Perennial Forage Systems Enhance Ecosystem Quality Variables Compared with Annual Forage Systems

Abstract

There is an intense argument about the environmental impact of annual vs. perennial forage production systems. In this study, a systematic review was employed to obtain 47 empirical studies from 13 published papers between the years 2017–2023 to help clarify the issue. The objective of this study was to determine how perennial and annual forage (business-as-usual, BAU) production systems affect dry matter yield (DM) and energy of production including specific environmental impact variables. Impact variables were classified into three main groups: human health, ecosystem quality, and resource consumption. Net energy of lactation (NEL) was considered as a functional unit. Overall, perennial forage production systems varied less in DM yield and energy production than annual monocrop systems, indicating stability in perennial production. There was no statistically significant difference in human health and resource consumption variables between perennial and annual forage production systems, except for ozone layer depletion potential. However, perennial forage systems significantly lowered variables within the ecosystem quality category. Ecotoxicity potential decreased by two and 18 times compared with BAU—control (only annual monoculture forages), and BAU—improved (any annual cropping system other than BAU—control), respectively. Perennial forage systems showed a significant effect size of −8.16, which was slightly less than the effect size of the BAU—improved system but two times less than BAU—control in terms of terrestrial acidification potential. While BAU—control showed an insignificant effect size in relation to eutrophication potential (EUP), perennial forage systems reduced EUP by approximately five and two times compared with BAU—control and BAU—improved, respectively. Therefore, this study highlights the importance of promoting perennial forage production system to foster resilience and stability in DM yield and energy production, with improvements in environmental human health (ozone layer depletion potential) and ecosystem quality variables.

1. Introduction

As global challenges like climate change, biodiversity loss, and resource depletion increase, adopting more sustainable agricultural practices becomes imperative. Modern agriculture has been a driving force behind the unprecedented growth in global food production, but concerns have also been raised about environmental consequences [1,2]. This is due to high reliance on chemical fertilizers, irrigation, animal farming, land clearing, and chemical pest control. These negative impacts of agriculture could be worsened by the already existing adverse effects of human-induced climate change if it is not properly managed [2]. Thus, practices in crop production, such as crop diversity, crop rotation, and the crop lifecycle, play a crucial role in shaping the environmental impact of cropping systems worldwide [3,4,5,6,7,8,9]. For the above reasons, there has been a growing interest in exploring sustainable alternatives to conventional annual cropping systems in recent years.

Conventional annual cropping systems regarded as business as usual (BAU), such as annual forages for hay or silage, often require more frequent tillage and do not provide soil cover in the winter months, leading to high carbon release and a high carbon footprint [10,11]. Therefore, one promising solution to mitigate impacts of climate change and contribute to the reduction of greenhouse gas (GHG) emissions is the diversification of cropping systems by integrating perennial forages into high-input, low-diversity cropping systems [10,12,13,14]. This is because perennial forage cropping offers several advantages over its annual counterparts. For instance, it reduces the need for frequent tillage and resowing. Also, perennial forages establish deeper root systems, providing year-round soil cover, enhancing soil health, and reducing soil erosion. Perennial cropping systems not only lead to increased carbon sequestration but also contribute to efficient water utilization [13]. By reducing the need for frequent tillage and pesticide application, perennial forage systems can help mitigate abiotic resource consumption and decrease the risk of soil degradation [15].

Furthermore, Picasso et al. [16] have documented that perennial cropping systems are more environmentally friendly and sustainable than conventional annual systems. However, different life cycle assessment (LCA) studies have provided contrasting findings due to inherent different assumptions about cropping systems, dry matter yield, energy productivity, specific environmental impact categories, functional units, and methodology [17,18]. For example, Zucali et al. [19] reported variability in environmental impacts between an annual system of silage maize (Zea mays L.) and Italian ryegrass (Lolium multiflorum L.) and a perennial system of alfalfa (Medicago sativa L.). The environmental benefits of alfalfa production without irrigation were offset when irrigation was employed, due to the electricity required to operate the irrigation systems. In a previous report [20], cup plant (Silphium perfoliatum L.) significantly reduced terrestrial acidification, climate change, marine eutrophication, freshwater eutrophication, human toxicity, terrestrial ecotoxicity fossil depletion, and freshwater ecotoxicity compared with maize. Yet, cup plant had a significant high environmental impact in terms of water depletion. Fathollahi et al. [21] reported higher ozone layer depletion and terrestrial ecotoxicity but lower global warming potential, marine aquatic ecotoxicity, acidification, and eutrophication potential for alfalfa hay compared with silage maize in forage production systems for dairy farming in Iran. Research from beef production systems in Uruguay also documented trade-offs between environmental impact categories for systems based on perennial forages vs. annual forages; GHG emissions were higher in perennial systems, but energy use, soil erosion, and pesticide ecotoxicity were lower compared with annual systems [22,23]. These findings call for a synthesis to understand the environmental impacts of perennial and annual forage production systems by considering net energy of lactation (NEL) as a functional unit [19]. Calculating environmental load in terms of NEL base units could serve as an interesting analysis to understand environmental impacts and forage quality and can be more easily connected to livestock production.

