Chaseyama and Chikukwa wards, located in the Chimanimani district of eastern Zimbabwe (Figure 1), were selected as case study sites due to their different bio-physical characteristics and agricultural potential. Chaseyama is on the western side of the district and is characterised by low rainfall (300–450 mm per annum) with periodic, seasonal droughts and with predominantly shallow, sandy soils, derived from granite, that are inherently infertile. The area is mostly unsuitable for dry-land crop production. Chikukwa, on the other hand, located on the eastern part of the district, is characterised by high rainfall of 1,000 mm per annum with much cooler temperatures. The area has deeply weathered red-clay soils that are highly suitable for diversified cropping and high-value crops such as coffee, tea, and potatoes (Mugandani et al., 2012; Oxfam-UNDP/GEF, 2015). Chikukwa is located in a very remote area, has a mountainous terrain not easy to navigate, and has few service centres like shops. Chaseyama, on the other hand, has a flat terrain transversed by major roads (from Mutare to Chimanimani and to Masvingo) and many service centres.

Farming communities in both areas have benefited from multiple seed assistance interventions such as seed aid, local seed production, seed fairs and seed banks. Seed aid includes seed directly supplied mostly by government programmes, through either the Presidential input support scheme, or via Command Agriculture (Pindiriri et al., 2021). The former targets smallholder farmers in communal areas with seed crops for household food security, while the latter targets large-scale farmers growing commercial crops such as maize, wheat, soya and sunflower, particularly in irrigation schemes. Local seed production is usually facilitated through contracts with seed companies or NGOs operating in communal areas. Seed fairs are supported by NGOs and seed companies, and aim to ensure that farmers access a diverse range of seed from local and external sources. Seed banks, mostly facilitated by NGOs, involve the significant participation of local farmer groups who aggregate a diverse range of crop varieties for storage, sharing and exchange.

Many scholars and development actors use seed security assessments as the basis of determining the extent of household seed insecurity and proposing appropriate responses (CIAT et al., 2012; FAO, 2016; Dalle and Westengen, 2020). The dimensions of household seed security include availability, access, varietal suitability, seed quality, and the resilience of the seed system (FAO, 2015a). According to Remington et al. (2002), seed availability means “having sufficient seed of desired crops within reasonable proximity (spatial availability) and in time for sowing (temporal availability).” It therefore refers to the farmer’s supply of seed from all sources. These sources can include own-saved seed, social networks, local markets, the formal seed sector and aid sources (FAO, 2015a; McGuire and Sperling, 2016).

Access to seed is defined as the ability to acquire seed through exchange, loan, barter, or use of influence in social networks (Sperling et al., 2013; FAO, 2015a). Sperling et al. (2013) also highlight the importance of access to relevant information about seed properties and management as a critical but under-appreciated part of seed access. Differentiating access from availability is vital because the seed might be available within the community or market, but some farmers might not have the resources or the social influence to acquire it. Access consists of social access and economic access (FAO, 2015a). Social access refers to acquiring seed through a household’s social network, while economic access means having the financial resources to acquire seed when needed.

Seed quality looks at the technical aspects of the seed, focusing on germination, physical purity, seed health and varietal purity (Walsh et al., 2014; FAO, 2015a). Seed germination requires tests to ascertain the percentage of seedlings that germinate from a given sample. Physical purity refers to seed that is clean and free from foreign material with no broken or immature grains. Seed health implies that the seed should be free from pests and diseases. Varietal suitability refers to seed with characteristics preferred by farmers (FAO, 2016). Shrestha (2020) further divides varietal suitability into adaptability to local agro-ecological conditions, and choice, which comprises the desired traits meeting farmers’ production, food, cultural and market needs.

The resilience of a seed system is the extent to which seed security is affected by stresses and shocks (Lin, 2011; McGuire and Sperling, 2013; FAO, 2015a). McGuire and Sperling (2013), emphasise the idea of “maintaining the functions of the elements of a system, giving attention to the institutions, relationships and knowledges of local and formal systems.” The FAO (2015a) argue that the resilience of a seed system can be measured by changes in the indicators of seed availability, access, and quality (i.e., the components of seed security). These indicators were used as the basis for determining the seed security of farming households in this study.

