*4.1. Explaining Water Use E*ffi*ciency*

#### 4.1.1. Water Use Efficiency and Soil Nutrient Supply

Although nutrient management was not a treatment in this study, field observations of possible nutrient deficiencies prompted investigation to the relationship between WUE and nutrient management. Proper soil nutrient supply improves plant growth, photosynthetic rate, transpiration, root growth and ultimately yield, all result in increased WUE [23]. For example, pearl millet, being tested under four levels of increasing N and irrigation supply, achieved 41 kg ha−<sup>1</sup> mm−<sup>1</sup> WUE in the most water stressed treatment with the highest level of N application, 225 kg N ha−<sup>1</sup> [20]. The authors of [37] also illustrated the effect that irrigation and N application can have on WUE in an experiment with irrigated and rainfed wheat when "adequate" N was applied with irrigation, mean WUE was 78% greater and grain production was 16% greater in irrigated than in rainfed wheat without N application. The 2016 and 2017 of this pearl millet study, DM was influenced by observed nutrient deficiency, especially in the no-till treatment of 2017. The authors of [8] conducted an experiment with forage pearl millet at the same time just adjacent to the plot where this forage pearl millet experiment took place. Agronomic management similar to the H irrigation, 19 cm row spacing, and tilled treatment was similar in both studies except that in [8] N was applied in both study years at rates of 84 kg ha−<sup>1</sup> and 78 kg ha−<sup>1</sup> in 2016 and 2017, respectively. Pearl millet DM reported by [8] used approximately 514 and 602 mm of water in 2016 and 2017, respectively, which produced 6287 and 9874 kg ha−1, with WUE of 12.2 and 16.4 kg ha−<sup>1</sup> mm−<sup>1</sup> in those years, respectively. The DM yields from [8] were approximately three times greater than the mean across all treatments of this study in 2016 and 2017 having used a maximum of 462 and 641 mm in those years without N application. The lack of N, demonstrated by comparison to [8], lead to lack of nutrients reduced plant productivity and negatively impacted WUE [38].

### 4.1.2. Weed Influence on Pearl Millet Water Use Efficiency

Native grass/weed DM may have impacted pearl millet WUE due to no weed control in either year of this study. In 2016 grass/weed DM was not influenced by row spacing, however grass/weed DM was different among irrigation and tillage levels (Table 7). In 2017, only row spacing influenced grass/weed DM. The authors of [39] demonstrated in forage sorghum, when weeds are present, there was no difference in DM yields among wide (76 cm) or narrow (25 cm) row spacing. Although weed control was not a treatment in this study, regression analysis showed that grass/weed DM did influence pearl millet WUE across treatments and years (Figure 6). As weed proliferation was not checked by high leaf area index and resulting light interception correlated with crop competitive ability [40], weed/grass DM accumulated, thus reducing pearl millet WUE.

**Table 7.** Grass/weed dry matter (DM) at pearl millet harvest by treatment near Canyon, TX in 2016 and 2017.


<sup>a</sup> Lowercase letters indicate mean separation within effect, *p* < 0.05.

**Figure 6.** Pearl millet water use efficiency (WUE) as a function of grass/weed dry matter (DM) across treatments and years.

#### 4.1.3. Water Use Efficiency Response to Agronomics

Pearl millet WUE reported in the literature has been consistent yet wide ranging. During a five-year rotation study by [27], mean pearl millet production from two seasons was 3670 kg ha−1. Although weather conditions were not considerably different between the two years, [27] reported WUE of 31 kg ha−<sup>1</sup> mm−<sup>1</sup> in 1995 and 4.5 kg ha−<sup>1</sup> mm−<sup>1</sup> 1996, most likely due to one month earlier planting date in 1995, allowing the crop to better utilize soil water prior to harvest. The authors of [13] found the WUE of pearl millet, in a study of various cultivars and plant populations in Niger, ranged from 7.6 to 8.7 kg ha−<sup>1</sup> mm−1. In Tunisia, [41] reported a pearl millet WUE range of 6.4 to 7.6 kg ha−<sup>1</sup> mm−<sup>1</sup> among four water stress treatments. The authors of [22] reported pearl millet DM values of 4000 to 8300 kg ha−<sup>1</sup> and WUE from 13.8 to 17.9 kg ha−<sup>1</sup> mm−1, respectively, in India. Whereas, in Akron, CO, among forage and grain, foxtail and proso millet cropping systems described by [33], DM production ranged from 562 to 4545 kg ha−<sup>1</sup> with WUE ranging from 4.8 to 27.5 kg ha−<sup>1</sup> mm−<sup>1</sup> during the five-year study. Additionally, [20] also reported high WUE ranging between 35.7 and 41 kg ha−<sup>1</sup> mm−<sup>1</sup> under four levels of irrigation. Grain sorghum WUE has been reported between 11 to 49.2 kg ha−<sup>1</sup> mm−<sup>1</sup> among locations throughout the Central and Southern Great Plains U.S. [42,43].

