Clarifying Grazing Management Methods: A Data-Driven Review

Abstract

Grasslands, particularly permanent grasslands, provide vital ecosystem services and, therefore, focus a number of management challenges. Grassland management revolves around organizing how livestock graze in both space and time, using various grazing methods. However, international research describes these grazing methods using diverse and sometimes inconsistent terminologies. This lack of standardization may create ambiguity and hinder comparative research on grazing methods. Here, to address this issue, we conducted a literature review aiming to identify common patterns of grazing methods based on shared grazing management criteria. Through multivariate analysis, we analyzed 249 experimental datapoints derived from 102 studies. We ran principal component analysis followed by hierarchical clustering on principal components on seven management criteria. This review identified 4 broad families of grazing methods: continuous grazing, conventional rotational grazing, deferred rotational grazing, adaptative multi-paddock grazing. This work distinguishes rotational from continuous grazing methods, as commonly described in the literature. Furthermore, it identifies adaptative multi-paddock grazing as a distinct and innovative group of rotational grazing. The approach developed here could serve as support to characterize and compare different grazing methods.

1. Introduction

There were 3.35 billion hectares of grassland worldwide in 2021, representing 70% of all agricultural land [1]. Grasslands serve as the predominant forage resource for grazing livestock around the world [2]. Grasslands also provide essential ecosystem services that are essential for sustainability, such as carbon storage, biodiversity conservation, water supply and flow regulation… [3,4]. The term ‘grassland’ encompasses a wide range of ecosystems (e.g., woody grasslands, savannas, pampas) that share a common vegetation mostly composed of grasses, legumes, and other herbaceous plants, and at times woody species [5,6,7]. In some parts of the world, there are ecosystems composed of indigenous vegetation [5,8] shaped by particular pedoclimatic conditions—soil and climate characteristics—(e.g., steppes, Alpine and Arctic grasslands, azonal grasslands) [9], wild herbivory [10] and fires (wild or prescribed) [11]. However, most grasslands are shaped by human influence designed to maintain native or introduce herbaceous flora via specific management practices [5,10,12] such as grazing or mowing [8]. Grazing is defined as the action of herbivores feeding in grasslands, which results in selective defoliation of the herbaceous cover [13] and prevents the vegetation succession as, for example, a shrubby stage [9]. This results in a complex interaction between plants, animals, and soil that directly impact vegetation and soil dynamics [14].

Human-managed grassland grazing has historically been nomadic, involving the movement of animals over large areas [15]. However, it is now mainly characterized by grazing in restricted areas, resulting in complex and dynamic systems called grazing systems, which Allan et al. [5] defined as “an integrated combination of soil, plant, animal, social, economic features, grazing method(s) and management objectives designed to achieve specific results or goals”. Setting up a grazing system is a decision-making process that factors in the farmer’s profitability and lifestyle objectives and the constraints imposed by landowners, farm’s local pedoclimatic conditions, water resources and land fragmentation [16]. The way the grazing system organizes the animals in both space and time is defined as the grazing method [5] and is characterized by grazing management criteria established by the farm manager, such as number of paddocks, rest period(s), stocking period(s) and stocking rate.

Grazing methods are generally structured around two main types of organization: continuous grazing, or rotational grazing [13,16,17,18]. Under continuous grazing, animals spend most of the stocking season on a small number of paddocks (one usually) but for very long stocking periods that typically stretch to the duration of the whole stocking season [13,17], resulting in a low instantaneous stocking density. Under rotational grazing, the farmer subdivides the grazing management unit (the total area to be grazed during the stocking season) into a variable number of paddocks and then moves their animals sequentially from paddock to paddock, creating differences in the state of the vegetation in terms of plant biomass and phenology. This approach provides rest periods enabling the in-paddock plants to recover from grazing [16]. Oftentimes, farmers set up rotational grazing methods to achieve their zootechnic objectives (production, reproduction, welfare, health …) by controlling the quantity and/or quality of forage on offer to the animals [13]. Sometimes though, farmers implement rotational methods as part of a more systemic approach that considers the provision of environmental services, such as carbon storage and biodiversity conservation, as seen in adaptive multi-paddock grazing [19,20].

