Dynamics of Woody Vegetation Patches in Semiarid Ecosystems in the Southeast of Iberian Peninsula

Dynamics of woody vegetation patches in semiarid ecosystems in the southeast of Iberian Peninsula

Beatriz Amat Martínez

Dynamics of woody vegetation patches in semiarid ecosystems in the southeast of Iberian Peninsula

Memoria presentada por

Beatriz Amat Martínez

para optar al grado de Doctor con Mención Internacional por la Universidad de Alicante

Director Dr. Jordi Cortina Segarra

Alicante, 8 Mayo 2015

Jordi Cortina Segarra, Catedrático y Profesor Titular de la Universidad de Alicante e Investigador del Instituto Multidisciplinar para el Estudio del Medio "Ramón Margalef"

HACE CONSTAR:

Que el trabajo descrito en la presente memoria, titulado: “Dynamics of woody vegetation patches in semiarid ecosystems in the southeast of Iberian Peninsula” ha sido realizado bajo su dirección por Beatriz Amat Martínez en la Universidad de Alicante, y reúne todos los requisitos necesarios para su aprobación como Tesis Doctoral con Mención Internacional.

Alicante, 8 de Mayo de 2015

Dr. Jordi Cortina Segarra DIRECTOR DE LA TESIS

AGRADECIMIENTOS

Esta tesis ha sido realizada gracias al apoyo financiero de distintas instituciones a las cuales quiero agradecer. Al Ministerio de Educación, por haberme otorgado una beca FPU, al proyecto UNCROACH (CGL2011-30581-C02- 01) financiado por el Ministerio de Ciencia e Innovación), al proyecto RECUVES (077/RN08/04.1) y ESTRES (063/SGTB/2007/7.1) financiados por el Ministerio de Medio Ambiente, Areas Marinas y Rurales, al proyecto GRACCIE (CSD2007-00067), financiado por el Ministerio de Ciencia e Innovación), a la Generalitat Valenciana por financiación de personal (Programa G. Forteza; FPA/2009/029) y provision de plantones para mis experimentos y al personal de VAERSA por su ayuda en las plantaciones de uno de mis experimentos. Finaliza mi tesis y una etapa de mi vida. Durante esta etapa no sólo mi tesis, sino yo misma he evolucionado, y ninguna era la misma al inicio que ahora. Durante la realización de esta tesis y muchas veces a causa de ella, han entrado personas en mi vida que me han marcado de una u otra forma. Y a todas estas personas quiero agradecerles la ayuda que me han dado en este tiempo. En primer lugar quiero agradecer a mi director de tesis, Jordi, por toda su confianza en mi. Gracias por haberme animado desde el primer momento a adentrarme en el mundo de la ciencia, y por creer en mí y en que yo era capaz de aportar algo en este mundillo. Muchas gracias por toda la paciencia que has tenido conmigo, por atender a mis exigencias y prisas en muchos momentos, por todas las ideas aportadas que sin ellas nunca me habría planteado el estudio de los dichosos “parches” y que he terminado cogiéndoles el gustillo, por enseñarme trabajar de forma científica y por siempre tener un momento para sentarnos a hablar. Gracias también por seguir mostrándome tu apoyo, a pesar de los más y los menos durante todo este tiempo. A Karen que ha sido como mi “madrina” guiándome desde el principio cuando más perdida me encontraba y hasta el final, dándome siempre algún consejo que me ha sido muy útil. Muchas gracias por todos esos ratos compartidos y por acercarte siempre para preguntarme cómo estaba y cómo iba mi tesis. A mis otros compañeros de despacho, a Román quien observaba mis inicios cuando él ya estaba finalizando, pero con quien he compartido ratos de despacho amenos. A Jorge, aunque ahora muy lejos, pero empezamos juntos este doctorado y compartimos quebraderos de cabeza mientras intentábamos escribir nuestro primer intento de paper o entender la estadística en R, e incluso te viniste conmigo a buscar parches alla por mis inicios. A Mchich, siempre tan cordial y con quien no solo compartimos despacho sino unos días por Marruecos conociendo otro interesante trabajo de campo. A Jaume, que además de también compartir Marruecos y hacer divertido el viaje, me ha hecho pasar muy buenos ratos durante dos años con todos esos cafés aderezados con charlas sobre gemelas, viajes, curriculum o, de nuevo, R, y por supuesto gracias por la ayuda en campo. Y a todos los que han pasado por ese despacho de forma más breve, a Victor, mi amigo fantasma, por compartir tanto conocimientos como alguna copita de vino. A David y Cristian, ahora cada uno en un lugar distinto del planeta pero que fueron unos compis de despacho muy majos. Gracias a Lucia por los ratos compartidos buscándole anillos a los arbustos, por su hospitalidad en Coimbra las veces que he estado por allí y por seguir haciéndome favores que le pido de vez en cuando. Gracias a toda esa gente que a pasado fugazmente por aquí y se ha venido al campo conmigo o me ha echado una mano en el laboratorio, Lorena, Nuria, Adela, Alejandro, Patricia, María, Elisa, y muy especialmente a María Joao, quien fue la primera en acompañarme al campo a muestrear parches hiciera frío o calor y en quien he encontrado una bonita amistad a pesar de la distancia que ahora nos separa. Gracias también por acogerme en tu casa en Porto. Sabes que te tengo mucho cariño. Gracias también al sector CEAM con quien además de colaborar de vez en cuando, he pasado buenos ratos. Especialmente a David y Alejandro, con quienes empecé a colaborar en mis inicios, mucho antes de comenzar esta tesis y siempre me han hecho sentir muy a gusto. A Jaime, Joan, Alberto, y los que pasaron por aquí Marina, Vanesa, Thanos, Esteban y que han tenido una palabra amable o un buen gesto hacia mí. Muy especialmente gracias a mi compi Cristian, o ricotí como hemos terminado bautizándolo, por esos cafés con cocacola casi a la hora de comer que sólo él entendía y gustosamente me acompañaba, por ayudarme a hacer mapitas y en la revisión de parte de esta tesis y por todos esos buenos ratos que hemos pasado más allá de la universidad, en playas, chat, Vallanca, un peñón en medio del mar o en la cima de una montaña, y que estoy segura que continuarán. Por supuesto también gracias a Nieves por esa transparencia y confianza que hace que parezca que nos conocemos de toda la vida. Gracias a los doctorandos, contratados, becarios y amantes del arte del departamento de Ecología que alguna vez en todos estos años me dieron una palabra de afecto, aunque a penas hayamos llegado a conocernos. A Rosario, Samantha, Bea, Soraya, Estrella, Encarni, Lorena, Azucena, Anna, Luna, Diana, Fran, Hassane y todos cuantos han pasado por estos despachos y me dedicaron una sonrisa. También gracias a los botánicos con quienes he compartido campo, risas y comidas, especialmente al principio de esta tesis. Gracias a mis dos paisanas Alma y Dinorah, que aunque ya están de vuelta al otro lado del charco, cuando estaban por aquí siempre fueron muy amables y dulces conmigo, y siempre hasta la fecha se han preocupado por saber cómo estaba. Ojalá nos veamos pronto! Gracias a Germán López, Andreu Bonet, Susana Bautista y Juan Bellot por atender más de una y dos veces mis peticiones por una cosa u otra y responderme siempre atentamente y con una sonrisa. Gracias a Paco por introducirme en R y atender mis constantes preguntas y visitas para descifrar los árboles de regresión. Especialmente también gracias a Jose Zubcoff por mostrarse tan abierto para ayudarme con la compleja revisión estadística de un paper que terminó dando buen fruto. Gracias también a los revisores externos de esta tesis y al tribunal por dedicar parte de su tiempo a mi trabajo. Santi, gracias especialmente por mostrarte tan colaborativo y dispuesto en estos años, y por los buenos ratos compartidos en Escocia. Gracias a Emma y María del Servicio de Lengua y Cultura por sus correcciones en las últimas versiones de esta tesis. A Fernando Maestre, gracias por sus consejos sobre mi trabajo, y por haber comenzado a describir la importancia de los parches, que de alguna manera fueron los trabajos predecesores de esta tesis. Gracias también a los técnicos de laboratorio, Fran y Jose, por ayudarme y pacientemente aguantar mis muestras y experimentos raros durante estos años. A los administrativos del departamento de Ecología e IMEM a quienes he mareado muchas veces con tickets, dietas y gestiones varias pero siempre me han atendido amablemente a Fina (E.P.D.), Emilio, Juanfra, Gema, Rosa y Ramón. Esta tesis afortunadamente me ha permitido viajar mucho y conocer el mundo de los congresos científicos, las reuniones relámpago y las estancias en países extranjeros que tanto ayudan a crecer. Y agradezco a las personas que han participado en esas estancias, desde su organización aquí en Alicante hasta aquellas personas que me he ido encontrando por allí, en Inglaterra y en Estados Unidos. Gracias a Scott Collins, quien ha sido atento y ha escuchado con interés mi trabajo desde el principio, y por supuesto, gracias por la revisión de esta tesis. Gracias a mis amigas Silvia y Ana, que son como mis hermanas y que aunque yo viviera en Alicante y ellas en Elda, se han venido muchas veces a desayunar conmigo a la Universidad con tal de verme un ratillo. Gracias a Cristina, por acogerme con los brazos abiertos más de una vez y por tus consejos como amiga y como colega sobre mi tesis. Y tampoco podía olvidarme de agradecer a quien debe ser mi amiga de más tiempo, que me ha demostrado que una amistad verdadera se sostiene a pesar de tiempos y distancias, gracias Ana Elena, y mucha suerte en tu tesis que acabas de comenzar. Gracias a todos los amigos que me han acompañado durante esta etapa aunque no supieran de que iba mi tesis pero que brindaban conmigo para sacarla adelante. Gracias también a mi familia, mis padres, tíos, prima y sobrinos, que aunque nunca han entendido muy bien qué hacía en la Universidad todos estos años siempre se interesaban y me preguntaban por mi trabajo (y por cuándo acabaría…). Gracias a mis hermanos por estar ahí, e incluso dejarse engañar alguna vez para venirse a descargar data-loggers al campo conmigo. Gracias también a mi primo Héctor que incluso me creó un programa para prepararme los datos para analizar las redes. Gracias a Jonás, por haber estado y seguir siempre ahí. Por haberme acompañado durante un largo recorrido y gran parte de mi tesis. Gracias por venirte a buscar parches, a analizar hojarasca, pasarme papers, reunirnos para discutir sobre el futuro de nuestras tesis, aunque sea con cervezas delante, y ayudarme siempre que te lo he pedido, que no ha sido poco. Y sobre todo por enseñarme a mi y a todos que cuando las cosas se hacen bien, dos personas son amigos a pesar de lo que sea. Finalmente, gracias a Pablo. Por compartir mi amor por la naturaleza, por hacérmelo todo tan fácil, por entender que muchas veces tocaba tesis aunque eso implicara comer a deshoras casi siempre, por hablar mi mismo idioma, por ayudarme en todo, por descubrirme tantas buenas experiencias y lugares pero sobre todo por quererme como compañera de viaje. Gracias porque ese entusiasmo tuyo me ha motivado para darle el último empujón a esta tesis, además ayudándome con los diseños que tan bien se te dan. Y aunque no sepamos a dónde nos llevará el viento ahora, gracias por aún así querer volar conmigo.

INDEX

INTRODUCTION ...... 1

1. Arid and semiarid areas. The threat of desertification...... 3 2. The semiarid Mediterranean: the case of Stipa tenacissima L. steppes...... 7 3. Use of patch-forming species in restoration...... 13 4. Unknown aspects of the dynamics of woody vegetation patches in semiarid ecosystems...... 14 5. Objectives and structure of the dissertation...... 15 CHAPTER 1 ...... 17

Overview of woody vegetation patches in Stipa tenacissima steppes

CHAPTER 2 ...... 61

Networks of plant-plant co-occurrence in semiarid steppes

CHAPTER 3 ...... 121

Endogenous and exogenous drivers of network structure in woody patches of semiarid steppes

CHAPTER 4 ...... 137

Community attributes determine facilitation potential in a semiarid steppe

CHAPTER 5 ...... 165

Litter as a filter for the recruitment of keystone species in Stipa tenacissima steppes GENERAL DISCUSSION ...... 197

1. What is a woody patch? ...... 199 2. How is a patch community structured? ...... 200 3. Emergent properties and community drivers ...... 204 4. Ecological role of woody patches ...... 207 5. Restrospective view of woody patches. Insights on the response to climate change ...... 210 6. Woody patches and steppe management ...... 212 CONCLUSIONS ...... 215

REFERENCES ...... 217

RESUMEN ...... 259

INTRODUCTION

Dynamics of woody vegetation patches in semiarid ecosystems

1. Arid and semiarid areas. The threat of desertification. Drylands cover more than 40% of the emerged land (Fig. 1; MEA, 2005). They include hyper-arid, arid, semiarid and sub-humid areas, according to the aridity index (AI), which is the ratio between mean annual precipitation and mean annual evapotranspiration (UNESCO, 1997). Generally speaking, drylands are areas where water scarcity limits primary productivity for crops and for the maintenance of a continuous plant cover (Whitford, 2002). Despite their low productivity, drylands have been inhabited for millennia, and currently they are estimated to hold 38% of the world population. In terms of surface area affected and their impact, grazing for domestic livestock production and, to a lesser extent, agriculture, are the main land uses in drylands (MEA, 2005).

Figure 1: Worldwide distribution of hyper-arid, arid, semiarid and dry subhumid areas according to FAO aridity index (AI). Source: Millennium Ecosystem Assessment, 2005.

Introduction

Ecosystems provide goods and services to humans through different processes such as primary production, production and decomposition of organic matter, nutrient cycling, redistribution and storage of water and soil erosion (Whitford, 2002). Goods offered by arid and semiarid ecosystems include food, fiber, fodder, fuel, medicines, building materials, biochemical compounds and water. They also provide a variety of services such as water regulation and purification, pollination and seed dispersal, local and global climate regulation (through vegetation and carbon sequestration, respectively), residue degradation, soil development, control of pathogens and parasites, places for recreation and tourism, signs of cultural identity, and spiritual and aesthetic services (Whitford, 2002; MEA, 2005).

It is estimated that between 10 and 20% of dryland areas are affected by desertification (MEA, 2005). This complex process is defined as "land degradation in arid, semiarid and dry sub-humid areas resulting from various factors, including climatic variations and human activities" (UNCCD, 1994). Climatic factors include drought and the reduction of available fresh water due to global warming. Human factors are both direct, such as activities related to land use and the use of services, and indirect, such as population pressure, political and socio-economic factors and changes in international market regulations.

Desertification results in the loss of biological and economic productivity, and the reduction of the provision of goods and services. It represents a greater threat to human wellbeing than degradation in other parts of the world, because of the shortage of water, the intense demand for services, and intrinsic vulnerability to climate change. For these reasons, measures to combat desertification and recover land potential to provide goods and services and sustain biodiversity are crucial (Reynolds, 2001; MEA, 2005; Rey-Benayas et al., 2009). Among them, the restoration of degraded areas is increasingly demanded.

1.1. Plant cover: a mosaic of vegetation patches Vegetation cover in arid and semiarid steppes and shrublands is low and discontinuous. Landscapes are often organized as a two-phase mosaic of vegetation patches and bare ground, at various spatial scales (Valentin et al., 1999). This spatial arrangement may be generated by geological factors, together with complex interactions between soil, climate, plants and animals (Whitford,

Dynamics of woody vegetation patches in semiarid ecosystems

2002). Still, the origin and consequences of the spatial patterns of vegetation in drylands is still a matter of discussion (Gilad et al, 2004; Pueyo and Alados, 2007; Thompson et al., 2008; Konings et al., 2011). Patch size and shape may reflect patch dynamics. For example, vegetation commonly forms bands or spots, depending on whether the driver of patch dynamics is water or wind, respectively (Aguiar and Sala, 1999; Valentin et al., 1999). Vegetation patterns have been used as indicators of ecosystem functioning in arid and semiarid areas (Maestre and Cortina, 2004b; Kefi et al, 2007).

Vegetation modulates patch formation, as it modifies environmental conditions and resource availability (Jones et al., 1994). This modification produces distinct microhabitats (Gilad et al., 2004, 2007), promoting spatial heterogeneity and the formation of "islands of fertility" around vegetation patches (Schlesinger and Pilmanis, 1998; Reynolds et al, 1999). These patches concentrate resources, and together with physical obstacles, act as sinks of resources as water, sediments, nutrients, litter and seeds, generated in bare or source areas (Tongway and Ludwig, 1994; Aguiar and Sala, 1999; Cortina et al, 2010).

1.2. Vegetation dynamics: Dependence on water and species interactions Vegetation dynamics greatly depend on the availability of soil resources as water and nutrients. In arid and semiarid systems, water is scarcely available, and its supply is intermittent and unpredictable. Thus, plant population dynamics depend on the response of individual plants to pulses of water ("auto-ecological hypothesis", Noy Meir, 1973; Collins et al., 2014). Precipitation events are isolated in time, and their impact is ephemeral (Timmermann et al, 1999; Abanades et al, 2007). Aridity promotes the presence of fast-growing species at the expense of slow-growing species, such as shrubs, as only the former can take advantage of sporadic rainfall events (Armesto et al., 1993; Cabido et al., 1993; Rey and Alcántara, 2000; Kutiel et al., 2000; Verdú and García-Fayos, 2002; Bochet et al, 2007). However, a recent study has shown that herbaceous systems are less resistant to extreme periods (Ruppert et al., 2014). Low efficiency of shrub recruitment may be offset by longevity (Bowers et al., 1995). Globally, climate pulses at various temporal scales, as El Niño Southern Oscillation (ENSO),

Introduction influence long-term vegetation dynamics in semiarid ecosystems, opening windows for recruitment of tree and shrub species (Holmgren et al., 2001).

Plant response to environmental conditions is strongly affected by species interactions (Jones et al, 1997; Callaway and Walker, 1997). It has been suggested that improved microhabitat conditions by one species may decrease the level of stress for other species (facilitation) in stressful environments. Conversely, when stress is low, competition prevails over facilitation ("stress-gradient hypothesis"; Bertness and Callaway, 1994). Numerous exceptions to this theory led to further refinements (Maestre et al, 2009a). Fauna is also an important factor affecting vegetation dynamics in arid and semiarid ecosystems. Small mammals, birds and arthropods play an important role in seed dispersal and predation, and physical modification of the environment (Debussche et al., 1982; Verdú and Gracía-Fayos, 1996; Gutiérrez et al., 1997; Barberá et al., 2006).

As a result of biotic and abiotic interactions in arid and semiarid ecosystems, populations of woody species are mainly dominated by adults forming isolated patches of relatively slow growth. Their recruitment is variable and largely dependent on environmental conditions. Thus, factors contributing to the creation of spatial heterogeneity and favorable microenvironments for plant establishment and growth in these ecosystems, such as availability of water and nutrients, concentration of propagules and soil availability, would positively influence recruitment of woody species.

1.3. Changes in plant cover: shrub encroachment Woody species density and cover have increased in semiarid areas worldwide over the last 150 years (Van Auken, 2000; Roques et al, 2001; Maestre et al, 2009b; Tighe et al, 2009). Invasion of grasslands by woody species was first described in southwest USA, and later in disparate areas such as Alaska, the Chihuahuan desert, southern Africa, South America, New Zealand and Australia (Van Auken, 2000; Briggs et al., 2005). In Spain, it has been reported in dehesas (Ramirez and Díaz, 2008), abandoned pastureland (Montané et al., 2007; De Soto et al., 2010), and rangelands (Alados et al., 2004; Maestre et al., 2009b). This phenomenon, known as ‘shrub encroachment’ and ‘woody encroachment’, has important implications for ecosystem functioning and the provision of goods and services (Jackson et al, 2002; Huxman et al., 2005). There is currently no

Dynamics of woody vegetation patches in semiarid ecosystems agreement on the causes of such large-scale process. In US grasslands, it may result from a combination of warming, high levels of herbivory and concomitant fire suppression (Van Auken, 2000; Tews et al., 2006; Brudwig, 2010). However, other factors such as increased atmospheric CO2 concentration, N deposition and cattle trampling may have contributed to the observed patterns (Van Auken, 2000; Briggs et al., 2005). Similarly, there is no consensus on the consequences of this phenomenon. An increase in cover of woody species could lead to an increase in evapotranspiration (Bellot et al, 2001; Huxman et al, 2005; Jackson et al, 2005). In addition, shrub encroachment may reduce forage production and nutritive value (Zarovali et al., 2007), and decrease community diversity and stability (Baez and Collins, 2008; Brudvig, 2010). Conversely, an increase in woody species cover has been associated with an increase in the richness of vascular plants, birds, fungi and microorganisms (López and Moro, 1997; Maestre et al., 2009b), as well as higher soil organic C and N content (Jackson et al., 2002; Asner and Martin, 2004). The previously described arguments have been used to both remove and re-introduce woody species in drylands worldwide (Cortina et al, 2004; Brudvig, 2010).

Despite the importance of shrub encroachment, our understanding of this process in Mediterranean areas is still scarce. Current evidence suggests that the presence of woody species in Mediterranean environments has a positive effect on community composition and ecosystem function (Maestre and Cortina, 2005; Maestre et al., 2009b). However, the composition and dynamics of woody patches, and their impact on the recruitment of key species and ecosystem services, are largely unknown.

2. The semiarid Mediterranean: the case of Stipa tenacissima L. steppes.

2.1. Extension, composition and land use history The semiarid Mediterranean is covered by perennial grasslands, low shrublands dominated by Fabaceae, Cistaceae and Labiaceae, tall shrublands dominated by resprouting species of the genera Quercus, Juniperus and Pistacia, and open forests of Pinus spp, Quercus spp, and Olea europaea L. (Quézel and Médail, 2003). In the western Mediterranean, semiarid areas are often covered by

Introduction open steppes of Stipa tenacissima L. (‘alfa grass’, Le Houérou, 1986). These steppes are abundant in North Africa, Spain and Italy, where they occupy 70,000 km2 (Fernández-Palazón, 1974). Current S. tenacissima steppes represent degradative stages of succession from open forest (wooded steppes) of Pinus halepensis Mill, Tetraclinis articulata (Vahl) Masters and Juniperus phoenicea L. (Le Houérou, 1986, 2001). In the Iberian Peninsula, steppes have been associated to degradative stages of oak (Quercus ilex L.) and pine (P. halepensis) woodlands, and thickets of kermes oak (Quercus coccifera L.) and sclerophyllous scrubs (Costa, 1973, 1999; Rivas-Martínez, 1987). Human pressure, together with periodic droughts and fires, are the main mechanisms responsible for degradation trajectories (Puigdefábregas and Mendizábal, 1998).

In Spain, S. tenacissima steppes are located in the southeast of the Iberian Peninsula, where they cover 6,500 km2 (Fernández-Palazón, 1974). These steppes have been affected by human activity for centuries, if not millennia (Barber et al., 1997). Until the mid 20th century, most land that was suitable for agriculture had been cultivated sometime in the history of the Mediterranean basin. As a result, shrub populations were probably reduced during periods of social integration and increasing demand for natural resources to few individuals that thrived on the margins of crop terraces and low-productivity areas, as rock outcrops. Removal and burning of woody species were common practices in these steppes until the mid-20th century (Servicio del Esparto, 1953; Fernández-Palazón, 1974). Later, as dryland crops were gradually abandoned, shrub recover was heterogeneous. On abandoned orchards, recruitment of dominant species has been commonly high and nucleated (Verdú and Garcia-Fayos, 1996; Bonet, 2004; Pausas et al., 2006). In contrast, spontaneous recruitment in abandoned S. tenacissima crops has been much slower (Barberá et al., 2006). For decades, the Forest Administration has fostered this process by planting trees and shrubs. For example, plantations carried out under the Spanish National Afforestation Plan (Plan Nacional de Repoblación Forestal, 1939) reached 3,820,000 Ha in the late 80’s (Peñuelas and Ocaña, 1996). Pinus halepensis was the most common species used in the semiarid southeast of the Iberian Peninsula (Pausas et al., 2004). Plantations in semiarid areas frequently failed, and repeated seeding and planting were needed to establish new populations (Cortina et al., 2004, 2006).

Dynamics of woody vegetation patches in semiarid ecosystems

Currently, S. tenacissima steppes are formed by a matrix of bare soil and S. tenacissima tussocks, interspersed by woody patches of resprouting shrubs, smaller shrub species mainly of the Labiaceae and Cistaceae family, and perennial grasses. Woody patches are mainly generated by Rhamnus lycioides L., Pistacia lentiscus L., Quercus coccifera, Juniperus oxycedrus and Ephedra fragilis L. Desf. In some cases, these steppes appear as low shrublands within open Pinus halepensis forests (Fig. 2).

Figure 2. Stipa tenacissima steppe in southwest of Iberian Peninsula. Dark green patches are large resprouting shrubs and Pinus halepensis adults.

2.2. Drivers of shrub colonization As in other arid and semiarid areas, low water availability is the main factor limiting ecosystem recovery and recruitment of woody species in S. tenacissima grasslands (Albaladejo et al., 1998; Gómez-Aparicio et al., 2004; Barberá et al., 2006; Mendoza et al., 2009). In these areas, water is frequently available only for a short period (Bochet et al., 2007), thus, it is not surprising that these habitats favor plant species that germinate fast and do not require high

Introduction levels of soil moisture, and species that rely on sporadic recruitment events (Holmgren et al., 2001; Schütz et al., 2002; Bochet et al., 2007).

In addition to climate, other factors control the establishment of resprouting shrubs in S. tenacissima steppes. Most of these species are dispersed by birds, thus nucleation occurs under the canopy of fleshy-fruited species and other physical structures as they act as resting perches for birds (Rey et al., 2002; Verdú and García-Fayos, 2003). Mammals commonly disperse fewer seeds than birds in Mediterranean rangelands (Herrera, 1995), but they are responsible for part of the long-distance dispersal (Jordano et al. 2007). Mammalian contribution to the regeneration of Mediterranean woody plant species through endozoochory has been highlighted in disturbed and undisturbed areas (Herrera, 1989; Herrera, 1995; Matias et al., 2008; Rosalino and Santos-Reis, 2009). Herbivore mammalians like rabbits (Oryctolagus cuniculus) can be important dispersers for fleshy-fruited shrub species in these environments (Muñoz Reinoso, 1993; Cerván and Pardo, 1997; Dellafiore et al., 2006; Delibes-Mateos et al., 2008). Mammals may disperse seeds from preferred habitats to less favorable areas as degraded shrublands and plantations (Matias et al., 2008), and may be significant vectors for the colonization of degraded S. tenacissima steppes by fleshy-fruited woody species.

2.3. Importance of woody patches in Stipa tenacissima steppes

2.3.1. Community assemblage ‘Woody patches’, as they will be referred to throughout this dissertation, are groups of large shrub species whose physiognomy is different from the adjacent matrix in Stipa tenacissima steppes. They may include up to six different resprouting species (hereafter ‘patch-forming species’), and are commonly accompanied by a large set of herbaceous species and small shrubs.

Community assembly in semiarid areas is mainly driven by facilitation (Pugnaire et al., 1996; Maestre et al., 2001; Tirado and Pugnaire, 2005; Valiente- Banuet and Verdú, 2007). However, the facilitative capacity is species-dependent (El-Bana et al., 2007; Maestre et al., 2009a), and the effect of patch-forming species on recruitment may be a function of patch composition and structure. Differences in litter accumulation, production of allelochemicals, and distribution and density of canopy and roots may affect the outcome of plant-plant

Dynamics of woody vegetation patches in semiarid ecosystems interactions within and between patches of large woody species (Beckage et al., 2000; Alrababah et al., 2009; Koorem and Moora, 2010; Rodríguez-Calcerrada, 2011). In addition, plant-plant interactions are modulated by resource availability and abiotic stress. It has been suggested that the frequency and intensity of facilitation may increase along a gradient of stress (Bertness and Callaway, 1994), although this theory is not general and has many specifications (Maestre et al., 2009a). These observations are consistent with a completely different view of woody patches as relicts of a past denser shrubland. We believe that both forces (colonization favored by positive interactions and historical fragmentation) have shaped S. tenacissima steppes for the last decades.

Patches of woody vegetation can be considered as networks of interactions between species. Patch-forming species may appear in different combinations varying geographically, as their composition depends on previous land-use, environmental conditions and patch structure. The relative importance of these factors in determining patch composition is unknown. Similarly, the presence and abundance of accompanying species (smaller species that are attached to the patches but do not form patches themselves) may reflect differences in environmental conditions and land-use history, but they may also result from interactions with patch-forming species and other accompanying species, resulting in different patterns of biodiversity. The analysis of species co- occurrence patterns provides useful information on medium to long-term interspecific interactions. Null-model randomization tests have proven very valuable to analyze these patterns and provide new insights on community assembly (Gotelli, 2000). Similarly, the analysis of ecological networks has widened our knowledge on species interactions, and community properties and dynamics (Bascompte et al., 2003; Blüthgen et al., 2006; Bastolla et al., 2009), and may be very helpful in understanding the drivers of patch formation and patch impact on community composition.

2.3.2. Biodiversity Despite the relatively low cover of patch-forming species, they play a key role in the composition, functioning and stability of Mediterranean semiarid steppes (Maestre and Cortina, 2004b; Cortina and Maestre, 2005; Maestre et al., 2009b). For example, the surface area covered by shrubs has been positively

Introduction related to the richness and diversity of perennial vascular plants (Maestre, 2004; Maestre and Cortina, 2005), although this relationship may be influenced by shrub age (Pugnaire and Lázaro, 2000). The presence of shrubs also affects composition of the biological soil crust and the relative abundance of fungi, bacteria and actinomycetes (Maestre et al., 2009b). The richness of birds and other animals such as small invertebrates has been positively related to the abundance and structure of shrub species (López and Moro, 1997; García-Pausas et al., 2004; Doblas-Miranda et al., 2009).

2.3.3. Soil properties Woody patches have been associated with higher values of soil organic C, total N and available N, water infiltration, and improved nutrient cycling and physical soil properties (Maestre and Cortina, 2004b; Maestre et al., 2009b). For example, soil beneath patches may show higher percentage of stable aggregates and greater ability to enhance the development of mycorrhizae than soil in bare areas or soil under grass cover (Caravaca et al., 2003). Furthermore, the density of soil meso- and micro-fauna may be higher under shrub patches, which contributes to increase soil porosity and aeration (García-Pausas et al., 2004). As observed in studies in arid and semiarid ecosystems (Geddes and Dunkerley, 1999; Schlesinger et al., 1999; Ludwig et al., 2005), water infiltration rate can be high under woody patches because of the protection against raindrop impact conferred by shrub canopies, and because of the high content of organic matter and increased density of soil fauna, which together contribute to improve soil structure.

2.3.4. Ecosystem services Resprouting shrubs contribute to the provision of ecosystem goods and services through their impact on biogeochemical processes, such as primary production, C sequestration, organic matter decomposition, nutrient cycling, redistribution and storage of water, and soil erosion. Thus, changes in the cover and composition of patch-forming species may affect the provision of ecosystem services provided by S. tenacissima steppes. Although the scientific community has come to a consensus on many aspects of the composition-function relationship, including potential effects on the provision of ecosystem services (e.g., Balvanera et al., 2001), ecosystem processes can respond to changes in composition or functional diversity in several ways. This response may vary for

Dynamics of woody vegetation patches in semiarid ecosystems different processes, different ecosystems, different degradation-integration states of the ecosystems, and even different compartments within the ecosystem (Cortina et al., 2006; Hughes et al., 2007), resulting in divergences in the provision of ecosystem services. For example, reforestation and revegetation have been traditionally promoted on the assumption that forests and woodlands maximize water availability and contribute to flood control. However, recent research suggests that tree canopies can indeed reduce water yield (Jackson et al., 2005; Van Dijk and Keenan, 2007), and may be unable to control flooding (O’Connell et al., 2005; FAO, 2005). Derak and Cortina (2014) recently showed that the level of services provided by shrublands in southeastern Spain was lower than the level of services provided by Pinus halepensis plantations, S. tenacissima steppes and grasslands, despite the relatively high values of soil organic matter, forage productivity and number of rare and endangered species observed in the former.

3. Use of patch-forming species in restoration In restoration programs in semiarid areas, the use of shrub species has been recently encouraged because of their role as keystone species, and evidences of low spontaneous recruitment (Fig. 3). The main objective of these interventions is to increase the cover of woody species in degraded S. tenacissima steppes (Cortina et al., 2004). The success of seeding plantations in these environments is often low and extra inputs in eco-technology are frequently needed for the establishment of new populations (Cortina et al., 2004, 2006). Traditional and innovative technologies have been employed to enhance shrub establishment in semiarid environments, including the optimization of seedling fertilization regimes in the nursery, the improvement of water harvesting techniques and soil preparation, and the use of species interactions to the benefit of introduced plants (Peñuelas and Ocaña, 1996; Maestre et al., 2001; Valdecantos, 2001; Maestre et al., 2002; Navarro et al., 2006).

Introduction

Figure 3. Use of resprouting shrubs in semiarid areas. Apical view of planted Quercus coccifera (left) and Pistacia lentiscus (right) seedlings growing inside treeshelters.

4. Unknown aspects of the dynamics of woody vegetation patches in semiarid ecosystems Despite the importance of resprouting shrubs in semiarid ecosystems, and their increasing use in restoration programs, there is little information on the structure and dynamics of their populations in the Iberian Peninsula. Few studies have evaluated the distribution and composition of shrub patches in semiarid ecosystems in the southeastern Iberian Peninsula. Similarly, patterns of species assemblage and emergent properties of patches when they are considered as separate communities, remain unknown. In addition, the attributes of these patches in relation to their ability to modify their surroundings and to modulate species assemblages, and how these attributes affect the interaction between species have not been evaluated either. For example, no studies have evaluated whether the low recruitment of some species is related to the presence of woody patches or, on the contrary, if woody patches promote the establishment of new individuals. In this context, organic horizons accumulated underneath shrubs could play a major role in the establishment of new individuals, especially in water-limited environments such as semiarid steppes, but there is little information on the accumulation of these horizons and their impact on recruitment.

Dynamics of woody vegetation patches in semiarid ecosystems

5. Objectives and structure of the dissertation In this PhD thesis entitled Dynamics of woody vegetation patches in semiarid ecosystems in the southeast of Iberian Peninsula I address different aspects of the composition and dynamics of woody patches in Stipa tenacissima steppes of southeastern Spain. First, I describe the main characteristics of woody patches, and the methodology used throughout the ensuing chapters. Second, I use a network approach to study patch composition and explore their properties. Third, I explore the biotic and abiotic drivers of network structure, and discuss their implication on steppe management. Fourth, I evaluate the ability of woody patches to facilitate the establishment of new individuals, and identify intrinsic and extrinsic factors modulating facilitation potential. And fifth, I assess the effect of litter of common woody species on the germination and early rooting of seeds of two key species in S. tenacissima steppes. Finally, I discuss the previous findings in the context of steppe ecology and management. These steps are structured into six chapters*:

Chapter 1: Overview of woody vegetation patches in Stipa tenacissima steppes. Chapter 2: Networks of plant-plant co-occurrence in semiarid steppes. Chapter 3: Endogenous and exogenous drivers of network structure in woody patches of semiarid steppes. Chapter 4: Community attributes determine facilitation potential in a semiarid steppe. Chapter 5: Litter as a filter for the recruitment of keystone species in Stipa tenacissima steppes. Chapter 6: General discussion and conclusions.

* Chapters 3 to 5 are, respectively, enlarged versions of the following articles:

Endogenous and exogenous drivers of network structure in woody patches of a semiarid steppe (Submitted) Community attributes determine facilitation potential in a semiarid steppe (Perspectives in Plant Ecology, Evolution and Systematics). Litter effects on seedling establishment of key species in Stipa tenacissima steppes (In review).

