The project employed a multi-stage sampling technique. In the first stage, the upper Yala catchment was divided into strata based on ecological and topographical characteristics. In the second stage, a random selection of sub-catchments was made from each stratum. In the third stage, households were selected from each sub-catchment using systematic random sampling.

The sample size was determined using the formula n = z 2 ∗ p ∗ ( 1 − p ) / e 2 , where “n” was the sample size, “z” was the standard normal deviation at the desired level of confidence (95% confidence level), “p” was the estimated proportion of households with conservation investments, and “e” was the desired level of precision (5% margin of error). Based on past studies, it was estimated that the proportion of households with conservation investments was 40%. Thus, the required sample size was 400 households.

The sampling procedure involved the following steps:

Ÿ Dividing the upper Yala catchment into strata based on ecological and topographical characteristics.

Ÿ Randomly selecting sub-catchments from each stratum.

Ÿ Using systematic random sampling to select households from each subcatchment.

The estimated sample size of 400 households was proportionally allocated to Bomas and then villages according to the targeted number of beneficiaries. Table 1 shows the detailed sample size distributions by Bomas. The sampled number of households per district was proportional to the size of the beneficiary population.

In a multi-stage sampling design with a given sample size, there is no prescriptive formula for determining how many clusters to choose and how many beneficiaries to choose within each cluster. There are competing interests in terms of what is most operationally expedient versus what is most statistically efficient. Due to movement restriction, access, and cost, it is optimal to select the smallest number of clusters possible, with a greater sample size per cluster (assuming a given sample size). When a smaller number of clusters are selected, the time and cost of transportation to, from, and in between the clusters are decreased―and potentially the number of data collectors can also be decreased.

It is possible, however, to provide a rule of thumb concerning the number of sampled beneficiaries to allocate to each sampled cluster. A range of 15 - 35 beneficiaries for each selected cluster is appropriate because, in most cases, this represents a logistically feasible number of beneficiaries per cluster to sample without inducing a design effect that is larger than roughly 2. Based on this “15 - 35

Note: PP-Proportional Piling.

beneficiaries per cluster” rule of thumb, this survey used the rule of thumb approach to decide on the actual number of clusters and beneficiaries per cluster to choose.

The household was carefully selected in consultation with village leaders. Since we don’t have a list of HHs per village, the team of data collectors directly went to the sampled villages and asked the village leader to take them to the center of the village. At the center of the village, they spun a pen to determine the direction to move. After they determined the direction, they went to the nearest household and that household was the first interviewee. Then the team moved to the next nearest and selected every 3rd household of the nearest household and collected data until the required number of households were interviewed.

On some occasions selected households may not fit the required criteria, participants may be busy or do not want to participate, or participants may not be at home. In this case, the sampled was household replaced by another household immediately next to the sampled household. The replacements were done in consultation with supervisors.

Data management commenced with the development of appropriate tools followed by the training of data collectors on data collection methods and data quality. During data collection, data quality was assessed at the point it was collected to ensure the correctness and completeness of the questionnaires. Data was collected by online questionnaires downloaded in Excel format and exported to SPSS software (statistical packages for social sciences), cleaned, and analyzed to ensure that any questionnaire with discrepancies could be identified and corrected. Qualitative data was transcribed daily to ensure the information was not lost and to help ensure that all required information was collected during the evaluation fieldwork. The lead consultant was also holding discussions every close of business with the data collectors to ensure they captured all the important information.

We analyzed quantitative data using the Statistical Package for Social Sciences (SPSS) Version 27 for Windows. Analysis was done by descriptive statistics. Descriptive statistics was in the form of frequencies and proportions since our data was mainly categorical. Ratio/interval data (scale) was put in bins and hence became categorical.

We transcribed all the qualitative data from the KIIs and FGDs. We used NVIVO-10 software to analyze the qualitative data. In context, qualitative data analysis adopted a thematic analysis approach. This entailed the use of a theory-driven approach where the analysis categories had been determined a priori based on the study objectives and project indicators derived from the project log frame. Following this, a data reduction method was carried out. The first stage involved identifying the common themes around which the analysis should be carried out. This entailed listing all the study indicators and sub-indicators where applicable.

Exploratory analysis of collected data was then carried out. Data (Focus Group Discussion (FGD) transcripts and Key Informant Interviews (KII) were structurally coded and then partitioned in line with the study indicators and sub-indicators where a set of string and numeric codes had been used to represent one given study indicator/sub-indicator and so on. Each discrete question and its probes, forming a domain that corresponded to an indicator, was assigned a code and then linked to the subsequent response text. Within each domain, questions were given code names which included a prefix for the domain and a numeric identifier for the question topic.

