Enset-based agricultural systems in Ethiopia: A systematic review of production trends, agronomy, processing and the wider food security applications of a neglected banana relative

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Plants, People, Planet © New Phytologist Trust *This article is dedicated to the memory of our colleague and friend Dr Mark Goodwin. 1Royal Botanic Gardens, Kew, Richmond, Surrey, UK 2Department of Genetics and Genome Biology, University of Leicester, Leicester, UK 3Bioversity International, c/o ILRI, Addis Ababa, Ethiopia 4Royal Museum for Central Africa, Tervuren, Belgium 5Department of Biology, Hawassa University, Hawassa, Ethiopia 6Center for Food Science and Nutrition, Addis Ababa University, Addis Ababa, Ethiopia 7Department of Plant Biology and Biodiversity Management, Addis Ababa University, Addis Ababa, Ethiopia


| INTRODUC TI ON
and enset (Ensete ventricosum (Welw.) Cheesman) (Harlan, 1969(Harlan, , 1971; Khoury et al., 2016). Whilst coffee has become one of the most highly traded global commodity crops (Ponte, 2002), and others have widespread and growing international importance (e.g., tef, [Cheng, Mayes, Dalle, Demissew, & Massawe, 2017]), enset has remained largely overlooked. It is perhaps the only major Ethiopian domesticated crop that has never been cultivated outside Ethiopia, despite wild populations of E. ventricosum occurring from Ethiopia, and eastern DR Congo (along the rift valley mountain ranges) to South Africa (Figure 1). This highly localized cultivation (though by many socio-cultural and ethnolinguistic groups) may in part explain limited research to date, despite indications that enset could play a key role in meeting food security challenges (Brandt et al., 1997), not only in Ethiopia, but also further afield.
Enset (Ensete ventricosum, Musaceae), is an unusual crop/plant from a sister genus to the bananas (Musa). Enset differs from domesticated bananas in that the mature plant does not produce edible fruit (these are filled with numerous large and hard seeds, similar to many other wild banana species). Instead the plant is grown for 3-12 years, before the pseudostem and corm are harvested and collectively processed into starchy food products (Shack, 1966). Enset serves as a staple food for about 20% of the Ethiopian population, over 20 million people, mainly in the south and south-west of the country (Borrell et al., 2018).
Under appropriate conditions it is estimated that 60 mature plants can provide enough food for a family of five to six people, over the course of a year, when consumed with other dietary components; typically meat, dairy, and cabbage (Demeke, 1986). use of this neglected starch staple. In conclusion, we discuss the challenges and opportunities for enset cultivation beyond its restricted distribution, and the regional food security potential it could afford smallholders elsewhere in Southern and East Africa.

K E Y W O R D S
agricultural systems, agroforestry, crop wild relative, food security, home gardens, indigenous knowledge, medicinal use, sustainable agriculture, tropical crop ecology F I G U R E 1 (a) The distribution of wild and domestic enset in Eastern and Southern Africa. It should be noted that domestic enset displays a range expansion east towards and across the Ethiopian rift valley system, beyond the current distribution of wild enset; (b) inset Ethiopia with the main study areas discussed in this article Enset farming systems contribute to the long-term sustainability of food production through several mechanisms. Enset is perceived to be relatively tolerant of drought (Garedew, Ayiza, Haile, & Kasaye, 2017;Zerfu, Gebre, Berecha, & Getahun, 2018), withstand heavy rain, tolerate flooding, and endure frost damage (Degu & Workayehu, 1990). Enset can be harvested at any time during the year, be harvested at any growth stage over a period of several years (up to an including the early flowering stage), and the fermented products can be stored for long periods (Brandt et al., 1997;Garedew et al., 2017;Sahle, Yeshitela, & Saito, 2018).
This combination of characteristics gives enset an important role during times of famine in the areas in which it is traditionally cultivated. Enset's resilience and versatility has earned it the name 'The Tree Against Hunger' (Brandt et al., 1997) and it already forms the basis of many households' food security (Negash & Niehof, 2004;Sahle et al., 2018).
We consider that there are two major challenges for the improvement of existing enset farming practices and the adoption of enset-based agriculture outside of the current distribution in south west Ethiopia. The first, is poor characterization of ecological requirements, genetic diversity, and the impact of changing climate, recently reviewed in a companion paper by Borrell et al., (2018). The second, addressed here, is poor documentation of the extensive ethnobotanical knowledge and practices associated with enset cultivation and recent enset household production trends in a national context. Smallholders account for 96% of the cultivated area in Ethiopia (Taffesse, Dorosh, & Gemessa, 2013), with at least 48 ethnic groups native to the Southern Nations, Nationalities and Peoples Region (the major enset growing region), 88 recorded languages (Population Census Commission, 2007) belonging to four major groups (Omotic, Semitic Nilotic, and Cushtic) and a corresponding multitude of ethnobotanic practices. Therefore, here we investigate and summarize reported uses and management of enset, recent trends in cultivation, emerging by-products and opportunities for improvement. We hope that by providing a framework within which to report enset ethnobotanic research, we can enable and catalyse development of this important Ethiopian food security crop.

