Forage Yield, Canopy Characteristics, and Radiation Interception of Ten Alfalfa Varieties in an Arid Environment

An increasing demand for new and improved livestock forage products is driving the development of forage systems in arid regions. Our study evaluated the productivity of 10 alfalfa (Medicago sativa L.) varieties and the relationship between forage yield and canopy structure traits, such as plant height, branch number, stem/leaf ratio, and leaf area index in the arid Hexi Corridor, north-west China. Here, plant height, primary branch number per plant, and stem/leaf ratio were positively correlated with forage yield. In terms of a two-year total yield, Gannong No. 5 produced the highest value (13,923 kg ha−1), followed by the WL342HQ (12,409 kg ha−1), Phabulous (11,928 kg ha−1), and Xinjiang Daye (11,416 kg ha−1) varieties. Therefore, these four alfalfa varieties are suitable for large-scale cultivation in the Hexi Corridor region and other arid areas where the effect of precipitation is even larger than that of temperature. These results provide valuable information for the selection and cultivation of alfalfa varieties, which could improve forage yield and the production of livestock in arid regions.


Introduction
Arid and semi-arid environments cover about 30% of the world's total land area and 52% of that in China [1]. Additionally, the area of semi-arid land increased by 25% from 1986 to 2010 [2]. Livestock production is a major economic contributor in arid and semi-arid regions [3]. However, water is often considered to be the most important determinant that restricts forage yield in these environments [4]. Forage crops usually suffer from drought stress in these regions, even under irrigation, owing to high annual evaporation. Therefore, it is imperative to screen suitable forage crops and varieties which are more adaptable to water-limited conditions to improve livestock production sustainability in arid and semi-arid regions.
Alfalfa (Medicago sativa L.) is a widely planted, perennial forage legume in arid and semi-arid areas because of its high productivity, nutritional value, nitrogen fixation capacity, and adaptability to various environments [5,6]. The canopy structure has a significant influence on the forage yield of alfalfa owing to the optimization of light energy utilization and accumulation. Some traits that influence canopy development and productivity are plant height (PH), the stem to leaf ratio, branch number, leaf area index (LAI), and light interception (LI) [7,8]. Previously, a study estimated the PH of different alfalfa crop canopies, and established that PH is positively correlated with forage yield [9]. The branch number is a vital yield component of forage crops [10,11]. It exerts a strong positive impact on forage yield [12,13], and has been included in plant growth models to improve predictive
The forage yield in each cutting decreased with harvest time (Figure 3). In 2017 proportion of first cutting yield to annual yield of 10 varieties ranged from 42 to whereas in 2018, it ranged from 35 to 60%. In addition, the first cutting yield of D (3163 kg ha −1 ) was significantly the greatest among the 10 varieties, followed by Gann No. 5 (2883 kg ha −1 ), Sanditi (2748 kg ha −1 ), and Phabulous (2716 kg ha −1 ) (p < 0.001) in 2 The second cutting yield varied from 615 kg ha −1 in WL168HQ to 1633 kg ha −1 in Gann No. 5 (p < 0.001) in 2017. On average, the third and fourth cutting yields were equiva to 27% (619 kg ha −1 ) and 20% (457 kg ha −1 ) of the first cutting yield (2334 kg ha −1 ) of th The forage yield in each cutting decreased with harvest time (Figure 3). In 2017, the proportion of first cutting yield to annual yield of 10 varieties ranged from 42 to 67%, whereas in 2018, it ranged from 35 to 60%. In addition, the first cutting yield of Derby (3163 kg ha −1 ) was significantly the greatest among the 10 varieties, followed by Gannong No. 5 (2883 kg ha −1 ), Sanditi (2748 kg ha −1 ), and Phabulous (2716 kg ha −1 ) (p < 0.001) in 2017. The second cutting yield varied from 615 kg ha −1 in WL168HQ to 1633 kg ha −1 in Gannong No. 5 (p < 0.001) in 2017. On average, the third and fourth cutting yields were equivalent to 27% (619 kg ha −1 ) and 20% (457 kg ha −1 ) of the first cutting yield (2334 kg ha −1 ) of the 10 alfalfa varieties in 2017. In 2018, the top two yields of the first cutting were WL343HQ (5251 kg ha −1 ) and Xinjiang Daye (3648 kg ha −1 ) varieties, whereas the lowest value was observed with Gannong No. 3 (1694 kg ha −1 ) and Wudi (1657 kg ha −1 ) varieties (p < 0.001). Here, we observed that the second cutting yield of Xinjiang Daye (2318 kg ha −1 ) was higher than those of the other nine varieties, while the lowest values occurred with Sanditi (713 kg ha −1 ) (p < 0.001).
Here, we observed that the second cutting yield of Xinjiang Daye (2318 kg ha −1 ) was higher than those of the other nine varieties, while the lowest values occurred with Sanditi (713 kg ha −1 ) (p < 0.001).

