Radiation Use Efficiency of Forage Resources: A Meta-Analysis

Livestock management requires monitoring forage production (Guido et al., 2014; Irisarri et al., 2014; Pacín and Oesterheld, 2015; Rahman et al., 2014). As forage production is difficult to measure in the field (Sala and Austin, 2000), individual farmers and organizations are increasingly relying on models based on remote sensing (Fritz et al., 2019; Grigera et al., 2007; Hammer et al., 2010; McCall and BishopHurley, 2003; Paruelo et al., 2000). Most of these models derive from Monteith’s equation (Monteith, 1972), which integrates three components: (i) incident photosynthetically active radiation, (ii) the fraction of the incident radiation absorbed by the canopy, and (iii) the conversion efficiency of absorbed radiation into aboveground plant biomass (aboveground radiation use efficiency, hereafter referred to as RUE) (Piñeiro et al., 2006; Rahman et al., 2014; Running et al., 2004). Data on incident photosynthetically active radiation may be obtained from weather stations or estimated through remote sensing. The fraction of the incident radiation absorbed by the canopy may be estimated from vegetation indices derived from remote sensing (Huete et al., 2002; Kumar and Monteith, 1982). Conversely, RUE is difficult to estimate, generating a lot of uncertainty into the models (Nouvellon et al., 2000).

At present, no study has synthesized the published values of RUE of forage resources. Therefore, the magnitude and sources of variability are unknown. A meta-analysis by Slattery et al. (2013) showed the environmental influences on RUE of crop species. Forage species, very little represented in Slattery et al. (2013), differ from annual crops in root to shoot ratio, longevity, and carbohydrate partitioning to reproductive structures, features that may affect RUE and its response to environmental changes (Friesen et al., 1997; Haynes and Francis, 1993). Thus, it is especially important to characterize the variation of RUE of forage species and to understand the sources of such variability (Kiniry et al., 2012).

There are at least three broad categories of variables that affect the RUE of forage species: genotype, resource availability, and phenology. Genotype includes both interspecific and intraspecific differences. For instance, C3 species are expected to exhibit lower RUE than C4 species (Gosse et al., 1986). Within the same photosynthetic pathway, RUE differences may also be contrasting, as shown previously for two C4 species (Kiniry et al., 2013). Cultivars of the same species growing in similar environments may exhibit RUE differences as high as 363% (Kiniry et al., 2013).

Resource availability also explains some variation of RUE because of its effects on photosynthetic rates and shoot to root ratio (Bélanger et al., 1994; Cruz, 1997; Sinclair and Horie, 1989). Nitrogen, water and light availability have been the most studied influences on RUE of forage species. The RUE responses to these resources have been extremely broad, ranging from negative to positive (Bat-Oyun et al., 2011; Cristiano et al., 2012; Fletcher et al., 2013; Gómez et al., 2012; Mills et al., 2009). Several studies reported higher RUE in shaded plants than in plants exposed to full light (Cruz, 1996; Varella et al., 2011), whereas others showed the opposite (Gómez et al., 2012). All these discrepancies may stem from species identity, and the extent and type of resource limitation (nutrients, water or light).

Regarding plant phenology, vegetative stages may show higher, equal or lower RUE than reproductive stages (Alexandrino et al., 2005; Cristiano et al., 2012; Fletcher et al., 2013; Mills et al., 2009; Teixeira et al., 2008). The contradictory responses could be explained by interactions between phenological stage and biomass partition between roots and shoots, level of leaf N content, changes in respiratory loading per unit leaf area and energy content of the biochemical constituents of the plant products (Gifford, 2003; Gosse et al., 1986; Muchow and Sinclair, 1994; Muchow et al., 1993; Sinclair and Muchow, 1999).

Our objective was to quantitatively synthesize the variation of RUE of forage resources and its main controls. Specifically, we attempted to answer two questions: (i) what is the average and variability of RUE of forage resources? and (ii) how genotype, resource availability and phenology affect RUE? Considering that forage production models require RUE of aboveground biomass of monospecific or polyspecific resources, we included studies that estimated RUE from direct measurements of aboveground biomass at the individual (monoculture) or community level (pastures and grasslands).