FAO’s seed security assessment tools were adapted and used to assess the seed security status of Chimanimani farming households (FAO, 2015a,b). Methods included a survey conducted in 2017 with 227 randomly selected households, and an agro-input dealer survey at 12 shops. The household survey captured key seed security aspects such as availability, accessibility, quality, and adequacy. The agro-input dealer survey collected information related to commercial seed stocking and seed sales. These surveys were complemented by two focus group discussions with farming households from Chaseyama and Chikukwa, as well as 10 individual life histories of farming households that provided in-depth knowledge of farmers’ characteristics, farming approaches, livelihoods, and information on how their use of different seed has shaped and contributed to seed security. The quantitative data were entered and analysed using SPSS software, while the qualitative data relied on thematic analysis. To validate and triangulate the information collected in 2017, literature and articles published before and after this period were analysed.

Indicators of seed security were grouped according to the FAO classification (FAO, 2015b) that considers availability, access, and quality (utilisation). Cluster analysis was done using SPSS to segregate households into groups based on their seed security characteristics. Specific variables were selected for the cluster analysis based on seed access, availability, and quality. For example, the seed availability component included seed proximity, timely provision, and seed presence at the level of the seed supplier. The seed access component included perceptions about seed affordability, the gender and age of the household head, and the number of times the household had received seed aid. The seed quality component included tests of germination and physical purity. Based on this initial variable selection, an algorithm was developed to determine the number and size of clusters under each domain of seed availability, access, and quality.

The crops grown included maize, sorghum, pearl millet, groundnuts, Bambara groundnut, cowpeas, beans, pumpkins, and yam. Table 1 compares the seeds of these crops, grouped into formal sources (mostly government-provided assistance and agro-input dealers), and local sources (own seed, social networks, and local markets). The results reveal that, with the exception of maize, local sources dominated, accounting for at least 70% of the seed supply. Local sources were thus essential in supplying both the bulk and diversity of seed crops needed by these farming households. This finding corroborates other studies which show that local seed channels are the main sources of seed for small grains and legumes, providing over 80% of the seed grown in Zimbabwe (McGuire and Sperling, 2016; Mazvimavi et al., 2017). Maize, being at staple crop, was mostly sourced from government aid (subsidies), agro-input dealers and own seed. Sorghum was mostly sourced from own seed, seed aid (from NGOs) and social networks as the seed (open-pollinated) can easily be saved and re-used in the subsequent seasons. Besides maize, own sources provided both the highest diversity and highest quantities of seed across crops.

Figure 2 compares the timing of seed from different sources (see also Table 1). The local seed sources included local market and social networks, while formal sources comprised agro-dealers and seed aid. The results show that most of the seed from local sources was obtained before the planting season, compared to seed from formal sources, which was obtained after the season had commenced. These results suggest that there was a higher probability of farmers acquiring their seed from local seed sources in time for the planting season. Local markets and social networks provided a readily available source of seed for farmers, while there were delays in supplying seed on time from agro-dealers. Seed aid was provided at the middle or end of the planting season, which tended to be too late to be useful. Own seed was not included in the analysis as farmers have more control over its timing for planting needs. One farmer from Chikukwa described the value of his own seed plot for seed security, explaining that:

The reliance of such farmers on own seed production ensures that they have adequate quantities of the required seed at the time of planting. The availability of seed in a timely manner is critical for smallholder farmers as most are reliant on rainfed agriculture, characterised by erratic and unpredictable rainfall that is further compounded by climate change effects. Although all farmers may not segregate their farming plots into seed and food crops, the presence of seed producers such as Christy contributes to wider seed availability for exchanges and sales.

The proximity of formal seed sources to farmers was compared to local seed sources. The results show that nearly all seed acquired from local sources was accessed within the village, while two- thirds of seed from formal sources was accessed within the village (Figure 3). Although both seed sources were found to be sufficiently close to farmers, the local seed sources provided the largest quantities of seed by volume. Seed obtained from far away districts was mostly from formal sources. As travelling to further locations has a financial cost and could be time consuming, access to seed from far locations may not be possible for all farmers. Local seed sources are thus the primary suppliers of most seed used by farmers, although there may be cases where they fail to supply the type of seed needed.

The data from agro-dealers serves as a useful illustration. Agro-dealers were not evenly spread across Chaseyama and Chikukwa but were in easy to reach central business centres along tarred roads. The mountainous terrain and unpaved roads in Chikukwa made the business centres less accessible than in Chaseyama. There were also fewer business centres in the more remote Chikukwa compared with Chaseyama. Famers who require certified seed were forced to commute to agro-dealers and seed company agents to obtain seed, even if it meant travelling to distant towns such as Mutare or Chipinge over 100 km away. Although there were similar numbers of farmers across Chaseyama and Chikukwa that accessed seed from agro-input dealers, the geographical location (i.e., terrain, infrastructure, and services available) of a community played an important role in shaping the nature of farmers’ seed availability.