The combinations of management criteria result in a complex array of grazing system configurations and consequently an array of names and terminologies used to describe grazing methods. In 1979, Lacey and Van Poollen [21] pointed out the need to standardize the terminology used for grazing methods. They proposed a dichotomous key that grouped grazing methods into 14 classes, based firstly on the distinction between year-round grazing methods and methods that integrate a winter break. Leray et al. [18] later grouped grazing methods into four categories: “intensive free-range grazing”, “rotational grazing”, “simplified rotational grazing”, and “dynamic rotational grazing”, and highlighted the need for methodological work to determine which criteria can effectively characterize grazing methods. Similarly, Allen et al. [5] defined 20 different methods (e.g., alternate stocking, continuous stocking, rotational stocking, variable stocking) that could overlap depending on the management criteria applied. Here, in an effort to evaluate grazing management practices, we identified 106 different names used to qualify different grazing methods (

Table 1. Terminologies used for the various grazing methods identified in international literature.

Grazing Method	Citation
10-paddock rotational grazing system	[22]
11-paddock rotational grazing	[23]
120-pasture, ultra-high stocking density system	[24]
15-paddock grazing system	[25]
2-pasture 1-herd deferred-rotation grazing system	[26]
3-pasture 2-herd deferred-rotation grazing system	[26]
30-paddock grazing system	[25]
4-pasture deferred-rotation system	[27]
4-pasture rotation system	[24]
4-pasture rotationally deferred grazing	[28,29]
4-pasture 3-herd deferred-rotation grazing	[26]
4-pasture 3-herd deferred-rotation system	[30]
6-paddock rotational grazing	[23]
8-paddock short-duration rotation grazing	[28]
8-paddock time-controlled rotational grazing	[29]
Adaptative multi-paddock grazing	[31,32,33,34]
Alternating stocking	[35]
Alternative rotational grazing	[36]
Cell stocking	[37]
Collaborative, adaptative rangeland management	[38]
Continuous grazing	[23,27,28,36,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77]
Continuous grazing management	[78]
Continuous grazing system	[22,25,44,79,80,81,82,83,84,85,86,87,88,89,90]
Continuous heavy grazing	[91]
Continuous large pasture	[92]
Continuous moderate grazing	[91]
Continuous set stocking grazing	[93]
Continuous small pastures	[92]
Continuous stocking	[35,37,94,95,96,97]
Continuous stocking system	[98]
Continuous turnout grazing	[99]
Continuous yearlong grazing	[100]
Continuous yearlong grazing system	[26]
Conventional grazing	[31,34,101]
Conventional grazing management	[33]
Conventional rotational grazing	[102]
Deferred rotation grazing	[58,103]
Deferred rotation grazing system	[83]
Deferred rotation system	[54,104]
Deferred system	[57]
Deferred-rotation grazing	[99,105]
Deferred-rotational grazing	[52,73]
Divisional rotation grazing	[53]
Flexible grazing system	[25]
Four-paddock rotational grazing	[47]
Free grazing	[106]
Free-range intensive grazing	[107]
Grazing in rotation	[45,108]
Grazing switchback system	[80]
Heavily stocked continuous grazing	[32]
Heavy continuous grazing	[30,90]
High-density short-duration grazing	[106]
High-intensity, low-frequency grazing	[109]
High-intensity-short duration rotational grazing	[93]
High-performance short-duration grazing	[110]
Hight intensity low frequency grazing system	[83]
Intensive free-range grazing in spring followed by simplified rotational grazing	[107]
Intensive rotational grazing management	[84]
Intensive short-duration rotational grazing	[62]
Intensive time-controlled rotation grazing system	[92]
Light continuous grazing	[90]
Merrill system	[80]
Moderate continuous grazing	[30]
Moderately stocked continuous grazing	[32]
Multi-paddock grazing	[90]
Permanent grazing	[111]
Planned rotational grazing	[112]
Rationed grazing	[107]
Regenerative rotational grazing	[102]
Repeated season-long grazing	[113]
Repeated seasonal grazing	[56,110]
Rest-rotation grazing	[105]
Rotated seasonal grazing	[56]
Rotation deferred grazing	[114]
Rotation grazing	[48]
Rotational grazing	[35,39,40,41,42,43,46,49,59,61,62,63,64,65,66,67,68,69,70,71,74,75,76,77,103,107,111,112,115,116,117]
Rotational grazing management	[78]
Rotational grazing system	[11,22,44,79,81,82,85,86,88,89]
Rotational heavy grazing	[91]
Rotational moderate grazing	[91]
Rotational plus reserve grazing	[59]
Rotational stocking	[94,95,96,97]
Rotational stocking system	[98]
Santa Rita grazing system	[100]
Season-long continuous grazing	[118]
Season-long grazing	[29,105,114,119,120]
Season-long stocking	[104]
Seasonal continuous heavy grazing	[121]
Seasonal continuous light grazing	[121]
Seasonal rotational heavy grazing	[121]
Set stocking grazing	[112,116,117]
Short duration grazing	[50,55,60,109,113,120]
Short-duration rotation grazing	[51]
Short-duration rotational grazing	[118]
Simplified rotational grazing	[107]
Strategic grazing	[72]
Summer-long grazing every year	[122]
Summer-long grazing in alternate years	[122]
Three-paddock deferred grazing	[47]
Three-unit rest-rotation grazing	[122]
Time-controlled grazing	[87]
Time-controlled rotation grazing	[99]
Traditional rangeland management	[38]
Twice-over rotation grazing	[119]
Yearlong continuous grazing	[103]
Yearlong continuously grazed treatment	[115]

