Introduction

Sorghum (Sorghum bicolor (L.) Moench) is the fifth most important cereal in terms of production in the world after maize, wheat, rice, and barley (FAOSTAT 2017). It is an important food security crop for millions of people in developing countries (Gerrano et al. 2014). Sorghum is one of the stable foods in many parts of Ethiopia, including Tigray, and is mainly used for enjera (Ethiopian pancake-like flat bread) and Swa (local beverage) (Teshome et al. 2007; Tsehaye et al. 2009). Ethiopia has been recognized as a Vavilovian center and/or origin of diversity for several crops, including sorghum (Vavilov 1951). The availability of several wild and weedy types of sorghum in several parts of the country including Tigray may support this hypothesis. It is one of the most diverse crops grown in Ethiopia and it can grow from an altitude of 400 up to 3000 masl by almost all the Ethiopian ethnic nationalities (Engels and Hawkes 1991; Tesso et al. 2011).

In the literature, several areas are hypothesized as the area for commencement of Sorghum domestication, including areas in present day Ethiopia (Doggett 1991); eastern Sahelian zone between Lake Chad and northwest Ethiopia (Harlan and Stemler 1976); and areas of northeastern Africa (Dillon et al. 2007; Patil 2016). Archeological evidence also indicated that far eastern Sahel is the potential candidate for sorghum domestication (Beldados et al. 2018; Fuller and Stevens 2018; Winchell et al. 2018; Venkateswaran et al. 2019). Thus, the exact place of sorghum domestication is still discussed, but it can be noted from these suggestions that sorghum may have multiple sites of domestication.

Based on their panicle morphology, sorghums are categorized into five major races (bicolor, durra, caudatum, guinea, and kafir) and ten intermediate races of the major races (Harlan and De Wet 1972). Although this classification is widely accepted, the incidence of some divisions within the races grown in different geographical areas showed different clustering pattern. For instance, West African Kaura (caudatum type) was grouped with guinea (Brown et al. 2011); Indian guinea type was grouped with durra sorghum (Morris et al. 2013); and the eastern Asian races of durra, caudatum, and bicolor were found to be close to each other (Billot et al. 2013). Ethiopia is recognized as the best relics of the development of durra sorghum and south-western Ethiopia, where the Konso people practicing ancient agriculture, provides the type of sites that sorghum (mainly guinea) could have been ennobled (Doggett 1988).

Phenotypic markers-based characterization of crop genetic resources is important for their conservation and utilization. Phenotypic markers have plentiful contribution in developing countries, where the availability of molecular resources is limited, though they are less accurate because they are limited in number and influenced by environmental factors (Adugna 2014; Nadeem et al. 2018). These markers have been applied for measuring the genetic diversity of Ethiopian sorghum revealing high diversity (Ayana and Bekelle 2000; Abdi et al. 2002; Geleta and Labuschagne 2005; Tsehaye et al. 2009; Adugna 2014). The existence of great sorghum genetic diversity in Ethiopia was also supported by molecular markers (Cuevas and Prom 2013; Adugna 2014; Cuevas et al. 2017; Girma et al. 2019).

Although Tigray is one of the important sorghums growing areas in Ethiopia, there is lack of references providing region-wide and detailed inventory of sorghum landraces and their diversity assessment. Little is also known on the race types and their distribution in the region. Besides, prediction and identification of suitable sites for sorghum production using future climate scenario is not yet done. To the best of our knowledge, this is the first intensive region wide collection and characterization work of sorghum landraces in the region. The current study is part of a continuous works on phenotypic, genomic, and nutritional diversity analysis of sorghum landraces in Tigray. Therefore, the objectives of this paper are (1) to assess the level of sorghum landrace richness and abundance; (2) to estimate the extent and patterns of phenotypic diversity with respect to zone of collection, altitude, and race; (3) to predict and identify suitable areas of sorghum cultivation using future climate data; and (4) to identify high diversity repository sites for in situ and/or ex situ conservation and utilization.

Materials and methods

Plant materials

In this study, a total of 358 sorghum landraces were assessed. All the sorghum growing areas of Tigray were surveyed, and panicles of these landraces were collected from 20 sorghum growing districts (‘Weredas’) that belongs to four zones of the Tigray region, northern Ethiopia (Fig. 1). The panicles of each sample were collected at maturity from 125 independent farmers’ field during the growing seasons of 2016/17 and 2017/18. Farmers were involved in tracing the vernacular names (local names) of the sorghum landrace collections and describing their characters including end use qualities. Through farmers’ participation and support, farms hosting unique landrace were surveyed and from each farm about three similar panicles having the same vernacular name were collected. In some farms two or more differently named varieties (in mixed state) were encountered and sampled accordingly. Apparently, a total of 358 sorghum landraces were sampled from the four zones: southern (105 samples), southeast (51 samples), central (122 samples), and northwest (80 samples).