Therefore, the objective of this study was to use a meta-analytical approach to determine how productivity (dry matter yield and energy production) and each environmental impact category affect annual and perennial forage cropping systems in different disparate regions, using NEL as a base functional unit. The cropping systems included annual (BAU—control and BAU—improved) and perennial systems. Environmental impact categories included human health variables such as global warming potential (GWP), GHG emissions, and ozone layer depletion (OLD); resource consumption variables such as fossil energy consumption and abiotic depletion potential (ADP); and ecosystem quality variables such as ecotoxicity potential (ECP) (terrestrial, freshwater, and marine), terrestrial acidification potential (AP), and eutrophication potential (EUP). This information can provide critical insight for building a resilient and environmentally responsible agricultural future as policymakers, farmers, and researchers strive to address challenges posed by modern agriculture.

2. Materials and Methods

2.1. Goal and Scope Definition

This meta-analysis was conducted on environmental impacts observed in perennial systems vs. annual cropping systems to identify a unifying lifecycle assessment (LCA) methodology for the disparate regions and ecosystems being examined. No geographical limitation was adopted, given that the focus of the review/meta-analysis was on specific annual vs. perennial forage cropping systems. It is important to note that the impact of these management practices can vary depending on the specific environmental conditions, climate, soil types, and local farming practices in different regions of the world. Consequently, site-specific assessments and tailored management strategies are essential for maximizing the benefits of forage production while minimizing its environmental footprint. Sustainable and adaptive management practices are another vital aspect that plays a crucial role in ensuring the long-term viability and health of forage production systems across the globe, which involves continuously monitoring and adjusting practices in response to changing conditions. This approach acknowledges that the environment and farming context are dynamic, and what works today may need modification in the future to remain effective and resilient. To ensure accuracy and reliability, data for this analysis were extracted from original peer-reviewed research papers that analyzed GHG emissions or carbon footprint and/or employed life cycle assessment (LCA), which are widely accepted methodologies in assessing climate changes and the environmental impacts of cropping systems. LCA is a holistic approach that uses a systematic set of procedures to convert the inputs and outputs of materials and energy that characterize a process into the associated environmental impact [1,24], and it usually follows the ISO standard 14040/44 methodology [25].

2.2. Data Collection and Management

Articles including direct measurements or estimates based on models or LCA were included in the assessment of the existing literature. While reviewing the articles, we collected data on treatment means of dry matter yield and net energy of lactation. Environmental impact categories were extracted from original national and international peer-reviewed articles reporting results estimated using LCA methodology. Priority was given to papers that analyzed impacts/parameters related to GWP or GHG emissions, OLD, fossil energy consumption, ADP, ECP, AP, and EUP. These data were grouped according to management practices related to forage production (grain, hay, silage, or grazing).

The number of observations (sample size) and information on dry matter yield, energy production, and several environmental variables, as well as features that may have affected forage cropping system responses were collected and categorized according to the cropping systems (Tables S1–S4). Data were collected using all available resources (e.g., Web of Science, Google Scholar, SCOPUS, etc.). The key words for the databases were life cycle assessment OR, agricultural LCA, OR environmental impact, OR greenhouse gas emission, OR global warming potential, combined using AND with the keywords forage, perennial forage, perennial cropping system, annual cropping system, mixed cropping systems, integrated cropping systems, OR crop rotation, OR yield.

The search of these records resulted in 769 articles (2017–2023) publications, examined by title and abstract using the keywords above. Then, the following selection criteria were applied: (i) article published in a peer-reviewed journal; (ii) study related to agricultural production (grain, pasture, silage, or hay); (iii) only studies performing impact assessment with more than one LCA indicator or GHG emissions were retained; and (iv) studies with lack of transparency, insufficient quantitative information, or units that made the data impossible to convert were excluded, leaving 154 articles. Furthermore, during the research process, the sample was complemented with nine articles identified by forward and backward tracking of citations and suggestions from co-authors and reviewers. The aim was to increase the quality of the sample and to find additional sources helpful for answering the research questions and the discussion [26,27]. This process resulted in 49 articles, of which 13 contained enough information for the meta-analysis. In these articles, data on 47 cropping systems were reported, including the differences between perennial and annual forages on environmental impact categories. In some instances, where data were provided in graphical form, means were extracted using GetData Graph Digitizer [28]. The collection and meta-analysis of the data followed the PRISMA (Preferred Reporting Items for System Review and Meta-Analysis) guidelines [29], as shown in Figure S1. A general overview of the selected studies is provided and their key parameters are listed in Table S5. The moderators used for the study were BAU—control, BAU—improved, and perennial. Definitions are indicated in

Table 1. Levels and definitions of moderators for forage cropping systems.