| National production and yield trends
Analysis of data from Ethiopia's Central Statistics Agency (CSA, 1995, shows the area of land under enset production has increased approximately 46% over the past twenty years, whilst yield has been reported to increase twelve-fold ( Figure 2; Dataset S1; see Methods in Supporting Information), implying substantial productivity improvements. We have concerns over the validity of these data for several reasons. First, these data are at odds with several studies reporting that farmers perceive enset production to be declining (Abebe, 2013;Negash, 2001;Yemataw et al., 2017;Zippel, 2005). Ethiopia, showing (a) hectares in production (circles) and yield (squares). Dashed lines indicate interpolation between incomplete data. Red circles denote values based on incomplete reported data. Yield data is not available prior to 2008. All data sourced from the Central Statistics Agency Agricultural Sample Survey ; (b) national yield for the ten highest yielding crops in Ethiopia (2017-18); and (c) yield per hectare for the ten most widely produced crops . Enset ranks second in yield per hectare and is the fourth highest yielding crop per hectare Second, over a similar period, (1996Second, over a similar period, ( -2017 the Ethiopian population increased 77% from 59 to 105 million (Center for International Earth Science Information Network, 2017). This disparity could be partly attributed to a proportional reduction of enset land area per person, as enset agriculture fails to keep pace with population growth, however, this is unlikely to be the case for overall yield. Third, there appears to be little evidence for policy or crop development-based drivers that could have contributed to the reported productivity increase. Instead Cochrane and Bekele (2018) report undocumented 'methodological changes' at the Central Statistics Agency and highlight concerns over quality, methods and politicization of the data, with inconsistencies over several other tuber crops such as taro and sweet potato.  (Table 1). Whilst much of this variation can be attributed to the challenges in surveying enset outlined previously (Box 1), there are also important differences in enset cultivation density and yield in different regions, likely attributable to local agroecological conditions (Shank & Ertiro, 1996). For example, lowland parts of the Gurage region are typically drier than other enset growing areas and this may be a contributing factor to it having one of the lowest enset planting densities, reducing competition between plants (Sahle et al., 2018) (see Table 1). In the same study, the authors show annual yields per hectare can vary from 2.9-7.7 tons when compared across local agroecological zones. Pijls, Timmer, Wolde-Gebriel, and West (1995), based on another survey in the Gurage region, measured the energy yield per area and unit of time for enset as 1,450 kcal/m 2 per year, substantially higher than that of other common Ethiopian staples such as cereals, potato, sweet potato and banana. Therefore, when considered per unit of space and time, in some localities enset earns the title of most efficient Ethiopian crop (Tsegaye & Struik, 2001). Despite this reputation, continuing to develop higher enset yields remains a key focus for Ethiopia's agricultural system as asserted by Taffesse et al. (2013) and many other researchers.

| Regional importance
The prevalence of enset agriculture across Ethiopia is highly heterogenous, depending on both amenable agroecological conditions and

