The location of the study is the Brahmaputra Valley in Assam. The Brahmaputra valley has a total drainage area of 580,000 km² (Debnath et al., 2023) and encompasses an area of 70,634 km² in Assam (GOA, 2023). For this study, we specifically focused on the Sonitpur district of Assam, which is part of the northern bank plains of the Brahmaputra (Chaturvedi et al., 2021). Apart from its location in the foothills of the Eastern Himalayas and the occurrence of diverse types of habitations (Srivastava et al., 2002), this district of Sonitpur is categorized as highly vulnerable to climate change (Ravindranath et al., 2011), which was one of the design criteria. The erstwhile Sonitpur district, with an area of 5,105 km² (Srivastava et al., 2002), was, however, bifurcated around the same time when this study was planned (between May 2021 and April 2022). Hence, the new Sonitpur district with an area of 2,109 km² (Assam, 2022) was selected for this study.

For demarcating different SEZs, we first performed a Land Use and Land Cover (LULC) classification using the satellite images Sentinel 2 (10 m resolution) of 19 November 2020 (less cloud cover) using the supervised classification (Campbell and Wynne, 2011) in ArcGIS version 10.8. The major land use types identified were river, protected/plantation area, builtup/settlement area, and arable land. Based on field verification (December–February 2020) and further literature review (Srivastava et al., 2002; NRSC, 2019; Chaturvedi et al., 2021; Mahato et al., 2021), four prominent SEZs, i.e., PA, RI, RM, and TE, were demarcated. The PA in the study site refers to three major classes, viz., the national parks (IUCN Category II), the wildlife sanctuary (IUCN Category IV), and the reserve forest (forests accorded a certain degree of protection according to the Indian Forest Act 1927) and comprises 45% of the study area (NRSC, 2019; Assam, 2022). The RI SEZs form another major dominant feature of the district, with three major rivers, i.e., Brahmaputra, Kameng, and Gabharu, and about 18% of the area of the district (NRSC, 2019). The TE SEZs are the areas under a larger tea estate plantation in the district. Tea cultivation started during the British colonial period in Assam, and at present, there are 799 TE in Assam and about 59 in Sonitpur district (DOTTAA, 2023). The current tea labor force in Assam is primarily composed of the descendants of people who were brought over from the areas that now constitute the tribal dominant states of Jharkhand, Orissa, and Chhattisgarh to work there during the colonial era. These individuals are now collectively referred to as tea tribes or Adivasis (Mahanta et al., 2015; DOTTAA, 2023). The cities (major built-in areas) were excluded as the study focused on HGs in rural areas. To commensurate with this, the RM zone was considered an important feature; the villages that had at least one major market a week, had facilities for banks or post offices, and were in close proximity to the National Highway (2 km) were considered as RM.

The SEZs PA, RI, and TE comprise 45, 18, and 12% of the total area of Sonitpur district (Census, 2011; Assam, 2022). A total of four replicate villages were then selected from each SEZ, i.e., 16 villages in total. The villages for PA were selected from the vicinity of Nameri National Park (Sonai Miri, Bhalukmari Pathar) and Sonai Rupai Wildlife Sanctuary (Naharani Basti Gaon, Urohiloga). These villages lie on the fringes of PAs (up to 5 km from the PA).

For RI, the villages on the bank or 2 km from the bank of the Brahmaputra River (Sithalmari, Bhomoraguri, Siddeswari) and Kameng or Jia Bharali River (Tow Bhanga) were selected. The villages were selected based on the number of flooding days that they experienced, i.e., a minimum of 60 days. The RM villages were selected based on the classification scheme mentioned earlier, and care was taken to select them from distinct blocks of the district (i.e., Pitha Khowa: Block—Tezpur; Thelamara Ghat: Block—Dhekiajuli; Jamugurihat: Block—Naduar; and Goraimari: Block—Balipara).

