Introduction

Arid and semi-arid rangelands are widely used for livestock husbandry in extensive production schemes, usually implementing continuous grazing. In these systems, livestock grazes freely throughout the year in large fenced paddocks with scarce watering points, often in coexistence with wildlife. From a production perspective, these grazing areas can only sustain a maximum number of herbivores, i.e. the carrying capacity, before the forage resources begin to deteriorate (Scarnecchia 1990). This carrying capacity is determined by the feeding area and forage availability resulting from the joint distribution of water sources, vegetation types and other landscape features such as topography, soil type or presence of predators, which can also influence the spatial distribution of the animals (Coughenour 1991). For example, because of their water requirements, the opportunity for animals to graze is defined by the intervals between drinking events. In this sense, in arid and semi-arid environments, livestock spends more time near the water points to maximize feeding and minimize travel time from the grazing sites to watering points (Pringle and Landsberg 2004). Thus, in the driest period of the year, with highest animal’s water requirements, distance to water points can totally or partially undermine the effects that other variables may have on animal distribution (Ogutu et al. 2014). On the other hand, water requirements also vary between species, conditioning their distribution and access to the resources in different ways (Derry and Dougill 2008; de Leeuw et al. 2001), thus their relationships regarding carrying capacity are expected to be different too. The restriction in the spatial distribution determined by the location of the water points generates a gradient of stocking pressure that translates into a gradient of forage use (i.e. piosphere) (Pringle and Landsberg 2004). The forage consumption decreases with increasing distance to watering points and extends to the maximum distance that an animal can travel before needing to drink again (Trash and Derry 1999; Derry 2004). Consequently, there is an inverse gradient of forage not consumed by livestock that can be potentially used by other herbivores with fewer water restrictions and capable of occupying these areas. Such complementarity in the use of space and vegetation between ungulates, both native and domestic, with different ecological characteristics has been described in diverse environments and proposed as an optimization mechanism in grassland use (Mentis 1978; Dekker 1997).

Although the carrying capacity estimation has become an important tool for rangeland planning and management, and a variety of methods have been developed to its assessment, they frequently assume a uniform livestock distribution in space. When several species coexist, these methods also assume equivalences between them based on body weight (Hardy 1996), and rarely incorporate the specific spatial distribution of the animals and/or differences in foraging behavior (Dekker 1997). This simplification can result in an overestimation of the carrying capacity if the distribution of animals is restricted (Holecheck 1988) or an underestimation if species with different diets are present in mixed systems (Dekker 1997). These shortcomings can have important consequences on environments that impose significant constraints on grazing distribution, or where species with very different adaptations coexist (such as domestic and native herbivores), contributing to land overgrazing, failure of economic development and decline in wildlife populations (Coughenour 1991).

In arid Patagonia, extensive livestock husbandry is carried out in natural grasslands and shrublands. With the introduction of sheep in the region at the beginning of the twentieth century, sheep-grazing heterogeneity and carrying capacity overestimation resulted in massive environmental degradation and progressive detriment of animal production indices (i. e. sheep recruitment, wool and meat production) (Golluscio et al. 1998; Coronato 2015). This situation stressed the need for technical support in the assessment of the carrying capacity and the corresponding livestock management proposals. Currently, several methods that were developed for the dominant vegetation type in different regions are used to assess carrying capacity and adjust sheep stocks (Golluscio 2009; Massara Paletto and Buono 2020). Although the authors of these methods acknowledge the heterogeneity in animal spatial distribution and the presence of wild herbivores in Patagonian environments, they do not incorporate any of these aspects when estimating carrying capacity.

The guanaco (Lama guanicoe) is the main native herbivore in the region and co-exists with livestock in rangelands. Since the European colonization, their populations were significantly reduced due to hunting, habitat loss and competition with sheep (Baldi et al. 2006), although, their numbers have started to recover in some locations during the last decade (Nugent et al. 2006). There is a severe conflict with regards to livestock production because ranchers perceive guanacos as the main livestock competitors and as a major overgrazing risk (Lichtenstein et al. 2022; Schroeder et al. 2022). The guanaco was pointed to out-compete livestock for forage, although this situation necessarily implies that there is a limiting amount of forage, a common use of space and dietary similarity between both herbivores (Pianka 1974). However, the analysis of the spatial segregation between guanacos and sheep, and the occupation patterns of guanacos in the presence and absence of sheep indicate that guanacos are displaced (i.e. competitive exclusion) to areas that livestock does not occupy (Marino and Rodríguez 2022; Schroeder et al. 2014, 2022). There is also wide variability in the corresponding diet overlap, and guanacos display a flexible diet according to plant species and biomass availability (Raedeke 1979), time of year (Puig et al. 2001; Baldi et al. 2004) and whether both herbivores are in sympatry or not (Raedeke 1979; Pontigo et al. 2020). The understanding of the spatial distribution patterns of guanacos and livestock and their relationship with their use of resources at the ranch scale is, therefore, essential in framing the arguments that support this conflict and developing adequate carrying capacity assessment and management tools for Patagonian rangelands.