In the studies of pearl millet and sorghum cited above, minimum or no-till was used by [27] and [43], however, tillage information was absent from the other studies. In this study, a WUE response to agronomic management was only observed from tillage practices (Figure 2), whereas WUE was not impacted in the same way in this study by the increasing DM and WU from irrigation treatments as previously observed in other studies [20,22,41]. In a review of soil management effects on WUE, [23] reported that in no-till, where plant residue is maintained on the soil surface, higher WUE is typically found due to reduced evaporative water loss (E) from soil, lower soil surface temperatures, and improved water infiltration. The authors of [23] analyzed nineteen studies of crop and soil management practices and synthesized their results into a graphical depiction of crop WUE responses to soil management and seasonal effects. They demonstrated that crop biomass and yield increased or decreased due to changes in soil management. In this study, pearl millet DM response to soil tillage reflects differences in biomass due to soil management changes as described by [23] (Figure 7). However, when, soil surface E is roughly approximated at the x-intercept of the DM/WU regression line (Figure 7) [38,44], the two-season average estimated E from tilled and no-till soil was approximately 114 and 131 mm, respectively. While these estimates are similar, the result is contrary to reports of less E from no-till soil [23]. Additionally, WU was not different between tillage or row spacing treatments in either year. Furthermore, the spread of DM responses observed, lack-of-fit to the DM/WU predicted line, and no differences in DM or WU detected between row space levels and WU between tillage levels indicates that WU was insensitive to changes in the management practices other than when water was added by irrigation (Figures 2, 7 and 8). The authors of [24] reported that pearl millet yield is largely independent of WU when LAI is low, as was observed in both seasons of this pear millet forage study, due to the majority of light being intercepted by the soil surface allowing water loss to E in the sparse canopied crop with LAI values < 2 [44]. In 2016, LAI was not different between the tillage levels, however, in 2017, LAI in the tilled treatment reached approximately 2 nearly 100 GDDs before no-till reached a LAI of approximately 2. As mentioned above, the N deficiency observed predominately in no-till explains slower growth, lower DM and WEU for both years even though till had lower early season plant count than in no-till during 2017. Thus, higher WUE in tilled soil was achieved as a result of earlier canopy development, more light interception and greater DM production.

**Figure 7.** Pearl millet forage dry matter (DM)/water use (WU) relationship with tillage levels indicating management impact on DM production for 2016 and 2017 at Canyon, TX.

**Figure 8.** Direct relationship between water use efficiency (WUE) and pearl millet dry matter (DM). Slopes show trends of increasing WUE as DM increases for each irrigation level.

The proceeding discussion explaining WUE differences in the tilled treatment also helps explain the absence of main effect interactions and lack of WUE response to decreasing irrigation. Poor soil nutrient supply and high weed competition reduced pearl millet crop growth thereby obscuring any possible interactions among the main effects and responses typically reported from no-till or water stress studies. The authors of [20,22,41] found that pearl millet WUE increased as water stress also increased, which was not observed in this study. However, the relationship between WUE and DM [26,37,38] affords direct comparison of WUE at each irrigation level (Figure 8). Statistical contrast analysis of the slopes and intercepts for the respective irrigation levels [26] show that L had greater positive gain in WUE than was achieved in the H or M. This analysis suggests that WUE at the L level had a higher rate of return for accumulated DM than the other water levels. The authors of [22] demonstrated that in water stressed pearl millet WUE increased 23% from well-watered conditions. Analysis of their results demonstrates that pearl millet maintains physiological functions even when constrained by water stress. For example, the ratio of net photosynthesis (g CO2 m−<sup>2</sup> h<sup>−</sup>1) to WU (mm) was greater in the limited water treatment than in the full water treatment (0.029, 0.002, respectively) [22]. This result can be explained by the biochemical pathway in warm season C4 plants used to continue photosynthesis when vapor pressure deficit is high in low density canopies [45,46]. Furthermore, [44] explained that increased WUE this is the result of a greater proportional increase in WUE for low LAI values than when LAI is >2. Thus, the direct relationship between WUE and DM illustrates an ecophysiological response from pearl millet interacting with the crop environment.

Although irrigation applied lead to increased DM, WU did not well predict the behavior of pearl millet forage production across both years and all treatments. The production function <sup>y</sup> <sup>=</sup> 6.68 <sup>×</sup> x (mm) <sup>−</sup> 837 kg ha−<sup>1</sup> (*R*<sup>2</sup> = 0.31, *n* = 96) cannot be readily utilized in other production situations that might have better weed and sol nutrient management. In contrast, for most DM/WU linear regressions, the coefficient of determination is much higher (>0.7) [33,42] and more confidence is given to estimating crop yield from WU given a set of WUE optimizing management strategies.