The TE-based (inside and around TEs) villages representing the distinctness of the eastern (i.e., Dhekialuji and Singri) and western parts (Phulbarie and Addabarie) of the district were selected for the study. This LULC classification scheme and the villages selected for the HG study are represented in Figure 1. While selecting the villages, it was made sure that they represented distinctive features of the district, i.e., they were selected from different blocks and along different directions. All the villages except the TE villages were revenue villages; the TE villages are generally the inhabitations inside the TE owned by the plantation companies, where respective company rules are followed (Mahanta et al., 2015). The permission to do the agrobiodiversity survey in all these villages was obtained from the District Collector of Sonitpur District. Permission was also obtained from the individual homegardeners in a consent form before participating in the study.

The sample size of households in each village was determined using the following formula (Corvar, 1974; Abdoellah et al., 2020):

where n is the total number of samples, N is the population size, z_y is the normal distribution in y quantile, π is the proportion of sub-population, and δ is the margin of error. According to the 2011 Census, there are 352,647 rural households in the erstwhile Sonitpur district. Assuming a 7.1% margin of error and a 95% confidence level, 191 households in 16 villages—approximately 192 were surveyed to ensure that the households were distributed equitably among the villages. With 48 villages per SEZ, 12 households per village were selected using the probability sampling technique of simple random sampling. At first, the map of the village with the major feature was obtained from the Gaon Bura (village head). Then the households were randomly selected, starting from one cardinal direction and entering the village.

Agrobiodiversity inventorying was performed by field visits (Avilez-López et al., 2020) between March 2022 and April 2023. For agrobiodiversity inventorying, the questionnaire was administered to the landowners of the HGs, and where they were not present, the questions were asked of the person responsible for most managerial decision-making in the HG (43% of the respondents were women). In addition, a questionnaire method was used to understand the general economic status, management practices followed, challenges in homegardening, and future plans for HGs. Apart from HGs attached to the residential plots, some respondents maintained land parcels away from the households too. However, considering the standard definition of the term (Kumar and Nair, 2004; Galhena et al., 2013), we considered only the cultivation practices within the fenced area of the households or very near the dwelling units. Das and Das (2005) reported that the average size of HGs in Assam ranged from 0.02 ha to 1.20 ha. However, in TE, where the average size of each worker’s quarter is 0.04 ha (Kar, 1984; Ahmmed and Hossain, 2016), we considered anything larger than 0.01 ha as an HG. Since the homegardeners in Sonitpur district generally used separately designated spaces within the HGs for the cultivation of vegetables, cereals, ornamentals, medicinal bamboo, boundary plants, plantation crops (e.g., Betel Nut [Areca catechu], Teak [Tectona grandis], Tea plant [Camellia sinensis], Rubber tree [Hevea brasiliensis], etc.), and livestock, we adopted whole-plot sampling and counted all plant and livestock species and their relative distribution in the HGs (Poot-Pool et al., 2012). All tree and shrub species having diameter at breast height (DBH) of 10 cm or more were enumerated. Since deliberate cultivation of vegetable crops and other herb species was done in land parcels assigned to these crops, their area was measured, and the name and type were noted. The names of various ornamental plants, climbers, and medicinal plants were recorded along with their frequency of occurrence. Homegarden age was recorded as stated by the respondent, which was cross-checked with the village head. The ethnicity/linguistic group to which particular homegardeners belonged was noted based on how they identified themselves and also as per the ethnicity classification given in the Census (2011).

Plant species diversity for each HG was computed using the Shannon diversity index.

where pi is the proportion of individuals belonging to a specific species determined by dividing the count of individuals from that species (n) by the total number of individuals observed (N). In this context, ln represents the natural logarithm, Σ denotes the summation of these calculations, and S represents the total number of species (Shannon, 1963).

Dominance index (Cd)

where pi is the proportion of individuals belonging to a specific species determined by dividing the count of individuals from that species (n) by the total number of individuals observed (N) following Simpson (1949).

The species evenness was calculated using the Pielou evenness index using.

Dp=H/lns

where H is the Shannon index, and S is the total number of species (Pielou, 1969).

Species richness was calculated using the Margalef Index.

where S is the total number of species, N is the total number of individuals, and ln is the natural logarithm (Margalef, 1958). The data for herb, shrub, and tree components were pooled to perform the above computations. Apart from the plant diversity estimate, the Margalef Index was also used to estimate the livestock diversification index (Mekuria and Mekonnen, 2018). Whether the difference between variables of HGs of different socio-ecological types was statistically significant was tested by ANOVA followed by the Duncan multiple range test (George and Christopher, 2020). The Sorenson test of similarity (S.I.) using.

