East Africa; most of the studies like those by M'Kaibi, Steyn, Ochola, and Du Plessis (2016) and Saaka, Osman, and Hoeschle-Zeledon (2017) have focused on linking production and consumption diversity, and it is no longer enough to keep examining the impact of agricultural interventions on food quality alone while neglecting where the food primarily comes from, land. Moreover, agriculture, land, and nutrition cannot be separated, but agriculture interventions have had gaps in contextualizing nutrition and overly lay emphasis on determining impacts of agricultural interventions on food consumption and diet quality (Herforth & Ballard, 2016).
This study examines the changes in agrobiodiversity for two distinct categories of households: those perceiving themselves secure and those perceiving themselves insecure with regards to land tenure.
The study further evaluates the association between the perception regarding land tenure and utilization of agrobiodiversity and the relationship with consumption patterns of vulnerable population groups, particularly women of reproductive age. The entire study areas were affected by the insurgency during the Lord's Resistance Army civil war that lasted 20 years  in northern Uganda and parts of eastern Uganda. Rehabilitation and total recovery of the affected communities are still ongoing in Teso subregion, eastern Uganda, and Acholi subregion in northern Uganda. During insurgency in the two subregions, land owners were displaced from their customary land and forcibly taken to camps by the government as a way of "protection." However, upon return to their land starting in 2005, the regions that were affected have suffered several land conflicts arising due to contested access to a land and disputes over land boundaries. Boundary

Key messages
• Despite 16% of households with women of reproductive age (WRA) perceived themselves to be insecure with regards to land tenure security, the studies small holder households in Teso and Acholi subregions of Uganda can access 55 cultivated plant species, eight domesticated animal species, and more than 150 wild plant and animal species for food.
• Almost half of the women (47%) were not meeting their minimum dietary diversity score and majority of those WRA who did not meet the MDD ate more of starchy staples (cereals, roots, tubers, and bananas) and vegetables and a lesser percentage ate pulses, nuts/seeds, dairy, meats, eggs, and vitamin A-rich vegetables and fruits. For all categories, consumption of milk and milk products, vitamin A-rich vegetables and fruits, and eggs was low.
• A unit increase in the number of species accessed for food by WRA is likely to increase their dietary diversity by 18%.
• The perception of being land tenure insecure has a likelihood of reducing diversity of diets consumed by WRA age by 26%.
disputes are also associated with traditional use of nonpermanent markers such as trees and shrubs which at times are lost because of various factors. Most of the people in Acholi and Teso are poor, and according to Fanzo (2018), the global population most affected by food insecurity and malnutrition includes the poor, rural, isolated, women and children, marginalized, and conflicted. The overriding hypothesis was that the household perception of land tenure insecurity could explain one of the food security dimensions-food quality-described in terms of dietary diversity. The study looks at land beyond its physicality in terms of size to understand how the unobserved fear of land being taken by another party affects the choices with regards to what food commodities the households choose to grow and rear for both food and income and any other practices related to conservation of agrobiodiversity thus influencing food options that in turn can have effect on consumption patterns for women of reproductive age (WRA).

| Study area and sampling procedure
This study was conducted in Acholi, which is a subregion in northern Multistage sampling involving mixed methods as shown in Figure 1 was used. Districts were selected based on historical perspective of insurgency and consultations done with local opinion leaders. The objective was to come up with two districts representing areas with more households potentially land tenure insecure and land tenure secure in each subregion. In Acholi, Nwoya and Lamwo districts were purposively selected to represent districts experiencing relatively higher incidence of land tenure insecurity and land tenure security, respectively. In Teso, Amuria and Bukedea districts were purposively selected to represent districts with higher incidence of land tenure insecurity and land tenure security, respectively. In the second sampling stage, two subcounties were randomly selected per district, and in each subcounty, one village chosen following simple random sampling.
The final sampling stage was done at household level to identify households with WRA and especially those with children 6-59 months.
The exact household sample size was calculated using Fisher's formula: where n is the required sample size, t is the confidence level at 95% (standard value of 1.96), p is the estimated proportion of children under 5 years in the respective four districts with regards to the total population, m is the margin of error at 5% (standard value of .05; Magnani, 1997 information), the scores on land tenure insecurity were derived from the set of responses. However, a dichotomous distinction was derived from the degree of insecurity at seven levels (1 = very insecure; 2 = moderately insecure; 3 = insecure; 4 = indifferent; 5 = secure; 6 = moderately secure; 7 = very secure), and generally, Levels 1-3 were considered relatively insecure, and Levels 5-7 were considered relatively secure, whereas farmers in Level 4 were left out of the analysis sample (Supporting information). The respondent was required to select a level based on the overall evaluation of responses on a set of 12 contextual question (Appendix A).

