Optimizing Sowing Depth of Tef for Irrigated Mediterranean Conditions: From Laboratory to Field Studies

: Developing new crops adapted to arid conditions is a promising approach to meet the increasing demand for food production under expanding aridity. Tef [ Eragrostis tef (Zucc.) Trotter] is a C4 cereal crop cultivated mainly in Ethiopia and the Horn of Africa, and known for its high resilience to stressful environments. Due to their tiny size, tef seeds are traditionally sown by broadcasting and lightly covering with soil. Under semiarid Mediterranean conditions, a deeper sowing may guarantee seedling establishment while saving on irrigation water. The objective of this study was, therefore, to determine the e ﬀ ects of sowing depth on tef emergence, development, lodging, and productivity. Tef seeds were sown at di ﬀ erent depths in test tubes and pots, and in two ﬁeld experiments. In tubes and pots, time from sowing to emergence increased about twofold and emergence rate decreased in the deepest sowing (3 cm) treatment compared to controls (0 cm). In the pot and ﬁeld trials, deep sowing (3 cm) signiﬁcantly reduced plant height, shoot and root biomass, and lodging. Sowing depths of 1–2 cm allowed successful plant establishment while not exacting penalties on plant emergence or development; hence this range appears to be optimal for sowing irrigated tef.


Introduction
Climate change toward higher temperatures, greater aridity, and more frequent erratic climate events is posing a major threat to agricultural sustainability [1]. The development of new crops adapted to arid conditions is among the promising approaches to meet the projected demand for food production under expanding aridity. The introduction of a crop into a new region requires comprehensive studies of the crop's responses to environmental factors. The adoption of a crop from a traditional rain-fed system in an intensive, mechanized, and irrigated cropping system is particularly challenging, requiring examination and adjustment of management practices, from field preparation, sowing, irrigation, mineral nutrition, and plant protection, to harvesting techniques.
Tef [Eragrostis tef (Zucc.) Trotter] is a cereal crop that is cultivated mainly in Ethiopia and the Horn of Africa [2]. Tef has a thin culm, long narrow leaves, and a thousand-kernel weight of 200-400 mg, making it the smallest-grained cereal [3]. Recently, tef has been gaining popularity in many Western countries due to its dual potential-as a gluten-free "super food" grain [3], and as high-quality forage [4,5]. In Israel, the demand for tef grain is further increased by the Ethiopian descent citizens who continue to consume it for their traditional cuisine [6].

Test-Tube Experiment
The effects of sowing depth on seedling emergence and establishment were studied in a test-tube experiment for all six genotypes. Seeds were sown in 50 mL test tubes filled with brown-red sandy loam soil composed of 76% sand, 8% silt, and 16% clay ( Figure 1).

Test-Tube Experiment
The effects of sowing depth on seedling emergence and establishment were studied in a testtube experiment for all six genotypes. Seeds were sown in 50 mL test tubes filled with brown-red sandy loam soil composed of 76% sand, 8% silt, and 16% clay ( Figure 1).
A hole was drilled at the bottom of each tube to allow drainage and covered with a thin layer of fabric to avoid soil leakage. A factorial experimental design was employed with the six genotypes, each sown at four depths: 0 cm (control) with 1-2 mm soil coverage, 1 cm, 2 cm, and 3 cm. Test tubes were filled with dry soil to the designated level, watered, and 20 seeds from a single genotype were scattered on the soil surface of each tube. To ensure an accurate sowing depth, preweighed soil was added to match a soil layer of 1, 2, and 3 cm in the 50 mL test tube, while the 0 cm treatment seeds were gently mixed into the top 1-2 mm of soil. After sowing, tubes were wrapped with aluminum foil to avoid root exposure to light. A total of 24 tubes-one tube for each of the six genotypes and four sowing depths-were sown in each of five growing cycles, referred to as experimental blocks.
The tubes were placed in a climate-controlled greenhouse at the Faculty of Agriculture, Food and Environment campus in Rehovot, Israel, under temperatures of 28/22 °C day/night, and were watered twice daily by microsprinklers. At the end of the growth cycle, after all plants reached the second leaf stage, plants were uprooted, and the soil was gently washed away to observe the root system. A hole was drilled at the bottom of each tube to allow drainage and covered with a thin layer of fabric to avoid soil leakage. A factorial experimental design was employed with the six genotypes, each sown at four depths: 0 cm (control) with 1-2 mm soil coverage, 1 cm, 2 cm, and 3 cm. Test tubes were filled with dry soil to the designated level, watered, and 20 seeds from a single genotype were scattered on the soil surface of each tube. To ensure an accurate sowing depth, preweighed soil was added to match a soil layer of 1, 2, and 3 cm in the 50 mL test tube, while the 0 cm treatment seeds were gently mixed into the top 1-2 mm of soil. After sowing, tubes were wrapped with aluminum foil to avoid root exposure to light. A total of 24 tubes-one tube for each of the six genotypes and four sowing depths-were sown in each of five growing cycles, referred to as experimental blocks.
The tubes were placed in a climate-controlled greenhouse at the Faculty of Agriculture, Food and Environment campus in Rehovot, Israel, under temperatures of 28/22 • C day/night, and were watered twice daily by microsprinklers. At the end of the growth cycle, after all plants reached the second leaf stage, plants were uprooted, and the soil was gently washed away to observe the root system. The number of emerged seedlings and number of seedlings that reached the second leaf stage were recorded daily for each test tube. Time from sowing to emergence (TSE), time from sowing to fully expanded second leaf (TSL2), and time from emergence to second leaf (TEL2) were calculated as follows: where n is the number of seedlings emerged at each counting, t is the time in hours from sowing to counting, and N is the total number of seedlings emerged.
where s is the number of seedlings with a fully unfolded second leaf at each counting, t is the time in hours from sowing to counting, and S is the total number of seedlings that reached the second leaf stage. TEL2 = TSL2 − TSE