), although not all of them are distinct. The purpose of this review was to analyze how a wide range of grazing methods identified in the literature relate to one another according to common management criteria, regardless of the terminology used.

2. Materials and Methods

2.1. Litterture Review: Identification of Studies Comparing Grazing Methods

We first conducted a review of the international literature to identify articles comparing at least two grazing methods in relation to one or more services provided. We initially used simple keyword searches (e.g., ‘grazing system’, ‘grazing method’, ‘stocking method’, ‘grazing management’) to determine the scope, and finalize the keywords for the study. This preliminary work allowed to identify a literature review comparing rotational and continuous grazing in terms of plant and animal production. We used the reference lists from this literature as a first source [123]. We then conducted a targeted search on the Web of Science platform, focusing on paper titles and a wide range of keywords, i.e., TI = ((grazing systems) OR (grazing management) OR (continuous grazing) OR (rotational grazing) OR (short-duration grazing) OR (adaptive multi-paddock grazing) OR (deferred rotation grazing)) AND TI = (“performance” OR “production” OR “productivity” OR “biomass” OR “standing crop” OR “nutritive value” OR “vegetation” OR “botanical composition” OR “biodiversity” OR “economy” OR “workload” OR “welfare” OR “behavior” OR “soil” OR “carbon” OR “water” OR “infiltration” OR “moisture”). Finally, 102 articles were selected, including 54 from the exploratory search (of which 43 were taken from the review by Briske et al., 2008 [123]) and 48 from the targeted search.

2.2. Grid Analysis: Extraction of Experimental Datapoints

A grid analysis was built to identify different variables in the selected articles. Firstly, we selected variables related to the study context, including article title and source, study location, climate factors (e.g., mean annual precipitation and temperature), and names for the grazing methods studied. Second, borrowing from Allen et al. [5], we established a list of management criteria linked to the grazing methods and allied calculated criteria, i.e.,

• Stocking season duration (days): A stocking season refers to the period of each year during which animals are present in the grazing management unit. This criterion is closely linked to climatic conditions, including seasonal grass growth, and soil bearing capacity.
• Number of paddocks: A paddock is a grazing area within the grazing management unit that is separated from other areas by fencing. When there are multiple paddocks, animals can be rotated within the system.
• Stocking period (days): The length of time that grazing livestock occupy a specific paddock without interruption.
• Rest period (days): Length of time a specific paddock is not stocked between stocking periods. If selected articles only contained two of the above three criteria, then we used the following formula to calculate the missing criterion: number of paddocks = (rest period/stocking period) + 1 [18].
• Stocking cycle duration (days): Time elapsed between stocking periods on a specified paddock. Stocking cycle duration is the sum of one stocking period plus one rest period.
• Number of stocking cycles per season: Calculated as stocking season duration divided by stocking cycle duration.
• Stocking ratio (%) (own criteria). This is the proportion of time that a paddock is grazed in one stocking cycle, calculated as stocking period divided by stocking cycle duration. For the purpose of simplicity, we name this variable ‘stocking ratio’. A high value of this indicator means less time for plants and soil to recover from grazing.

A dataset was then created in an Excel spreadsheet (version 2503) with all the variables integrated into columns, where each row represents an experimented grazing method within an experiment, with at least two grazing methods per experiment.

2.3. Statistical Analysis: Clustering of Experimental Datapoints

We conducted a principal component analysis (PCA) [124,125] followed by hierarchical Clustering of Principal Components (HCPC) [126], using 7 variables collected from the grid analysis. PCA reduces the dimensionality of data by transforming correlated original variables into 2 new uncorrelated variables, called principal components. HCPC groups individual datapoints into homogeneous groups according to the new dimensions created by these 2 components. During this process, we used two types of variables. First, variables defined as “active” were used to calculate the 2 principal components. Then, variables defined as “illustrative” are projected a posteriori onto the newly created dimensions providing more detailed information to help the description and interpretation of the groups created, without influencing the structure of the clustering.