CHAPTER 1 Overview of woody vegetation patches in Stipa tenacissima steppes

Overview of woody vegetation patches in Stipa tenacissima steppes

INTRODUCTION

Stipa tenacissima steppes cover 70,000 km2 in the western Mediterranean basin. They frequently form a mosaic of woody vegetation patches immersed in a matrix of S. tenacissima tussocks, small sub-shrubs and bare soil. Some studies suggest that large woody species enhance vascular plant richness and affect the composition and functioning of soil microflora in these steppes (Maestre and Cortina, 2004b, 2005; Doblas-Miranda et al., 2009; Maestre et al., 2009a), acting as keystone species (sensu Hulbert, 1997). Thus, patch-forming species have been frequently used to restore degraded S. tenacissima steppes (Cortina et el., 2004). However, information on the composition and dynamics of these species and their impact on community composition, ecosystem processes and the provision of goods and services is scarce.

In the southeast of the Iberian Peninsula, S. tenacissima steppes cover 6,500 km2 in semiarid areas. They are largely the remnants of fiber crops that were tended and harvested until the second half of the 20th century. These steppes frequently cover the slopes of small catchments, where terraced old crops occupy the bottom and the ridges are often covered by rock outcrops (Fig. 1). As for S. tenacissima, most crops were abandoned during the second half of the 20th century. Woody patches are present throughout the area, and they are frequently aggregated around old crop margins (stone walls) and rock outcrops. This is probably the result of previous land use, when rainfed fruit crops in the terraces and fiber crops in the slopes were the main priority, and shrubs and trees were removed (Servicio del Esparto, 1953; Fernández-Palazón, 1974). Only areas of low accessibility and unsuitable for cropping were left aside and acted as local biodiversity refugia.

Chapter 1

Figure 1. Aerial view of a semiarid catchment in Campello (Alicante, southeastern Spain). Slopes are dominated by Stipa tenacissima, and valley bottoms are occupied by abandoned rainfed crops. Woody vegetation patches (small dark green spots) are dispersed on the slopes, and agreggated in ridges (white line) and abandoned terraces (red area).

Woody patches are composed by large resprouting shrubs, and they can be easily distinguished from the matrix (Fig. 2). They include one to a few shrub species accompanied by a larger set of herbaceous species and smaller shrubs. In this study, I define these two groups of plant species as “dominant” and “accompanying” species, respectively (Fig. 3). Dominant species are large resprouting shrubs that form patches themselves. Six dominant species are found

Overview of woody vegetation patches in Stipa tenacissima steppes in the area: Pistacia lentiscus L., Quercus coccifera L., Rhamnus lycioides L., Juniperus oxycedrus L., Ephedra fragilis Desf. and Osyris lanceolata Hochst. & Steud.

Figure 2. Woody vegetation patches in a semiarid catchment (Orihuela, Alicante, southeastern Spain).

Figure 3. A characteristic woody patch in a semiarid Stipa tenacissima steppe formed by individuals of Quercus coccifera and Rhamnus lycioides. Some accompanying species are also identifiable in the picture: Anthyllis cytisoides, Brachypodium retusum, Fumana ericoides.

Chapter 1

Despite the importance of woody patches for ecosystem functioning and biodiversity (as previously described in the Introduction section of the present dissertation), woody patches in S. tenacissima steppes have not been characterized. To know species composition and their distribution within the patch and along the steppes give valuable information for the study of the role of this vegetation in the ecosystem and it represents a starting point for management and conservation studies. In this chapter, I describe the physical structure and the biotic composition of 450 woody patches distributed in 15 semiarid catchments in southeastern Spain. I also describe biotic and abiotic attributes of these catchments. METHODOLOGY I selected 15 semiarid catchments along a 60 km-transect in a semiarid area in southern Alicante, Spain (Fig. 4, Table 1). As described above, catchment slopes are covered by Stipa tenacissima steppes, and they frequently show abandoned agricultural terraces at their bottom. Site characterization was carried out at catchment level, at unit level (see description below), and in the terraces to encompass the whole spectrum of spatial heterogeneity.

12 1 5 11 9 14 4 3 10

8 13 2 15 6 7

Figure 4. Distribution of the 15 catchments studied in Alicante province, southeast Spain. Numbers correspond to those shown in Table 1.

Overview of woody vegetation patches in Stipa tenacissima steppes

Table 1. Location and main properties of the 15 study catchments dominated by Stipa tenacissima. Total Old terraces No. of Catchment Tempe- Precipitation No. Code surface surface woody name rature (ºC) (mm) units area (m2) area (m2) patche 1 Aigües 16 417 26484 9130 2 56s 2 Aspe 3 18 282 29522 2166 2 310 3 Ballestera 18 343 23624 2841 2 350 4 Ballestera1 18 343 27688 3641 1 363 5 Campello2 18 357 75858 3150 1 271 6 Colmenar 18 300 13851 170 2 219 7 Colmenar2 18 300 41128 1619 2 211 8 Orihuela3 16 361 97317 0 1 242 9 Porxa 18 368 51235 9976 2 491 10 Torreón 18 343 30647 3049 3 286 11 Ventós 1 15 293 34194 1390 1 129 12 Vila 17 525 50626 11596 3 323 13 Crevillente 16 382 25183 0 1 228 14 Ventós 3 16 288 28209 0 3 107 15 Aspe 5 18 306 11998 968 1 163

Catchment characterization

We recorded five variables at catchment level (Table 2). Mean annual precipitation and mean annual temperature were obtained from Atlas Climático Digital de la Península Ibérica (Ninyerola et al., 2005). Total surface area and total number of woody patches was estimated using aerial photographs and in situ verification of catchment boundaries and woody patches bigger than 1 m of canopy diameter. These procedures excluded trees (mainly P. halepensis and Ceratonia siliqua), small patches (<1 m canopy diameter) and other confounding landscape features. We measured soil availability in 30 randomly selected points over the whole catchment area. At each point, we measured soil availability as the depth at which we could hammer a 1 cm-diameter iron rod into the soil.

Unit characterization

We selected 3-6 homogeneous units per catchment in terms of exposure, topography and plant cover by using aerial photographs and on-site verification (Table 1). Units were limited to S. tenacissima slopes, accounting for a total of 27

Chapter 1 units. In each unit we set one to three 15 x 21 m plots, depending on the unit area, and established two 21 m-transects per plot, 8 m apart, following the maximum slope. In each transect, we quantified plant cover, and the cover of loose rocks and rock outcrops, by visual estimation in 14 consecutive quadrats of 1.5 x 1.5 m (Table 2). Total plant cover was estimated as the sum of all species present in the transects, and was higher than 100% when canopies overlapped. In addition, in each plot we recorded plant richness, number of rabbit latrines, and number and size (2 orthogonal diameters of the projected canopy and maximum height) of recruits of patch-forming species. We considered an individual as a recruit when the diameter of its canopy projection area was less than 1 m. All variables obtained in the plots were averaged per unit.

In the same 21 m-transects, we estimated indicators of slope functionality, such as mean patch width, density of patches, total patch area and the average interpatch length, following the directions provided by the Landscape Functional Analysis (LFA; Tongway and Hidley, 2004; Table 2). In this method, a “patch” is every long-lived feature that acts as a sink of resources by obstructing or diverting water flow and thus collecting and filtering water, nutrients, seeds (e.g., grass tussocks, stones, branches and litter). Accordingly, “interpatches” are gaps between patches acting as source of resources such as bare soil, gravel and plants whose structure is unable to retain resources.

Patch characterization

We identified and geo-referenced all woody patches in each catchment by using aerial photographs and in situ verification. Then, we randomly selected 30 patches per catchment for patch characterization. As patches were chosen at random, we consider that patch characterization represents the actual status of woody vegetation patches in each catchment.

We recorded the aspect, slope and specific location of each woody patch (i.e. patches located on slopes vs. rock outcrops and troughs; Table 2). We measured the maximum canopy height, maximum canopy diameter and orthogonal diameter for the whole patch and for each dominant species in the patch. Litter depth and soil availability were measured in 6-10 randomly distributed points, depending on canopy projected area. Soil availability was

Overview of woody vegetation patches in Stipa tenacissima steppes considered as the depth at which we could hammer a 1 cm-diameter iron rod into the soil. We also recorded the identity and cover of all accompanying species underneath the canopy of each patch. For this, we set a transect underneath the patch and centered, parallel to the main slope. Transects extended from 1 m upslope to 1 m downslope of the patch. We visually estimated cover of each accompanying species in consecutive 50 cm-quadrats along the transects.

Old crop terraces characterization

In catchments with old crop terraces, we estimated the total area occupied by terraces using aerial photographs and in situ verification of their boundaries (Table 2). In a randomly defined set of terraces corresponding to one third of the total crop area we estimated dominant and accompanying species cover. We set upslope-downslope transects using the point-intercept method recording the identity of dominant and accompanying species each 1 m.

Note: Plant survey lasted several months and annual species composition varied during this period. Thus, only perennial species were taken into account for the analyses. Perennials represent the majority of plant cover in these steppes (J. Tormo, University of Alicante, pers. comm.).

Statistical analysis

I performed Student’s t tests to compare the cover and number of accompanying species underneath and in the periphery of the patches. I used linear regression to analyze the relation between patch area and the number of species in a patch. Non-linear fit was also evaluated but discarded as linear regression fit best to the data. The significance of differences between patches dominated by different species was assessed using ANOVA. I used Non-metric Multi Dimensional Scaling (NMDS) to analyze species composition in patches. NMDS has been recommended over other ordination techniques for community analysis because it does not ignore community structure that is unrelated to environmental variables and it does not assume multivariate normality (McCune and Grace, 2002). I performed NMDS to reduce the dimensionality of the species composition matrix using cover of dominant and accompanying species. I used Bray-Curtis distance measure with random starting configurations for NMDS.

Chapter 1

Finally, I used Pearson correlation analysis to explore the relationship between variables measured at unit level. All analyses were performed using R 3.1.0 statistics software (R Development Core Team, 2014).

Table 2. Summary of the variables recorded at each level. Catchment level (N=15) Mean annual precipitation

Mean annual temperature

Total surface area

Total number of woody patches

Soil availability Unit level (N=27) Total plant cover

Cover of perennial species

Total rock (<5cm) cover

Total rock outcrop cover

Plant richness

Number of rabbit latrines

Number of recruits of patch-forming species

Size of recruits Unit level-LFA indices Mean patch width (cm)

Nº patches/10m

Mean interpatch length (m)

Interpatch length (%)

Mean patch length (m) 2 Total patch area (m ) Patch level (N=450) UTM co-ordinates

Aspect

Slope

Patch location within the catchment

Patch area

Area of each dominant species

Patch height

Each dominant species height

Litter depth

Soil availability

Accompanying species richness

Cover of each accompanying species Terraces (N=15) Area covered by old crop terraces

Dominant species cover Accompanying species cover

Overview of woody vegetation patches in Stipa tenacissima steppes

RESULTS and DISCUSSION

Patch characterization

Physical structure of woody patches

Woody vegetation patches were 1.5 m high on average. They covered 11 m2, although the range of projected patch area was substantially wide (Table 3). Litter accumulation was maximum under patches dominated by Quercus coccifera. Soils were narrow, as mean soil depth underneath woody patches ranged between 1.4 and 50 cm. Woody patches were formed by 1 to 5 dominant species, and between 1 and 26 accompanying species.

Table 3. Main physical and biological attributes of 450 woody patches in Stipa tenacissima steppes of southeastern Spain. Patch attribute Mean ± SE Range Patch height (m) 1.61 ± 0.02 0.33 - 3.15 Patch area (m2) 11.2 ± 0.6 0.5 - 103.7 Litter depth (cm) 1.3 ± 0.1 0.1 - 7.0 Soil depth (cm) 20.0 ± 0.5 1.4 - 50 Number dominant species 2 ± 1 1 - 5 Number accompanying species underneath 8 ± 0 1 - 26 Number accompanying species periphery 7 ± 0 1 - 17 Total richness underneath 10 ± 0 2 - 29 Cover accompanying species underneath (%) 52.7 ± 1.1 0.3 - 113.3 Cover accompanying species periphery (%) 45.7 ± 1.2 1.5 - 131.3

Patches dominated by Q. coccifera were the biggest, reaching a maximum of 104 m2 (ANOVA and post-hoc Tukey-HSD test, F=50.13, df=5, p <0.05; Fig.5). Patches of E. fragilis, J. oxycedrus, O. lanceolata and R. lycioides were the smallest ones, and did not differ between them.

Chapter 1

30 c

)

b 15

10 a a a a Patch projectedcanopy area (m 5

Ephedra fragilis Osyris lanceolataPistacia lentiscus Quercus cocciferaRhamnus lycioides Juniperus oxycedrus

Dominant species Figure 5. Size of patches dominated by different patch-forming species in Stipa tenacissima steppes. Mean ± 1 SE are shown. Different letters indicate significant differences between species (Tukey-HSD test).

Biotic structure of woody patches

I defined the most dominant species in a patch as those species whose canopy projection area was the biggest. Patches dominated by R. lycioides were the most abundant (199 patches out of 450; Fig. 6). Juniperus oxycedrus and O. lanceolata patches were the least abundant. Within each group of patches dominated by one species, patches were formed mainly by one or two dominant species, and they rarely reached four or five species (16 patches out of 450).

Overview of woody vegetation patches in Stipa tenacissima steppes

250

1 dominant species 2 dominant species 199 200 3 dominant species 4 dominant species 5 dominant species

150

100 76 Number ofNumber patches 69

49 50 29 28

Ephedra fragilis Osyris lanceolataPistacia lentiscus Quercus cocciferaRhamnus lycioides Juniperus oxycedrus

Dominant species

Figure 6. Distribution of the 450 woody patches used in this study according to the dominant species (numbers at the top of the bars). Differences in shade indicate the number of patches formed by 1, 2, 3, 4 and 5 dominant species within each group.

The cover of accompanying species in the periphery of the patches was lower than underneath them (t-test, t=4.306, df=898, p<0.05; Fig. 7). This pattern was not consistent when I segregated the database by dominant species. Patches dominated by E. fragilis, Q. coccifera and R. lycioides showed higher cover of accompanying species underneath them than in their periphery (t-test p<0.05: t=2.716, df=91.9 for E. fragilis; t=2.506, df=146.6 for Q. coccifera; t=2.803, df=395.7 for R. lycioides), whilst patches dominated by J. oxycedrus, P. lentiscus and O. lanceolata had similar accompanying species cover in both microsites (t- test, p>0.05). The number of accompanying species was also lower in the periphery of the patches than underneath them (t-test, t=3.880, df=869, p<0.001).

Chapter 1

This pattern also depended on the dominant species but it was different from the cover pattern: P. lentiscus, Q. coccifera and O. lanceolata had more accompanying species underneath the patches than in their periphery (t-test p<0.05: t=3.700, df=117.5 for P. lentiscus; t=4.710, df=122.2 for Q. coccifera; t=3.346, df=53.4 for O. lanceolata), and there were no differences between microsites in patches dominated by the rest of the species (t-test, p>0.05).

100 Underneath patches Patch periphery

Cover ofCover accompanying species (%)

Ephedra fragilis Osyris lanceolataPistacia lentiscus Quercus cocciferaRhamnus lycioides Juniperus oxycedrus

Dominant species

Figure 7. Total cover of accompanying species underneath woody patches and in their periphery for each dominant species. Plant cover underneath E. fragilis, Q. coccifera and R. lycioides was significantly higher than in their periphery. Differences were not statistically significant for other species.

In addition to species cover, the identity of accompanying species was also different underneath patches and in their periphery (Table 4). Most species were slightly more abundant underneath the former microsite. Species that were apparently favored by woody patches included shade-tolerant mesic species whose presence may be strongly dependent on their existence, such as Carex

Overview of woody vegetation patches in Stipa tenacissima steppes humilis, Polygala rupestris, and Rubia peregrina. Surprisingly, species as Anthyllis cytisoides, Asparagus horridus, Ballota hirsuta, Dorycnium pentaphyllum, Helichrysum stoechas, and Sedum sediforme showed the same pattern, despite the fact that they may endure high levels of aridity. In contrast, a few species were more abundant in the periphery of the patches than inside them; these included Fumana ericoides, F. tymifolia, Globularia alypum, Plantago albicans and S. tenacissima. Other species, mostly rare species, appeared exclusively underneath (18% of accompanying species) or in the periphery (3% of accompanying species) of the patches. Finally, the frequency of species as Erica multiflora, Helianthemum spp., Cistus spp, Phagnalon spp., Rosmariuns officinalis, Sedum album, Sideritis leucantha, Stipa parviflora, Teucrium spp. and Thymus vulgaris was remarkably similar in both microsites.

Differences in the abundance of some species between microsites suggest that microsite conditions may be different enough to affect their performance. Vegetated patches may promote positive inter-specific interactions (Cuesta et al., 2010; Pugnaire et al., 2011), but no study has specifically assessed the difference between species abundance and composition underneath patches and in their immediate vicinity. Micro-environmental conditions in the area surrounding woody patches may be intermediate between those underneath the patches and in open areas (Gallardo, 2003). Thus, roots and mycrorrhizae are likely to spread beyond the projected area of woody patches, and the same may be applicable to soil faunal activity (Maestre and Cortina, 2002), and aboveground features as shadow and litterfall (Martens et al., 2000; see Chapters 4 and 5). In contrast, raindrop is frequently higher in the periphery of the canopy of Mediterranean woody species. It has been widely shown that woody patches in Mediterranean steppes increase the richness of vascular plants (Maestre, 2004; Maestre and Cortina, 2005). My results show that micro-environmental heterogeneity generated by woody patches may be partly responsible for the increase in biodiversity. On the other hand, the particular conditions created in the periphery of the patches have strong implications on the establishment of new individuals and species, and the enlargement of woody patches.

The density of some species was very low. In some cases, as Phyllirea angustifolia and Chamaerops humilis, S. tenacissima steppes represent their

Chapter 1 distribution limit. In others (e.g., Olea europaea, Salsola genistoides) this was unexpected, as these species are frequently used in restoration programs in semiarid steppes where they commonly show good performance. Finally, Pinus halepensis may be underrepresented, as we intentionally excluded catchments where this species was abundant. In contrast, some species were widespread in these steppes. Stipa tenacissima and, particularly, Brachypodium retusum, where present in more than 50% of the area, and were only followed by Asparagus horridus, Fagonia cretica, Fumana ericoides, Globularia alypum and Helianthemum violaceum as they were present in more than one third of the sampled sites.

Table 4. Proportion of patches where accompanying species were present underneath and in the periphery of woody patches in 15 Stipa tenacissima steppes.

Accompanying species Underneath (%) Periphery (%)

Ajuga iva 0.7 - Anthyllis cytisoides 14.9 7.1 Anthyllis terniflora 3.1 3.8 Artemisia lucentica 1.1 1.6 Asparagus acutifolius 1.6 - Asparagus horridus 35.3 16.2 Asparagus officinalis 0.4 - Asperula aristata subsp. scabra 0.2 - Asphodelus fistulosus 4.0 3.8 Astragalus hispanicus 0.9 - Astragalus incanus 1.1 0.7 Atractylis humilis 8.7 10.0 Ballota hirsuta 3.8 0.7 Brachypodium retusum 96.4 84.9 Bupleurum fruticescens 6.2 3.1 Carex humilis 23.8 13.3 Centaurea aspera subsp. stenophyla 6.2 5.1 Centaurium quadrifolium subsp. barrelieri 5.6 5.3 Ceratonia siliqua 0.2 - Chamaerops humilis 2.0 0.2 Cheirolophus intybaceus 5.8 2.4 Chiliadenus glutinosus 0.4 -

Overview of woody vegetation patches in Stipa tenacissima steppes

Cistus albidus 7.6 8.4 Cistus clusii 4.0 4.9 Convolvulus altaeoides 1.8 1.6 Convolvulus lanuginosus 1.3 0.9 Coris monspeliensis 1.1 0.9 Coronilla juncea 1.6 0.9 Coronilla minima subsp. lotoides 16.0 9.3 Dactylis glomerata 1.1 1.6 Dianthus broteroi - 0.4 Diplotaxis harra subsp. lagascana 0.7 1.1 Dorycnium pentaphyllum 4.4 1.6 Echium humile 2.7 2.2 Elaeoselinum asclepium 0.9 1.8 Elaeoselinum tenuifolium 0.2 0.7 Erica multiflora 5.1 6.0 Eryngium campestre 1.3 2.0 Fagonia cretica 39.3 25.6 Fumana ericoides 39.6 44.2 Fumana laevipes 11.3 12.7 Fumana thymifolia 20.7 32.7 Galium setaceum 1.8 1.6 Globularia alypum 30.7 44.4 Haplophyllum linifolium 4.2 4.7 Hedysarum boveanum 0.4 - Helianthemum cinereum 9.1 9.6 Helianthemum syriacum 8.0 7.6 Helianthemum violaceum 32.2 38.4 Helichrysum stoechas 20.2 14.4 Heteropogon contortus 1.3 0.2 Hyparrhenia hirta 2.7 3.3 Hyparrhenia sinaica 0.2 0.4 Hypericum ericoides 0.4 - Lavandula dentata 1.6 0.4 Lithodora fruticosa 0.2 - Lobularia maritima 1.6 1.3 Lonicera etrusca - 0.2 Matthiola fruticulosa 4.7 2.4 Olea europaea - 0.2 Ononis minutissima 1.3 0.7 Pallenis spinosa 0.2 -

Chapter 1

Paronychia suffruticosa 2.4 2.4 Phagnalon rupestre 19.1 19.3 Phagnalon saxatile 20.4 18.0 Phillyrea angustifolia 0.2 - Phlomis lychinitis 0.2 - Pinus halepensis 0.9 0.2 Plantago albicans 5.8 10.0 Polygala rupestris 21.8 11.3 Retama sphaerocarpa 0.2 - Rhamnus alaternus 0.4 - Rosmarinus officinalis 9.8 12.0 Rubia longifolia 0.2 - Rubia peregrina 6.4 0.4 Ruta angustifolia 14.4 6.0 Salsola genistoides 0.7 1.3 Santolina chamaecyparisus subsp. squarrosa 0.7 0.7 Satureja obovata 4.7 4.2 Sedum album 8.7 7.1 Sedum sediforme 29.3 18.2 Sideritis leucantha 18.4 14.4 Staehelina dubia 0.7 0.4 Stipa parviflora 18.2 22.4 Stipa tenacissima 51.3 67.6 Teucrium buxifolium subsp. rivasii 0.4 0.4 Teucrium capitatum 22.9 22.4 Teucrium carolipaui 9.8 8.4 Teucrium pseudochamaepitys 18.4 13.8 Teucrium ronnigeri 8.0 10.4 Thymelaea argentata 0.2 - Thymelaea hirsuta 2.0 0.4 Thymus moroderi 2.0 1.3 Thymus vulgaris 21.6 20.4 Viola arborescens 1.6 1.8

The composition of woody patches was mainly determined by the dominant species (Fig.8, Table 5). NMDS analysis including all species (92 accompanying species and 6 dominant species) showed the prevailing effect of dominant species over accompanying species on species ordination (Appendix).

Overview of woody vegetation patches in Stipa tenacissima steppes

However, some accompanying species were also correlated with NMDS axes and influenced patch ordination. Thus, to explore the composition of the community of accompanying species in further detail I performed separated NMDS analyses for dominant and accompanying species (Fig.8, Table 5). The first NMDS axis of dominant species ordination was positively correlated to Rhamnus lycioides and negatively correlated to Pistacia lentiscus, and the second NMDS axis was positively correlated to Quercus coccifera and negatively correlated to Osyris lanceolata. These results agree with a study in a Mediterranean ecosystem in Israel (Blank and Carmel, 2012) where the community of herbaceous species growing under different woody dominant species was strongly determined by the identity of the dominant species. Dominant species abundance may substantially respond to thermic and aridity gradients. Particularly, R. lycioides abundance was strongly and positively correlated with high mean annual temperature and relatively low mean annual precipitation. On the contrary, Q. coccifera is more abundant under lower temperatures, even when the range of mean annual temperature in the studied steppes was narrow (15-18ºC). High mean annual precipitation was positively correlated to Juniperus oxycedrus abundance and negatively correlated with Ephedra fragilis abundance (Fig. 8).

In accompanying species ordination, Stipa tenacissima and Brachypodium retusum were the most influential accompanying species correlated with the same axis (NMDS 1; Fig. 8, Table 5) but in opposite directions. They are both perennial grasses and very common in these areas. Stipa tenacissima forms tall tussocks, and it is the most abundant species in these steppes in terms of cover. Brachypodium retusum is a small and widespread resprouting grass. Both species were equally abundant underneath the patches as in their periphery. The second NMDS axis was mainly defined by Asparagus horridus and Carex humilis, in opposite directions. These species are much less abundant in S. tenacissima steppes than S. tenacissima and Brachypodium retusum. Mean annual temperature and precipitation were weakly correlated with both NMDS axes, although S. tenacissima abundance was positively correlated to high mean annual temperatures and B. retusum showed the opposite correlation. Carex humilis was associated to low temperatures while A. horridus was more abundant in warmer sites.

Chapter 1

Figure 8. Correlation of dominant (top) and accompanying (bottom) species with the first two NMDS (Non-metric Multi-Dimensional Scaling) axes. The correlation of each axis with mean annual temperature and precipitation is also shown. NMDS analyses were performed separately, and thus axes for dominant species are not equivalent to axes for accompanying species. Only significant accompanying species are depicted for clarity (p<0.05, see Table 5). Species abbreviations are shown in Table 5.

Overview of woody vegetation patches in Stipa tenacissima steppes

Table 5. Pearson correlation coefficients relating species cover and NMDS axes 1 and 2. NMDS analyses for dominant and accompanying species were performed separately. Significant p-values (p<0.05) are shown in bold. Species NMDS NMDS p-value p-value Species abbreviation axis 1 axis 2 NMDS 1 NMDS 2 ACCOMPANYING SPECIES

Ajuga iva Ajuiva 0.060 -0.055 0.207 0.243 Anthyllis cytisoides Antcyt 0.022 0.013 0.644 0.782 Anthyllis terniflora Antter 0.055 -0.017 0.241 0.719 Artemisia lucentica Artluc -0.148 -0.015 0.002 0.747 Asparagus acutifolius Aspacu -0.063 -0.025 0.183 0.590 Asparagus horridus Asphor 0.040 -0.212 0.393 <0.001 Asparagus officinalis Aspoff 0.090 0.110 0.057 0.020 Asperula aristata subsp. scabra Aspari 0.078 0.043 0.099 0.361 Asphodelus fistulosus Aspfis 0.003 0.030 0.950 0.527 Astragalus hispanicus Asthis -0.037 0.059 0.435 0.215 Astragalus incanus Astinc -0.055 0.003 0.241 0.957 Atractylis humilis Atrhum 0.140 -0.022 0.003 0.647 Ballota hirsuta Balhir -0.083 0.001 0.080 0.982 Brachypodium retusum Braret 0.320 0.013 <0.001 0.778 Bupleurum fruticescens Bupfru 0.244 0.070 <0.001 0.136 Carex humilis Carhum 0.311 0.172 <0.001 <0.001 Centaurea aspera subsp. Cenasp -0.005 -0.026 0.912 0.586 stenophyla Centaurium quadrifolium subsp. Cenqua -0.062 0.023 0.189 0.623 barrelieri Ceratonia siliqua Cersil 0.065 -0.135 0.169 0.004 Chamaerops humilis Chahum -0.115 -0.014 0.015 0.773 Cheirolophus intybaceus Cheint 0.021 -0.005 0.658 0.914 Chiliadenus glutinosus Chiglu 0.037 -0.087 0.429 0.066 Cistus albidus Cisalb 0.049 -0.143 0.303 0.002 Cistus clusii Cisclu 0.125 0.005 0.008 0.920 Convolvulus altaeoides Conalt 0.019 -0.001 0.684 0.985 Convolvulus lanuginosus Conlag 0.082 0.015 0.083 0.750 Coris monspeliensis Cormon -0.035 -0.023 0.453 0.628 Coronilla juncea Corjun -0.003 -0.080 0.956 0.091 Coronilla minima subsp. lotoides Cormin 0.098 0.044 0.037 0.351

Chapter 1

Dactylis glomerata Dacglo -0.144 -0.011 0.002 0.808 Diplotaxis harra subsp. Diphar 0.086 -0.092 0.067 0.051 lagascana Dorycnium pentaphyllum Dorpen 0.181 0.064 <0.001 0.173 Echium humile Echhum -0.014 -0.055 0.763 0.245 Elaeoselinum asclepium Elaasc 0.109 0.020 0.020 0.678 Elaeoselinum tenuifolium Elaten 0.058 -0.023 0.217 0.630 Erica multiflora Erimul 0.013 0.037 0.783 0.439 Eryngium campestre Erycam 0.087 -0.010 0.064 0.840 Fagonia cretica Fagcre -0.015 -0.124 0.745 0.009 Fumana ericoides Fumeri 0.072 -0.102 0.128 0.030 Fumana laevipes Fumlae 0.062 0.076 0.186 0.106 Fumana thymifolia Fumthy 0.190 -0.012 <0.001 0.800 Galium setaceum Galset 0.160 0.030 0.001 0.525 Globularia alypum Gloaly -0.107 0.164 0.023 <0.001 Haplophyllum linifolium Haplin -0.024 0.009 0.616 0.857 Hedysarum boveanum Hedbov 0.104 -0.133 0.028 0.005 Helianthemum cinereum Helcin 0.033 0.167 0.49 <0.001 Helianthemum syriacum Helsyr 0.125 -0.072 0.008 0.125 Helianthemum violaceum Helvio 0.275 -0.073 <0.001 0.121 Helichrysum stoechas Helicsto 0.036 -0.004 0.447 0.925 Heteropogon contortus Hetcon 0.030 -0.053 0.520 0.264 Hyparrhenia hirta Hyphir 0.076 -0.005 0.106 0.915 Hyparrhenia sinaica Hypsir -0.105 -0.006 0.026 0.900 Hypericum ericoides Hyperi -0.004 -0.047 0.932 0.317 Lavandula dentata Lavden 0.099 0.068 0.036 0.148 Lithodora fruticosa Litfru 0.047 0.109 0.318 0.020 Lobularia maritima Lobmar 0.134 -0.086 0.004 0.068 Matthiola fruticulosa Matfru -0.014 0.049 0.762 0.298 Ononis minutissima Onomin 0.050 0.078 0.285 0.100 Pallenis spinosa Palspi -0.065 0.103 0.170 0.028 Paronychia suffruticosa Parsuf 0.110 0.050 0.020 0.288 Phagnalon rupestre Pharup -0.070 0.074 0.140 0.119 Phagnalon saxatile Phasax 0.175 -0.090 <0.001 0.057 Phillyrea angustifolia Phiang -0.057 -0.020 0.227 0.666 Phlomis lychinitis Phllyc 0.064 -0.020 0.175 0.674 Pinus halepensis Pinhal 0.083 0.112 0.078 0.017 Plantago albicans Plaalb 0.082 -0.019 0.082 0.686

Overview of woody vegetation patches in Stipa tenacissima steppes

Polygala rupestris Polrup 0.270 0.100 <0.001 0.033 Retama sphaerocarpa Retsph -0.002 0.035 0.966 0.455 Rhamnus alaternus Rhaala 0.144 -0.072 0.002 0.127 Rosmarinus officinalis Rosoff 0.038 -0.049 0.420 0.302 Rubia longifolia Rublon -0.080 0.078 0.090 0.098 Rubia peregrina Rubper 0.056 0.108 0.233 0.021 Ruta angustifolia Rutang -0.034 0.004 0.475 0.937 Salsola genistoides Salgen -0.022 -0.109 0.648 0.021 Santolina chamaecyparisus Sancha -0.016 0.063 0.739 0.179 subsp. squarrosa Satureja obovata Sataov 0.047 0.099 0.317 0.036 Sedum album Sedalb -0.154 -0.076 0.001 0.109 Sedum sediforme Sedsed 0.085 -0.067 0.072 0.156 Sideritis leucantha Sidleu 0.162 -0.166 0.001 <0.001 Staehelina dubia Stadub 0.063 -0.059 0.183 0.215 Stipa parviflora Stipar 0.157 0.142 0.001 0.003 Stipa tenacissima Stiten -0.448 0.070 <0.001 0.141 Teucrium buxifolium subsp. Teubux -0.110 -0.012 0.020 0.800 rivasii Teucrium capitatum Teucap 0.005 -0.037 0.914 0.436 Teucrium carolipaui Teucar 0.014 -0.030 0.760 0.521 Teucrium pseudochamaepitys Teupse 0.234 0.240 <0.001 <0.001 Teucrium ronnigeri Teuron 0.184 -0.007 <0.001 0.879 Thymelaea argentata Thyarg -0.003 0.012 0.944 0.805 Thymelaea hirsuta Thyhir 0.018 0.095 0.702 0.043 Thymus moroderi Thymor 0.051 0.002 0.283 0.974 Thymus vulgaris Thyvul 0.204 0.088 <0.001 0.062 Viola arborescens Vioarb -0.044 -0.089 0.350 0.060 DOMINANT SPECIES

Quercus coccifera Quecoc -0.193 0.721 <0.001 <0.001 Pistacia lentiscus Pislen -0.526 0.109 <0.001 0.021 Rhamnus lycioides Rhalyc 0.646 -0.521 <0.001 <0.001 Juniperus oxycedrus Junoxy -0.477 -0.194 <0.001 <0.001 Ephedra fragilis Ephfra 0.470 0.366 <0.001 <0.001 Osyris lanceolata Osylan -0.285 -0.449 <0.001 <0.001

Chapter 1

Relationship between physical and biotic structure of woody patches

I found that patch richness was positively related to patch area when I considered all species together, or dominant and accompanying species separately (Fig. 9, Table 6). Conversely, this relation was not always significant when I analyzed the data separating the patches by its dominant species (Fig.10, Table 6). The positive relation between the richness of accompanying species and patch area was significant for patches dominated by J. oxycedrus, Q. coccifera and R. lycioides, and marginally significant for P. lentiscus. Because of this relation, Q. coccifera patches (the biggest ones; Fig. 5) had the highest richness of accompanying species. The slope of the relationship between patch size and richness of accompanying species was higher in patches dominated by J. oxycedrus than in patches dominated by other species.

35 Dominant species Accompanying species 30 Total species

Number of species 10

0 0 20 40 60 80 100

2 Patch projected area (m )

Figure 9. Number of perennial species as a function of patch area. Lines correspond to linear regressions of total number of species (upper line), accompanying species (middle line) and dominant species (lower line). See Table 6 for estimated parameters of the regressions.

Overview of woody vegetation patches in Stipa tenacissima steppes

20 20 Juniperus oxycedrus Ephedra fragilis 18 18 16 16 14 14 12 12 10 10 8 8 6 6 4 4 2 2

Number of accompanying species accompanying of Number Number of accompanying species accompanying of Number 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Patch canopy projected area (m 2) Patch canopy projected area (m 2)

20 20 Rhamnus lycioides 18 18 Osyris lanceolata 16 16 14 14 12 12 10 10 8 8 6 6 4 4 2 2 Number of accompanying species accompanying of Number 0 species accompanying of Number 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Patch canopy projected area (m 2) Patch canopy projected area (m 2)

30 30 Quercus coccifera Pistacia lentiscus 25 25

20 20

15 15

10 10

5 5

Number of accompanying species accompanying of Number 0 species accompanying of Number 0 0 20 40 60 80 100 120 0 10 20 30 40 50 60 Patch canopy projected area (m 2) Patch canopy projected area (m 2)

Figure 10. Number of accompanying species underneath patches as a function of patch area, for patches dominated by different patch-forming species. Note the different scale for Q. coccifera and P. lentiscus. Lines correspond to linear regressions. See Table 6 for estimated parameters of the regressions.

Chapter 1

Table 6. Estimated parameters of linear regression fits between patch size (P) and number (No.) of species (S; (S=a+bP). p= p-value, N= sample size. Dominant species refer to linear regressions for patches where that species was the most dominant species in the patch. Original data is presented in Figures 9 and 10. S a b R2 p N Dominant species Total number of species 8.42 0.15 0.247 <0.001 450 - No. of dominant species 1.53 0.02 0.132 <0.001 450 - No. of accompanying species 6.89 0.12 0.200 <0.001 450 - No. of accompanying species 7.72 0.01 0.000 0.903 49 Ephedra fragilis No. of accompanying species 4.08 0.47 0.374 <0.001 29 Juniperus oxycedrus No. of accompanying species 7.18 0.18 0.038 0.321 28 Osyris lanceolata No. of accompanying species 7.22 0.09 0.047 0.072 69 Pistacia lentiscus No. of accompanying species 7.33 0.11 0.326 <0.001 76 Quercus coccifera No. of accompanying species 6.88 0.11 0.032 0.011 199 Rhamnus lycioides

Accompanying species density was negatively related to patch area fitting an exponential decay (Fig. 11) suggesting that an increase in patch size does not increase the probability of admitting new species at the same rate, and thus small patches are relatively richer than bigger ones.

