Response of eight sorghum varieties to plant density and nitrogen fertilization in the Sudano-Sahelian zone in Mali

1 Centre d’Etude Régional pour l’Amélioration de l’Adaptation à la Sécheresse (CERAAS), Institut Sénégalais de Recherches Agricoles (ISRA), Route de Khombole, BP 3320, Thiès, Sénégal. 2 Institut d’Economie Rurale (IER), LABOSEP de Sotuba, BP 262, Bamako, Mali. 3 Laboratoire Campus de Biotechnologies Végétales, Département de Biologie Végétale, Faculté des Sciences et Techniques, Université Cheikh Anta Diop (UCAD), BP 5005, Code postal 10700, Dakar-Fann, Dakar, Sénégal. 4 CIRAD, UMR AGAP, BP 1813, Bamako, Mali. 5 CIRAD, UMR AGAP, BP 3320, Thiès, Sénégal. 6 AGAP, University of Montpellier, CIRAD, INRA, Montpellier SupAgro, Montpellier, France.


INTRODUCTION
Sorghum is one of the staple cereals grown in the semi-arid and arid regions of Africa and Asia (Srinivas et al., 2009;Borrell et al., 2014). It ranks fifth in the world in terms of production and growing area and the fourth most cultivated cereal in Mali during the raining season for human consumption and animal feeding. Despite its importance, sorghum yield remains low with less than one ton per hectare at national scale (Trouche et al., 2001). This low yield is mainly attributed to spatial and temporal variability in rainfall, poor soil fertility and extensive traditional agronomic management practices (Trouche et al., 2001;Leibman et al., 2014) . . Until now, to meet the food demand of the growing population, increase in production has been mainly achieved by expanding the areas dedicated to crop cultivation (Hanak-Freud, 2000). This strategy is limited by urbanization and the saturation of the rural space leading the farmers to use intensification method (Brocke et al., 2002;Xie et al., 2019). In addition, sorghum cultivation is highly competitive by potentially productive maize in areas of intensification in Mali Vaksmann et al., 2008). Presently, it is well documented that grain yield depends both on crop genetic potential and agronomic practices such as plant density and mineral fertilization (Moosavi et al., 2013;Kondombo et al., 2017). Numerous studies have shown the importance of plant density and nitrogen fertilization on sorghum production (Bayu et al., 2005;Akmal et al., 2010;Arunakumari and Rekha, 2016). In addition, previous studies reported that the optimum plant density depends on each crop (Biswas and Ahmed, 2014), beyond which the competition between plants for light, water and nutrients becomes important and can lead to decreased crop yields (Berenguer and Faci, 2001;Çalifikan et al., 2007;Li et al., 2016). Nitrogen is also one of the most important nutrients which must be used in an optimal quantity depending on plant density as its lack or excess can reduce crop productions (Fischera and Wilsonb, 1975;Ferraris and Charles, 1986;Tajul et al., 2013;Sher et al., 2016). Recently in Mali, to increase sorghum production, mineral fertilization studies were experimented and diffused on sorghum varieties (Kouyate and Wendt, 1991;Zougmoré et al., 2003;Coulibaly et al., 2019). However, plant density of 25.000 hills/ha and the doses of 100 kg ha -1 diammonium phosphate (DAP) at sowing and 50 kg ha -1 urea before the panicle initiation were recommended for sorghum cultivation (Kouyate and Wendt, 1991;Coulibaly et al., 2019). These agronomic practices advised in the growing areas were disseminated separately either on local sorghum or on improved sorghum. Nowadays, little research has been done to understand the performance of newly improved sorghum varieties to respond to plant density and nitrogen fertilization to intensify grain and straw production. A better knowledge on the effect of these techniques on sorghum productivity (grain and straw) should contribute to a better understanding of the constraints related to sorghum intensification in the Sahel. The objective of this study was to identify sorghum varieties that respond to plant density and nitrogen fertilization, and to determine agro-morphological and physiological traits involved in the variation in plant density and nitrogen fertilization.