Materials and methods

Database Selection

We selected articles recorded by ISI Web of Knowledge until November 2016. We searched for ‘radiation use efficiency’ OR ‘light use efficiency’, AND forage OR grassland OR meadow OR prairie OR steppe OR tundra OR names of the main forage species. Additionally, we retrieved studies cited in these references. We only included studies that: a) directly measured RUE of mono or polyspecific resources at plot or pot level, based on aboveground dry matter and total or photosynthetically active radiation intercepted or absorbed by the canopy, b) assessed effects of any biotic or abiotic factor on RUE, and c) reported means, standard errors or standard deviations, and sample sizes of each treatment (in a few occasions we obtained this information from the authors). As a result, we selected 48 studies, which reported 496 RUE values. Nearly 96% of these values corresponded to monocultures of 43 forage species. The remaining 4% were polyspecific resources (one humid savanna, one semiarid grassland and one grass-clover pasture). Monocultured species were 52% C4 and 48% C3. According to lifespan, they were 82% perennial and 18% annual species.

RUE values from the studies included in the meta-analysis are detailed in the Supplemental Tables S1 and S2. Additionally, the tables include 37 values from studies that did not meet criterion “b” (see above). Although they were not included in our meta-analysis, they may be helpful for future RUE data searches.

We assessed the effects of treatments related to plant genotype, resource availability and phenology (Table 1). Genotypic treatments were photosynthetic pathway (C3 vs. C4), species identity within C3 and C4 groups, and cultivar identity within species. Resource availability treatments were direct and indirect manipulations of resources. Direct manipulations were nitrogen fertilization, irrigation and shading, whereas indirect manipulations were defoliation frequency and inter-annual environmental variation. Although these indirect manipulations usually involve variation of regulators (e.g., temperature), we included them in the resource-availability group because in both cases plants are expected to be exposed to different water, nutrient and light availability. In the inter-annual variability analysis, we did not include the sowing year in perennial species. The phenological treatment was vegetative vs. reproductive stages.

For each treatment, we analyzed the variation of effects among studies (i.e., in meta-analysis terminology, heterogeneity among effect sizes). We explored the causes of this heterogeneity through the effects of various moderators (Koricheva et al., 2014). For this purpose, we collected the level of those moderators reported in each study. The levels of discrete moderators (i.e., photosynthetic pathway, life-cycle, ability to fix N symbiotically, water condition and soil texture) were taken as reported, whereas the levels of continuous moderators (i.e., level of N fertilization, site elevation) were arbitrarily categorized (Table 1).

When available, we extracted RUE directly from the articles. Other studies reported regressions of dry biomass versus a measure of either intercepted or absorbed radiation by the canopy. In these cases, we estimated RUE for each regression data point as the ratio between biomass and intercepted radiation. This ratio is more accurate than the regression slope because the slope equals RUE only when the intercept is zero (Verón et al., 2005). Because most RUE estimates considered intercepted photosynthetically active radiation (PAR), we converted all the other estimates to this basis. Estimations of RUE based on absorbed PAR were multiplied by 0.85 and estimates based on total solar radiation were divided by 0.5 (Sinclair and Muchow, 1999). Therefore, all analyses were made on RUE estimated on intercepted PAR (IPAR) and will be expressed in grams of dry matter (DM) per unit of IPAR (g DM MJ IPAR1). We extracted values from tables, text or figures (for the latter, we used GetData Graph Digitizer 2.26 software, Fedorov, 2013). We used the Median Absolute Deviation method to detect outliers (Leys et al., 2013), common at very low IPAR (Sinclair and Muchow, 1999). These outliers represented 5.9% of total data and were excluded. We considered results from different species, cultivars, growing periods, or treatment levels within the same experiment as independent measurements. Although this criterion makes some of the data statistically dependent, loss of information by excluding them would have likely resulted in serious distortion of the results (Gurevitch et al., 1992).

Meta-Analysis Response Ratio and Statistics

We calculated the effect sizes of treatments as follows (Hedges et al., 1999):

Table 1. Treatments, moderators and moderator levels considered for each variable included in the meta-analysis.