S.I.=2l×100/2a+b

where L is the number of species two samples have in common, a is the number of species in the first sample, and b is the number of species in the second sample was used to find the similarity between HG of different ethnicities and different SEZs. Furthermore, cluster analysis was performed on the variables Shannon Index and Livestock Diversification Index. In addition, regression and correlation coefficients were derived to understand the relationship between different variables. All the data analysis is performed in R, Excel, OriginPro2023, and PAST software.

There was a preponderance of HGs in all the SEZs evaluated (Table 1). However, a large variation in size and main purpose was noticed. The preference for different crop types based on the dominant features of the SEZs and the cultural and economic background of the respondents led to four major types of HGs in the study area. The PA-type HG, which were mostly documented from the PA, were HGs with usually large sizes (>0.5 ha) often incorporating plantation species. Here, the size ranged from 0.07 ha to 1.806 ha. These HGs also had the highest number of stems of plants greater than 10 cm DBH. In addition, they had the maximum land, i.e., 53% devoted to plantation crops like Areca catechu, Hevea brasiliensis, and Camellia sinensis, and 47.73% of HG in the PA also had ponds with as many as five varieties of fish. The majority of HGs reported here are old HGs (age > 40 years) and had designated spaces allotted to vegetables, ornamentals, plantation species, and ponds. Intercropping species like pineapple (Ananas comosus) with lemon (Citrus limon) and Areca catechu was fairly common. Seventy-nine percent of HGs had bamboo species, with almost 50% of HGs having more than one species of bamboo. Often, bamboo was planted at the end of HGs, connecting them to the field. These HGs also had a sizeable number of livestock (and a high livestock diversification index) and were also more commercialized. The major characteristics and composition of these HGs are shown in Figures 2A, 3A,B.

The RI type HG majorly documented from the RI SEZs were the ones that were more prone to floods (almost 60 days), had large variations in size (ranging from 0.013 ha to 0.67 ha), often had less livestock reared, and had more preference for vegetable cultivation. Although the RI zones were vulnerable and prone to floods, homegardening was enthusiastically pursued by the farmers in this zone. In RI HG, the maximum land is for vegetable cultivation (51%). The major vegetables cultivated were okra (Abelmoschus esculentus), brinjal (Solanum melongena), potato (Solanum tuberosum), radish (Raphanus sativus), cabbage (Brassica oleracea), and cauliflower (Brassica oleracea) as shown in Figures 2D, 3C,D. In the RI zones, apart from the HGs attached to the household, the cultivation was also carried out in the Char (floodplain sediment island) areas. The most widely cultivated plants on the Char lands were jute (Corchorus olitorius), vegetables, and black lentil (Vigna mungo).

The RM type HG had a relatively average size ranging between 0.013 ha and 0.47 ha, and since RM SEZs are multigenerational old (as they are among the oldest habituated areas in the study site), even though they are reduced in size, they represent high cultural values, and they seem to have the highest concentration of culturally important species like Orchid Rhynchostylis retusa (Kopu Ful; Figure 3A); GI-tagged Citrus limon (Kaji nemu); and Litchi chinensis Sonn (Lichu; Figure 3F). In addition, almost 56% of the HGs had more than three varieties of banana. Most common among them were Musa chinensis (Jahaji Kol), Musa champa (Cheni-champa Kol), Musa assamica (Malbhog Kol), Musa paradisiaca (Kach Kol), and Musa gigantea (Bhim Kol). These all play important roles in Assamese cuisine. Vegetable cultivation and plantation were allotted approximately the same extent of land in rural market-based HG, i.e., 45 and 47%, respectively, as shown in Figure 2B. Ponds are present in 21% of RM HGs. In addition to being most proximate to the market, the species diversity was also very dominated by ornamental species.