Box 1 Challenges and sources of variation in estimating area under production and yield for enset agriculture in Ethiopia
Estimating hectares under production • Transplanting occurs up to five times, at increasing spacing meaning that enset density varies with age class.
• Final mature spacing differs in different regions, potentially dependent on variables such as precipitation and soil moisture.
• Enset is sometimes intercropped with other food or cash crops such as coffee (Coffea arabica); in some areas such as Gedeo and Sidama, enset is also grown in the understory of agroforestry systems.
• Enset is harvested most frequently at 4-6 years old, but may be harvested from as little as 2 years, until flowering, which may be 12 years in some conditions.
• A variable portion of total enset hectarage (often 15%-25%) harvested each year, depending on household requirements and performance of other crops.
• Annual mortality, including impact of pests and diseases is poorly known.

Estimating yield
• Yield varies due to varying performance in highly heterogeneous agroecological conditions across the crops' distribution.
• Yield likely varies due to differing performance of genetically distinct landraces.
• Yield varies based on age at harvest.
• Yield per hectare sometimes comprises a hectare of harvested enset, or alternatively the yield of enset that can be sustainably removed from a hectare each year (for example, ~1/6th of the total area).
• Yield and ratio of different food products (kocho, bulla and amicho) vary depending on cultivated landrace.
• The number of plants harvested and yield may be withheld or misreported.
• Water content of derived products rarely quantified.
• Local unit names and weights for enset products also vary, even across small spatial scales (e.g. neighbouring communities).

| Household utilization, supply and demand
Household utilization of enset (available CSA data averaged for the period 2008-16) also indicates its important role in household food security, with enset having the highest consumption rate for both humans and livestock, the most used crop for 'wages in kind' and the lowest sale rate (Figure 4). It is perhaps due to this high degree of TA B L E 1 Enset yield per plant, yield, and number of plants per hectare reported in the literature

| Propagation and tissue culture
Domesticated enset is exclusively propagated vegetatively using suckers from the appropriately prepared corm of a young plant. As enset plants are generally harvested before full emergence of the inflorescence, seeds are rarely available to farmers, although preference for vegetative reproduction is normally attributed to the increased vigor of the suckers/plantlets (Alemu & Sandford, 1991).
There is a single domesticated enset landrace from the Ari region, known to spontaneously sucker from the leaf petiole base. Enset seeds from cultivated landraces are characterized by low germination rates (Negash, 2001), while higher germination rates have been observed for wild enset (Z. Yemataw, pers. comm.). Wild enset is assumed to reproduce exclusively via seeds, because spontaneous suckering has not been observed.
Suckers initiate from growing buds found in concentric circles on the upper section of the corm. While some variation in corm preparation are reported between different ethnic groups (Diro, Haile, & Tabogie, 1996) the principal aspects are consistent.
To produce suckers for planting, a farmer will harvest the pseudostem and save the corm of a desired plant to initiate suckers.
In contrast to banana, naturally occurring sucker production is rare due to the very high dominance of the apical meristem, which must be destroyed/removed to stimulate sucker production. After the removal of the apical meristem, corms are kept in its entirety or are split into two or four equal parts. After preparation, the corm/corm pieces are planted 20-30 cm deep in loosened soil, mixed with manure. Corms from plants of 2-4 years old and about 10-35 cm in diameter are preferred (Bezuneh & Feleke, 1966;Yemataw, Mohamed, Diro, Addis, & Blomme, 2014). Each corm (piece) will produce between 40 and 200 suckers, depending on the cultivar, size and age of the mother plant, soil characteristics, rainfall, land preparation and time of planting (Shumbulo, Gecho, & Tora, 2012). Suckers appear 2-3 months after planting (Tsegaye & Struik, 2002). Time to sucker emergence is shorter for split corms and a higher number of seedlings are generated this way compared with entire corms (Karlsson, Dalbato, Tamado, & Mikias, 2015). Early emergence is associated with more vigorous suckers, which promotes success of establishment, higher and earlier yield (Tabogie & Diro, 1992). In areas where extended droughts are common and watering is difficult, it may be preferable to use whole corms to reduce corm desiccation; in other areas watering and application of manure is advised (Karlsson et al., 2015).
Various technologies have been developed to increase sucker production . Makiso (1996) refers to a macro-propagation method to regenerate enset plants by cutting the corm, with leaf bases intact, vertically into small pieces which are then planted in plastic tubes or bags and raised in a growth chamber with proper temperature and humidity (20°C and moist medium of soil or any other material). Other rapid propagation methods have been reported by Diro, van Staden, and Bornman (2004), including zygotic embryo culture (Bezuneh, 1980;Diro, van Staden, & Bornman, 2003;Negash, 2001;Negash, Puite, Schaart, Visser, & Krens, 2000), shoot tip culture (Afza, van Duren, & Morpurgo, 1996;Negash et al., 2000;Zeweldu, 1997), and callus culture and somatic embryogenesis (Mathew, Manuel, & Philip, 2000;Mathew & Philip, 2003). Micro-propagated enset plants are prone to various issues, including blackening resulting from phenolic oxidation of explants, necrosis and the formation of unwanted callus. Low levels of multiple shoot formation are also observed from shoot tip explants (Diro et al., 2004), thus these approaches do not yet provide an alternative to sucker propagation.