Of the 7 variables collected from the grid analysis, 3 were used as active variables and 4 as illustrative (

Table 2. Summary of the 7 management criteria used for clustering grazing methods.

Clustering Application	Criteria Used in the Clustering	n	Median	Mean	Min	Max
Management criteria used as active variables	Number of paddocks	249	3	9.46 ± 21.709	1	150
	Rest period (days)	249	26	38.6 ± 60.61	0	450
	Stocking ratio (%)	249	50	56.7 ± 40.43	0.667	100
Management criteria used as illustrative variables	Stocking season duration (days)	237	200	238 ± 113.2	47	365
	Stocking period (days)	249	70	126 ± 139.7	0.500	365
	Stocking cycle duration (days)	249	120	164 ± 141.5	2	781
	Number of cycles in one stocking season (nb)	237	1	2.61 ± 2.652	0.300	15.9

).

The three active variables were chosen, based on their relevance and the fact that they were consistently mentioned in the reviewed articles. The three active variables chosen were number of paddocks (94% of the included papers reported this criterion), duration of the rest period (98%), and stocking ratio (92%). Note too that these three variables are weakly correlated to each other on the correlation matrix (average R² was 0.38), and thus provide a new dimension of information. The remaining four variables (i.e., stocking season duration, stocking period, stocking cycle duration, number of cycles per stocking season) were used as illustrative for descriptive analysis of our clustering.

The PCA and HCPC were conducted using R version 4.3.1 and RStudio version 2023.12.1 software. The FactoMineR package (version 2.10) [127] was used with the PCA() function to perform the PCA and the HCPC function to perform the HCPC. The factoextra (version 1.0.7) and ggplot2 (version 3.5.0) packages were used to create graphs. Descriptive analysis of the different groups obtained from the HCPC used the active and illustrative variables that significantly explained (or tended to explain) each group. A variable significantly explains a group if the average of the individuals in the group is significantly different from the overall average of all individuals [128]. This level of significance, calculated directly by the HCPC function, was set at as a p-value (p) < 0.05.

The number of groups was determined graphically using a dendrogram to illustrate the improvement in inertia explained with each new group created.

3. Results

3.1. Descriptive Statistics of Experimental Datapoints

Analysis of the literature found that analyzed studies had been conducted in North America (60 studies) and Europe (16 studies), followed by Oceania (9), South America (7), Africa (6), Asia (2) and the Middle East (2). Grid analysis on these articles identified 269 method tested in an experiment (experimental datapoints), each presenting distinct method names and management criteria. We decided not to retain datapoints for which data on the three active variables used in the PCA was missing, which resulted in a final sample of 249 methods. The clustering exercise thus focused on these 249 separate datapoints.

For each of the seven selected management criteria, individual datapoints were widely dispersed (Figure 1) into an asymmetrical distribution.

For instance, the average number of paddocks was 9.46 ± 21.709, ranging from 1 to 150 paddocks with an average rest period of 38.6 ± 60.61 days, ranging from 0 to 450 days per cycle. For both criteria, the majority of datapoints (75%) had low values, with less than 7 paddocks and 58 days of rest period. Stocking ratio had a more even distribution between its two extremes (median of 50%), with an average of 56.7 ± 40.43%. The stocking season duration ranged from 47 to 365 days and distinguished between methods where grazing was year-round (98 methods) and methods that integrated a seasonal break (stocking season duration < 365 days, n = 139) which had a symmetrical distribution. The median stocking period was 70.0 days, clustered into two extremes (0.5 and 365 days), whereas 50% of methods applied stocking periods that were either shorter than 7 days or longer than 183 days. The two stocking cycle variables also showed a disparate asymmetrical distribution: median stocking cycle duration was 120 days, but 9 methods had a stocking cycle duration exceeding 1 year. On average, paddocks were grazed at 2.61 ± 2.652 cycles per year, with 52% of the methods using one cycle or less per season, resulting in an overall median of 1.

3.2. Clustering of Experimental Datapoints in 6 Groups

PCA analyses performed on seven main variables and three used as active variables showed that the first two dimensions explained 85% of the total variance (Figure 2).