Fig. 1
figure 1

Map of the Tigray region (study site) and Ethiopia (upper left). The color highlighted maps are Woredas of Tigray from where sorghum landraces were collected

Full size image

Morphological characterization and evaluation

Before characterizations, the landraces were classified according to their zone of origin/collection and altitude to understand the zonal and altitudinal patterns of diversity. It is important to note here that the names of zones used in this study may not be the same with the administrative demarcation of zones used by the government. This is because some sorghum landraces were grouped beyond their administratively demarcated zone of the region. Following the altitude-based classification described by Ayana and Bekele (1998), the sorghum collections were classified into lowland (< = 1600), intermediate (1601–1900), and highland (> 1901) categories. Consequently, 165, 69, and 124 sorghum landraces were grouped into the lowland, intermediate, and highland category, respectively. The climatic descriptions of the study districts mainly annual temperature, annual precipitation and altitude is respectively, 18.49–23.29 °C, 598–820 mm, and 614–2551 masl. The dominantly cultivated crops in the southern zone are sorghum, wheat, barley, and beans while in the southeast are wheat, barley. In the central Tigray the widely grown crops are sorghum wheat, barley, maize, and teff whereas in the western parts are sorghum, finger millet, maize, and sesame.

Morphological characterizations and scoring of all sampled panicles were performed as per the sorghum descriptor list developed by the International Board for Plant Genetic Resources (IBPGR 1993) and farmers’ notion with some modifications. A total of 17 morphological descriptors (characters) were used for the characterizations of the samples. Traits that are important in the folk taxonomy and farmers’ varietal identification and selection schemes were also included in the list of descriptors. The code and descriptions of the morphological characters are shown in Table 1. Visual observation, color chart, and hand lens with 10X magnification were used for the characterization works. The sorghum collections were also subjected to racial classification following the botanical classification of Harlan and de Wet (1972) that recognizes five basic races and ten intermediate races.

Table 1 Sorghum reproductive morphological characters used in the study, their codes and descriptions
Full size table

Statistical data analysis

Landrace names were considered as a unit of diversity for estimating varietal richness and thus, landrace richness was estimated based on the count data of the named landraces aggregated on zonal basis. Two common species richness measures, Margalef’s index (DMg) (Eq. 1) and Menhinick’s index (DMn) (Eq. 2) (Magurran 1988) were used for measuring landrace richness (Tsehaye et al. 2006), as follows:

$$D_{Mg} = {\raise0.7ex\hbox{${(S - 1)}$} \!\mathord{\left/ {\vphantom {{(S - 1)} {\ln \;N}}}\right.\kern-0pt} \!\lower0.7ex\hbox{${\ln \;N}$}}$$
(1)
$$D_{Mn} = {S \mathord{\left/ {\vphantom {S {\sqrt N }}} \right. \kern-0pt} {\sqrt N }}$$
(2)

where S is number of landraces, N is total number of individuals sampled.

To understand the extent and patterns of diversity, frequency of characters scored from each panicle for each landrace were summarized (pooled) in the zonal and altitudinal class frameworks. Thus, phenotypic diversity was estimated using frequency as an input using the software IBM-SPSS version 23. Shannon–Weaver diversity index (H′) (Eq. 3) was estimated on the frequency data as described by Hutchenson (1970) as follows.

$$H^{\prime } = - \sum\limits_{i = 1}^{n} {\frac{pi\;\ln \;pi}{{\ln \;n}}}$$
(3)

where pi is the proportion of the total number of individuals in the ith class and n is the number of phenotypic classes for a given character.

Suitability analysis for the cultivation of sorghum under a novel climate change state was predicted using the future climate scenario. The prediction was made using the ecological niche model that was performed by domain binary method in DIVA-GIS version 7.5 using future annual average temperature and precipitation as described in Hijmans et al. (2012). Accordingly, possible areas for future cultivation were identified and mapped.

Results

Sorghum landrace richness and abundance

Taking local name as a diversity unit and reference (excluding double counting), 140 distinctly named landraces were identified and collected. However, different morphotypes with similar vernacular names were also considered for further analysis. The list of the locally named landraces with their character states and races is illustrated in supporting Table 1.

The collected landraces belong to three major cultivated sorghum races (durra, bicolor and caudatum), and one intermediate race (durra-bicolor) as well as a weedy race (wild type). The race durra was the most abundant race with the highest spatial coverage, stretching from the southern to the western part of Tigray. In this collecting mission, sweet-stalk sorghum (‘Tinkish’) and the wild relative of sorghum (‘Zerie-Sheytan’ [the Devil’s sorghum]) that belongs to the Sorghum bicolor subspecies verticilliflorum were found. As shown in Fig. 2, a morphologically diverse panicles of sorghum landraces were identified.

Fig. 2
figure 2

Picture of some of the morphological diversity in sorghum in the study

Full size image

In the south zone of Tigray, in the district of Raya Alamata, a maximum of 22 locally named landraces were encountered and collected from a single farmer’s plot. Almost all the sorghum fields were devoted to landraces (testifying to yet limited success of the formal breeding scheme and low adoption of improved varieties) and the race durra was the dominant race, accounting for 95% of the collections. Nevertheless, apart from the race durra, other races were less frequent. Apparently, 1% each of the races bicolor and caudatum, and 3% of the intermediate race durra-bicolor were found. In the lowland part of this zone, ‘Kodem’, ‘Dengele’, ‘Degalit’ and ‘Amarica’ (the latter [Amarica] was probably introduced via the formal seed system long time ago and now considered as adapted local landrace) were the most common (abundant) landraces. These landraces have wider spatial distribution. Other landraces such as ‘Aba’are’ and ‘Gano’ were cultivated in large farm plots (relatively large farm area) but had specific agro-ecological niches (limited spatial expansion). The latter two landraces (Aba’are and Gano) are characterized as having very compact panicle, yellow seed color, goosed peduncle and late maturity attributes.