The pastoral value method (Elissalde et al. 2002) is one of the most widespread methods used for the assessment of rangeland’s carrying capacity in central Patagonia. This method assumes that domestic herbivores are distributed across the entire paddock proportionally to the forage supply [“ideal free distribution” (Tyler and Hargrove 1997)]. This postulation is reflected in the computation of total available forage, which is carried out by compounding the forage consumed by sheep the year before the assessment, as well as the forage quality and availability measured at the field in “key areas” (i.e. reference areas with intermediate forage use) (Holechek 1988) at each paddock (Escobar et al. 2020). However, this method does not take into account the spatial distribution of the animals, even though its inclusion could determine areas highly underused, reducing real paddock receptivity; and it also neglects the potential presence of wild herbivores, such as guanacos. To address these issues, we selected a sheep ranch with a typical extensive grazing system, in which a previous assessment of the carrying capacity according to the pastoral value method was available. In this ranch, livestock and guanacos showed a heterogeneous spatial distribution in relation to their conditioning factors which determined species segregation mediated by competitive exclusion (Marino and Rodríguez 2022), so we hypothesized that (i) sheep distribution is spatially limited and shows a gradient of pasture use that result in over- and under-grazed sites; (ii) guanacos are also distributed heterogeneously and mainly occupy the areas underutilized by sheep; and (iii) the density of guanacos is determined by the portion of rangeland-carrying capacity unused by sheep; (iv) the inclusion of these patterns in CC assessment and herbivore stocks analysis modifies classical outcomes respect to the use of the forage and detrimental effects on the rangelands. Under this framework, we aimed to address (1) how the assessment of the carrying capacity deviates from the classic outcome if the effective space use by sheep is considered in the analysis; (2) how the carrying capacity assessment varies if the forage consumption by guanacos present in paddocks is included and the guanaco-livestock equivalent is corrected for the diet overlap reported for both species; and to evaluate (3) the guanaco densities regarding unused forage by livestock based on the estimated mixed carrying capacity and the distribution of sheep and guanacos.

Materials and methods

Study area

The study was carried out in five paddocks of a traditional sheep ranch located in the southeast of the province of Chubut, Argentina. This ranch is devoted to Merino sheep husbandry and is located in one of the most productive regions of the province. The ranch is subdivided into large paddocks by fences and has a considerable number of artificial and natural water points (Fig. 1a, Table 1). Sheep are distributed by classes (ewes, lambs, wethers) in these paddocks and graze freely and continuous throughout the year. Animals graze exclusively forage of natural grasslands and shrublands, without food supplementation. Sheep stocks are adjusted according to each paddock receptivity assessed by pastoral value method (Elissalde et al. 2002)(Table 1).

Fig. 1
figure 1

Study area with paddock numbers and sampling units (a), and vegetation units (b). See “Study area” for vegetation unit (VU) description

Full size image

The geomorphology of the area corresponds to rocky hills with reduced associated foothills. Canyons shape a subdendritic drainage network and coastal cords. The altitude of this unit with respect to sea level varies between 0 to 200 m. The soils are Calciorthids, Haplargids (Aridisol) and Torriorthents (Entisol). The mean annual precipitation is 250 mm concentrated in the autumn–winter (Beeskow et al. 1987) but the region shows high inter and intra-annual variability typical for the arid zones (Colombani 2012). The vegetation of the area is characteristic of the Patagonian phytogeographic province which is composed of grasslands and shrublands. For this region the dry season, which is coupled with higher temperatures, occurs from December to March. It is considered a critical moment for sheep husbandry mainly because vegetation has null or very low growth rates (Campanella and Bertiller 2008). The forage availability in this period would influence the amount of wool produced per individual, wool fineness and ewe’s body condition at rut time (Colombani 2012).

Vegetation units’ description and pastoral value

Based on a classified Landsat TM satellite image and field vegetation surveys, Gauna et al. (2012) identified and mapped five dominant vegetation units (VU) within paddocks (Fig. 1b) for which they assigned a relative measure of forage quality and availability defined as the pastoral value (PV) (Daget and Poissonet 1971). The PV varies between 1 and 100 and it is computed from specific plant cover, a specific quality index and the total forage cover (Elissalde et al. 2002). The VU and associated PV were classified as follows:

VU 1 is rocky or bare areas, without vegetation or very scarce. PV = 0.