Dimension 1 (X-axis) was mainly explained by stocking ratio (black arrow at left), with a 42% contribution to that dimension. For Dimension 2 (Y-axis), rest period (black arrow at top) contributed 51%, and number of paddocks (black arrow at bottom) contributed 49%. The fact that the three arrows are not parallel confirms that each of the variables explained a new dimension of information (Figures S1 and S2). However, the four illustrative variables (see Table 1) were less prominently represented, as indicated with their shorter arrows. Stocking season duration was positively correlated with rest period (both arrows point in the same direction), and stocking period was positively correlated with stocking ratio. The two stocking cycle variables show weaker correlations with the three active variables and were negatively correlated with each other (the two arrows point in opposite directions).

HCPC analysis grouped the 249 experimental datapoints into six distinct groups (Figure 3, Table S1).

Group 1 (n = 110) was significantly explained by a low number of paddocks (1.24 ± 0.823; p < 0.05) with almost no rest periods (1.39 ± 6.706 days; p < 0.05) and a stocking ratio close to 1 (98.7 ± 6.14%; p < 0.05) (

Table 3. Summary of the six groups of grazing methods created according to the seven management criteria used for the clustering.

		Group 1 (n = 110)	Group 2 (n = 83)	Group 3 (n = 30)	Group 4 (n = 17)	Group 5 (n = 4)	Group 6 (n = 5)
Grazing method family		Continuous grazing	Conventional rotational grazing	Deferred rotational grazing	Adaptative multi-paddock grazing	Deferred rotational grazing	Adaptative multi-paddock grazing
Number of paddocks	Mean	1.24	7.26	4.76	46.0	3.03	136
	Sd	0.823	4.901	3.151	13.18	2.696	14.3
	Min	1	2	2	25.5	1	120
	Max	5	25	15	61	7	150
Rest period (days)	Mean	1.39	36.7	104	73.0	398	93.2
	Sd	6.706	16.57	32.5	20.90	42.5	41.83
	Min	0	1	60	45	362	44.8
	Max	41.5	84	225	120	450	149
Stocking ratio (%)	Mean	98.7	20.3	46.5	3.42	28.0	1.05
	Sd	6.14	11.92	21.86	3.235	24.23	0.303
	Min	66.8	4.65	6.67	1.52	0.822	0.667
	Max	100	50	76.5	15.4	50	1.41
Stocking season duration (days)	Mean	231	208	283	342	365	263
	Sd	113.6	106.7	104.4	55.6	0.0	176.1
	Min	47	47	90	217	365	60
	Max	365	365	365	365	365	365
Stocking period (days)	Mean	234	9.74	132	2.73	202	0.900
	Sd	117.6	9.147	126.2	3.268	190.5	0.2236
	Min	43.5	1	8	1	3	0.5
	Max	365	47	365	15	365	1
Stocking cycle duration (days)	Mean	235	46.4	236	75.7	600	94.1
	Sd	115.9	22.11	138.7	21.91	191.9	41.97
	Min	47	2	113	47	365	45.3
	Max	365	100	485	122	781	150
Number of stocking cycles per stocking season	Mean	1.02	4.86	1.45	4.72	0.675	2.23
	Sd	0.235	3.172	0.807	1.359	0.2363	0.862
	Min	0.3	1	0.7	3	0.5	1.3
	Max	3.2	15.9	3.2	6.3	1	3

). In terms of other criteria used as illustrative variables, these grazing methods used fewer stocking cycles per stocking season (1.02 ± 0.235; p < 0.05) and long stocking periods (234 ± 117.6 days; p < 0.05) approaching the entire length of the stocking season (231 ± 113.6 days; trend). Representative datapoints for the group are named: permanent grazing, continuous grazing, and continuous stocking.

Group 2 (n = 83) was significantly explained by an intermediate stocking ratio (20.3 ± 11.92%; p < 0.05), related to rotational methods using an average of 7.26 ± 4.901 paddocks (trend) and rest periods of 36.7 ± 16.57 days (trend). Group-2 methods had shorter stocking periods (9.74 ± 9.147 days; p < 0.05) than the overall average methods (126 ± 139.7 days) and a higher number of cycles per season (4.86 ± 3.172; p < 0.05) than the overall average methods (2.61 ± 2.652). Representative datapoints for the group are named: rotational grazing, 6-paddock rotational grazing, and rotational grazing management.