The TE type HG was the HGs mostly documented among the TE workers were smaller (size <0.03 ha), ranging from 0.013 to 0.134 ha, had few tree species, mostly ornamental, fruit, and vegetable species, and had small places for worship and poultry or small avian species (67% of HGs) reared for meat and eggs; however, the number of large cattle reared were few (15%). In TE-based HGs, the maximum area was used for vegetable and ornamental plant cultivation, as shown in Figure 2C. Ponds were present in less than 5% of HGs in TEs. Since the homegardening area was small, it was therefore judiciously used for ornamental and vegetable cultivation, along with scattered tree species for cultural values and subsistence (Figures 3G,H).

In addition, the mean HG age is found to be 59 years in PA HG, 45 in RI zones, 47.4 in HGs adjacent to RMs, and 34.5 in the TEs. In addition, though the average age of HGs does not seem to be significantly different (p > 0.005) in each SEZ, the oldest HG of 110 years was reported from the RM locations, and the youngest of 5 years was observed in the RI as well as RM types.

The plant composition in the HGs of the PA was found to be very diverse, with an average number of plant species of 43 (range 26–75). The number of plant species was found in the order PA > RM > RI > TE. In addition, the total stems of plants greater than 10 cm DBH were highest in PA and were observed in the same order. Overall, 101 tree species were identified from HGs in the study area. Areca Catechu (betel nut) was the most frequently observed tree species in all SEZs. In addition, three major species of bamboo were observed in our study area, viz., Bambusa balcooa (Bhaluka Bah), Bambusa pallida Munro (Bijuli Bah), and Bambusa nutans (Mokal Banh). Sixty-four percent of HGs surveyed had one or more species of bamboo. Moreover, four varieties of bananas were found to be cultivated, and 76% of HGs were found to have at least one variety of banana. Although diverse functional groups of plants were observed, most of the plant species obtained in the HGs of each SEZ were food plants (36.6, 45.7, 35.8, and 36.5% in the HGs of PA, RI, RM, and TE, respectively). The species classified for food include vegetables, fruits, and other edible species. Figure 4 shows the distribution of plant species in different use categories and their distribution in corresponding SEZs.

The statistical distribution of plant species diversity observations in the HG of different villages in diverse SEZs is depicted in Table 2. The Shannon diversity index was estimated to be highest at 1.48 in TE HGs and 1.44, 1.39, and 1.09 in HGs of RI, RM, and PA HG, respectively. The Simpson diversity index was also estimated at the highest values of 0.66 in TEs HG and 0.62, 0.51, and 0.45 in RI, RM, and PA HGs, respectively. The Margalef index was highest at 2.76 in RM and 2.58, 2.57, and 2.39 in TEs, RI, and PA HG, respectively. The Evenness index was highest at 0.70 in TE and 0.61, 0.40, and 0.31 in RI, RM, and PA HG, respectively. Table 2 shows that diversity indices, i.e., Shannon, Simpson, Margalef, or Evenness, were significantly different in each SEZ. However, there were no significant differences among the villages in the same SEZs except for those villages in the TE or RM.

Livestock formed the most important part of 68% of all HGs surveyed; if we also take into consideration poultry farming and rearing avian birds for eggs and meat, this number becomes 85%. The common livestock observed in the field were Bos taurus (cattle), Bubalus bubalis (buffalo), Capra aegagrus hircus (goat), Sus scrofa domesticus (pig), and Ovis aries (sheep). In birds rearing for meat and eggs, the common birds reared were Gallus gallus domesticus (chicken), Anas platyrhyncos (ducks), and Columba livia (domestic pigeon). In addition, 20% of the homegardeners had more than one large animal. The highest average livestock variety and numbers were found in the HGs of PA (range 3–16) and least in the HGs of TE (range 1–3), shown in Table 2. A total of 21.87% of HG had a pond with an average of three fish varieties and a maximum of six varieties. The most common of them are Rohu (Labeo rohita), Catla (Catla catla), Mrigal carp (Cirrhinus cirrhosis), and Bariala (Aspidoparia morar). The SEZ with the highest number of ponds attached to HG is the PA (47.73%). The HG in the study area also reportedly had quite a high average livestock diversity index (0.67), and the PA had the highest (1.04) as shown in Table 3. Often, fodder tree species like Gmelina arborea Roxb were found to be cultivated in the HG to meet the requirements of this livestock. In addition, a small, outside, separated kitchen-like space was found to be present in almost all HGs where the food for livestock was prepared. For preparing the fodder, the fuelwood tree species cultivated in the HGs were used. Table 3 gives the agrobiodiversity distribution in different villages of diverse SEZs. The Sorenson similarity index among different SEZs for plant and livestock diversity was found to be highest between the HGs of the PA and RM (0.82) and least between the HGs of the TE and RI areas (0.58).