)

2 Density = 7.005 e-0.315 area 2 25 R =0.420 p>0.001

Density of accompanying species (species m (species species of accompanying Density

0 20 40 60 80 100 120

2 Patch canopy projected area (m ) Figure 11. Density of accompanying species as a function of patch canopy projected area. Line corresponds to exponential fitting.

Overview of woody vegetation patches in Stipa tenacissima steppes

Catchment and unit characterization

Mean annual temperature was negatively correlated to altitude and average woody patch area at unit level (Table 7). This relationship may reflect the prevalence of P. lentiscus and Q. coccifera patches at higher altitudes, as these species formed the biggest patches. Mean annual precipitation and plant cover were positively related at this level. Units with relatively high cover of loose rock and rock outcrops showed higher distance between resource sinks (interpatch length) than units where the soil surface was less rocky. In units where the distance between resource sinks was relatively high, the cover and length of these resource sinks were smaller. In units with high plant cover, resource sink size (length and width) was also high. However, woody patch area was negatively correlated to resource sink width, which might suggest that in units with big woody patches (those dominated by Q. coccifera and P. lentiscus) interpatch area is basically bare, whereas in units with smaller woody patches (those dominated by R. lyciodes, E. fragilis, J. oxycedrus and O. lanceolata) the surrounding area is covered by small shrubs and perennial grasses, which may act as resource sinks.

Table 7. Pearson correlation coefficients between unit features. N=27 in all cases, except for resource sink cover and resource sink width (N=26). P-values are shown in parenthesis. Significant correlations (p<0.05) are highlighted in bold. “Resource sink” corresponds to every long-lived feature that obstructs or diverts water flow and collects/filters out material from runoff (e.g., grass tussocks, stones, branches and litter).

cover cover length length Altitude Interpatch Interpatch Patch area Patch Rock cove Rock Plant Plant cover Precipitation Temperature Rock outcrop outcrop Rock sink Resource sink Resource Temperature -0.148 ------(0.461) ------Altitude -0.069 -0.913 ------(0.732) (<0.001) ------Patch area -0.162 -0.694 0.762 ------(0.418) (<0.001) (<0.001) ------Rock cover -0.292 0.121 -0.027 -0.063 ------(0.140) (0.547) (0.892) (0.753) ------Rock outcrop -0.132 0.101 -0.051 -0.048 0.315 - - - - - cover (0.513) (0.617) (0.799) (0.814) (0.110) - - - - - Plant cover 0.680 0.043 -0.107 -0.118 -0.089 -0.005 - - - - (<0.001) (0.832) (0.595) (0.558) (0.659) (0.981) - - - - Interpatch length -0.307 -0.074 0.225 0.275 0.386 0.467 -0.321 - - - (0.120) (0.714) (0.259) (0.165) (0.046) (0.014) (0.103) - - - Resource sink -0.010 0.348 -0.400 -0.422 -0.249 -0.036 0.146 -0.568 - - cover (0.960) (0.082) (0.043) (0.032) (0.221) (0.863) (0.477) (0.002) - - Resource sink 0.302 0.311 -0.420 -0.266 -0.235 -0.121 0.610 -0.530 0.740 - length (0.125) (0.114) (0.029) (0.180) (0.239) (0.547) (0.001) (0.005) (<0.001) - Resource sink 0.176 0.370 -0.385 -0.414 -0.181 -0.172 0.422 -0.387 0.818 0.695 width (0.391) (0.063) (0.052) (0.036) (0.377) (0.401) (0.032) (0.051) (<0.001) (<0.001)

Overview of woody vegetation patches in Stipa tenacissima steppes

The relation between the abundance of woody patches and climatic variables was highly dependent on the dominant species (Fig. 12 and 13). Under warmer conditions, R. lycioides patches were the most abundant, and as mean annual temperature decreased, the abundance of R. lycioides patches decreased (Fig. 12). In contrast, Q. coccifera and J. oxycedrus patches were dominant at the lowest temperatures. Rhamnus lycioides patches were also abundant in the driest conditions, together with patches of Q. coccifera, J. oxycedrus and E. fragilis. In the wettest sites, O. lanceolata patches were more abundant than patches dominated by other species. Although all dominant species were present in many catchments, patches dominated by O. lanceolata, Q. coccifera and J. oxycedrus were more abundant at specific catchments, and thus, specific climatic conditions (Fig. 13). O. lanceolata-dominated patches were more abundant under warm conditions, but in a wide range of precipitation. J. oxycedrus-dominated patches were more abundant in colder sites and at a relative wide range of precipitation. Quercus coccifera-dominated patches showed their maximum abundance at low temperatures but also at low precipitation (288-361 mm), which is surprising as this species forms dense shrublands in sub-humid and humid sites (>400 mm annual precipitation) in the Mediterranean region (E. Pastor, University of Alicante pers. comm.). Taking into account the fact that the range of variation in mean annual temperature in the studied steppes was narrow (15-18ºC), differences in species abundance suggest that these species may be very sensitive to climatic variations. Average temperatures in southeastern Spain are expected to increase by 2.5-3.5 in the next 75 years (IPCC, 2007). Most predictions on average precipitation point towards a decrease in average precipitation in this area, although the degree of certainty is much lower in this case (Machado et al., 2011). Thus, considering that the model of species dominance along gradients of temperature and precipitation described would hold true under new climate scenarios, catchments covered by patches of Q. coccifera and J. oxycedrus would gradually be dominated by R. lycioides and E. fragilis over the next decades. Species shift may have significant implications on the functioning of S. tenacissima steppes, as patch function and ecological interactions are strongly related to the dominant species (see Chapters 3 and 4).

Chapter 1

E. fragilis 1.0 J. oxycedrus O. lanceolata P. lentiscus Q. coccifera 0.8 R. lycioides

0.6

0.4

Proportion ofProportion patches

0.2

0.0 15 16 17 18 Mean annual temperature (ºC)

1.0

0.8

0.6

0.4

Proportion ofProportion patches

0.2

0.0 250 300 350 400 450 500 550 Mean annual precipitation (mm)

Figure. 12. Proportion of patches dominated by each patch-forming species as a function of mean annual temperature (top) and mean annual precipitation (bottom).

Overview of woody vegetation patches in Stipa tenacissima steppes

Figure 13. Distribution of woody patches along temperature and precipitation gradients. Circle size corresponds to the relative abundance of patches of a given species with respect to the total number of patches dominated by this species in all catchments. As a reference: 43% of O. lanceolata patches were present at the catchment with 525 mm and 17ºC, 20% of Q. coccifera patches were present at the catchment with 288 mm and 16ºC, and 6% of P. lentiscus patches were present at the catchment with 417 mm and 16ºC.

Chapter 1

Variability in the recruitment of patch-forming species was huge. Seedling density in the different catchments ranged from 1 to 49 seedlings Ha-1 (Fig. 14 and 15). Recruitment of patch-forming species depended on the considered species, Rhamnus lycioides showed the largest recruitment, with an average of 11 ± 3 seedlings Ha-1. On the contrary, recruitment of other species was lower, and did not significantly differ between them (ANOVA and post-hoc Tukey-HSD test, p<0.05; Fig. 14). The pattern is similar when considering only catchments where adult species were present. There were not juveniles of certain dominant species in catchments where that species was not present. However, with respect to the abundance of adult individuals, the pattern is different, J. oxycedrus had higher recruitment than O. lanceolata, P. lentiscus and Q. coccifera, with no significant differences with E. fragilis and R. lycioides (ANOVA and post-hoc Tukey-HSD test, F=3.12, df=5, p <0.05, Fig. 15). These results suggest that population dynamics of patch-forming species are heterogeneous, and some species may be stagnant or may recruit in pulses.

Overview of woody vegetation patches in Stipa tenacissima steppes

)

-1 16 14 All catchments Catchments where the species was present 14

12 6

10 6 8

7 6 3

4 2

Recruitment density (Number of seedlings Ha 0

E. fragilis R. lycioides J. oxycedrus P. lentiscusO. lanceolataQ. coccifera Woody shrub species

)

-1 50

Recruitment density (Number of seedlings Ha 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Catchment

Figure 14. Recruitment of patch-forming species (top) and total recruitment in each of the 15 studied catchments (bottom). In the upper figure, black bars correspond to density recruitment considering all catchments (15), and grey bars correspond to density recruitment only in the catchments where juveniles of each species were present (numbers at top of the bars). Catchment features are described in Table 1.

Chapter 1

)

-1 0.8

0.6

0.4

ab ab 0.2

b b b

0.0

Turnover rate (Number recruits · Number adults

Ephedra fragilis Osyris lanceolataPistacia lentiscusQuercus coccifera Juniperus oxycedrus Rhamnus lycioides

Woody shrub species

Figure 15. Estimation of the turnover rate of patch-forming species in catchments where adults were present. Mean ± 1 SE are shown. Different letters indicate significant differences between species (Tukey-HSD test).

The relation between recruitment, and mean annual temperature and precipitation was positive in both cases, although the variability between catchments was very high (Fig. 16). It is worth mentioning that the relation remained positive when outliers were removed (data not shown). The increase in recruit density with temperature may be a result of the increasing relative dominance of the more thermophilous species R. lycioides and E. fragilis. These results suggest that the reduction in recruit density with decreasing precipitation predicted by climate models may be counterbalanced by the effect of the increasing average temperature. However, recruitment rates would be maintained at the expenses of the shift in species composition mentioned above It is worth to recall here that R. lycioides was present in 44% of the patches and all catchments. The overwhelming presence of this species and its large recruitment performance indicate that it may become more widespread in S. tenacissima

Overview of woody vegetation patches in Stipa tenacissima steppes steppes as climate becomes warmer and drier. Yet, other species whose presence is currently scarce or are indeed absent may expand as a result of changing climatic conditions (Bakkenes, et al., 2002).

) 60

-1 y = -45.47 + 3.87 x 2 50 R = 0.070 p = 0.343

Density of recruitmentDensity (number of Ha seedlings 15 16 17 18 Mean annual temperature (ºC)

) 60

-1 y = -9.37 + 0.08 x 2 50 R = 0.120 p = 0.205

Density of recruitmentDensity (number of Ha seedlings 250 300 350 400 450 500 550 Mean annual precipitation (mm)

Figure 16. Recruitment of patch-forming species in Stipa tenacissima steppes as a function of mean annual temperature (top) and precipitation (bottom).

Chapter 1

The cover of patch-forming species in old crop terraces was highly variable (Fig. 17). Rhamnus lycioides was present in the terraces of all catchments, being the only species in catchment 6 and 11. The mean cover of this species was 15 ± 3 %, which more than doubled the cover of other patch-forming species (6 ± 2 % and 2 ± 1% for P. lentiscus and O. lanceolata, respectively). The cover of Q. coccifera, E. fragilis and J. oxycedrus was remarkably low in the terraces of all catchments.

60 E. fragilis J. oxycedrus 50 O. lanceolata P. lentiscus Q. coccifera R. lycioides 40

Cover of patch-forming species (%) of species patch-forming Cover

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Catchment

Figure 17. Cover of the six patch-forming species in old crop terraces of the 15 evaluated catchments in Stipa tenacissima steppes. No old crop terraces were present in catchments 8, 13 and 14.

Overview of woody vegetation patches in Stipa tenacissima steppes

The number of woody patches and the recruitment of patch-forming species in the slopes were positively related to the number of woody patches in old crop terraces (Fig. 18). Inferring causal relationships from observational data is risky. However, I may hypothesize that abandoned terraces, and particularly stone walls used to retain soil, acted as local refugia for patch-forming species. As S. tenacissima crops in slopes were abandoned (mostly during the second half of the 20th century; Fernández-Palazón, 1974), species present in terraces probably expanded along the terraces and along the slopes. Chronosequences of aerial photographs, and genetic and dendrological analysis are not conclusive in this respect (V. Rolo, K. Disante, L. De Soto, pers. comm.). On the one hand, expansion trajectories from abandoned terraces may be blurred by the presence of old isolated individuals in slopes, and particularly in ridges which could protect these species during the peak of farming activities. Besides, most patch-forming species are zoochorous, and long-distance transport from inside and outside the catchments is likely.

Chapter 1

600 y = 171.1 + 4.1 x R2 = 0.351 500 p < 0.05

400

300

200

100

Total number ofpatches in slopes woody

0 0 10 20 30 40 50 60 Cover of dominant species in old crop terraces (%)

70 y = 19.2 + 0.2 x R2 = 0.035 60 p = 0.503

Number of ofseedlings dominant in species slopes 0 10 20 30 40 50 60 Cover of dominant species in old crop terraces (%)

Figure 18. Number of woody patches and recruitment of patch-forming species in the slopes as a function of the cover of dominant species in old crop terraces of Stipa tenacissima steppes.

Overview of woody vegetation patches in Stipa tenacissima steppes

CONCLUSIONS Woody patches in S. tenacissima steppes can be very large. As they develop on extremely narrow soils, patches may tap resources from wider areas. Still, I have found that patch effect on accompanying vegetation declines beyond its canopy limits, which emphasizes the importance of aboveground features in determining patch ecological engineering capacity. Patches are dominated by only a few species but complex communities organize around them. The rules governing the assembly of dominant and accompanying species are largely unknown. Considering the relationship between both groups of species described in this study, I anticipate that complex interspecific and multi-specific interactions may contribute to shape community composition in woody patches and hence, patch functioning. My results suggest that patches facilitate the establishment of accompanying species, and may promote the recruitment of patch-forming species. Manipulative studies could help to evaluate patch capacity to facilitate patch-forming and accompanying species, and identify patch features supporting facilitation capacity. Despite that large re-sprouting shrub species have the ability to form patches, their effects on patch properties differ. Recruitment rate in open areas was also dependent on the identity of the woody species, as it was the relationship between recruitment density and environmental conditions. While some species (particularly R. lycioides) are particularly abundant and successful in incorporating new individuals, other (mainly Q. coccifera) show a much more conservative strategy, and may be unable to cope with future climate scenarios. These observations should be validated by further observational studies and manipulative experiments, as their implications on the composition and functioning of S. tenacissima steppes can be crucial. For example, my study suggests that patch persistence and expansion may be guaranteed when R. lycioides is present, and this species may increase its presence under warmer and drier conditions. In contrast, efforts to establish Q. coccifera may be doomed if not restricted to particularly suitable microsites (N facing slopes, deep soils). My results suggest that abandoned terraces and ridges acted as local refugia and sustained colonization of abandoned S. tenacissima crops in slopes. However, these evidences are weak in this respect, and further studies are needed to validate this theory.

Chapter 1

APPENDIX Pearson correlation coefficients relating species cover and NMDS axes 1 and 2. The first six species correspond to patch-forming large resprouting shrubs. Significant p-values (p<0.05) are in bold.

NMDS NMDS p-value p-value Species axis 1 axis 2 NMDS 1 NMDS 2 Quercus coccifera -0.648 0.035 0.000 0.457 Pistacia lentiscus -0.234 -0.485 0.000 0.000 Rhamnus lycioides 0.740 -0.231 0.000 0.000 Juniperus oxycedrus -0.237 0.320 0.000 0.000 Ephedra fragilis 0.217 0.504 0.000 0.000 Osyris lanceolata -0.101 0.166 0.032 0.000 Ajuga iva 0.089 -0.027 0.061 0.562 Anthyllis cytisoides -0.023 -0.016 0.629 0.736 Anthyllis terniflora 0.058 -0.029 0.219 0.535 Artemisia lucentica 0.074 -0.099 0.119 0.036 Asparagus acutifolius -0.038 0.046 0.421 0.326 Asparagus horridus 0.086 0.139 0.068 0.003 Asparagus officinalis -0.081 -0.112 0.086 0.017 Asperula aristata subsp. scabra -0.091 0.000 0.053 0.998 Asphodelus fistulosus -0.095 0.128 0.045 0.007 Astragalus hispanicus -0.060 -0.093 0.207 0.049 Astragalus incanus 0.032 -0.079 0.492 0.092 Atractylis humilis 0.093 0.000 0.049 0.998 Ballota hirsuta 0.123 -0.009 0.009 0.848 Brachypodium retusum -0.045 0.041 0.342 0.386 Bupleurum fruticescens -0.161 -0.042 0.001 0.373 Carex humilis -0.069 -0.100 0.144 0.033 Centaurea aspera subsp. 0.179 0.034 0.000 0.474 stenophyla Centaurium quadrifolium 0.030 -0.102 0.525 0.030 subsp. barrelieri Ceratonia siliqua -0.062 -0.129 0.193 0.006

Overview of woody vegetation patches in Stipa tenacissima steppes

Chamaerops humilis -0.014 -0.112 0.768 0.018 Cheirolophus intybaceus 0.036 0.064 0.448 0.175 Chiliadenus glutinosus 0.074 -0.040 0.118 0.401 Cistus albidus -0.054 0.091 0.255 0.053 Cistus clusii 0.002 -0.060 0.971 0.201 Convolvulus altaeoides 0.094 0.010 0.047 0.824 Convolvulus lanuginosus -0.092 0.014 0.051 0.761 Coris monspeliensis -0.070 -0.107 0.135 0.023 Coronilla juncea -0.063 -0.069 0.184 0.146 Coronilla minima subsp. 0.065 -0.054 0.168 0.251 lotoides Dactylis glomerata 0.071 -0.096 0.132 0.041 Diplotaxis harra subsp. 0.124 0.029 0.008 0.535 lagascana Dorycnium pentaphyllum -0.128 -0.057 0.006 0.225 Echium humile 0.053 0.012 0.260 0.808 Elaeoselinum asclepium 0.076 0.015 0.110 0.743 Elaeoselinum tenuifolium -0.071 -0.014 0.132 0.764 Erica multiflora -0.184 -0.049 0.000 0.298 Eryngium campestre 0.037 -0.064 0.430 0.177 Fagonia cretica 0.196 0.084 0.000 0.075 Fumana ericoides 0.100 0.045 0.033 0.342 Fumana laevipes -0.031 -0.061 0.516 0.194 Fumana thymifolia 0.106 -0.061 0.025 0.194 Galium setaceum -0.036 0.007 0.443 0.876 Globularia alypum -0.043 -0.110 0.365 0.019 Haplophyllum linifolium -0.056 0.030 0.237 0.526 Hedysarum boveanum -0.054 -0.026 0.250 0.585 Helianthemum cinereum 0.031 -0.078 0.510 0.098 Helianthemum syriacum 0.027 0.018 0.563 0.709 Helianthemum violaceum 0.181 0.220 0.000 0.000 Helichrysum stoechas -0.052 0.066 0.271 0.163 Heteropogon contortus -0.063 -0.078 0.182 0.100 Hyparrhenia hirta -0.015 0.066 0.753 0.165 Hyparrhenia sinaica -0.007 -0.123 0.889 0.009

Chapter 1

Hypericum ericoides 0.031 -0.096 0.507 0.042 Lavandula dentata 0.000 0.037 0.994 0.440 Lithodora fruticosa -0.112 0.016 0.017 0.738 Lobularia maritima 0.127 0.159 0.007 0.001 Matthiola fruticulosa -0.084 0.063 0.076 0.183 Ononis minutissima -0.103 -0.089 0.029 0.060 Pallenis spinosa -0.112 -0.043 0.018 0.368 Paronychia suffruticosa 0.045 0.050 0.340 0.291 Phagnalon rupestre 0.080 -0.018 0.090 0.699 Phagnalon saxatile 0.192 0.108 0.000 0.022 Phillyrea angustifolia 0.008 0.081 0.869 0.086 Phlomis lychinitis -0.079 0.006 0.094 0.894 Pinus halepensis -0.076 -0.113 0.106 0.016 Plantago albicans 0.222 0.054 0.000 0.249 Polygala rupestris 0.010 -0.118 0.827 0.012 Retama sphaerocarpa -0.026 -0.041 0.589 0.384 Rhamnus alaternus -0.148 0.009 0.002 0.845 Rosmarinus officinalis -0.105 0.051 0.026 0.279 Rubia longifolia -0.086 -0.067 0.069 0.158 Rubia peregrina -0.134 -0.136 0.005 0.004 Ruta angustifolia 0.011 0.068 0.811 0.149 Salsola genistoides -0.017 0.114 0.715 0.015 Santolina chamaecyparisus 0.043 -0.107 0.362 0.024 subsp. squarrosa Satureja obovata -0.073 -0.041 0.120 0.390 Sedum album 0.120 0.128 0.011 0.007 Sedum sediforme 0.157 0.228 0.001 0.000 Sideritis leucantha 0.252 0.096 0.000 0.042 Staehelina dubia -0.103 0.019 0.029 0.685 Stipa parviflora 0.018 -0.051 0.708 0.283 Stipa tenacissima 0.094 -0.185 0.045 0.000 Teucrium buxifolium subsp. -0.049 0.009 0.299 0.846 rivasii Teucrium capitatum 0.241 0.089 0.000 0.060 Teucrium carolipaui 0.071 -0.089 0.133 0.059

Overview of woody vegetation patches in Stipa tenacissima steppes

Teucrium pseudochamaepitys -0.118 -0.064 0.012 0.174 Teucrium ronnigeri 0.125 0.017 0.008 0.726

Thymelaea argentata -0.016 -0.004 0.738 0.935 Thymelaea hirsuta 0.100 0.011 0.034 0.824 Thymus moroderi -0.042 -0.009 0.371 0.845 Thymus vulgaris -0.010 -0.095 0.836 0.043 Viola arborescens -0.109 -0.088 0.020 0.064

CHAPTER 2 Networks of plant-plant co-occurrence in semiarid steppes

Networks of plant-plant co-ocurrence in semiarid steppes

INTRODUCTION

In semiarid areas, trees and large shrubs frequently form woody patches immersed in a matrix of bare soil and smaller annual and perennial plants (Aguiar and Sala, 1999). The expansion of woody patches on areas previously dominated by herbaceous vegetation has been referred to as shrub encroachment, an important and largely studied phenomenon in semiarid areas worldwide (Van Auken, 2000; Eldridge et al., 2011). The effect of shrub encroachment on community composition and ecosystem functioning depends on many factors, including patch traits (Maestre et al. 2009a; Maestre et al. 2009b). These reflect the morpho-functional traits of patch-forming species and the outcome of their multiple interactions. However, despite its importance on patch and ecosystem functioning, patch composition has been largely ignored in woody encroachment studies.

In Chapter 1, I described the identity of dominant and accompanying species forming woody vegetation patches in semiarid Stipa tenacissima steppes, and classified them according to their composition (Non-metric Multi-Dimensional Scaling analysis). However, these approaches provide few insights on the wide array of interactions explaining patch formation and maintenance. The use of network analysis could help to increase our knowledge on the underlying mechanisms governing these processes.

Network theory has been used in ecological studies for decades (Paine, 1966), but only recently, network analysis has become a common approach to study ecological communities (Jordano et al., 2003; Montoya et al., 2006; Estrada, 2007). Network analysis is a powerful tool that allows for the study of the physical structure of the community, taking into account the fact that the interaction between two species not only affects actors in the pair-wise relation, but the complete network. In ecology, network theory has been mainly employed to understand mutualistic networks and food webs (Pascual and Dunne, 2006; Bascompte and Jordano, 2007; Thébault and Fontaine, 2010). In these networks, the interaction between elements is usually uni-directional (e.g. pollinators- pollinated plants, predators-preys, etc.; Bascompte and Jordano, 2007). In contrast, network analysis has scarcely been applied to evaluate the composition of plant communities. In these communities, interaction between species may be

Chapter 2 uni-directional (e.g. benefactor-beneficiary species in facilitative interactions; Verdú and Valiente-Banuet, 2008). However, because of the complex nature of plant-plant interactions and the wide range of intensity and directionality in species interactions, they may be more accurately described by bi-directional networks (such as species co-ocurrence networks). Uni-directional networks are necessarily bipartite networks, as interactions between their elements occur between nodes from different categories (‘two-mode networks’; Bascompte and Jordano, 2007). On the contrary, bi-directional networks, where all their elements interact may be studied through both approaches (bipartite and unipartite networks), as their elements may belong to different or the same level. As most studies on ecological communities focus on bipartite networks, network indices have been mostly described and implemented for these uni-directional bipartite networks. In contrast, few studies have employed these indices to gain further insight on unipartite networks (but see Saiz and Alados, 2011a, 2011b), and none has compared the suitability of both approaches for the study of co-occurrence ecological networks. Furthermore, to my knowledge, no studies of plant-plant networks have taken into account the intensity of inter-specific links (weights) to describe network structure. The use of interaction weights provides more and more precise information about the functioning of the system. Substantial differences are found when using quantitative versus qualitative interaction data (Scotti et al., 2007).

The use of a network approach to study plant community structure allows for new insights into emergent properties and patterns of the whole community. This information can be crucial to optimize management decisions, and improve conservation and restoration actions. For example, network analysis provides tools to define species roles within the community, and thus identify priority species in conservation plans (Jordán and Scheuring, 2002). In addition, ecological restoration has gradually shifted its emphasis from restoring species to restoring species interactions, and thereby ecosystem functions and services (Henson et al. 2009; Heleno et al. 2010; Tylianakis et al., 2010; Cortina et al., 2011; Devoto et al., 2012). Ecological interactions are key drivers of biodiversity, as the prevalence of all organisms relies upon interactions with other individuals.

Networks of plant-plant co-ocurrence in semiarid steppes

Here, I used network theory to explore the properties of woody patch communities in S. tenacissima steppes. These patches represent a suitable environment to test hypotheses on community organization and network theory as they are physically isolated, and they show heterogeneous physical and biotic structures. In addition, patches are organized at two levels: dominant patch- forming species and accompanying species, which allows for the use of a bipartite approach. As I have previously explained (Chapter 1), dominant species are resprouting shrubs which form clearly defined patches, because of their large size and their ability to create new microhabitats. Accompanying species are smaller shrubs, perennial grasses and annuals that thrive underneath the dominant species, and do not form patches by themselves. The intensity and directionality of the interaction between these two groups of species may be heterogeneous, which impairs the use of unidirectional networks. I evaluated 27 bipartite quantitative plant co-occurrence networks to explore the structure of woody patch communities in semiarid steppes. First, I assessed network structure diversity across contrasting sites by using four commonly used indices for bipartite networks: connectance, nestedness, modularity and network specialization. Second, I discussed the suitability of these network indices for the description of co-occurrence networks and compared these results with a unipartite approach. Third, I examined diversity within communities by assessing species specialization and their linkage level, and compared these interaction patterns across sites. Finally, I identified the contribution of each species to the maintenance of community structure.

MATERIALS AND METHODS

Study site

The same 15 catchments of S. tenacissima steppes in SE Spain described in Chapter 1 were studied in this Chapter. As explained above, catchments were divided into 3-6 homogeneous units, according to exposure, topography and plant cover, by using aerial photographs and field surveys, giving a total of 27 homogeneous units. In addition, in each catchment, 30 woody patches were randomly selected, uo to a total of 450 patches. Each patch was assigned to its

Chapter 2 corresponding unit. In each patch, I recorded the presence of all perennial species that occurred in more than 5% of the patches, as recommended by McCune and Grace (2002) for community analysis. In total, there were 6 dominant species and 92 accompanying species in all catchments (See Chapter 1 for details on the methodology).

Network analysis

Each homogeneous unit was considered a network whose nodes were species, and interactions corresponded to co-occurrences of two species in the same patch of a given unit (Appendix 1). The most commonly used interaction strength surrogate for bipartite networks is interaction frequency (proportion of cases where a given interaction is observed; Vázquez et al., 2005). However, this surrogate is affected by the abundance of the participating species. I took this influence into account to assign link weights, and generate an index of interaction strength using the frequency of the co-occurrence between species weighted by the abundance of each of the co-occurring species. This index was calculated as

(Number of patches where spi and spj co-occur * Total number of patches) *

(Number of patches where spi is present * Number of patches where spj is present)-1. Values from this index were multiplied by 10 and rounded to satisfy the requirements of network indices algorithms to be integers. This index is analogous to those used in mutualistic weighted networks, which are standardized by abundance of flowers and census time (Castro-Urgal et. al, 2012).

I described the structure of the 27 bipartite networks at two levels: network level and species level. At network level, I estimated connectance, nestedness, modularity and network specialization. At species level, I estimated species specialization, weighted betweenness, generalization level and the role of each species in network structure. Connectance, nestedness and modularity indices were also estimated for unipartite networks. Network and species specialization have not been developed for unipartite networks yet. This set of network indices were selected because they are widely used, and they provide complementary information that is relevant to understand woody patches and their assembly rules. Each index is described in detail below.

Networks of plant-plant co-ocurrence in semiarid steppes

Network level

Connectance (C) is the number of actual links over all possible links in the network (Dunne et al. 2002).

Nestedness (N) refers to the way that elements of a particular set are linked to elements of a second set, dominant and accompanying species in this study. Networks are highly nested when species with low connectance interact with species that form perfect subsets of the species with which highly connected species interact. Nestedness was calculated using the weighted-interaction nestedness estimator (WINE; Galeano et al., 2008). WINE is based on the calculation of Manhattan distance between cells that assigns positions to each cell with the lowest number of links. Then, a new distance is calculated for each cell, and its mean is the weighted-interaction distance (WIN). WIN is tested for significance using a null model generated by 100 random matrices, and then it is normalized into WINE to allow comparison between network matrices. WINE ranges between cero (equivalent to average WIN of random matrices) and one (equivalent to maximal nestedness matrix), although negative values may be possible (Almeida-Neto et al., 2007; Galeano et al., 2008).

Modularity index (Q) is a measure of the degree to which a network is organized into clearly separated modules. Nodes in a module are highly connected between them, and poorly connected with nodes from other modules. Modularity index was calculated using the QuanBiMo algorithm (Dormann and Strauss, 2014) for quantitative bipartite networks based on the Louvain method for weighted networks (Newman, 2004; Blondel et al., 2008). The value of Q is 0 when the number of within-module links is similar to random. When Q approaches 1, it indicates strong community structure. I tested Q for significance by comparing each empirical network with 100 random networks with the same marginal totals.

Network specialization index (H2’) measures the degree of niche divergence among species, obtained by comparing the observed value with an expected probability distribution of interaction frequencies which assumes that all species interact with their partners in proportion to their observed total frequencies (Blüthgen et al., 2006). It ranges from 0 (low specialization, high niche

Chapter 2 overlap) to 1 (high specialization, low niche overlap). It is important to note the differences between H2’ and the interaction strength surrogate (“interaction weights”) that I used in this study (see above). Whilst both of them take into account species abundance, interaction weights only consider the two species involved in each interaction, and H2’ (and d’) consider the potential links that all species can establish with certain species. The first one is a measure of the community interaction, and the second one is a measure for each species.

Species level

Species specialization index (d’), as the previous network specialization index (H’2), measures the degree of species specialization/generalization, taking into account their abundance and the frequency of interactions (Blüthgen et al., 2006).

Weighted betweenness is a measure of the position of the node within the network, and describes the importance of a node as a connector between different parts of the network by estimating the proportion of shortest paths between pairs of species passing through each species, considering the number and the weight of links (Freeman, 1979; Martín-González et al., 2010).

The generalization level of a species is defined as the proportion of species it interacts with, out of the total possible in the network (normalized degree; Martín-González et al., 2010).

Finally, the role of each species in each network was assessed following a simplification of the criteria of Guimerà and Amaral (2005) by Olesen et al. (2007). While the former identified seven roles, the later distinguished only four. I sorted all species into peripherals, connectors, module hubs and network hubs. The last three are termed generalists. A peripheral species has a few links inside its own module and rarely any to other modules. A connector species links modules together and is thus important for network coherence. A module hub is important for the coherence of its own module. A network hub is important to the coherence of both the network and its own module.

Networks of plant-plant co-ocurrence in semiarid steppes

Statistical analysis

All indices for bipartite networks except the species role identification, were obtained using R 3.1.1 statistics software, bipartite package (R Development Core Team, 2014). To obtain the modularity and nestedness of unipartite networks I used PAJEK (Batagelj and Mrvar, 1998) and BINMATNEST (Rodríguez- Gironés and Santamaría, 2006), respectively. The role of species was calculated with software NETCARTO (Guimerà and Amaral, 2005). NETCARTO and BINMATNEST are only implemented for unweighted networks; thus, only in these cases, I considered presence/absence of species co-occurrences, irrespective of the frequency of the co-occurrence and its abundance. Network indices were tested for significance by generating 100 randomizations for each network, using fixed marginal totals. Indices for bipartite and unipartite networks were compared using Student’s-t test for paired samples. I estimated Pearson correlation between network indices. These analyses were done using R 3.1.1 statistics software (R Development Core Team, 2014)

RESULTS

Network level

In the bipartite approach of plant co-occurrence networks, connectance ranged between 0.50 and 0.89 (Table 1). Modularity ranged from 0.10 to 0.41, all networks were significantly modular, i.e., their modularity index was significantly higher than the mean modularity index of 100 random networks (p<0.05). Each network had 2-4 modules and each module had between 3 and 34 species (Fig. 1). Only networks with more than 32 species had 4 modules. Nestedness ranged from -0.33 to 0.55, but only 15 units were significantly nested. Network specialization ranged from 0.08 to 0.47 and it was significant for all networks (p<0.05). The unipartite approach showed lower connectance and higher number of modules and nestedness, and a marginally significant increase in modularity (Student’s t test, p<0.05; Table 1).

Chapter 2

Table 1. Network indices estimated for 27 networks of plant-plant co-occurrence in semiarid steppes. Dominant and accompanying species were considered as different levels for the bipartite approach. Significant values for modularity, nestedness and network specialization are shown in bold (100 random simulations, p<0.05). A summary of each index and Student's t test results from comparing bipartite indices with their unipartite counterpart are shown. C=Connectance, M=Modularity, nM=number of modules, N=Nestedness, H'2=Network specialization, Std. Error= Standard error, Min=minimum value, Max=maximum value, d.f.=degrees of freedom. Accom- Dominant Bipartite approach Unipartite approach Network panying species C M nM N H' C M nM N species 2 1 6 36 0.70 0.28 3 0.24 0.26 0.75 0.16 4 0.71 2 6 20 0.61 0.32 3 -0.08 0.35 0.64 0.34 4 0.61 3 4 30 0.67 0.35 3 0.27 0.35 0.56 0.27 4 0.81 4 2 31 0.85 0.18 2 0.36 0.20 0.50 0.40 5 0.71 5 3 32 0.73 0.29 3 0.35 0.30 0.70 0.25 4 0.71 6 3 13 0.67 0.30 3 0.53 0.31 0.65 0.46 2 0.74 7 4 20 0.76 0.25 3 0.05 0.24 0.68 0.29 3 0.77 8 4 30 0.64 0.34 3 -0.02 0.41 0.58 0.33 4 0.77 9 5 38 0.64 0.33 3 0.43 0.31 0.59 0.30 5 0.71 10 3 26 0.78 0.22 3 0.27 0.22 0.76 0.18 4 0.72 11 3 35 0.64 0.36 2 0.50 0.39 0.49 0.40 5 0.66 12 2 22 0.82 0.20 2 0.20 0.25 0.58 0.40 3 0.70 13 3 38 0.67 0.32 3 0.18 0.46 0.42 0.27 4 0.83 14 5 39 0.50 0.41 3 0.55 0.47 0.50 0.32 5 0.79 15 4 28 0.76 0.30 2 0.19 0.29 0.65 0.30 3 0.73 16 6 39 0.81 0.22 3 -0.06 0.16 0.68 0.28 3 0.70 17 5 24 0.78 0.24 3 -0.27 0.24 0.70 0.27 3 0.76 18 4 39 0.87 0.21 3 0.11 0.15 0.74 0.29 3 0.72 19 6 36 0.68 0.27 4 0.26 0.25 0.68 0.26 4 0.72 20 6 20 0.73 0.22 3 0.13 0.19 0.66 0.38 2 0.74 21 4 18 0.65 0.33 3 -0.14 0.38 0.58 0.42 3 0.59 22 6 27 0.72 0.23 4 -0.10 0.24 0.52 0.35 4 0.81 23 5 16 0.89 0.10 3 -0.33 0.08 0.78 0.19 3 0.72 24 5 38 0.67 0.29 4 0.25 0.30 0.61 0.31 4 0.73 25 4 28 0.68 0.24 4 0.42 0.25 0.62 0.26 5 0.67 26 4 29 0.67 0.33 3 0.08 0.36 0.59 0.30 3 0.78 27 3 27 0.73 0.30 2 0.36 0.31 0.60 0.38 4 0.73 Mean 4 29 0.71 0.28 3 0.18 0.29 0.62 0.31 4 0.73 Std. Error 0.2 1.5 0.02 0.01 0.1 0.04 0.02 0.02 0.01 0.2 0.01 Min 2 13 0.50 0.10 2 -0.33 0.08 0.42 0.16 2 0.59 Max 6 39 0.89 0.41 4 0.55 0.47 0.78 0.46 5 0.83 Median 4 29 0.70 0.29 3 0.19 0.29 0.62 0.30 4 0.72 Student's t 5.43 -2.06 -3.76 -11.95 d.f. 26 26 26 26 p-value <0.001 0.050 0.001 <0.001

Networks of plant-plant co-ocurrence in semiarid steppes

50 module 1 45 module 2 module 3 40 module 4

Number ofNumber species 15

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Network

Figure 1. Number of modules and species per module of the 27 plant co-occurrence networks studied in woody patches of Stipa tenacissima steppes.