VU 2 is a shrub steppe of Chuquiraga avellanedae. It is found mainly in the sectors of hills with slight to pronounced slopes. The differences between VU 2 and other VUs are mainly associated with characteristics of the soil, i.e. a greater amount of gravel on the surface and rocky outcrops. Chuquiraga avellanedae predominates among the shrubs, although some sectors have a more herbaceous environment with the presence of Stipa humilis and Festuca argentina. PV = 9.49.

VU 3 is a tall shrub steppe. It has a good percentage of Ch. avellanedae cover, while other tall shrub species such as Lycium ameghinoi, Lycium chilense, Schynus poligamus and Trevoa patagonica also stand out. This unit has both good total cover and forage cover. It is observed mainly in sectors of intermediate slopes. PV = 10.43.

VU 4 is a shrub steppe. Its floristic composition is similar to VU 3 but with a higher cover of Nassella tenuis. There are areas with predominant scattered bushes and more defined sub-shrub strata, which in some cases are associated with lowlands, foothills with low cover and runoff sectors connected to temporary channels that flow into the sea. PV = 11.38.

VU 5 is an herbaceous-shrub steppe. This vegetation community has a high cover of N. tenuis. PV = 13.62.

Herbivore surveys

Large herbivore surveys were conducted by traveling on available roads and tracks (Fig. 1a). The guanaco and sheep populations at the ranch were surveyed on four occasions: in December and March 2017, March 2019 and January 2020. Some transects had to be shortened due to gravel roads blockage after heavy rains thus we traveled 43, 63, 64 and 75 km respectively, surveying between 17 and 30% of the study area. Data collection was performed by two observers located at the back of a pick-up vehicle and was based on line-transect surveys in which species (sheep or guanaco), group sizes and composition, as well as other relevant data including location, azimuth and distance to the vehicle by a laser range finder, were recorded.

Predictor variables, animal distribution and carrying capacity analysis

Sampling units and explanatory variables

In order to assess the main factors shaping sheep and guanaco distribution within the study area, we used QGIS 3.8 software to convert survey transects into bands of 500 m width (250 m on each transect side) that we subsequently chopped them into 500 m long segments, obtaining a series of contiguous polygons of approximately 0.25 km2 (Fig. 1a). These polygons were the sampling units that allowed us to assess the relationships between the counts of herbivore groups and the explanatory variables. Based on a bibliographic information (Marino and Rodríguez 2022), a preliminary set of five explanatory variables was explored to predict sheep and guanaco distribution: distance to drinking water, PV of vegetation, valley depth, and average terrain slope and paddock area.

We estimated the PV for each sampling unit on the basis of the VUs map built according to the Landsat TM satellite image (30 × 30 m pixel) classified by Gauna et al (2012) (Fig. 1b). We computed a weighted average accounting for the PV score of each VU included in the polygon and its corresponding area. We obtained terrain slope and valley depth data from a digital elevation model (25 × 25 m pixel) (Macandza 2009) using GIS tools from the GDAL and SAGA libraries, and computed average values per polygon.

Only the water points that maintain water throughout the year (natural springs, pumping waterholes) were considered in the present analysis. Although in these paddocks there are temporary water points, their effect on livestock distribution in the dry season would be limited due to their short availability period, even if the animals were able to re-visit their location (Western 1975; Derry and Dougill 2008; Naidoo et al. 2020). The water points were geolocated using handheld GPS (Garmin Etrex), which allowed us to calculate Euclidean distances from polygon centroids to the nearest water point within each paddock.

Estimation of the distribution and densities of sheep and guanacos

Sheep and guanaco distribution

To assess the factors affecting sheep and guanaco distribution within paddocks, we fitted Negative Binomial models separately to the counts of sheep and guanaco groups observed per sampling unit. This approach has been successfully used to assess the intensity of use in other species (Nielson et al. 2016). The detection probability for both guanacos and sheep during post-reproductive surveys was more than 80% at 250 m away from the transect line. Sheep detection decreased considerably beyond this point, and therefore, observations beyond 250 m were discarded to minimize imperfect detection effects. We also discarded counts of groups closer than 150 m from water points to avoid the effect of groups that were drinking water instead of grazing (sacrifice areas). Although sampling-unit (polygons) sizes were roughly similar, the log-transformed area of the polygon was included as an offset to compensate for the known variation in the response resulting from differing polygon sizes (Kéry 2010). The fixed effects tested included PV, terrain slope, valley depth, distance to the nearest water point in the paddock and paddock area. Survey date was also included as a fixed factor to account for potential differences between surveys. The polygon ID was included as a random effect to account for the lack of independence among observations recorded on the same polygon in successive surveys (Yirga et al. 2020). After model fitting, a preliminary ranking of the relative performance of competing models according to Akaike Information Criterion (AIC) scores (Burnham et al. 2010) was conducted using the MuMIn R package (R Core Team 2022, www.R-project.org), keeping all reasonably well-fitting models (e.g. ∆AIC < 2). The Moran I test (Moran 1950) was used to discard models with significant spatial auto-correlation among Pearson residuals. We averaged across the subset of well-fitting models (Bolker et al 2009), accounting for model weight, using the MuMin R package (Barton 2009) to predict the spatial distribution of sheep and guanacos across a prediction grid of the explanatory variables (see next section). In the case of sheep, we used the prediction grid to guide the distribution of the known stock (ranch records) across the cells of each corresponding paddock on a proportional basis. Guanaco densities recorded during surveys included an average 14% (DE = 2) chulengos (i.e. young of the year) which were considered as 0.34 of an adult weight (Von Thüngen 2003), yearlings were considered as adult guanacos. Model fitting and statistical tests were performed with glmr, MuMin and spdep R packages (R Core Team 2022, www.R-project.org).