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 1

Where RUE, represents the mean radiation use efficiency in treated plants and RUEc the mean radiation use efficiency in control plants. We transformed the ratio using the natural logarithm, because non-transformed ratios generally have poor statistical properties (Rosenberg et al., 2013). For resource availability treatments, RUEc was the least manipulated condition (i.e., the lowest amount of N or water applied, the highest light availability, and the lowest defoliation frequency). For genotype treatments, RUEc was the RUE of C3 species for C3 vs. C4 treatments and the minimum RUE for the species or cultivar identity treatments. For interannual variation RUEc corresponded to the minimum RUE. For phenological treatments, RUEc corresponded to vegetative stages. Treatments were significant when 95% confidence intervals did not overlap zero. We analyzed the heterogeneity among effect sizes of each treatment by the Q test. If total heterogeneity (QT) was significant (α = 0.05), we assessed the heterogeneity among the levels of each moderator (QB). We only reported significant effects of moderators. We used the absolute value of the mean effect size to rank the impact of treatments on RUE, regardless of their sign. We performed all meta-analytical calculations and statistical comparisons with MetaWin software, version 2.1 (Rosenberg et al., 2000).

Results

Overall Variation of RUE

Mean RUE of the entire dataset (n = 496) was 1.93 ± 1.2 g DM MJ IPAR–1 (±SD). Ninety–eight percent of the values came from field–plot experiments (n = 484), and the remaining came from pot experiments. Eighty-four percent of the data ranged between 0.5 and 3 g DM MJ IPAR–1 (Fig. 1). Three percent were below 0.5, and 13% were above 3 g DM MJ IPAR–1 (Fig. 1). Studies focused mostly on C4 species and cultivar identity effects, and on N fertilization and irrigation (Table 2). Many studies included several years of data, which allowed exploring inter-annual environmental variation. Overall, the treatments that produced greater changes of RUE regardless of their sign (absolute value of effect size) were related to genotype (C4 species identity, cultivar identity, and C3 species identity) and resource availability (shading, N fertilization and defoliation frequency). The smallest absolute value of effect size on RUE corresponded to contrasts between photosynthetic pathway and irrigation levels. The transition from vegetative to reproductive state generated absolute values of effect sizes of intermediate magnitude (Table 2).

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 2

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 3

Influence of Treatments Related to Genotype on RUE

RUE widely varied among forage species and cultivars within species (Fig. 2). Some of this variation changed with environmental conditions. C3 species identity generated a mean RUE change of 42%, with significant heterogeneity (QT = 64, p = 0.00158). However, the available moderators (Table 1) did not explain it. On average, C4 species identity generated a 61% change in RUE (Fig. 2). These interspecific differences were higher under rainfed condition (65%) than under irrigated condition (36%, heterogeneity QT = 150, p = 0.0087, QB = 4.31, p = 0.0378, respectively). Mean effect size of cultivar identity on RUE was 58% (Fig. 2) and varied with photosynthetic pathway and N fertilization (QT = 104, p = 0.0409; QB = 4.45, p = 0.0347). Cultivar differed much more within non-fertilized C4 species than within fertilized C3 species (Fig. 2). On average, plant photosynthetic pathway did not influence RUE and the response was homogeneous (confidence interval overlapped zero, QT = 6.9, p = 0.9946).

Influence of Resource Availability on RUE

Direct manipulations of resource availability greatly influenced RUE. N fertilization caused wide variations that depended on the applied dose, the photosynthetic pathway of the species, soil texture and water condition (Fig. 3). On average, N fertilization increased RUE by 43% (Fig. 3). Effect sizes presented significant heterogeneity (QT = 140, p = 0.0150), related to N dose, plant photosynthetic pathway, and soil and water condition. The effect of N fertilization increased with the amount of fertilizer applied (QB = 51, p < 0.0001; Fig. 3). It was also much larger when applied to C3 than to C4 species (QB = 24, p < 0.0001; Fig. 3). RUE
increased more by N fertilization on fine than on coarse textured soils (QB = 25, p < 0.0001; Fig. 3). Finally, the effect of N fertilization on RUE increased as water condition of the experiments became more favorable for plant growth (QB = 10, p = 0.0056; Fig. 3). In addition to these N fertilization effects, irrigation and shading affected RUE (Fig. 4). On average, irrigation increased RUE by 23% and shading increased it by 40% (Fig. 4). In both cases, total heterogeneity was significant among effect sizes (QT = 150, p < 0.0001 and QT = 322, p < 0.0001, respectively). RUE also widely responded to indirect manipulations of resource availability. The mean variation between the RUE of a year and the minimum RUE in the period under study was 40%, with no heterogeneity among effect sizes (QT = 151, p = 0.1538). Higher defoliation frequency increased RUE by 23% (Fig. 5), with significant heterogeneity among effect sizes (QT = 47, p = 0.0324). Defoliation frequency increased RUE under rainfed conditions and did not affect it under irrigation conditions (Fig. 5) (QB = 18, p < 0.0001). Similarly, the positive effect of defoliation frequency was observed in plants with the ability to fix nitrogen symbiotically (QB = 8, p = 0.0049) and in annuals more than in perennials (QB = 32, p < 0.0001; Fig. 5).

