Finger millet is one of the most nutritious food crops extensively grown in Asia and Africa (Rodríguez et al., 2020). The crop covers 12% of millets that are in the world and is ranked fourth after sorghum, pearl millet and foxtail millet (Vetriventhan et al., 2016). In arid and semi-arid regions, soil moisture stress is the major abiotic constraint that adversely affects crop productivity (Choudhary and Padaria, 2015). Finger millet has been reported to be drought resilient as compared to other cereals such as maize (Zea mays) (Gupta et al., 2017).

Studies carried out on finger millet genotypes showed that there exists genotypic variation in the degree of drought tolerance among different varieties (Bartwal et al., 2016; Bartwal and Arora 2017). Finger millet is well adapted to temperature ranges of 11 to 28°C. However, it can thrive well under hot conditions where temperatures are as high as 35°C. Although finger millet is drought tolerant, its growth is adversely affected by both intermittent and terminal droughts. The crop is largely grown by subsistence farmers who rely on rain fed agriculture, hence prone to the risk of economic yield loss due to drought.

Feeding the fast-growing human population with balanced nutritional diet under unpredictable severe weather events is a challenging task globally. The climate change crisis is expected to cause shifts in food production and yield loss, causing a severe threat to food security (Dhankher and Foyer, 2018). A key strategy to adapt to a changing climate is to develop and promote elite germplasms with stable yields that can survive under changing weather conditions (Bhat et al., 2018). There exist great potential in underutilized crops such as finger millet that are well adapted to extreme weather conditions and can act as an alternative food resource towards ensuring food and nutritional security (Mabhaudhi et al., 2019). Despite the many advantages offered by the cultivation of finger millet in Africa including, Kenya, there is limited research on tolerance to drought in finger millet. The production of finger millet is restricted to low yielding and poorly adapted genotypes (Mgonja et al., 2013). However, there is great potential to increase production through screening and selection of well adapted genotypes to low soil moisture with better grain yield.

Numerous morpho-physiological and biochemical traits such as shoot length, root length, shoot to root ratio, relative water content and stomatal conductance among others are considered important under drought stress conditions (Murtaza et al., 2016). In a related study, Mude et al. (2020) reported that water use efficiency, harvest index and biomass are important for resilience to drought in cereal crops. In contrast, decrease in root growth, relative water content and lipid peroxidation was found to show a considerable level of tolerance to drought stress (Mukami et al., 2020). Finger millet improvement in Kenya in the past has laid emphasis on selecting for high yielding lines with little regard on drought tolerance traits (Mukami et al., 2020). Drought tolerant finger millet lines have not yet been developed in Kenya where arid and semi-arid land covers 80%. Therefore, the present investigation was conducted to identify finger millet lines with enhanced tolerance to drought based on morpho-physiological traits with the intention to be used in future breeding programmes to develop improved drought tolerant cultivars.

Description of the experimental sites

The study was conducted in the field at two locations; Agricultural Training Centre (ATC) Koibatek in Baringo County and ATC Soin in Kericho County in 2020. ATC Koibatek is located at 1°35?S, 36°66?E and elevated at an altitude of 1890 meters above sea level and falls in the Upper Midland zone 4 (UM4) agro-ecological zone (AEZ). ATC Soin is located between latitude 0^o 23?S and longitude 35º 02?E with an altitude of about 2002 m above the sea level and falls in the Lower Midland zone 3 (LM3) AEZ. The climatic conditions of the respective study sites are represented in Table 1.

Finger millet genotypes

The planting material used in this study consisted of 25 genotypes (24 advanced finger millet lines and one commercial check cultivar, P224). These genotypes were obtained from International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), Kenya Agricultural and Livestock Research Organization (KALRO) and Egerton University Seed Units (Table 2).