Plant Height
Plant height at spring (PHS) and at flowering (PHF) were measured in this study. These parameters were found to be significantly affected by the varieties and the interaction between years and varieties (p < 0.001) ( Table 1). In 2017, the PHSs of Gannong No. 5 (26.6 cm) and Derby (25.3 cm) were significantly greater than those of other alfalfa varieties, whereas the lowest values were observed in WL168HQ (17.6 cm) and Wudi (17.8 cm). In 2018, the greatest and lowest PHSs occurred in Gannong No. 5 (29.2 cm) and Sanditi (16.8 cm), respectively. Overall, Phabulous and WL343HQ presented a significant increase in PHS from 2017 to 2018 (Table S1, p = 0.047, and =0.012, respectively). In contrast, Sanditi showed a significant decrease (p = 0.02).

Plant Height
Plant height at spring (PHS) and at flowering (PHF) were measured in this study. These parameters were found to be significantly affected by the varieties and the interaction between years and varieties (p < 0.001) ( Table 1). In 2017, the PHSs of Gannong No. 5 (26.6 cm) and Derby (25.3 cm) were significantly greater than those of other alfalfa varieties, whereas the lowest values were observed in WL168HQ (17.6 cm) and Wudi (17.8 cm). In 2018, the greatest and lowest PHSs occurred in Gannong No. 5 (29.2 cm) and Sanditi (16.8 cm), respectively. Overall, Phabulous and WL343HQ presented a significant increase in PHS from 2017 to 2018 (Table S1, p = 0.047, and =0.012, respectively). In contrast, Sanditi showed a significant decrease (p = 0.02).

Branch Number
As illustrated in Figure

Branch Number
As illustrated in Figure

Leaf Area Index and Light Interception
The LAI varied significantly among the 10 alfalfa varieties (p < 0.001) ( Figure 6).

Stem to Leaf Ratio
As shown in Figure 7, the stem to leaf ratio varied among the 10 varieties in 2017. The stem/leaf ratio of Gannong No. 3 (3.69) was significantly greater than those of the other eight alfalfa varieties, except Phabulous (p < 0.001). The lowest ratio was observed in WL168HQ (1.71) (p < 0.001).

Leaf Area Index and Light Interception
The LAI varied significantly among the 10 alfalfa varieties (p < 0.001) ( Figure 6). The

Leaf Area Index and Light Interception
The LAI varied significantly among the 10 alfalfa varieties (p < 0.001) ( Figure 6). The

Stem to Leaf Ratio
As shown in Figure 7, the stem to leaf ratio varied among the 10 varieties in 2017. The stem/leaf ratio of Gannong No. 3 (3.69) was significantly greater than those of the other eight alfalfa varieties, except Phabulous (p < 0.001). The lowest ratio was observed in WL168HQ (1.71) (p < 0.001).

Correlation Analysis for First Cutting Yield
Considering that forage yield was influenced by the morphological traits of the plants, we analyzed correlations between forage yield and the above-mentioned traits using data (sample size of 30 for each trait) in the same growing year (2017), which are summarized in the correlation coefficient matrix ( Table 2) and PCA biplot based on the correlation matrix (Figure 8). Before starting PCA, the applicability of PCA was tested using the Kaiser-Meyer-Olkin (KMO) and Barlett tests. The calculated results were KMO = 0.66 (>0.5) and Barlett test value = 0 (<0.05), indicating that the dataset was appropriate for a factor analysis.

