The Role of Feral Goats in Maintaining Firebreaks by Using Attractants

: The threat of large forest ﬁres is increasing, and the main causes are the depopulation of rural areas, along with the e ﬀ ects of climate change. To counter this threat in recent decades, there have been numerous proposals and actions aimed at promoting grazing in the forest as a tool for controlling biomass fuel. However, the continued disappearance of traditional herds makes this activity di ﬃ cult. Rural depopulation has also meant that domestic species become feral, being habitual in the case of goats. Currently, little is known about the role that feral goats can play in the ﬁght against forest ﬁres. In this work, an analysis is made on the e ﬀ ect of feral goats on the control of the vegetation in ﬁrebreak areas. Furthermore, the e ﬀ ect of attractants, such as water, salt, or food, on goat behavior is also studied. The study was carried out on the island of Mallorca, where a population of feral goats occupies the mountain areas, and where it is common for them to graze on the network of ﬁrebreaks. The results showed that these areas in themselves exert an attractive e ﬀ ect with respect to the neighboring forest, and that the herbaceous biomass is reduced. This e ﬀ ect was enhanced with the implementation of water and salt points, although only in certain periods of the year. In general, it was possible to reduce the phytovolume of many species without a ﬀ ecting biodiversity in the short or medium term. Therefore, strategic management of feral animals, aimed at ﬁrebreak areas, could contribute not only to reducing the risk of ﬁres and, consequently, to the mitigation of climate change, but also to attracting these animals to the forests, thus avoiding their dispersion to conﬂictive places such as roads, residences, agricultural ﬁelds, and gardens.


Introduction
In recent decades, Europe, North America and parts of Asia have experienced an increase in their forest area [1]. This increase corresponds to the abandonment of agricultural land in remote and poorly productive areas, coinciding with the increase in urban area [2]. The decrease in agricultural, livestock, and forestry activity leads to the loss of the mosaic structure of the landscape, reducing its heterogeneity [3]. One of the effects of this change in land use is colonization by woody species of abandoned land [4]. In Mediterranean forests, this transformation leads to increasingly dense undergrowth, with a greater accumulation of biomass that, in turn, increases the risk of forest fires [5]. The increase in fuel continuity at landscape scale has led to the generation of large fires in the last decade, as has occurred in the Mediterranean area [6], including the Balearic Islands [7]. Furthermore, the existence of an understory composed of highly flammable sclerophyllous species, or large tufts

Study Area
The work was carried out in the Sierra de Na Burguesa (39 • 35 11 N 2 • 33 55 E) located in the municipality of Calvià in the southwestern part of the island of Mallorca (Balearic Islands, Spain). The altitude of the study area ranges between 400 and 503 m above sea level (asl). The climate is dry supra-Mediterranean, with a mean temperature of 17 • C, with January being the coldest month and August being the warmest. Annual rainfall is irregular, distributed mainly in spring and autumn, with annual lows of 313 mm and highs of 863 mm. The mean annual temperature was 18 • C for the two years of study, and the annual rainfall was 528 and 326 mm for the first and second year, respectively [24].
The vegetation is characterized by the dominance of Aleppo pine (Pinus halepensis M.) forests with an understory of Mediterranean scrubs, where species such as Erica multiflora L., Pistacia lentiscus L., Phillyrea angustifolia L., Cistus albidus L., Cistus monspeliensis L., and Anthyllis cytisoides L. are common. Perennial grasses predominate in the herbaceous layer, including Ampelodesmos mauritanica (Poiret) and Brachypodium retusum (Pers.), as well as other annual species.
The study area is part of the network of fire prevention firebreak areas on the island of Mallorca. The study area was located in a firebreak strip of variable width ranging between 30 and 50 m and a length of 2 km. Maintenance is the responsibility of the Council of Agriculture, Environment, and Territory of the Balearic Islands. It is carried out by means of selective mechanical or manual brush clearing, in addition to thinning and the occasional pruning of trees, with a frequency of 5-6 years, stipulated in the IV General Plan of defense against forest fires of the Balearic Islands [25], with the last clearing being carried out in early spring 2017. According to reports from the Ministry of Agriculture, Fisheries and Food of Spain, the last forest fire in the study area was in 1993, with a total surface of 492 hectares affected.
The area is frequented by feral goats (Capra hircus), weighing between 30-50 kg, generally forming small groups that do not exceed 50 animals, with one or more males. Their behavior is territorial, with foraging areas of between 45 and 170 hectares [26]. These goat populations probably come from the gradual abandoning of livestock from old traditional extensive farms, and as they currently lack management, they graze all over the Na Burguesa mountains.