Experimental design and agronomic practices

The field experiment was conducted under rain-fed conditions during the long rainy season (June to November 2020). Land preparations were done according to ICRISAT recommendations (ICRISAT, 1992). The seeds were planted on June 13, 2020 and June 14, 2020 in Soin and Koibatek locations, respectively. Lattice design with five blocks consisting of five plots per block with three replications was used to carry out the experiment. The plot size was 4 m² with four rows, 2-m length. The seeds were drilled by hand at a depth of 2 cm in rows, 15 cm apart, with seed rate of 3.2 kg . At planting, Di-ammonium phosphate (DAP) fertilizer was applied at 20 kg ha^-1 to each experimental plot to supply a basal fertilizer dose of 10 kg . Two weeks after emergence, the plants were thinned to one plant per hill. Topdressing was done using Calcium ammonium nitrate (CAN) at the rate of 30 kg ha^-1 to supply 8 kg N , applied in three split doses, (50% two weeks after emergence, 25% at five leaf and 25% at the time of flowering). Weeding was done twice by hand, two weeks after emergence and two weeks after the first weeding. Insect pest and disease control was carried out as required.

Data collection

Three plants were randomly selected and tagged from the two middle rows in each experimental plot and data collected on morphological, physiological, yield and yield parameters. For the morphological parameters, seedling vigour, plant height, total number of tillers and productive tillers, finger number and finger size were recorded following the International Board for Plant Genetic Resources (IBPGR, 2011) for finger millet. Root to shoot ratio, total biomass (measured as sum mass of the weight of above ground parts of the plant and root), and harvest index (measured as ratio of grain yield to the total biomass) were taken at harvesting where the plants were uprooted and the biomass was divided into shoot and root. The shoot was oven dried; whereas the root was washed using tap water and dried in the oven at 70°C for 24 h. The biomass dry weight was taken using an electronic balance.

Where; W= Critical difference, P= number of treatment means, fe=error degrees of freedom, α= level of significance, MSE =mean square error and r= number of replicates.

In this study environmental and genotypic effects were significant for the agronomic, morphological and physiological traits among the evaluated finger millet genotypes. Plant responses to drought stress have been shown to vary depending on drought level, plant species and plant growth stage (Mukami et al., 2019). Therefore, for drought tolerance agronomic, morphological and physiological traits can provide important information to improve crop production in arid and semi-arid area. Drought adapted crop genotypes may be considered to have various mechanisms such as avoidance, escape or tolerance. However, genotypes that possess these adaptive mechanisms hardly express desirable agronomic characteristics, such as grain yield (Dhami et al., 2018). Evaluation of crops for traits related to drought adaptation has been shown to be limited (Nadeem et al., 2020). The reason being that most of the approaches used for screening drought tolerance are below ground, which are tedious and may involve destructive sampling (Gebreyohannes et al., 2021).

The results from this study revealed a significant variation among the finger millet genotypes for the morphological and physiological traits, with greater implication on the differences under drought stress conditions. These results can be useful in the selection of parental stock for breeding in drought improvement programmes and possible release for commercial production of promising lines. In similar studies, drought tolerance have been reported to vary among finger millet genotypes evaluated in Uganda and Ethiopia (Owere et al., 2016). The variation in the agronomic traits observed across the two study locations for the number of productive tillers, number of fingers, finger length and yield could be attributed to genotypic and environmental differences (Dramadri, 2018). High grain yields observed in Koibatek can be directly associated with high number of fingers, finger length, number of productive tillers and early flowering. In a similar study, improved performance for agronomic traits under drought stress was positively correlated with grain yield (Shanker and Shanker, 2016). According to Bennani et al. (2016) reduced number of days to flowering and heading was considered as one of vulnerabilities of plants to drought stress. Drought stress severity, plant species and crop growth stage as well have been attributed to influence grain yield (Demirevska et al. 2009).