| Transplanting and harvest
Harvested suckers are planted in a nursery plot (0.5 to 1 m 2 plant −1 ) where they grow for about one year. Subsequently, suckers are consecutively transplanted into ever more widely spaced arrangements, with a final minimal spacing of 2 to 4 m 2 plant −1 (Bezuneh & Feleke, 1966;Hiebsch, 1996;Tsegaye & Struik, 2002), with wider spacing more common in areas of lower soil moisture e.g., Gurage (Sahle et al., 2018; Table 1). Since the corm and pseudostem are harvested, higher yields can be expected by successive transplanting steps which delay flowering, thus allowing a longer period of time for vegetative growth, which includes assimilation of starch in the pseudostem and corm. Therefore, direct transplanting is advised when early yields are the objective, but more frequent transplanting will result in higher yields per plant (Tsegaye & Struik, 2000). Once equilibrium (planting = harvesting) has been achieved, annual yields will be higher for twice transplanted plants compared to once transplanted as this encourages greater partitioning of dry matter to the harvestable parts Tsegaye, 2007).
Systems involving up to five transplanting stages are used in some areas (J. Borrell, pers. obs.). An additional benefit of repeated transplanting is the smaller overall space required (i.e., all plots with varying plant size/density are grown together). Full maturity of the enset plant is reached after 4 to 12 years, depending on the landrace and altitude of the farm, with higher locations significantly increasing cropping cycle duration (Negash, 2001). The ideal moment to harvest is at inflorescence emergence. At that time, dry matter yield is highest.
After flowering, assimilates are redirected toward the inflorescence and away from the pseudostem and corm (Tsegaye & Struik, 2000). A summary of growth stages is presented in Figure 5.

| Enset farm management and fertilization
As a perennial crop enset fields generally do not require tilling (depending on intercropping practices), though sequentially spaced and sized holes are dug for subsequent transplanting stages and then reused for successive individuals. Due to large leaf surfaces rainfall is intercepted, limiting erosion. Enset leaf bases also appear to be adapted to trap water close to the pseudostem, with pools of water observed even several months into the dry season (Figure 5h; J. Borrell, pers. obs.). Irrigation practices have not been observed. Enset appears to have a larger root system and thicker root cords compared to banana (possibly aiding plant stability in its preferred natural habitat of steep riverbanks and slopes), with 89%-96% of roots in the upper 40 cm soil layer (Blomme, Sebuwufu, Addis, & Turyagyenda, 2008). Enset leaves and discarded parts of the pseudostem are frequently used as a mulch, which enhances soil moisture and organic matter.
Enset farming systems are reported to require relatively little off-farm input, however manure application is widespread and considered essential by enset farmers. Manure is derived from livestock which are traditionally housed immediately adjacent to the enset for ease of transfer. As a result soil, fertility in enset fields is reported to be greater than in surrounding fields or pastures (Shank & Ertiro, 1996). The rate, timing, and method of manure application varies among households and depends on the growth stage of the plantation and the availability of manure (Tsegaye & Struik, 2002).
Compared to other crops, there are very few fertilizer response trials for enset. Bezuneh (1984)  "Tuzuma" and "Adow" were achieved, however control comparisons were not reported. Uloro and Mengel (1996)  yields in the range of 10-12 t/ha, which is three to four times higher than without fertilizer application.