Sweet-stalk sorghum (‘Tinkish’) often planted within the sorghum fields (at low frequency) were also found in only the southern zone of Tigray, particularly in the district of Raya-Alamata, and apparently five sweet stalk sorghum landraces were identified by farmers. A landrace locally known as ‘Zengada’ was the only and widely cultivated sorghum landrace dominantly grown in the higher altitude (> 2400 masl) of southern Tigray mainly in the locality near to the lake Hashenge (a highland lake) that attributed to its culinary value and stress tolerance like frost (see Fig. 3). All farms in this area often covered with Zengada year after year, with some crop rotation schemes particularly with faba bean and maize. Zengada has red seed color and intermediate panicle compactness and belong to the intermediate race durra-bicolor.

Fig. 3
figure 3

Zengada, the only landrace cultivated in the higher elevation areas of southern Tigray

Full size image

In the southeast zone, in some farms only single, while in others more than two co-existing landraces were encountered. In this zone, in a single event, about seven sorghum landraces were collected from an individual farmer’s plot. The preponderance of landraces having intermediate panicle compactness and spreading rachis branching were quite high. The yellow seeded landrace, Kodem (in some localities within this zone called with different dialect “Koden’), was the most cited and dominant landrace in this zone. In terms of race classification, three sorghum races (durra, bicolor and durra-bicolor) were predominant, where the race durra was the most frequent that accounted for 83% of the collections. Sorghum landraces that belong to the intermediate race, durra-bicolor, were also relatively abundant (14%) as compared to race bicolor that accounts for only 3%.

In the central Tigray, relatively large geographic areas were visited. The number of landraces collected per farm ranged from one to eight. The sorghum landraces were collected from relatively wider elevation arrays ranging from 1369 to 2211 masl. The three most abundant local varieties (landraces) of this zone were ‘Lequa’, ‘Shulkit’ and ‘Gumbil’. The former was a typical bicolor sorghum that is characterized by its open and/or dropping inflorescences and long clasping glumes that are at least three-fourths as long as the grain. However, other variants of Lequa, having compact panicles and short glumes and that belong to race durra and durra-bicolor were identified. Two more variants of Lequa were also identified by farmers and named as ‘Lequa Hatsiro’ (to mean short stature Lequa) and ‘Lequa Mshela Barya’ (meaning sorghum of a slave). Lequa Mshela Barya may be eaten by poor farmers during drought period and crop failure. The landrace Gumbil is a typical durra race with a compact panicle and appressed rachis branch position. According to the farmers account, Gumbil is very popular for its yield and enjera quality. Generally, in the central zone, race durra was the most abundant (71%), followed by durra-bicolor (16%), bicolor (13%), caudatum (3%) and wild type (3%). The latter was concentrated in the intermediate altitude category and identified by farmers as Devil’s sorghum (‘Zerie-Sheytan’).

In the northwest zone of Tigray, agro-ecologies that range between 614 to 1814 masl were covered. In this zone, five races including the wild type were identified in different frequencies where race durra accounted for 65% followed by race caudatum (21%), bicolor (6%), durra-bicolor (5%) and wild type (3%). The economically important race, caudatum was more concentrated in this zone and collected in the altitude range of 614–1500 masl. The local variety, ‘Merewey’ was the most abundant landrace in this zone. Landraces with chalky white seed colors were dominant in this zone.

Based on the alpha diversity (measured using landrace count as a unit) estimate, southern zone of Tigray had the highest value score (DMg = 10.74, DMn = 5.00) followed by central (DMg = 8.54, DMn = 3.80). The level of landrace richness in northwest Tigray (DMg = 6.16, DMn = 3.13) was relatively higher than the southeast part (DMg = 4.58, DMn = 2.66).

Zonal distribution of characters

Zone level percentage frequencies of all the sorghum morphological characters are summarized in Table 2. Marked differences in proportion and zonal distribution were found for the considered morphological traits. Higher percentage frequency (79%) of the compact type of panicle occurred at southern zone followed by northwest (58%). High number of rachis branch number per whorl was concentrated in southern zone. Of the 358 samples, 294 (82%) had an erect type peduncle position. In the south zone, most of the samples have recurved and inclined peduncle position with a frequency of 43% and 41%, respectively. The recurved type of sorghum has a sickle like structure which is a typical character of the sorghum landrace, locally known as ‘Aba’are’ (Fig. 4). The curved panicle sorghum landrace types (typical character of the race durra) were more frequent in the southern part of Tigray and that is may be an adaptation to lowland high temperature areas particularly the Mekhoni area and its environs.

Table 2 Percentage frequency for each morphological character at each zone (S = South, SE = Southeast, C = Central, NW = Northwest) and region wide (R)
Full size table
Fig. 4
figure 4

Aba’are, a recurved type of sorghum in southern zone of Tigray

Full size image

Starchy endosperm was the most frequent variant that reaches 52% followed by intermediate (40%), while only 3% of the landraces had corneous endosperm. Among the seed color character states, yellow and chalky white were found with maximum (34%) and minimum (4%) frequency, respectively. In the lowland areas of south Tigray, yellow seeded landraces were largely concentrated in different panicle forms, stature and maturity (Table 2). In the other hand, the landraces with chalky white seed color such as Wedi-Aker, Arfia’agdm, and Deber were concentrated in the northwest zone of Tigray.