Influence of Phenology on RUE

The RUE varied less between vegetative and reproductive stages than between some of the genotypic and resource availability contrasts shown above. The mean effect size did not differ significantly from zero. There were however broad differences among studies (QT = 644, p < 0.0001), which were not accounted for by any of the available moderators (photosynthetic pathway, soil texture, site elevation, water condition and level of N fertilization, p > 0.05).

Discussion

This study provides the first quantitative synthesis on the variation of RUE of forage resources and its main controls. Our results indicate that (i) the pattern of variation of RUE of forage resources differ from the patterns observed in crops, (ii) the major drivers of this variation were both genotypic and environmental, and frequently interactive, and (iii) satellite-based models aimed at modeling forage production face the great challenge of incorporating all this relevant variation.

Regarding the differences of RUE patterns, mean RUE of forage resources was lower than typical RUE of the main crops (Sinclair and Muchow, 1999) and the magnitude of RUE change generated by environmental factors also varied between these two groups (Slattery et al., 2013). Forage resources presented a higher sensibility to changes in light and N availability than crops. Furthermore, the higher response of RUE to N fertilization in C3 than in C4 species contradicts the pattern observed in crops (Slattery et al., 2013). These discrepancies could be explained not only by the expected differences in morphophysiological traits between species of each group, but also by the processes of artificial selection and genetic improvement of crop species. Genetic manipulation has allowed increasing tolerance to different types of stress, modifying morphological and physiological characteristics, such as root system architecture, stomatal conductance, nutrient use efficiency, and herbivore resistance (Cattivelli et al., 2008; de Dorlodot et al., 2007; Deng et al., 2008; Kant et al., 2010). We also found similarities between forage resources and crops. For instance, both groups presented a positive relationship between RUE and N dose, and similar percentages of RUE change according to water availability (Slattery et al., 2013).

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 4

Regarding the drivers of RUE variation, we found stronger effects of genotypic and resource-availability treatments than phenological stage treatment. A large part of RUE variability was explained by genotypic factors, such as identity of C3 and C4 species and cultivar identity within the same species. Differences in morphophysiological traits among species or cultivars of the same species (e.g., shoot to root ratio, photosynthesis rate, energy content of tissues) may explain this response (Sinclair and Muchow, 1999; Stockle and Kiniry, 1990). Variation in RUE among C4 species was higher in studies performed under rainfed than under irrigated conditions. This result suggests that forage species widely differ in their tolerance to water stress. Photosynthesis pathway did not influence RUE of forage species. This result agrees with work by Sinclair and Muchow (1999) and Albrizio and Steduto (2005), who observed that the expected difference in RUE between C3 and C4 species (Monteith, 1978) is rarely demonstrated experimentally.

As reported for crops by Slattery et al. (2013), N fertilization increased RUE and the magnitude of this response was positively correlated with N dose. Positive effect of N fertilization on RUE can be related to two mechanisms. On the one hand, fertilization may enhance the partitioning to shoots rather than roots (Kobe et al., 2010; Lestienne et al., 2006). On the other hand, fertilization often increases leaf N content, which increases CO2 assimilation rates (Sinclair and Horie, 1989). The lack of response of RUE to N fertilization on coarse-texture soils (i.e., soils with low water holding capacity) or under drought conditions suggests that water availability imposes a higher order limitation.

Under water stress, stomatal closure reduces water loss via transpiration, but the downregulation of photosynthesis also reduces RUE (Cornic and Massacci, 1996; Chaves et al., 2009; Earl and Davis, 2003; Flexas et al., 2007). Additionally, water stress increases the partitioning of carbon to the roots, and reduces shoot growth more than total growth, with the concomitant reduction of aboveground RUE (Stockle and Kiniry, 1990). Indeed, the consequent reduction in RUE is usually more sensitive to water stress than IPAR (e.g., Bat-Oyun et al., 2011; Manderscheid et al., 2014). We detected a significant heterogeneity among effect sizes of water stress, but it could not be explained by the available moderators. It is possible that part of the heterogeneity is related to the timing and duration of drought treatments (Jamieson et al., 1995; Sinclair and Muchow, 1999).

