RELATIONSHIPS AMONG WATER USE EFFICIENCY AND THE PHYSIO-AGRONOMIC TRAITS IN DURUM WHEAT (TRITICUM DURUM DESF.) CULTIVARS ASSESSED UNDER RAINFED CONDITIONS OF THE EASTERN HIGH PLATEAUS OF ALGERIA

Genetic advances in grain yield under rainfed conditions haves been low, slowed by genotype x environment interaction arising from unpredictable rainfall in drought prone areas. A good understanding of factors regulating yield provides the opportunity to identify and select for physiological and agronomic traits that increase both water use efficiency and grain yield under rainfed conditions. The results of this investigation exhibited large variation for physiological and agronomic traits among varieties and cropping seasons. Modern varieties had high harvest index, grain yield, and leaf chlorophyll content, low leaf relative water content, and were shorter than varieties derived from land races. Total dry matter and specific leaf area differences, among groups of varieties, were not significant. Water use efficiency for total dry matter showed no significant correlations with the measured physiological and agronomic traits, while water use efficiency for grain yield was significantly correlated with harvest index, plant height and to a lesser extent with leaf chlorophyll content. Path analysis, based on phenotypic correlations, showed the consistent direct and indirect effects of harvest index and to a lesser extent those of plant height. Selecting for plant height and harvest index could improve both water use efficiency and grain yield under drought prone environments.


INTRODUCTION
Durum wheat cultivation, in Algeria, is practiced in a fallow-wheat rotation, relying on stored water during the fallow period, in addition to the cropping season's rainfall.Annual precipitations, inherently low in amount, varied quantitatively and qualitatively, mainly on the high plateaus area, where nearly 70% are receipted during the cold winter months.Under such growing conditions, the occurrence of intermittent drought stress limits grain yield and renders water scarcity as the most penalizing production factor (Chennafi et al., 2006).The high plateaus area belongs to a vast geographical region where agriculture has been forecast to be at greater risk due to an increase in the frequency and severity of drought episodes (Sahnoune et al., 2013).Selection of drought tolerant cultivars is sought to minimize the effects of water scarcity and to sustain crop production.The release of improved cultivars requiring lower amounts of water per unit yield and characterized by high yield potential is essential for more sustainable agricultural practices, particularly in rainfed, drought prone areas.Water conserving breeding strategy could combine high yield, high WUE and good drought resistance traits in one variety (Zhang et al., 2004).Water use efficiency (WUE) is seen as an important determinant of yield under stress and as a component of crop drought resistance (Ehdaie, 1995 ;Kirda et al., 1999).This trait remains among the most appropriate strategies to cope with drought stress under rainfed conditions.Several studies have shown that selection based on this trait improved grain yield potential (Rebetzke et al., 2002, Franks et al., 2015).Zhang et al. (2005) reported that grain yield improved by 50%, resulting in significant WUE increases.Studies are needed to focus on plant traits that are beneficial to both grain yield and WUE improvement.
Besides crop husbandry, numerous plant characteristics are reported to affect WUE and grain yield (GY).In fact GY and WUE, due to their close association with harvest index (HI), could be improved by manipulating this trait (Ehdaie and Waines, 1993;Zhang et al., 2008).Siddique et al. (1990) reported that WUE of modern cultivars was higher than old cultivars among Australian tested wheat varieties, because of significant changes in plant stature and crop cycle duration, leading to improved HI and stress escaping.Slafer and Araus (1998) reported that the improved crop performance may be achieved by improvements in water use (WU), WUE and HI.Several plant traits such as chlorophyll content, osmotic adjustment, relative water content, translocation of stem stored carbohydrate, stay green, early seedling vigor, earliness, canopy temperature, carbon isotopic discrimination, coleoptile length, stem and leaf waxiness, leaf and root architecture as well as the amount of soil moisture available to the crop and its partitioning between evaporation and transpiration are reported to related to WUE and GY (Quin et al., 2013;Richard et al., 2015;Farjam et al., 2015;Nakhforoosh et al., 2016;Christy et al., 2018;Rashid et al., 2018;Abdolahi et al,2018).The present investigation aimed to analyze the association between some physio-agronomic traits and WUE in eight durum wheat (Triticum durum Desf.) varieties, belonging to two different eras, evaluated under semi-arid conditions during three cropping seasons.