Majority of the samples had transverse depression in their glumes although the adaptive significance of this attribute is not known. Most of the samples (about 75%) had hairy glumes. However, considerable amounts of sorghum landraces collected from the northwest zone lacks glume hair and were glabrous. Four phenotypic states of glume color (white, red, brown, and black) were identified and landraces with white glume color were the most abundant (54%) followed by brown (33%). Red glume color was a rare variant that had a very limited distribution (2%) in the region. Most of the studied landraces in the region (290 out of 358 samples) had visible seeds (open seeds with limited glume cover, less than 25% coverage), while 6% of the samples had covered seeds (Table 2).

The single seeded sorghum was the dominant attribute and more frequent variant in the region. Nevertheless, twin seeded sorghum (worldwide rare phenotype of sorghum) was observed with a frequency of 18%. The twin seeded sorghum was found in some pocket sites of the region, particularly around the areas of Sheraro (northwest Tigray) where the ethnic Kunama resides. This variant was predominantly cultivated by the Kunama community, a minor community among the three ethnic groups (Tegaru, Irob [Saho], and Kunama) living in Tigray. Three of the five major races (durra, bicolor, and caudatum) were detected with variable proportions, while from the intermediate races only durra-bicolor was found. Majority of the landraces (79%) collected and examined in this study were identified as durra sorghum.

Altitudinal distribution of characters

Altitudinal variations for all the evaluated morphological characters of sorghum are illustrated in Table 3. The distribution of most of the investigated sorghum morphological character states did not show a clinal variation with altitude. Higher frequency of bicolor and durra-bicolor was observed in highland than lowlands, while race caudatum was dominant in the lowland. Relatively higher frequency of race durra occurred in the lowland elevation than highland. The frequency of race durra and caudatum had shown a decreasing pattern with increasing altitude, while bicolor and durra-bicolor increased with increase in elevation.

Table 3 Percentage frequency of each morphological characters for each altitude class.
Full size table

Erect type peduncle growth habit was concentrated (96%) at higher altitude, while recurved and inclined peduncle positions had widespread occurrence in the lowland. Brown seed color and presence of glume hair were more frequent attributes in the highland. White, chalky white and red seed colors were more frequent in the lowland than highland. The frequency of yellow seed color was higher in the intermediate elevation, while almost similar in the lowland and highland altitudes. White glume color was the most frequent in intermediate altitude class and higher in highland than lowland, while red glume color was found in the same proportion (2%) in all altitude classes. Starchy endosperm was dominantly concentrated equally in the lowland and intermediate altitude ranges. Landraces with open seed (< 25% seed covering) were concentrated in intermediate altitude ranges than in the lowland or highland. The landraces having covered grains were highest in the highland.

Estimates of diversity

The estimates of diversity, as estimated by Shannon–Weaver diversity index (H′), was done for individual characters and pooled across zone of collection and altitude. The value of H′ for individual characters varied from 0.24 (seed form) to 0.95 (rachis branch length). Panicle compactness, seed color, rachis branch position, and rachis branch number are among the characters with higher diversity (H′), while traits such as seed spot, seed covering, and peduncle position scored relatively lower H′. The pooled mean diversity estimate for all traits was 0.70 (Table 4). Maximum H′ (0.71 ± 0.06) was recorded in northwest Tigray followed by central (H′ = 0.64 ± 0.08) (Table 4). The southern (H′ = 0.51 ± 0.08) and southeast (H′ = 0.52 ± 0.08) have relatively lower diversity index as compared to other zones of the region. A diversity index of 0.70, 0.68, and 0.61 were recorded in the lowland, intermediate and highland elevation classes showing decreasing trend from lowland to highland elevation (Table 5).

Table 4 Estimates of Shannon–Weaver diversity index (H′) for sorghum morphological characters for each zone (S = South, SE = Southeast, C = Central, NW = Northwest) and Region wide (R)
Full size table
Table 5 Altitudinal (L = Lowland, I = Intermediate, and H = Highland) estimates of Shannon–Weaver diversity index (H′) for sorghum morphological characters
Full size table

Ecological niche modeling

Future prevalence and potential distribution of sorghum (about 50 years from the current scenario) and future climate suitability under the changing climate was predicted using the DIVA-GIS. Future regional distribution of sorghum races and identification of suitable sites for sorghum cultivation were predicted based on the future climatic data, mainly annual precipitation and temperature provided in the DIVA-GIS (Fig. 5). Many sites in the region are found to be suitable for sorghum cultivation in the future and several areas that are not covered in the current study too are found to be appropriate for sorghum cultivation and production, even at small scale farmers’ condition. The racial distribution analysis in the future climate scenario showed that durra race of sorghum will have a wider array of abundance and will be dominantly distributed in the region.

Fig. 5
figure 5

Future prediction of sorghum cultivation and racial distribution in the Tigray region using future climate scenario. Green color shows suitable areas of cultivation for sorghum in the future. Colored symbols indicate the sorghum races (Red = Durra; Yellow = Caudatum; Green = Durra-bicolor; and Blue = bicolor)

Full size image

Discussion

Landrace distribution and abundance

In the current study, most of the important sorghum producing areas of the Tigray region were surveyed and the existing landraces were inventoried and documented. The distribution and abundance of the sorghum landraces appeared to vary within and between localities and altitudinal ranges.