The evaluation was conducted on respondents who had provided informed consent. The eligible respondents were given verbal consent to participate. The purpose and objectives of the study were explained to the respondents in adequate detail to enable them to decide on participation. No respondent was coerced, threatened, or intimidated to participate. They were all respected regardless of their decision to participate or not participate.

When asked about the nature of the water resource, almost all households (97.0%) correctly identified it as natural. A tiny fraction of the respondents either believed it to be man-made (2.0%) or admitted they didn’t know (1.0%). This suggests a broad understanding of the river/wetland’s natural origin, which could potentially influence attitudes towards its conservation. Approximately three-quarters of the households (74.6%) live within 2 kilometers of the wetland, while the remaining quarter (25.4%) lives 3 kilometers or more away (Table 3). This proximity could influence the households’ interaction with and reliance on the wetland and its resources. In terms of residency duration in the area, a majority

Concerning the water source’s availability, a vast majority of households (94.5%) understood it to be a permanent source. This perception may reflect their reliance on the river/wetland for consistent water supply. A small percentage considered it to be seasonal (4.2%), while an even smaller proportion (1.2%) was unsure about its availability. Results from the qualitative data finding showed that participants expressed varied levels of understanding regarding the seasonality of the water resource. Some households demonstrated accurate understanding, while others had misconceptions or lacked awareness. (KIIs: Department of Agriculture Nandi, Water Resource Association)

A vast majority (86.3%) of the respondents own the land where they reside or farm, while a smaller proportion (13.7%) do not. Among those who do not own their land, most (83.6%) indicated that the land is owned by a native person from the area. A few identified the owner as a non-native person (9.1%), the community (5.4%), or the national government (1.8%). Despite not owning the land, most of these respondents (67.3%) have some form of authority over its use granted by the owner. In terms of how they acquired the land they own, the respondents mentioned various means, including inheritance (37.8%), buying (10.8%), leasing (10.8%), and other means (37.8%). This diversity in land acquisition methods reflects the complexity of land tenure arrangements in the area. When asked about the duration of their stay on the land, most respondents (63.4%) have resided there for over 15 years. Others have lived on the land for 10 to 15 years (18.4%), 5 to 10 years (12.4%), and less than 5 years (5.7%) (Table 4). Qualitative data findings showed that the number of households owning the land where they are living or farming varied across the catchment area. Some households owned the land, while others were renting or had insecure tenure. (KIIs: Department of Agriculture Nandi, Water Resource Association)

Most households use their land for agriculture (93.0%), emphasizing the prominence of farming as the main occupation in the region. Land use for recreation (0.2%), as a water source (6.2%), for worship or sacrifices (0.5%), and for other purposes (5.0%) was far less common. These findings indicate a strong connection between the local community and the land, primarily through ownership and agricultural use. This information is vital for understanding local land management practices and shaping interventions that align with these realities to ensure the project’s success.

Qualitative data findings showed that participants identified various uses of the land by households, including agriculture, residential purposes, and small-scale businesses. The number and types of land uses varied across the catchment area. (FDGs: Farmers’ Group, KANAWASCO). The percentage of households having authority over the use of the land differed among participants. Some households had full control and decision-making power, while others

faced restrictions or lacked control due to communal land systems or external factors. (KIIs: Green Belt Movement, Kenya Platform for Climate Governance NOREB).

The majority of respondents (64.2%) reported that their land stretches into the wetland area. For those whose land doesn’t stretch into the wetland area (35.6%), a significant portion (28.6%) still has access to the river/wetland. Out of the respondents who do not have access to the river/wetland (6.5%), the main hindrances include lack of an access road (29.2%), restraint by neighboring people (33.3%), the wetland being private property (16.7%), and restraint by the management (4.2%) (Table 5). Qualitative data findings showed that the number of households that have access to the river/wetland differed among participants. Some households reported easy access, while others faced challenges such as distance, physical barriers, or restrictions imposed by landowners. (FDGs: Community Members, Center for Community Dialogue and Development)

A majority reported the greatest danger they faced was flooding (62.4%). However, encounters with wild animals are relatively rare, with only 4.7% reporting such incidents. About 16.9% of respondents reported facing frequent diseases due to their interaction with the river/wetland, and a small number (3.9%) reported conflicts over their presence in the river/wetland. Additionally, 22.7% of respondents reported facing other unspecified dangers. Qualitative data findings revealed that participants identified various dangers faced by households in their interaction with the river/wetland, including waterborne diseases, accidents, and encounters with dangerous wildlife. The number of households facing such dangers varied across the catchment area. (KIIs: Green Belt Movement, Kenya Forest Service)