Correlation Analysis for First Cutting Yield
Considering that forage yield was influenced by the morphological traits of the plants, we analyzed correlations between forage yield and the above-mentioned traits using data (sample size of 30 for each trait) in the same growing year (2017), which are summarized in the correlation coefficient matrix ( Table 2) and PCA biplot based on the correlation matrix ( Figure 8). Before starting PCA, the applicability of PCA was tested using the Kaiser-Meyer-Olkin (KMO) and Barlett tests. The calculated results were KMO = 0.66 (>0.5) and Barlett test value = 0 (<0.05), indicating that the dataset was appropriate for a factor analysis.    The first two principal components explained 70.8% of the first yield variation ( Figure 8). Thus, we focused on these two components to represent the PCA ordination biplot loadings. The angles between the vectors represent the correlations, where small angles (close to 0 • ) indicate high positive correlations; wide angles (close to 180 • ) indicate high negative correlations; and angles close to 90 • indicate a lack of correlation [38]. As illustrated in Figure 8 and Table 2, PCA and correlation analysis showed that the yield of first cutting had a significant positive association with PHS, PHF, and TB (p = 0.006, 0.002, and <0.001, respectively) in 2017. The total yield of 2017 was strongly related to PHS, PHF, and stem/leaf ratio (p < 0.001, =0.005, and <0.001, respectively). The two-year total yield was significant positive affected by PHF, LAI, and LI (p < 0.001, =0.002, and <0.001, respectively). Moreover, there was a highly significant positive relationship between LAI and LI (r = 0.8745, p < 0.001) ( Table 2). The stem/leaf ratio showed a significant positive association with PHF and PB (p < 0.001 and =0.027, respectively).