Bipartite network indices were highly correlated. Connectance was strongly and negatively correlated to modularity and network specialization, and slightly and negatively correlated to nestedness (Table 2). Modularity was also strongly and positively correlated to network specialization, and slightly and positively correlated to nestedness. Only connectance, on one side, and modularity and number of modules, on the other, were negatively correlated in unipartite networks. Network size was correlated with the number of modules in the unipartite representation (Pearson’s r=0.539, d.f.= 25, p-value=0.004). No correlations were found between network size and bipartite network indices.

Chapter 2

Table 2. Pearson correlation of network indices for bipartite and unipartite plant co- occurrence networks (N=27 networks corresponding to woody patches in 27 homogeneous Stipa teenacissima steppes). p-values are shown in parenthesis. Significant correlations (p<0.05) are in bold.

bipartite bipartite bipartite bipartite bipartite unipartite unipartite unipartite Modularity Modularity Modularity mber modules modules mber Nestedness Connectance Connectance Connectance Specialization Specialization

Nu modules Number Modularity -0.892 ------bipartite (<0.001) ------Number -0.252 -0.034 ------modules bipartite (0.206) (0.867) ------Nestedness -0.396 0.433 -0.168 - - - - - bipartite (0.041) (0.024) (0.403) - - - - - Specialization -0.852 0.920 -0.081 0.359 - - - - bipartite (<0.001) (<0.001) (0.690) (0.066) - - - - Connectance 0.494 -0.516 0.169 -0.325 -0.686 - - - unipartite (0.009) (0.006) (0.398) (0.098) (<0.001) - - - Modularity -0.199 0.232 -0.347 0.204 0.284 -0.538 - - unipartite (0.319) (0.244) (0.076) (0.308) (0.152) (0.004) - - Number -0.394 0.335 0.053 0.471 0.361 -0.453 -0.158 - modules unipartite (0.042) (0.088) (0.793) (0.013) (0.064) (0.018) (0.431) - Nestedness -0.033 0.088 0.151 0.043 0.146 -0.246 -0.201 -0.057 unipartite (0.871) (0.663) (0.452) (0.830) (0.468) (0.217) (0.315) (0.779)

Species level

Considering all units, the normalized degree (the degree scaled according to the maximum number of possible links) for dominant species ranged between 0.23 and 1, and between 0.17 and 1 for accompanying species (Appendix 2). Species specialization (d’) ranged from 0 to 0.78, with 98% of the cases (all species in all units) showing values under 0.5. Only four cases with d’ >0.5 corresponded to accompanying species (Centaurea aspera, Rosmarinus officinalis, Asparragus horridus and A. acutifolius). The remaining 16 cases showing high levels of

Networks of plant-plant co-ocurrence in semiarid steppes specialization were dominant species. Rhamnus lycioides always behaved as a generalist, as maximum d’ value for this species was 0.37. Rhamnus lycioides was also the dominant species with the highest degree (ANOVA and post-hoc Tukey-

HSD test, F=10.7, d.f.=5, p<0.05; Fig. 2). Species specialization was negatively correlated to normalized degree in all units (Pearson’s r< -0.4, p<0.005 in all units). Weighted betweenness was cero in 83% of the cases (all species in all units); otherwise, it ranged between 0.004 and 1. Because of the high amount of zeroes, I did not perform correlation analysis with this variable.

1.0 b Normalized degree Species specialization (d')

0.8 a a a

a 0.6 a

A A 0.4 Relative value Relative AB AB AB

0.2 B

0.0

Ephedra fragilis Osyris lanceolataPistacia lentiscus Quercus coccifera Rhamnus lycioides Juniperus oxycedrus

Figure 2. Normalized degree and species specialization index (d’) for dominant species in bipartite plant co-occurrence networks in woody patches of Stipa tenacissima steppes. Mean and SE for each dominant species across all networks are shown. Different letters indicate significant differences between species for each index (ANOVA and post-hoc Tukey-HSD test, F=10.70, d.f.=5, p<0.05 for normalized degree and F=7.22, d.f.=5, p<0.05 for species specialization).

Rhamnus lycioides (100%) and Pistacia lentiscus (93%) were the two dominant species present in most networks (Fig. 3). Only two accompanying species were present in all networks: Stipa tenacissima and Brachypodium

Chapter 2 retusum. They both had high normalized degree (0.87 and 0.99, respectively) and low specialization (0.07 and 0.01, respectively). In total, 12 accompanying species out of 92 were present in >75% of the networks. Accompanying species only showed peripheral and connector roles, and none of them acted as module or network hubs (Appendix 2). Conversely, all dominant species played all roles, except for Ephedra fragilis, which was never a connector, and Osyris lanceolata, which was never a network hub (Fig. 3; Appendix 2). In most cases where E. fragilis or O. lanceolata were present in a network, they acted as peripheral nodes. Rhamnus lycioides and P. lentiscus were hubs, either network or module hubs, in most networks where they were present. Juniperus oxycedrus and Quercus coccifera were the less abundant dominant species, and their role in the networks was similar: they were predominantly peripherals and module hubs.

18 16 25 27 14 13 100

Network hub Module hub 60 Peripheral Connector

Proportion of units (%)

Ephedra fragilis Osyris lanceolataPistacia lentiscus Quercus coccifera Rhamnus lycioides Juniperus oxycedrus

Dominant species Figure 3. Network roles of dominant species in bipartite co-occurrence networks of dominant and accompanying species forming woody patches in semiarid Stipa tenacissima steppes. Numbers on top of the bars indicate the number of networks where each dominant species was present.

Networks of plant-plant co-ocurrence in semiarid steppes

In three networks, I only found peripheral species, and most networks (14/27) only comprised two types of species, mostly module hubs and peripherals (Fig. 4). Network hubs were only present in networks with three or more roles. Peripheral species dominated in most units, with the exception of four of them, where connector was the most abundant network role. The maximum number of network hub species in a network was three, and the maximum number of module hub species was four.

100

60 Network hub Module hub Peripheral 40 Connector

Proportion of (%) Proportion species 20

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Network Figure 4. Proportion of species role in each plant co-occurrence network of woody patches in Stipa tenacissima steppes.

Chapter 2

DISCUSSION

Network structure

The indices used in this study to explain network structure –connectance, nestedness, modularity and specialization- are consistent to describe the global structure of a network, as they provide complementary information about network cohesion, specificity of link establishment, and community segregation into sub-communities (Fortuna et al., 2010). Species in woody patches were highly connected, even more than in other plant networks (Verdú et al., 2010; Saiz and Alados, 2011b). Plant co-occurrence networks may have fewer restrictions to establish links than mutualistic networks and food webs, because there are fewer morphological, phenological and physiological requirements to satisfy in the former. High connectance may contribute to increase network robustness against disturbance (Dunne et al., 2002), which is probably the result of high redundancy (species that overlap in the species they interact with; Devoto et al., 2012). Modularity in woody vegetation patches was moderate, similar to that found in mutualistic networks and food webs (Thébault and Fontaine, 2010). However, the number of modules per network was low, similar to some antagonistic networks but lower than in many mutualistic networks (Krause et al., 2003; Olesen et al., 2007; Travesset et al., 2013). Although there was no relation between network size and modularity or number of modules, the maximum number of modules (4 modules) was only reached in those networks with relatively high number of species (more than 32 species). In general, the size of plant co-occurrence networks was in the lower range of those described for antagonistic and mutualistic networks (Jordano et al., 2003; Saiz and Alados, 2011b). In addition to their low specificity, compartmentalization in plant co-occurrence networks seems unlikely. Low compartmentalization may facilitate the spread of disturbances (Krause et al., 2003 and references therein). However, due to the nature of plant co-occurrence networks, the effects of low modularity may be attenuated, as link specificity is more relaxed in these than in mutualistic and antagonistic networks.

In this study, nestedness reached a maximum of 0.55, significantly lower than the values found in mutualistic networks and food webs (Bascompte et al., 2003). Low values of nestedness reflect low specificity in dominant-accompanying

Networks of plant-plant co-ocurrence in semiarid steppes species interactions (Bascompte et al., 2003). In addition, seven of the studied communities showed a trend towards negative nestedness or ‘anti-nestedness’, which corresponds to networks that are less nested than expected by chance (Galeano et al., 2008). No specific model can be assigned to anti-nested matrices as they may reflect various ecological contexts. Thus, co-ocurrence matrices that depart from perfect nestedness may be arranged in different structural patterns, described as checkerboard, high-turnover, compartmented and random (Almeida- Neto et al., 2007). In this study, communities with negative nestedness showed contrasted values of modularity and connectance, which hampers ascription to any of the structural patterns described for anti-nested matrices. The low values of nestedness observed in woody patches suggest that species in these communities are mostly generalists and the level of specificity in plant-plant interactions may be low. Low nestedness may be related to a decrease in the stability and robustness in response to disturbances (Memmot et al., 2004, 2007; Okuyama and Holland, 2008; Verdú and Valiente-Banuet, 2008; Thébault and Fontaine, 2010), and a reduction in community ability to support high levels of biodiversity (Bastolla et al., 2009). Woody patches act as ecosystem engineers, modifying the microenvironment around them and favoring the establishment of new species (Maestre and Cortina, 2005; Maestre et al., 2009b; Amat et al., 2015). However, the balance between positive and negative interactions is dynamic, and may change with time, distance, identity of the dominant species, and patch community composition (Castro et al., 2002; Maestre and Cortina, 2004a; Pugnaire et al., 2011; Amat et al., 2015). Results of the present study suggest that, in general terms, microhabitats created underneath different woody patches acted as generalist filters, and did not select for particular sets of accompanying species. In agreement with the results obtained for nestedness, network specialization was also low, indeed lower than specialization levels found in mutualistic networks (Blüthgen et al., 2006; Traveset et al., 2013).

In woody patches, connectance was negatively correlated to nestedness, modularity and network specialization. The lack of structure of highly connected communities may reflect the difficulty to organize large numbers of interactions, added to the fact that interactions change overtime (Fortuna et al., 2010). In drylands, temporal variation in the intensity and direction of plant-plant interactions results from changes in stress severity, phenology and ontogeny

Chapter 2

(Morris and Wood, 1989; Maestre and Cortina, 2004a; Amat et al., 2015). Networks with different levels of specialization had similar values of connectance, which indicates that the weight of the interactions was not evenly distributed between networks (Blüthgen et al., 2006). The strong and negative correlation between network specialization and connectance, and the strong and positive correlation between network specialization and modularity, suggest that a higher ability to form sub-communities was linked to more specialized interactions. This may allow the coexistence of different types of woody patches in the same area. However, we may note that modularity and specialization were both relatively low, and woody patch communities should be considered overall generalists and scarcely compartmented. My results partly agree with those of Thébault and Fontaine (2010) who found that connectance was consistently and positively related to nestedness, and negatively related to modularity, for both mutualistic networks and food webs. These authors claimed that relationships between connectance, nestedness and modularity could be used as surrogates of community stability, as they are related to species persistence after disturbance.

Bipartite vs. unipartite approaches for co-occurrence networks

I applied four indices commonly used in the analysis of ecological bipartite networks -mutualistic networks and food webs- for the study of plant co- occurrence networks. In bipartite networks interactions are directed, and no interaction between elements belonging to the same level is possible. In plant co- occurrence networks, interactions are un-directed and possible between elements from the same level (whether or not there is more than one level). However, a bipartite approach to co-occurrence networks may provide insight on community structure, specifically the structure of interactions. This approach can be particularly valuable in studies focusing on the effect of a particular group of species on another (e.g., benefactor and beneficiary species in plant facilitation studies; Verdú and Valiente-Banuet, 2008; Verdú et al., 2010). Furthermore, this approach may be useful in co-occurrence networks when two or more groups can be discriminated, based on morphological, functional or ontogenetic traits. In my study, I focused on the relationship between dominant and accompanying species. While all species co-occur in woody patches (i.e., unipartite networks), their morphology, function, distribution and impact on community composition of

Networks of plant-plant co-ocurrence in semiarid steppes the two groups of species differ: large resprouting shrubs, which form independent entities and predominantly occur in patches vs. small shrubs and grasses, which are favored by woody patches but can thrive outside them.

The main disadvantage in using the bipartite approach for the study of plant co-occurrence networks is that it misses information about co-occurrence of species of the same level. For example, interaction between S. tenacissima and Cistus clusii, two accompanying species, may change from positive to neutral, depending on environmental conditions, which generates contrasting patterns of spatial distribution (Armas and Pugnaire, 2005). Similarly, positive effects between two patch-forming species, Juniperus sabina and J. communis, have been described in the bibliography (Verdú and García-Fayos, 2003). Finally, studies on bipartite and tripartite networks are more abundant than studies on unipartite networks, which has fostered the development of powerful tools for their analysis (Bascompte, 2009; Henson et al., 2009; Thébault and Fontaine 2010; Fontaine et al., 2011; Chagnon et al., 2012; Heleno et al., 2013 for bipartite networks, compared to Martín-González et al., 2010; Gómez et al., 2011; Saiz and Alados, 2011a, 2011b, 2014 for unipartite networks).

The analysis of co-occurrence networks has been surprisingly underexplored, despite the huge amount of data available. The network approach, either bipartite or unipartite, fosters our understanding of community assembly rules by providing an integrative view of community structure and composition. In this way, it complements co-occurrence studies based on pairwise interactions. In addition, network indices are associated with functional properties as robustness, stability against disturbances and extinctions, and spread of disturbances (Dunne et al., 2002; Verdú and Valiente-Banuet, 2008; Thébault and Fontaine, 2010; Gómez et al., 2011; Fontaine et al., 2011), which makes them useful tools for ecosystem management, conservation and restoration.

In my study, I found substantial differences when comparing bipartite and unipartite network indices for the same communities. In the unipartite approach, connectance was lower than in the bipartite approach, which was not unexpected, as the number of possible links is much higher when all species are pooled. Nestedness increased substantially when using a the unipartite vs. the bipartite approach. It is important to note that the algorithm used to estimate

Chapter 2 nestedness in unipartite networks (BINMATNEST; Rodríguez-Gironés and Santamaría, 2006) only deals with unweighted interactions. To my knowledge, no other algorithm has been implemented for nestedness estimation in unipartite networks. The fact that all interactions are considered equally important in the unipartite approach may overestimate the heterogeneity of interactions, which explains the increased nestedness in the unipartite version. This is interesting, because nestedness estimation in weighted networks has only been recently developed (WINE; Galeano et al., 2008), and scarcely used in ecological studies, despite the importance of interaction weights in community structure (Scotti et al., 2007).

In contrast, modularity was not affected by network type, despite the fact that the number of interactions considered in bipartite networks represents only a subset of their unipartite versions. These results suggest that the structure of links between dominant and accompanying species is key in determining community segregation. The addition of further links may thus induce the organization of new modules that are formed exclusively by accompanying species, as it is reflected in the slight increase in the number of modules observed in unipartite networks.

Diversity of woody patches communities

Communities of woody vegetation patches differed across units. Species composition was relatively similar, but the way species were connected to each other was different, which emphasizes the importance of considering species interactions in the study of community structure. For example, the size of units 4, 22 and 26 was identical, and they shared many species, but their degree, specialization and species role differed. On the contrary, other communities with similar levels of modularity or nestedness differed in size or composition.

Only dominant species acted as module hubs or network hubs, which confirms their importance as key species in maintaining community structure and diversity of S. tenacissima steppes. Pairs of dominant species played similar roles. Thus, P. lentiscus and R. lycioides were mostly hubs, whilst E. fragilis and O. lanceolata were mostly peripherals, and Q. coccifera and J. oxycedrus played different roles in similar proportions. This classification may partly respond to

Networks of plant-plant co-ocurrence in semiarid steppes their abundance. Rhamnus lycioides was present in all units, and it was the most abundant patch-forming species (see Chapter 1). Hence, it is not surprising that, as a hub, most accompanying species interacted with it. However, abundance may not be the sole factor behind hubs, as the abundance of P. lentiscus, which was present in most units, was much lower than that of R. lycioides. Indeed, Q. coccifera was present in only half of the units, but in those units it occurred in all patches. As a result, its abundance was similar to that of P. lentiscus, but its role as hub was much less predominant. The fact that equally abundant dominant species played different roles in ecological networks denotes contrasting ability to assemble communities. For example, P. lentiscus and R. lycioides may form patches with less restrictions than other dominant species. Indeed, a parallel study in S. tenacissima steppes showed that the potential to facilitate the establishment of new individuals depended on the composition of the communities of accompanying species and traits associated to the composition of dominant species, such as litter accumulation and phylogenetic distance between dominant species community and new seedlings (Amat et al., 2015; Chapter 4). It is important to note that the identification of species roles does not take into account the weight of the interactions, as no algorithm has been developed for weighted networks yet. The importance of species as hubs or connectors could change if this information was taken into account. This is particularly true for species that were present in most networks and showed high normalized degree, but were not identified as hubs in this study (B. retusum, Fumana ericoides, Globularia alypum). The observation that some species are disproportionately well linked to many other species may help to identify key species for restoration, when restoration aims at promoting biodiversity (Pockok et al., 2012).

All networks evaluated in this study corresponded to a similar habitat (S. tenacissima semiarid steppes). However, specialization, linkage and species role differed across networks. Also network-scale indices showed some degree of variability. This suggests that biotic and abiotic environmental factors, probably playing at different scales, are important in defining community composition and structure.

Chapter 2

CONCLUSION

I have demonstrated that network analysis is a powerful tool to uncover emergent properties of plant communities. Compared to antagonistic and mutualistic networks, co-occurrence networks are highly connected and less structured. Yet, network traits in the studied semiarid steppes are heterogeneous, and may be sensitive to both endogenous (species composition) and exogenous factors (environmental conditions, previous land use). Identifying the drivers of community assembly and network structure will improve our understanding of the relationship between community composition and ecosystem function, and enhance our ability to manage semiarid steppes. My findings are consistent with studies on plant-plant interactions and analysis of other types of ecological networks. Given the vast amount of information on plant co-occurrence that is currently available at different scales, the use of network analysis on these data sets could provide new insights on the underlying mechanisms of plant community assembly rules, and contribute to our understanding of ecological networks.

Networks of plant-plant co-ocurrence in semiarid steppes

APPENDIX 1. Bipartite representation of the 27 plant co-occurrence networks. Upper level corresponds to dominant species (large shrubs that form patches by themselves) and the lower level corresponds to accompanying species (smaller species present in patches). Complete species names are given in Appendix 3.

Network 1

Network 2

Network 3

Chapter 2

Network 4

Network 5

Network 6

Networks of plant-plant co-ocurrence in semiarid steppes

Network 7

Network 8

Network 9

Chapter 2

Network 10

Network 11

Network 12

Networks of plant-plant co-ocurrence in semiarid steppes

Network 13

Network 14

Network 15

Chapter 2

Network 16

Network 17

Network 18

Networks of plant-plant co-ocurrence in semiarid steppes

Network 19

Network 20

Network 21

Chapter 2

Network 22

Network 23

Network 24

Networks of plant-plant co-ocurrence in semiarid steppes

Network 25

Network 26

Network 27

Chapter 2

APPENDIX 2: Network indices at species level for the 27 plant co-occurrence networks studied in Stipa tenacissima steppes. A list of species forming each network, the degree, normalized degree, weighted betweenness, species specialization and the role of species is given. Complete species names are given in Appendix 3. See text for further details on the indices.

Network 1 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 33 0.92 0.00 0.11 network hub Pislen 32 0.89 0.00 0.13 network hub Rhalyc 34 0.94 1.00 0.06 connector Junoxy 17 0.47 0.00 0.33 peripheral Ephfra 26 0.72 0.00 0.31 network hub Osylan 10 0.28 0.00 0.66 peripheral Antcyt 5 0.83 0.00 0.06 connector Asphor 6 1.00 0.00 0.01 connector Aspoff 3 0.50 0.00 0.14 peripheral Atrhum 3 0.50 0.00 0.13 peripheral Braret 6 1.00 0.00 0.01 connector Carhum 5 0.83 0.00 0.06 connector Cenasp 2 0.33 0.00 0.60 peripheral Cenqua 4 0.67 0.00 0.09 connector Cheint 5 0.83 0.00 0.07 connector Cormin 5 0.83 0.00 0.05 connector Dorpen 3 0.50 0.00 0.14 peripheral Echihum 5 0.83 0.64 0.13 connector Erimul 4 0.67 0.00 0.13 connector Fumeri 6 1.00 0.03 0.01 connector Fumthy 3 0.50 0.00 0.13 peripheral Gloaly 5 0.83 0.00 0.08 connector Helcin 5 0.83 0.00 0.05 connector Helsyr 4 0.67 0.00 0.08 connector Helvio 5 0.83 0.00 0.06 connector Helicsto 6 1.00 0.08 0.01 connector Matfru 1 0.17 0.00 0.47 peripheral Phasax 4 0.67 0.00 0.18 connector Phiang 1 0.17 0.00 0.47 peripheral Pinhal 3 0.50 0.00 0.13 peripheral Rhaala 3 0.50 0.00 0.14 peripheral Rubper 5 0.83 0.00 0.06 connector

Networks of plant-plant co-ocurrence in semiarid steppes

Rutang 6 1.00 0.25 0.07 connector Sedalb 4 0.67 0.00 0.07 connector Sidleu 5 0.83 0.00 0.05 connector Stiten 5 0.83 0.00 0.05 connector Teucap 4 0.67 0.00 0.08 connector Teucar 4 0.67 0.00 0.08 connector Teupse 5 0.83 0.00 0.06 connector Thyhir 3 0.50 0.00 0.11 peripheral Thymor 3 0.50 0.00 0.14 peripheral Thyvul 6 1.00 0.00 0.05 connector

Network 2 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 12 0.60 0.00 0.45 connector Pislen 10 0.50 0.00 0.34 connector Rhalyc 10 0.50 0.00 0.33 peripheral Junoxy 17 0.85 0.00 0.21 connector Ephfra 11 0.55 0.00 0.37 peripheral Osylan 13 0.65 0.00 0.26 connector Antcyt 5 0.83 0.00 0.08 connector Asphor 4 0.67 0.09 0.14 peripheral Braret 6 1.00 0.00 0.01 connector Carhum 2 0.33 0.00 0.39 peripheral Cenqua 2 0.33 0.00 0.39 peripheral Cheint 2 0.33 0.00 0.39 peripheral Cormin 6 1.00 0.11 0.03 connector Dorpen 1 0.17 0.00 0.44 peripheral Fumeri 6 1.00 0.11 0.01 connector Gloaly 6 1.00 0.00 0.01 connector Helvio 1 0.17 0.00 0.50 peripheral Helicsto 4 0.67 0.05 0.17 peripheral Phasax 2 0.33 0.07 0.34 peripheral Rutang 4 0.67 0.07 0.17 peripheral Sedalb 3 0.50 0.00 0.24 peripheral Stiten 4 0.67 0.00 0.17 peripheral Teucap 5 0.83 0.38 0.14 connector Teucar 3 0.50 0.01 0.36 peripheral Teupse 5 0.83 0.11 0.07 connector Thyvul 2 0.33 0.00 0.31 peripheral

Chapter 2

Network 3 Species Normalized Weighted Species Degree specialization index Species role degree betweenness (d’) Rhalyc 30 1.00 0.00 0.17 module hub Junoxy 11 0.37 0.00 0.52 peripheral Ephfra 27 0.90 0.00 0.16 module hub Osylan 12 0.40 0.00 0.53 peripheral Antter 3 0.75 0.00 0.27 peripheral Asphor 4 1.00 0.78 0.03 peripheral Astinc 2 0.50 0.00 0.20 peripheral Atrhum 3 0.75 0.22 0.15 peripheral Balhir 2 0.50 0.00 0.20 peripheral Braret 4 1.00 0.00 0.01 peripheral Carhum 3 0.75 0.00 0.26 peripheral Cenasp 3 0.75 0.00 0.14 peripheral Conalt 2 0.50 0.00 0.20 peripheral Cormin 3 0.75 0.00 0.34 peripheral Diphar 1 0.25 0.00 0.31 peripheral Elaasc 2 0.50 0.00 0.20 peripheral Fagcre 4 1.00 0.00 0.00 peripheral Fumeri 4 1.00 0.00 0.02 peripheral Fumthy 2 0.50 0.00 0.20 peripheral Helvio 4 1.00 0.00 0.00 peripheral Matfru 1 0.25 0.00 0.31 peripheral Parsuf 2 0.50 0.00 0.20 peripheral Pharup 2 0.50 0.00 0.19 peripheral Phasax 3 0.75 0.00 0.13 peripheral Plaalb 3 0.75 0.00 0.07 peripheral Polrup 3 0.75 0.00 0.16 peripheral Sedsed 4 1.00 0.00 0.02 peripheral Sidleu 4 1.00 0.00 0.02 peripheral Stipar 2 0.50 0.00 0.19 peripheral Stiten 1 0.25 0.00 0.31 peripheral Teucap 2 0.50 0.00 0.20 peripheral Teupse 2 0.50 0.00 0.20 peripheral Teuron 3 0.75 0.00 0.12 peripheral Thyvul 2 0.50 0.00 0.20 peripheral

Networks of plant-plant co-ocurrence in semiarid steppes

Network 4 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Rhalyc 31 1.00 0.00 0.14 module hub Ephfra 22 0.71 0.00 0.26 module hub Antter 1 0.50 0.00 0.16 peripheral Asphor 2 1.00 0.00 0.01 peripheral Aspfis 2 1.00 0.00 0.03 peripheral Balhir 2 1.00 0.00 0.01 peripheral Braret 2 1.00 0.00 0.00 peripheral Cenasp 2 1.00 0.00 0.01 peripheral Conalt 2 1.00 0.00 0.00 peripheral Cormin 1 0.50 0.00 0.16 peripheral Dacglo 2 1.00 0.00 0.00 peripheral Diphar 1 0.50 0.00 0.16 peripheral Echihum 1 0.50 0.00 0.16 peripheral Erycam 2 1.00 0.00 0.03 peripheral Fagcre 2 1.00 0.00 0.00 peripheral Gloaly 1 0.50 0.00 0.16 peripheral Helvio 2 1.00 0.00 0.00 peripheral Hyphir 2 1.00 0.00 0.00 peripheral Parsuf 1 0.50 0.00 0.16 peripheral Pharup 2 1.00 0.00 0.03 peripheral Phasax 2 1.00 0.00 0.00 peripheral Plaalb 2 1.00 0.00 0.01 peripheral Polrup 2 1.00 0.00 0.00 peripheral Rubper 2 1.00 0.00 0.03 peripheral Rutang 1 0.50 0.00 0.16 peripheral Sedsed 2 1.00 0.00 0.00 peripheral Sidleu 2 1.00 0.00 0.01 peripheral Stipar 2 1.00 0.00 0.01 peripheral Stiten 2 1.00 0.00 0.00 peripheral Teucap 1 0.50 0.00 0.16 peripheral Teupse 1 0.50 0.00 0.16 peripheral Teuron 2 1.00 0.00 0.01 peripheral Thyhir 2 1.00 1.00 0.03 peripheral

Chapter 2

Network 5 Species Normalized Weighted Species Degree specialization index Species role degree betweenness (d’) Pislen 12 0.38 0.00 0.51 module hub Rhalyc 31 0.97 0.00 0.17 network hub Ephfra 27 0.84 0.00 0.23 network hub Antter 1 0.33 0.00 0.23 peripheral Asphor 3 1.00 0.02 0.05 connector Aspfis 1 0.33 0.00 0.29 peripheral Atrhum 1 0.33 0.00 0.23 peripheral Balhir 2 0.67 0.13 0.10 peripheral Braret 3 1.00 0.00 0.00 connector Bupfru 2 0.67 0.00 0.10 peripheral Carhum 2 0.67 0.13 0.14 peripheral Cenasp 2 0.67 0.00 0.10 peripheral Cormin 3 1.00 0.00 0.08 connector Fagcre 3 1.00 0.31 0.00 connector Fumeri 3 1.00 0.00 0.05 connector Fumthy 2 0.67 0.00 0.09 peripheral Galset 2 0.67 0.00 0.10 peripheral Helsyr 2 0.67 0.04 0.10 peripheral Helvio 3 1.00 0.00 0.00 connector Hyphir 2 0.67 0.00 0.10 peripheral Lobmar 2 0.67 0.00 0.11 peripheral Parsuf 2 0.67 0.00 0.09 peripheral Pharup 2 0.67 0.15 0.15 peripheral Phasax 2 0.67 0.00 0.14 peripheral Plaalb 3 1.00 0.00 0.08 connector Polrup 3 1.00 0.00 0.04 connector Sedsed 3 1.00 0.01 0.01 connector Sidleu 3 1.00 0.00 0.06 connector Stipar 2 0.67 0.00 0.12 peripheral Stiten 2 0.67 0.00 0.11 peripheral Teucap 2 0.67 0.00 0.10 peripheral Teucar 1 0.33 0.00 0.23 peripheral Teupse 2 0.67 0.00 0.10 peripheral Teuron 3 1.00 0.22 0.15 connector Thyvul 1 0.33 0.00 0.23 peripheral

Networks of plant-plant co-ocurrence in semiarid steppes

Network 6 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 5 0.38 0.00 0.43 peripheral Rhalyc 13 1.00 0.00 0.20 module hub Osylan 8 0.62 0.00 0.31 peripheral Asphor 1 0.33 0.00 0.29 peripheral Braret 3 1.00 0.00 0.04 peripheral Dorpen 1 0.33 0.00 0.29 peripheral Fagcre 3 1.00 0.00 0.04 peripheral Fumeri 3 1.00 0.00 0.04 peripheral Gloaly 3 1.00 0.00 0.05 peripheral Helvio 2 0.67 0.00 0.14 peripheral Pharup 1 0.33 0.00 0.29 peripheral Phasax 1 0.33 0.00 0.29 peripheral Stiten 3 1.00 0.00 0.08 peripheral Teucap 2 0.67 0.00 0.14 peripheral Thymor 1 0.33 0.00 0.29 peripheral Thyvul 2 0.67 0.00 0.14 peripheral

Network 7 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 18.00 0.90 1.00 0.16 peripheral Rhalyc 20.00 1.00 0.00 0.04 peripheral Ephfra 7.00 0.35 0.00 0.49 peripheral Osylan 16.00 0.80 0.00 0.30 peripheral Antcyt 3.00 0.75 0.09 0.19 peripheral Asphor 4.00 1.00 0.55 0.11 peripheral Braret 4.00 1.00 0.00 0.01 peripheral Cenqua 2.00 0.50 0.00 0.22 peripheral Chahum 2.00 0.50 0.00 0.29 peripheral Erycam 2.00 0.50 0.00 0.20 peripheral Fagcre 3.00 0.75 0.00 0.09 peripheral Fumeri 4.00 1.00 0.00 0.04 peripheral Fumthy 3.00 0.75 0.18 0.10 peripheral Gloaly 3.00 0.75 0.00 0.06 peripheral Helvio 3.00 0.75 0.09 0.07 peripheral Helicsto 3.00 0.75 0.00 0.17 peripheral Pharup 3.00 0.75 0.00 0.06 peripheral Phasax 3.00 0.75 0.00 0.07 peripheral

Chapter 2

Rutang 3.00 0.75 0.09 0.07 peripheral Sedalb 3.00 0.75 0.00 0.08 peripheral Stiten 4.00 1.00 0.00 0.02 peripheral Teucap 4.00 1.00 0.00 0.02 peripheral Teucar 3.00 0.75 0.00 0.06 peripheral Thymor 2.00 0.50 0.00 0.20 peripheral

Network 8 Species Normalized Weighted Species Degree specialization Species role degree betweenness index (d’) Pislen 21 0.70 0.00 0.44 network hub Rhalyc 26 0.87 0.00 0.22 network hub Ephfra 13 0.43 0.00 0.48 module hub Osylan 17 0.57 0.00 0.49 peripheral Antcyt 4 1.00 0.13 0.04 connector Asphor 2 0.50 0.00 0.19 peripheral Asthis 1 0.25 0.00 0.36 peripheral Atrhum 3 0.75 0.09 0.15 connector Braret 4 1.00 0.00 0.03 connector Cenqua 3 0.75 0.00 0.06 peripheral Chahum 2 0.50 0.03 0.20 peripheral Cheint 2 0.50 0.00 0.18 peripheral Cormon 2 0.50 0.09 0.25 peripheral Cormin 2 0.50 0.00 0.21 peripheral Dorpen 1 0.25 0.00 0.36 peripheral Echihum 3 0.75 0.02 0.14 peripheral Fagcre 3 0.75 0.00 0.15 connector Fumeri 3 0.75 0.00 0.05 peripheral Fumthy 3 0.75 0.00 0.05 peripheral Gloaly 4 1.00 0.00 0.01 connector Helvio 4 1.00 0.10 0.02 connector Helicsto 4 1.00 0.19 0.11 connector Hypsir 1 0.25 0.00 0.37 peripheral Pharup 2 0.50 0.00 0.33 peripheral Phasax 3 0.75 0.00 0.13 connector Polrup 2 0.50 0.00 0.22 peripheral Rosoff 1 0.25 0.00 0.36 peripheral Rutang 2 0.50 0.03 0.20 peripheral Sedalb 3 0.75 0.00 0.12 peripheral Stiten 4 1.00 0.00 0.00 connector

Networks of plant-plant co-ocurrence in semiarid steppes

Teucap 3 0.75 0.00 0.13 connector Teucar 1 0.25 0.00 0.36 peripheral Thymor 3 0.75 0.14 0.12 peripheral Thyvul 2 0.50 0.18 0.14 peripheral

Network 9 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 9 0.24 0.00 0.52 peripheral Pislen 38 1.00 0.00 0.13 module hub Rhalyc 38 1.00 0.00 0.13 module hub Ephfra 8 0.21 0.00 0.62 peripheral Osylan 28 0.74 1.00 0.20 module hub Antcyt 5 1.00 0.00 0.06 peripheral Aspacu 2 0.40 0.00 0.19 peripheral Asphor 3 0.60 0.00 0.11 peripheral Asthis 2 0.40 0.00 0.19 peripheral Astinc 3 0.60 0.00 0.10 peripheral Atrhum 3 0.60 0.00 0.10 peripheral Braret 5 1.00 0.00 0.01 peripheral Carhum 2 0.40 0.00 0.19 peripheral Cenqua 4 0.80 0.00 0.22 peripheral Chahum 4 0.80 0.00 0.19 peripheral Cormon 4 0.80 0.00 0.21 peripheral Cormin 3 0.60 0.00 0.12 peripheral Dorpen 2 0.40 0.00 0.19 peripheral Echihum 2 0.40 0.00 0.19 peripheral Fagcre 3 0.60 0.00 0.10 peripheral Fumeri 4 0.80 0.00 0.08 peripheral Fumlae 4 0.80 0.66 0.22 peripheral Fumthy 3 0.60 0.00 0.11 peripheral Gloaly 5 1.00 0.00 0.04 peripheral Helsyr 3 0.60 0.00 0.15 peripheral Helvio 4 0.80 0.00 0.19 peripheral Helicsto 3 0.60 0.00 0.10 peripheral Hetcon 3 0.60 0.00 0.10 peripheral Hyphir 2 0.40 0.00 0.19 peripheral Pharup 3 0.60 0.00 0.10 peripheral Phasax 4 0.80 0.00 0.09 peripheral Polrup 3 0.60 0.00 0.15 peripheral Rubper 5 1.00 0.34 0.15 peripheral