Group 3 (n = 30) was significantly characterized by long rest periods (104 ± 32.5 days; p < 0.05), resulting in systems where animals rotate on a relatively low number of paddocks (4.76 ± 3.151; trend) in extended stocking periods (132 ± 126.2 days; trend) that maintain a balanced stocking ratio (46.5 ± 21.86%; trend) and fewer cycles per season (1.45 ± 0.807; p < 0.05). Representative datapoints for the group are named: Santa Rita grazing system, Deferred-rotation grazing and 2-pasture 1-herd deferred-rotation grazing system.

Group 4 (n = 17) was characterized by a high number of paddocks (46.0 ± 13.18; p < 0.05), long rest periods (73.0 ± 20.90 days; p < 0.05), low stocking ratios (3.42 ± 3.235%; p < 0.05), and short stocking periods (2.73 ± 3.268 days; p < 0.05). Methods in Group 4 had more paddocks than the overall average methods (9.46 ± 21.709), enabling them to apply shorter stocking periods and longer rest periods than the overall average methods (38.6 ± 60.61 days). Representative datapoints for the group are named: adaptative multi-paddock grazing.

Group 5 (n = 4) used long rest periods (398 ± 42.5 days; p < 0.05) and was also characterized by long stocking periods (202 ± 190.5 days; trend), which translates into stocking cycles scheduled over several years (600 ± 191.9 days; p < 0.05). Representative datapoints for the group are named: high-density short duration grazing or conventional grazing.

Group 6 (n = 5) was significantly explained by a large number of paddocks (136 ± 14.3; p < 0.05) and long rest periods (93.2 ± 41.83 days; p < 0.05), leading to a low stocking ratio (1.05 ± 0.303%; p < 0.05) and resulting in short stocking periods (0.900 ± 0.2236 days; p < 0.05). Representative datapoints for the group are named: 120-pasture, ultra-high stocking density system or adaptative multi-paddock grazing.

This work correctly separated continuous methods (Group 1) from rotational methods (Groups 2 to 6), as commonly recommended in the literature [13,16,17,18]. Among rotational methods, based on management criteria and the names associated with each data point, we identified three main families of grazing methods: conventional rotational grazing (Group 2), deferred rotational grazing (Groups 3 and 5), and adaptive multi-paddock grazing (Groups 4 and 6). We further discuss below the rationale behind this clustering.

4. Discussion

This work grouped 249 descriptor-names of grazing methods into six groups and four broad families according to grazing management criteria. This highlights that common patterns can be identified across the diverse range of grazing method terminologies worldwide, thus reducing the complexity of the diverse terms used globally.

4.1. Definition of the Four Grazing Method Families

Separation between continuous methods (Group 1) from rotational methods (Groups 2 to 6 is illustrated by the average number of paddocks, which is low for Group 1 methods (1.24 ± 0.823) compared with the overall average for the rotational groups (16.0 ± 27.39). Furthermore, in the literature, 75% of Group 1 methods were named using the word “continuous’’. Therefore, based on their management criteria, we can define Group 1 as “continuous grazing”. Group 1 appears in the studies here cited, defined by different terms: “intensive free grazing” by Leray et al. [18], “continuous stocking” by Allen et al. [5], and “yearlong” or “continuous-seasonal” by Lacey and Poollen [21].

Datapoints in Groups 2 and 3 used a close average number of paddocks (respectively, 7.26 ± 4.901 and 4.76 ± 3.151) but differed in their other management criteria. Group 3 methods were characterized by longer alternating rest periods and stocking periods than Group 2 methods, but also than overall rotational methods (respectively, 68.1 ± 67.73 days of rest and 40.4 ± 89.94 days of stocking). Furthermore, in the literature, 47% of Group 3 methods were named using the word “deferred”. This practice, where part of the system is seasonally removed from the grazing management unit [5,129], results in extended rest periods. These methods do not necessarily rotate systematically between paddocks [5], and sometimes multiple paddocks are grazed simultaneously [80], which is consistent with the longer stocking periods observed in this clustering. We defined the 30 methods in Group 3 as “deferred rotational grazing”. This group corresponds to the definition of “deferred stocking” as proposed by Allen et al. [5] as “A method to defer grazing on land units that may or may not be in a systematic rotation with other land units”. However, as these methods are not commonly employed in Europe, they are not developed by Leray et al. [18] for the French context.