Experimental design and crop management
A split-split-plot design was used to study three factors including two plant densities (D1: 26666 plants ha -1 and D2: 53333 plants ha -1 ) as the main plot, three nitrogen levels (0, 41 and 82 kg ha -1 ) as the subplot and varieties as the sub-subplot with three replications. The dose of nitrogen recommended in Mali for the sorghum cultivation is 41 kg ha -1 (Kante et al., 2017). A total of 144 treatments were used. The area for each elementary plot was 18 m 2 (6 lines 4.5 m long and 4 m wide). The seeding spacing was 0.75 m x 0.5 m for the low density (D1) and 0.75 m x 0.25 m for the high density (D2). The soil was ploughed to a depth of about 30 cm. Sowing was done on June 18 th , 2018 and on July 5 th , 2019 after a good rainfall (25 mm) at a rate of 5 to 6 seeds per hill. Around 15 days after emergence, the plots were thinned to one plant per hole in wet condition. Nitrogen was applied in the form of urea in two fractions, three weeks after thinning (50%) of the plants and before   the panicle initiation (50%). Basal application of Phosphorus (46 kg P 2 O 5 ) was homogeneously made in all plots before sowing as "phosphate naturel de Tilemsi" (PNT) granulated (31% P 2 O 5 ). Two manual weeding were carried out in each experiment and crop ridging was performed after the second nitrogen application. Breakouts were realized between blocks and between replications to reduce nitrogen exchange. All plots were treated with EMACOT 050 WG insecticide in the vegetative phase of the crops according to the infestation level to control the attacks of legionary caterpillars (Spodoptera exempta).

Measurements and sampling
The phyllochron was calculated according to the ratio of number of days necessary for the appearance of the flag leaf to the total number of leaves that appeared on the main stem. Its measurement was carried out on three plants randomly selected from each elemental plot. Physiological measurements were realized at flowering and carried on the leaf area index and chlorophyll estimation. The leaf area index was estimated with a Sunscan Septometer (Delta-T Device Ltd) equipped with an external BF5 sensor on plots (1 m 2 ) delimited in each elementary plot. The chlorophyll estimation was done with a SPAD-502 device. It was carried on the 3 rd ligulated leaf of the main stalk from the top of the plant and was performed on the three plants used for the phyllochron determination. An area of 2.25 m² was used for plant height measurements, yield components, straw biomass and grain yield at physiological maturity. The seedlings of 6 hills for the low density and 12 hills for the high density were collected. Plant height was measured with a ruler and the panicule number per m² was assessed by manual counting. After harvesting (panicle and dry straw) and sun drying for one month, panicle weight per m² and straw yield were determined.
The dried panicles were threshed to determine grain yield. The 1000-grains weight was obtained by counting with an electric counter (NUMIGRAL) and the grain numbers per panicle was also calculated according to the ratio of average panicle weight to grain weight.

Data statistical analysis
The combined variance analysis of the two trials was performed with the "agricolae" package for the environment (Venables, Smith and the R Core Team, 2019) according to the split-split-plot model developed by Carmer et al. (1989). Shapiro-Wilk normality and Bartlett homogeneity tests were performed to identify and exclude aberrant data induced by soil heterogeneity for different measured variables. The Tukey test (smallest significant difference, LSD) at the 5% threshold was used to compare the means of the studied factors. Pearson correlation analyses were performed with the "Hmisc" package and the principal component analyses (PCAbiplot) were performed with the "FactoMineR" and "factoextra" packages in the same R software.

Grain yield and straw yield
The analysis of variance showed significant effects of the year, density, nitrogen dosage and variety on grain and straw yields. The interaction effects between year x density and year x nitrogen were also significant on grain and straw yields; and the year x variety interaction was significant for straw yield ( Table 2). The interaction effects between nitrogen x variety and density x nitrogen x variety were likewise significant on grain yield and straw yield.