Chapter 2

Rutang 2 0.40 0.00 0.19 peripheral Salgen 3 0.60 0.00 0.12 peripheral Sataov 2 0.40 0.00 0.19 peripheral Sedsed 4 0.80 0.00 0.11 peripheral Sidleu 2 0.40 0.00 0.19 peripheral Stiten 4 0.80 0.00 0.07 peripheral Teucap 3 0.60 0.00 0.10 peripheral Teupse 2 0.40 0.00 0.19 peripheral Thyhir 3 0.60 0.00 0.12 peripheral Thyvul 3 0.60 0.00 0.12 peripheral

Network 10 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 13 0.50 0.00 0.40 module hub Rhalyc 26 1.00 0.00 0.08 network hub Ephfra 22 0.85 0.00 0.18 network hub Ajuiva 2 0.67 0.00 0.15 peripheral Antcyt 2 0.67 0.00 0.12 peripheral Asphor 3 1.00 0.00 0.01 connector Balhir 3 1.00 0.00 0.03 connector Braret 3 1.00 0.00 0.01 connector Cenasp 2 0.67 0.00 0.09 peripheral Conalt 1 0.33 0.00 0.25 peripheral Dorpen 2 0.67 0.00 0.10 peripheral Fagcre 3 1.00 0.00 0.01 connector Fumeri 3 1.00 0.13 0.08 connector Fumthy 2 0.67 0.00 0.10 peripheral Helvio 2 0.67 0.00 0.09 peripheral Helicsto 3 1.00 0.88 0.05 connector Pharup 3 1.00 0.00 0.00 connector Phasax 1 0.33 0.00 0.25 peripheral Plaalb 2 0.67 0.00 0.09 peripheral Polrup 1 0.33 0.00 0.25 peripheral Rutang 2 0.67 0.00 0.10 peripheral Sedalb 3 1.00 0.00 0.00 connector Sidleu 2 0.67 0.00 0.10 peripheral Stipar 2 0.67 0.00 0.22 peripheral Stiten 3 1.00 0.00 0.01 connector Teucap 3 1.00 0.00 0.01 connector

100

Networks of plant-plant co-ocurrence in semiarid steppes

Teucar 3 1.00 0.00 0.03 connector Teupse 2 0.67 0.00 0.09 peripheral Thyvul 3 1.00 0.00 0.02 connector

Network 11 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 8 0.23 0.00 0.65 module hub Rhalyc 35 1.00 0.00 0.24 network hub Ephfra 24 0.69 0.00 0.30 module hub Ajuiva 1 0.33 0.00 0.23 peripheral Antcyt 2 0.67 0.00 0.14 peripheral Artluc 3 1.00 0.00 0.17 connector Asphor 3 1.00 0.00 0.05 connector Atrhum 1 0.33 0.00 0.23 peripheral Braret 2 0.67 0.00 0.09 peripheral Carhum 1 0.33 0.00 0.23 peripheral Cenasp 2 0.67 0.00 0.09 peripheral Chiglu 1 0.33 0.00 0.23 peripheral Dacglo 3 1.00 0.00 0.17 connector Echihum 2 0.67 0.00 0.14 peripheral Fagcre 3 1.00 0.00 0.00 connector Fumeri 2 0.67 0.00 0.08 peripheral Fumlae 1 0.33 0.00 0.23 peripheral Fumthy 2 0.67 0.00 0.09 peripheral Gloaly 1 0.33 0.00 0.23 peripheral Helvio 2 0.67 0.00 0.10 peripheral Helicsto 2 0.67 0.00 0.10 peripheral Hyphir 2 0.67 0.00 0.14 peripheral Hyperi 1 0.33 0.00 0.23 peripheral Parsuf 1 0.33 0.00 0.23 peripheral Pharup 2 0.67 0.00 0.15 peripheral Phasax 2 0.67 0.00 0.14 peripheral Plaalb 3 1.00 0.66 0.08 connector Rutang 2 0.67 0.00 0.12 peripheral Sedalb 3 1.00 0.00 0.04 connector Sidleu 2 0.67 0.34 0.09 peripheral Stipar 2 0.67 0.00 0.11 peripheral Stiten 3 1.00 0.00 0.05 connector Teucap 2 0.67 0.00 0.11 peripheral Teucar 1 0.33 0.00 0.23 peripheral

101

Chapter 2

Teupse 1 0.33 0.00 0.23 peripheral Thyhir 2 0.67 0.00 0.10 peripheral Thyvul 3 1.00 0.00 0.15 connector Vioarb 1 0.33 0.00 0.23 peripheral

Network 12 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 14 0.64 0.00 0.33 module hub Rhalyc 22 1.00 0.00 0.19 module hub Artluc 1 0.50 0.00 0.22 peripheral Asphor 2 1.00 0.00 0.02 peripheral Balhir 2 1.00 0.00 0.03 peripheral Braret 2 1.00 0.00 0.00 peripheral Chahum 2 1.00 0.00 0.03 peripheral Dacglo 1 0.50 0.00 0.22 peripheral Fagcre 2 1.00 0.00 0.00 peripheral Fumeri 2 1.00 0.00 0.00 peripheral Fumthy 1 0.50 0.00 0.22 peripheral Helvio 1 0.50 0.00 0.22 peripheral Hyperi 2 1.00 0.00 0.03 peripheral Pharup 1 0.50 0.00 0.22 peripheral Rosoff 2 1.00 0.00 0.03 peripheral Rutang 1 0.50 0.00 0.22 peripheral Sedalb 1 0.50 0.00 0.22 peripheral Sidleu 2 1.00 0.00 0.01 peripheral Stipar 1 0.50 0.00 0.22 peripheral Stiten 2 1.00 0.00 0.00 peripheral Teubux 2 1.00 0.00 0.03 peripheral Teucap 2 1.00 0.00 0.03 peripheral Teucar 2 1.00 0.00 0.03 peripheral Thyvul 2 1.00 0.00 0.00 peripheral

Network 13 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 25 0.66 0.00 0.45 module hub Rhalyc 35 0.92 1.00 0.25 network hub Junoxy 16 0.42 0.00 0.67 module hub Ajuiva 1 0.33 0.00 0.25 peripheral

102

Networks of plant-plant co-ocurrence in semiarid steppes

Antcyt 3 1.00 0.00 0.00 connector Artluc 2 0.67 0.00 0.16 peripheral Asphor 2 0.67 0.00 0.09 peripheral Atrhum 2 0.67 0.00 0.18 peripheral Balhir 2 0.67 0.00 0.11 peripheral Braret 3 1.00 0.00 0.01 connector Carhum 2 0.67 0.00 0.10 peripheral Cenasp 1 0.33 0.00 0.25 peripheral Cenqua 2 0.67 0.00 0.11 peripheral Chiglu 2 0.67 0.00 0.13 peripheral Cisalb 3 1.00 0.49 0.01 connector Cormin 2 0.67 0.00 0.19 peripheral Diphar 1 0.33 0.00 0.25 peripheral Echihum 1 0.33 0.00 0.25 peripheral Erycam 2 0.67 0.00 0.13 peripheral Fagcre 3 1.00 0.00 0.01 connector Fumeri 3 1.00 0.00 0.00 connector Fumthy 3 1.00 0.00 0.01 connector Gloaly 2 0.67 0.00 0.13 peripheral Helvio 1 0.33 0.00 0.25 peripheral Helicsto 2 0.67 0.00 0.10 peripheral Pharup 1 0.33 0.00 0.25 peripheral Phasax 2 0.67 0.00 0.13 peripheral Plaalb 1 0.33 0.00 0.25 peripheral Polrup 1 0.33 0.00 0.37 peripheral Rosoff 3 1.00 0.00 0.00 connector Rutang 2 0.67 0.31 0.23 peripheral Sedalb 2 0.67 0.04 0.15 peripheral Sidleu 2 0.67 0.00 0.10 peripheral Stiten 3 1.00 0.00 0.01 connector Teubux 1 0.33 0.00 0.37 peripheral Teucap 2 0.67 0.00 0.11 peripheral Teucar 3 1.00 0.00 0.01 connector Teupse 2 0.67 0.00 0.13 peripheral Thymor 2 0.67 0.00 0.13 peripheral Thyvul 3 1.00 0.16 0.02 connector Vioarb 1 0.33 0.00 0.37 peripheral

103

Chapter 2

Network 14

Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 14 0.36 0.00 0.50 module hub Pislen 34 0.87 0.60 0.33 network hub Rhalyc 38 0.97 0.20 0.23 network hub Junoxy 8 0.21 0.20 0.74 peripheral Ephfra 3 0.08 0.00 0.67 peripheral Antcyt 4 0.80 0.07 0.26 connector Asphor 4 0.80 0.03 0.04 connector Aspfis 2 0.40 0.00 0.16 peripheral Atrhum 3 0.60 0.19 0.26 peripheral Balhir 1 0.20 0.00 0.28 peripheral Braret 5 1.00 0.08 0.07 connector Bupfru 2 0.40 0.01 0.18 peripheral Carhum 3 0.60 0.00 0.25 peripheral Cisclu 3 0.60 0.00 0.19 peripheral Conalt 2 0.40 0.00 0.16 peripheral Cormin 1 0.20 0.00 0.28 peripheral Dorpen 2 0.40 0.01 0.18 peripheral Elaasc 2 0.40 0.00 0.18 peripheral Erycam 1 0.20 0.00 0.30 peripheral Fumeri 3 0.60 0.14 0.05 peripheral Fumlae 1 0.20 0.00 0.28 peripheral Fumthy 4 0.80 0.00 0.02 connector Helcin 4 0.80 0.31 0.04 connector Helsyr 2 0.40 0.00 0.16 peripheral Helvio 3 0.60 0.00 0.14 peripheral Helicsto 3 0.60 0.00 0.13 peripheral Matfru 2 0.40 0.00 0.16 peripheral Parsuf 2 0.40 0.00 0.17 peripheral Pharup 2 0.40 0.00 0.16 peripheral Phasax 2 0.40 0.00 0.16 peripheral Plaalb 1 0.20 0.00 0.28 peripheral Polrup 3 0.60 0.00 0.13 peripheral Rosoff 2 0.40 0.00 0.53 peripheral Rutang 2 0.40 0.00 0.16 peripheral Sataov 2 0.40 0.03 0.18 peripheral Sedsed 3 0.60 0.00 0.18 peripheral Sidleu 3 0.60 0.00 0.14 peripheral Stipar 3 0.60 0.02 0.14 peripheral

104

Networks of plant-plant co-ocurrence in semiarid steppes

Stiten 4 0.80 0.00 0.28 connector Teucap 2 0.40 0.03 0.16 peripheral Teupse 3 0.60 0.00 0.16 peripheral Teuron 2 0.40 0.00 0.18 peripheral Thyvul 2 0.40 0.00 0.18 peripheral Vioarb 2 0.40 0.08 0.18 peripheral

Network 15 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 23 0.82 0.00 0.27 module hub Rhalyc 26 0.93 1.00 0.09 module hub Ephfra 8 0.29 0.00 0.67 peripheral Osylan 28 1.00 0.00 0.11 module hub Antcyt 3 0.75 0.00 0.32 peripheral Asphor 4 1.00 0.10 0.01 peripheral Braret 4 1.00 0.10 0.00 peripheral Carhum 3 0.75 0.00 0.10 peripheral Corjun 2 0.50 0.00 0.20 peripheral Cormin 3 0.75 0.00 0.09 peripheral Echihum 3 0.75 0.00 0.09 peripheral Fagcre 3 0.75 0.00 0.09 peripheral Fumeri 3 0.75 0.00 0.09 peripheral Fumlae 3 0.75 0.00 0.10 peripheral Fumthy 2 0.50 0.00 0.17 peripheral Gloaly 3 0.75 0.00 0.10 peripheral Helvio 4 1.00 0.40 0.05 peripheral Hyphir 3 0.75 0.03 0.11 peripheral Lavden 3 0.75 0.00 0.09 peripheral Pharup 3 0.75 0.00 0.09 peripheral Rosoff 4 1.00 0.00 0.00 peripheral Rubper 3 0.75 0.00 0.09 peripheral Rutang 3 0.75 0.03 0.11 peripheral Salgen 1 0.25 0.00 0.30 peripheral Sedsed 4 1.00 0.00 0.03 peripheral Sidleu 3 0.75 0.00 0.32 peripheral Stipar 3 0.75 0.20 0.10 peripheral Stiten 4 1.00 0.00 0.00 peripheral Teucap 3 0.75 0.00 0.09 peripheral Teucar 3 0.75 0.03 0.11 peripheral Teuron 3 0.75 0.13 0.12 peripheral

105

Chapter 2

Thyvul 2 0.50 0.00 0.17 peripheral

Network 16 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 30 0.77 0.00 0.16 peripheral Pislen 36 0.92 0.00 0.08 module hub Rhalyc 38 0.97 0.00 0.06 module hub Junoxy 25 0.64 0.00 0.27 peripheral Ephfra 22 0.56 0.00 0.36 peripheral Osylan 39 1.00 0.00 0.03 module hub Antcyt 3 0.50 0.00 0.18 peripheral Aspacu 5 0.83 0.00 0.09 peripheral Asphor 3 0.50 0.00 0.18 peripheral Atrhum 5 0.83 0.01 0.05 peripheral Braret 6 1.00 0.00 0.00 peripheral Bupfru 6 1.00 0.35 0.01 peripheral Carhum 6 1.00 0.00 0.02 peripheral Cersil 4 0.67 0.00 0.21 peripheral Cisalb 6 1.00 0.00 0.04 peripheral Corjun 4 0.67 0.00 0.12 peripheral Cormin 6 1.00 0.00 0.00 peripheral Elaten 4 0.67 0.00 0.14 peripheral Erimul 5 0.83 0.00 0.10 peripheral Fagcre 5 0.83 0.00 0.07 peripheral Fumeri 6 1.00 0.00 0.01 peripheral Fumlae 3 0.50 0.00 0.17 peripheral Fumthy 5 0.83 0.00 0.06 peripheral Gloaly 6 1.00 0.00 0.01 peripheral Hedbov 4 0.67 0.00 0.21 peripheral Helcin 2 0.33 0.00 0.25 peripheral Helsyr 6 1.00 0.00 0.04 peripheral Helvio 6 1.00 0.00 0.01 peripheral Helicsto 5 0.83 0.00 0.08 peripheral Lavden 5 0.83 0.00 0.07 peripheral Matfru 6 1.00 0.00 0.02 peripheral Onomin 5 0.83 0.00 0.06 peripheral Palspi 5 0.83 0.22 0.08 peripheral Pharup 5 0.83 0.00 0.10 peripheral Rosoff 6 1.00 0.00 0.02 peripheral Rubper 4 0.67 0.00 0.12 peripheral

106

Networks of plant-plant co-ocurrence in semiarid steppes

Rutang 6 1.00 0.00 0.00 peripheral Sedsed 4 0.67 0.00 0.11 peripheral Stipar 6 1.00 0.00 0.01 peripheral Stiten 4 0.67 0.00 0.10 peripheral Teucar 3 0.50 0.00 0.21 peripheral Teupse 5 0.83 0.14 0.08 peripheral Teuron 6 1.00 0.00 0.01 peripheral Thyvul 3 0.50 0.00 0.22 peripheral Vioarb 6 1.00 0.28 0.02 peripheral

Network 17 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 24 1.00 1.00 0.02 connector Rhalyc 15 0.63 0.00 0.37 peripheral Junoxy 16 0.67 0.00 0.34 connector Ephfra 14 0.58 0.00 0.41 peripheral Osylan 24 1.00 0.00 0.01 connector Asphor 5 1.00 0.00 0.00 connector Atrhum 3 0.60 0.00 0.20 connector Braret 5 1.00 0.00 0.01 connector Carhum 3 0.60 0.00 0.19 peripheral Cisalb 3 0.60 0.00 0.20 peripheral Corjun 4 0.80 0.43 0.05 connector Cormin 5 1.00 0.03 0.00 connector Fagcre 5 1.00 0.00 0.00 connector Fumeri 5 1.00 0.00 0.00 connector Fumlae 4 0.80 0.00 0.11 connector Fumthy 3 0.60 0.00 0.20 peripheral Gloaly 4 0.80 0.20 0.05 connector Hedbov 3 0.60 0.00 0.20 peripheral Helvio 5 1.00 0.03 0.00 connector Helicsto 3 0.60 0.00 0.24 peripheral Lavden 4 0.80 0.00 0.11 connector Onomin 3 0.60 0.00 0.24 peripheral Pharup 3 0.60 0.00 0.20 peripheral Rosoff 4 0.80 0.00 0.11 connector Rutang 4 0.80 0.31 0.05 connector Sedsed 5 1.00 0.00 0.01 connector Stipar 4 0.80 0.00 0.11 connector Stiten 3 0.60 0.00 0.20 connector

107

Chapter 2

Vioarb 3 0.60 0.00 0.24 peripheral

Network 18

Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 39 1.00 0.00 0.12 module hub Pislen 32 0.82 0.00 0.09 module hub Rhalyc 38 0.97 0.00 0.10 module hub Junoxy 26 0.67 0.00 0.27 module hub Antcyt 4 1.00 0.00 0.01 peripheral Asphor 3 0.75 0.00 0.11 peripheral Aspari 2 0.50 0.00 0.20 peripheral Aspfis 3 0.75 0.00 0.14 peripheral Atrhum 4 1.00 0.00 0.04 peripheral Braret 4 1.00 0.00 0.01 peripheral Bupfru 2 0.50 0.03 0.20 peripheral Carhum 4 1.00 0.00 0.00 peripheral Cisalb 4 1.00 0.00 0.01 peripheral Cisclu 4 1.00 0.00 0.02 peripheral Conlag 2 0.50 0.00 0.20 peripheral Cormin 4 1.00 0.20 0.07 peripheral Dorpen 2 0.50 0.00 0.19 peripheral Erycam 4 1.00 0.00 0.07 peripheral Fumeri 4 1.00 0.00 0.01 peripheral Fumlae 4 1.00 0.00 0.01 peripheral Fumthy 4 1.00 0.00 0.00 peripheral Helcin 4 1.00 0.29 0.00 peripheral Helsyr 4 1.00 0.00 0.01 peripheral Helvio 4 1.00 0.00 0.00 peripheral Helicsto 4 1.00 0.00 0.03 peripheral Hyphir 4 1.00 0.00 0.01 peripheral Litfru 1 0.25 0.00 0.31 peripheral Onomin 3 0.75 0.00 0.15 peripheral Pharup 3 0.75 0.00 0.11 peripheral Phasax 3 0.75 0.00 0.11 peripheral Phllyc 4 1.00 0.43 0.07 peripheral Polrup 4 1.00 0.00 0.02 peripheral Rosoff 3 0.75 0.00 0.11 peripheral Rubper 4 1.00 0.06 0.07 peripheral Sataov 4 1.00 0.00 0.02 peripheral

108

Networks of plant-plant co-ocurrence in semiarid steppes

Sedalb 2 0.50 0.00 0.20 peripheral Sedsed 4 1.00 0.00 0.02 peripheral Sidleu 2 0.50 0.00 0.20 peripheral Stipar 4 1.00 0.00 0.02 peripheral Stiten 4 1.00 0.00 0.02 peripheral Teucap 4 1.00 0.00 0.01 peripheral Teupse 4 1.00 0.00 0.02 peripheral Thyvul 4 1.00 0.00 0.00 peripheral

Network 19 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 24 0.67 0.00 0.25 connector Pislen 35 0.97 1.00 0.07 network hub Rhalyc 36 1.00 0.00 0.06 network hub Junoxy 7 0.19 0.00 0.63 peripheral Ephfra 21 0.58 0.00 0.36 peripheral Osylan 23 0.64 0.00 0.26 peripheral Antcyt 1 0.17 0.00 0.33 peripheral Antter 5 0.83 0.00 0.03 connector Asphor 3 0.50 0.00 0.24 peripheral Aspfis 2 0.33 0.00 0.21 peripheral Asthis 3 0.50 0.00 0.21 peripheral Atrhum 5 0.83 0.43 0.04 connector Braret 6 1.00 0.00 0.01 connector Carhum 4 0.67 0.02 0.09 connector Cheint 6 1.00 0.06 0.05 connector Conlag 2 0.33 0.00 0.23 peripheral Cormin 4 0.67 0.02 0.14 peripheral Echihum 3 0.50 0.00 0.23 peripheral Erimul 4 0.67 0.09 0.11 connector Fagcre 6 1.00 0.04 0.03 connector Fumeri 5 0.83 0.00 0.06 connector Fumlae 5 0.83 0.00 0.04 connector Fumthy 5 0.83 0.02 0.07 connector Gloaly 5 0.83 0.00 0.05 connector Helsyr 5 0.83 0.03 0.05 connector Helvio 6 1.00 0.01 0.04 connector Helicsto 5 0.83 0.06 0.04 connector Matfru 3 0.50 0.00 0.21 peripheral Pharup 4 0.67 0.00 0.10 peripheral

109

Chapter 2

Phasax 5 0.83 0.00 0.05 connector Polrup 4 0.67 0.00 0.33 peripheral Rosoff 2 0.33 0.00 0.23 peripheral Rubper 4 0.67 0.00 0.08 connector Rutang 3 0.50 0.00 0.27 peripheral Sedsed 4 0.67 0.00 0.13 peripheral Sidleu 4 0.67 0.00 0.09 connector Stipar 3 0.50 0.00 0.15 peripheral Stiten 6 1.00 0.00 0.05 connector Teucap 4 0.67 0.01 0.12 connector Teucar 5 0.83 0.20 0.23 connector Teupse 2 0.33 0.00 0.23 peripheral Thyvul 3 0.50 0.00 0.21 peripheral

Network 20 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 12 0.60 0.00 0.20 pheripheral Pislen 15 0.75 0.00 0.09 pheripheral Rhalyc 20 1.00 0.00 0.04 pheripheral Junoxy 12 0.60 0.00 0.22 pheripheral Ephfra 17 0.85 0.00 0.27 pheripheral Osylan 11 0.55 0.00 0.31 pheripheral Braret 6 1.00 0.00 0.00 pheripheral Carhum 6 1.00 0.08 0.00 pheripheral Cenqua 5 0.83 0.30 0.09 pheripheral Cheint 2 0.33 0.00 0.29 pheripheral Cisalb 6 1.00 0.00 0.02 pheripheral Cormin 6 1.00 0.00 0.03 pheripheral Fagcre 2 0.33 0.00 0.29 pheripheral Fumeri 6 1.00 0.00 0.04 pheripheral Gloaly 5 0.83 0.00 0.06 pheripheral Helsyr 6 1.00 0.03 0.03 pheripheral Helvio 4 0.67 0.00 0.17 pheripheral Helicsto 5 0.83 0.33 0.09 pheripheral Pharup 2 0.33 0.00 0.29 pheripheral Phasax 4 0.67 0.00 0.12 pheripheral Polrup 2 0.33 0.00 0.29 pheripheral Rubper 3 0.50 0.00 0.32 pheripheral Sedsed 2 0.33 0.00 0.29 pheripheral Stiten 6 1.00 0.00 0.01 pheripheral

110

Networks of plant-plant co-ocurrence in semiarid steppes

Teucar 6 1.00 0.25 0.02 pheripheral Thyarg 3 0.50 0.00 0.27 pheripheral

Network 21 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Pislen 13 0.72 0.00 0.22 peripheral Rhalyc 13 0.72 0.00 0.22 peripheral Ephfra 7 0.39 0.00 0.78 peripheral Osylan 14 0.78 0.00 0.29 peripheral Astinc 3 0.75 0.00 0.12 peripheral Braret 4 1.00 0.22 0.00 connector Carhum 3 0.75 0.04 0.11 peripheral Cenasp 1 0.25 0.00 0.48 peripheral Chahum 1 0.25 0.00 0.48 peripheral Fagcre 3 0.75 0.00 0.10 peripheral Fumeri 3 0.75 0.04 0.12 peripheral Fumthy 3 0.75 0.02 0.12 peripheral Gloaly 4 1.00 0.00 0.00 connector Haplin 4 1.00 0.43 0.04 connector Helvio 3 0.75 0.02 0.12 peripheral Rubper 3 0.75 0.10 0.12 peripheral Sedsed 1 0.25 0.00 0.36 peripheral Stipar 3 0.75 0.00 0.13 peripheral Stiten 4 1.00 0.12 0.00 connector Teucap 1 0.25 0.00 0.36 peripheral Teupse 2 0.50 0.00 0.24 peripheral Teuron 1 0.25 0.00 0.48 peripheral

Network 22 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 19 0.70 0.00 0.35 connector Pislen 21 0.78 0.00 0.16 peripheral Rhalyc 27 1.00 0.00 0.02 connector Junoxy 11 0.41 0.00 0.49 connector Ephfra 18 0.67 0.00 0.26 peripheral Osylan 21 0.78 0.00 0.15 peripheral Antter 3 0.50 0.00 0.20 peripheral Asphor 2 0.33 0.00 0.56 peripheral

111

Chapter 2

Braret 6 1.00 0.00 0.01 peripheral Carhum 6 1.00 0.00 0.04 peripheral Cheint 4 0.67 0.00 0.13 peripheral Cisalb 6 1.00 0.07 0.00 peripheral Dorpen 2 0.33 0.00 0.31 peripheral Erimul 2 0.33 0.00 0.31 peripheral Fagcre 5 0.83 0.00 0.04 peripheral Fumeri 3 0.50 0.00 0.22 peripheral Gloaly 6 1.00 0.00 0.04 peripheral Haplin 6 1.00 0.00 0.00 peripheral Helsyr 3 0.50 0.00 0.19 peripheral Helvio 5 0.83 0.03 0.09 peripheral Helicsto 3 0.50 0.00 0.22 peripheral Matfru 6 1.00 0.01 0.03 peripheral Phasax 6 1.00 0.01 0.02 peripheral Polrup 4 0.67 0.00 0.12 peripheral Retsph 2 0.33 0.00 0.28 peripheral Rubper 3 0.50 0.00 0.23 peripheral Sidleu 4 0.67 0.00 0.13 peripheral Stipar 6 1.00 0.00 0.01 peripheral Stiten 5 0.83 0.00 0.04 peripheral Teucap 4 0.67 0.00 0.13 peripheral Teucar 6 1.00 0.13 0.04 peripheral Teupse 4 0.67 0.00 0.13 peripheral Thyvul 5 0.83 0.76 0.06 peripheral

Network 23 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 13 0.81 0.00 0.13 peripheral Pislen 16 1.00 0.00 0.02 peripheral Rhalyc 16 1.00 0.00 0.02 peripheral Ephfra 13 0.81 0.00 0.11 peripheral Osylan 13 0.81 0.00 0.11 peripheral Antcyt 3 0.60 0.00 0.20 peripheral Braret 5 1.00 0.00 0.00 peripheral Carhum 5 1.00 0.00 0.00 peripheral Cisalb 5 1.00 0.00 0.01 peripheral Conlag 3 0.60 0.00 0.18 peripheral Fagcre 3 0.60 0.00 0.18 peripheral Fumlae 3 0.60 0.00 0.19 peripheral

112

Networks of plant-plant co-ocurrence in semiarid steppes

Gloaly 5 1.00 0.00 0.00 peripheral Haplin 4 0.80 0.00 0.08 peripheral Phasax 5 1.00 0.00 0.01 peripheral Polrup 5 1.00 0.00 0.01 peripheral Rublon 5 1.00 1.00 0.01 peripheral Rubper 5 1.00 0.00 0.01 peripheral Stipar 5 1.00 0.00 0.01 peripheral Stiten 5 1.00 0.00 0.00 peripheral Teucar 5 1.00 0.00 0.01 peripheral

Network 24 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 30 0.79 0.00 0.25 network hub Pislen 24 0.63 0.00 0.29 module hub Rhalyc 35 0.92 0.00 0.15 network hub Junoxy 29 0.76 0.00 0.20 network hub Osylan 10 0.26 0.00 0.62 peripheral Aspacu 3 0.60 0.00 0.56 peripheral Asphor 4 0.80 0.00 0.05 connector Aspfis 3 0.60 0.00 0.13 connector Atrhum 1 0.20 0.00 0.32 peripheral Braret 5 1.00 0.00 0.00 connector Bupfru 4 0.80 0.00 0.06 connector Carhum 5 1.00 0.09 0.01 connector Cormin 2 0.40 0.01 0.22 peripheral Dorpen 1 0.20 0.00 0.32 peripheral Elaasc 2 0.40 0.01 0.25 peripheral Erimul 4 0.80 0.05 0.07 connector Erycam 1 0.20 0.00 0.32 peripheral Fumeri 4 0.80 0.00 0.06 connector Fumlae 3 0.60 0.00 0.14 connector Fumthy 4 0.80 0.04 0.09 connector Gloaly 5 1.00 0.00 0.03 connector Haplin 2 0.40 0.00 0.33 peripheral Helcin 4 0.80 0.00 0.06 connector Helsyr 1 0.20 0.00 0.32 peripheral Helvio 5 1.00 0.00 0.02 connector Helicsto 4 0.80 0.00 0.09 connector Matfru 4 0.80 0.62 0.12 connector Onomin 2 0.40 0.00 0.21 peripheral

113

Chapter 2

Parsuf 2 0.40 0.02 0.21 peripheral Pharup 4 0.80 0.00 0.06 connector Polrup 5 1.00 0.00 0.02 connector Rhaala 1 0.20 0.00 0.39 peripheral Rubper 2 0.40 0.00 0.23 peripheral Rutang 5 1.00 0.07 0.05 connector Sancha 2 0.40 0.00 0.22 peripheral Sedalb 3 0.60 0.02 0.16 connector Sedsed 5 1.00 0.00 0.02 connector Stadub 4 0.80 0.00 0.08 connector Stipar 4 0.80 0.00 0.06 connector Stiten 5 1.00 0.00 0.02 connector Teupse 4 0.80 0.00 0.06 connector Teuron 4 0.80 0.05 0.06 connector Thyvul 5 1.00 0.01 0.02 connector

Network 25 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 20 0.71 1.00 0.15 peripheral Pislen 21 0.75 0.00 0.22 peripheral Rhalyc 28 1.00 0.00 0.12 network hub Junoxy 7 0.25 0.00 0.56 module hub Asphor 2 0.50 0.00 0.19 peripheral Aspfis 3 0.75 0.00 0.31 connector Braret 4 1.00 0.00 0.01 connector Bupfru 1 0.25 0.00 0.28 peripheral Carhum 3 0.75 0.00 0.05 peripheral Cheint 2 0.50 0.00 0.14 peripheral Cormin 3 0.75 0.00 0.05 peripheral Dorpen 3 0.75 0.03 0.08 peripheral Elaasc 1 0.25 0.00 0.28 peripheral Erimul 1 0.25 0.00 0.28 peripheral Fumeri 4 1.00 0.03 0.07 connector Fumlae 3 0.75 0.00 0.05 peripheral Fumthy 3 0.75 0.13 0.08 peripheral Galset 3 0.75 0.00 0.05 peripheral Gloaly 3 0.75 0.00 0.05 peripheral Helcin 2 0.50 0.00 0.18 peripheral Helsyr 1 0.25 0.00 0.28 peripheral Helvio 4 1.00 0.00 0.08 connector

114

Networks of plant-plant co-ocurrence in semiarid steppes

Parsuf 3 0.75 0.13 0.08 peripheral Phasax 2 0.50 0.00 0.19 peripheral Polrup 3 0.75 0.00 0.05 peripheral Rutang 3 0.75 0.00 0.05 peripheral Sedsed 4 1.00 0.00 0.09 connector Stiten 4 1.00 0.20 0.04 connector Teucap 1 0.25 0.00 0.28 peripheral Teupse 3 0.75 0.00 0.05 peripheral Teuron 3 0.75 0.00 0.05 peripheral Thyvul 4 1.00 0.47 0.04 connector

Network 26 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 21 0.72 0.00 0.37 module hub Pislen 12 0.41 0.00 0.48 peripheral Rhalyc 29 1.00 0.00 0.08 module hub Junoxy 16 0.55 0.00 0.49 module hub Asphor 3 0.75 0.00 0.06 peripheral Aspfis 2 0.50 0.00 0.23 peripheral Braret 4 1.00 0.00 0.01 peripheral Bupfru 3 0.75 0.00 0.17 peripheral Carhum 4 1.00 0.00 0.02 peripheral Cormin 2 0.50 0.00 0.23 peripheral Erimul 2 0.50 0.00 0.23 peripheral Fumeri 2 0.50 0.00 0.19 peripheral Fumlae 3 0.75 0.20 0.08 peripheral Fumthy 4 1.00 0.19 0.07 peripheral Galset 2 0.50 0.00 0.24 peripheral Gloaly 3 0.75 0.00 0.11 peripheral Helcin 3 0.75 0.00 0.07 peripheral Helsyr 2 0.50 0.00 0.24 peripheral Helvio 3 0.75 0.20 0.06 peripheral Helicsto 2 0.50 0.00 0.24 peripheral Pharup 1 0.25 0.00 0.31 peripheral Phasax 2 0.50 0.00 0.24 peripheral Plaalb 3 0.75 0.00 0.22 peripheral Polrup 4 1.00 0.00 0.04 peripheral Rutang 2 0.50 0.00 0.20 peripheral Sedsed 3 0.75 0.26 0.13 peripheral Stipar 2 0.50 0.00 0.23 peripheral

115

Chapter 2

Stiten 3 0.75 0.00 0.14 peripheral Teucap 3 0.75 0.00 0.15 peripheral Teucar 2 0.50 0.00 0.20 peripheral Teupse 4 1.00 0.14 0.01 peripheral Teuron 1 0.25 0.00 0.31 peripheral Thyvul 4 1.00 0.00 0.01 peripheral

Network 27 Normalized Weighted Species specialization Species Degree Species role degree betweenness index (d’) Quecoc 26 0.96 0.00 0.17 module hub Pislen 7 0.26 0.00 0.62 module hub Rhalyc 26 0.96 0.00 0.14 module hub Asphor 3 1.00 0.13 0.01 peripheral Atrhum 2 0.67 0.00 0.10 peripheral Braret 3 1.00 0.00 0.00 peripheral Bupfru 3 1.00 0.00 0.01 peripheral Carhum 2 0.67 0.25 0.10 peripheral Cheint 1 0.33 0.00 0.27 peripheral Conlag 2 0.67 0.00 0.10 peripheral Cormin 3 1.00 0.00 0.17 peripheral Erimul 3 1.00 0.00 0.17 peripheral Fumeri 2 0.67 0.00 0.10 peripheral Fumlae 2 0.67 0.00 0.11 peripheral Fumthy 2 0.67 0.25 0.11 peripheral Galset 2 0.67 0.00 0.11 peripheral Helcin 2 0.67 0.00 0.10 peripheral Helsyr 2 0.67 0.00 0.10 peripheral Helvio 2 0.67 0.00 0.10 peripheral Helicsto 2 0.67 0.00 0.10 peripheral Pharup 3 1.00 0.00 0.17 peripheral Phasax 2 0.67 0.00 0.10 peripheral Plaalb 1 0.33 0.00 0.23 peripheral Polrup 2 0.67 0.00 0.10 peripheral Rubper 2 0.67 0.13 0.10 peripheral Rutang 2 0.67 0.13 0.10 peripheral Sedsed 2 0.67 0.13 0.11 peripheral Stiten 2 0.67 0.00 0.10 peripheral Teupse 3 1.00 0.00 0.00 peripheral Thyvul 2 0.67 0.00 0.10 peripheral

116

Networks of plant-plant co-ocurrence in semiarid steppes

APPENDIX 3. Species names for the species codes used through this chapter. Dominant species are in bold.