The temporal availability, spatial availability and sufficiency of seed (in terms of quantity) available from different sources have been identified as key elements for assessing seed availability (Remington et al., 2002; FAO, 2016). Based on three key cultivated crops, a cluster analysis was conducted using seed availability variables to group households with similar characteristics. Household-level seed availability was “on time” if all required seed across key crops was available in a timely manner (i.e., before or at the start of the planting season). If at least one of the key crops was only available late (i.e., mid-season or at the end of the season), the household was classified as “late.” For the proximity variable, a household obtained a classification of “near” if all three major seed sources were obtained from their own village and/or neighbouring villages; “far” was when seed was further away for at least one of the three crops. For the adequacy variable, a household was classified as having “adequate” seed if they responded “yes” to the question of whether seeds were available from all their sources. If the household responded “no” to any of their seed sourced they were classified as “no”/inadequate. Table 2 illustrates the three distinct clusters of seed availability categories.

Cluster 1 consisted of 45% of households interviewed who accessed their seed on time, in proximity and who had adequate seed. These households were categorised as being “more seed secure” in terms of seed availability. Seed availability was not a challenge for these households as all received their seed before or at the start of the season. Their source of seed was nearby (either from their own village or from a neighbouring village). Over half of these households (57%) also mentioned that the seed from all their seed sources was adequate.

Thirty six percent of households fell into Cluster 2, which comprised households that accessed their seed late, with most sourcing seed from far away (57%), and 78% reporting that the seed available from their sources was inadequate. The households falling under Cluster 2 were categorised as “less seed secure” in terms of seed availability since the seed availability indicators were negative. Seed availability was a challenge as all received their seed either mid-season or towards the end of the planting season.

Cluster 3 constituted 19% of sampled households. These households accessed their seed on time (100%), although the distance travelled was far for 95% of them. Sixty three percent indicated that seed available was adequate from their sources. The households falling under this cluster were categorised as “diverse/inconclusive” or varied in terms of seed availability since they gave a mixed picture. Seed availability was not uniform (in terms of timeliness, proximity and adequacy of the quantity) across their crops. While most of this group accessed seed from far, it was acquired on time. A higher proportion viewed their seed as adequate compared to those that did not. This suggests that while proximity is an important aspect for accessing seed, for this group distance was a less important factor affecting timely availability of adequate seed.

These three clusters suggest that there is a wide range of seed availability between households (from more seed secure, less secure, and in between), and that such variations depend on the farmers’ crops, varieties, seed sources, location and circumstances. The following account from a farmer in Chaseyama illustrates this variability, suggesting that although he acquired his seed from agro-dealers, NGOs and a government-supported scheme, the seeds were not adequate for his needs.

Critical issues raised by this farmer included the cost of seed, the timeliness of seed provision, and the reliability of the seed sources. In contrast Tabeth, also from Chaseyama, relied on her own seed which was sufficient for her needs. She explained:

There were more insecure households in Chikukwa (63%) than in Chaseyama (9%). Overall, a higher proportion (61%) of households from Chaseyama were seed secure, compared to Chikukwa (29%). The farmers’ location determined which seed sources were available, the timeliness of the seed provision, and the distance travelled to obtain seed. Most seed sources for Chikukwa farmers were formal while most for Chaseyama farmers were local. The farmers that accessed most of their seed from local sources were more likely to be seed secure compared to those that accessed their seed from formal sources.

Almost all the seed from local sources (93%) was perceived to be affordable compared to seed from formal sources (32%) (Figure 4). Sixty three percent of seed planted and accessed from formal seed sources was considered to be expensive. The “affordability” and “expensive” indicators were based on farmers’ perceptions of the cost of seed purchased, thus excluding seed aid. These were seeds purchased from social networks, local markets, agro-input dealers, and local agents of seed companies.

These results suggest that there was a higher probability of farmers acquiring affordable seed from social networks and local markets for their planting needs than from agro-input dealers. The affordability of seed was also analysed by looking at the cost of seed sold by agro-input dealers. Data from agro-input dealers for example, showed that a 10 kg bag of hybrid maize seed sold in Chaseyama ranged from US$18 to US$22, while the same seed quantity sold in Chikukwa shops ranged from US$30 to US$35. This could be related to the remoteness of Chikukwa which is further than Chaseyama from Mutare, the nearest city that provides most of the formal seed. Such factors emphasise the importance of local seed sources for resource-constrained smallholder farmers.