Of the total 140 locally named landraces, 51 were identified in southern zone, 42 in central, 28 in northwest, and 19 in southeast. This figure was relatively higher than previously reported from northeastern (Seboka and van Hintum 2006; Teshome et al. 2007) and Eastern Ethiopia (Mekbib 2007). This may indicate that the informal seed system that accounts for 80% of the seed procurement in Ethiopia (Bishaw and Turner 2008) remains as the most valuable option as a seed source and seed/grain markets, particularly in Tigray. The farmers’ dependency on the local varieties of sorghum from the informal system rather than the formal system, may be viewed as caused by the lack of availability of new varieties with superior agronomic adaptation as well as quality attributes desired by farmers and varieties that fit to the existing complex cropping system. Further, the provision of limited attention for the development of varieties that suits to specific agro-ecological niches and farmers concern may be one of the chief reasons for reliance on locally available landraces and the limited success in the adoption of improved varieties.

In the study districts/zones, some landraces are spatially abundant (e.g., Kodem) that covering large geographic areas and others are rare and that having specific niches (e.g., Gano, Aba’are). The dominance of some landraces could be associated with their end use attributes, adaptation to the existing agro-ecologies and farming systems, the existence of high proportion of locally common alleles of adaptive significance (Tsehaye et al. 2009) and market demand (as grain and seed) linked to the locally adapted landraces in the rural settings. Besides, the active traditional seed network and exchange through different mechanisms is also instrumental for the observed high level of landrace richness. From farmers’ account of the study area, producing and retaining own seed on-farm or acquiring seed via the local exchange or purchasing seeds of similar variety from the existing pockets of small local markets is the most viable economic approach. Moreover, cultural settings, landscape heterogeneity as well as climate variability are also some of the important factors that influence inter- and intra-specific diversity on-farm (Teshome et al. 2007).

In the lowland areas of the southern zone of Tigray, five landraces (Dengele, Degalit, Kodem, Aba’are and Gano) that are characterized as yellow seeded with very compact panicle and all belong to the race durra, were most frequently encountered and cited. The first three landraces (Dengele, Degalit, Kodem) have wider distribution and adaptation across the lowland part of the southern zone (Raya valley). Unlike the long-standing notion that stated that landraces have a narrow geographic adaptation and niche (Zeven 1996) these landraces have relatively wider spatial expansion that grow in large geographic areas. The high yielding and early maturing landrace ‘Kodem’ grows well in areas of poor fertility status and low moisture and has relatively wider adaptability that stretched from the lowland areas of southern to the southeast parts of Tigray. The distribution of Kodem was also reported to extend up to the central part of Tigray, with closely related dialect ’Koden’. From this we can understand that as long as the growing environment (climate and soil properties) are conducive, landraces can do well and move to distant geographic areas via the traditional seed system and network. For example, in North-Eastern Ethiopia, particularly south Welo (neighboring Amhara/Oromia region), Teshome (1996) and Seboka and van Hintum (2006) listed some landraces (like Zengada, Degalit and Aba’are), although in low frequency, that have exactly similar vernacular names and are widely cultivated in south Tigray. The similarity in vernacular names and likely morphology, from the two linguistically different farming communities (Amhara/Oromo versus Tigray) may show practice of exchanging seed lots (specifically exchange of unthreshed panicles often a common practice in sorghum) and subsequent adaptation across large spatial scales (Almekinders et al. 1994; Tsehaye et al. 2006). Nonetheless, deeper assessment of the direction of the seed flow, the structure of the genetic diversity within and between the landraces as well as the ethnobotanical knowledge associated with them at the respective sites is important for future on-farm conservation efforts and utilization of the landraces (Tsehaye et al. 2009).

Some landraces (e.g., Aba’are and Gano) have limited spatial expansion and concentrated in a defined eco-geographic area within the southern part of Tigray. These two sorghum landraces, which are long maturing and well preferred for making enjera, are grown by farmers in relatively large farms but they are adapted to specific agro-ecologies within this area. These two landraces are adapted to waterlogged soils and naturally fertile lowland areas (i.e., mountain bottoms having micronutrients rich soils accumulated from the mountain hills). Both Aba’are and Gano have a tapering panicle form. Farmers in the southern part of Tigray have a collective generic name for such type of varieties and identify them as ‘Afincho’ (a local word to mean nose-like), to indicate the nose like structure of the panicle tip. This attribute is very important in the folk taxonomy and classification of sorghum in the farmer’s criteria.