Code Species name Ajuiva Ajuga iva Antcyt Anthyllis cytisoides Antter Anthyllis terniflora Artluc Artemisia lucentica Aspacu Asparagus acutifolius Asphor Asparagus horridus Aspoff Asparagus officinalis Aspari Asperula aristata subsp. scabra Aspfis Asphodelus fistulosus Aspram Asphodelus ramosus Asthis Astragalus hispanicus Astinc Astragalus incanus Atrhum Atractylis humilis Atrsem Atriplex semibaccata Balhir Ballota hirsuta Braret Brachypodium retusum Bupfru Bupleurum fruticescens Carhum Carex humilis Cenasp Centaurea aspera subsp. stenophyla Cenqua Centaurium quadrifolium subsp. barrelieri Cersil Ceratonia siliqua Chahum Chamaerops humilis Cheint Cheirolophus intybaceus Chiglu Chiliadenus glutinosus Cisalb Cistus albidus Cisclu Cistus clusii Conalt Convolvulus altaeoides Conlag Convolvulus lanuginosus Cormon Coris monspeliensis Corjun Coronilla juncea Cormin Coronilla minima subsp. lotoides Dacglo Dactylis glomerata

117

Chapter 2

Diabro Dianthus broteroi Diphar Diplotaxis harra subsp. lagascana Dorpen Dorycnium pentaphyllum Echihum Echium humile Elaasc Elaeoselinum asclepium Elaten Elaeoselinum tenuifolium Ephfra Ephedra fragilis Erimul Erica multiflora Erycam Eryngium campestre Fagcre Fagonia cretica Fumeri Fumana ericoides Fumlae Fumana laevipes Fumthy Fumana thymifolia Galset Galium setaceum Gloaly Globularia alypum Haplin Haplophyllum linifolium Hedbov Hedysarum boveanum Helcin Helianthemum cinereum Helsyr Helianthemum syriacum Helvio Helianthemum violaceum Helsto Helichrysum stoechas Hetcon Heteropogon contortus Hyphir Hyparrhenia hirta Hypsin Hyparrhenia sinaica Hyperi Hypericum ericoides Junoxy Juniperus oxycedrus Lavden Lavandula dentata Linsuf Linum suffruticosum Litfru Lithodora fruticosa Lobmar Lobularia maritima Lonetr Lonicera etrusca Matfru Matthiola fruticulosa Onomin Ononis minutissima Osylan Osyris lanceolata Palspi Pallenis spinosa Parsuf Paronychia suffruticosa

118

Networks of plant-plant co-ocurrence in semiarid steppes

Pharup Phagnalon rupestre Phasax Phagnalon saxatile Phiang Phillyrea angustifolia Phllyc Phlomis lychinitis Pinhal Pinus halepensis Pislen Pistacia lentiscus Plaalb Plantago albicans Polrup Polygala rupestris Quecoc Quercus coccifera Retsph Retama sphaerocarpa Rhaala Rhamnus alaternus Rhalyc Rhamnus lycioides Rosoff Rosmarinus officinalis Rublon Rubia longifolia Rubper Rubia peregrina Rutang Ruta angustifolia Salgen Salsola genistoides Sancha Santolina chamaecyparisus subsp. squarrosa Sataov Satureja obovata Sedalb Sedum album Sedsed Sedum sediforme Sidleu Sideritis leucantha Stadub Staehelina dubia Stipar Stipa parviflora Stiten Stipa tenacissima Teubux Teucrium buxifolium subsp. rivasii Teucap Teucrium capitatum Teucar Teucrium carolipaui Teupse Teucrium pseudochamaepitys Teuron Teucrium ronnigeri Thyarg Thymelaea argentata Thyhir Thymelaea hirsuta Thymor Thymus moroderi Thyvul Thymus vulgaris Vioarb Viola arborescens

119

CHAPTER 3 Endogenous and exogenous drivers of network structure in woody patches of semiarid steppes

Endogenous and exogenous drivers of network structure

INTRODUCTION

Network theory provides powerful tools to understand community structure by revealing interaction patterns and emergent properties of the whole community (Jordano et al., 2003; Montoya et al., 2006; Estrada, 2007; Verdú and Valiente-Banuet, 2008; Thébault and Fontaine, 2010). In Chapter 2 of this dissertation, I used network theory to describe plant communities formed by woody patches in semiarid steppes. In these communities, I differentiated dominant and accompanying species and studied co-occurrence networks using a bipartite perspective. Taking into account the presence and the intensity of co- occurrence links, I found that these communities are highly connected, poorly nested, scarcely specialized, and showed moderate modularity, with few sub- communities. Dominant species emerged as key species for the maintenance of network structure, as they played the role of network hubs in most communities. This role may be associated to their ability to create patches and facilitate the establishment of new individuals (Amat et al., 2015; Chapter 4). The presence and abundance of accompanying species varied across communities as did the indices of network structure. Variation in network indices may depend on endogenous and exogenous factors. Increasing our knowledge of the drivers of network structure can significantly improve our understanding of community functioning, and contribute to the implementation of sound management practices (Devoto et al., 2012).

Studies on the factors controlling mutualistic networks and food webs have largely focused on endogenous factors (e.g., community age, species morpho-functional traits and species richness), while exogenous factors have received less attention (Heleno et al., 2010; Devoto et al., 2012). Yet, several studies have examined the role of climatic conditions, disturbance regime, size and density of neighboring vegetation, and the structure of understory vegetation on network structure of mutualistic networks (Memmot et al., 2007; Devoto et al., 2012). Conversely, drivers of plant community networks have seldom been studied (Saiz and Alados, 2011a, 2014). It has been found that the abundance of a key plant species directly affects the number of links they establish in the network (Saiz and Alados, 2011a) and the stocking rate may influence the complexity of a plant community (Saiz and Alados, 2014).

123

Chapter 3

Network indices describe the global structure of a network in terms of cohesion, specificity of link establishment, and community segregation (Fortuna et al., 2010), and they are also associated with important network properties. For example, connectance has been related to network robustness (Dunne et al., 2002), nestedness has been associated to stability against disturbances and extinctions (Okuyama and Holland, 2008; Verdú and Valiente-Banuet, 2008; Thébault and Fontaine, 2010), and modularity strongly influences the spread of disturbances (Krause et al., 2003; Fontaine et al., 2011). Network architecture is also related to network functioning, in terms of the performance of their populations (Gómez et al., 2011). Increasing our knowledge on network properties and the drivers of those properties may help us understand community assembly rules and community variability in time and space, optimize management decisions, and improve the efficiency of conservation and restoration actions (Jordán and Scheuring, 2002; Henson et al. 2009; Heleno et al. 2010; Devoto et al., 2012).

In this chapter, I assess the exogenous and endogenous drivers of the structure of woody patch communities in semiarid Stipa tenacissima steppes. I studied 27 plant co-occurrence networks, and related them to climatic variables, and physical and biotic attributes of sites and woody patches.

MATERIALS AND METHODS

Co-occurrence networks

I studied the bipartite co-occurrence networks described in Chapter 2. As previously explained, each network was built by considering species forming woody patches present in 27 environmental units. Nodes corresponded to species, and interactions to species co-occurrences in woody patches. Interaction strength was estimated as the frequency of co-occurrence between two species weighted by their respective abundance. Dominant and accompanying species formed the two levels of the bipartite networks. I used the values of connectance, nestedness and modularity estimated in Chapter 2 as network structure indices (Fig. 1 and Table 1).

124

Endogenous and exogenous drivers of network structure

1.0

0.8

0.6

0.4

0.2

Relative value Relative

0.0

-0.2

-0.4 Connectance Nestedness Modularity

Figure 1. Network structure indices in communities of woody patches in semiarid steppes. The horizontal line within each box is the median of 27 bipartite quantitative networks of dominant and accompanying species forming patches. The bottom and top limits of each box represent the 25th and 75th percentiles, respectively. Error bars represent the 10th and 90th percentiles.

125

Table 1. Descriptors and indices of the surveyed plant-plant co-occurrence quantitative networks. All modularity values were statistically significant (p-value <0.05). Significant p-values for nestedness are shown in bold. Network Catch- Unit area Number Network Number of Connectance Modularity Nestedness (unit) ment (m2) of patches size modules 1 1 11100 20 42 0.704 0.276 3 0.242 2 1 15100 10 26 0.608 0.317 3 -0.079 3 2 15600 15 34 0.667 0.353 3 0.269 4 2 11500 15 33 0.855 0.176 2 0.362 5 3 9700 30 35 0.729 0.289 3 0.351 6 4 1900 4 16 0.667 0.301 3 0.533 7 4 17700 26 24 0.763 0.249 3 0.054 8 5 29100 30 34 0.642 0.340 3 -0.019 9 6 75900 30 43 0.637 0.326 3 0.430 10 7 9900 16 29 0.782 0.218 3 0.266 11 7 6500 14 38 0.638 0.362 2 0.504 12 8 16700 6 24 0.818 0.205 2 0.197 13 8 22800 24 41 0.667 0.324 3 0.178 14 9 25200 30 44 0.497 0.410 3 0.551 15 10 24400 13 32 0.759 0.303 2 0.186 16 10 10000 12 45 0.812 0.216 3 -0.061 17 10 6300 5 29 0.775 0.240 3 -0.269 18 11 77800 30 43 0.865 0.211 3 0.113 19 12 34100 23 42 0.676 0.272 4 0.257 20 12 7200 7 26 0.725 0.220 3 0.130 21 13 5900 5 22 0.653 0.327 3 -0.142 22 13 17500 21 33 0.722 0.235 4 -0.096 23 13 4800 4 21 0.888 0.103 3 -0.334 24 14 34200 30 43 0.674 0.290 4 0.253 25 15 8200 9 32 0.679 0.240 4 0.415 26 15 14900 14 33 0.672 0.333 3 0.079 27 15 5900 7 30 0.728 0.305 2 0.364

Endogenous and exogenous drivers of network structure

Patch characterization and site variables

We measured patch size (canopy projected area) and patch species composition (weighed by the cover of each species present) to test the influence of these variables on network structure (see further details on the methodology in Chapter 1, Patch characterization).

For each network (homogeneous environmental unit) we estimated the richness of vascular plant species, woody patch density, total plant cover, loose rock cover, rock outcrop cover and indicators of slope functionality such as mean patch width, density of patches, total patch area and the average interpatch length (LFA; Tongway and Hidley, 2004). Because of the significant correlation between these indicators, and between interpatch length and rock outcrop cover (Table 4), I only kept interpatch length for further analyses. In addition, interpatch length has been described as an accurate indicator of slope functionality (Tongway and Hindley, 2004). I used published records to estimate local mean annual precipitation and mean annual air temperature in each unit (Ninyerola et al., 2005). For methodological details see Catchment characterization and Unit characterization in Chapter 1.

Statistical analyses

I used Non-metric Multi Dimensional Scaling (NMDS) to analyze species composition in patches. NMDS has been recommended over other ordination techniques for community analysis because it does not ignore community structure that is unrelated to environmental variables and it does not assume multivariate normality (McCune and Grace, 2002). I performed NMDS to reduce dimensionality of the species composition matrix using cover of dominant and accompanying species. I used Bray-Curtis distance measure with random starting configurations for NMDS. The scores of the first two NMDS axis were averaged for patches from the same unit. Explanatory variables were tested for correlation at unit level using Pearson correlation coefficients.

Ten environmental and biotic variables were included in the models to evaluate their effect on network structure (Table 2). I built a beta-regression model for connectance and modularity because these variables are restricted to

127

Chapter 3 the interval (0-1). Beta-regression is a flexible analysis that allows us to interpret results in terms of the variable of interest; instead of transformations, it accounts for heteroskedasticity and easily accommodates asymmetries (Ferrari and Cribari- Nieto, 2004). I conducted a general linear model (GLM) for nestedness, as this variable was not constrained to any interval.

Table 2. Site and patch descriptors included in the models to explain network structure. Interpatch length is a measure of the distance between consecutive sinks to surface resource flow. Patch values correspond to the average of studied patches present in a given unit. na = not applicable. Site Range Mean ± SE Mean annual precipitation (mm) 282 - 525 353 ± 14 Mean annual air temperature (°C) 15 - 18 17 ± 0 Plant cover (%) 38 - 154 70 ± 6 Rock cover (%) 0 - 24 10 ± 1 Interpatch length (m) 0.7 - 2.3 1.6 ± 0.1 Woody patches Patch density (woody patches 100 m-2) 0.04 - 0.31 0.11 ± 0.01 Species richness (species 100 m-2) 0.06 - 0.84 0.29 ± 0.04 Patch size (m2) 4.1 - 23.9 10.5 ± 1.0 NMDS axis 1 -0.1294 - 0.1451 na NMDS axis 2 0.0921 - 0.1135 na

I simplified the models by removing the least significant covariates one by one. For each model, I estimated Akaike’s information criterion corrected for small samples (AICc), as the relation between sample size (N=27) and the number of estimated parameters (K=10) was less than 40 (Burnham and Anderson, 2002). Then, I selected the model having the highest relative Akaike weight (Table 3). It is important to note that for each response variable, there were similar models with relatively high relative Akaike weights, and thus with a high probability to be the best models; however, I only chose and discussed the model with the highest weight for each response variable, following a commonly used criterion (Burnham and Anderson, 2002). All analyses were performed using R 3.1.1 statistics software (R Development Core Team, 2011). Network indices were estimated with bipartite package. NMDS was performed with vegan package, and beta- regression with betareg package.

128

Table 3. Model selection process using Akaike’s information criterion corrected for small samples (AICc) and relative Akaike weights. For each response variable, the model with the maximum relative Akaike weight (in bold) was selected for discussion. ΔAICc is the difference between each AICc and the smallest AICc. CONNECTANCE Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10 Number of parameters 10 9 8 7 6 5 4 3 2 1 AICc -29.79 -34.94 -39.48 -43.50 -46.90 -49.69 -50.24 -49.96 -51.45 -50.05 ΔAICc 21.66 16.51 11.97 7.94 4.55 1.76 1.21 1.49 0.00 1.40 Relative Akaike weight <0.01 <0.01 <0.01 0.01 0.03 0.14 0.18 0.16 0.33 0.16 MODULARITY Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Number of parameters 10 9 8 7 6 5 4 3

AICc -48.19 -53.06 -57.07 -60.57 -63.66 -65.72 -67.40 -67.56

ΔAICc 19.37 14.50 10.49 6.99 3.90 1.84 0.16 0.00

Relative Akaike weight <0.01 <0.01 <0.01 0.01 0.06 0.16 0.37 0.40 NESTEDNESS Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Number of parameters 10 9 8 7 6 5 4

AICc 11.42 6.28 1.88 -1.81 -4.75 -7.00 -8.37

ΔAICc 19.79 14.65 10.24 6.55 3.61 1.36 0.00

Relative Akaike weight <0.01 <0.01 <0.01 0.02 0.10 0.29 0.58

Chapter 3

RESULTS

Large resprouting shrubs were the most influential species for community ordination in woody patches of S. tenacissima steppes (See Appendix in Chapter 1). NMDS stress was 0.3, and Shepard’s regression showed a non-metric fit of R2= 0.91 between the distances of each pair of patches in the final configuration and their original dissimilarities. Axis 1 of the NMDS analysis was positively correlated to Rhamnus lycioides cover, and negatively correlated to Quercus coccifera cover. Dominant species Ephedra fragilis and Pistacia lentiscus showed the strongest positive and negative correlation with NMDS axis 2, respectively. The three accompanying species that were more correlated to NMDS axis 1 were Sideritis leucantha, Teucrium capitatum and Plantago albicans (all of them positively correlated). The three accompanying species that were more correlated to NMDS axis 2 were Sedum sediforme, Helianthemum violaceum and S. tenacissima.

Mean annual temperature was strongly and negatively correlated to altitude and woody patch size (Table 4). Mean annual precipitation was positively correlated to plant cover. Interpatch length was positively correlated to rock cover and rock outcrop cover. Various LFA indices were correlated. Thus, resource sink cover and resource sink length were negatively correlated to interpatch length, and positively correlated to resource sink width. Resource sink cover was negatively correlated to altitude and woody patch size. Scores of NMDS axis 1 were positively correlated to mean annual temperature and negatively correlated to altitude and woody patch size.

130

Table 4. Pearson correlation of the measured variables. N=27 for all cases except resource sink cover and resource sink width (n=26). P-values are shown in parentheses. Significant correlations (p<0.05) are in bold.

size cover cover width length length Altitude Interpatch Interpatch Rock cover Rock Plant cover Plant NMDS axis 1 NMDS axis Precipitation Temperature Rock outcrop Rock outcrop Woody patch patch Woody sink Resource sink Resource sink Resource -0.148 ------Temperature (0.461) ------0.069 -0.913 ------Altitude (0.732) (<0.001) ------Woody -0.162 -0.694 0.762 ------patch size (0.418) (<0.001) (<0.001) ------0.292 0.121 -0.027 -0.063 ------Rock cover (0.140) (0.547) (0.892) (0.753) ------Rock outcrop -0.132 0.101 -0.051 -0.048 0.315 ------cover (0.513) (0.617) (0.799) (0.814) (0.110) ------0.680 0.043 -0.107 -0.118 -0.089 -0.005 ------Plant cover (<0.001) (0.832) (0.595) (0.558) (0.659) (0.981) ------Interpatch -0.307 -0.074 0.225 0.275 0.386 0.467 -0.321 - - - - - length (0.120) (0.714) (0.259) (0.165) (0.046) (0.014) (0.103) - - - - - Resource -0.010 0.348 -0.400 -0.422 -0.249 -0.036 0.146 -0.568 - - - - sink cover (0.960) (0.082) (0.043) (0.032) (0.221) (0.863) (0.477) (0.002) - - - - Resource 0.302 0.311 -0.420 -0.266 -0.235 -0.121 0.610 -0.530 0.740 - - - sink length (0.125) (0.114) (0.029) (0.180) (0.239) (0.547) (0.001) (0.005) (<0.001) - - - Resource 0.176 0.370 -0.385 -0.414 -0.181 -0.172 0.422 -0.387 0.818 0.695 - - sink width (0.391) (0.063) (0.052) (0.036) (0.377) (0.401) (0.032) (0.051) (<0.001) (<0.001) - - -0.370 0.646 -0.458 -0.584 0.314 0.224 -0.299 0.149 0.116 -0.166 0.040 - NMDS axis 1 (0.058) (<0.001) (0.016) (0.001) (0.110) (0.262) (0.130) (0.460) (0.574) (0.408) (0.845) - 0.017 -0.043 0.138 -0.180 0.225 0.263 0.130 0.224 -0.206 -0.041 -0.167 0.251 NMDS axis 2 (0.933) (0.832) (0.494) (0.368) (0.259) (0.186) (0.518) (0.261) (0.313) (0.840) (0.416) (0.207)

Chapter 3

According to the beta-regression model, mean annual air temperature was positively related to connectance. Average patch size was also positively related to connectance, but this effect was marginally significant (Table 5). This model explained 16% of the variability in connectance. Modularity was negatively related to mean annual air temperature and average patch size, and positively related to rock cover (Table 5). The beta-regression model explained 31% of the variability in modularity. Nestedness was negatively related to mean annual air temperature and NMDS axis 2, positively related to NMDS axis 1, and marginally and positively related to species richness (Table 5). The GLM explained 52% of the variability in nestedness.

Table 5. Parameter estimates of regression models to evaluate the effect of biotic and abiotic factors on three network indices. Z-value is given for beta-regression models and t-value is given for GLM. Model selection was made according to the highest relative Akaike weights. z value/ Estimate Standard error p-value t value Connectance (beta regression model)

(Intercept) -3.462 2.019 -1.715 0.0864 Air temperture 0.229 0.108 2.126 0.0335 Average woody patch size 0.041 0.021 1.918 0.0551 Phi (precision coefficient) 29.976 8.051 3.723 0.0002 Pseudo R2 0.155

Modularity (beta regression model)

(Intercept) 2.829 1.460 1.938 0.0526 Air temperture -0.208 0.078 -2.662 0.0078 Rock cover 0.018 0.009 2.033 0.0420 Average woody patch size -0.038 0.016 -2.408 0.0160 Phi (precision coefficient) 60.080 16.24 3.699 0.0002 Pseudo R2 0.306

Nestedness (GLM)

(Intercept) 3.215 0.835 3.850 0.0009 Air temperture -0.181 0.049 -3.719 0.0012 Species richness density 0.004 0.002 2.051 0.0524 NMDS axis 1 2.639 0.566 4.665 0.0001 NMDS axis 2 -1.764 0.752 -2.346 0.0284 Deviance explained 0.522

132

Endogenous and exogenous drivers of network structure

DISCUSSION

Both endogenous and exogenous factors affected community structure. Bigger patches favored highly connected networks. This may be explained in two ways. Firstly, as patch size increases, space and niche diversity also increase, and so does the probability that an increasing number of accompanying species will co-occur. Secondly, biggest patches are also the oldest (L. De Soto, University of Coimbra and V. Rolo, Mendel University, pers. comm.), and time increases the chances of new accompanying species getting established. Surprisingly, patch composition did not affect connectance, despite its correlation to patch size and the contrasted ability of different dominant species to accept new individuals (Amat et al., 2015). This suggests that patch size effect on connectance is not related to dominant species composition. Bigger patches also reduced modularity, which is in agreement with the positive effect of patch size on connectance, as higher connected communities hamper species segregation. In addition, the higher probability that less species will co-occur in smaller patches than in bigger ones also explains the increase in modularity when average patch size is low, as the pool of species results inevitably segregated.

High rock cover promoted modularity. Species in less rocky sites were present in most patches, but in rocky sites, species pool was segregated resulting in different community assemblages. The presence of rock fragments may affect community structure in various ways. On the one hand, plant dispersal and establishment may be hampered by soil stoniness (Poesen and Lavee, 1994; Maestre et al., 2003b) and thus rock fragments may hinder species colonization of nearby patches. On the other hand, heterogeneity in soil stoniness may affect micro-scale species assemblages (Noy-Meir, 1973; Peters et al., 2008), and increase beta diversity.

In contrast to what I found for connectance and modularity, patch composition strongly affected network nestedness. Patches dominated by Rhamnus lyciodes and Pistacia lentiscus were the most nested. Nestedness reflects the way species are related to each other and reveal specificities in these relationships. In our study, species relations were asymmetrical when R. lyciodes and P. lentiscus dominated patch composition: some accompanying species co- occurred with all dominant species, while other species only appeared under

133

Chapter 3 specific dominant species. In contrast, most accompanying species co-occurred with all dominant species (low nestedness) in units where Quercus coccifera and Ephedra fragilis dominated. Explanation for the relation between species composition and nestedness is not straightforward. The first NMDS axis (R. lycioides – Q. coccifera gradient) showed a correlation to mean annual temperature, altitude and woody patch size, which makes sense as patches dominated by R. lycioides are smaller than those dominated by Q. coccifera, and are associated to warmer and lower sites. The relationship between climatic variables and nestedness is not trivial, and has been scarcely addressed in the bibliography, particularly for plant co-occurrence networks (Selva and Fortuna, 2007). In contrast, the second NMDS axis (P. lentiscus – E. fragilis gradient) was not related to any environmental variable. Morphofunctional attributes of patches dominated by different species may partially explain differences in nestedness. For example, R. lycioides increased nestedness while Q. coccifera decreased it (first NMDS axis). The physiognomy of these two types of patches differs substantially. Patches dominated by R. lycioides are frequently small (average projected area 6.4 ± 0.3 m2, N=199), they have funnel-like and sparse canopies (Leaf Area Index=1.9 ± 0.2 m2 m-2, N=10), and they do not accumulate thick litter layers (0.9 ± 0.2 cm, N=8). In contrast, patches dominated by Q. coccifera are commonly bigger (average projected area 27.1 ± 2.5 m2, N=76), their dense canopies absorb most photosynthetic active radiation (Leaf Area Index=2.6 ± 0.2 m2 m-2, N=19), and they accumulate thick litter layers (6.1 ± 0.8 cm, N=8), which may hamper seed germination (Rotundo and Aguiar, 2005; Chapter 5 of the present dissertation). Thus, in units where R. lycioides dominated, most accompanying species thrived under their canopies, and patches with small number of species represented subsets of those. In contrast, the presence of Q. coccifera may act as a stronger filter for species establishment, either increasing randomness in species composition or selecting for a few accompanying species. In a random subset of 50 woody patches dominated by R. lyciodes there were more accompanying species (1.1 ± 0.1 species m-1) than in a similar subset of patches dominated by Q. coccifera (0.6 ± 0.1 species m-1). However, this argument is not valid when we look at the second NMDS axis: E. fragilis patches, whose physiognomy is similar to R. lycioides patches, decreased nestedness, while P. lentiscus patches, whose physiognomy is similar to Q. coccifera patches, increased it. This suggests that the effect of composition on nestedness goes beyond the

134

Endogenous and exogenous drivers of network structure dominant species composition of patches, even when dominant species were the most influential for patch composition. These results support previous observations of a two-level organization of patch communities, dominant and accompanying species, on the basis of their role in patch formation and their contribution to patch structure (Maestre and Cortina, 2005; Amat et al., 2015). The accompanying species that were correlated to NMDS axes did not share easily recognizable morphological or functional attributes that may affect their potential to affect nestedness. Unfortunately, no studies have explored the relationship between species traits and network attributes in plant co-occurrence networks. Our results suggest that the importance of patch composition in structuring nested communities relies on the whole pool of species and not on the attributes of particular species, even if certain species have a stronger influence than others in community ordination.

Species richness around woody patches positively affected community nestedness. This result suggests a positive relation between the size of species pool and community structure, and may reflect the increased probability of occurrence of accompanying species requiring specific dominant species as species pool enlarges. It is important to note that network size (number of species) was not related to any network index (see Chapter 2), suggesting that high species richness around the patches may promote nestedness by increasing the probability of patches incorporating specialists, but not by increasing network size. In addition, high species richness around woody patches suggests high microenvironment heterogeneity, which may also favor community structuring. These results are in agreement with those of Saiz and Alados (2014) where low richness, promoted by high stocking rate, decreased network complexity in plant- plant networks in a semiarid Mediterranean area.

Mean annual air temperature affected the three network indices. These relationships were somewhat surprising, as the range of air temperature in the 27 units was rather narrow. Higher temperatures favored species co-occurrence, and hampered modularity and nestedness. A direct relationship between network indices and small variations in air temperature is not straightforward. On one hand, increases in average air temperature are associated with increases in evaporative demand, which can be crucial for plant survival in semiarid areas

135

Chapter 3

(Valiente-Banuet and Ezcurra, 1991; Vallejo et al., 2012). However, in this case, we would expect water stress to be primarily reflected in average precipitation, and I found no relationship between average rainfall and network indices. On the other hand, small changes in average air temperature may indeed reflect significant changes in extreme temperatures. These may affect network structure, albeit in non-linear ways. Average air temperature was strongly correlated to altitude, and probably to the risk of freezing. These extreme events may act as filters, decreasing the probability of co-occurrence, favoring specialization and promoting the segregation of different dominant-accompanying species assemblages. Our results suggest that the structure of plant communities is strongly related to air temperature, and thus it may be sensitive to climate change. Aridity will increase in semiarid areas over the next decades (IPCC, 2013). Considering current species pool, warming may promote connectance, while reducing network nestedness and modularity. These changes may increase the vulnerability of woody patches to further disturbances, and deeply modify the composition and function of semiarid steppes (Maestre et al., 2009b).

CONCLUSION

Plant communities showed higher levels of connectance and lower levels of nestedness and modularity than mutualistic and trophic networks, which is congruent with the complex and weakly structured nature of most plant-plant interactions. Furthermore, I found that both endogenous and exogenous factors were major drivers of network structure in woody patches of semiarid steppes. Patch size, rock cover, average air temperature and the composition of patch- forming species were related to various aspects of community organization. Information about drivers on network structure of plant communities can be crucial to assess community ability to withstand disturbances and to prescribe management practices to foster biodiversity and the provision of ecosystem services.

136

CHAPTER 4 Community attributes determine facilitation potential in a semiarid steppe

Community attributes determine facilitation potential

INTRODUCTION Biotic interactions determine community assembly and ecosystem functioning. Competition has been traditionally considered as the main driver in structuring plant communities (Grime, 1974). However, over the last decades, an increasing number of studies have emphasized the importance of facilitation as a major ecological interaction (Bertness and Callaway, 1994; Callaway, 1995, 2007; Brooker et al., 2008). Studies on plant-plant relations have traditionally focused on pair-wise interactions, paying scarce attention to other co-existing species (Brooker et al., 2008). This topic was reviewed by Jones and Callaway (2007), where they discussed the context-dependency of plant interactions and emphasized the role of third species. When many benefactor and beneficiary species co-occur, a complex network of interactions arises, leading to indirect effects, such as indirect facilitation (Callaway, 2007; Brooker et al., 2008; Gross, 2008). Thus, the net outcome of multi-species interactions may not necessarily be the additive effect of pair-wise relationships (Weigelt et al., 2007; Zhang et al., 2011). In recent years, there has been an increasing number of studies employing the community approach to study facilitation (Verdú and Valiente- Banuet, 2008; Cavieres and Badano, 2009; Gilbert et al., 2009; Soliveres et al., 2011a; Granda et al., 2012). However, most of those studies are either extrapolations of individual responses to a community scale, or studies focused on the effect of nurse species on biodiversity, but not on the opposite, i.e., on the capacity of species assemblages to act as nurses (but see Castillo et al., 2010 for an analysis of the effect of phylogenetic distance of the nurse community on seedling establishment). Drylands are not the exception and pairwise approaches dominate the study of positive interactions (Pugnaire et al., 2011). However, species in drylands do not occur in isolation. Harsh conditions, especially water scarcity, promote spatially aggregated vegetation patterns (Aguiar and Sala, 1999). In drylands, vegetation is frequently arranged in multi-species patches where direct and indirect facilitation promotes species coexistence (Flores and Jurado, 2003; Callaway, 2007). Attributes defining the physical and biotic structure of these patches, such as patch size, litter accumulation, canopy structure, and species number and identity may affect seedling establishment. 139

Chapter 4

Communities of putative benefactor species in vegetation patches may promote species coexistence by favoring seedling performance. As not all species co- occurring in a patch act as nurses, patches with high species richness are more likely to contain benefactor species than those with fewer species (i.e., sampling effect), and thus species richness may enhance the recruitment of new individuals. Conversely, as species richness increases, the probability that species with contrasted functional traits co-occur may also increase. Due to niche saturation, patch biotic and physical dimensions may be limited, and the establishment of new individuals may be hindered by patch size and age (MacArthur and Levins, 1967; Case, 1991). In this case, increased species richness would reduce patch potential to accept new individuals. The positive and negative relationship between species richness and community capacity to accept further species, together with the influence of spatial scale, has been discussed around the theory of the invasion paradox (Levine, 2000; Stohlgren et al., 2003; Friedley et al., 2007). According to this theory, community diversity and invasibility should be negatively related at small spatial scales. However, facilitation often promotes invasive species richness (Von Holle, 2005; Friedley et al., 2007; Vellend, 2008; Altieri et al., 2010), and accordingly, we would expect increased establishment rates in species-rich patches, particularly under the stressful conditions of semiarid areas (Von Holle, 2005). Species identity may also be crucial for the establishment of new individuals, because of the similarity of ecological niches, species competitive ability and species capacity to enhance microclimate conditions for the newcomers. Phylogenetic distance between species may directly affect the net outcome of the interaction, as closely related species are likely to share important ecological traits and therefore compete among each other (Webb et al., 2002; Valiente-Banuet and Verdú, 2007). However, in multi-specific assemblages, phylogenetic composition of the entire community, rather than the phylogeny of a dominant species, may control seedling establishment. The mechanisms of competition in multi-species communities in dry grasslands showed that non- additive effects of pairwise interactions drive the net outcome of the interaction (Weigelt et al., 2007). However, to our knowledge, no study has evaluated the facilitative effect of whole communities in seedling establishment in drylands. Neither the attributes of communities that are involved in seedling recruitment

140

Community attributes determine facilitation potential have been explored. Still, species identity, patch composition and the physical structure of woody vegetation patches may be crucial for seedling establishment. Woody patches may affect seedling establishment in various ways, including microclimate regulation, changes in water and nutrient availability, and the presence of symbiotic fungi, to mention a few (Vetaas, 1992; Nara and Hogetsu, 2004; Smith and Reed, 2008; Cable et al., 2009; Soliveres et al., 2011a; Anthelme et al., 2012). Their combined effect cannot be easily predicted, as interactions are common and complex. For example, nutrient and organic matter accumulation, and the formation of a thick litter layer depend on plant size and species identity (Vivanco and Austin, 2006). However, whilst higher soil fertility may enhance seedling establishment, litter frequently hinders seed germination and rooting (Rotundo and Aguiar, 2005; Chapter 5). The outcome of the interaction between established patches and seedlings may also depend on the particular location within the patch where the new individual thrives. For example, the relative importance of aboveground interactions (e.g., competition for light, excess radiation, herbivory) vs. belowground interactions (allelopathy, competition for nutrients and water, mycorrhizae) may substantially change if seedlings germinate underneath the patches or on their periphery. I studied the facilitative potential of whole communities of woody vegetation patches on the establishment of a keystone species in Stipa tenacissima L. steppes. These semiarid steppes show a combination of bare soil, S. tenacissima tussocks and woody patches, formed by large resprouting shrubs (dominant species), accompanied by small shrubs and perennial grasses (accompanying species). These woody patches improve ecosystem functionality, and favor the presence of frugivorous birds and vascular plants (López and Moro, 1997; Maestre and Cortina, 2004b; Maestre et al., 2009). Positive interactions between S. tenacissima and large woody species have been widely described (Maestre et al., 2001, 2003a; Soliveres et al., 2011a), and dominant species of woody patches act as benefactors in other areas (Castro et al., 2004), but little is known about the facilitative potential of patch forming species and patch communities in these semiarid steppes. As woody patches in S. tenacissima steppes form spatially delimited plant communities, they are suitable environments to test the facilitative effects of communities and explore the role of different community attributes in that interaction. The aims of our study are (1)

141

Chapter 4 to evaluate how patches modify their immediate environment, (2) to assess the net effect of woody patches on the performance of new individuals, (3) to quantify the relative importance of physical and biotic community (patch) attributes as drivers of seedling recruitment, and (4) to explore the underlying mechanisms of such interactions. To achieve this, I characterized biotic and physical attributes of multi-specific woody vegetation patches in S. tenacissima steppes, and evaluated the performance of shrub seedlings planted underneath them compared to seedlings planted in open areas. I expected substantial differences in aboveground and belowground interactions in different parts of the patch. To take this into account, I evaluated seedling performance in three different locations within the patch.