Agronomical and physiological parameters
The analysis of variance showed significant effects of the year, density, nitrogen and variety on the measured parameters ( Table 3). The density x variety and nitrogen x variety interactions effects were significant on panicle number m-², panicle weight m -² and plant height. The interaction effect between nitrogen x variety was also significant on SPAD value.
The panicle number m-² increased significantly with the increasing in plant density and nitrogen application. FADDA and TIEBILE varieties obtained on average 6.81 panicle m -² in D2 (53 333 plants ha -1 ). In addition, FADDA variety produced 7.41 panicle m -² in N2 (178 kg Nha -1 ) compare to the other varieties. FADDA variety obtained a panicle weight m -² of 675 g in D2 and 7.94 g in N2 than the other varieties. FADDA statistically had the same panicle weight m -² as PABLO and A12-79 in D2. For 1000WG, PABLO and TIEBIELE varieties recorded on average 24 g and the A12-79 variety was the least performing. The D1 density produced 457 more grains per panicle than the D2 density. It varied from 2803 grains in N0 to 3165 grains in N1 (89 kg N ha -1 ) (statistically equal to N2). A12-79 variety had significantly the highest grain per panicle (3475). For plant height, PABLO variety was the longest in D2 (437 cm). It varied from N0 (269 cm) to N1 (279 cm) (statistically identical to N2).
The leaf area index value was 2.33 and 2.99 respectively in D1 and D2. Nitrogen input increased the vegetation cover from 2.49 in N0 to 2.89 in N2. FADDA had significantly the highest leaf area index (2.99) and A19-79 obtained lower value of 2.27. PABLO and FADDA varieties performed with SPAD values of 49.9 and 49.7 respectively in N2. It was 44.77 in D1 and 43.45 in D2. Phyllochron ranged from 3.01 days in D1 to 3.20 days in D2. Nitrogen application shortened phyllochron from 0.20 days in N2 to 0.08 days in N1. FADDA and PABLO varieties recorded a short phyllochron with an average of 2.96 days than the C2_007-03 and TIEBILE varieties, which averaged 3.23 days.

Contribution of variables to grain and straw production in sorghum
To better understand the variable contributions to the increase in grain yield and straw yield, a correlation   matrix was carried out on the average of N0D1 (Table 4) and N2D2 (Table 5) treatments. Grain yield was significantly and positively correlated with the leaf area index, grain number per panicle, panicle number per m 2 , panicle weight per m 2 and straw yield in N0D1 and N2D2; and negatively correlated to phyllochron in N2D2. Straw yield was significantly and positively correlated to leaf area index, grain number per panicle, panicle number per m 2 and panicle weight per m 2 in N0D1 and N2D2. It was negatively correlated with plant height in N0D1 and
positively correlated with plant height in N2D2. The variables correlated to grain and straw yields will be used in variety Characterization.

Characterization of eight varieties for the traits studied
Principal Component Analysis (PCA-biplot) based on eight sorghum varieties in treatments N0D1 ( Figure 3A) and N2D2 was conducted. In N0D1, the ACP-biplot shows three homogeneous groups. Dimensions 1 and 2 explain respectively 47.6% and 29.9% of total variation. Group 1 includes SOUMBA, FADDA and A12-79 varieties. It is characterized by less important variables. Group 2 is determined by a plant height and 1000-grain weight and is composed of TIEBILE and PABLO. Group 3 involves the varieties of type caudatum GRINKAN, C2_075-15 and C2_007-03 and is characterized by grain yield, straw yield, leaf area index, weight per m 2 and Grain number per panicle raised. In N2D2, the ACP-biplot indicates four homogeneous groups. Dimensions 1 and 2 explain respectively 49.2% and 26.6% of total variability ( Figure 3B). Group 1, which includes the GRINKAN, SOUMBA and C2_007-03 varieties is characterized by a long phyllochron. Group 2, which involves the TIEBILE and PABLO varieties is defined by a large 1000-grain weight and plant height. Varieties of Group 3 are characterized by high grain number per panicle and consist of C2_075-15 and A12-79. Group 4, which is a single FADDA variety of guinea type is characterized by better grain yield, straw yield, leaf area index, panicle number per m 2 and panicle weight per m 2 .