| Measurement of agrobiodiversity using plant and animal species diversity
In this study, species diversity is considered as a count of the number of different species available in the communities and accessed by the household for use as food. This includes crops grown; animals reared; wild plants and animals (including insects) trapped, gathered, or hunted; and other food commodities obtained through purchase. Thus, this was measured by summing species that a respondent reported to have accessed during the 12 months preceding the survey. For each species (plant or animal) mentioned, the respondents had to provide details such as its source, prime use, its availability, and, where applicable, the level of production in terms of area or number of trees/animals.

| Analytical technique
Maximum likelihood estimation of the Poisson family was used in this study. The class of Poisson regression applies for outcomes of count data (Long & Freese, 2001;Greene, 2002) (2012), it is hard to achieve in practice. In situations where the variance is lower than the mean, the data are said to be under dispersed. Modelling under dispersed count data using inappropriate models can lead to overestimated standard errors and misleading inference (Harris et al., 2012). According to Husain and Bagmar (2015), the GP regression has statistical advantages over both standard Poisson regression and negative binomial regression models and is suitable for analysis of count data that exhibit either overdispersion or underdispersion. The GP regression is a generalized event count model that is appropriate for both overdispersed and underdispersed count data (Consul & Jain, 1973;Winkelmann & Zimmermann, 1994).
In our case, the distribution of dietary diversity score of WRA exhibited under dispersion. This was also justified by the Akaike information criterion factor that was lower for the GP compared with the standard Poisson regression. The primary equation of the Poisson regression model is shown in Equation (1).
where the expectation of the number of foods groups eaten is denoted by Y outcome that takes on a set of dietary score integers J = {1,2,3,…,j} dependent on a vector of explanatory variables x 0 whose quantitative influence is estimated by a vector of parameters β.
The model was operationalized through estimation of GP regression following the expression below showing dietary diversity (y i ) of WRA i, and y i is hypothesized to be dependent on a set of explanatory variables (x 1 to x n ) whose respective effect is quantified by the variable specific parameter (β).
Explanatory variables x 1 to x n included variables categorized in three sets: household socioeconomic factors, land factors, and location factors. Household characteristics included age of household head and spouse, dependency ratio, education of WRA, and income. Age influences household dietary diversity (Mango, Bryon Zamasiya, Nyikahadzoi, & Siziba, 2014), and this is likely to affect dietary diversity of household individuals and information on traditional feeding practices. However, formal schooling is also important because it enhance one's knowledge on diet quality. Additionally, socioeconomic factors such as income levels and household dependency burden can negatively impact on dietary quality (Bouis, Eozenou, & Rahman, 2011). Wealth is a determinant of dietary diversity (Powell, Kerr, Young, & Johns, 2017). Income is also an indicator for the household standard of living and in this case, is meant to control for poverty following the international poverty line of US$ 1.9/day given income also affects food access.
Additionally, location specific factors were also controlled for because species diversity in an area and isolation away from the food markets can potentially reduce access to a wide range of food choices.
Distance from the market was therefore added to control for isolation, whereas the dummy for subregional location not only controls for heterogeneity in agro-ecological conditions (which also influences variations in species diversity and agricultural system) but also controls for variation in access to several species by inhabitants of Acholi and Teso. Ritzema et al. (2019) showed that crop and livestock diversity influence dietary diversity, but also, Guo et al. (2019) observed that production diversity alone is not enough in explaining dietary diversity. Additionally, although wild vegetables may be consumed in small quantities, they influence intake of cereal staples, manage hunger, and play a central role in household food security for the rural poor (Mavengahama, McLachlan, & Clercq, 2013;Walsh & Rooyen, 2014).
In this case, the sum of plant and animal species accessed for food is used as a measure of species diversity for food.
The variable land tenure insecurity is measured as a score generated from principle component analysis. The regression was done in two steps. First, with several factors of which others were eliminated after ascertaining that they caused multicollinearity, and others with p values less than 0.1 in order to achieve the most parsimonious model.
Based on measures of central tendencies and proportions, data on characteristics of household, WRA is presented, whereas the regression was done in the spirit of multivariate analysis. Generally, data analysis was done using STATA version 14.