| Enset agrisystems
Enset-based agriculture is considered one of the most sustainable of the indigenous farming systems in Ethiopia, with a human carrying capacity higher than other crops and cropping systems for the same agroecology and inputs (Brandt et al., 1997), although further empirical evaluation is necessary. Many enset-based farms are derived from forest, whereby farmers clear away the undergrowth to plant enset, coffee, and other crops, leaving the upper story trees, resulting in multi-story agroforestry systems which are thought to have remained relatively stable for centuries (Kippe, 2002).
Ethiopian agricultural systems are typically highly diverse. Asfaw and Nigatu (1995) identified a total of 162 crop species cultivated throughout the highlands of southern, western, eastern and central Ethiopia. On average, enset farms produce more than 10 different crop and livestock species (Sibhatu, Krishna, & Qaim, 2015), with farmers growing up to 20 different enset landraces within one plantation (Zippel, 2005). With the lowest proportion sold of any Ethiopian crop (Figure 4), cash crops and livestock are integral to enset-based production systems (Kandari, Yadav, Thakur, & Kandari, 2014;Tsegaye, 2002). However, enset does provide a high level of flexibility and security, in that larger enset plants can be sold as a standing plant before being processed or a portion of fermented kocho or bulla may be taken from the storage pit for sale at any time.
Distinctive enset production systems can be categorized based on environmental, agronomic, and cultural criteria, and the importance given to enset within each cropping system (Brandt et al., 1997;Shank, 1994;Westphal, 1975), for example enset-coffeemaize is the dominant system of Sidama zone. Additional variation is observed due to wealth of farming households, farming skills, landholding size, availability of resources, access to a highway and altitude (Abebe, 2005). Mellisse, Descheemaeker, Giller, Abebe, and Ven (2018), in an analysis of home gardens, found that overall crop productivity was lowest in the traditional enset-coffee systems (1,820 kg Dry Matter [DM] ha −1 ) and highest in the newly evolved enset-cereal-vegetable systems (3,020 DM kg ha −1 ).
Energy productivity from food crops was higher in enset-based systems (43 Gigajoules (GJ) ha −1 ) compared to other systems, but the revenue was lowest in enset-based systems (719 US$ ha −1 ) and highest in newly evolved chat-based systems (6,817 US$ ha −1 ).
Thus many farmers may need to adapt to new agronomic systems and crop combinations that maintain enset's role as a sustainable staple, whilst ensuring sufficient cash crop income. labor-intensive and tedious (Hunduma & Ashenafi, 2011). Garedew et al. (2017), in an analysis of the indigenous knowledge of enset management in the Shekicho people of the SNNPR, highlights the danger of indigenous knowledge being lost as younger people turn away from enset-based agriculture, and thus the documentation of the diversity of agronomic practices, and especially processing methods, is a priority.

| Enset food products
At harvest, starch is decorticated ( Borrell, pers. Obs.), although there are varieties that are claimed to produce good quality amicho even at maturity (e.g., landraces 'Nifo' and 'Zoober'). It is the least commonly encountered preparation method. Similarly, bulla is only available in much lower quantities than kocho, and is generally more expensive because of the additional processing involved. As a result, a premium price is paid as it is pure starch, with no fiber or other stem-derived materials.