Seedling vigour is considered as one of the reliable phenotypic traits towards selection of drought tolerance at the seedling stage. Among the evaluated finger millet lines, there was a significant variation for the seedling vigour, signalling potential tolerance to soil moisture deficit at the seedling stage. In a related study by Struik et al. (2007), seedling vigour was included in the evaluation of wheat genotypes for drought tolerance at the early growth stage. Vigorous and fast growing plant seedlings can compete against weeds at an early stage, which is critical for better grain yield (Zhang et al., 2015). Ahmad et al. (2015)evaluated 50 wheat genotypes for different seedling traits including seedling vigour, and successfully identified eight potentially drought-tolerant genotypes.

Plant height is one of the morphological traits which can be used for selecting drought tolerance among crop genotypes. In previous studies, plant height was directly linked to grain yield, where short plants had higher grain yield compared to taller plants (Mohammadi et al., 2012; Koocheki et al., 2014). Short plants were found to reduce moisture demand and prevent plant moisture loss due to transpiration (Zhang et al., 2018). In wheat (Triticum aestivum), reduced plant height was reported to reduce photosynthesis and nutrient translocation, especially during the stem elongation stage due to low moisture content (Sarto et al., 2017). Reduced plant height has also been associated with increased partitioning of assimilates to the ear (Grover et al., 2018). Short plants may also result in higher HI and lodging resistance (Divashuk et al., 2013).

Increased number of productive tillers could be a desirable trait to higher grain yield. However, under drought stress, this trait can be detrimental due increased competition for assimilate partitioning (Geleta et al., 2019). In this study, the number of productive tillers was negatively correlated with harvest index. Similar findings were reported by Lule et al. (2012), where grain yield low was registered in finger millet genotypes, which had high number of productive tillers under low soil moisture.

Positive correlations were registered between days to flowering and grain yield. Similar results were reported by Ganapathy et al. (2011) and Chandra et al. (2013), who found that late maturity was associated with grain yield and yield components. In this study, 1,000-grain weight, finger number, finger length, days to maturity and harvest index were positively correlated with grain yield; this was in accordance with the results of Bezaweletaw et al. (2006), who reported a positive association of 1,000-grain weight with finger number, finger length, days to maturity, harvest index and grain yield per plant. Moreover, Wolie et al. (2013) and Kumar et al. (2016) found grain yield to be positively correlated with biomass and harvest index in finger millet.

Harvest index (HI) can influence yield, as it is the proportion of the whole plant mass that is partitioned to the seed (Pachepsky et al., 2011). Harvest index is the partitioning of dry matter into the reproductive parts; hence, it can be used as an important indicator for drought tolerance. In this study, results showed a significant positive relationship between HI and grain yield. Similar results have been also reported byJyothsna et al. ( 2016)and Reddy (2020), where harvest index was positively correlated with number of tillers per plant, finger length and grain yield. Grain yield is considered a complex trait that is highly influenced by genotypic and environmental factors. Therefore, high variation observed for yield among the finger millets can be attributed to genotypic and environmental difference across the two locations. However, lines ICFX 1420314-2-1-1-1 (7), KNE 814 X Ex Alupe (P) P8-1-1-1-1 (24) and ICFX 1420415-3-1-1-2 (14) displayed consistency with relatively high grain yield both in Koibatek and Soin. Similar findings reported that high variation in grain yield was attributed to both genetic and environmental factors (Malambane and Jaisil, 2015; Mukami et al., 2019).

Physiological traits were noted to vary across the finger millet genotypes and location. Similarly, a change in physiological traits has been demonstrated to be triggered by both genetic and environmental conditions (Anjum et al., 2011; Mukami et al., 2019). Reduced photosynthetic rate and chlorophyll content have been widely associated with soil moisture deficits. Drought influences nutrient uptake such as nitrogen, which affects chlorophyll content that regulates photosynthetic activities (Fathi and Tari, 2016). The reduction in chlorophyll content is a mechanism that responds to drought in order to reduce the light absorbed by chloroplasts (Gu et al., 2017).