Moderator	Definition
BAU—control	Cropping systems that are considered BAU—control include only annual monoculture forages (single crops used as a control in our study).
BAU—improved	Cropping systems that are considered BAU—improved include all kinds of cropping systems except those used as a control (annual monoculture forages). This includes diversified monoculture systems without perennial crops in the system.
Perennial	Crops that can regrow and continue to produce grains, seeds, and biomass (forage) after a single year. They can be harvested numerous times for many years. Includes diverse, perennial, circular systems (DPCSs) as defined by Picasso et al. [16] and the three dimensions of diversity (more crop species), perenniality (a perennial crop), and circularity (legume crop or integrated systems).

.

2.3. Functional Unit (FU) and Assumptions

The selection of the functional unit (FU) plays a crucial role among the various steps defined in the ISO standards, as it ensures a fair comparison with previous studies, and different FUs can lead to varying results and conclusions [17,30]. As specified in ISO 14040 [25] the functional unit is described as the quantified performance of a product system in one year, serving as the reference unit in an LCA. According to Bacenetti et al. [24], historically, when it comes to agricultural processes, commonly chosen FUs have been based on (i) mass, such as the mass of grain, fruit, milk, or meat; (ii) area, where usually for agricultural processes, 1 hectare (ha) is considered; and (iii) energy, such as the energy produced (e.g., electricity from biogas) or the energetic value of the product (e.g., metabolizable energy content). In this study, considering that the focus was on management practices related to forage production including grain, hay, silage, or grazing, which are used as animal feed, especially for cattle consumption for meat and milk production, the following FUs were chosen:

• Mass-based FU: $1 \mathrm{M}\mathrm{g}$ of dry matter $\left(\mathrm{D}\mathrm{M}\right)$ of the product (grain, silage, hay, or forage) per year $(\mathrm{M}\mathrm{g} \mathrm{D}\mathrm{M} {\mathrm{y}}^{-1})$ for animal feed production. This FU aims to evidence the production efficiency of different cropping systems [17];
• Area-based FU: $1 \mathrm{h}\mathrm{a}$ of land occupied per year $(\mathrm{h}\mathrm{a} {\mathrm{y}}^{-1})$. Considering that environmental impacts are related to a specific amount of land, the concern of a land-based unit is considered complementary to the mass-based unit. However, both usually lead to different results [31];
• Energy-based FU: $1 \mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l}$ of net energy of lactation $(\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{L})$ per year. The net energy of lactation $\left(\mathrm{N}\mathrm{E}\mathrm{L}\right)$ is an interesting FU to compare forages with different energy content, since the NEL unit includes the product’s quality, and can possibly be more easily connected to livestock production [19,32].

To ensure a fair comparison between results from different studies, some of the FUs from the original articles were converted into the FUs chosen for this study [17,32]. In this phase, knowing the yields of the different crops analyzed in terms of the amount of $\mathrm{D}\mathrm{M}$ of product per hectare per year $\left({\mathrm{M}\mathrm{g} \mathrm{h}\mathrm{a}}^{-1}{\mathrm{y}}^{-1}\right)$ and $\mathrm{N}\mathrm{E}\mathrm{L}$ per amount of $\mathrm{D}\mathrm{M}$ of the product $\left({\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{L} \mathrm{M}\mathrm{g} \mathrm{D}\mathrm{M}}^{-1}\right)$ was crucial. However, most of the papers analyzed reported the yield in terms of the amount of product per unit of area, but not in terms of the energetic value of the product, such as $\mathrm{N}\mathrm{E}\mathrm{L}.$ Thus, when the $\mathrm{N}\mathrm{E}\mathrm{L}$ data were presented in a paper, the conversion of the original results of the environmental impact categories (IC) to $\mathrm{N}\mathrm{E}\mathrm{l}$ was calculated using Equation (1):

$$\left(\mathrm{I}\mathrm{C} \mathrm{u}\mathrm{n}\mathrm{i}\mathrm{t}\right){\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{L}}^{-1}=\left(\mathrm{I}\mathrm{C} \mathrm{f}\mathrm{o}\mathrm{r} \mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c} \mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\right){\mathrm{M}\mathrm{g} \mathrm{D}\mathrm{M}}^{-1} {\mathrm{y}}^{-1}/{\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{L} \mathrm{M}\mathrm{g} \mathrm{D}\mathrm{M}}^{-1} {\mathrm{y}}^{-1}$$

(1)

When the $\mathrm{N}\mathrm{E}\mathrm{L}$ data were not mentioned in a paper, the conversion of the original results to $\mathrm{N}\mathrm{E}\mathrm{L}$ was conducted according to the equation presented in [32], considering total digestible nutrients $\left(\mathrm{T}\mathrm{D}\mathrm{N}\right)$ of feed, Equation (2), or diet (forage, silage, hay, or concentrates) and fat and crude protein contents in milk, Equation (3). When the $\mathrm{T}\mathrm{D}\mathrm{N}$ data were not mentioned in a paper, we used the standard data from [33], considering the specific feedstuff.