Discussion
Stem traits, including plant height (PH), branch number, and stem to leaf ratio, are essential indicators related to the growth and development of forage, and are associated with the distribution of solar radiation in the canopy of forage crops [39]. In this study, PH (including PHS and PHF) was shown to significantly affect forage yield, which is consistent with previous studies [9,20,40]. In addition, PH demonstrated a positive correlation to the central leaflet area and stem to leaf ratio, but no effects on LAI and LI.
Branch number is an important morphological characteristic that could enhance the forage yield in alfalfa grown in arid environments [16]. In this study, the results indicated that PB was significantly corelated with forage yield, whereas SB had no effect on yield. Additionally, SB expressed less internal variation among the 10 alfalfa varieties. These results demonstrated that PB plays a greater role in plant architecture construction, determining forage yield. These findings are consistent with those of previous studies [41]. Therefore, we suggest that PB should be viewed as a crucial indicator of high productivity varieties for plant breeding research in the future.
The PCA and correlation analysis results revealed that a higher stem/leaf ratio indicates a better yield performance. This could be attributed to a higher stem/leaf ratio accompanied by higher PH and PB. These results are consistent with those reported by other researchers [17,18,42].
The leaf is considered to be an important functional unit of a plant, contributing to its yield formation. Traditionally, LAI and LI demonstrated a positive correlation with forage yield [25]. In this study, LAI and LI had a significant positive correlation with two-year total yield. However, they did not affect the yield of first cutting and total yield from four cuttings in 2017. The first reason could be that the effects of leaflet density and number to the LAI and LI were more significant than the central leaflet size in an arid environment [43]. Thus, in arid environments, the positive yield effect of canopy structure traits should not be attributed simply to the leaf traits, such as LAI and LI; rather, it could also be due to its effects of stem traits, like PH and branch number, via an increase in the stem dry weight ratio, as proposed by Hakl (2021) [7] and Ta (2020) [8]. Secondly, forage yield is related to light interception that accumulates throughout the entire growth cycle. Therefore, only one LAI and LI measurement before harvest does not truly reflect the efficiency of light energy utilization. It is suggested that researchers consider the significance of dynamic and continuous measurements of LAI and LI in future experiments.
Choosing an alfalfa variety is arguably one of the most important decisions to make in an alfalfa operation. Alfalfa yield reveals the suitability of a given variety to arid regions. On average, the forage annual yield in the second harvest year increased by 30% (1326 kg ha −1 ) compared with that of the first harvest year. Similarly, a study showed that the forage yield of alfalfa reached peak performance in the second and third harvest years [44]. Therefore, the advantage of accumulation of perennial alfalfa biomass appeared in the second harvest year. From our perspective, we should not only consider the yield of first harvest year, but also pay close attention to long-term productivity.
As one of the most widely planted varieties in the Hexi Corridor, Xinjiang Daye showed good forage yield in 2017 and 2018. In addition, each yield of its four cuttings significantly increased between the two growing years. In addition, it showed outstanding performance regarding the accumulation of light energy owing to its high leaf area index and excellent light interception.
Three varieties, i.e., Gannong No. 5, WL343HQ, and Phabulous, performed better in terms of their two-year total yield than Xinjiang Daye. Gannong No. 5 gave the highest value, i.e., 22% (2507 kg ha −1 ) higher than that of Xinjiang Daye. This could be due to its excellent PH, LAI, LI, central leaflet area, and stem to leaf ratio. WL343HQ presented the most significant increase in forage yield from 2017 to 2018 among the 10 alfalfa varieties, and reached the highest yield in 2018. The two-year total yield of WL343HQ was more than 9% (993 kg ha −1 ) higher than that of Xinjiang Daye. WL343HQ had a low PH, low PB, low stem/leaf ratio and low central leaflet area, implying that this cultivar has substantial nutritional potential. A higher leaf/stem ratio has been largely used as a positive indicator of alfalfa forage quality in previous studies [19], owing to its strict positive correlation with digestibility and intake of forage [18] that arises from the higher digestibility of leaves relative to stems [45,46]. The two-year yield of Phabulous was 4% (512 kg ha −1 ) more than that of Xinjiang Daye. Phabulous possessed a high potential for forage yield because of its high stem/leaf ratio and PH features. However, the two-year total yield of Derby was slightly below 14% (1653 kg ha −1 ) compared with that of Xinjiang Daye.
Considering its high branch number, PH, and the central leaflet area, Derby could be used as a supplementary variety to reduce forage shortages for animals.
In 2017 and 2018, Wudi displayed poor performances in its annual yield. Gannong No. 3 showed poor productivity, significantly reducing each cutting yield from 2017 to 2018. Although the yield of Xinmu No. 2 and WL168HQ increased from 2017 to 2018, they presented poor performances in their two-year total yields. Moreover, Wudi, Sanditi, Gannong No. 3, Xinmu No. 2, and WL168HQ showed poor over-wintering performances (data not shown) in this study. When winter hardiness is poor for alfalfa, the productivity is reduced in an arid environment [47]. Generally, the yield should be the greatest determining factor for variety selection. In this study, Wudi, Gannong No. 3, Xinmu No. 2, Sanditi, and WL168HQ were found to be the least suitable varieties and are not recommended for planting in the Hexi Corridor region.
The study evaluated the correlation of forage yield with canopy structure traits and analyzed the productivity of 10 alfalfa varieties in an arid environment. Our results suggest that we should focus more intensively on stem traits, such as PH, PB, and stem/leaf ratio, which contribute significantly to canopy structure, and consequently, produce more forage yield in an arid environment. Moreover, Gannong No. 5, WL343HQ, Phabulous, and Xinjiang Daye achieved the highest forage yields. However, more field experiments with multiple locations are necessary to test these varieties over a broader area. In addition, a closer investigation of forage quality and related indicators of the selected varieties is necessary in follow-up research. These results provide more insights into alfalfa characteristics and could be used to improve forage yield and livestock production sustainability in the world's ever-expanding arid regions.

Site Description
The field experiment was conducted from 2016 to 2018 at the Linze experimental station of Lanzhou University in the Hexi Corridor, Gansu Province, China (100 • 02 E, 39 • 15 N, elevation 1390 m a.s.l.), a typical temperate desert climate area. We collected the climate datasets from 2005 to 2020. The annual mean air temperature was 7.6 • C, absolute maximum and minimum were 39.1 • C and −27 • C, and average annual precipitation was 117 mm. Moreover, the mean annual pan evaporation was 2390 mm, i.e., 20 times greater than the annual average precipitation. The frost-free season was 165 days, on average. The annual accumulated temperature of ≥10 • C was 3088 • C, on average [48].
The experimental site has a meadow soil mapped as an Entisol according to the U.S. soil classification system [49], a pH of 7.6, 0.806g kg −1 total nitrogen, and 0.703 g kg −1 total phosphorus at sowing in 2016. Previously, the experimental site had been utilized to grow yellow sweet clover [Melilotus officinalis (L.) Pall.] and maize (Zea mays L.).

