The Sierra de Guadarrama National Park (SGNP) is located in the Spanish Central Range that was formed during the Variscan Orogeny (Peinado et al. 1981). Intrusions of plutonic bodies in diverse phases occurred at the end of this period (Pérez-Soba and Villaseca 2010). After a long period of erosion, the Variscan Orogeny followed, defining different erosion surface levels (Pedraza 1978). Later, the Alpine Orogeny elevated the current Spanish Central Range, reactivating the Variscan fractures and creating a relief with horst and graben morphology.
Most materials present in the SGNP are granites and gneisses, which define various geomorphological environments (Pedraza et al. 2014). The landscape mainly includes elements comprising large exposed rock outcrops and mixed areas of large rock blocks standing in sandy substrates (Centeno 1989; García-Rodríguez 2015). It also includes other areas of loose materials deriving from alterations of the geological substrate. The SGNP shows a strong altitudinal gradient, with height ranging from around 900 m to slightly above 2400 m at the top of Peñalara Peak (Fig. 1).
The climate in the area is of continental Mediterranean type, with cold winters and hot dry summers. The climatic conditions vary markedly with altitude, with three zones showing different characteristics (Comunidad de Madrid 2010). The lowest zone lies between 900 and 1400 m, with 750 mm annual average rainfall, annual average temperature of 10 °C, and maximum and minimum of 28 °C in summer and −6 °C in winter, respectively. Between 1400 m and 2000 m, the annual average rainfall is around 950 mm, with annual average temperature of 8 °C and average maximum and minimum temperature of 25 °C in summer and −8 °C in winter, respectively. Precipitation regularly falls as snow between December and April. In the highest mountain zone, above 2000 m, the annual rainfall is highly variable, ranging from 1200 to over 1500 mm while the average annual temperature fluctuates from 6 to 7 °C, with a summer maximum of 22 °C and winter minimum of −12 °C. In this zone, precipitation usually falls as snow between November and May.
The vegetation in the SGNP comprises a wide variety of species showing an extraordinary degree of adaptation to the difficult rocky environment with poorly developed soils. This includes pastures and shrubs in the highest areas, and forests on the slopes and in the valleys below 1800 m (Morales 2003).
At least 323 moss species have been catalogued in the Sierra de Guadarrama (Blanco Castro and Acón Remacha 1984; Vicente and Ron 1989; Luna and Estébanez 2008), accounting for 12.6% of all the plant species in the Community of Madrid (Morales 2003). Some examples of common species in the Sierra de Guadarrama include (Lara et al. 2005): Andreaea rupestris, Brachythecium populeum, Bryum gemmiparum, Fissidens pusillus, Funaria convexa, Amblyodon dealbatus, and Encalypta streptocarpa.
The variety and richness of plant species in the Sierra de Guadarrama have provided a huge food source that has enabled the proliferation of the Iberian wild goat (Capra pyrenaica Schinz, 1838) in the absence of natural predators for more than two decades (Fig. 2).
Impact of Iberian wild goat on mosses (im indicator)
The proposed indicator assesses the disappearance of rock moss cover due to Iberian wild goat grazing and/or trampling. Rocks cleared of moss due to Iberian wild goat presence exhibit pale areas indicating the exact distribution and surface area of the patches previously covered by mosses (Fig. 3). Lichen typically surrounds moss areas. This facilitates the localization of disappeared moss patches areas, because their perimeter is clearly defined by a change in color and texture (García-Rodríguez 2015).
The value of the im indicator ranges from 0 to 1. A value of 0 indicates no moss loss, linked to no pressure exerted by Iberian wild goat. A value of 1 indicates complete moss loss, because of grazing and/or trampling. Up to values of 0.2, moss loss can be due to other natural causes. This threshold is derived from a group of areas with im values from 0 to 0.2 where there is no Iberian wild goat presence (Table 1).
Concerning the application of the im indicator, three geomorphological scenarios (Pedraza et al. 2014) were described: (a) rock outcrops, (b) screes, and (c) mixed terrain with both loose materials and rocks. Additionally, wide sandy areas of loose materials occur in the Sierra de Guadarrama, where the indicator cannot be applied.
Rock outcrop (R)
These are represented exclusively by granites and gneisses, covered by lichens and mosses (Fig. 4a). These rocks may form fractured and stepped walls, smooth rounded surfaces, or large rock blocks of tens of meters in size. Rock outcrops showing open fractures tend to be filled with muddy or sandy materials, leading to soil formation. This soil supports a wide range of plant species, including trees (Izquierdo 2007; Centeno and García-Rodríguez 2008).
This term refers to the accumulation of rocks on mountain slopes (Fig. 4b). These screes form due to the breakup of rocks forming the mountain peaks, which fall and move down the slopes. These are frequent in the high mountain zone of the Sierra de Guadarrama (Palacios et al. 2003; Carrasco et al. 2016). The rock fragment sizes in screes vary from a few centimeters to more than 1 m long.
Mixed terrain of loose materials and rocks (M)
Rock outcrops and loose materials are combined (Scarciglia et al. 2007; Lucía et al. 2011). This is the commonest scenario at Sierra de Guadarrama. This scenario clearly shows the relationship between rock moss loss and erosion processes surroundings outcrops with runnels, gullies, or mass material movements.