MATERIALS AND METHODS Study site This study was established in a semiarid area in southeastern Spain. Mean annual temperature is 18ºC and mean annual precipitation ranges between 286 and 330 mm (Ninyerola et al., 2005). Soil is Lithic Calciorthid developed from marl and limestone (Soil Survey Staff, 1994). The area is covered by Stipa tenacissima steppes with sparse patches of woody vegetation. I selected 8 to 13 patches in five independent sites (53 patches in total). Patches were selected to provide a balanced representation of number and identity of dominant species. Seedling plantation and monitoring In November 2008 we planted 1-year-old seedlings of P. lentiscus in four locations per patch (212 seedlings in total): (1) underneath the patch, where aboveground and belowground interactions take place, (2) south of the patch, outside the canopy projection, where only belowground interactions are expected, (3) north of the patch, outside the canopy projection area where belowground interactions occur and seedlings are partially shaded; and (4) in open areas, at least 2 m apart from any woody patch (Fig. 1). Seedlings in this location (open areas) were independent from each other and from any woody patch, but we sampled them as “belonging” to a patch only for logistic purposes. In this way, we took spatial heterogeneity in biotic and abiotic conditions

142

Community attributes determine facilitation potential occurring within the patches into account. Planting holes were dug manually with a 5 x 5 x 20 cm soil auger to minimize soil disturbance. We recorded seedling survival and stem height 2 months after planting, before and after summer 2009 and 2010. During the first survey, we noticed that almost all seedlings from one of the study sites were dug up or eaten by rabbits, thus we replaced those plants, and protected all seedlings with plastic mesh cages. These cages have a negligible effect on seedling microclimate. I considered these records as baseline survival (191 alive seedlings: 51 seedlings underneath, 48 seedlings north, 48 seedlings south, 44 seedlings in open areas), once transplant shock was excluded and predated seedlings were replaced. We assessed water use efficiency (WUE), integrated transpiration rate and N source using 13C, 18O and 15N enrichment, respectively, from leaf samples collected during first pre-summer survey (June 2009). The number of (alive) seedlings for this survey at each location of plantation was: 38 seedlings underneath, 41 seedlings north, 37 seedlings south, 35 seedlings in open areas. Leaf samples were ground in a MM200 Retsch ball mill and isotope concentration together with foliar C and N concentration were determined at the University of California-Davis Stable Isotope Facility (CA, USA) using a continuous flow isotope ratio mass spectrometer (Europa 20-20 Scientific, Sercon Ltd., Cheshire, UK) in the dual-isotope mode, interfaced with a CN elemental analyzer and Hekatech HT Oxygen Analyzer. Results for isotopes enrichment are expressed as δ13C, δ15N, δ18O, relative to the Pee Dee Belemnite

(PDB), atmospheric N2 and Vienna-Standard Mean Ocean Water (V-SMOW), respectively.

143

Chapter 4

Figure 1. Schematic representation of the experimental unit. Seedlings (solid black) were planted in three locations per woody patch (octagon): underneath them and on their northern and southern edges, all of them being putatively affected by underground interactions with the patch (in grey). Northern seedlings were under the shade for part of the day (hatched area). One additional seedling was planted in an adjacent open area (Open), separated at least 2 m from any woody patch.

Patch characterization Patch biotic structure was described by characterizing community composition and measuring species richness of dominant and accompanying species under the canopy of each patch, and determining the phylogenetic distance between P. lentiscus seedlings and the patch community. Dominant species composition was estimated in terms of species cover within the patch, by measuring two orthogonal diameters of the canopy of each species, and estimating canopy projected area as an ellipse. The cover of each accompanying species was estimated by using transects of consecutive 50 cm-quadrats underneath each patch along the canopy projected area, in upslope-downslope direction (see Methods in Chapter 1). An abundance weighted measure (M) of phylogenetic distance between dominant species present in a patch and P. lentiscus was estimated according to Violle et al. (2007). This measure estimates the mean trait value of the community weighted by the relative abundance of

144

Community attributes determine facilitation potential each species present in the community: M = . Where pi is the abundance of species i and Ti is the phylogenetic distance between P. lentiscus and each patch-forming species i obtained from the TimeTree database (Hedges et al., 2006). I estimated patch size (area and maximum height), patch canopy density, litter accumulation and total cover of accompanying species under the patches to describe patch physical structure. Patch area was estimated as described for the estimation of the cover of dominant species. As a measure of canopy density, I estimated the leaf area index (LAI) with a plant canopy analyzer LAI2000 (Li-Cor Inc., Nebraska, USA) in the same four locations where seedlings were planted. We measured litter depth in 6-10 points underneath patch canopy, depending on patch size. We also evaluated soil fertility in the same four locations per patch where seedlings were planted by analyzing oxidable C and total N. We collected 5 x 5 x 5 cm soil samples in the four locations, ground them in a MM200 Retsch ball mill, and analyzed them for oxidable C (modified Moebius method, P. Rovira, Centre Tecnològic Forestal de Catalunya, pers. comm.) and total N content (semi-micro- Kjeldahl distillation in a Tecator Kjeltec Auto 1030 analyzer, Hogana, Sweden). We recorded soil moisture at 0-10 cm depth underneath 7 patches in a size gradient from 4 to 30 m2 of canopy projection and in nearby areas beyond the influence of the patches. Soil moisture was recorded every hour by soil moisture sensors (10HS) and Decagon ECH2O data-logger (ECHO-10, Decagon Devices, Inc., Pullman, Washington, USA) from July 2009 to June 2011.

Statistical analyses I performed Non-metric Multi-Dimensional Scaling (NMDS) to reduce the dimensionality of the species composition matrix. NMDS has been recommended over other ordination techniques for community analysis because it does not ignore community structure that is unrelated to environmental variables and it does not assume multivariate normality (McCune and Grace, 2002). I used Bray- Curtis distance measure with random starting configurations. I performed NMDS analyses for dominant and accompanying species separately and used the first two axes of each for the next analyses.

145

Chapter 4

I evaluated the effect of location on soil organic C, total N and C:N ratio in samples from the four locations by using General Linear Mixed Models (GLMM) with site as random effect and location as fixed effect. To avoid data dependence, I selected a random dataset that included only one patch-dependent location per patch (underneath, north or south), plus the 53 independent samples from open locations. I used the same procedure to analyze the effect of location on seedling performance (survival, relative growth rate, and stable isotope enrichment, C and N concentration in leaves of alive seedlings). For all dependent variables except for survival, samples in the four locations were selected to optimize the number of experimental units, while ensuring sample independence. Short-term seedling survival and relative growth rate (RGR) were low, thus, I only kept seedling survival as a measure of seedling performance to evaluate its relationship with the physical and biotic structure of woody patches. I used GLMM with site and patch nested within site as random effects with a binomial error distribution. Patch size, LAI, patch composition (NMDS axes), cover and richness of dominant and accompanying species, soil organic C, soil total N, phylogenetic distance of the community of dominant species, and litter depth were included as covariates in the initial model. I made a simplification of the maximal model by removing the least significant covariates one by one. For each model, I estimated Akaike’s information criterion corrected for small samples

(AICc) as the relation between sample size (N=159) and the number of estimated parameters (K=13) was less than 40 (Burnham and Anderson, 2002). Then, I selected the GLMM having the lowest AICc to determine the best model (model 9, Appendix). Only seedlings affected by patch properties (i.e., those planted under or close to the patch), were included in this analysis. Location could not be included in this analysis as a result of the lack of replication at this level. Thus, I fit one GLMM for each location (N=53) with site as random factor, to examine the drivers of seedling establishment in further detail. The model that minimized AICc was selected for each location. All analyses were performed in R 3.1.0 statistics software, package vegan for NMDS analyses and package lme4 for GLMMs (R Development Core Team, 2014).

146

Community attributes determine facilitation potential

RESULTS Patch attributes Patch area ranged from 3 to 44 m2 and 72% of the patches were smaller than 20 m2. Soil organic C and total N were significantly higher underneath the patches than in other locations (Fig. 2, Table 1). In contrast, I found no effect of location on the C:N ratio.

0.5 16 6 Underneath Open North 14 South 5 0.4 12

4 10 0.3

3 8

0.2 ratio C:N 6 2

Soil C concentration (%) C concentration Soil (%) N concentration Soil 4 0.1 1 2

0 0.0 0 Organic C Total N C:N

Figure 2. Organic C, total N and C:N ratio in soil samples taken at different locations within woody patches (underneath, north and south) and in open areas in Stipa tenacissima steppes. The number of analyzed patches was 53. Mean and standard error bars are shown. Soil sampled underneath the patches differed from soil sampled on other locations for C and N, but not for the C:N ratio (p<0.05).

147

Chapter 4

Table 1. Parameter estimates of the GLMM to estimate the effect of location within woody patches on soil organic C, soil total N and the C:N ratio. LocationU, locationN and locationS correspond to underneath, north and south edges of the patch, respectively. Fixed effects Estimate Std. error d.f. t-value p-value C (Intercept) 4.942 0.552 98 8.948 <0.001 locationU -2.154 0.401 98 -5.375 <0.001 locationN -2.223 0.499 98 -4.454 <0.001 locationS -1.667 0.496 98 -3.362 0.001 N (Intercept) 0.347 0.043 98 8.108 <0.001 locationU -0.126 0.022 98 -5.679 <0.001 locationN -0.129 0.028 98 -4.658 <0.001 locationS -0.084 0.028 98 -3.057 0.003 C:N (Intercept) 14.046 1.374 98 10.220 <0.001 locationU -0.780 1.420 98 -0.550 0.584 locationN -1.338 1.757 98 -0.761 0.448 locationS -1.537 1.758 98 -0.874 0.384 Random effects: site Std. deviation C (Intercept) 0.961 Residual 1.453 N (Intercept) 0.085 Residual 0.081 C:N (Intercept) 1.377 Residual 5.167

Soil water content was consistently higher in open areas than inside the patches (Fig. 3). Average litter depth ranged from 0 to 6.4 cm and LAI values from 0 to 3.8. Phylogenetic distance of the dominant species ranged from zero (where the only dominant species was Pistacia lentiscus) to 331 million years (where Juniperus oxycedrus and Ephedra fragilis were the dominant species).

148

Community attributes determine facilitation potential

0,35 Underneath

North )

3 South - Open

m 0,30 3 0,25

0,20

0,15

0,10

0,05

Soil volumetric water content (m content watervolumetric Soil 0,00

Figure 3. Temporal changes in soil water content in different locations with respect to woody patches. Measurements were taken daily at 12:00 pm at 0-10 cm depth. Lines correspond to means of 7 patches.

Patches hosted between 4 and 20 species. Each patch was formed by up to 5 dominant species, and 1 to 18 accompanying species. Of the six patch- forming species, Osyris lanceolata was not dominant in any patch. Patch community composition was strongly influenced by the dominant species (Table 2). NMDS axis 1 was strongly and positively correlated with P. lentiscus cover, and negatively correlated with J. oxycedrus cover. NMDS axis 2 was strongly and positively correlated with Rhamnus lycioides cover and negatively correlated with Quercus coccifera cover. Fifteen accompanying species showed a significant correlation with NMDS axes (p<0.05; Table 2), but correlations were strong for only five of them (Pearson correlation coefficients r>0.5). Thus, Avenula murcica Holub, Helichrysum stoechas (L.) Moench and Globularia alypum L. correlated with NMDS axis 1, and Brachypodium retusum (Pers.) P. Beauv. and Stipa tenacissima correlated with NMDS axis 2.

149

Chapter 4

Table 2. Pearson correlation coefficients between species cover and NMDS axes 1 and 2. NMDS analyses for dominant and accompanying species were performed separately. Significant p-values (p<0.05) are in bold. NMDS NMDS p-value p-value axis 1 axis 2 NMDS1 NMDS2 DOMINANT SPECIES Quercus coccifera -0.441 -0.773 0.001 0.000 Rhamnus lycioides 0.094 0.629 0.532 0.000 Pistacia lentiscus 0.757 -0.198 0.000 0.158 Juniperus oxycedrus -0.615 0.384 0.001 0.005 Ephedra fragilis 0.376 0.389 0.008 0.004 Osyris lanceolata 0.261 0.212 0.061 0.145 ACCOMPANYING SPECIES

Rosmarinus officinalis -0.100 -0.101 0.478 0.472 Avenula murcica -0.582 0.271 0.000 0.050 Thymus vulgaris 0.426 0.325 0.001 0.017 Brachypodium retusum 0.194 0.534 0.163 0.000 Teucrium capitatum -0.061 -0.128 0.666 0.362 Fumana ericoides -0.250 0.214 0.071 0.123 Asparragus horridus 0.171 0.078 0.221 0.580 Dorycnium pentaphyllum 0.034 0.090 0.808 0.523 Helichrysum stoechas 0.543 0.013 0.000 0.928 Sedum sediforme 0.408 0.171 0.002 0.222 Helianthemum violaceum -0.227 0.073 0.102 0.606 Sideritis leucantha -0.097 -0.268 0.490 0.052 Phagnalon saxatile 0.133 -0.437 0.342 0.001 Globularia alypum -0.538 -0.164 0.000 0.239 Anthyllis cytisoides 0.341 -0.144 0.013 0.303 Stipa tenacissima -0.081 -0.604 0.564 0.000 Teucrium carolipaui -0.268 0.204 0.052 0.143 Teucrium pseudochamaepitys 0.166 0.156 0.236 0.265 Coronilla minima -0.036 0.285 0.800 0.038 Rubia peregrina 0.203 -0.017 0.144 0.905 Bupleurum fruticescens 0.296 -0.119 0.031 0.397 Carex humilis -0.109 0.103 0.435 0.462 Erica multiflora -0.113 0.245 0.419 0.077 Polygala rupestris 0.337 0.229 0.014 0.099 Cistus clusii -0.065 -0.066 0.646 0.641 Asparagus acutifolius 0.248 -0.073 0.073 0.605 Lapiedra martinezii -0.002 -0.281 0.988 0.042 Satureja obovata 0.422 0.230 0.002 0.097 Fumana laevipes -0.305 0.279 0.027 0.043

150

Community attributes determine facilitation potential

Seedling performance Patches had a positive effect on seedling survival as only open areas significantly decreased its probability (Fig. 3, Table 3). This effect was stronger after the first summer (10 months after planting). Two years after planting, seedling survival underneath the patches was 55% compared to 18% in open areas. Relative growth rate (RGR) was estimated in 151, 72 and 65 seedlings in total, corresponding to surveys at 5, 8 and 22 months after plantation. RGR was highly variable and showed no significant location effect in any of the surveys (Fig. 4, Table 3).

151

Chapter 4

100 Underneath North South 80 Open

Survival (%) Survival 40

0 0 5 10 15 20 25

) 0.04

-1

0.02

0.00

-0.02

-0.04

Relative Growth Rate (stem height, cm month cm height, (stem Rate Growth Relative 0 5 10 15 20 25 Time after planting (months) Figure 4. Survival and growth of Pistacia lentiscus seedlings planted on different locations within woody patches (underneath, north and south) and in open areas in Stipa tenacissima steppes. Survival baseline was considered 2.5 months after planting, once transplant shock was excluded and seedlings predated immediately after planting were replaced. RGR was only measured in those seedlings that were alive 22 months after planting. Means ± 1 SE are shown for seedling growth of living individuals.

152

Community attributes determine facilitation potential

Table 3. Parameter estimates of the GLMM to estimate the effect of location on seedling survival and relative growth rate (RGR) 22 months after planting. Z-values and t-values are given for survival and RGR, respectively. LocationU, locationN and locationS correspond to underneath, north and south edges of the patch, respectively. Fixed effects Estimate Std. error d.f. z/t-value p-value Survival (Intercept) 0.397 0.751 - 0.528 0.598 locationU -2.472 0.711 - -3.477 0.001 locationN -1.024 0.824 - -1.243 0.214 locationS -1.144 0.791 - -1.447 0.148 RGR (Intercept) 0.012 0.009 34 1.352 0.185 locationU -0.007 0.013 34 -0.554 0.584 locationN -0.015 0.012 34 -1.292 0.205 locationS -0.019 0.012 34 -1.531 0.135 Random effects: site Std. deviation Survival (Intercept) 1.088 Residual 0.931 RGR (Intercept) 0.005 Residual 0.028

Seedlings planted underneath the patches showed lower integrated WUE and higher foliar 15N enrichment than seedlings planted in open locations (Fig. 5, Table 4). Integrated transpiration rates, as estimated by 18O analysis, were similar in all locations. There were no differences between seedlings planted in the northern and southern parts of the patches compared to seedlings planted underneath patches for the three stable isotopes analyzed. Seedlings planted underneath the patches showed higher foliar N concentration than seedlings planted on their northern edge, whereas there were no significant differences in N concentration between the former and seedlings planted in other locations (Fig. 5, Table 4). Foliar C concentration was similar in all locations.

153

Chapter 4

Underneath Open 50 2.0 North

South Foliar C concentration (%) concentration C Foliar 40

1.5 30 * 1.0 20

Foliar N concentration(%) 0.5 10

0.0 0 Nitrogen Carbon

30 1.5

15 N isotope enrichment enrichment isotope N ( value) 20 1.0

10 0.5

0 0.0

-10 -0.5 value)

O isotope enrichment ( isotope enrichment O

18 -20 -1.0 *

C and 13 -30 * -1.5

d13C13C d18O18O d15N15 N

Figure 5. Foliar C and N concentration and isotope enrichment of Pistacia lentiscus seedlings planted on different locations within woody patches (underneath, north and south) and in open areas in Stipa tenacissima steppes. Only seedlings that were alive 7 months after planting were analyzed. Asterisks indicate significant differences with seedlings planted underneath the patches for a given element or isotope. Note the different scale for different axes.

154

Community attributes determine facilitation potential

Table 4. Parameter estimates of the GLMM to estimate the effect of location on isotope enrichment and foliar concentration of C and N. LocationU, locationN and locationS correspond to underneath, north and south edges of the patch, respectively. Fixed effects Estimate Std. error t-value p-value δ13C (Intercept) -28.733 0.497 -57.801 <0.001 locationU 1.256 0.435 2.888 0.005 locationN 0.452 0.508 0.888 0.377 locationS 0.696 0.505 1.378 0.172 δ15N (Intercept) 0.503 0.555 0.906 0.368 locationU -1.052 0.450 -2.338 0.022 locationN -0.156 0.526 -0.296 0.768 locationS -0.552 0.523 -1.056 0.294 δ18O (Intercept) 30.105 0.662 45.460 <0.001 locationU 0.914 0.582 1.569 0.121 locationN -0.244 0.681 -0.358 0.722 locationS 0.316 0.676 0.467 0.642 C (Intercept) 47.693 1.457 32.740 <0.001 locationU 0.604 1.740 0.347 0.730 locationN -2.702 1.978 -1.366 0.176 locationS 1.002 2.009 0.499 0.619 N (Intercept) 1.404 0.133 10.530 <0.001 locationU -0.158 0.125 -1.271 0.208 locationN -0.364 0.142 -2.573 0.012 locationS -0.184 0.144 -1.277 0.206 Random effects: site Std. deviation δ13C (Intercept) 0.775 Residual 1.442 δ15N (Intercept) 0.927 Residual 1.491 δ18O (Intercept) 1.027 Residual 1.931 C (Intercept) 0.455 Residual 5.571 N (Intercept) 0.188 Residual 0.391

155

Chapter 4

Drivers of seedling performance Seedling survival was predicted by both physical and biotic patch attributes. Composition of the community of accompanying species affected seedling survival (Fig. 6, Table 5). The increase in NMDS axis 2 (and thus, the increase in Brachypodium retusum cover and decrease in Stipa tenacissima cover) decreased the probability of survival. In addition, phylogenetic distance to the dominant species, cover of accompanying species and litter depth increased the probability of seedling survival, (Fig. 6, Table 5). Dominant species composition was included in the final model, but its effect was not statistically significant (Table 5). For each patch nested within site, the estimates of the model increased and decreased by a random value with expected mean of zero and variance of 0.2928.

Table 5. Parameter estimates for GLMM to evaluate the effect of community attributes on seedling survival (N=159). Sites (5) and patches (53) were included as random effects. The additional variance in estimates ascribed to sites was 0.1292. Considering patches nested within sites the variance was 0.2928. Fixed effects Estimate Std. Error z value p-value (Intercept) -2.629 0.465 -5.654 <0.001 Cover accompanying species 0.016 0.004 4.480 <0.001 Litter depth 0.980 0.374 2.619 0.009 NMDS axis 1 dominant species 0.633 0.405 1.564 0.118 NMDS axis 2 accompanying species -6.602 2.164 -3.051 0.002 Phylogenetic distance of dominant species 0.006 0.002 3.010 0.003

156

Community attributes determine facilitation potential

Figure 6. Survival probability of Pistacia lentiscus seedlings (N=159) against accompanying species composition, phylogenetic distance to dominant species, litter depth and cover of accompanying species in woody patches in Stipa tenacissima steppes. Histograms show number of seedlings (right axis) of alive (superior histograms) or dead (inferior histograms) individuals. Data range for both alive and dead seedlings is represented by the dotted lines, and the box plots represent data between the first and third quartiles, and the median (black line within the box). Solid line is the logistic regression fit of survival probability against each variable of X-axis.

Considering only seedlings planted underneath the patches, the composition of accompanying species and soil organic C successfully predicted seedling survival (Table 6). As for the whole dataset, as NMDS axis 2 of

157

Chapter 4 accompanying species increased, the probability of seedling survival under the patches decreased. An increase in soil organic C increased the probability of survival. Comparatively, survival probability in seedlings planted in the northern edge of the patches depended on a larger number of explanatory variables: number of accompanying species, both NMDS axes of dominant species, NMDS axis 2 of accompanying species and phylogenetic distance of the dominant species (Table 6). This second NMDS axis of accompanying species decreased the probability of survival. In addition, an increase in the number of accompanying species decreased seedling survival. The identity of the dominant species also affected seedling performance in this location: NMDS axis 1 increased and NMDS axis 2 decreased the probability of survival. I found a direct relationship between phylogenetic distance of the dominant species and the probability of survival in seedlings planted in the northern edge of patches. Finally, the probability of survival of seedlings planted in the south of the patches was not significantly predicted by any patch attribute, but the number of accompanying species and the NMDS axis 2 of dominant species were included in the best model (Table 6).

Table 6. Parameter estimates of GLMM on the effect of community attributes on seedling survival planted in three different locations within woody patches (N=53). Only models with the lowest AICc are shown. Fixed effects Estimate Std. Error z value p-value Underneath (Intercept) -1.493 0.869 -1.718 0.086 NMDS axis 2 accompanying species -7.741 3.137 -2.468 0.014 Soil organic C 0.302 0.147 2.060 0.039 North (Intercept) -1.270 1.694 -0.749 0.454 Number accompanying species -0.628 0.309 -2.036 0.042 NMDS axis 1 dominant species 2.130 0.985 2.163 0.031 NMDS axis 2 dominant species -2.092 1.208 -1.731 0.083 NMDS axis 2 accompanying species -16.150 6.268 -2.577 0.010 Phylogenetic distance of dominant species 0.025 0.009 2.796 0.005 South (Intercept) -2.447 1.095 -2.234 0.026 Number accompanying species 0.231 0.149 1.551 0.121 NMDS axis 2 dominant species 0.829 0.506 1.639 0.101 Random effects: site Std. deviation Underneath (Intercept) 0.274 North (Intercept) 0.328 South (Intercept) 0.000

158

Community attributes determine facilitation potential

DISCUSSION Patch effect on seedling performance I found evidence that patches positively affected seedling survival. This effect was more intense during the first summer, when mortality was higher, and it was larger in seedlings planted underneath the patches than in seedlings planted on their periphery. The restricted spatial extent of facilitation illustrates the spatial-temporal dynamics of the interaction, as slight modifications in environmental conditions may shift the relationship from positive to neutral or negative. This is in agreement with the temporal variability of positive interactions observed in S. tenacissima steppes (Maestre and Cortina, 2004a). On the other hand, the limited span of the interaction has strong implications for the spatial distribution of plant cover and the functioning of S. tenacissima steppes, and the restoration of degraded steppes (Puigdefábregas, 2005; Cortina et al., 2011). Seedlings planted under the patches were less efficient in using water than those planted outside them, suggesting that water stress was lower in the former. However, I found that the decrease in water use efficiency (WUE) resulted from a decrease in photosynthetic rate rather than an increase in transpiration rate, as 18O enrichment was not affected by location. Thus, the reduction in evaporative demand under the patches was probably offset by the decrease in soil water availability and, as a result, competition for light was indeed the main responsible for the decrease in WUE. The combination of shade and water stress may have deleterious effects on seedling survival (Valladares and Pearcy, 2002). However, this was not the case in Pistacia lentiscus seedlings, probably because other factors, including the reduction in irradiance stress, compensated for low water availability. Higher 15N enrichment in seedlings planted near patches may indicate rapid N cycling and relatively high N losses, as major pathways of N loss such as nitrification, denitrification and ammonia volatilization promote soil 15N enrichment (Shearer et al., 1974; Peñuelas et al., 1999). In agreement with this, I found higher foliar N concentration in seedlings planted underneath the patches than in seedlings planted north of them, suggesting that N availability was higher in the former. In addition, soil total N and organic C were higher underneath the

159

Chapter 4 patches than in any other location. Thus, the positive effect of patches on seedling performance was probably the result of amelioration in microsite conditions resulting from both the reduction in irradiance and the improvement in soil fertility. This is in agreement with studies in semiarid areas where microsite has been identified as one of the major drivers of facilitation (Barberá et al., 2006; Soliveres et al., 2011a). On the other hand, the reduction in photosynthetic rate may partly explain the lack of a significant effect of patches on seedling growth. Seedling protection against predation was not evident, as two months after planting rabbits affected some seedlings regardless of their location. This observation contrasts with Soliveres et al. (2011a), as these authors found that tussocks provided protection against rabbit predation in a Stipa tenacissima steppe in central Spain. Soliveres et al. (2012) also showed that protection against predation was more likely below 450 mm annual average precipitation.

Community attributes driving seedling survival Our study showed that the physical and biotic structure of woody patch communities and, particularly, their composition, explained their potential for facilitation. It also showed that the importance of community attributes as drivers of seedling performance depended on location within the patch. The general model, that included seedlings from all locations, identified the composition and cover of accompanying species, litter depth and phylogenetic distance of the dominant species as major drivers of facilitation. The composition of dominant patch species completed the model, but its effect was not statistically significant. The relatively high variance ascribed to the random effects of patches nested in sites emphasizes the importance of site conditions, above microsite properties, as determinant of seedling performance (Maestre et al., 2003b; Cortina et al., 2011). The second axis of the NMDS for accompanying species had a negative effect on seedling survival, and thus on patch potential for facilitation. This axis was correlated with the cover of several species. For example, NMDS axis 2 was correlated with the cover of Brachypodium retusum, a species that may interfere with P. lentiscus seedlings by reducing irradiance and competing for soil resources (Maestre et al., 2004). The second axis of the NMDS analysis of accompanying species was also related to S. tenacissima. The cover of this species was related

160

Community attributes determine facilitation potential with an increase in seedling survival, which may be associated to the facilitative effect of S. tenacissima on P. lentiscus that has been previously described in the literature (Maestre et al., 2003a). The cover of accompanying species also increased survival probability. The positive effect of plant cover on seedling performance has been associated to protection from excessive radiation (Rey- Benayas et al., 2002; Sánchez-Gómez et al., 2006; Soliveres et al., 2011b), which may disappear under mesic conditions (George and Bazzaz, 1999; Beckage and Clark, 2003). The positive effect of litter on the probability of seedling survival may result from its capacity to act as mulch. Litter increases soil moisture by reducing runoff, increasing water infiltration and reducing evaporative losses through shading and lowering soil temperatures (Facelli and Pickett, 1991; Guevara- Escobar et al., 2007; Goldin and Brookhouse, 2014). Phylogenetic distance of the dominant species also increased the probability of survival, which is in agreement with studies suggesting that closely related species are likely to share important ecological traits, and thus compete more severely than distant species (Webb et al., 2002). In addition, phylogenetically distant species frequently differ in their ecological traits and the environmental conditions they can cope with, and thus facilitative interactions among them are more likely to occur (Valiente-Banuet et al. 2006, Valiente-Banuet and Verdú 2007, Castillo et al. 2010). The analysis of community attributes affecting patch potential for facilitation at each of the three patch locations was consistent with the general analysis, and revealed different sensitivity to community attributes. While survival of seedlings planted underneath or in the northern edge of the patches was driven by one or several biotic community attributes, respectively, no community attribute was related to the survival of seedlings planted in the southern edge of the patches. As overall seedling performance underneath the patches was better than in their periphery, our results suggest that seedlings planted in the latter experienced higher level of stress than those planted under the patches, but they also indicate that the source of stress differed at the different locations. Thus, biotic factors related to the composition of the communities of dominant and accompanying species were major drivers of the performance of seedlings planted in the northern edge. In contrast, biotic factors did not affect the survival of seedlings planted in the southern edge, suggesting that the source of stress

161

Chapter 4 may have been predominantly abiotic in this location. The composition of the accompanying species community and soil organic C affected the potential for facilitation underneath the patches. In models where the former variable emerged (the general model, and the models for seedlings planted underneath and in the northern edge of the patches), it had a negative effect on seedling survival. The mechanisms underlying this effect may be similar to those described for the general model. There is no clear ascription of NMDS axes for accompanying species to single species, thus I consider the effect of species composition on facilitative potential as a global effect of the community. The composition of the community of dominant species was only included in the model for seedlings planted on the northern edge of the patches. There, high cover of P. lentiscus and low cover of J. oxycedrus positively affected seedling survival. This result contrasts with the positive effect of phylogenetic distance of the dominant species on seedling survival described above, as P. lentiscus and J. oxycedrus are the phylogenetically closest and most distant species to planted seedlings, respectively. There are some potential explanations for this apparent mismatch. On the one hand, the effect of community composition was not significant in other locations, and it may disappear when all locations are pooled. On the other hand, patches were not monospecific, and differences in the cover of dominant species, other than P. lentiscus, may have been indeed responsible for this effect. A high number of accompanying species decreased the potential for facilitation in the northern edge of the patches. These results support the hypothesis that in this location, patches act as finite communities, in which niche saturation occurs (MacArthur and Levins, 1967; Case, 1991), and are in agreement with the low invasibility of species-rich native communities observed at small spatial scales (Levine, 2000; Stohlgren et al., 2003; Friedley et al., 2007), but not with studies emphasizing the prevalence of positive interactions in species-rich communities (Von Holle, 2005). Thus, a high number of accompanying species may increase the probability that existing species occupy a similar niche as P. lentiscus seedlings and outcompete them, compared to the probability that a number of species in the extant pool facilitate P. lentiscus. Finally, we should bear in mind that I have evaluated the short-term effects of communities on facilitation. Ecological niches may change with 162

Community attributes determine facilitation potential ontogenic development (Mediavilla and Escudero, 2004), and so may do the net outcome of the interactions. I have demonstrated important ecological interactions, but we should be careful extrapolating our results to predict long- term dynamics.

CONCLUSIONS Woody patches had positive effects on the incorporation of new individuals of a keystone species in S. tenacissima steppes. The potential of woody patches to behave as benefactor communities was mainly determined by the cover and composition of the community of accompanying species, litter depth and phylogenetic distance of the dominant species. Furthermore, within- patch differences in the drivers of seedling survival revealed the existence of spatial heterogeneity in stress type and intensity, and facilitation potential, at this scale. Our study highlights the importance of taking into consideration community attributes over pair-wise interactions to predict facilitation potential in multi- specific communities. Community-wide interactions and within-patch heterogeneity should be taken into account when evaluating the outcome of ecological interactions, as they may have profound implications for the composition, function and management of these semiarid steppes.

163

APPENDIX

Significance (p-value) of the effect of community attributes on seedling survival for each GLMM tested. Degrees of freedom (Df) and Akaike's Information Criterion corrected for small samples (AICc), AICc differences (Delta AICc) and Akaike weights for each model are shown. Significant p-values of the covariates are in bold (p<0.05). Model

Community attributes 1 2 3 4 5 6 7 8 9 10 11 12 13 (Intercept) 0.0092 0.0087 0.0082 0.0063 0.0113 0.0155 0.0142 0.0003 <0.0001 <0.0001 0.0055 0.0060 0.0223 Litter depth 0.3687 0.3662 0.3494 0.3667 0.1782 0.1181 0.1073 0.1147 0.0088 <0.0001 0.0112 0.0064 0.0078 NMDS axis 2 0.0498 0.0489 0.0496 0.0445 0.0139 0.0131 0.0169 0.0098 0.0023 <0.0001 0.0357 0.0507 - accompanying spp Cover accompanying 0.1598 0.1596 0.1594 0.1691 0.1216 0.1212 0.1308 0.1178 <0.0001 <0.0001 0.1524 - - spp Phylogenetic distance 0.0716 0.0264 0.0275 0.0292 0.0238 0.0140 0.0117 0.0185 0.0026 0.0100 - - - of dominant species NMDS axis 1 dominant 0.1715 0.1637 0.1683 0.1530 0.1642 0.1167 0.0936 0.1421 0.1179 - - - - spp Soil organic C 0.2378 0.2371 0.2368 0.1660 0.1551 0.2543 0.2758 0.2795 - - - - - Number dominant spp 0.1529 0.1543 0.1587 0.1521 0.2060 0.1509 0.3224 ------Patch size 0.2708 0.2460 0.2488 0.2488 0.2414 0.2581 ------NMDS axis 1 0.2916 0.2906 0.2946 0.2778 0.3214 ------accompanying spp Number 0.3609 0.3494 0.3558 0.3826 ------accompanying spp Soil total N 0.7270 0.7362 0.7248 ------Leaf area index 0.8673 0.8650 ------NMDS axis 2 dominant 0.9475 ------spp Df 16 15 14 13 12 11 10 9 8 7 6 5 4 AICc 201.3 198.9 196.6 194.5 193.0 191.7 190.8 189.6 189.2 189.4 191.7 191.7 193.5 Delta AICc 12.1 9.8 7.4 5.3 3.8 2.5 1.6 0.4 0.0 0.2 2.5 2.5 4.3 Akaike weights 0.00 0.00 0.01 0.02 0.03 0.06 0.10 0.19 0.23 0.21 0.06 0.06 0.03

CHAPTER 5 Litter as a filter for the recruitment of keystone species in Stipa tenacissima steppes

Litter as a filter for the recruitment of keystone species

INTRODUCTION Litter accumulation and ecosystem functioning Carbon accumulation in organic horizons is highly variable. It may represent 10-25% of the organic C accumulated in vegetation in Mediterranean ecosystems (Fons, 1995; García-Cano, 1998). However, heterogeneity in the spatial distribution of litter can be substantial, since it is affected by species distribution, faunal activity, wind, runoff and landform (Welbourn et al., 1981; Boeken and Orenstein, 2001). Soil litter accumulation and the buildup of organic horizons result from the dynamic balance between litterfall inputs and organic matter decomposition and transport. Primary productivity and its drivers, as water availability and soil fertility, are the main determinants of litter deposition at a large scale (Meentemeyer et al., 1982). The main factors controlling litter decomposition are climate, litter quality and activity of soil fauna (Aerts, 1997; Wardle and Lavelle, 1997; Austin and Vitousek, 2000). Climatic factors such as temperature, water availability and radiation directly and indirectly control decomposition rate (Austin and Vivanco, 2006; Kurz-Besson et al., 2006). Litter quality, which has been often associated with lignin, cellulose and N content, also affects decomposition rate (Taylor et al., 1989; Hobbie, 1992; Northup et al, 1995; Talbot and Treseder, 2012; Walela et al., 2014). Finally, soil fauna contributes to litter decomposition and transport by fragmenting plant debris, mixing soil and affecting microorganism activity (Hanlon and Anderson, 1980; Hassall et al, 1987; Fons, 1995; Garcia-Pausas et al, 2004). In fact, recent studies have demonstrated that the stability of soil organic matter depends predominantly on environmental and biological factors more than its molecular structure (Schmidt et al., 2011). Litter is involved in various functional processes such as soil protection, water infiltration and retention, carbon sequestration and nutrient cycling. The litter layer protects the soil surface from raindrop impact (Geddes and Durkeley, 1999), it improves water infiltration and reduces surface runoff and sediment yield (Benkobi et al., 1993, France, 1997; Guevara-Escobar et al., 2007). Moreover, litter is a major source of organic matter, it affects the structure, stability, porosity and aeration of the soil surface (Marshall et al., 1996), and it is a source of carbon, energy and nutrients for soil fauna (González and Zou, 1999;