Prediction grid

Within a GIS platform, we built a prediction grid of 500 × 500 m cells. This grid covered all the studied paddocks and each cell included information about the same explanatory variables considered for the sampling units. In all cases, cells smaller than 0.1 km2 were discarded. The final grid was used to map sheep and guanaco densities predicted by the distribution models and to calculate the carrying capacity at different spatial scales.

Carrying capacity assessment, percentage of grazing and residual carrying capacity

We computed the carrying capacity for each prediction grid’s cell according to the equations of the pastoral value method (CCpv) (Elissalde et al. 2002; Escobar et al. 2020) as:

$$ CCpv \, \left( {OU} \right) = TUF \, \left( {kg \, DM} \right)/330 \, \left( {kg \, DM*OU^{ - 1} } \right) $$

where TUF is the total usable forage and OU represents one ovine unit equal to one 40 kgLW Merino wether consuming 330 kg dry matter (DM) per year (Massara Paletto and Buono 2020). The TUF was calculated as:

$$ TUF \, \left( {kg \, DM} \right) = \left( {FA + Consumed \, forage} \right)*Use \, factor $$

where the use factor is 35% (Escobar et al. 2020) and FA is the forage availability at the field, calculated as:

$$ FA \, \left( {kg \, DM} \right) = 12.26*PV $$

where 12.26 is a value obtained from linear regression models between the forage evaluation transects (cover) with forage biomass (harvest) for the ecological area of the coast of Chubut (Elissalde et al. 2002; Escobar et al. 2020). The PV is the pastoral value (adimensional: 0–100) and is calculated from field data as:

$$ PV = 0.2*\sum {\left( {CFSi* \, SIi} \right)*TFC} $$

where CFSi is plant cover for each forage species i (0–100%) and SIi is the specific quality index of each species i (adimensional: 0–5) (Massara Paletto and Buono 2020) and TFC is total forage cover (%).

The constant 0.2 is used to keep the range of PV between 0 and 100.

Finally, the consumed forage is calculated as:

$$ {\text{Consumed forage }}\left( {\text{kg DM}} \right){\text{ = sheep stock during the previous year }}\left( {{\text{OU}}} \right) \, * \, 330 \, \left( {{\text{kg DM }}*{\text{ OU}}^{ - 1} } \right) $$

We used the VUs map as a raster layer with their associated PV to estimate FA at each grid cell. The PVs used were estimated by Gauna et al. (2012) with field data for an average year. The consumed forage was calculated using sheep stocks present in each paddock during the previous year according to the information provided by the rancher. In order to account for differences in sheep classes among paddocks we express sheep stocks and the CC in ovine units (OU).

Since we estimated CCpv for an average year, we accounted for inter-annual rainfall variability by considering a 15% range around the estimated value as proposed by Buono (2020).

Based on sheep stock at each cell predicted by the distribution models and CCpv assessed for each cell (CCcell), we computed the percentage of sheep grazing as:

$$ \% {\text{ Grazing}} = \left( {{\text{sheep stock in cell}} - {\text{CCcell}}} \right)/{\text{CCcell}}*100 $$

This measure led us to evaluate the relationship of herbivore stocks with respect to CC values. Values greater than 15% were considered as overgrazing, and values lower than − 15% as undergrazing. Intermediate values were considered at the CC.

Based on the % grazing, we then computed the CC corrected for the degree of livestock use (CCu) in each cell according to the following criteria:

  • For cells with % grazing ≥ 0 (all available cell forage is used for livestock), the CCu is equal to the CCpv of the cell.