In total, we encountered HG of eight linguistic and ethnic groups in the study region: the Assamese, Bodo, Bengali, Hindi Speaking, Mishing, the Gorkhalis, the Tea Tribes or Adivasis (staying outside the TEs), and the Tea Garden Workers (staying inside the TEs). Except for two SEZs, which were dominated by one ethnic group, others were more heterogeneous. The RM area is dominated by Assamese, the TE area is found to be dominated by the tea tribe, and the PA and RI are found to be dominated by a mixture of tribal and non-tribal groups. The largest size of HG was found to be in the Mishing tribe-managed HGs (0.268 ha) and the smallest size was found in the HGs of tea tribes (0.013 ha). The oldest unaltered HG was found in the RM (110 years) in the Assamese household, and the youngest HG was also found in the same zone among the Assamese household (5 years).

The average diversity indices among different ethnic and linguistic groups are shown in Table 4. The Shannon diversity index was found to be highest at 1.56 for the tea tribe. The average Simpson diversity index was also 0.66, with the highest for the tea tribe and the lowest 0.48 for Bengali. The Margalef Index was found to be highest at 3.01 for Adivasi and lowest at 2.43 for the Bengali linguistic group. This tendency to have high diversity indices in TE HG could be because there is no scope for planting plantation species because of their small size. This could be further verified from the observation that though diversity indices are higher for HGs of the tea tribe, the number of plant species observed is the least there (28).

The highest mean number of plant species was found in the HGs maintained by Mishing Tribe (51) followed by Bodo Tribe (42), Assamese (41), Gorkhali (40), and Adivasi (33). The highest average variety (up to 4) and several livestock (up to 8) were found to be reared in Mishing HGs. Bos taurus and Capra aegagrus hircus were found to be raised in HGs of all communities, whereas Sus scrofa domesticus was mostly reared in the tribal communities, i.e., Bodo and Mishing HGs. For poultry and bird rearing, it was noticed that Anas platyrhyncos and Columba livia were predominantly reared in the Assamese HGs, and another form of small poultry was predominantly observed in the TE HGs. Though big livestock was comparatively fewer in the TE HGs, however, they had small birds and poultry more frequently. HG age is often reported as an important characteristic that tells about diversity, usage trends, and carbon stock. The highest Sorenson similarity index among plant species of HGs of an ethnic group is between Mishing and Bodo (0.78) and the lowest between Tea Tribe and Mishing (0.41). The comparative difference in these parameters among diverse ethnic groups is shown in Table 4.

HGs as an important site for the conservation of both floral and faunal diversity have been mentioned in the literature (Sharma et al., 2022). Similar to the HG study (Das and Das, 2015; Barbhuiya et al., 2016), we found the critically endangered species Agarwood (Aquilaria malaccensis) to be frequently cultivated in the HGs in all three SEZs except TEs. In addition, the endangered tree species Livistona jenkinsiana Griff (Fan palm) and Mesua ferrea L. (Cobra saffron) were observed. The IUCN-vulnerable species like Canarium strictum Roxb (Black Dhup) was also observed.

Apart from these tree species, many culturally important species like the Orchid Rhynchostylis retusa (Kopo ful; Figure 3E), a very commonly used plant for the Bihu festival (majorly by the Assamese family), Euphorbia splendens (Bathou), a very auspicious tree for the Batho religion (majorly of Bodo people), and Citrus limon (frequently used in Assamese cuisine) were frequently observed. This is similar to literature that mentions HGs as the site for the conservation of culturally important species (Galluzzi et al., 2010).

In addition, unique management practices were observed in the HGs, some of them being the use of homemade biopesticide and fertilizer, mixed cropping, and the use of indigenous seeds. Moreover, the majority of knowledge exchanges for plant species selection and techniques of homegardening were family- or community-based. Only 7% of homegardeners reported having received any formal training.