Root and shoot biomass accumulation has been used as an indicator of drought tolerance. Genotypes allocate biomass differently between roots and shoots (Weiner, 2004); and there are indications that drought tolerance can be improved via traits, such as root length, shoot and root biomass accumulation (Paustian et al., 2016; Griffiths and Paul, 2017). Drought-tolerant genotypes have been reported to have higher root dry matter per unit of leaf area, signalling that they would invest more in deeper rooting for water absorption. Increased root biomass has also been linked to drought avoidance in which plants accumulate more root biomass compared to above ground biomass (Zhou et al., 2018). Root biomass is directly associated with the root length and number of root hairs, which are important for increased water uptake. Therefore, increased root biomass displayed among the selected finger millet genotypes in this study can be attributed to an increase in one, or combinations of, these root system components.

These results agree with those of Chen et al. (2020), who established that biomass allocation pattern influences drought tolerance in wheat. Plants that invest significantly in root biomass increase their potential for water and nutrient absorption, which directly influences their growth potential (Wasaya et al., 2018). Large root biomass is important in dryland farming conditions where crops have to explore large volumes of soil to extract enough moisture for growth (Ehdaie et al., 2012).

Changes in stomata conductance cause changes in leaf water potential by changing the transpiration rate. High photosynthesis and stomatal conductance among the evaluated finger millet genotypes indicated that photosynthetic CO₂ fixation in the genotypes was not affected by a water stress condition across the two locations. Furthermore, the high photosynthesis and stomatal conductance exhibited by these genotypes can be an indicator for improved water use efficiency. In a related study by Chen and Hao (2015), transpiration rate, stomatal conductance and water use efficiency (WUE) were found to have no correlation with grain yield in wheat. In contrast, Sharma et al. (2015) observed a positive correlation between water use efficiency and grain yield in pearl millet. Photosynthetic and transpiration rates, which depend on stomatal conductance, have been widely proven to be regulated by soil moisture levels (Frih et al., 2021).

The leaf area index showed a significant positive relationship with shoot biomass as well as grain yield. Reduced leaf area resulted in a lower shoot biomass and grain yield in all the genotypes. Low light intensity reduces the leaf expansion rates and delays the complete expansion of a leaf; thus, leaf area per plant is decreased under shade conditions (Fan et al., 2018). In the present study, the leaf area was reduced under low light intensities, which might be due to higher allocation of biomass towards stem elongation than to leaf expansion (Wu et al., 2017). Furthermore, low light intensity reduces the photosynthesis rate, slowing down other physiological processes in plants (Anjum et al., 2020).

Morphological and physiological traits have been widely used in the screening and selection for drought among the cultivated crops. Morphological traits (such as the number of root hairs, lateral roots, root volume, root length, root density and root surface area) have been directly linked to higher water uptake from water-deficient soils. Deep and proliferate root systems avoid drought stress due to their ability to acquire more water from deeper soil horizons.

Improved water uptake is considered a key strategy towards drought tolerance in crops. Therefore, the development and distribution of root systems can be regarded as key factors for more efficient water uptake; and thereby is a means for managing the performance of finger millet under drought stress. Physiological traits as well have been widely exploited for drought tolerance. Results of this study revealed the genotypic and environmental differences for the physiological traits assessed. The wide variability that existed among the finger millet genotypes and in locational differences could be used to generate important information towards selection for drought tolerance among the evaluated finger millet genotypes. Finger millet lines ICFX 1420314-2-1-1-1 (7), KNE 814 X Ex Alupe (P) P8-1-1-1-1 (24) and ICFX 1420415-3-1-1-2 (14) proved to be consistent for better morpho-physiological traits across the two locations. Therefore, the named finger millet lines can be considered for further evaluations and breeding programs towards drought tolerance.

The authors have not declared any conflicts of interests.

The authors are grateful for Centre of Excellence in Sustainable Agriculture and Agribusiness Management (CESAAM) project, Egerton University for funding the study. They also acknowledge Egerton University for providing institutional support.