  • For cells with % grazing < 0 (available cell forage is underused by livestock due to limited spatial distribution), the CCpv was adjusted for the percentage of utilization (Holecheck 1988, Valentine 1947) based on sheep stock predicted by models for each cell, as follows:

$$ {\text{CCu}}\, = \,{\text{CCpv }}* \, \left( {{{\text{predicted sheep stock}} \mathord{\left/ {\vphantom {{\text{predicted sheep stock}} {{\text{CCpv}}}}} \right. \kern-0pt} {{\text{CCpv}}}}} \right)\, = \,{\text{predicted sheep stock}} $$

To calculate the CCu for the entire paddock, we added the CCu of each cell within the paddock.

The residual CC is the OUs estimated in the CCpv but that are not being used by the sheep according to the estimated CCu:

$$ {\text{Residual\,CC}}\, = \,{\text{CCpv}}{-\!\!-}{\text{CCu}} $$

Estimation of the carrying capacity considering the guanacos present in each paddock

Using the pastoral value method previously described, we calculated a mixed carrying capacity considering the forage consumed by sheep and guanacos (CCm) as:

$$ {{CCm}} \, = {{TUFm}} \, \,\left( {{{kg}} \, {{DM}}} \right) \, \,/\,{{330}} \, \,({{kg}} \, {{DM}}*{{OU}}^{{ - {{1}}}} ) $$

where TUFm is the total usable forage for sheep and guanacos, calculated as:

$$ {\text{TUFm}}\left( {{\text{kgDM}}} \right) = \left( {{\text{FA at field}} +\,{\text{Consumed forage by guanacos}} +\,{\text{Consumed forage by livestock}}} \right) \times {\text{Use factor}} $$

To calculate the forage consumed by guanacos, we used values of guanaco density estimated from a survey carried out in 2008 since it is a date closer to CC assessment made by Gauna et al (2012) for the ranch. Guanaco density was 1.9, 44.1, 33.2, 2.4 and 2.2 guanacos*km−2 for paddock 1, 2, 3, 4 and 5 respectively. Periodical assessments suggest that guanaco density at the ranch has been stable for at least two decades (Baldi et al. 2001; Marino and Rodríguez 2022). We calculated an animal equivalent adjusted for dietary similarity (Johnson 1979) of 1 guanaco = 1.15 OU, assuming a dietary overlap of 71% (computed for the site by Baldi et al 2004), a live weight of an adult guanaco of 75 kg and a daily consumption equivalent to 2% of the guanaco live weight (San Martin 1987). Using the CCm estimation, we calculated an adjusted mixed carrying capacity for livestock (CCmu) incorporating the degree of use in each cell by sheep as described in the previous section.

Our methodological approach with the main characteristics of the different CC estimated in this work is summarized in Fig. 2.

Fig. 2
figure 2

Summary of the main characteristics of the different carrying capacity estimations (CC) computed in this work

Full size image

To explore the relationship between guanacos and sheep grazing, we evaluated graphically the guanaco density regarding the percentage of sheep grazing derived from the distribution model. In addition, we evaluated graphically the relative portion of the CCm used by guanacos and sheep through the relationship between the percentage of the total grazing (total OU (guanacos + sheep)—CCm)/CCm * 100 and percentage of sheep grazing in each of the cells (see Sect. “Carrying capacity assessment, percentage of grazing and residual carrying capacity”).

Results

Spatial distribution of sheep and guanacos

The best models subset (delta AIC < 2) regarding sheep counts included the distance to drinking water and paddock area, and variable combinations of PV, valley depth and terrain slope (Online Appendix 1). Guanaco counts were affected by the same variables that affected sheep but with opposite relationships (Online Appendix 1). Figure 3 shows the variability of these variables in the prediction grid for the study area. Among the variables tested, distance to drinking water and paddock size were present in all the models fitted for sheep counts, and almost all of the guanaco models, indicating the strongly influence of these factors in their spatial distribution (Online Appendix 1).

Fig. 3
figure 3

Vector grids for the surveyed paddocks. Distance to the water point, pastoral value, valley depth and mean terrain slope. The light blue circles indicate the position of water points and the dotted lines indicate the traveled transects

Full size image
Table 1 Main metrics of sampled paddocks
Full size table

Sheep density decreased with increasing distance to water and paddock size, whereas it increased in areas with higher PV or bottom valleys (Table 2, Fig. 4a). On the contrary, the guanaco counts increased far from drinking water and in larger paddocks whereas decreased at areas with higher PV or bottom valleys (Table 2, Fig. 4b). While sheep counts increased with slope, guanaco counts decreased. Survey date did not seem to affect sheep or guanaco distribution pattern (Online Appendix 1, Table 1). The guanaco model estimated 1242 guanacos (1428 OU) for the study area.