Data Collection
To determine the effect of goats on plant species in the firebreak areas, 12 plots of 5 × 5 m were marked out. Plots were distributed in 6 pairs (one fenced plot and its respective unfenced plot). An attempt was made for each pair of plots to have the same ecological conditions (same inclination, orientation, and height), and they were less than 5 m apart. Fencing of the plots was performed immediately after the clearing maintenance of the firebreak areas, using a galvanized metal mesh 25 mm thick and 1.80 m high, thus guaranteeing the complete exclusion of goats. Each pair of plots was randomly distributed in the firebreak areas. The phytovolume of the bushes (including A. mauritanica) and the biodiversity of the vegetation in these plots were estimated before the clearing and at 2 years after clearing. In order to calculate the phytovolume (V) of each individual, the height (h), largest diameter (D), and smallest diameter (d) were measured, and then the inverted cone formula was used: To determine the biodiversity, a variant of the point interception method was used [27], which consisted of tracing 2 linear transects, following the two diagonals of the plot. Every 10 cm, a 1 m rod was driven into the ground and all species that came into contact with it were noted.
The biomass of the herbaceous group was determined after two years of the clearing. The herbaceous biomass was obtained by collecting samples from the herbaceous stratum in 5 quadrants of 50 × 50 cm distributed randomly in each plot at the end of the study period. The samples were dried to constant weight at 103 • C.
The effect of attractants on the habitat used by feral goats was studied from fecal group counts, assuming that the greater the use of a territory, the greater the amount of deposited feces [28,29]. A total of 32 strip transects of 20 × 1 m were established in two areas, 16 in each, separated by a distance of 490 m. In each zone, eight transects were distributed in parallel in the firebreak area and another eight in the adjacent pine forest. These transects were cleared of excrement at the start of the study, and the accumulated excrement was counted and removed periodically. Six fecal pellets together were established as a minimum fecal group [30], with a similar appearance in color, consistency, and humidity [31]. The sampling period was two consecutive years (from July 2017 to July 2019). During the first year, no type of attractant was used, and for the second year, two water points were installed in aluminum drinking troughs, 16 × 14.5 × 6 cm with outlet regulators, connected to a 210-L water container and attached to a stake 60 cm above the ground. The containers were filled every  15 days on average, guaranteeing the availability of water. Also, from the second year, a 10 kg block of salt was placed about 3 m from each trough, at a height of approximately 80 cm. In each water replacement, 1 kg of corn grain was spread around each water container. The 16 transects closest to the attractants were placed at an average distance of 70 m, and the 16 farthest were placed 490 m away.

Statistical Data Processing
To evaluate habitat use, the accumulation of fecal groups in transects was analyzed as an indirect measurement of the number of animals present. Two study periods were established to analyze the effect of the attractants: "before", referring to the first year prior to the use of attractants, and "after", referring to the second year after placement. The total fecal group count was divided by the total days of accumulation and multiplied by 30 to obtain a mean value of fecal groups per month. The mean monthly accumulation of fecal groups was analyzed by means of a generalized linear mixed model (GLMM), using the transect as a random factor and using the following as fixed factors: the period, referring to before and after the placement of attractants; the season, referring to the four seasons of the year; the area, split between near and far with respect to the distance towards the attractants; and the habitat, differentiating between the firebreaks area and the forest.
The effect of the exclusion of goats on the accumulation of dry biomass of grasses was analyzed by means of a mixed general model with the treatment (inside referring to the fenced plot, and outside referring to the unfenced plot), and the area (near and far with respect to the distance from the attractant) as fixed factors and plot as a random factor.
As expected, the height, cover, and phytovolume of the vegetation were highly correlated (Spearman ρ > 0.77 p < 0.0001), so only the phytovolume was analyzed. This was normalized by a logarithmic transformation. To analyze the effect of the exclusion of goats on the phytovolume of woody species (including A. mauritanica), a general mixed model was performed, with plot and species as random factors and treatment, time (before clearing and two years after clearing), and zone as fixed factors, including interactions when they were significant. For the analysis, data from species that did not have a minimum of 10 individuals were excluded.
Similar analyses were carried out separately for the species groups according to the flammability index, since this index evaluates the ease of igniting and producing a flame when subjected to constant heat energy [32]. Table 1 shows the flammability index of the main species in the study area. Statistical analyses were adjusted with mixed models, with the plot as a random factor and treatment, period, and area as fixed factors. In some cases, it was only possible to test some of the interactions since there were not enough replicates at the different levels of each factor. For the most abundant species, this same model was re-adjusted considering only the individuals of the species. All the aforementioned mixed models were adjusted using the lmer function of the lme4 package [35] by R, R Core Team [36]. The best model was selected following the Akaike information criterion (AIC) (models with lower AIC value were selected), and the residual distribution graphs were reviewed to ensure the adequacy of the models [37].
For the analysis of the effect of exclusion on diversity, the Shannon diversity index was calculated for each of the plots. The effect of the treatment and area as fixed factors was analyzed using a two-way ANOVA with the avo function by R, R Core Team [36].