Plant material and experimental design
The experiment was carried out at the Field Crop Institute-Agricultural Experimental Station of Setif (ITGC-AES, 36 °12 ' N and 05°24 ' E, 1080 masl, Algeria), under rainfed conditions during three growing seasons (2013/14-2015/16).Eight durum wheat varieties were evaluated (Table 1).Waha and Gaviota durum are selections from Cimmyt-Icarda joint durum wheat breeding program.Simeto is an Italian cultivar while Megress is an ITGC-AES Setif selection.These varieties proved to be well adapted to the Setif region and are classified as early-heading genotypes (Haddad et al., 2016).Mohamed Ben Bachir (MBB), Hedba 3 , Guemgoum Rkhem, and Oued Zenati 368 are old varieties selected from land races.MBB is selected from a land race native to the Setif region.Hedba 3 , alias Pelissier, is a drought tolerant cultivar.Guemgoum Rkhem is native from Tiaret region (Western Algeria), while Oued Zenati 368 is a selection from a population native to the Guelma region (Eastern Algeria).Varieties derived from landraces are taller and late maturing compared to recently released ones (Nouar et al. 2012).Recommended cultural practices for the area were followed to raise a good crop.Monoammonium phosphate (52% P 2 O 5 + 12% N) with 80 kg ha −1 was applied just before sowing and 80 kg ha −1 of urea (46%) were broadcasted at the tillering stage.Weeds were controlled chemically by application of 150 g ha −1 of Zoom [Dicamba 66% Triasulfuron 4%] and 1.2 L ha −1 of Traxos [22.5 g/l de Pinoxaden, 22.5 g/l Clodinafop-propargyl, 6.5g/l de Cloquintocet-méxyl] herbicides.

Measurements
At the heading stage, leaf relative water content (LRWC), leaf chlorophyll content (LCHC) and specific leaf area (SLA) were measured.LRWC was determined by the method of Barrs and Weartherly (1962) described by Pask et al., (2012).Four leaves were sampled per plot and immediately weighed to obtain the fresh weight.Leaf samples were then placed in test tubes containing distilled water, and let to stand for four hours, under dim light at laboratory ambient temperature.Leaf samples were then reweighed to obtain the leaf turgid weight.Leaf samples were then oven dried at 80°C for 48 h for leaf dry weight determination.The LRWC was calculated according to the following formulae reported by Pask et al., (2012): where FW is the sample fresh weight, TW is the sample turgid weight, and DW is the sample dry weight.SPAD chlorophyll meter (Minolta SPAD-502 meter, Tokyo, Japan) was used to estimate leaf chlorophyll content.Three readings were taken per leaf from a sample of five fully expanded flag leaves per plot.Readings were averaged to get the plot mean SPAD value.The same leaf samples were used to estimate the specific leaf area, which was measured with an image scanner software (Mesurim pro, version 3.4).Leaf dry weight (LDW) was determined after oven-drying at 80 °C for 48 hours.SLA, derived as leaf area (LA) per unit leaf dry weight (cm 2 .g -1 ), was calculated using the following formulae reported by Rashid et al., (2018): ( ²) ( )