Table 2 Weighted average of coefficients across models with delta AIC < 2 for sheep and guanacos and adjusted standard errors (in brackets)
Full size table
Fig. 4
figure 4

Sheep (a) and guanaco (b) distribution prediction based on models considering the distance from the water point, pastoral value, depth valley, mean terrain slope and paddock area. Dotted lines represent surveyed tracks. Light blue circles represent permanent water points

Full size image

Sheep grazing heterogeneity

Sheep distribution within paddocks was heterogeneous and therefore sheep grazing pressure too. The distribution model indicated areas with sheep stock adjusted to CCpv and areas with a variable degree of over and undergrazing which determined a lower CC for the evaluated paddocks (CCu) (Table 3). The guanaco density decreased with increasing sheep grazing. This pattern was more intense in the largest paddocks, which hold the highest guanaco density (Fig. 5).

Table 3 Areas with overgrazing, undergrazing and at CCpv
Full size table
Fig. 5
figure 5

Relationship between the predicted guanaco density and sheep grazing in each cell of the five evaluated paddocks

Full size image

Mixed carrying capacity

The carrying capacity calculated by the PV method but accounting for guanaco forage consumption (CCm) was considerably higher than the CCpv. The difference between CCpv and CCu or between CCm and CCmu (residual CC) is a measure of the forage not used by sheep and that could be used by other herbivores. This metrics varied among paddocks according to sheep grazing distribution which determined CCmu (Table 4a). The residual CC estimated with the CCmu would be enough to include 83.5% of the guanacos in the study area in an average year (Table 4b).

Table 4 a- Mixed carrying capacity, mixed carrying capacity adjusted by sheep distribution and ovine units not used by sheep (residual CC) for each sampling paddocks in an average year; b- Ovine units available for sheep (Sheep OU), ovine units not used by sheep (residual CC), and percentages of the guanaco population that could be included considering the residual CC calculated according to the different carrying capacity estimation methods in dry, average and rainy years
Full size table

Considering the CCm and the total grazing (sheep + guanacos) in each cell, in most of the cells where sheep grazing was below the CCm, the guanaco grazing was complementary, reaching values close to CCm (Fig. 6, grey area). On the other hand, as sheep grazing increased up to cases with sheep overgrazing, the difference between the sheep grazing and total grazing decreased, indicating a lower density of guanacos in these cells (Fig. 6).

Fig. 6
figure 6

Sheep grazing and total (sheep + guanacos) grazing in each cell. X-axis represent each grid cell. Grey area indicates grazing values < 15% (no overgrazing values)

Full size image

Discussion

Accordingly to our hypothesis, the livestock distribution showed a clear pattern defined mainly by other environmental variables than vegetation that defines over and under-grazed areas. In contrast, the guanaco distribution opposed livestock density and showed higher densities in areas sub-utilized by sheep. However, in arid Patagonia, as in other regions of the world, the methods currently used to assess rangeland’s carrying capacity omit the effective use of space by livestock (Cibils and Coughenour 2001; Golluscio 2009) and do not account for the presence of wild herbivores and their forage consumption. Based on our results, we discuss the consequences of these omissions on the carrying capacity estimates and suggest possible alternatives for adjusting them according to the particular management objective to be achieved.

Heterogeneity in spatial use by livestock

Sheep showed a highly heterogeneous spatial distribution affected mainly by water location. Sheep density decreased as the distance to the water points increased, it also decreased with terrain slope and in elevated ground areas with lower PV. The paddocks have a limited number of water points thus, the area with low or null sheep densities was greater in larger paddocks. This distribution pattern defined by environmental variables other than vegetation coincides with that described for other environments, mainly arid and semi-arid lands (Squires 1973; Holechek 1988; Stuth 1991; Adler and Hall 2005; Hunt et al. 2007; Derry and Dougill 2008). At the same time, it opposes predictions derived from the classical paradigm based on an ideal free distribution, where animals use the landscape proportionally to the forage availability (Coughenour 1991; Tyler and Hargrove 1997). In the present study, the heterogeneity in livestock distribution implied that most of the area was over or undergrazed, despite global stocks being relatively adjusted to the CC estimated for each paddock under the pastoral value method. According to these results, in an average year, around 45% of the study area would present a variable amount of forage that would not be used by livestock due to physiological restrictions (water requirements). Therefore, the real CC of the paddocks would be lower than the one estimated. In contrast, areas near drinking water showed densities higher than the estimated CCpv. The difference between values of CC estimated by the current method (CCpv) and CC adjusted by sheep grazing heterogeneity (CCu) suggested an overstock situation in paddocks with potential degradation effect in the areas effectively occupied by livestock. This type of scenario with marked heterogeneity in the spatial use by livestock can result in negative consequences on the nutritional status of animals and, therefore, on productive indices (Hunt et al. 2007). This would become especially critical in dry years if paddocks are not destocked (Holechek 1988; Sullivan and Rohde 2002). Different ways of dealing with this situation have been proposed. On one hand, some ways exist to reduce this heterogeneity through the installation of additional permanent water points or artificial shading sites, paddock subdivisions, movement of animals with herdsmen or attractors, etc. (Bailey 2004; Hunt et al. 2007). On the other hand, in cases where livestock distribution cannot be modified, the stocks should be adjusted to the real CC, accounting for this heterogeneity in its calculation (Valentine 1947; Coughenour 1991; Cowley et al. 2015), even if the estimation method used is based on the concept of key area (Holechek 1988). The CCu adjusted for sheep distribution proposed in this work is in accordance with this idea.