Feed or diet:

$${\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{l} \mathrm{k}\mathrm{g} \mathrm{D}\mathrm{M}}^{-1}=0.0245\times \%\mathrm{T}\mathrm{D}\mathrm{N}-0.12$$

(2)

Milk:

$${\mathrm{M}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{N}\mathrm{E}\mathrm{l} \mathrm{k}\mathrm{g} \mathrm{D}\mathrm{M}}^{-1}=0.0929\times \%\mathrm{f}\mathrm{a}\mathrm{t}+0.0547\times \%\mathrm{c}\mathrm{r}\mathrm{u}\mathrm{d}\mathrm{e} \mathrm{p}\mathrm{r}\mathrm{o}\mathrm{t}\mathrm{e}\mathrm{i}\mathrm{n}+0.193$$

(3)

The $\mathrm{N}\mathrm{E}\mathrm{L}$ of cup plant and $\mathrm{T}\mathrm{D}\mathrm{N}$ of switchgrass (Panicum virgatum L.) were obtained from [34,35], respectively. The methods, factors, and equations used in the conversion process are listed in Table S6. The average yield of products (grain, pasture, silage, and/or hay) in Mg ha⁻¹ y⁻¹ and energy production in Mcal NEL Mg DM⁻¹ in each cropping system analyzed are shown in Table S1.

2.4. Statistical Analysis of Data

2.4.1. Moderator Variables

Various variables affecting the forage cropping systems were employed as moderators to ascertain their level of environmental impact. In addition to the parameter means recorded from each empirical study, we considered the available information on the variables and characteristics that may have affected the forage cropping system response, to understand and interpret results and determine whether management of perennial forages had less environmental impact than annual forage crops.

2.4.2. Effect Size and Meta-Analysis

For the meta-analysis, we adopted the methodology followed by [36]. We used Comprehensive Meta-Analysis (CMA) software, version 4 [37], to estimate the effect of the selected cropping systems on the effect size (ES), using mean ES across studies of yield, energy production, and categories of impact of interest. The random-effects model was adopted, the variance of the natural log of response ratios was calculated based on sample size (non-parametric variance) of different moderator levels (Table 1), and the number of observations (sample size) was used in the meta-analysis. The mean effect size of selected variables was analyzed with a 95% confidence interval for each moderator level considering the summary effect (mean $ES$ across studies) on all variables of interest.

Meta-analysis of yield, energy production, and the environmental impact categories selected was conducted to analyze the effect of perennial forage crops compared with annual forage crops on reducing resource consumption, increasing human health, and ecosystem quality, by forage production, considering different types of cropping systems. Due to the diversity of the variables of interest and their measurement units impacting the variation in results and magnitude between those variables, it was necessary to transform data, which allowed the comparison and interpretation of data in terms of $ES$ of the treatments compared with the control treatments. This method has been employed in similar studies [36,38,39,40] because it gives a standardized expression of treatment-induced changes that have biological significance [36]. We used a log transformation to balance positive and negative treatment effects across response ratios $\left(RR\right)$, while the symmetry between data was maintained. Therefore, treatment $ES$ was evaluated in terms of the natural logarithm of the $RR\left(lnRR\right)$ of the treated moderators to the control means, as expressed by Equation (4).

$$\mathrm{E}\mathrm{S}=\mathrm{l}\mathrm{n}\mathrm{R}\mathrm{R}=\mathrm{l}\mathrm{n}\left({\displaystyle \raisebox{1ex}{${\mathrm{Y}}_{\mathrm{t}\mathrm{r}\mathrm{t}}$}\!\left/ \!\raisebox{-1ex}{${\mathrm{Y}}_{\mathrm{c}\mathrm{t}\mathrm{r}}$}\right.}\right)$$

(4)

where $Y_{trt}$and $Y_{ctr}$ are means of treated $\left(trt\right)$ and control $\left(ctr\right)$ treatments for each variable of interest and represent the cumulative $ES$ of the cultivation of perennial forage crops compared with annual monoculture forage cropping systems (BAU—control and BAU—improved) according to the method adopted by Ashworth et al. [36].

For the forest plot analysis, weighted, overall summary effect sizes (response ratios) did not include null hypothetical significance testing (NHST) alone. Estimation and confidence intervals were used accordingly in the analysis of the effect sizes [41]. Negative values indicated that the specific moderator subgroup induced a decrease in the parameter of interest, and positive values indicated a positive effect on the parameter of interest.