Variables were selected to cluster households based on seed access, and the ability of farming households to acquire seed. These included seed price (affordable or expensive), access to seed aid, the number of times a farmer accessed seed aid, income, asset variables (livestock, homestead assets and land), and demographic variables (gender and age of household head). Seed affordability was assessed in terms of seed prices being affordable, expensive, and very expensive across the different seed sources. A household was given a code of “0” (expensive) if at least one of their seed purchases was perceived to be expensive. A household was given a code of “1” (affordable) if all their seed was considered affordable. This demarcation, though extreme, serves to segregate households by the seed access challenges they experienced. The households that did not incur any challenges of seed access were classified as seed secure. Those households that had at least one challenge were classified as seed insecure. Table 3 shows the results from the cluster analysis after removing variables that did not significantly contribute to the clustering.

Cluster 1 comprised households who perceived their seed to be affordable (42%). These households were categorised as “more seed secure” in terms of seed access since the price of all seed planted was perceived to be affordable by all. This group also received seed aid the least number of times on average (statistically significant, p < 0.05). The fact that all household heads under this cluster were older males (mean age of 52.7 years) may serve to support assertions that access to seed could be influenced by the gender of the household head (Beshir and Nishikawa, 2012).

Cluster 2 comprised 35% of the farming households surveyed, which perceived at least one of their seed crops to be expensive. Those in this cluster were categorised as “less seed secure” because of this expense, and also received seed aid the largest number of times. Compared to Cluster 1, male respondents were younger (mean age of 49 years) and less seed secure in terms of seed access. Age in this case could be a contributor to seed access and/or adoption depending on the type of seed as argued by Kinuthia and Mabaya (2017).

Cluster 3 consisted of 23% of households surveyed. Seed was perceived to be “affordable” by 54% and “expensive” by 46% of this group. They were categorised as “diverse/inconclusive” in terms of seed access as the households experienced mixed conditions of seed accessibility. This mixed group comprised only female headed households. The most important variable in clustering the households was gender, followed by affordability (Table 3). Compared to the Cluster 1 male group, the Cluster 3 group had fewer households that perceived their seed to be affordable as they were more resource constrained (shown by a significantly smaller household size, fewer income sources and fewer household assets). However, they received seed aid less frequently than Cluster 2 (less seed secure group). The three cluster categories serve to illustrate the diversity of farmers’ seed security (access) as this relates to affordability, age and gender of household head, and seed aid.

The highest proportion of households that received seed aid once or twice over a five-year period was from Chaseyama. In contrast, the highest proportion of households that received seed aid every year was from Chikukwa. Over 5 years, Chikukwa households benefited on average 3.6 times from seed aid compared with Chaseyama farmers (1.9). In the 2016–17 season, about 48% of the 3,894 kg of maize planted in both areas and 17% of sorghum seed, were derived from seed aid. Seed aid is thus an important source of seed for smallholder farmers in Chimanimani. While maize seed was supplied mostly by government, sorghum seed aid was supplied by NGOs such as World Vision. This shows the strong influence of government support towards enabling the formal seed system to supply the main staple cereals to farming households. However, as described, although seed aid has increased seed availability, this is usually provided too late. The modes of seed aid assistance were largely through free distribution (88%), with some through vouchers and seed loans (Figure 5).

Households believed that at least 96% of seed planted from both formal and local seed sources did not have physical impurities or any physical damage. These results show that farmers perceived that seed from local sources had comparatively good quality characteristics to that from formal sources. About 96% of seed from formal and local sources was perceived to have good germination. The germination performance of seed from local sources was thus understood to be comparable with seed from formal sources.

A cluster analysis on seed quality was not performed as most farmers perceived their seed to be of good quality. Table 4 presents methods used for seed preservation and storage to ensure good seed quality. The methods mentioned by most farmers in Chaseyama included treating the seeds with ashes and smoking. In Chikukwa, farmers noted the importance of smoking seeds in the kitchen and treating them with a storage chemical (copper shumba). A 48 years-old farmer from Jinga village in Chaseyama gave the following account of the seed preservation techniques he used, based on those used historically by his parents:

This finding concurred with an account from Vivian, a 42 years-old woman residing in the village of Munaka in Chikukwa:

These accounts of traditional methods of seed preservation and storage corroborate farmers’ perceptions of having quality seeds of their own, stored varieties. Farmers with access to adequate land were able to separate food crops from seed crops and prevent cross-pollination of their seed crops. This is a particularly important element for maintaining seed quality for farmers engaged in seed production processes (Wojciech et al., 2022).