The drivers of livestock spatial distribution as well as their relative magnitude can vary between regions, making it necessary to account for specific situations to evaluate the effect of the different variables (Valentine 1947; Holechek 1988). Moreover, seasonal changes can also occur. Particularly, spatial limits for use due to water restriction can be broadened out during periods of higher humidity where animals can take advantage of the temporary water sources or can obtain part of their hydric requirements from the consumed vegetation, allowing them greater use of the paddock area (Owen-Smith 1996). Consequently, we limit the scope of our results to the hot-dry season, being able to vary the length and the magnitude of the period according to each year’s rainfall and temperature. However, in the middle fringe of extra-Andean Patagonia where this work was carried out, the dry season is considered a critical moment for sheep husbandry (Colombani 2012), therefore, accounting for livestock spatial use in CC estimates, at least during the dry season, constitutes the first step in carrying out a stock adjustment according to the real livestock receptivity which would mitigate grazing impact on the environment, livestock nutritional restrictions and falls in ranches production rates.

Presence of wild herbivores

Abundance and distribution of guanacos

According to our model higher guanaco densities were observed in areas with lesser sheep densities mainly far from the water points and in the largest paddocks. The intensity of this pattern varied between paddocks as a function of the joint distribution of environmental variables in the landscape. For example, in Paddock 2, its size, combined with a single water point located on the northern margin, determines a large area distant from it. As a result, this paddock shows the highest guanaco density for the ranch. On the other hand, in Paddocks 4 and 5, the water points are better distributed, considering the size and shape of the paddocks. This allows for a more homogeneous distribution of sheep grazing and, consequently, lower densities of guanacos. The segregation pattern between sheep and guanaco within paddocks was analyzed for the study area by Marino and Rodríguez (2022). It seems to be mediated by restrictions on livestock grazing and a competitive exclusion mechanism, where guanacos are displaced to areas underused by sheep. Similar distribution patterns have been reported in numerous studies with guanacos and livestock for different areas and scales (Schroeder et al. 2022). In this sense, guanaco adaptations to lower water requirements, dietary flexibility and greater efficiency in the digestion of lower-quality forage would allow them to use environments and resources that are not suitable for livestock. In this way, the guanacos seem to be distributed mainly according to the supply of forage unused by sheep, so the CC for livestock should not be affected by the presence of guanacos within grazing paddocks. On this regard, it is important to highlight that the proportion of underused forage by sheep varies gradually according to the conditioning factor, for example, it gradually increases when moving away from the water points (Valentine 1947; Holechek 1988; Adler and Hall 2005). Therefore, it would be expected that in large areas of the sampled paddocks, livestock and guanacos may overlap, not necessarily indicating that guanacos take forage away from sheep. Additionally, there is a portion of available forage resources, such as some tall shrubs, which are only accessible to guanacos and seem to facilitate guanacos and sheep coexistence when grazing in the same areas (Marino and Rodríguez 2022). In times of higher humidity and when forage is not limiting (mainly in peaks of availability of annual grasses and herbs), segregation patterns are also expected to diminish, while the spatial and diet overlap are expected to increase (Saba et al. 1995; Linares et al. 2010; Ogutu et al. 2014). In accordance with our results, it would be expected that modifications in livestock grazing distribution due to infrastructure improvements, or other options, would produce a landslide (exclusion) of the guanacos settled in previously underused areas. Finally, although the environmental determinants of livestock grazing distribution may vary between regions of Patagonia (Ormaechea et al. 2018), the previous approach would be valid in any situation with a similar segregation pattern among herbivores.

Forage consumption by guanacos and mixed carrying capacity

The pastoral value method does not include consumption by wild herbivores in the estimation of available forage, nor does incorporate diet overlap correction to set animal equivalents. This methodological approach would be valid for livestock in situations where grazing by wild herbivores is not significant. In situations with relevant guanaco abundance, such as those reported in our study area, the inclusion of the forage consumed by guanacos allows us to estimate a mixed CC, including both species of herbivores. According to our results, the inclusion of the guanaco intake in the pastoral value calculations significantly increased the total forage availability and paddocks CC (approximately 30%) compared to the CC estimated only considering sheep intake. The proportion of the mixed CC that can be used by guanacos or sheep will depend on the factors determining livestock grazing distribution.