Pot Experiment
Three genotypes, RTC-119 (large seeds), RTC-400 (medium seeds), and RTC-19 (small seeds), were selected for the pot experiment (Table 1). To enable accurate sowing depth, transparent cylindrical pots (8.2 cm diameter, 18 cm height), made from plastic soft-drink bottles, were used for this experiment. Holes were drilled at the bottom of each pot to allow drainage and covered with a thin layer of fabric to avoid soil leakage. A factorial experimental design was employed with three genotypes, each sown at four depths (0 cm (control), 1 cm, 2 cm, and 3 cm), and six replicates (a total of 72 pots). The type of soil used, number of seeds per experimental unit, sowing methodology, and scoring of seedling emergence and second leaf stage were as in the test-tube experiment. Once all seedlings reached the second leaf stage, all but one randomly selected seedling were removed from each bottle and the remaining plant continued to grow until harvest at flowering onset. Just before harvest, the lodging angle (between the main culm and the vertical) was measured. The plants were then uprooted, the soil was washed away, and number of tillers, crown diameter, number and diameter of crown roots, and shoot and root dry weight were assessed following the protocols in Ben-Zeev et al. [21].
Soil preparation included shallow tilling and flattening with a heavy-duty leveler and crumble roller tool to obtain a smooth and even seedbed. Seeds were tested to confirm uniform germination rates and were treated with Vitavax ® fungicide (thiram + carboxin, Gadot Argo, Israel) at a rate of 100 µL/10 g seeds. Experimental plots were mechanically sown using a Plotseed S seeder (Wintersteiger, Ried, Austria). Ten rows were sown in each 8 m × 1.6 m plot at two sowing depths (1 or 3 cm) using a sowing rate of 6 kg/ha (Hula) or 4 kg/ha (Revadim).
Seedling density was assessed 10 days after emergence. A metal frame with inner dimensions of 25 cm × 100 cm was placed across the seedling rows in two locations per plot and the number of seedlings within this frame area was counted.
Lodging was assessed visually by two independent surveyors twice a week, starting from six weeks after emergence (WAE), using a scoring method adapted from Caldicott and Nuttall [22].
Lodging was scored on a 10-level severity scale (0 being a vertical nonlodging plant and 9 being a horizontal fully lodging plant), and lodging prevalence was determined (percentage of the entire plot area). Lodging index was calculated as the product of lodging severity × lodging score.
Plant height and panicle length assessments and harvest were performed at grain maturity, 12 WAE. Three plants, randomly selected from various parts of each plot, were measured from the soil surface to the bottom and top of the panicle and used to calculate the average culm length and panicle length, respectively. A metal frame of 25 cm × 100 cm was placed in the middle of the plot across six rows (excluding the two outer rows from each side) and all of the biomass bordered by the frame was manually harvested and collected into paper bags. Biomass samples were dried in a hot glasshouse (maximum temperature~55 • C) for two weeks, weighed for total dry matter, and threshed using a LD350 thresher (Wintersteiger, Ried, Austria). Seed samples were cleaned of debris and weighed to determine grain yield (GY).