| Household characteristics
Results in Table 1 show that 72% of the WRA in Acholi and Teso were in male-headed households that is 4% higher than the national estimated average for rural areas of Uganda. The results also show that 60% of the household were in monogamous marriages, a figure close to the national level of 57% as per the

| Access to land and land tenure insecurity
The findings shown in Table 2 indicate that land size and land use has changed between the present (in 2017) and the past years (before 1997). Average land holding was found to have dropped by 50% from an average of 12 acres to an average of 6 acres. Area of land farmed has also dropped by 50% from an average of about 8 acres to an average of 4 acres. Additionally, land under range (covered by grass, shrubs, and trees, left for grazing of either domestic livestock or wild animals or both) has not changed for Acholi, but it has significantly reduced for Teso (Table 2). Figure 2 shows that 16% of the households perceived themselves to be land tenure insecure and the level of land insecurity was higher in Acholi (20%) compared with Teso subregion (12%).
Generally, 17% of the WRA were living in households with perceived land tenure insecurity. Figure 2 further shows that land tenure insecurity in Acholi was higher in Nwoya District, and in Teso, Amuria District had the highest percentage of land tenure insecure households.

| Plant species diversity
More than 35 plant species were recorded as cultivated by the interviewed household members. The most popular cultivated plant species included sorghum and cassava that were cultivated by more than 50% of the households. Other popular food crops included maize, millet, white-fleshed sweet potato, groundnuts, beans, cowpeas, and sesame seeds (Table 3).

| Livestock species diversity
Results showed that the animal species most kept by households in Acholi and Teso include cattle (of local breed) followed by goats, chickens, sheep, pigs, and turkeys. A significantly higher percentage of households in Teso kept cattle compared with Acholi, whereas the reverse was true for chickens.

| Wild species diversity
Overall, 69% and 41% of households reported to have gathered some plant species and/or hunted some animal species, respectively. More than 100 wild species that could be used for food were reported by the 1,279 respondents. Approximately 78 animal species, 10 bird species, 14 fish species, 23 fruit species, 13 insect species, and more than 80 wild plant species were reported. It is shown that plant and animal species are the most gathered and hunted (Table 5). Plant species most maintained within their natural habitats were mainly fruits and they included mangoes, oranges, pawpaw, and jackfruit maintained by 19%, 15%, 9%, and 7% of the households, respectively.
Also, despite the large diversity in edible wild species, the level of their utilization varies in the two subregions, but utilization of fish, bird, and insect species is generally very low in both Acholi and Teso subregions. Table 5 shows that 61% of households in Acholi gather one to two species from the wild, whereas about 78% of households in Teso gather three to four species from the wild. Thus, households in Teso gathered/hunted more wild species compared with households in Acholi. Generally, the insecure households had better access to wild species compared with households that perceived themselves land tenure secure.

| Crop grown and overall species diversity
The results further reveal that irrespective of the subregion, most households grew three to four different crops, and particularly, 66% of the households in Teso compared with 55% of households in Acholi grew three to four crops. On account of perceived land tenure insecurity vis-a-vis land tenure security, it was found that for both Acholi and Teso subregions, households that perceived themselves to be secure had more crop diversity compared with those that perceived themselves land tenure insecure (Table 6). For instance, 55% of the land tenure insecure households grow three to four crops compared with 61% of the land secure households. Additionally, 59% of land insecure households had access to five to nine wild species compared with 57% of the land secure counterparts.   Table 7 shows that the most common food groups consumed were cereals, roots, and tubers with more than 89% of all the WRA having consumed foods from these groups. Other popular food groups included leafy vegetables (80%), meat/fish (55%), pulses/legumes (55%), and nuts and seeds (43%). The findings also indicate that consumption of cereals, roots, tubers, and bananas; pulses/legumes; and nuts and seeds was not statistically different among WRA from households that perceived themselves to be land tenure insecure versus the land tenure secure. Consumption of vegetables and fruits (either vitamin A rich or not) was significantly higher among households that considered themselves land tenure insecure compared to the land tenure secure. This is explained by the fact that most households considering themselves to be land insecure were also observed to be living in areas closer to conservation areas and could have had better access to foods from wild habitats compared with their counterparts.
T A B L E 6 Crop and overall species diversity among land tenure insecure and land tenure secure households