DISCUSSION
The study on the performance of eight sorghum varieties at different plant densities and nitrogen fertilization enabled an understanding of the effect of intensification factors on sorghum grain and straw production in the Sudan-Sahelian zone in Mali. The response of the studied factors during the two-year trials may be due to rainfall distribution (Turgut et al., 2005;Oikeh et al., 2009) and soil heterogeneity. This could explain the decrease in grain yield (by 10%) and straw yield (by 72%) in the first trial as compared to the second. In this study, grain and straw yields increased with increasing plant density from D1 ((26666 plants ha -1 )) to D2 (53 333 plants ha -1 ) for all varieties tested. Nitrogen application also increased grain and straw yields from N0 (0 kg N ha -1 ) to N2 (178 kg N ha -1 ). Our results are similar to those reported by Moosavi et al. (2013) and Shrestha et al. (2018). However, the results also showed that grain yield and straw yield varied for all varieties at different plant density and nitrogen fertilization combinations. They also increased from N0D1 ( Figure 2B) and straw yield. These results suggest that response of varieties to different plant densities and nitrogen fertilization for grain and straw yields is highly variable and could be genetic. This shows that each variety or group of varieties needs an optimum nitrogen level and plant density to produce maximum grain and straw. Our results are consistent with studies conducted by Zhou et al. (2019). Shahrajabian et al. (2011) and Soleymani et al. (2011) also confirmed these findings in their study on sorghum. Grain yield depends on the variety and growing conditions, in particular plant density and nitrogen fertilization. Its improvement depends on its components but also on physiological and growth traits. In this study, the panicle weight per m 2 , leaf area index, straw yield and grain number per panicle parameters were most expressed in N0D1 (Table 4) and N2D2 (Table 5). These results clearly show that through these traits it is possible to increase grain yield under less intensive (N0D1) and intensive (N2D2) conditions. Researches conducted by Ogunlela and Okoh (1989), Buah et al. (2009), andAjeigbe et al. (2018) reported similar results. But in N2D2, grain yield was strongly explained by panicle number per m 2 . Moosavi et al. (2013) believed that at high plant density, emphasis should be placed on panicle number per m 2 , because at high plant density, the grain number per panicle decreases even if grain yield per unit area increases. Straw yield in N0D1 and N2D2 was positively explained by leaf area index and panicle weight m -². There was also a positive correlation between straw yield, plant height and panicle number per m 2 in N2D2, but it was positively correlated by grain number per panicle in N0D1. This finding show that a selection made in favor of these traits can help us increase production of sorghum straw. According to Sahu et al. (2018), nitrogen application increases plant height and leaf area index in sorghum, which would be involved in increasing straw yield.
The variability observed under N0D1 and N2D2 treatments enabled the classification of eight varieties according to the traits studied. In N0D1, GRINKAN and C2_075-15 and C2_007-03 caudatum varieties (short size) produce the highest grain yield and are characterized by panicle weight per m 2 , leaf area index, straw yield and grain number per panicle. In N2D2, FADDA (large size) guinea hybrid variety is the most performing and is characterized by grain yield, straw yield, leaf area index, panicle number m -² and panicle weight per m 2 . One of the specificities of this variety is its capacity to valorize nutrients and to develop an important tillering, a trait probably inherited from its parent Lata3. This would explain an increase in his traits in FADDA in N2D2. According to Lafarge et al. (2002) and Zand and Shakiba (2013), tillering is an important trait that leads to increased grain and straw yields in sorghum. In general, this trait has not been of much interest to the sorghum selection programs. However, it should now be one of the priorities of breeding programs to develop productive varieties with large tillering to intensify the crop sorghum.

Conclusion
This study showed that plant density and nitrogen fertilization on sorghum varieties significantly increased grain yield and straw yield. Grain yield in N0D1 and N2D2 was associated with panicle weight per m 2 , leaf area index, straw yield, panicle number per m 2 and grain number per panicle. GRINKAN, C2_075-15 and C2_007-03 varieties produced maximum production of grains and straws in N0D1 (0 kg N ha -1 and 26666 plants ha -1 ). These caudatum-type varieties may be recommended in less intensive sorghum production areas in Mali. FADDA variety produced the highest of grains and straws in N2D2 (178 kg N ha -1 and 53 333 plants ha -1 ). Indeed, FADDA being a guinea-type hybrid variety could be recommended to the farmers for grain and fodder production because it is the variety that was better adapted to intensification.