Elevation is one of the criteria in the folk taxonomy used to classify and identify farmers’ varieties (within a crop species) with respect to their ecological niches (e.g., lowland and highland sorghum) (Brhane 1981). Some sorghum landraces are specifically adapted to the highland environments, where pulses like faba bean are dominantly grown. In the highland (> 2400 masl) area of the southern zone of Tigray (cool and wet most of the year particularly during the summer), only the sorghum landrace named as Zengada is growing in almost all farms (planted on contiguous farms) allotted to sorghum. This variety is characterized by having a compact panicle and dark brown seed color. Despite the competition for space with bread wheat, barley and highland pulses (such as faba bean and field pea), Zengada still persist and farmers are still willing to cultivate it in this area. Farmers’ explanations indicate that the continued cultivation of such a single variety (Zengada) in the highland area is attributed to its special characteristics, mainly specific adaptation to frost (common stress) and other quality attributes (that may not fulfill by other highland cereal crops such as wheat and barley). The preponderance of Zengada to the highland agro-ecology may also be associated with its adaptation to cooler environment (Harlan 1975) and phenotypic plasticity (Teshome et al. 1999). New varieties of sorghum with superior agronomic adaptation to such highland areas as well as quality attributes desired by farmers were not available in the formal system and on the market. Farmers further noted that producing and retaining own seed on-farm was the most viable economic approach particularly for Zengada. In the highland areas of the southern part of Tigray, thus, Zengada (intra-specifically buffered local variety) will continue to be the only viable option in the near future unless the formal breeding scheme supports the farmer’s effort by enhancing more options.

In the lowland to mid-highland areas of the central Tigray, ‘Lequa’ (brown seeded, that resembles the wild and weedy relatives of sorghum), Gumbil (yellow seeded) and Shulkit (white seeded) are the most frequent and abundant sorghum landraces. Lequa is very popular for making local beer (‘Swa’) and unleavened bread (‘Kicha’). Lequa has spreading to dropping panicle forms and the seeds are completely covered by glumes. Most of the collected samples goes with this name (Lequa) belong to the race bicolor. However, in a specific lowland area of the central Tigray (Werei’-lekhe districts of the study area), some lequa samples that belong to the race durra (having short glumes, big and open seeds), an attribute different from what was seen in the common bicolor race Lequa, were also identified. As the cultivated and the wild relatives of sorghum races coexist in most sorghum growing fields in Ethiopia (Ayana et al. 2001; Tesso et al. 2008) including Tigray, the occurrence of mixed morphotypes (lequa of mixed races, bicolor and durra types) observed in the current expedition may show possible gene flow and introgression in either direction and the likely hybridization and cross fertilization scenarios between the two forms. Adugna et al. (2012) also reported possible cross fertilization (under natural conditions) and introgression between the cultivated and wild sorghum forms via SSR markers analysis. Furthermore, planting of deliberately mixed wild and cultivated forms, to enhance diversity via gene flow (Worede and Mekbib 1993; Teshome et al. 1999) and to utilize both forms (the leaves and stalks) as a cattle feed and firewood (Seboka and van Hintum 2006) were reported.

In the northwest zone of Tigray, sorghum richness was quite high and comparable to the southern part. However, a landrace named as Merewey (yellow seeded durra) was the most dominant and abundant landrace that has a wide array of adaptation and planted to several farm plots each year. The peculiar characteristics of this variety is its yield potential and enjera quality. This landrace has relatively big and compact panicle as well as long stature and is often being used for thatching roofs in the rural areas and firewood making as well as cattle feed. Apart from Merewey, two other frequently cultivated sorghum landraces (Dagnew and Chemroy), were also identified in the northwest zone. Both landraces have a very narrow geographical niches, specifically adapted to the clay soils of the Sheraro area (northwest zone), where the ethnic Kunama community (a community with a language belonging to the Nilo-Saharan language family) resides. Chemroy is a twin-seeded variety (a rare trait with low frequency of occurrence) that was specifically sampled in farms cultivated by the Kunama community. The twin-seeded attribute has very low frequency of occurrence in Ethiopian sorghum collections and so far, only 6% (Abdi et al. 2002) from north Shewa collections and 0.7% (Ayana and Bekele 1998) from the Ethiopian gene bank collections were reported. This indicates that the twin-seeded attribute may not be well represented in the national seed repository. Apart from the agro-ecological adaptation, the specific abundance of the twin-seed landrace, Chemroy, in the Sheraro area may be linked with the culinary culture and farming system of the Kunama people. Strong association between different ethnolinguistic groups and genetic diversity have been reported in sorghum (Deu et al. 2008; Westengen et al. 2014) and maize (Perales et al. 2005). Westengen et al. (2014) also reported strong ties between sorghum genetic patterns and broader scale distribution of language families in Africa.

Regional distribution of characters

Sorghum landraces of starchy endosperm types were more frequent and abundant in the present collections. This was in line with the findings of Ayana and Bekele (1998) and Kebede (1991) in Ethiopian sorghum ex situ collections. From farmers stand point, they tend to maintain sorghum landraces having starchy endosperm because the starchy sorghum contains a floury region that have high milling quality which is preferable for making enjera (pancake-like Ethiopian flat bread). Conscious selection may be instrumental in the preponderance of the starchy endosperm sorghum landraces as starchy grain is very important attribute in making enjera (Ayana and Bekele 1998). Besides, high milling and nutritional quality may also be associated with the observed high frequency of the starchy endosperm (Abdi et al. 2002). Thus, endosperm texture is one of the important farmers’ selection criteria particularly linked with end use quality (gastronomic attributes), enjera quality and test.