| Enset fiber and packaging products
Fiber is extracted from the pseudostem and leaves, largely as a byproduct of kocho production ( Figure 6) . The fibers are dried to make sacks, ropes, sieves and mats.
The fibers can be 4 m or more in length, depending on the height of the pseudostem, and are strong and flexible. Fiber can also be extracted from the leaf sheath, petioles and midrib, which are commonly used for animal feed, compost and fuel for fires. Teli and Terega (2017) analyzed the characteristics of enset fiber, including tensile strength, elongation at breaking point and thermal stability, and found it to be comparable with many other natural fibers such as abaca, flax, sisal, hemp, and jute.  found enset fiber to have a tensile strength of 9.8-17.5 kg g −1 m −1 .
Enset fiber has the potential to be used for a range of novel applications beyond traditional uses, and there is likely to be variation in performance across genetically diverse landraces.
Enset leaves are used for many purposes, including lining the fermentation pits for kocho, wrapping kocho and other traditional breads during baking, transporting butter and honey to market, making mattresses and cushions, and as fuel (Tedla & Abebe, 1994).
Leaves and dried midribs are also used as thatching for houses and fences, as well as for mulch. It should be noted that wild enset plants are also commonly harvested for these purposes in western regions where both wild and domestic enset co-occur.

| Nutritional content
Considering the importance of enset as a dietary staple food, compared to other regional crops only a relatively small number of nutritional analyses have been undertaken on raw enset plant tissues (Fanta & Satheesh, 2019;Fekadu & Ledin, 1997;Mohammed et al., 2013;, covering a small selection of varieties (max n = 10), with a low number of replications (max n = 3), summarized in Table 2. Overall, the calorie content of kocho per 100 g of edible material is reported 200 kcal or 57% lower than the corresponding value for 100 g of food grains (Urga, Fite, & Biratu, 1996). Other studies have investigated the nutritional value of the resulting enset food products (Atlabachew & Chandravanshi, 2008;Forsido, Rupasinghe, & Astatkie, 2013) or as a component of animal feeds (Afele, 2014;Nurfeta et al., 2009;Talore, 2015).
Only one recent paper by Bosha et al. (2016) included an assessment of the nutritional content of wild enset individuals in comparison with domestic varieties. They analyzed kocho produced from three wild and three cultivated varieties after fermentation for 30 or 90 days. In general, the cultivated plants scored better than the wild ones for protein, fat, sugar, and mineral content, but the wild plants contained greater starch. It is important to note that farmers partly identify different enset landraces based on the qualities of the kocho, bulla, and amicho they produce (Tsegaye, 2002).
In a dietary survey of 39 households containing 237 people, Pijls et al. (1995) showed that the average daily intake of 0.55 kg of enset products provided 68% of the total energy intake, 20% of the protein and 28% of the iron, but no detectable level of Vitamin A. The low protein content is often addressed by adding alternative protein sources to the diet. For example, kidney beans or cabbage are often eaten with enset products (Abebe, Stoecker, Hinds, & Gates, 2006). Indeed the enset farming system accommodates many food plants including legumes, vegetables, and fruits in addition to animal products and the culture of adding these in food preparations that can effectively supplement the nutritional balance of the enset food system.
Enset is reported to provide important dietary micronutrients although sample sizes are low, and results highly variable. Abebe

et al. (2007) in a study of subsistence farming in households in
Sidama in southern Ethiopia, found that cereals (mainly unrefined maize) contributed the major source of energy, protein, iron and zinc, whereas starchy foods prepared from enset were the primary source of calcium and an additional source of iron. Atlabachew and

Analysis Unit
Pseudostem Corm Note: Raw data collated from (Fekadu & Ledin, 1997;Mohammed et al., 2013;Nurfeta et al., 2009;Talore, 2015;Zewdie, Olsson, & Fetene, 2008). Chandravanshi (2008) found that the potassium concentration in kocho and bulla was the highest, followed by sodium, calcium, and magnesium. In general, kocho contained more minerals than bulla, but both are rich in calcium and zinc compared to other locally available starchy foodstuffs and contain comparable concentration of copper, iron and manganese. Mohammed et al. (2013) reported that the corm contains 17 of 20 amino acids and has similar or higher concentrations than potato for 12 of them.