2.4.3. Variance

For studies reporting multiple annual monoculture forages as controls, effect sizes were computed using the control treatment average. Thus, as explained above, controls were represented by annual monoculture forages. In studies not reporting the number of replicates, we assumed $n=1$. Multiple treatment combinations from a single paper were treated as independent studies and are represented as individual units in this meta-analysis; thus, each treatment mean recorded was considered independent (47 treatments in total). Considering that many papers do not explicitly report the standard deviation or standard errors, according to [42], it is usual to omit measures of dispersion from some publications, and this requires the calculation of weight based on sample size (nonparametric variance). Exclusion of studies that report sample size but not some measure of dispersion would represent a substantial loss of analytical power [42]. Thus, the variance estimation was based on the treatment sample sizes (non-parametric variance), according to Equation (5):

$$V_{In\left(RR\right)}={\displaystyle \frac{{(n}_{trt}+n_{ctr})}{(n_{trt}\ast n_{ctr})}}$$

(5)

where $V_{In\left(RR\right)}$ is the variance of the natural log of the response ratio RR, and $n_{trt}$and $n_{ctr}$ are the sample sizes of treated $\left(trt\right)$ and control $\left(ctr\right)$ treatments, considering a simple non-parametric variance for a single trial effect size. In this study, $n_{ctr}$ was represented by the average sample size from the control group.

2.4.4. Test for Heterogeneity Using the Q Statistic

The $\mathrm{Q} \mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}$ was used to determine whether the true effect varied across all studies. Heterogeneity is a yardstick for measuring the $\mathrm{Q} \mathrm{s}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}$ (or multiple significance testing across means; weighted squared deviations) and ascertaining using ${\mathrm{I}}^2$ an index that estimates ratios of true variation to total variation across effect sizes [36,43]. ${\mathrm{I}}^2$ is defined by Equation (5). An ${\mathrm{I}}^2$ value of $0\%=\mathrm{n}\mathrm{o} \mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e} \mathrm{h}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{i}\mathrm{t}\mathrm{y}; >0 \mathrm{i}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{s} \mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e} \mathrm{h}\mathrm{e}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{n}\mathrm{e}\mathrm{i}\mathrm{t}\mathrm{y}$; and larger values suggests variation due to true heterogeneity among studies. When the $p\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}$ for the $\mathrm{Q}$ test $\left(P_{hetero}\right)$ is less than the alpha heterogeneity criterion of $0.1$, the null hypothesis of among-study variance across moderator subgroups is rejected [43].

$${\mathrm{I}}^2={\displaystyle \frac{{(\mathrm{Q}}_{\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}}-\mathrm{d}\mathrm{f})}{{\mathrm{Q}}_{\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}}}}\times 100$$

(6)

where ${(\mathrm{Q}}_{\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}})$ is total variation, degrees of freedom $\left(\mathrm{d}\mathrm{f}\right)$ represent expected within-study variation, and ${\mathrm{Q}}_{\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}}-\mathrm{d}\mathrm{f}$ is true heterogeneity or between-study variation ${(\mathrm{Q}}_{\mathrm{t}\mathrm{r}\mathrm{u}\mathrm{e}}$).

2.4.5. Publication Bias and Sensitivity Analysis

Publication bias is an assessment of a systematic literature review, to check a holistic representation of all completed studies on a particular research topic. The publication bias in this study was analyzed using a dataset for energy production because it contained all 47 studies employed in the meta-analysis (Table S7). However, publication bias issues are common with meta-analysis due to the quantitative and comprehensive approach that it involves [43]. It is important to know whether there has been a trade-off or failure to publish non-significant findings with regard to significant treatments in a study. Although it is good to carry out publication bias analyses on reviewed studies, obtaining a justifiable result is difficult [44,45]. The relationship between effect size and precision is used to analyze publication bias; the funnel plot is employed to measure the study size, and most often it plots standard error or precision on the vertical axis against the effect size on the horizontal axis [43]. This is based on the principle that studies with smaller sample sizes or large variances tend to have higher effect sizes compared with larger studies with better precision. Begg and Mazumdar [46] used rank order correlation of “Kendall’s tau b” to determine publication bias via a funnel plot. An improved Rosenthal fail-safe N method [47], Orwin’s fail-safe N method [43], was used to determine whether the summary effect could be associated with publication bias. Trimming and filling via the Duval and Tweedie iteration approach was used to assess how the missing studies would shift the effect size [48]. Sensitivity analyses were conducted on each of the subgroup analyses to determine how the overall summary effects varied relative to the weight of the samples.