Seed color is one of the main morphological characters used by local farmers to identify their varieties and often they associate it with end use attributes. A clear spatial pattern was observed in the distribution of different seed color variants. In the south and south-east part of Tigray, the yellow seeded durra sorghum varieties were more frequent, unlike the north-west, where the white seeded caudatum race varieties were highly concentrated. In the lowland (< 1000 masl) areas of the north-western part of Tigray, the short statured chalky white seeded local varieties that belong to the race caudatum were more frequent unlike in the highlands where brown seeded varieties were concentrated. The lack of certain polyphenolics in the white seeded and the presence of much of it in the brown seeded may explain for the concentration of the variants in the lowland and highland, respectively. Besides, brown seeded varieties particularly those that have high tannin content have an agronomic advantage over the low tannin varieties (mainly the white seeded types) that are resistant to grain molds and pre-harvest germination in humid highland areas (Asante 1995; Harris and Burns 1970). The spatial pattern of the sorghum landrace seed color variants may be attributed to niche-specific compartment of the sorghum landraces, not only agro-ecology per se but also cultural niches, as described by Tunstall et al. (2001) and Teshome et al. (2007). The cultural and end-use niches linked with some landraces may be the chief driver of farmer’s decision making in maintaining different bundles of landraces with certain common attributes.

In Tigray and elsewhere in the country, enjera is the common daily dish, apart from the use of tiny cereal tef for this purpose (particularly in the urban areas). In sorghum growing areas, sorghum varieties with different color attributes could also fit in depending on the test and preferences of the farming community for enjera. In the study region (Tigray), sorghum is mainly utilized for human consumption as enjera in which yellow and white seeded are ideal, whereas sorghum with red seed color is used for making ‘Swa’ and ‘areqi’ (unfermented local beer and strong distilled local liquor, respectively). Similarly, Awika and Rooney (2004) reported that red seeded sorghum is preferred for brewing of traditional beer. The differences in the end use of different seed colors of finger millet landraces, in which black seeded is used to prepare local drinks, while white seeded for making enjera, was also reported in the same study area (Tsehaye et al. 2006). Such cultural and end use niches associated with different colored sorghum variants could have conservation, breeding and market implications. The existence of association between kernel color and human consumption and cultural use value for different traditional purposes was also stated in Ethiopian barley (Asfaw 1988).

Glume hairiness is a heritable, genetically controlled character (Kebebew et al. 2001). Farmers in north Shewa and south Welo of Ethiopia use glume hairiness to differentiate sorghum landraces in the folk taxonomy (Teshome et al. 1997). In the current work, sorghum landraces with hairy glumes were relatively frequent in the higher altitudes. The prevalence of glume hair in the highland could contribute in lessening mold because glume could reduce mold in high rainfall and humid areas as it dries quickly (Dogget 1991; Desmae et al. 2016). Less colonization and fewer propagules of grain mold in varieties with hairy glume were reported (Mansuetus 1990). It has also been stated that this character is associated with resistant to kernel bunt in wheat, rye, and barley (Warham 1988) and powdery mildew (Negassa 1985).

In the Tigray region, naked seeded (< 25% seed coverage) sorghum landraces were the most abundant types (ranged from 66% in north-western to 99% in southern Tigray). The higher frequency of naked grained sorghum in the southern zone of Tigray implies the predominance of race durra. In the current collections, sorghum landraces which have covered seeds (> 75% seed coverage) were less frequent. The low frequency of the 75% seed coverage found in this study disagrees with Ayana and Bekele (1998) who reported higher frequency of covered seeds (75% glume coverage) on sorghum landraces sampled from Tigray although their samples were small and that covers limited geographic areas. The low frequency and low abundance of the covered sorghum landraces (a typical characterstics of the race bicolor) in the current collections also implies the low preponderance of the race bicolor in the Tigray region as compared to the more abundant race durra. Geleta and Labuschagne (2005) showed that open seeded (less than 50% grain coverage) sorghum varieties were the most frequent variant in sorghum samples collected from eastern highlands of Ethiopia. Respectively, naked and covered seeded sorghums were more abundant in the lower and higher elevation areas of Tigray and this agrees with what Ayana and Bekele (1998) had reported.

The higher frequency of covered landraces in the highlands may attribute to human selection as covered grains are important in protecting grains from fungal disease, which is prevalent in highland as this contains high rainfall that favors fungi. Further, the prevalence of covered grains in the highland areas of Tigray may also imply the concentration of the race bicolor in humid highland environment. The frequency of the race bicolor showed a clinal variation that its abundance linearly increases with altitude and this may indicate that this race is adapted to higher elevation where moisture is plenty and plant diseases including mold is severe. Doggett (1991) also noted that covered glume sorghum landraces (race bicolor) can escape bird attack and resist grain mould prevalent in humid highland areas.

Diversity estimates

In the current sorghum landrace collections, high level of pooled mean diversity (H′ = 0.70 ± 0.05) across all traits was observed. However, this estimate was slightly lower than the previous finding by Ayana and Bekele (1998), who reported relatively higher estimate of diversity (H′ = 0.76) in Tigray sorghum ex situ collections (genebank accession). The minor discrepancy may be attributed to differences in sample size, number and type of traits and trait classes scored in both studies. Moreover, in the current research, vernacular name was used as a reference and the level of diversity (intraspecific diversity) within a given landrace represented by a given local name was almost none, unlike the study by Ayana and Bekele (1998) where a given accession (likely population of mixed genotypes) was used as a sample for estimating the diversity. Nevertheless, the high level of diversity estimate observed in the current collections may attributed to the existence of diverse agro-climatic landscapes that hosts different sorghum landraces with variable agronomic significance. In addition, farmers’ willingness to maintain diverse bundles of landraces with different attributes that suites to different end use qualities and cultural niches may also contributed for the high diversity estimate. The result may also suggest the likely availability of sorghum landraces having important genes of adaptive significance for future sorghum improvement work against the increasing effect of climate change. Further, the observed high level of diversity in the sorghum landraces shades light on the presence of some specific important sites in the Tigray region where high sorghum gene reservoirs exist that can be a potential candidate sites for on-farm conservation.

