Impacts of sheep versus cattle livestock systems on birds of Mediterranean grasslands

Mediterranean pastures are experiencing strong changes in management, involving shifts from sheep to cattle-based livestock systems. The impacts of such shifts on biodiversity are still poorly understood. Here, we sought to contrast the grazing regime, vegetation structure, bird species richness and abundance, between sheep and cattle grazed parcels, to understand the mechanisms through which management decisions impact farmland birds. During spring 2019, we characterized livestock management, bird populations and sward structure in 23 cattle and 27 sheep grazed parcels. We used a Structural Equation Model to infer the direct and indirect effects of sheep and cattle grazing on birds. Although no effects were found on overall species richness, there were species-specific responses to sheep and cattle grazed systems. Grazing pressure (variable integrating stocking rate and the number of days in the parcel) had negative impacts on the prevalence/abundance of Zitting Cisticola, Corn Bunting and Little Bustard, either directly or indirectly, through the effects of grazing pressure on vegetation height. Animal density and vegetation cover had direct positive effects in Galerida spp. and Common Quail, respectively. Zitting Cisticola and Little Bustard also showed a direct response to livestock type. Our study emphasizes the importance of grazing pressure as a driver of negative impacts for bird populations in Mediterranean grasslands. Since the ongoing transition from sheep to cattle-based systems involves increases in stocking rate, and therefore potentially higher grazing pressure, we propose a policy change to cap the maximum allowed grazing pressure. At the landscape scale, a mix of sheep and cattle grazed fields would be beneficial for maintaining bird diversity.


Path construction
In order to understand how two different grazing regimes, namely sheep and cattle grazing, may affect grassland birds we used path analysis based on Structural Equation Models (SEM), as it allows us to look for direct and indirect effects of grazing regimes on bird density and presence [1] .
We start by choosing our variables of interest (see methods section) and then generated a theoretical model [2] (Figure 2) with all relevant paths being based on our general predictions and previous knowledge. Our model can be summarized in 3 main groups, which account for the direct or indirect effect of livestock on bird species density and occurrence.
The first group accounts for the indirect effect of livestock type on birds, via the impacts of the resulting grazing pressure and other associated management decisions (e.g. fertilizer use or improved pastures) on vegetation structure (vegetation height and cover) (A). The second and third groups account for the direct effect of livestock on birds, either through the effect livestock-specific (sheep or cattle) behavior (trampling patterns, impacts of feeding mode on food resources for birds, potential egg predation) on bird abundance (B) or via disturbance impacts of livestock on birds (C). Bellow, we describe the supporting theory that allowed us to structure our model.

A) Influence of livestock type indirectly via the impacts of the resulting grazing pressure and associated management on vegetation structure
A link is expected between the type of livestock (sheep or cattle) and vegetation cover and height, either because the resulting grazing pressure might be different, and/or there are differences in associated management (e.g. fertilizer use, improvement of pastures) [3,4] ( Supplementary Fig. S1, path A). Moreover, from a management point-of-view and assuming that different types of livestock have different requirements, the animal density on a given parcel might be directly affected by the type of livestock, for example if sheep parcels tend to hold a higher number of animals than a cattle parcel ( Supplementary Fig. S1, path A).
Grazing pressure is a measurement of livestock pressure on the pasture based on a widely used measurement of livestock management: the stocking rate [5] . However, this variable also considers the mean number of days, during the studied period, that animals grazed the parcel [6] , allowing for a more accurate index of livestock effect on the parcel. The stocking rate and, consequently, the grazing pressure measurements are obtained considering the livestock units (LU) conversation of each livestock type. The LU is a management unit that allows to compare different livestock types and is usually derived in terms of relative feed requirements [7,8] . The conversion ratios are generally based on metabolised energy requirements of an adult cow and in Portugal is converted as follow: adult bovine = 1 LU, young bovine (less than 6 months) = 0.4 LU and adult sheep = 0.15 LU [9] . The stocking rate and the grazing pressure calculations are, consequentially, translated as LU per hectare (LU/ha). Considering these conversions, the livestock type is restricting directly the grazing pressure in each parcel, since 1 adult bovine is equivalent to 6.7 adult sheep. Grazing pressure, which directly depends on the management decisions regarding the LU/ha and the number of grazing days, can likely affect bird species occurrence and densities, since it can significantly alter vegetation and soil features [10] .
Regarding vegetation, both livestock type and grazing pressure will influence vegetation structure ( Supplementary Fig. S1, path A). Livestock type has a direct effect on vegetation, since the trampling effect and feeding behaviour of cattle and sheep are significantly different [11,12] , and those differences can shape vegetation structure and limit vegetation height. Additionally, since sheep and cattle have different feeding requirements, expressed in livestock units, and management decisions like the number of days which a given herd graze a parcel, it is expected that grazing pressure also conditions vegetation [3,4] .
Since vegetation structure may influence food availability, nesting sites occurrence and refuge, which are three main ecologic features for grassland birds and passerines [10,13] , we expect that vegetation structure have an impact on bird's occurrence and densities, either by vegetation cover or vegetation height. For instance, vegetation cover, plays an important role in providing a heterogeneous landscape, with patches with more vegetation where birds can incubate their eggs and open patches where they can display, forage and look for predators [14,15,16,17] . In addition, moderate to higher vegetation allow for more nesting sites [18,19] , while lower vegetation can benefit mating rituals (eg. Little bustards displays [20] ).

B) Influence of livestock type directly through other non-measured species-specific (sheep or cattle) effects on birds
Livestock type is likely to influence the occurrence of some bird species due to the different behaviour of cattle and sheep. These differences can significantly alter the habitat, either by promoting different food sources (eg. higher invertebrate abundance in cattle dung) [10,21,22] , altering soil composition [4] and causing different intensities of trampling [23] or even nest predation [24] .

C) Influence of livestock directly through the disturbance impacts of livestock on birds, regardless of livestock type
The presence of livestock at different densities (regardless of being sheep or cattle) can have a direct impact on bird species densities and occurrence [11,12] (Supplementary Fig. S1, path C). In critical periods, such as breeding season, the impact of livestock disturbance on bird species presence and occurrence can be a serious threat, as a higher animal density can increase the probability of trampling of the nests [11,14,25,26] . However, disturbance can also have a positive impact on more generalist species, which can adapt to highly disturbed habitats [27] .
Supplementary Figure S1 -Theoretical model of the confirmatory-exploratory path analysis, where A) represents the paths of the indirect effect of livestock type via impacts on vegetation structure; B) represents the livestock type specific direct effects; C) represents the direct effect of animal density through disturbance impacts on birds.

Additional results
Supplementary Table S1 -Number of individual detections of bird species during the point counts. Grey background indicates the species with frequency of detection > 30%, for which the density and occurrence analysis were performed. i.

Species
k. l.
Supplementary Figure S2 -Path diagrams of density and occurrence for Zitting Cisticola (a. and b.), Common Quail (c. and d.), Corn Bunting (e.), Galerida spp. (f. and g.), Calandra Lark (h. and i.), Little Bustard (j. and k.) and for species richness (l.). Conditional R 2 is shown in the top corner of each response variable. Double-headed arrows indicate correlated errors. Thickness of black (positive) and red (negative) paths is proportional to standardized path coefficients. Path transparency is proportional to the p-value significance level. Standardized path coefficients are shown according to the criteria: p < 0.01***; 0.01 < p < 0.05**; 0.05 < p < 0.10*; p > 0.1, no value.