The most common sources of water mentioned were springs, natural springs, and spring water, with 19.4% of respondents mentioning them. Other sources mentioned include rivers, streams, boreholes, dams, forests, and various streams, among others. With regards to specific names of Rivers/Streams/Drainages, among the respondents who indicated the presence of rivers or streams in the catchment area, the most frequently mentioned river was Kimondi (14.4% of respondents), followed by Kingwal (3.2% of respondents). Other rivers and streams mentioned include Chemogonja, Cheptaburbur, Yala, Budalangi, and St. Mary’s, among others. Qualitative data findings revealed that the percentage of households able to name the source(s) of the water in the river/wetland varied. Some households demonstrated good knowledge of the water sources, while others had limited awareness or misconceptions. (KIIs: Kenya Wildlife Service, Department of Health - Public Health Officer)

The majority of respondents (92.3%) understand the river/wetland to be perennial, and a large majority (80.3%) confirmed that the river/wetland has an outflow (Table 6). Among these, the seasonality of the outflow/drainage is perceived as perennial by 85.8% of respondents. Regarding the management of the river/wetland, the data reveals that 88% of respondents do not think the person managing the river/wetland is a native of the area. Almost all respondents (99.8%) do not think the entity managing the river/wetland is an organization or institution. However, a majority (60.1%) believe the community manages the river/wetland, whereas only a small fraction believe the county government

The sociodemographic characteristics of the population provide a comprehensive understanding of the community’s composition. The balanced age distribution, with a slight inclination towards younger individuals, suggests a dynamic community with the potential for both experienced and youthful perspectives. This age distribution is consistent with global trends where younger populations are predominant in many regions (United Nations, 2019) [10] . The predominance of males as household heads aligns with traditional societal structures observed in many parts of the world, where men often assume leadership roles within families (Doss, 2013 [11] ; Agarwal, 1997 [12] ). The high percentage of married household heads could indicate a community rooted in family structures, which can influence communal decision-making processes (Smith, 2017 [13] ; Becker, 1981 [14] ). The educational attainment, with a significant portion completing secondary education, suggests a community with a basic level of literacy, which can be instrumental in awareness campaigns and community engagement initiatives (World Bank, 2018 [15] ; Sen, 1999 [16] ).

The overwhelming recognition of the river/wetland as a natural resource underscores the community’s inherent understanding of their environment. Such awareness is crucial as it can influence conservation attitudes and practices (Turner et al., 2015 [17] ; Berkes, 2004 [18] ). The proximity of a majority of households to the wetland might suggest a direct reliance on its resources, which can be both an opportunity and a challenge for conservation efforts. Long-term residency in the area further indicates a deep-rooted connection with the environment, which can be leveraged for conservation initiatives (Brown & Raymond, 2007 [19] ; Adger, 2003 [20] ). The qualitative findings from NEMA and the Kenya Forest Service emphasize the importance of accurate knowledge dissemination to address any misconceptions.

The high percentage of land ownership among respondents indicates a sense of permanence and investment in the area. Land tenure security can influence land management practices, with landowners more likely to invest in sustainable practices (Deininger & Jin, 2006 [21] ; Feder & Feeny, 1991 [22] ). The diverse methods of land acquisition, from inheritance to leasing, highlight the multifaceted nature of land tenure in the region. The predominant use of land for agriculture aligns with the occupation data and underscores the importance of sustainable agricultural practices to ensure the health of the river/wetland (Pretty et al., 2006 [23] ; Barrett et al., 2001 [24] ).

The data suggests a strong interaction between the community and the river/wetland, with a majority having access to it. The dangers faced, predominantly flooding, highlight the need for risk mitigation strategies and awareness campaigns. The qualitative insights from the Green Belt Movement and the Kenya Forest Service emphasize the importance of understanding local challenges to tailor conservation strategies effectively (Folke et al., 2005 [25] ; Lebel et al., 2006 [26] ).

The identification of various water sources, with springs being predominant, provides insights into the hydrological dynamics of the area. The recognition of rivers like Kimondi and Kingwal can be instrumental in targeted conservation efforts. The perception of the river/wetland as perennial suggests a consistent water supply, which can be a crucial resource for the community. However, the management perceptions, with a majority believing in community management, indicate a potential gap between formal governance structures and community perceptions. This aligns with global discussions on the effectiveness of community-based resource management versus centralized approaches (Ostrom, 2009 [27] ; Dietz et al., 2003 [28] ).