| Enset traditional medicine
Enset is widely considered an important medicinal plant in Ethiopia, with remarkable consistency across diverse ethnic groups in its reported uses, and the phenotypic traits that characterize medicinal values. Primarily, but not always, plants used medicinally have red or reddish-purple leaf blades with midribs and pseudostems that are red to a varying degree (Alemu & Sandford, 1991;Assefa & Fitamo, 2016;Tsehaye & Kebebew, 2006).
There are two principal medicinal properties commonly attributed to enset. First, boiled corm (amicho) of various varieties, often consumed with milk, allegedly cures fractured or broken bones (Tsehaye & Kebebew, 2006). We hypothesize that this could be attributed to high calcium content of certain varieties speeding recovery. Second, the amicho of several varieties are consumed with milk and butter to facilitate placental discharge in both humans and livestock (Assefa & Fitamo, 2016;Tsehaye & Kebebew, 2006). In some cases, it may also induce abortion (Tsegaye, 2002).
In addition, Pijls et al. (1995) reported the use of enset root to treat nematode worms. This is supported by a report from Hölscher and Schneider (1998), who identified a novel phenylphenalenone in enset, as well as compounds already known from other Musaceae.
Phenylphenalenones may have anti-bacterial, anti-cancer, and nematicidic properties Hölscher and Schneider (1998). Tsegaye (2002) found that farmers in Sidama, Wolaita, and Hadiya used the corm or other parts of specific enset landraces to cure cirrhosis, diarrhoea or venereal diseases.

| Enset microbiome and fermentation
Traditionally, the scrapped parenchymatic tissue of the pseudostem and pulverized corm of enset are buried in an earthen pit wrapped by enset leaves and subjected to fermentation, to increase the organoleptic property and shelf life. The fermentation procedure reduces toxicity of plant raw materials, while contributing to flavor (Urga et al., 1996). Some loss of protein and dry matter is associated with the fermentation process (Besrat, Mehansho, & Bezuneh, 1979;Tsegaye, 2002), possibly due to the permeability and long duration of storage for fermentation (2-3 months) in the pit, which allows leaching of water-soluble proteins and amino acids (Tsegaye, 2002). Before fermentation, kocho has high moisture content, neutral pH and several microorganisms, including aerobic and anaerobic spore formers, lactic acid bacteria, and others in the Ethiopia, while there is substantial variation in processing techniques and potentially yeast cultures across the region (Gashe, 1987;Urga et al., 1996). More recently Birmeta, Bakeeva, and Passoth (2018) isolated cultural microbes across different stages of fermentation identifying 12 yeast and 17 bacteria species using rDNA sequencing, showing that microbial composition changes through the fermentation process.
There is significant potential to improve the efficiency and effectiveness of enset processing, which is currently a labor-intensive process. The Shakacho people, for example, add Mandillo (Crassocephalum macropappum) stem during enset fermentation, which appears to result in a pH lower than natural fermentation. As a result Mandillo is thought to better reduce spoilage and increase shelf life (Gonfa, 2016). Ashenafi (2008) and Tafere (2015), in general reviews of fermented products in Ethiopia, highlighted the value of controlled fermentation studies with selected cultures and starters to optimize the processes involved. Both Hunduma and Ashenafi (2011) and Gizaw et al. (2016) recommended that this could reduce labor for women, help to avoid spoilage during fermentation, improve the long-term storage and product quality, and reduce waste and increase food security.

| Cultural importance
Enset also plays an important cultural role for several Ethiopian ethnic groups who have traditionally cultivated Enset (Assefa & Fitamo, 2016;Negash & Niehof, 2004;Olango, Tesfaye, Catellani, & Pè, 2014;Shank, 1994). Tsehaye and Kebebew (2006) found that farmers in Kaffa Zone (SNNPR) are aware of, and cultivate, a number of different varieties specifically in relation to myths, poems and beliefs about their medicinal and ritual significance. Assefa and Fitamo (2016) similarly report songs and rituals associated with specific aspects of enset in the Sidama culture. The Gurage people, to whom enset is an important staple, label themselves the "people of enset" (Shank & Ertiro, 1996). Sahle et al. (2018) described the role of enset in Gurage economic and social life, which involves the extensive and well-planned cultivation and storage of enset alongside other crops such as coffee and khat in a mixed horticulture.
Within the family, women have the major role in processing and cooking of foods from enset and it is often described as a "women crop" on account of women's significant role in the processing, cooking, and selling of enset products (MacEntee, Thompson, Forsido, & Jihad, 2013).