Altitude must be considered while assessing im. Above 2100 m, moss presence is naturally less frequent than lower down, making this indicator more difficult to apply. Above 2000 m, accumulated winter snow persists for several months, frequently making it hard for mosses to develop (Palacios and García 1997).
Another factor to consider is rock type. In the ridge area of the SGNP, granites and gneisses alternate. Gneisses have a more irregular surface texture where the moss cover is also more irregular than on granitic surfaces. That is why granitic outcrops are recommended for data collection and indicator analysis.
In this sense, Pedriza de Manzanares is the area where this indicator is most easily and suitably applied in the SGNP because of the high frequency of granites and rock outcrops, as both rock faces and isolated blocks.
Regarding the Iberian wild goat factor, several considerations must be noted. In those areas of the Sierra de Guadarrama without Iberian wild goats, rock moss cover is virtually complete (Lara et al. 2005; García-Rodríguez 2019). Moss loss may be caused by long dry periods. Nevertheless, it was shown that moss cover does not show significant losses in areas without Iberian wild goat, even when south facing, at altitudes under 900 m, and subject to high temperature and low rainfall. This provides a baseline before the impact of the reintroduction of the Iberian wild goat.
To link moss loss with Iberian wild goat pressure, a series of requirements are needed: (a) identify similar rocks areas that are accessible and inaccessible of Iberian wild goat, which effectively justify moss presence, (b) presence of browsing evidence on woody and herbaceous plant species, and (c) presence of Iberian wild goat excrement in the area.
This indicator can be applied to any rock accessible to Iberian wild goat and with moss cover. This includes the three geomorphological scenarios described above.
A total of 95 different points were selected for this study (Table 1), distributed in areas of the SGNP with different abundance of Iberian wild goat (Refoyo et al. 2019).
At each selected point, 300 m2 of rock surface was surveyed. When rock outcrops were smaller, a series of dispersed outcrops were surveyed until this total surface area was reached. The presence of Iberian wild goat at each of these 95 survey points was determined following Perea et al. (2015), where wild goat droppings, pellet groups (pg), were counted at each point corresponding to a sampling area of 314 m2 (10-m-radius circle). The dropping counting surveys took place at the same location and during the same dates as the surveying for moss cover and rock erosion. The estimate of the habitat use, via counting of the groups of Iberian wild goat excrement (pg), followed the procedures of Pfeifer et al. (1996) and Smart et al. (2004). Rock outcrop im measurements were done at the same surface where pellet groups were assessed. Quantification of the indicator was based on studying rock surface moss loss percentage. This moss-free surface was identified visually and quantified with the help of photographs. The im value for each selected point was averaged from the individual im values for that spot.
Based on the Iberian wild goat pressure, rock-type geomorphological scenarios, and altitude, eight types of zone were predefined to enable comparisons and validate the proposed indicator. These areas were, from low to high altitude: La Pedriza, Siete Picos, Navafría (East and West), Cuerda Larga (East and West), Bola del Mundo and Valdemartín, and Peñalara Massif (Table 2).
Statistics and data analysis
The following tools were employed for data analysis: Python, version 3.7.3.; Anaconda client, version 1.7.2.; Conda, version 4.7.5.; Jupyter, version 1.0.0.; Jupyter client, version 5.2.4.; IPython, version 7.4.0.; Numpy, version 1.18.2.; Scipy, version 1.2.1.; Pandas, version 0.24.2.; Matplotlib, version 3.1.1.
The approach encompassed both data analysis and statistical analysis. Classical statistics and newer data science techniques were used for either data exploration or hypothesis testing. This focus is supported by both a proposal of systemization (Karpatne et al. 2017), and its application in several scientific fields such as geology (García-Rodríguez et al. 2015), psychology (Manzanero et al. 2019), and education (Aroztegui Vélez, et al. 2020). The specific techniques used were: (a) exploratory data analysis, (b) scatter plots, (c) box plots, (d) class I analysis of variance (ANOVA), (e) null hypothesis significance test (p-value) of ANOVA results, (f) coefficient of determination, used as an effect size measurement, (g) null hypothesis significance test (p-value) linked to coefficient of determination results, (h) least-squares function fit. The curve_fit function from the scipy.optimize library was used for function adjustments. The curve_fit function uses the squared minima procedure as detailed by Branch et al. (1999), (i) the goodness of fit function for the data through the coefficient of determination, and (j) graphical representation of functions.
Scatter plots show the values and distribution of the raw data collected. Some key data distribution patterns are only accessible through these types of analyses, as has been known for a long time (Anscombe 1973). Awareness of their relevance has continued to grow over time (Matejka and Fitzmaurice 2017).
Class I analysis of variance was used to test hypotheses concerning statistically significant differences between groups being compared. Analysis of variance is a relatively robust technique even in the face of noncompliance of its assumptions, including both a normal distribution (Blanca et al. 2017) and homoscedasticity (Kohr and Games 2014). It is selected, given its clarity, even in cases where all these assumptions are not met.