On average, shading increased RUE of forage resources. This response may be related to the reduction of incident radiation that can increase shoot to root ratio, leaf to stem ratio and leaf N concentration, and reduce the carbon dioxide limitation for photosynthesis of the leaves (Healey et al., 1998; Wilson and Ludlow, 1991). It could also be caused by increases in the proportion of diffuse radiation, which allows a greater light penetration into the canopy, stimulating canopy photosynthesis (Roderick et al., 2001; Sinclair et al., 1992). In this case, we detected significant heterogeneity in the response of RUE to shading, but none of the analyzed moderators (water condition and N fertilization level) explained it.

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 5

The positive effect of defoliation frequency on RUE may be explained by the mobilization of C and N stored in organs and the increase of Rubisco activity after defoliation (Avice et al., 1997; Harrison et al., 2010; Strullu et al., 2013; Teixeira et al., 2008). However, we found heterogeneity among effect sizes. One causal factor could be that defoliation frequency increased RUE only in species with the ability to fix N symbiotically, which have a specific organ for reserves (i.e., crowns). Another factor could be that defoliation frequency increased RUE only under rainfed conditions. This could be attributed to reduced leaf area index and canopy evaporative demand in defoliated canopies, which allow a partial alleviation of leaf water stress (Harrison et al., 2010).

Decreases in RUE during the transition from vegetative to reproductive stage caused by a reduction in leaf N concentration and the generation of high-energy tissues were reported in crop species (Campbell et al., 2001; Hall et al., 1995; Muchow and Sinclair, 1994). This pattern was evidenced only in some forage species growing in certain environments, but there were also opposite responses. This situation led to a nonsignificant effect of phenological stage on RUE when considering all the data. This discrepancy could be partially due to the fact that forage species have not undergone the process of selection and genetic improvement that many crop species have, which tend to increase the flux of carbohydrates and nitrogen to reproductive structures (Slafer et al., 2005).

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 6

Radiation Use Efficiency of Forage Resources: A Meta-Analysis - Image 7

The variability of RUE shown by this meta-analysis, implies a great challenge to estimate aboveground net primary productivity (ANPP) through remote sensing. According to Monteith (1972), any variation of RUE proportionally translates into a change of ANPP. Thus, assigning a value of RUE to the equation is a key decision. Our results show that ANPP estimations of a certain forage species could vary on average by 58% depending on the RUE of the cultivar (Fig. 2). For example, estimated ANPP of Phleum pratense cultivar ‘Nike’ would be 60% higher than ‘Farol’ (RUE values are respectively 2.89 and 1.8 g MJ IPAR–1; Supplemental Table S1). Similarly, the resource availability status of the forage resource will strongly influence RUE and thus ANPP. We showed that N fertilization level can modify the ANPP on average by 43% due to its effects on RUE (Fig. 3). For example, ANPP of Festuca arundinacea ‘Clarine’ in a fertilized plot (120 kg N ha–1) would be 67% higher than in unfertilized plots (RUE values are respectively 1.92 and 1.15 g MJ IPAR–1; see Supplemental material). The case of shading is different. Our results suggest that under cloudy conditions higher RUE will compensate at least partially for the reduction of incident PAR. As a consequence, ANPP estimates will be higher than if Monteith’s equation were used with a constant RUE.

Conclusions

Our meta-analysis significantly advances our understanding of the influence of genotype, phenology and resource availability on RUE of forage resources. The information generated will increase our certainty when selecting RUE values as input to estimate productivity through remote sensing. As a result, estimates of secondary production can also be improved, and the impact of different management practices and the effect of climate change on such productivity also can be estimated. The large RUE variability also highlights the margin for increasing RUE through breeding. We also identified some knowledge gaps (Koricheva et al., 2014). For instance, the heterogeneity of the effects of phenology and C3 species identity on RUE remains to be explained. In addition, the literature lacks enough data to evaluate the influence of some factors related to the environment (e.g., temperature, presence of pathogens or symbionts) or to management practices (e.g., phosphorous fertilization, plant density, row spacing, and defoliation intensity) on the RUE of forage resources.

This article was originally published in Agronomy Journal. 111:1–9 (2019). doi:10.2134/agronj2018.10. 645. Reproduced here with permission from the author.