| WRAs that met MDD requirement
Despite the observed relatively high percentage of women consuming starchy staples, legumes/pulses, nuts/seeds, green leafy vegetables, and meats and fish, considering the overall number of women of reproductive age, 47% were not meeting their MDD score. The findings also showed that a significantly higher proportion of women in Acholi subregion (57.1%) were not meeting MDD as compared with those not meeting the same in Teso (35.3%). Additionally, WRA who did not meet the MDD ate more of starchy staples (cereals, roots, tubers, and bananas) and vegetables and a lesser percentage ate pulses, nuts/seeds, dairy, meats, eggs, vitamin A-rich vegetables and fruits.
Women who met their MDD requirement ate most of the foods in the food basked (Table 8). However, for all categories, consumption of milk and milk products, vitamin A-rich vegetables and fruits, and eggs was low. This is attributed to the high price often placed on such products to the extent that even when produced at household level, they are highly regarded for income generation rather than consumption.

| Drivers of dietary diversity
The results presented are in the spirit of a generalized Poisson multivariate regression of DD i and a set of regressors (x) for WRA, and the results are presented in two panels, standard regression coefficients and after computation of marginal effects.
The results show that dietary diversity of WRAs exhibits a positive correlation with a male-headed household, age of household head, species diversity, daily disposable income, and education of WRA, whereas increase in WRA age, dependency ratio, land tenure insecurity, and distance to the market negatively correlate with dietary diversity (Table 9).
After controlling for the various factors including age, sex of household head, land size, household income, and distance to the market, we find that a unit increase in the number of species accessed for food is likely to increase dietary diversity of WRA by 18% (p < .000), other factors notwithstanding. Additionally, the perception of being land tenure insecure has a likelihood of reducing diversity of diets consumed by women of reproductive age by 26% (p < .05).

| Limitations of the study
The study tried to control for variation in species diversity due to geographical and cultural differences; however, variation in soil quality, which is an important determinant of what a farmer can produce, was not controlled for basing on the scope of the study. Additionally, this was a cross-sectional study, where data were collected at only one   Wald χ 2 = 161.86, p > χ 2 = .000 Observations (N) = 1,072 Note. All the variables listed in the table were the ones included in the final multivariate Poisson regression analysis. Robust standard errors in parentheses. *** p < .01. ** p < .05. * p < .1.
between land tenure insecurity and dietary diversity of WRA in this study. The negative correlation between land tenure insecurity and dietary diversity is attributed to the likelihood of it, limiting investment and production on the land more so, the type of animal species reared, and crops planted. It may also affect what is maintained/conserved on the land much as not domesticated. In that regard, the number of species available for home consumption is negatively affected, leading to a significant downward impact on the dietary diversity of the WRAs.

| CONCLUSION
From the findings that dietary diversity of WRA is consistently, positively correlated with species diversity and negatively with land tenure insecurity, interventions with an aim to increase species diversity can deliver positive dividends for food and nutrition security, whereas land policy reforms and interventions that strengthen land tenure security for both men and women are more likely to contribute positively to biodiversity and dietary diversity. This would lead to improved food and nutrition security of vulnerable communities in rural areas.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

CONTRIBUTIONS
BE designed the study, initiated the publication, and coordinated and led the improvement of the manuscript during review process. RMA developed the protocols, performed data collection, data analysis, played a major role in putting together the results and the methodologies section, and revised the manuscript during the review process.
GK provided technical backstopping during the study and provided great input during development of this manuscript especially the methods and analysis work and has been involved in revision of the manuscript during its review. ID played a major role in leading development of protocols for collecting data on land tenure in/security and actively engaged in the development of the manuscript through regular feedback on methodology and results. MB actively engaged in development of protocols for collecting data on land tenure in/security and actively engaged in the development of the manuscript through regular feedback on Introduction, Methods, Results, and Discussion with special focus on land related issues.