LA cm SLA LDW g =
At crop maturity, 2-row segments, 2 m long, were sampled per plot to estimate plant height, measured from ground level to the tip of the terminal spikelet, awns excluded; total dry matter, grain yield, and harvest index, derived as the ratio of grain yield over total dry matter yield.The amount of water evaporated and that transpired by each variety during the cropping cycle (water used =WU) was determined as the sum of the soil moisture available at seeding minus soil moisture available at harvest, plus the accumulated rainfall, from seeding to harvest.Soil available moisture (ASM, mm), at sowing and at harvest was deduced by the following formulae: ASM (mm) = [(H%-WP) x h x ρb]/100, where H% = 100(wet soil weight-dry soil weight)/dry soil weight, WP =wilting point =12%, average of the soil of the experimental site, h = soil profile depth in mm(600 mm), and ρb = bulk density = 1.23 (Chennafi et al. 2011;Belagrouz et al.,2016).Water use efficiency for total dry matter (WUE TDM, kg ha -1 mm -1 ) and grain yield (WUE GY, kg ha -1 mm -1 ) were derived according to Cheikh M 'hamed et al., (2015) as follow: Where TDM= total dry matter (kg ha -1 ) and GY=grain yield (kg ha -1 ).

Data analysis
Collected data were subjected to a combined analysis of variance using balanced anova subroutine implemented in Cropstat software package (Cropstat, 2007).Years, replications within years, and genotype by year interaction effects were considered as random and genotype effect was considered as fixed.Year main effect was tested against the replication hierarchized within years, while the genotype main effect was tested against the interaction which was tested against the residual.Mean comparisons were performed using the Fisher's protected least significant difference test at 5% probability level.Relationships among the measured traits were computed using Pearson's simple correlation test implemented in Past software (Hammer et al., 2001).Path coefficient analysis was performed to divide the correlation coefficient between WUE and the physio-agronomic traits (r iy ) into direct (p iy ) and indirect effects (rij pjy) according to the following equation reported by Garcıa del Moral et al., (2003): . iy iy ij jy r P r P = +

1.Physiological characteristics
The combined analysis of variance indicated significant year main effect for leaf chlorophyll content and leaf relative water content, but not for specific leaf area.Genotype main effect was significant only for leaf chlorophyll content, while the genotype x year interaction was significant for the three measured physiological traits (Table 2).The significant interaction indicated that ranking order of the varieties changed between years suggesting that differences existed for the same trait between varieties within year and varied significantly also for the same variety among years.Differences among extreme mean values were statistically significant as indicated by the significant genotype x cropping season interaction (Tables 2 and  3).Differences among cropping seasons (average over varieties) and among varieties (average over cropping seasons) main effects were not statistically significant for specific leaf area, whose mean values ranged from 8.5 to 9.6 cm² g -1 , among cropping seasons and from 7.1 to 11.2 cm² g -1 among varieties main effect.Per cropping season, GMG, in 2013/14, (10.8 cm² g -1 ), SMT, in 2014/15, (83.8%), and MGS, in 2015/16, (77.7%), expressed the lowest leaf relative water content.Meanwhile MBB, in 2013/14, (78.6%),GTA, in 2014/15, (97.2%), and H3, 215/16, (91.3%), showed the highest leaf relative water content mean values.Differences among extreme varieties mean values were statistically significant as indicated by the significant genotype x cropping season interaction (Tables 2 and  3).These results indicated that the expression of the physiological traits was strongly affected by the environment and to a lesser extent by the genotype.