Group 2 methods have intermediate values for 5 on the 7 criteria studies (number of paddocks, rest periods duration, stocking ratio, stocking period, and stocking cycle duration). This is particularly evident for the stocking cycle (20.3 ± 11.92%), which is close to the overall average of rotational methods (23.4 ± 19.71%). Therefore, we have defined this group as “conventional rotational grazing”. These methods align with the definition of “rotational grazing” as described by Leray et al. [18] or “rotational stocking” as proposed by Allen et al. [5] as “method that utilizes recurring periods of grazing and rest among three or more paddocks in a grazing management unit throughout the time when grazing is allowed”. However, in this article, alternative definitions such as “set stocking”—“a method that allows a specific, non-variable number of animals on a specific, non-variable area of land during the time when grazing is allowed”—or “mob stocking”—“a method of stocking at a high grazing pressure for a short time to remove forage rapidly as a management strategy”—may be analogous to our Group 2. Group 4 methods are characterized by extended rest periods. However, unlike Group 3 methods, which have the same characteristic, this was achieved without deferment. To achieve this, a high number of paddocks were used (46.0 ± 13.18), which exceeds the overall average for rotational methods (16.0 ± 27.39). This results in shorter stocking periods (2.73 ± 3.268 days) compared to the overall average for rotational methods (40.4 ± 86.94 days). In the literature here used, 53% of Group 4 methods are referred to as “adaptive multi-paddock grazing”, and the management criteria cited are consistent with the principles underpinning adaptive multi-paddock grazing presented in various papers [20,34,130]. This method aims to support grassland ecosystem functions—such as soil and plant health and animal well-being—by relying on high stocking densities, a large number of paddocks, short grazing periods during which plant consumption is monitored, and extended rest periods to allow vegetation to regrow. An dynamic adjustment of livestock number or/and paddock size is central to the method to not exceed available forage and to avoid overstocking and overgrazing [34]. We, therefore, defined this family as “adaptative multi-paddock grazing”. Group 4 has not been precisely defined in previous studies [5,18,21]. This group may align with the “short duration-seasonal” or “short duration-rotation” categories proposed by Lacey and Poollen [21]. Notably, the methods currently referred to as “adaptive multi-paddock grazing” were previously designated as “short duration grazing” by their proponent, Allan Savory [131]. However, the definitions proposed by Lacey and Poollen [21] could also be applicable to all Group 2 methods, provided that the number of cycles per stocking season exceeds 2 (short duration-seasonal) or 3 (short duration-rotation). The fact that our analysis highlights this family of methods supports that they offer an innovative approach to grazing management compared to other rotational methods.

Groups 5 and 6 appeared relatively anecdotal, as they only counted a small number of datapoints (4% of rotational methods) coming from a very low number of articles (just two articles for Group 5 and three articles for Group 6), and so we elected not to define them independently. Group 5 can be aligned to Group 3, as it follows a similar strategy (low number of paddocks, long stocking and rest periods) but on a multi-year scale as it has an average stocking cycle duration of 600 days. Group 6 can be aligned to Group 4 as it again follows a similar strategy (high number of paddocks, short stocking and rest periods) but with an even higher number of paddocks and even shorter stocking periods.

4.2. Robustness of Analysis and Factors Influencing Clustering

Our analysis is based on management criteria that are objective variables, allowing it to work on an international scale where other approaches could be limited by pedoclimatic contexts. For example, studies generally use grazing intensity to compare different methods on their service provision performance [62,91,92,109]. However, the definitions can vary significantly between different locations according to stocking season duration and the annual production of pasture biomass [132]. For example, two studies conducted in 2017 in New South Wales, Australia, reported different stocking rates depending on local climate conditions. Badgery et al. [25] applied a stocking rate of 8.8 to 13.6 DSE/ha in a wet area (777 mm annual rainfall, 13.8 °C), while Orgill et al. [77] used 0.9 DSE/ha in a dry area (382 mm, 20.8 °C).

Regarding the results of our work, by considering only three active variables, we may have overestimated the robustness of the group boundaries (Table 2). That involves assessing the extent of overlap and the potential for confusion between groups. Looking at the projection of the six groups on the correlation circle (Figure 3), the datapoint clouds of Groups 1, 4, 5, and 6 are distinct and do not overlap, whereas Groups 2 and 3 share overlap. The dendrogram used to define the number of clusters in the HCPC reveals that Groups 2 and 3 were the last to be separated during the transition from 5 to 6 clusters. This indicates that their separation contributes less to the gain in explained inertia than the earlier splits. Therefore, there is a risk of confusion between these groups, especially at the boundaries of certain management criteria. This is the case for the rest period, where the maximum value for Group 2 was 84 days and the minimum for Group 3 was 60 days, whereas Group 3 was statistically characterized by a high average rest period value. The datapoint cloud of Group 1 was clustered on the circle of correlation, which means that the management criteria of Group 1 methods are not widely dispersed. For example, number of paddocks in Group 1 varied from 1 to 5. Conversely, the datapoint cloud of Group 2 was very scattered, with 2 to 25 paddocks. It could be argued that Group 2 should be subdivided, but the inertia gained by adding groups was very low after six groups, and so more individuals would be needed for further clustering.