3. Results

3.1. Heterogeneity, Sensitivity, and Publication Bias

Table 2 shows significant heterogeneity p values (P_hetero) of <0.100 in most subgroups’ summary effect sizes. These results are significant, as P_hetero falls below the alpha criterion of 0.1, except for fossil energy consumption (p = 0.230), abiotic depletion potential (p = 0.377), and global warming potential (p = 0.43). Any P_hetero > 0.100 fails to reject the null hypotheses, the I² values rendered unfit to conclude that there are differences between the true effect sizes among the study populations. Our sensitivity analysis also showed no evidence of substantial variation in the effect sizes when single studies were deliberately removed to test the robustness of the meta-analysis. The overall summary effects and their p-values did not vary relative to the weights of individual groups of studies or datasets. The funnel plot analysis conducted using a dataset for energy production showed no evidence of publication bias (Figure S2). The two-tailed p-value of the Eggers intercept test was 0.482, indicating no evidence of publication bias. Orwin’s fail-safe N was 149 (i.e., 149 ‘null’ studies needed for a combined two-tailed p-value > 0.05).

Table 2. Heterogeneity test for moderators.

Moderator	n	Qvalue †	df	Phetero	I2
Dry matter yield	43	24.09	2	<0.100	91.7
Net energy of lactation	47	9.08	2	0.0110	77.9
Fossil energy consumption	21	2.93	2	0.2300	31.8
Abiotic depletion potential	34	1.95	2	0.3770	−2.5
Global warming potential	42	1.46	2	0.4810	−36.6
Ozone layer depletion potential	19	12.82	2	0.0020	84.4
Ecotoxicity potential	33	33.51	2	<0.100	94.0
Terrestrial acidification	38	15.40	2	<0.100	87.0
Eutrophication potential	28	14.78	2	0.0010	86.5

^† $Q_{value}$ (between-study variation); $n$, number of studies; $df,$ degrees of freedom, levels within a moderator; $I^2$, the ratio of true variation (heterogeneity) to total variation; $p\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}$ = $P_{hetero}>0.100$ shows that all observed (total) variation is due to sampling error (within-study variation). Dry matter yield, energy production, fossil energy consumption, abiotic depletion potential, global warming potential, ozone layer depletion potential, ecotoxicity potential, terrestrial acidification, and eutrophication potential effect size analyses were conducted on log-transformed values $\left(lnRR\right)$ from each study. The levels of each moderator with their summary effect sizes, confidence intervals, intercrop-induced change as a percentage, number of studies, and significance values are given in Figure 1, Figure 2, Figure 3 and Figure 4.

Table 2. Heterogeneity test for moderators.

Moderator	n	Qvalue †	df	Phetero	I2
Dry matter yield	43	24.09	2	<0.100	91.7
Net energy of lactation	47	9.08	2	0.0110	77.9
Fossil energy consumption	21	2.93	2	0.2300	31.8
Abiotic depletion potential	34	1.95	2	0.3770	−2.5
Global warming potential	42	1.46	2	0.4810	−36.6
Ozone layer depletion potential	19	12.82	2	0.0020	84.4
Ecotoxicity potential	33	33.51	2	<0.100	94.0
Terrestrial acidification	38	15.40	2	<0.100	87.0
Eutrophication potential	28	14.78	2	0.0010	86.5

^† $Q_{value}$ (between-study variation); $n$, number of studies; $df,$ degrees of freedom, levels within a moderator; $I^2$, the ratio of true variation (heterogeneity) to total variation; $p\mathrm{v}\mathrm{a}\mathrm{l}\mathrm{u}\mathrm{e}$ = $P_{hetero}>0.100$ shows that all observed (total) variation is due to sampling error (within-study variation). Dry matter yield, energy production, fossil energy consumption, abiotic depletion potential, global warming potential, ozone layer depletion potential, ecotoxicity potential, terrestrial acidification, and eutrophication potential effect size analyses were conducted on log-transformed values $\left(lnRR\right)$ from each study. The levels of each moderator with their summary effect sizes, confidence intervals, intercrop-induced change as a percentage, number of studies, and significance values are given in Figure 1, Figure 2, Figure 3 and Figure 4.

3.2. Effect of Cropping System Levels on Dry Matter Yield and Energy Production

For the dry matter yield, overall, perennial, and BAU—improved showed significant effect sizes with p value ≤ 0.05 (Figure 1). However, effect sizes for overall and perennial systems showed decreases in dry matter yield. In contrast, BAU—improved showed an increase in dry matter yield. BAU—control showed a non-significant effect size with a p value greater than 0.05. Regarding energy production, BAU—control and BAU—improved both showed a non-significant effect size (p value ≥ 0.05). This means that the dry matter yield of BAU—improved did not translate to equivalent significant energy production. However, the overall and perennial systems maintained significant decreases in effect size in relation to energy production, similar to their effects on dry matter yield. The results in Figure 1 show a decrease in dry matter yield and energy production in with perennial cropping systems; however, there is no evidence to show that BAU—control had better dry matter yield, failing to reject the null hypothesis at a 95% level of confidence.