2.Agronomic performances
The combined analysis of variance indicated significant year main effect for the four measured agronomic traits.Plant height, grain yield and harvest index showed significant genotype main effect.The genotype x year interaction was significant for the four measured agronomic traits (Table 2).The 2015/16 cropping season was the most favorable environment for the expression of the potential of plant height, total dry matter and grain yield.The less favorable environment for the expression of these traits were the 2014/15 for plant height and the 2013/14 for both grain yield and total dry matter.Plant height was reduced from the favorable to less favorable environments by 23.4 cm which represents 29.3% of plant height mean value recorded under favorable environment (Table 4).Total dry matter and grain yield were reduced by 56.6 and 17.5 q ha -1 , respectively, which represents 59.8 and 57.4 % of the mean values recorded under favorable environment for total dry matter and grain yield (Table 4).The best mean value of harvest index (34.3%) was expressed under the 2013/14 cropping season, which was less favorable to the expression of grain yield and total dry matter.The lowest harvest index mean value (22.9%) was recorded in 2014/15 cropping season.These results suggested that the measured respectively (Table 4).
Even though the small set of varieties assessed, the results showed the presence of variability for all the measured traits.Globally, newly released varieties were shorter, high grain yielding and allocating more dry matter to the grain than old varieties.Difference in terms of total dry matter produced was not significant.This corroborated results of Waddington et al., (1987) whom mentioned that increases in HI have accounted, in many instances, for the grain yield improvement in wheat since new high-yielding wheat varieties have higher HI than older ones.Samarrai et al. (1987) reported that HI is influenced by environment, as the results of the present study suggested.PHT = Plant height,(cm) TDM= Total dry matter,(q ha -1) GY= Grain yield, ,(q ha -1) , HI = Harvest index, (%) WUE TDM = Water use efficiency for total dry matter, (kg ha -1 mm -1 ) WUEGY= Water use efficiency for grain yield (kg ha -1 mm -1 ).GMG =Guemgoum Rkhem, OZ 368 = Oued Zenati 368, H 3 = Hedba3, MBB= Mohammed ben Bachir, SMT= Simeto, WAH= Waha, GTA= Gaviota, MGS= Megress, LSD5%= Least significant difference at the 5% probability level.

3.Water use efficiency
Total rainfall, accumulated from sowing to harvest, reached 251.9, 299.4 and 237.7 mm in 2013/2014, 2014/2015 and 2015/16, respectively.Compared to the long term average of 321.2 mm reported by Mekhlouf et al., (2006), these figures appeared to be very low, mainly during the 2013/14 and 2015/16 cropping seasons, suggesting a strong drought stress effect during the course of the experiment.At sowing, soil relative humidity, in the 600 mm profile, reached 19.0, 18.3 and 18.6%, in 2013/14, 2014/15 and 2015/16 cropping seasons, respectively.These figures are the equivalents of 51.6, 46.7 and 48.9 mm soil moisture available to the plant.This soil moisture resulted from early autumn rain showers and from moisture stored during the fallow season.Soil relative humidity, measured at harvest, was below the wilting point and thus the available moisture left in the soil was assumed to be nil.Water available for use (evapotranspiration) by the crop during the growing cycle reached 303.6, 346.1 and 286.7 mm, in 2013/14, 2014/15 and 2015/16 cropping seasons, respectively.Table 5. Mean values of water use efficiency for total dry matter and for grain yield, averaged over years (variety main effect), averaged over varieties (year main effect), variety mean value per cropping season (year) and the least significant difference at 5% probability level.1.6 0.4 WUE TDM = Water use efficiency for total dry matter, (kg ha -1 mm -1 ) WUEGY= Water use efficiency for grain yield (kg ha -1 mm -1 ).GMG =Guemgoum Rkhem, OZ 368 = Oued Zenati 368, H 3 = Hedba3, MBB= Mohammed ben Bachir, SMT= Simeto, WAH= Waha, GTA= Gaviota, MGS= Megress, LSD5%= Least significant difference at the 5% probability level.