Our analysis was based only on grazing methods that had served for comparative analyses in the literature, which limited the number of methods clustered. The sample could have been extended by including studies that focused on a single grazing method and described the associated management criteria. This would have enabled us to expand the dataset, but also to obtain data for methods established in major grazing nations such as Ireland and New Zealand for which we found practically no comparative analyses (zero for Ireland and one for New Zealand). The same hypothesis can be made concerning the underrepresentation of Asian studies (two studies), despite Asia being a major contributor to grazing research, with extensive grassland areas and long-term grazing management histories. However, processing all these studies on grazing would have been unfeasible. For example, a Web of Science keyword search on “grazing” gives 89,336 results on 25 June 2024.

Note that in some studies, the stocking season criteria aligned with the duration of the study-trial. Non-experimental groups typically grazed on the paddocks either before or after the experimental period [41,74], which limited the development of the notion of seasonal break (seasonal break = 365 days—stocking season duration) in our analysis. In fact, when the animals are removed from the grazing management unit, the system enters a state of rest, generally due to unfavorable climatic conditions slowing pasture growth (winter in temperate climates, or dry/humid seasons for other climates). This criterion was not used in our analysis but can be useful to investigate [133] and could be complementary with the rest period applied during the stocking season [134].

The approach developed here remains limited given the complexity of each grazing system. We can assume that a grazing system is explain not only by the grazing method applied but also by environmental factors (such as climate and vegetation) as well as social and geographical norms [135,136] and by the interactions between them. Climates less favorable to pasture growth may prompt technical and research institutes to test out more innovative methods. Likewise, some grazing methods may be more suited to regions where pasture growth is abundant or lasts longer, whereas other grazing methods may be better suited to areas with limited or sporadic grass growth. This geographical and environmental specificity may, therefore, influence the applicability and effectiveness of different grazing methods. For instance, here, the lowest average annual cumulative rainfall was at sites where grazing method was defined as “deferred rotational grazing” (Group 3) and 9% lower than the average across all study over the period 1990–2020. In Ireland and New Zealand, on the other hand, where we found few studies comparing grazing methods, grazing is organized around a main strategy that can be summarized as ‘one paddock per day’ [137,138,139,140]. Both Ireland and New Zealand have a climate that is favorable to pasture growth [141,142], and so as their model works, there is little incentive for technical and research institutes to test out new methods.

4.3. Perspective of the Review

Grasslands are important supports for sustainability, notably through the ecosystem services they provide, such as carbon storage, biodiversity conservation, water supply, and flow regulation [3,4]. Poor grazing management leads to the degradation of grasslands worldwide [143], particularly due to overgrazing, which can compromise their ability to deliver essential ecosystem services [144]. In this context, it is important to understand which grazing practices promote sustainable land management. More specifically, it is relevant to study in detail the links between grazing method families and the ecosystem services they provide, using a multiservice approach. Multiservice approaches are especially relevant today given global climate change and efforts to reduce the environmental footprint of agroecosystems [145,146,147].

This work has not been conducted at an international scale and is challenging to carry out due to some extent to the lack of homogeneity in the definition of grazing methods. The review developed here could serve as a support for characterizing and clustering different grazing methods developed in different contexts into a reduced number of groups in order to conduct this future study.

5. Conclusions

This review, conducted across a wide range of grazing methods from the literature, identified six groups of grazing methods based on multivariate analysis of their management criteria (number of paddocks, rest period, stocking ratio). The six groups created reflect the separation found in several studies between continuous and rotational grazing, and, in particular, highlight four broad families of methods that we defined as “continuous grazing”, “conventional rotational grazing”, “deferred rotational grazing”, and “adaptative multi-paddock grazing”. This analysis highlights that common patterns can be found across the diverse grazing methods used worldwide when consistent criteria are applied. Grazing is central to major grassland management issues, particularly due to the diversity of ecosystem services they provide (i.e., carbon sequestration, biodiversity, water cycle). This approach can provide a support to help researchers and practitioners understand and incorporate grazing methods into their studies. More specifically, it could provide an objective basis to study the relationship between the chosen grazing method and the level of ecosystem services provided, even though this work may be limited by complexities associated with environmental and social norms.