Effect of Habitat and Attractants on the Accumulation of Fecal Groups
Statistically significant differences were found in the accumulation of fecal groups according to the period (p = 0.0005 GLMM), the season (p = 0.0002 GLMM), and the habitat (p < 0.0001 GLMM). The interactions between the period and habitat (p = 0.002 GLMM) and period and zone (p < 0.0001 GLMM) were also significant. Figure 1 illustrates the habitat effect, showing the number of fecal groups in the firebreaks area being 70% higher in relation to the forest (note the difference in scale between Figure 1a,b). Before the attractants were placed, the presence of fecal groups in the area furthest from attractants versus those closest was the same in most seasons of the year, except in autumn in the forest, and autumn and winter in the firebreaks, where it was greater in the furthest (Figure 1). After the attractants were placed, there was a general decrease in the accumulation of fecal groups in the forest habitat, except in autumn and winter in the nearby area, which increased. On the other hand, within the firebreak areas, after the placement of the attractants, fecal groups always decreased in the distant zone, whereas in the near zone they only decreased in spring, remaining stable in summer and winter and increasing in autumn.

Effect of the Exclusion of Goats on the Evolution of the Vegetation in the Firebreaks
Significant differences were observed in the biomass of the herbaceous stratum, both between treatments (p = 0.02 GLMM) and between areas (p = 0.01 GLMM). The biomass of grasses in the fenced plot is 53% higher compared with the unfenced plot ( Figure 2). This difference is more noticeable in plots near the attractant where it reaches 76%, while in the fenced plots far from the attractant, the biomass difference was only 37%.  The results of the analysis of the variation of the phytovolume of the bushes and scrubs, including A. mauritanica and excluding species with less than 10 individuals in the study plots, are shown in Figure 3. There are significant differences regarding the treatment (p = 0.01 GLMM), with the phytovolume being greater within the fenced plots. There were also differences regarding time (p < 0.0001 GLMM), with greater phytovolume before the clearing. In contrast, no differences were found with respect to the area (p > 0.05 GLMM). A significant effect (p = 0.0002 GLMM) was observed in the interaction between the time and zone, showing that, after two years, the phytovolume within the plots far from the attractant was similar to the phytovolume before the clearing, whereas this effect was not noticeable in fenced plots near the attractant.  The species with flammability index 2 did not show significant differences with respect to treatment and area (p > 0.05 GLMM), while the time and treatment*time interaction showed significant differences (p = 0.001 GLMM) and (p = 0.002 GLMM), respectively, indicating that variation of the phytovolume is less within fenced plots after clearing. The species with flammability index 3 showed significant differences for the three factors: treatment (p = 0.005 GLMM), zone (p = 0.003 GLMM), and time (p = 0.02 GLMM), indicating differences in phytovolume inside and outside the fenced plots, as well as the effect of the distance to the attractant and differences after two years of clearing. In this case, the interaction between factors could not be calculated due to lack of replicates. For species with flammability index 4, no significant differences were found with any of the three factors (p > 0.05 GLMM). However, significant differences were observed for the treatment and time interaction (p < 0.05 GLMM). In this case, a significant increase in the phytovolume of these species was observed after two years, although only outside the fenced plots of the nearby area. well as the effect of the distance to the attractant and differences after two years of clearing. In this case, the interaction between factors could not be calculated due to lack of replicates. For species with flammability index 4, no significant differences were found with any of the three factors (p > 0.05 GLMM). However, significant differences were observed for the treatment and time interaction (p < 0.05 GLMM). In this case, a significant increase in the phytovolume of these species was observed after two years, although only outside the fenced plots of the nearby area.  Figure 5 shows the plant volume of some of the most abundant species in the study area. In general, in the nearby plots a lower recovery of the phytovolume is observed in the unfenced plots (outside) compared with the fenced ones (inside). The phytovolume in some fenced plots, whether near or far from the attractant, showed a phytovolume higher than the initial record (two previous years), such as A. unedo (near), A. mauritanica (far), and P. angustifolia (far). Observing the phytovolume from outside the fenced plots, differences were observed both near and far from the attractant, except in P. angustifolia, with phytovolume around zero.
The Shannon biodiversity index (H) did not show significant differences between inside and outside the exclusions, nor between study areas (p > 0.05 ANOVA). H in the area near the attractants  Figure 5 shows the plant volume of some of the most abundant species in the study area. In general, in the nearby plots a lower recovery of the phytovolume is observed in the unfenced plots (outside) compared with the fenced ones (inside). The phytovolume in some fenced plots, whether near or far from the attractant, showed a phytovolume higher than the initial record (two previous years), such as A. unedo (near), A. mauritanica (far), and P. angustifolia (far). Observing the phytovolume from outside the fenced plots, differences were observed both near and far from the attractant, except in P. angustifolia, with phytovolume around zero.
The Shannon biodiversity index (H) did not show significant differences between inside and outside the exclusions, nor between study areas (p > 0.05 ANOVA). H in the area near the attractants within the exclusions was 1.98 ± 0.04, while outside it was 1.85 ± 0.38. In the case of the exclusions of the area far from the attractants, within exclusion H was 1.99 ± 0.16 and outside was 1.91 ± 0.05.