4.Relationships between WUE and the physio-agronomic traits
The correlation coefficients relating WUE TDM to the measured physioagronomic traits were statistically no significant, except the correlation coefficient between LCHC and WUE TDM , measured in 2015/16, which reached significance and had a negative sign (Table 6).-0.435 PHT = Plant height,(cm), HI = Harvest index, (%) WUE TDM = Water use efficiency for total dry matter, (kg ha -1 mm -1 ) WUEGY= Water use efficiency for grain yield (kg ha -1 mm -1 ), LCHC= Leaf chlorophyll content, LRWC= Leaf relative water content, SLA = Specific leaf area, r 5% = 0.666.These results suggested that, among the measured physio-agronomic traits, no one could be able to predict WUE TDM , and to be used as selection criterion for screening purposes.Correlation coefficients of SLA and TDM with WUE GY were non-significant, suggesting that these two traits were of little value for WUE GY prediction.Results about SLA didn't supported findings of van den Boogaard et al. (1997) whom studied wheat plant growth and water-use efficiency and found that WUE was higher for plants with higher leaf area per unit plant weight.Richards et al. (2002) suggested using specific leaf area as an indirect selection criterion for yield potential in wheat.Atta (2013) found that specific leaf area was negatively correlated with WUE and grain yield and suggested that selection against this trait may be effective in raising grain yield.The relationship between WUE GY and LRWC was unreliable, being dependent on the environment for its expression.However PHT and HI, and to a lesser extent LCHC were reproducible and significantly correlated with WUE GY .These traits appeared to be useful for WUE GY improvement (Table 6).In this context, Zhang et al., (2016) found no significant correlations between WUE GY and LCHC, or LRWC, but significant correlations were found between WUE GY and HI.Through multiple regression analysis Atta (2013) identified several key traits that contribute to improve WUE among which leaf traits, plant height, total dry matter at maturity, harvest index and grain yield which corroborated partially the results of this study.
Taking LCHC, LRWC, PHT and HI as causing traits and WUE GY as caused trait, path analysis indicated that direct and indirect effects were inconsistent and varied from one environment to another (Table 7).Hence LCHC exhibited a large positive direct effect (0.450) in 2016/14, which lessened in the second cropping season (0.200) then vanished (-0.080) in the third one.This trait acted indirectly via HI during the three cropping seasons (0.164, 0.152, 0.520), via LRWC, in one season (0.170) and via PHT during two seasons (0.196, 0.315).The positive sign of the direct and indirect effects of LCHC suggested that higher LCHC was desirable to improve WUE GY , either directly (but depending on the environment) or indirectly via HI and to lesser extend via PHT.Recently released cultivars expressed consistently high LCHC and HI compared to old ones which explain their observed high WUE GY (Table 3).Similarly LRWC exhibited a large direct effect (-0.475) associated to sizeable indirect effects via LCHC (-0.161) and HI (-0.196) in one season, and both direct and indirect effects vanished during the two other seasons (Table 7).High LRWC was expressed by local varieties which had lower HI and LCHC, but the effect of this trait were inconsistent depending on the environment.PHT expressed a direct effect variable which was lower than the consistent indirect effects via HI.Taller varieties tended to have low HI and WUE GY .HI expressed a consistent positive direct effect; the indirect effects, either via PHT or via LCHC and LRWC, were inconsistent (Table 7).

CONCLUSIONS
Experiment results revealed that modern cultivars are more efficient users of rain water than all others in semi-arid conditions, It is also revealed that those varieties, which use more water, produce hi harvest index value and give more grain yields.Our studies demonstrated that LCHC, LRWC, PHT and HI are more important traits linked to the WUEg in semi-arid regions.Thus, path analysis, based on phenotypic correlations between WUE GY and HI, PHT, LRWC, LCHC, showed the consistent direct and indirect effects of HI and to a lesser extent those of PHT.Selecting for PHT and HI could improve both WUE GY and grain yield under variable environments.The high WUE GY genotypes identified in the current study can be used to develop more efficient cultivars that increase grain yield per unit of water used, in drought prone areas.However, selection for high HI, to improve GY and WUE GY , will reduce plant height and biomass production under severe drought conditions.Conversely grain yield, at excessive crop height, can be reduced because of poor HI and increased lodging.It is suggested to select for tall, high-yielding plants within dwarf segregating populations.
*, ** = Significant effect at the 5 and 1% probability level, respectively; LCHC= Leaf chlorophyll content, LRWC= Leaf relative water content, SLA = Specific leaf area, PHT = Plant height, TDM= Total dry matter, GY= Grain yield, HI = Harvest index, WUE TDM = Water use efficiency for total dry matter, WUEGY= Water use efficiency for grain yield.Leaf

Table 4 .
Mean values of the four measured agronomic traits, averaged over years (variety main effect), averaged over varieties (year main effect), variety mean value per cropping season (year) and the least significant difference at 5% probability level.

Table 6 .
Simple correlation coefficients between water use efficiency for total dry and grain yield and physio-agronomic traits.

Table 7 .
Direct and indirect effects of the physio-agronomic traits on WUE GY .