Experimental Design and Agronomic Management
As outlined in Table 3, a total of 10 alfalfa cultivars were used in this experiment (Table 3). Alfalfa seeds were sown in nursery substrate (containing more than 40% organic matter and 15% humic acid) and acclimatized in a greenhouse for 45 days. Each plant was then transplanted into the experimental site on 20 June 2016. All experimental plots were arranged in a randomized complete block design with three replicates. The plot size was 3.6 m × 2.4 m, and the seedlings were planted at a spacing of 60 cm within and between rows.
In 2016, the experimental field received rotary tillage to a depth of 0.2 m before sowing. In addition, no fertilizer was applied at sowing or during the growing seasons of this experiment. The field received 150 mm of irrigation on 24 April, 25 June, 25 August, and 25 October during each growing season based on conventional local practices. However, irrigation could not compensate for the water loss caused by high pan evaporation (over 2390 mm per year). Weeds were controlled manually as necessary.

Plant Height
Plant Height was measured one month after the spring emergence and at flowering (first cutting). Three plants were randomly selected from each plot, and their height was measured as the distance from ground to the top end using a ruler (accuracy 1 mm).

Branch Number
During the initial growth phase, a primary axis develops from the seed, and secondary shoots subsequently develop from the axillary buds of cotyledons and primary axes leaves [51]. The primary branch number per plant (PB) and secondary branch number per stem (SB) were counted on five randomly selected plants from each plot at the flowering stage before the first cutting in 2017. Lastly, the total branch number per plant (TB) was calculated as the product of PB and SB.

Leaf Area Index and Light Interception
The leaf area index (LAI) and light interception (LI) were measured at the flowering stage before the first cutting in 2017 during clear sunshine from 11:00 to 13:00 using a Ceptometer (AccuPAR LP-80, Decagon Devices Inc., Pullman, WA, USA). One above and four below the canopy (close to the ground) randomly chosen readings on points were collected from each plot. Notably, the average below canopy reading was used. The LI was calculated [52] as: where PAR bl and PAR ab are photosynthetically active radiation below and above the canopy, respectively.

Forage Yield and Stem to Leaf Ratio
Samples were harvested manually using scythes on the plant base and measured four times annually in 2017 and 2018 at an interval of 4-5 weeks, starting at the initial flowering stage, specifically, 2 June, 8 July, 9 August, and 14 September in 2017, and 12 June, 14 July, 17 August, and 18 September in 2018. In the first cutting of 2017, the stem/leaf ratio was recorded on a single plant basis (after separation and oven drying of the stems and leaves). Subsequently, 10 plants were randomly chosen from each plot, cut at ground level and immediately weighed, before being oven-dried at 64 • C for 48 h to a constant weight in order to determine the dry weight. The central leaflet was a mature leaf taken from the main stem. To measure the length, width, and area, 20 randomly selected central leaflets from each plot (each species a total of 60 leaflets) were scanned, and the image was analyzed with an Epson GT-15000 and imaging software (WinSEEDLETM, Quebec City, QC, Canada).

Statistical Analysis
All the statistical analyses described below were performed using the R software (version 4.0.3). Plots were produced using the ggplot2 [53] and Factoextra [54] packages. A one-way analysis of variance (ANOVA) was conducted to assess differences of all measured traits among the 10 alfalfa varieties during the experimental years. The main effects and interactions of year and cultivar were obtained using two-way ANOVA. All tests were considered significantly different at p < 0.05, and the means were compared using Duncan's multiple comparison test. In addition, an independent samples t-test was conducted to assess the differences of traits between 2017 and 2018. Finally, we evaluated the relationships between canopy traits and forage yield using principal component analysis (PCA) based on a correlation matrix and Pearson's correlation analysis. Before starting PCA, we checked the sampling adequacy to detect whether or not the data would factor well. Sampling adequacy was measured by the Kaiser-Meyer-Olkin criterion (KMO). KMO varied from 0 to 1 and the overall KMO had to be not less than 0.5 to proceed to the principal component analysis. The Bartlett Test is another criterion in this sampling procedure which is likely also significant [55].