The residual CC resulting from the difference between the mixed CC and mixed CC adjusted by livestock distribution accounted for the portion of mixed CC that was not used by sheep. It would, therefore, function as an approximation of CC for guanacos in situations of coexistence with sheep. In our case, the residual CC values would sustain 83.5% of the guanaco population in average years, even without considering the forage exclusively consumed by guanacos. In dry years, more information about reduced sheep distribution due to increased hydric requirements (reduced CCu), guanaco dispersion to other sites, diet overlap and changes in guanaco digestive efficiency is required to modify the relationship between the CCu and residual CC. Therefore, the relationship between residual CC and the guanaco population is a fundamental issue when evaluating overgrazing risk and productivity losses due to competition. In addition, it would be a suitable approach for mixed production schemes with wild guanacos as a complement to sheep husbandry, and as a management strategy aimed at optimizing production based on the spatial heterogeneity of these environments (Fuhlendorf et al. 2017).

The adjustment of animal equivalents by the dietary similarity between herbivore species is essential because the pastoral value and other current methods only consider the plant species consumed by sheep. This adjustment may be more or less important depending on the degree of diet overlap between species that will depend on the area, time of year and guanaco diet flexibility in a particular plant community (Raedeke 1979; Pontigo et al. 2020; Linares et al. 2010).

In summary, the combined analysis of sheep and guanaco densities, using an adjusted animal equivalent, respect to the portion of mixed CC that can be used by each one (adjusted mixed CC vs residual CC) would be a first approximation to assess overgrazing risk. In areas far away from water points predominantly used by guanacos, the grazing values considering the total herbivores were closer to or under the estimated mixed CC value. This result may be explained by the density self-regulation mechanism through territoriality described for sedentary populations of guanacos (Marino et al. 2016) and is in agreement with the better vegetation states reported for these sites by Marino and Rodríguez (2022). These results highlight the importance of considering the heterogeneity in the grazing distribution of each herbivore when evaluating their impact on vegetation and degradation due to overgrazing.

Although the proposed approach represents an improvement regarding the simple analysis of global stocks, it has the limitation of not including spatial and temporal variations in diet overlap, as well as forage resources that are exclusively consumed by guanacos. It is also expected that models that include more explanatory variables can better explain the spatial segregation between guanacos and sheep (Marino et al. 2020) but more parsimonious models as showed in Online Appendix 1 also would be valid. Thus, the trade-off between the model explanatory power and the operational disadvantages of a too complex one should be driven by a cost-effective analysis and the objectives underlying each particular assessment. This must be taken into account when making decisions, considering that the proposed tools trade simplicity to be practical and manageable by a technical-productive sector, but also complexity to maintain a certain explanatory power and representativeness of the assessed situations. Therefore, is necessary to work on the evaluation and monitoring of the effectiveness of different tools based on impact indicators measured in the field. Finally, regarding the risk assessment of guanaco overgrazing and its impact on sheep husbandry, it is important to highlight some key aspects and particular considerations for wild herbivores, which may differ from those used for livestock management, such as the concept of carrying capacity to be used (Mysterud 2006).

Conclusions

In a context such as the conflict between livestock ranchers and guanacos, the possible adjustments proposed to the currently used method for CC assessment can contribute to generating more realistic information and diagnoses. The results of this work reject the arguments of forage loss and guanaco impacts on grasslands on which the conflict is based. Therefore, for ranches or areas of arid Patagonia where these results could be extrapolated, it is urgent to rethink the role of the guanaco as the cause of livestock crisis and rangelands deterioration and, instead, focus on working on the heterogeneity of livestock grazing. In turn, these adjustments would result in more accurate tools with the potential to enhance sustainable productive and economic projections both for exclusively domestic livestock production and for the development of mixed productions including guanaco exploitation. Identifying the objective to be addressed (livestock production, mixed production, risk assessment of overgrazing, etc.) and the conditions of the area in terms of environment type, herbivore densities and distribution, and infrastructure, would allow defining proper methods for the assessment of the carrying capacity. On this basis, an adaptive management approach could be applied (Allen et al. 2017) that works on sources of uncertainty and information gaps, monitoring responses to management decisions and finally improving proposed tools and/or developing new ones. Hopefully, the development of methodologies with a higher predictive power but maintaining a manageable degree of complexity for technicians and ranchers could result in greater adoption of stock adjustment technology.