Discussion
The present study shows the important role that feral goats play in the maintenance of firebreak areas and how the use of attractants increases these effects. The results show that feral goats use the firebreak area more than the adjacent forest, putting significant pressure on its vegetation. This is in agreement with other results in which it was observed that firebreak areas have herbaceous covers of higher quality and abundance for feral ungulates and extensive livestock, compared with woody formations [18,34]. For many ungulates, modified linear areas such as firebreaks and roads act as attractants, either because of their current biodiversity, the quantity or quality of existing food, or because they are corridors for movement [13,18,38]. In the case of goats, it should be added that, being prey animals, they usually prefer open spaces where they have a wide visual field to monitor possible predators [39].
Other authors also conclude that it is possible to reduce phytovolume and biomass by different species of domestic livestock, such as cows, sheep, and goats, and highlight the potentiating effect of the attractants [40,41]. The results obtained also show the effect on the distribution of feral goats in the autumn and winter seasons, with a greater effect on vegetation near the attractants.
It is worth noting the scarce difference detected in the number of fecal groups in the summer before the use of attractants and after their placement, when water shortage could have intensified its effect. This could be due to hunting carried out 2 km from the firebreak between the months of April and June, when 257 feral goats were killed, as well as the hunting in September, when another 67 goats were killed. This would explain the general decrease in excrement accumulation in the second year of study compared with the first. Across the island, feral goat population management systems are highly heterogeneous, from administration controlled areas to completely abandoned areas, including hunting areas [18]. This makes population management very difficult since there are significant fluctuations in populations at a local level. Public administration will have to make a coordinated effort to harness the potential of these herds in maintaining firewall areas [21].
The behavior of feral goats, in regard to the range areas, was modified after placing attractants during the summer, winter, and autumn seasons, as the use of the area increased compared with before their placement. This phenomenon has also been observed in elk management, where lumps of salt were distributed at strategic points in their territories to avoid collisions with vehicles on the roads [42,43]. These results are of special relevance for the management of the territory, since extensive livestock farming is in decline, and the populations of feral animals and the forest biomass are increasing [18].
In this work, the differences in the biomass of herbaceous plants within the fenced plots compared with the biomass of those outside them were particularly notable, indicating that grazing decreases their biomass, especially in areas near water and salt supply points. Similar results have been obtained with equine cattle in Galicia (Spain), where it was evident that the continuous and rotational management of the grazing system modifies the productivity of herbaceous plants and the evolution of understory biodiversity [44]. This has also been observed with sheep in Andalusia (Spain), where a decrease in biomass of between 53% and 72% was observed [45]. In the case of domestic goats in mountain areas where grazed and ungrazed areas were compared, a decrease in understory biomass was also observed [46].
It should be noted that the herbaceous layer of the study area is dominated by Brachypodium retusum, which is a common grass in the Mediterranean basin with flammable characteristics and early post-fire regeneration [47]. On the other hand, this grass constitutes the fine fuel for the initiation of many forest fires [48], so its reduction must be considered a success in the management of feral goats. From the results obtained, it also follows that the use of supplementary attractants, such as water and salt, increase the clearing effect of animals within a discrete radius of the supply point, in the order of tens of meters in this case, a distance similar to that obtained in other studies [13,49].