Statistical Analysis
Statistical analysis was conducted using JMP ® Pro, Version 15, software (SAS Institute Inc., Cary, NC, USA, 1989-2019), and included two-way analysis of variance (ANOVA), Student t test, and Tukey HSD test (for factors showing a significant F ratio).

Test-Tube Experiment: Seedling Emergence and Establishment
The six tef genotypes, sown at four depths (0, 1, 2, 3 cm) in test tubes, presented highly significant effects of both sowing depth and genotype on average TSE and emergence rate, with no significant interaction between the two ( Table 2). Table 2. Test-tube experiment. ANOVA for average time from sowing to emergence (TSE), emergence rate, time from emergence to second leaf (TEL2), and time from sowing to second leaf (TSL2).

Source of Variation (d.f.) TSE Emergence Rate TEL2 TSL2
F ratio Sowing depth (SD) (3) 193.4 *** 8.2 *** 16.18 *** 9.38 *** Genotype (G) (5) 6.0 *** 26.4 *** 3.12 * 7.42 *** SD*G interaction (15) 1.1 0.9 0.58 0.48 Block (4) 0.4 0.2 8.14 *** 8.13 *** *, ***, significant at p < 0.05, 0.001, respectively. TSE increased twofold from 55 h at the shallowest sowing depth (0 cm) to 109 h at the deepest sowing depth (3 cm) ( Figure 2A); the other treatments increased in similar increments across this range, with all treatments differing significantly from one another. TEL2 exhibited the opposite trend, decreasing from 134 h at the shallowest sowing depth to 104 h at the deepest sowing ( Figure 2A). This trend compensated for a substantial part of the delay in emergence from deeper sowings, and therefore TSL2 ranged between 189 h for the shallowest treatment to 214 h for the deepest treatment. Emergence rates recorded at 0 and 1 cm depths were 47%, with no apparent difference between them. Emergence rate was significantly reduced to 33% in the 3 cm treatment, whereas the 2 cm depth exhibited intermediate values, not significantly different from the other treatments ( Figure 2B). The delayed emergence and lower emergence rate induced by deeper sowing can be clearly visualized in Figures 1 and 3.  The six genotypes tested in this study presented a rather uniform TSE, with only RTC-19 showing a significantly higher value (slower emergence) compared to most other genotypes ( Figure   Figure 2. Test-tube experiment. Effects of sowing depth (A) and genotype (C) on average time from sowing to emergence (TSE, brown segment), time from emergence to second leaf (TEL2, green segment), and time from sowing to second leaf (TSL2, both segments combined). Effects of sowing depth (B) and genotype (D) on average emergence rate. Different letters indicate significant differences (p < 0.05).  The six genotypes tested in this study presented a rather uniform TSE, with only RTC-19 showing a significantly higher value (slower emergence) compared to most other genotypes (Figure The six genotypes tested in this study presented a rather uniform TSE, with only RTC-19 showing a significantly higher value (slower emergence) compared to most other genotypes ( Figure 2C). There were no significant differences between genotypes for TEL2, hence TSL2 of the various genotypes exhibited a trend that was rather similar to that of TSE. An interesting trend was observed with the emergence rates: the small-seeded genotypes (RTC-176 and 19) exhibited significantly higher emergence rates (56-60%) ( Figure 2D) than the medium-and large-seeded genotypes (26-42%), with no consistent trend among the latter. However, the number of genotypes tested in this study is too small to extrapolate from these data to a relationship between seed size and emergence rate.

Pot Experiment: Emergence and Plant Performance
The three tef genotypes sown in pots at 0, 1, 2, and 3 cm depths exhibited significant effects of both sowing depth and genotype on TSE and emergence rate ( Table 3). TSE increased significantly with every increase in sowing depth from 0 to 3 cm ( Figure 4A), similar to the test-tube experiment. Emergence rate was significantly reduced at the deepest sowing depth compared to the other depths ( Figure 4B). Genotype-by-sowing depth interaction was of low significance for TSE and nonsignificant for emergence rate (Table 3); whereas TSE for all three genotypes responded similarly to increased sowing depth ( Figure 4C), emergence rate of RTC-19 and 119 decreased with sowing depth, while no effect was recorded for RTC-400 ( Figure 4D). Sowing depth significantly affected plant height and shoot and root biomass, and thus also the root-to-shoot weight ratio. Plant height ranged from 45.5 to 65.5 cm for RTC-19 at 3 cm and RTC-400 at 0 cm, respectively, and was generally reduced as sowing depth increased, with the exception of RTC-119 ( Figure 5). Root and shoot biomass (range 0.28-0.6 g and 0.2-0.39 g, respectively) decreased with increasing sowing depth. Root-to-shoot ratio generally increased with increasing sowing depth, with RTC-119 as the only exception (data not shown).