3.3. Effect of Cropping Systems on Environmental Impact Categories

Perennial forage systems and BAU—improved systems significantly decreased global warming potential, with a p value ≤ 0.050 for the overall effect (Figure 2). However, the heterogeneity test results in Table 2 failed to support the hypothesis that there is true variation in the effect sizes among cropping systems for global warming potential. This result is true because of the overlapping of the confidence intervals. Perennial forage cropping systems decreased the ozone layer depletion potential more than BAU—control and BAU—improved, including the overall effect size (Figure 2). This decrease accounted for 8.5% less ozone layer depletion by perennial forage systems than BAU—control and BAU—improved.

Similar to the lack of heterogeneity in global warming potential variables, there was also no evidence of variation among the cropping systems to describe the environmental impacts associated with fossil energy consumption and abiotic depletion potential (Figure 3). The cropping systems describing the environmental impact of fossil energy consumption also failed to reject the null hypothesis, with non-significant p values > 0.050.

For abiotic depletion potential, there were significant p values of less than 0.05; however, the result failed to establish evidence of true differences among the cropping systems due to the overlapping of the confidence intervals, except for BAU—improved (Figure 3). Figure 4 shows a significant decrease in ecotoxicity potential associated with perennial forage systems and BAU—control cropping systems, compared with the overall effect. BAU—improved had no significant effect, with a p value of 0.128. There was an overall significant effect size of −0.878 for the cropping system variables describing the environmental impact of ecotoxicity potential. Perennial forage systems had a significant effect size of −8.268, representing decreases in ecotoxicity potential of approximately nine, two, and eighteen times compared with the overall effect size, BAU—control, and BAU—improved, respectively (Figure 4).

However, BAU—improved showed an insignificant effect size in terms of ecotoxicity potential. Overall, perennial forage systems, BAU—control, and BAU—improved had significant decreases in terrestrial acidification potential compared with the overall effect size. Perennial forage systems had a significant effect size of −8.159, which was slightly less than the effect size of overall and BAU—improved but half that of BAU—control in terms of terrestrial acidification potential. Furthermore, eutrophication potential was significantly decreased overall by perennial forage systems and BAU—improved, while BAU—control had an insignificant effect size in relation to eutrophication potential (Figure 4). Perennial forage systems decreased eutrophication potential approximately five and two times compared with BAU—control and BAU—improved, respectively.

4. Discussion

The lack of statistical heterogeneity despite the consistency in the results of this study might be due to sampling error in the studied populations or the limited number of studies that our samples included. Thus, it was very challenging from a statistical perspective to establish whether there was true effect size variation across the whole range of studies with low sample size. However, this meta-analysis addressed this challenge of low statistical power by using new statistical methods of analysis, such as NHST, estimation, and confidence intervals, to analyze the forest plots [41]. Combining all these new statistical methods of interpreting forest plots made it possible to analyze moderators, thereby accounting for more elaborate details of each impact category via a non-dichotomous approach.

The differences in dry matter yield and energy production in annual forage cropping systems are likely to have depended on the season, harvest management, or whether fertilizer input was hindered in a nitrogen-limited environment [49], since dry matter yield and nutritive value depend primarily on harvest management and available nitrogen [50,51,52]. Perennials’ dry matter yield and energy production may also have decreased inputs compared with more intensive fertilizer input systems such as BAU—improved or BAU—control [16,53]. The results for BAU—improved included similar reports [54,55,56] where a high diversity of forage species did not increase biomass yield and energy, because the correlation between species richness and biomass relies on a combination of species that are well-adapted to a particular environment. The authors also suggested that for a diverse perennial system to have a competitive higher biomass yield, dominant “driver” species in each community should be grown.