Key data in firebreaks management refer to the time it takes for shrub vegetation to recover after mechanical cutting. These data will determine the frequency of clearing and, therefore, the efficiency of prevention tasks. In this work, it has been observed that part of the parcels in exclusion recover the phytovolume that existed before clearing in two years, while others require more time. It is also observed that the non-fenced plots (with free grazing), adjacent to those that recover the phytovolume in two years, have not recovered it in this period. This indicates that feral goats could delay the frequency of clearing, thereby saving management resources for fuel treatment. This has already been observed using domestic livestock, especially goats [50][51][52]. However, in the case of feral goats, this effect does not occur in all the plots, possibly due to lack of management that forces them to remain in the area. In the case of this study, reduction in plant volume depends on the area and the shrub species.
The results of the analysis of phytovolume variations as a function of flammability showed that, in general, outside the exclusion plots there is a reduction in phytovolume for all classes of flammability, although in the group of non-flammable species (1) this reduction is not significant. However, in the case of the highly flammable group of species (4), it was observed that in plots near the supplementary attractants the phenomenon was reversed, with a significant increase in the phytovolume of this group of species outside the fences. Aleppo pine (P. halepensis), a dominant tree in the area, is found in this group and is not very appealing to animals due to its high content of secondary compounds [53]. In this particular case, the browsing effect is likely to reduce competition with other species and favor the growth of highly flammable and poorly palatable species, such as pine. Something similar could happen with P. angustifolia, which is also highly flammable but much more palatable.
When the variability of the phytovolume by species in the grazed plots is analyzed after the two years of treatment, a lower recovery is observed in most of them. The effectiveness of feral goats in reducing fuel in firebreaks is thus verified. In addition, the use of feral goats also offers the potential to reduce fuel used in mechanical brush clearing, with a likely reduction in greenhouse gas emission. However, as previously discussed for pine, in some species the effect seems to be the opposite. This would be the case for P. angustifolia, which increased the phytovolume in the ungrazed plots near the attractant supply points. When browsing, animals are likely to reduce the competition of other species, thus favoring their growth.
Regarding biodiversity, grazing treatment and supply of attractants had no effect in the short-medium term. After two years of clearing and one year with attractants, no differences were observed in the Shannon indices. The values were low, below 2, probably due to low floristic richness and the dominance of some species. Two years is likely to be too short a period for changes in vegetation biodiversity to occur, especially in perennial species [13].

Conclusions
From the results obtained, it can be concluded that firebreak areas are places of attraction for feral goats, and their browsing activity contributes to slowing down the recovery of vegetation after mechanical clearing. This effect can be enhanced by supplying attractants, such as water and salt, at certain times of year and is especially evident in the reduction of the biomass of fine fuel, mainly Brachypodium retusum. As for the control of woody species, the effect depends on each area and each species, although, in general, it is possible to reduce the phytovolume of many species without affecting biodiversity in the short or medium term. The effect is probably less forceful than that achieved by directed grazing with domesticated herds. In this sense, support and recovery policies for rural pastoral communities would be necessary. However, this livestock has been decreasing since giving territory to populations of feral animals is a management strategy to attract feral goat populations to firebreak areas. This would contribute not only to reducing the risk of fires and, consequently, to the mitigation of climate change but also to the distribution of these animals to less conflictive places that do not cause damage in particularly fragile areas.