Pot Experiment: Lodging-Related Traits
Lodging-related traits, which included lodging angle, number of internodes and tillers, crown diameter, and crown root diameter and number, were assessed 12 weeks after sowing. ANOVA for lodging-related traits revealed significant effects of both sowing depth and genotype on the number of internodes, and significant effects of genotype on root diameter, whereas sowing depth-by-genotype interactions were not significant (Table 4). Crown root number and tiller number were not affected by sowing depth, genotype, or their interaction. Sowing depth had a moderate effect on lodging angle (angle between the main culm and the vertical), which fell below the statistical threshold by ANOVA, but did show significant differences between treatments by Tukey HSD test. The average lodging angle for the 3 cm treatment (40 • ) was significantly smaller (lower lodging) than for the 0 cm treatment (62 • ) ( Table 4). Another variable affected by sowing depth was the number of internodes, which decreased incrementally with the increase in sowing depth, resulting in a significant difference between the 0 cm treatment and both 2 and 3 cm treatments ( Table 4).
The tested genotypes did not have a significant effect on lodging angle. Significant differences between genotypes were manifested in three other lodging-related traits, with RTC-400 showing a significantly higher number of internodes and significantly lower crown and root diameters, compared to the two other genotypes, which did not differ from one another (Table 4).

Field Experiments: Emergence and Plant Performance
Two and three genotypes were mechanically sown in the field to depths of 1 or 3 cm in the 2019 and 2020 seasons, respectively. Sowing depth had a significant effect on number of seedlings/m 2 and plant height in both years (Table 5), whereas panicle length and GY were only affected in 2019. The genotypes had significant effects on GY in both years, and on the number of seedlings/m 2 in 2020.
Deeper sowing reduced seedling density by about 15% in both years (Table 5), and reduced plant height by 25% in 2019 and 7% in 2020. In 2019, panicle length and GY under the 3 cm sowing treatment were 21% and 39% lower, respectively, than for the 1 cm depth treatment, whereas no such effects were evident in 2020.
Seedling number did not differ between genotypes in 2019; however, in 2020, RTC-19 (not included in the 2019 trial) had a significantly greater number of seedlings than the other two genotypes (Table 5). This result corresponded with the greater germination rate of RTC-19 in our test-tube experiment ( Figure 2). GY was significantly affected by genotype, with no consistent trends across years; RTC-119 exhibited the highest GY in 2019 and the lowest in 2020 (Table 5).

Field Experiments: Lodging
The weekly lodging index value, based on two evaluations per week, was subjected to ANOVA. In 2019, lodging indices at 7 and 8 WAE were significantly affected by sowing depth; the genotype effect was significant throughout 7-10 WAE and depth-by-genotype interactions were not significant ( Table 6). In 2020, lodging index was not significantly affected by sowing depth, genotype effect was significant only at 8 WAE, and depth-by-genotype interactions were significant at 7 and 10 WAE. The whole-season analysis showed a significant effect of time (Week) and sowing depth in both years, a significant effect of genotype in 2019, and significant depth-by-genotype interactions in 2020 (Table 6). A plot of lodging indices vs. WAE showed a similar trend in both years, with the deep sowing treatment exhibiting lower lodging 7 and 8 WAE, whereas at 9 and 10 WAE, this advantage was diminished ( Figure 6). *, **, ***, significant at p < 0.05, 0.01, 0.001, respectively. Different uppercase letters indicate significant differences between sowing depths or genotypes.

Emergence Time and Rate
All three experimental setups presented similar trends: deeper sowing of tef resulted in later emergence and lower emergence rates in test tubes (Figure 1, Figure 2A,B and Figure 3), pots ( Figure 4A,B), and field trials ( Table 5). A sowing depth of 1 cm led to a slight but significant increase in TSE in both test tubes and pots. Larger effects on germination were recorded at a sowing depth of 2 cm, and the greatest effects were evident at a sowing depth of 3 cm (Figures 2A and 4A). In both the test-tube and pot experiments, TSE increased incrementally with increasing sowing depth, reflecting the longer distance from the seed to the soil surface (Figures 2A and 4A). TEL2, measured only in the test-tube experiment, exhibited a slight decrease with increasing sowing depth, thus partially compensating for the effect of sowing depth on TSE (Figure 2A). A rather clear breaking point can be observed with respect to germination rate, with a significant reduction for the 3 cm depth treatment in both test tubes ( Figure 2B) and pots ( Figure 4B), which was confirmed in the two field trials (Table 5). These trends were, for the most part, uniform across genotypes, in both the test-tube experiment (data not shown) and pot experiments ( Figure 4C,D).
These results suggest that sowing tef at 1 or 2 cm depth has no or only minor effects on TSE or TSL2, whereas sowing at 3 cm depth has significant negative effects on these variables. In wheat (Triticum spp.) and barley (Hordeum vulgare), increasing sowing depths along the ranges of 0.8-14.3 cm [23] and 5-20 cm [24], respectively, reduced emergence rate and increased TSE. Previous studies on tef, all conducted in pots, have reported a reduction in the rate of and time to germination for sowing depths of 0 or greater than 2 cm [5,[13][14][15]18]. Our test-tube and pot experiments are generally in agreement with these studies, although we did not observe a reduction in the 0 cm sowing treatment and did observe an increase in TSE at 2 cm sowing. These results were further validated in our two field experiments in two different locations with different soil types. We are not aware of any prior study on tef sowing depth conducted under field conditions.