Sorghum race identification and distribution

Panicle compactness was very well distributed in the study areas indicating the presence of different race types of sorghum in the region because this character is one of the major criteria for race identification and classification in sorghum (Harlan and De Wet 1972; Stemler et al. 1977; Doggett 1988). Thus, variation in the panicle compactness found in this study is an opportunity for sorghum breeders to develop varieties for different argo-ecological conditions. Humidity of the environment at flowering and ripening time is a factor for panicle compactness in that those with compact panicle (e.g. race durra) are confined to in dry conditions, whereas the open type (e.g., race bicolor) is most common in areas of high rainfall (Harlan and De Wet 1972).

Races of durra, bicolor, caudatum and durra-bicolor were identified in the current study. Most of the landraces investigated in this study were durra sorghum. This supports the fact that the race durra is widely adapted to low precipitation (Harlan and De Wet 1972), which is one of the typical characteristics of the study area, Tigray. In addition, Stemler et al. (1977) noted that durra is a higher share holder in the annual production of sorghum in Ethiopia. Other reason for the wide spread of this race is because in the region sorghum is largely used for food consumption in which farmers prefer durra type as it usually has no pigmentation on its pericarp and yields high quality white flour (Abdi et al. 2002) that is preferred for making enjera. It was observed that durra has extensive adaptability from lowland to the highlands up to 2400 masl. This race was more dominant in the southern than in other zones of Tigray. Sorghum landraces with recurved peduncle position, which is one of the attributes of race durra, is also prevalent in this zone. Amelework et al. (2016) has defined this area as one of the main sites for the race durra. The race caudatum is economically important race and is widely used in breeding programs. Doggett (1988) had reported that caudatum sorghum is often associated with pastoralists. However, the race caudatum was very common in northwest part of Tigray where the agricultural system is different from pastoralists.

Prediction of sorghum suitability in the face of climate change

Identification of suitable areas for sorghum production in the future was conducted using the software DIVA-GIS (Hijmans et al. 2012). DIVA-GIS is a free computer program for mapping and geographic data analysis as well as analysis of climatic data (Hijmans et al. 2001). It is useful to analyze plant genetic resources to elucidate ecological and geographic patterns of distribution of crops. Here, the DIVA-GIS software tool was used for the prediction of suitable areas for sorghum production using the climatic data annual precipitation and temperature. Accordingly, many areas that were surveyed in this study and other sites in the Tigray region are found to be suitable for sorghum cultivation in the future. This reflects that sorghum is a climate smart crop that can tolerate the most immediate impacts of climate change. Thus, several areas in the region that are currently covered by other cereals may be replaced by sorghum suggesting sorghum will have the potential in changing the crop production scenario and food system of the region. In agreement with this, Msongaleli et al. (2014) revealed that sorghum is resilient to projected changes in climate by 2050s. With respect to race type, durra sorghum will still be consistent in the future climate scenario and the cultivation of durra will not be affected by future climate changes. This prediction will help in identification of important sites to set conservation strategies considering the possible future novel climate situations. Based on this prediction, comprehensive collection and expedition mission can be launched to thoroughly collect the adapted genetic resources for future breeding efforts.

Conclusions

A total of 358 landraces, out of which 140 are distinctly named, that are maintained and managed by farmers were identified and deposited ex-situ in the national genebank. High degree of phenotypic variation among the sorghum landraces was found though the amount of variability is not similar in all zones and altitude classes. Landraces with unique properties such as twin seeded sorghum were found which is essential to include in the breeding programme of sorghum. Races durra, bicolor, caudatum and durra-bicolor were identified in which the durra race was highly dominant. Based on the future climate data, all the currently surveyed areas and others are predicted to be suitable for sorghum cultivation suggesting sorghum is climate smart crop and may dominate the crop production system and change food consumption tradition of the region.

In the view of genetic resources conservation, the lowlands of northwest, central, and southern parts of the Tigray region could be used as potential sites for in-situ (on-farm) as well as ex-situ conservation and utilization. The high level of sorghum landrace diversity observed in the regional collections can be in the sorghum improvement programme of the region. Further, in light of farmers' rights platform, as recognized in the International Treaty on Plant Genetic Resources for Food and Agriculture (ITPGRFA), suitable mechanisms need to be established on how benefits from the use of the genetic resources and the associated knowledge will be shared with the farming communities. For the safety of this wealth of landrace diversity against disasters including war, establishment of duplicate gene bank at regional level is important as storage only in the national genebank in the capital does not provide sufficient security. Though the current phenotypic marker-based analysis provided good insight of the level of diversity, this alone may not be enough to reveal all the diversity in the material. Thus, complementing the findings with the ongoing SNP based molecular marker analysis (Semere et al. in prep) is crucial to capture the overall genetic variability in sorghum landraces in Tigray.