| CON CLUS I ON S AND A FR AME WORK FOR SUS TAINAB LE DE VELOPMENT
Due to the complexity of enset propagation, management, harvesting, and processing, enset agricultural practice is currently inseparable from the ethnobotanical knowledge housed by numerous Ethiopian ethnic groups for which enset is a starch staple. Therefore it is unsurprising that compared to other crops, enset has not transitioned at any scale to industrial production.
Only 99 tons were reported from commercial farms in the period 2011-2015, and we consider these data and the existence of commercial farms questionable (Central Statistics Agency, n.d.).
A combination of complex ethnobotanic knowledge, the relative isolation of highland agroecologies and long-term cultural boundaries is likely to have impeded the spread of enset historically.
Indeed from field observations by the authors, the distribution of enset agriculture appears, in many areas, to more closely match the distribution of specific cultural groups, rather than environmental conditions-although empirical evaluation of enset distribution remains to be performed. Future sustainable development of enset will depend on appropriate and equitable documentation and exploitation of this extensive knowledge. For example, in the context of increasing disease incidence in the Musaceae and the impact of climate change on East African agriculture, there are considerable opportunities in recording farmers perceptions of disease resistant, drought tolerant and early maturing landraces to guide breeding and genomic analyses (Borrell et al., 2018).
There is also likely to be considerable undocumented ethnobotanic knowledge on cultivated enset that is in danger of being lost. For this reason, we outline in Box 2, a framework within which to report data from enset ethnobotanic studies, to helps ensure that comprehensive quantitative information is gathered and documented to facilitate use and interpretation by other researchers. In the longer term, shifting the range of current agricultural systems such as enset-chat, enset-cereal-vegetable, and enset-tuber, or development of novel crop combinations, rather than enset-coffee, may be important for adaptation to changing climate. This, combined with genetic surveys, modern crop breeding, improved processing and selection of appropriate landraces is likely to position enset as an important climate-smart starch staple for Ethiopia and beyond. • Communication with other researchers, particularly those that may be working in the area or wider region is crucial to avoid duplicated work effort and survey fatigue-particularly among communities close to agricultural universities, field stations or readily accessible on major roads.

Box 2 A framework for ethnobotanic surveys of enset across diverse communities and agrisystems in Ethiopia
• Hypothesis generation and study design should be clearly defined and aim to minimise confounding variables. For example, analyses are frequently reported that compare farms among Ethiopian agro-ecological zones, however if differences are found it is difficult to conclude whether the driver is, for example, climatic, edaphic or socio-cultural.
• Location (GPS) should be reported in addition to local village, kebele and woreda names. Grid references can be reported at ~1 km resolution with data aggregated within study sites to protect farmer anonymity.
• The method used to select farmer respondents is important to reduce risk of bias. Many studies use model farmer or key informant interviews, which are appropriate for some questions, whilst comparatively few have used random surveys.
• Key data to ensure usability of survey data by other researchers include; location, responded selection method, landrace names, the local language, the social or cultural group and if appropriate, sample sizes and processing method.
• Additional metadata should be collected and reported to maximize the utility of data for other researchers, such as; alternative uses, medicinal applications, disease or frost tolerance, pest and pathogen incidence, identifying morphological features, co-cultivated crops, harvesting cycle, processing methods or other indigenous knowledge reported to the investigator.
• Raw data should be made readily available and accessible, for example in theses, academic paper appendices, supplementary materials or online repositories. Open data and research is critical to accelerate the food security potential of enset.