Plant Development and Performance
In other tropical grasses, such as maize (Zea mays L.) and sorghum (Sorghum bicolor L.), the mesocotyl elongates until the coleoptile is exposed to light [25], thus locating the apical meristem at a depth corresponding to the coleoptile length, which is associated with seed size [26]. Being the smallest-seeded cereal [3], tef has a short coleoptile which locates the apical meristem just below the soil surface. Indeed, regardless of sowing depth, the root-shoot junction of tef seedlings for all treatments was located, upon emergence, just below the soil surface ( Figure 3). Nevertheless, plants sown at 2 and 3 cm depths developed into shorter, lower biomass plants in both pots and field trials, with shorter panicles and a lower GY (2019 field trial) than plants in the shallower sowing treatments ( Figure 5, Tables 3 and 5). These differences in shoot traits, manifested at a later plant age, can be attributed to a penalty incurred from the deeper sowing. Similar penalties have been reported for wheat and barley [23,24], including a reduction in number of leaves and internode lengths, plant height, tiller number (no significant effect found in the present study), biomass, and GY. Confounding trends have been reported for sorghum, with a nonsignificant effect on plant biomass and a significant increase in plant height at deeper sowing depth [27].

Effect of Sowing Depth on Lodging and Related Traits
Lodging is a major yield-restricting factor in tef [18,19]. Deeper sowing has been shown to reduce root lodging in wheat [17], possibly due to the formation of a deeper root plate [28], which improves plant anchoring to the soil. In our pot experiment, deeper sowing of tef induced a significant effect on number of internodes, whereas its effects on lodging angle, crown diameter, and crown root diameter fell below the statistical threshold (Table 4). In our previous study, lodging of four genotypes grown in the field under various plant densities was associated with crown diameter, crown root diameter, and crown root number [21]; in the pot experiment, crown diameter and crown root diameter showed similar effects.
In both field trials, sowing tef seeds at 3 cm depth significantly reduced the seasonal average lodging index compared to 1 cm sowing depth (Table 6), with a particularly pronounced effect at 7-8 WAE ( Figure 6). The effect of sowing depth on tef lodging has not been reported in previous studies. Recently, we reported that lower seedling density reduces lodging index [21]. Published studies on tef [20,29] and other cereals [28] have shown an association between lower plant height and reduced lodging. Both seedling density and plant height were reduced by the deeper sowing in the field (Table 5). Therefore, we cannot rule out the possibility that these two factors contributed to the reduced lodging observed in our field trials. However, the significant increase in root-to-shoot ratio (not presented) and reduced number of internodes, as well as the nearly significant modifications in crown and crown root diameters in the pot experiment, suggest that additional mechanisms contribute to reduced lodging under the deeper sowing treatment in the field.

Conclusions
This study was motivated by the need to reduce the amount of irrigation water required for tef establishment under semiarid Mediterranean conditions. An optimal sowing depth should allow sufficient soil above the seed to maintain the required moisture for germination, while not inflicting penalties on seedling emergence or plant development. Despite the tiny size of tef seeds, which was thought to require a very shallow sowing depth, we showed that tef can successfully emerge from a depth of 1-2 cm without any apparent penalty. Tef even emerged from 3 cm depth, while showing reduced lodging, but this was coupled with a penalty on plant development and productivity. Therefore, we conclude that a 1-2 cm depth is optimal for sowing irrigated tef. Further research combining optimal sowing depth, sowing rate, and irrigation regimes with a wide genotypic range to evaluate plant development, lodging and productivity, could lead to advances in tef production in Mediterranean and other similar regions around the globe.