Also, the lack of significant increase in energy production despite higher dry matter yield in BAU—improved systems suggests that the composition of the biomass may depend on other factors, such as lower calorific value and higher moisture content. This means that biomass with high dry matter yield may not produce equivalent energy. Therefore, the synergetic relationship between DM yield, nutritive value, and environmental impacts requires strategic management of one-crop systems. However, the challenge is to achieve the best balance by maximizing yield and nutritive value in the context of environmental sustainability. These analyses are crucial because they relate to the various environmental impact categories. For instance, when livestock consume high-quality feeds, they excrete less waste, which reduces the emission of pollutants such as ammonia methane and dinitrogen monoxide [19,24]. Hence, this information will help in the efficient management of dairy nutrient balance, feeding levels, and the energy efficiency of diets, therefore influencing the excretion associated with nutrients and greenhouse gas emissions [19]. In the environmental impact categories, perennial forage systems significantly reduced ozone layer depletion potential, ecotoxicity, terrestrial acidification, and eutrophication potential compared with annual monoculture systems BAU and BAU—improved. These categories represent human health and ecosystem quality variables. However, their impact was more prominent in ecosystem quality variables. Franco et al. [57] reported that lower benefits are expected from ecosystem services in annual single-genotype monoculture systems compared with landscapes with perennial crops and greater interspecies and intraspecies diversity. This may also be connected with the numerous biogeochemical and biochemical ecosystem services benefits associated with diverse perennial circular systems, such as low material and energy input compared with annual cropping systems [16]. Similarly, a synopsis of Berti et al. [58] and Nemecek et al. [59] shows that low-input cropping systems could reduce impacts in categories such as acidification, eutrophication, and human toxicity. This is due to decreased chemical fertilizer application, which reduces N-NO₃ leaching, p run-off, and N₂O field emission, including reduced terrestrial acidification due to reduction in NH₃ application. Also, cropping diversification has been demonstrated to minimize eutrophication potential [1]. These ecosystem service benefits associated with perennial and diverse perennial circular system are exemplified in rotation or intercropping of annual and perennial species [20,60]. Intercropping perennial forages with maize reduced the environmental burden, while rotating annual crops with perennial crops had the highest mitigation of environmental burdens compared with rotation with other annual crops (BAU—improved). This demonstrates the true definition of nutrient circularity [16] in a diverse perennial circular system. Therefore, perennial diversified cropping systems have more resilience and stability and provide more ecosystem services than annual monocultures [15]. For this reason, Liao et al. [8] reported that integrating alfalfa into a diverse monoculture system improved all environmental impact categories.

BAU and BAU—improved systems did not reduce the potential environmental impacts indicated above as perennial forage systems did. This was because, according to findings in Solinas et al. [61], the cropping methods for BAU—improved systems are closely related to the conventional annual monocultural system. This results in loss of soil organic carbon, inefficient use of water, low nutrient retention, and high prevalence of pests inherent to low-diversity ecosystems. Also, agricultural monoculture systems are associated with polluting the groundwater via pesticides and fertilizers [62]. However, in our study, there was no statistical difference between the effect of cropping systems on resource consumption variables, including global warming potential. This is because the environmental benefits that would have been attributed to the perennial forage system over the annual system are offset by its agrotechnological operational requirements, such as excess use of fossil fuel due to multiple harvests per season. There is a direct relationship between global warming and fossil fuel consumption [7]. In a similar trend [63,64,65], there were no differences in GHG balances when alfalfa was integrated into annual or perennial cropping systems. This was in addition to the findings of Guareschi et al. [6] and Yang et al. [2], where fuels used for mowing, raking, and baling in multiple alfalfa harvest cuts and transportation added to the CO₂ emissions. All these facts and findings show that field operations influence GWP and fossil fuel consumption in perennial forage cropping systems. Therefore, progress should be made to adopt any processes that can improve and establish more environmentally friendly approaches to the field operation of any forage cropping system.

The results of perennial diversification shed light on the lack of a broader understanding of whether diversification practices can support biodiversity and multiple ecosystem services [66]. For instance, while diversification practices such as crop rotation and the application of organic amendments enhance soil bacteria and the functions of arbuscular mycorrhizal fungi, they also increase soil respiration and breakdown of organic matter, resulting in increased emissions of CO₂ and nitrous oxide, a potent greenhouse gas. For this reason, the positive environmental effects of diversification in agroecosystems can be offset by greenhouse emissions. Considering this, it is essential to holistically assess and adopt DPCSs with perennial legumes or forages (in rotation with annual crops) and spatial diversity (e.g., intercropping multiple crop species), which can significantly impact land use change and environmental impact categories relating to ozone layer depletion potential and ecosystem quality.

5. Conclusions

Our meta-analysis synthesizes results to reveal that perennial forage dry matter yield did not vary significantly like BAU—control and BAU—improved. Among the environmental impact categories, perennial forage production systems decreased negative impacts on ecosystem quality variables such as terrestrial acidification, ecotoxicity, and eutrophication potential compared with annual monoculture systems. In spite of their yearly fossil energy usage during multiple harvests, perennial systems had similar global warming potential to annual monoculture systems. The highest gains in the perennial crop lifecycle are attributed to its ability to reduce the use of chemical nitrogen fertilization. This further underscores the importance of biologically fixing N₂ with leguminous perennial forages like alfalfa. Therefore, a perennial forage cropping system can foster resilience and stability by improving ecosystem service delivery.

Finally, most of the data used in this study were products of LCA studies that used different models and functional units to analyze environmental burdens. Hence, the diverse cropping systems and operations of agricultural systems and the limited number of studies among the study samples made interpretations challenging due to the various scenarios used by multiple studies. However, after a holistic look at the inherent assumptions made and considering the different allocation of impacts, this work concludes that all the moderators employed in this study had a specific effect on mitigating the environmental burdens in perennial cropping systems. However, functional units must be duly considered when conducting environmental impact assessments in agriculture.