167

Chapter 5

García-Pausas et al., 2004). Litter also controls soil respiration, and the composition and biomass of the microbial community (Fontaine, 2003; Sayer, 2006). Nutrient cycling through litterfall may account for over 90% of plant nitrogen and phosphorus demand (Chapin et al., 2002). Nitrogen and phosphorus concentration in litter may be strongly reduced by retraslocation and leaching in late stages of leaf and needle lifespan, and thus it is highly dependent on species identity and site fertility (Vitousek, 1982; Del Arco et al., 1991; Aerts, 1996). Low nutrient concentration hampers saprophytic activity and slows down decomposition. Nutrient depleted litter immobilizes nutrients, delays nutrient flow and hinders nutrient losses (Schlesinger et al., 1999). The litter layer may also affect wildfire intensity and severity by either protecting the soil or acting as fuel (Burgan and Rothermel, 1984; Fulé and Covington, 1994; Williams and Wardle, 2007). Litter protection against erosion depends on fire severity, species identity and their interaction, and litterfall recovery rate (García-Cano, 1998; Maestre and Cortina, 2005; Maret and Wilson, 2005; Laughlin and Fulé, 2008). Litter effects on vegetation Litter accumulation may have different effects on plants, depending on the amount and type of litter, on the species involved, ecosystem type and latitude. In general, short-term effects of litter on vegetation are usually negative (Xiong and Nilsson, 1999). Litter may inhibit seed germination and seedling emergence by reducing radiation amount and quality (Vázquez-Yanes et al., 1990; Facelli and Pickett, 1991c). Litter induces latency of seeds requiring certain wavelengths for germination, and this may affect the composition of the soil seed bank (Vázquez-Yanes and Orozco-Segovia, 1992; Metcalfe and Turner, 1998; Milberg et al., 2000). Litter can also hamper seed-soil contact and imbibition required for germination (Rotundo and Aguiar, 2005). By contrast, in shade- tolerant species with larger seeds, litter may enhance germination and seedling establishment by protecting seeds from excessive radiation, reducing evaporation and water stress, and protecting seeds against predation (Reader, 1993; Cintra, 1997; López-Barrera and González-Espinosa, 2001). Seedling survival and growth may be favored by nutrient mineralization, shade and water retention, especially in dry areas (Ginter et al., 1979; Facelli et

168

Litter as a filter for the recruitment of keystone species al., 1999; Xiong and Nilsson, 1999; Boeken and Orenstein, 2001; Lopez-Zamora et al., 2001; Brearley et al., 2003). In contrast, nutrient immobilization in decomposing litter may impair seedling nutritional balance (Noble and Randall, 1999). Furthermore, litter modifies the radiation regime, which may represent a strong filter for seedling emergence, particularly for plants with scarce seed resources (Vázquez-Yanes and Orozco-Segovia, 1992; Eckstein and Donath, 2005; Navarro-Cano, 2007). Litter may be the habitat of herbivorous fauna and create suitable environments for pathogenic fungi (García-Guzmán and Benitez-Malvido, 2003). Finally, litter may release phytotoxic compounds with deleterious effects on both seed germination and seedling establishment (Inderjit and Duke, 2003; Bonanomi et al., 2006; Herranz et al., 2006). At a community level, litter affects species diversity and community structure by favoring some species and inhibiting others (Facelli and Pickett, 1991a, 1991b; Xiong and Nilsson, 1999; Lamb, 2008; Gazol and Ibáñez, 2010). Moreover, disturbances affecting the litter layer, such as changes in litterfall inputs and quality, litter removal, erosion and burning, affect community composition (Tilman, 1993; Sayer, 2006; Ruprecht et al., 2010). Litter also allows plant interactions. Its effect may be either direct (i.e., when litter from one species affects the performance of another species), or indirect, when litter produced by one species influences the outcome of the interaction between two different species (Facelli and Pickett, 1991a, 1991b; Wardle et al., 2004). The balance between facilitation and inhibition can be determined by the quantity and quality of accumulated litter (Xiong and Nilsson, 1999; Eckstein and Donath, 2005). Therefore, plants, by depositing litter, act as ecosystem engineers, physically and chemically modifying the environment, and directly and indirectly controlling the availability of resources for other organisms (Jones et al., 1994). Thus, litter accumulation represents an evolutionary pressure (Stinchcombe and Schmitt, 2006), and may be considered part of the expanded phenotype, since they exert a genetic control beyond the individuals themselves and their life span (Dawkins, 1982; Whitham et al., 2003; Hastings et al., 2007). Litter accumulation in Mediterranean ecosystems Several studies have described litter accumulation in temperate and tropical areas (Covington, 1981; Vitousek et al., 1995; Xiong and Nilsson, 1997; Dames et al., 1998; Schlatter et al., 2006). Conversely, few studies have 169

Chapter 5 comprehensively quantified litter accumulation in drylands (but see Cañellas and San Miguel, 1998; Kavvadias et al., 2001; Descheemaeker et al., 2006). In regions with arid and semiarid climate, vegetation cover is patchy and largely dominated by herbaceous species, shrubs and small trees (Valentin et al., 1999; Whitford, 2002). In these areas, litter accumulation is commonly low due to low litterfall inputs, and relatively high rates of decomposition and vertical transfer (Vallejo et al., 1998; but see Casals et al., 2000 for exceptions). Litter accumulation in humid and dry sub-humid Mediterranean areas is somewhat higher, particularly under species such as pines (Pinus halepensis, P. pinaster Ait., P. nigra Arn.), cork oak (Quercus suber L.), oak (Q. ilex), kermes oak (Q. coccifera), rockrose (Cistus albidus L.) and rosemary (Rosmarinus officinalis L.; Kavvadias et al., 2001; Maestre and Cortina, 2005; Baeza et al., unpublished data). In Mediterranean areas, litter decomposition is strongly driven by water availability and exposition to direct solar radiation (Vallejo et al., 1998; Kemp et al., 2003; Vivanco and Austin, 2006). As a result, decomposition rates can be higher in deep soil horizons than in the soil surface (Rovira and Vallejo, 1997). These differences have been attributed to increased water content and stability at lower soil horizons. There, litter quality may be a major driver of organic matter decomposition (Cortina and Vallejo, 1994; Austin and Ballaré, 2010). Species from arid and semiarid areas often show high mechanical resistance (Méndez-Alonso et al., 2013), which hampers decomposition (Gallardo and Merino, 1993). As discussed above, under Mediterranean conditions, litter may increase water availability by promoting infiltration and reducing evaporation, thus improving plant establishment and growth (Facelli and Pickett, 1991b). Yet, when rainfall becomes too scarce, rainfall events are small and evaporation is high, a large proportion of incoming rainwater may be intercepted by litter, and litter accumulation may have a negative effect on plant performance (Navarro et al., 2009; Villegas et al., 2010). However, few studies have evaluated the effects of litter on recruitment in Mediterranean arid and semiarid areas. For example, Rebollo et al. (2001) found a negative effect of litter on germination and emergence of annual species in a semiarid Mediterranean area. This effect was inversely related to seed size, suggesting the existence of a physical filter. In similar areas, the experimental addition of litter promoted an increase in species richness, annual biomass production and animal activity, and a decrease in runoff

170

Litter as a filter for the recruitment of keystone species

(Boeken and Orenstein, 2001). Navarro et al. (2009) found that germination and early growth of Stipa tenacissima was affected by the amount of P. halepensis litter and the location of the seed within this layer. In addition, allelopathic effects of litter on seed germination and seedling growth have been described in woody Mediterranean species as P. halepensis, Quercus douglasii Hook & Am., Q. coccifera, C. ladanifer L., Adenostoma fasciculatum Hook. & Arn. and some ericaceous species (Christensen and Muller, 1975; Hobbs, 1984; Callaway et al., 1991; Chaves and Escudero, 1997; Fernández et al., 2008;. Alrababah et al., 2009). In previous chapters, I have discussed the key role of resprouting shrub species in S. tenacissima steppes. Despite of their low cover, their presence has been associated with significant changes in ecosystem functioning and community composition. I have shown that community composition, complexity and facilitation ability are affected by the presence and composition of woody patches. Spatial heterogeneity in the performance of accompanying species and the recruitment of patch-forming species suggest that interactions between dominant and accompanying species are more complex than suggested by the competition-facilitation theory. Litter may play an important role in plant-plant interactions in woody patches, as patch-forming species show strong contrasts in their ability to build up organic layers. The outcome of these interactions is hardly predictable, as physical and chemical processes may act concurrently at the various stages of plant establishment. Thus, thick and compacted litter layers, may hinder seed contact with the mineral soil and subject seeds to strong fluctuations in temperature and moisture, and expose seeds to direct radiation, while reducing contact with allelopathic compounds, compared to seeds lying on the mineral soil (Fig. 1). In addition, litter may favor pathogen growth because of its higher moisture, but it may reduce the probability of predation.

171

Chapter 5

Allelochemical impact - Pathogen contact + Protection from predation

Temperature variability Moisture variability + Exposure to radiation -

Distance to topsoil - + -

Figure 1. Schematic representation of the biotic and abiotic drivers of germination in seeds located in different soil microhabitats. Yellow ellipses represent seeds, dark brown elements represent litter and light brown rectangles represent mineral soil.

In this chapter I evaluate the effect of a range of litter types on the germination of two key widespread species in S. tenacissima steppes: a patch- forming woody species and a perennial grass. My objective is to deepen our understanding on litter accumulation under Mediterranean woody species, assess its role in plant-plant interactions, and discuss its importance as part of the expanded phenotype of these species. I evaluated mechanical and chemical effects of litter by performing three laboratory experiments in which I evaluated 1) seed germination on top of the litter layer, 2) seed germination underneath the litter layer, and 3) the effect of litter extracts on seed germination.

172

Litter as a filter for the recruitment of keystone species

MATERIAL AND METHODS Study site and species I characterized and collected litter samples under five common woody species with contrasted litter accumulation patterns: Quercus coccifera L., Pistacia lentiscus L., Rosmarinus officinalis L., Pinus halepensis Mill and Rhamnus lyciodes L. For each species, I selected 8 adults of average size forming monospecific patches, in order to achieve homogeneity in litter composition. The selected individuals were spaced at least 50 m, and were scattered in slopes and abandoned crop fields in an area covering 10 km2 in Alicante (southeast Spain). This area has semiarid Mediterranean climate, with an average annual rainfall of 332 mm and an average temperature of 17.1ºC (San Vicente del Raspeig Station; Rivas-Martínez and Rivas-Sáenz, 2002). For the germination experiments, I selected two common species with contrasting morphology and functionality, both during seed and in adult stages. The two species, Brachypodium retusum Pers. (Beauv.) and P. lentiscus (Fig. 2), were obtained from local provenances (Intersemillas S.A. and Forest Seed Bank, Generalitat Valenciana, respectively). Brachypodium retusum is a perennial grass that produces rhizomes. It is distributed in the central and western Mediterranean basin, forming open grasslands and under low shrubland and pine forests in thermo- and meso-Mediterranean areas, and abandoned crop fields (Bolós, 2001). It is very abundant in S. tenacissima steppes, where it may form dense mats in the vicinity of patches of woody vegetation. Brachypodium retusum is a barochorous species, which seeds are elongated with an average length of 9 mm and an average width of 1.5 mm. Pistacia lentiscus is a dioecious, perennial resprouting shrub, common in shrublands and abandoned crop fields in thermo- and meso-Mediterranean areas (Verdú and García-Fayos, 1996). It is one of the most abundant patch-forming species in S. tenacissima steppes (Chapter 1). Fruits of P. lentiscus are round and fleshy, containing one 4 mm diameter seed, and they are usually dispersed by birds.

173

Chapter 5

Figure 2. Seeds used for the germination experiment: Pistacia lentiscus (left) and Brachypodium retusum (right). A centimeter scale is shown.

Effects of litter on germination Between December 2007 and January 2008, we collected two samples of organic horizons and four samples of mineral soil under 8 individuals per species. Organic horizons were defined as the litter layer and the underlying layer formed by decomposing organic debris. The organic-mineral transition was easily identifiable by changes in texture and color, due to high carbonate content and low organic matter content of the mineral soil. Organic horizons were removed intact by using PVC cylinders of 5 cm high and 9 cm in diameter. Once the mineral soil was reached, the cylinder containing the organic layers on the inside was removed with a spatula, depositing the assembly in a plastic bag, and transported to the laboratory. We used 8 cm high cylinders to collect organic horizons of P. halepensis as litter depth was larger under this species. Finally, we dug the underlying mineral soil with a shovel, to a depth of 15 cm, and transported the samples in plastic bags to the laboratory, where soil was air-dried. Additionally to the previous samples, we collected 10 cylinders of litter per species from different individuals and combined all litter samples from a given species for allelopathical tests. Litter was transported to the laboratory and air- dried.

174

Litter as a filter for the recruitment of keystone species

In the laboratory, we filled 32 PVC cylinders of 5 cm high and 9 cm in diameter for each species with the collected mineral soil. On 16 of them, we deposited the cylinders containing the organic horizons of the corresponding individuals on top of the mineral soil. Cylinders containing mineral soil and organic horizons were attached with tape. On the additional 16 cylinders, empty cylinders were attached on top. All cylinders were placed on plastic trays and its position was changed over the course of the experiment. Two consecutive experiments were conducted to assess the prevalence of physical or chemical litter effects: 1. Germination on litter layer. Brachypodium retusum was sown on half of the cylinders of each type (with and without litter), and P. lentiscus was sown on the rest. Fifty seeds per cylinder were placed directly on the surface of the mineral soil or litter layer. At the onset of the experiment, all cylinders were watered with distilled water to saturation. During the experiment, they were kept moist by spraying distilled water daily during the first three days, and every 2-3 days thereafter. The experiment lasted eight weeks and was conducted under laboratory conditions: temperature ranged from 18.5 to 24.0ºC and relative humidity between 34-63%. At the end of this period, the number of germinated seeds per cylinder was recorded. I considered germinated seeds when the radicle exceeded 0.5 cm in length. 2. Germination under litter layer. Once the previous experiment concluded, I removed seedlings and seeds from each cylinder, and separated the litter layer from the mineral soil with a spatula. Then, I sowed 50 seeds of the same two species per cylinder, and placed the litter layer back on top of the mineral soil and the seeds. I also sowed the cylinders containing no litter. Due to a lack of B. retusum seeds from the same batch, the number of cylinders with mineral soil used in this experiment was halved (4 cylinders per litter species). At the onset of the experiment all cylinders were watered with distilled water to saturation. During the experiment, they were kept moist by spraying distilled water daily during the first two weeks, and every two days during the following two weeks. The experiment lasted for four weeks, and was conducted under laboratory conditions: temperature ranged from 24.9 to 30.1ºC and relative humidity between 42-54%. At the end of this period the number of germinated seeds was counted in each cylinder.

175

Chapter 5

Effects of litter extracts on germination When the second experiment ended, I removed all seeds and seedlings from the cylinders. I used 20 cylinders with mineral soil per species, and sowed 20 seeds per cylinder of either B. retusum or P. lentiscus. Half of the cylinders sown with each species were watered with distilled water, and the other half were watered with a litter extract every 2-3 days. The extract was obtained by shaking a mixture of litter and distilled water for 48h in an orbital shaker (Garnett et al., 2004). The amount of litter used was proportional to litter accumulation under each species. Proportions ranged between 0.02 and 0.2 g litter l-1. The resulting solution was filtered through Whatman filters Nº4 and stored at 4ºC until use. The solution was prepared along the experiment as needed, to avoid storage for more than 5 days. The number of germinated seeds was recorded daily. The experiment lasted for seven weeks, when germinations finished, and it was conducted in an incubation chamber with 12-h photoperiod, 18-24ºC temperature and 62-79% relative humidity. Extracts were less concentrated than in other studies on allelopathical interactions (100-200 g l-1) but closer to natural conditions. Other methods extract allelochemical compounds more efficiently and exhaustively (Inderjit and Nilsen, 2003), but they are less suitable for simulating natural conditions. Characterization of organic horizons We quantified the area covered by litter under each individual in the field (N=8) by measuring the largest diameter and its perpendicular diameter and approximated it to an ellipse. In these environments, litter accumulates mostly under the projection of the canopy. We measured litter depth in the center of the litter patches, in their periphery and in the intermediate area (N=4 sampling points per location). After the germination experiments, we inserted a 1.3 mm-plastic mesh, with the help of a spatula, underneath the litter layer to facilitate repeated weighting. This setting allowed measuring water-holding capacity and short-term evaporation rate of litter. Cylinders were watered to saturation and weighed. Then, they were placed in an incubation chamber in the dark, where temperature was 19-23ºC and relative humidity 60-80%. Cylinders were weighed at increasing time intervals, from 1 h to 132 h, when weight stabilized. Water retention

176

Litter as a filter for the recruitment of keystone species capacity was estimated as the difference between the weight to saturation of the litter layer and its dry weight. Short-term evaporation rate was estimated as the ratio of the negative exponential equation for the adjusted water content over time for each type of litter. Carbon content of the litter was estimated as half their dry weight (Nelson and Sommers, 1982), considering litter density measured in laboratory samples. Finally, I estimated hydrophobicity of the organic horizons by the water drop penetration time test (Dekker and Ritsema, 1994). I put three drops of distilled water on different parts of the litter surface chosen at random, and recorded the time for the second drop to infiltrate. Statistical analysis The effect of species on litter properties and germination rate was tested by analysis of variance (ANOVA) with one fixed factor and five levels. Differences between means were compared using Tukey-HSD test. Litter depth and carbon content were logarithmically transformed to meet criteria of normality and homoscedasticity. The net effect of litter on germination rate was evaluated in both experiments by estimating the relative interaction intensity index, RII = (Gh - Gs ) / (Gh + Gs), where Gh is the number of germinated seeds in the presence of litter and Gs is the number of germinated seeds without litter (Armas et al., 2004). This ratio quantifies the extent of an interaction, with defined boundaries between -1 and 1, and it is symmetrical on zero. In this study, the effect of litter on germination is either positive or negative as RII approaches 1 and -1, respectively. Student’s t-test was used to assess the significance of differences between RII and zero. Germination percentages were arcsine transformed prior to analysis. All analyses were performed with the R statistical package for Windows 3.1.0 (R Development Core Team, 2014).

177

Chapter 5

RESULTS Characterization of organic horizons Litter accumulation and properties differed across the five studied species. Litter depth in the inner, middle and outer part of the litter surface under all species showed a similar pattern (Fig. 3). Rhamnus lyciodes had less litter than the other species, whereas Q. coccifera and P. halepensis showed the highest accumulation. Carbon density was higher under Q. coccifera and P. halepensis than under other species, exceeding by more than twice the amount of carbon accumulated underneath R. lycioides and R. officinalis (Table 1).

inner middle 8 outer

a a 6

Litter depth (cm)

b 2 b c

0 P. halepensis P. lentiscus Q. coccifera R. lycioides R. officinalis Species Figure 3. Litter depth measured in the inner, middle and outer part of litter patches. Means ± 1SE for N=8 individuals are shown. Different letters indicate significant differences between species for the middle location, while inner and outer locations showed the same pattern (ANOVA and post-hoc Tukey HSD test, p<0.05).

178

Litter as a filter for the recruitment of keystone species

Table 1. Litter properties of five Mediterranean woody species. Means ± 1SE, number of samples (N) and results of ANOVA are shown. Different letters indicate significant differences between species for the same variable (post-hoc Tukey HSD test, p<0.05). d.f.=degrees of freedom.

Litter surface Litter Moisture Hydropho- Species area per accumulation content at bicity (min) individual (m2) (g C m-2) saturation (g g-1)

Rosmarinus officinalis 1.5±0.3 a 859±870 bc 0.83±0.07 b 49±14 ab

Quercus coccifera 14.8±3.6 b 2335±240 a 1.02±0.10 b 113±11 a Pinus halepensis 28.2±4.0 c 2321±397 a 0.92±0.09 b 113±18 a

Rhamnus lyciodes 3.1±0.7 a 616±119 b 1.44±0.16 a 5±10 b Pistacia lentiscus 7.8±1.3 ab 1541±133 ac 1.14±0.09 a 35±10 b N 8 8 10 10*

F 18.790 15.574 5.098 11.885

d.f. 4 4 4 4

p-value <0.001 <0.001 0.0018 <0.001

* Except for R. lyciodes where N=6.

179

Chapter 5

Water retention capacity also varied between species. Rhamnus lycioides and P. lentiscus litter retained more water and showed less hydrophobicity than other species. The highest hydrophobicity was observed in Q. coccifera and P. halepensis. On the other hand, short-term evaporation rate was highest in R. lycioides and R. officinalis, and lowest in P. halepensis (Fig. 4, Table 2).

1.8 R. officinalis 1.6 R. lycioides Q. coccifera 1.4 P. halepensis -1 P. lentiscus 1.2

1.0

0.8

0.6

0.4

Water content (g water) (g litter) (g Waterwater) (g content 0.2

0.0

0 20 40 60 80 100 120

Time after water saturation (hours) Figure 4. Reduction in water content of different types of litter at 19-23ºC and HR = 60- 80%. Curve fitting parameters and estimated short-term evaporation rate are shown in Table 2.

180

Litter as a filter for the recruitment of keystone species

Table 2. Short-term evaporation rate of different litter types estimated from the equation y=a·e-kt, where y is water content (g·g-1), t is drying time (hours), k is evaporation coefficient and a is a constant. The goodness of fit and its probability are also shown. N=10 in all cases.

2 Species Evaporation coefficient (k) R F1,18 p-value

Rosmarinus officinalis 0.0173 0.977 776.847 <0.001

Rhamnus lycioides 0.0285 0.843 96.687 <0.001

Quercus coccifera 0.0056 0.892 148.951 <0.001

Pinus halepensis 0.0030 0.926 223.857 <0.001

Pistacia lentiscus 0.0084 0.969 570.601 <0.001

Litter effects on germination: seeds on litter layer Germination was higher when seeds of both species were sown on R. lycioides litter than on litter of other species (Fig. 5A, Table S1). Brachypodium retusum germination was also higher on P. lentiscus litter than on litter of other species. I observed no effect of the mineral soil on germination, except for a reduction in P. lentiscus germination when sown on P. halepensis soil compared to other types of soil (Fig. 5B).

181

Chapter 5

A) 100 R. officinalis Q. coccifera P. halepensis a 80 R. lycioides P. lentiscus

a 60 b

40 b c b Seed germination (%) c b c 20 b

0 Pistacia lentiscus Brachypodium retusum Sown species B) 100

a a a a

Seed germination (%)

0 Pistacia lentiscus Brachypodium retusum Sown species Figure 5. Germination of Pistacia lentiscus and Brachypodium retusum on the litter layer (A) and on the mineral soil (B) collected under five woody species. Means ± 1SE for N = 8 are shown. Different letters indicate significant differences for each sown species and seed location (ANOVA and Tukey-HSD test, p <0.05).

182

Litter as a filter for the recruitment of keystone species

The net effect of litter of R. officinalis, Q. coccifera and P. halepensis on P. lentiscus germination was negative when seeds were sown on the litter layer (Fig. 6, Table S2). In contrast, the effect of R. lycioides litter on P. lentiscus germination was positive. Litter had a negative effect on B. retusum germination, except for R. lycioides litter, which had no effect. Moreover, the negative effect of litter on the germination of B. retusum was greater when seeds were sown on litter of R. officinalis, Q. coccifera and P. halepensis than litter of P. lentiscus.

1.0 R. officinalis 0.8 Q.coccifera P. halepensis 0.6 R. lycioides P. lentiscus 0.4 * b 0.2

0.0 RII b -0.2 b -0.4 ab * a a -0.6 * * a a * * -0.8 a a * * -1.0 Pistacia lentiscus Brachypodium retusum Sown species Figure 6. Net effect (RII) of litter on germination of Pistacia lentiscus and Brachypodium retusum when seeds were sown on top of the litter layer. Means ± 1SE for N=8 are shown. Different letters indicate significant differences for each sown species and seed location (ANOVA and Tukey-HSD test, p <0.05). Asterisks indicate significant differences from zero (Student’s t-test, p <0.05).

183

Chapter 5

Litter depth and weight were negatively correlated to the number of seedlings for both seeded species (Fig. 7).

A) B) 100 100 P. halepensis R2 = 0.28 P. lentiscus p < 0.001

(%) 80 Q. coccifera 80 R. lycioides

(%) 60 R. officinalis 60 R2 = 0.26 40 p = 0.002 40

20 20

Pistacia lentiscus Pistacia

Germination on litter litter on Germination

Germination on litter on Germination of Brachypodium retusum retusum Brachypodium 0 0

0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 Litter depth (cm) Litter depth (cm) C) D) 100 100 2 R2 = 0.13 R = 0.21 p = 0.082 80 p = 0.059 (%) 80

(%) 60 60 40 40 20

Pistacia lentiscus Pistacia

Germination on litter litter on Germination

Germination on litter on Germination of

Brachypodium retusum retusum Brachypodium 20 0

0 0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 Litter weight (g) Litter weight (g) Figure 7. Relationship between litter accumulation (litter depth A-B and litter weight C-D) and seed germination when seeds were sown on the litter layer. Linear regression parameters correspond to the ensemble of all litter types for a given seeded species.

Litter effects on germination: seeds under litter layer I observed no effect of litter type on B. retusum germination when seeds were sown under litter layer (Fig. 8A, Table S1). The number of germinated P. lentiscus seeds was lower under P. halepensis litter than under R. lycioides litter. Other species showed no differences in their effect on P. lentiscus germination. In this experiment, I found no effect of the type of mineral soil on the germination of P. lentiscus and B. retusum (Figure 8B).

184

Litter as a filter for the recruitment of keystone species

A) 100 R. officinalis Q. coccifera P. halepensis 80 R. lycioides P. lentiscus

Seed (%) germination

b ab ab ab a 0 Pistacia lentiscus Brachypodium retusum Sown species B) 100

Seed (%) germination

0 Pistacia lentiscus Brachypodium retusum Sown species Figure 8. Germination of Pistacia lentiscus and Brachypodium retusum under the litter layer (A) and on mineral soil (B), collected under five woody species. Means ± 1SE for N = 8 are shown. Different letters indicate significant differences for each sown species and seed location (ANOVA and Tukey-HSD test, p <0.05).

185

Chapter 5

The net effect of litter on the germination of both species was negative, except for R. lycioides litter, which had no effect on any species, and P. halepensis litter, that had no effect on B. retusum (Fig. 9, Table S2). The magnitude of the negative effect was similar among the different types of litter (ANOVA, p> 0.05). The relation between the net effect when seeds were sown on and under the litter layer was positive but weak (R2=0.147, p-value=0.014 for P. lentiscus seedlings, and R2=0.083, p-value 0.219 for B. retusum).

1.0 R. officinalis Q. coccifera P. halepensis R. lycioides P. lentiscus 0.5

0.0

RII

(*) -0.5 (*) (*) (*) (*) * * -1.0 Pistacia lentiscus Brachypodium retusum Sown species Figure 9. Net effect (RII) of litter on germination of Pistacia lentiscus and Brachypodium retusum when seeds were sown under the litter layer. Means ± 1SE for N=8 are shown. Asterisks indicate significant differences from zero (Student’s t-test, p <0.05) or marginally significant differences (Student’s t-test, p <0.15, in parenthesis).

186

Litter as a filter for the recruitment of keystone species

There was a significant relationship between the number of germinated seeds and litter accumulation (litter depth and weight) when seeds were sown under the litter layer for both seeded species (Fig. 10).

A) B) 100 100 P. halepensis R2 = 0.49 P. lentiscus p < 0.001

(%) 80 Q. coccifera 80 R. lycioides

(%) 60 R. officinalis 60 R2 = 0.11 40 p = 0.047 40

20 20

Pistacia lentiscus Pistacia

Germination on litter litter on Germination

Germination on litter on Germination of Brachypodium retusum retusum Brachypodium 0 0

0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 Litter depth (cm) Litter depth (cm) C) D) 100 100 2 R2 = 0.16 R = 0.40 p = 0.051 80 p = 0.005

(%) 80

(%) 60 60

40 40

20 20

Pistacia lentiscus Pistacia

Germination on litter litter on Germination

Germination on litter on Germination of Brachypodium retusum retusum Brachypodium 0 0

0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 Litter weight (g) Litter weight (g) Figure 10. Relationship between litter accumulation (litter depth A-B and litter weight C- D) and seed germination when seeds were sown under the litter layer. Linear regression parameters correspond to the ensemble of all litter types for a given seeded species.

187

Chapter 5

Litter effects on germination: non-physical effects Germination was very high when seeds of both sown species where watered with litter extract (Fig. 11). These values were higher than those recorded in the previous experiments. In this case, I found no effect of litter type on seed germination (ANOVA p>0.05, Table S1).

100 R. officinalis Q. coccifera P. halepensis R. lycioides 80 P. lentiscus

Seed germination (%)

0 Pistacia lentiscus Brachypodium retusum Sown species Figure 11. Germination of Pistacia lentiscus and Brachypodium retusum sown on mineral soil and watered with litter extract from five woody species. Means ± 1SE for N = 8 are shown.

188

Litter as a filter for the recruitment of keystone species

The net effect of watering with litter extract with respect to distilled water was negligible for all types of litter extract and for both sown species (Fig. 12, Table S2).

1.0 R. officinalis Q. coccifera P. halepensis R. lycioides 0.5 P. lentiscus

0.0

RII

-0.5

-1.0 Pistacia lentiscus Brachypodium retusum Sown species Figure 12. Net effect (RII) of litter extract from five woody species on germination of Pistacia lentiscus and Brachypodium retusum. Means ± 1SE for N=8 are shown.

DISCUSSION

Litter accumulation Litter accumulation was higher than observed in other Mediterranean areas (Ferran, 1996; Cañellas and San Miguel, 1998). For example, Ferran (1996) found 1.7 kg m-2 of litter in communities dominated by Q. coccifera, which contrasts with the 4.7 kg m-2 observed in this study. Discrepancies may partly result from differences in the sampling method. I estimated organic matter accumulation by considering litter density and volume under each individual

189

Chapter 5 patch, while in other studies, litter sampled from known surface areas is directly weighted. The later approach integrates spatial heterogeneity at a larger scale. My results are, however, in agreement with studies using similar methods (Fons, 1995; Baeza et al. Fundación CEAM, unpublished data). Interspecific differences in litter accumulation may reflect differences in age and balance between litterfall and litter decay. For example, R. officinalis, one of the species with less litter accumulation, as other small shrubs, has a shorter lifespan (Capitanio and Carcaillet, 2008; Eugenio et al., 2012) than Q. coccifera and P. halepensis, the two species showing highest ability to build up thick litter layers. Furthermore, the balance between inputs and outputs may differ for each species. For example, R. lycioides, which is a resprouting and conspicuous species, shows low litter accumulation, which may be related to low litterfall inputs and relatively high decomposition rates. In contrast, P. lentiscus may accumulate more litter because litterfall inputs are higher in this species than in R. lycioides (Lavado et al., 1989). The decay rate of R. officinalis litter is higher than that of P. halepensis (Almagro and Martinez-Mena, 2012). Effects of litter on germination The effect of litter on germination was strongly dependent on litter type. In species with high litter accumulation, as P. halepensis and Q. coccifera, litter had a negative effect on the germination of seeds located either on top or underneath the litter layer, which suggests the presence of a physical barrier for germination. The litter layer may hamper soil-seed contact and seed imbibition needed to initiate this process (Rotundo and Aguiar, 2005). In contrast, in species with lower litter accumulation as R. officinalis, P. lentiscus and R. lycioides, the effect of litter on germination was heterogeneous. I found a negative relationship between litter accumulation and seed germination. However, the relationship was weak, suggesting that other factors may be more important in controlling seed germination. For example, the physical structure of the litter layer may affect seed germination (Donath and Eckstein, 2008). This probably explains the contrasted effect of litter on seed germination in R. lycioides vs. R. officinalis and P. lentiscus. Thus, unlike R. lycioides, litter of R. officinalis and P. lentiscus showed a negative effect on seed germination. Rosmarinus officinalis litter is strongly compacted, forming an intertwined network of litter fragments at different stages of decomposition. Similarly, Pistacia lentiscus develops a dense litter mat. In 190

Litter as a filter for the recruitment of keystone species contrast, the litter layer of R. lycioides is loose. Thus, despite low levels of litter accumulation under these species, litter density may hinder seed movement and prevent contact with the mineral soil in R. officinalis and P. lentiscus, compared to R. lycioides. Furthermore, the scarce abundance and loose morphology of the litter layer under R. lycioides may allow seedling contact with the mineral soil, and protect germinating seeds from desiccation. In addition to the physical effect associated with litter accumulation and structure, water retention capacity may also influence seed germination. One would expect that a wet litter layer constitutes an optimal environment for germination. However, this layer may also have negative effects on seed germination by modifying radiation income and quality, and intercepting rainfall. This later effect depends on rainfall amount and distribution, and seed location (Navarro-Cano et al., 2009; Villegas et al., 2010). As observed in this study, R. lycioides litter is highly hydrophilic and has great water-holding capacity, but moisture accumulation in this type of litter is scarce. In contrast, P. halepensis and Q. coccifera litter, show low water-holding capacity, high hydrophobicity, and high water accumulation capacity. In general, germination was negatively related to litter accumulation and positively related to water-holding capacity. However, it must be noted that in this study, litter was regularly watered, and seeds located under the litter layer never dried out. These results contrast with studies showing no effect of litter cover on seed germination under constant humidity (Eckstein and Donath, 2005). Other authors have reported a negative effect of litter on field germination (intermittent drought) and a positive effect in the laboratory (constant humidity), which they attributed to water interception under conditions of low water availability (Navarro-Cano et al., 2009). I cannot infer the effect of litter under low water availability from my data, but my results show a negative effect of litter on germination, and suggest that the effect is predominantly mechanical, and related to litter abundance and structure. These two factors, together with litter water-holding capacity may be critical for germination in the absence of regular water inputs. Allelopathic effects on seed germination depend on seed species and may be proportional to litter abundance (Alrababah et al., 2009). Allelopathic effects of P. halepensis and Q. coccifera on germination have been previously described (Fernández et al., 2008; Alrababah et al., 2009; Navarro-Cano et al., 2009).

191

Chapter 5

However, the negative effect of litter on seed germination observed in this study was apparently unrelated to allelopathy. Indeed, the effect of litter was more intense when seeds were sown on top of the litter, compared to seeds sown under this layer or watered with litter extracts. My results agree with other studies showing no allelopathic effects of P. halepensis litter on P. lentiscus growth (Maestre et al. 2004). Moreover, in this study I evaluated seed germination from a patch forming species (P. lentiscus) and an accompanying species (B. retusum), whose seeds are morphologically distinct and with different germination requirements (García-Fayos et al., 2001; Espejo, 2007). Yet, the observed effects were similar for both types of seeds, suggesting that the effect of litter of the studied species does not depend on seed type. In addition, the weak correlation between the net effect of litter when seeds were sown on top and under the litter layer for both seed species suggests that the effect of the litter species on germination is not independent of seed location and there may be an interaction between these factors. Implications on population dynamics, community composition and management The negative effect of litter on the recruitment of new individuals may have important consequences on population dynamics, species evolution and community composition. As a result of this interference, the progeny of those species lacking mechanisms for long-distance seed dispersal may be reduced under adult individuals. Moreover, a negative effect of litter may represent a filter for the establishment of other species (Tilman, 1993; Gazol and Ibáñez, 2010). According to my results, the studied woody plants have the ability to modulate resource flow and ecological conditions by means of litter accumulation, affecting seed germination. Hence, the studied species behave as ecosystem engineers (sensu Jones et al., 1994) and litter accumulation contributes to their expanded phenotype (Hastings et al., 2007). In addition, litter interference may have significant implications on competitive exclusion and the spatial distribution of vegetation. My results contrast with studies showing a direct and positive relationship between woody resprouting plants and richness of vascular plants in Mediterranean steppes and shrublands (Verdú and García-Fayos, 1996; Maestre and Cortina, 2004b; Chapter 1). Various processes may explain this disagreement. On the one hand, 192

Litter as a filter for the recruitment of keystone species germination and recruitment could precede litter accumulation. For example, fire, wildlife and humans may remove litter from a given microsite for months to decades (Lavado et al., 1989; Serrasolsas et al., 1989; Cañellas and San Miguel, 1998; Gutiérrez et al., 1997; Sayer, 2006; Dunham, 2011), and open windows for germination and seedling establishment. If this was the case, and if the disturbance was intense enough to remove the aboveground parts of resprouting woody species, the age of resprouts and accompanying vegetation should be correlated, which is not common (Capitanio and Carcaillet, 2008). On the other hand, the negative effect of litter accumulation on germination described in this chapter could be offset by a positive effect of woody species on other phases of plant establishment (Amat et al., 2015; Chapter 4). Thus, woody species promote seed rain, particularly from ornitochorous species (A. Castillo, pers. comm.), and generate favorable microenvironments for seedling establishment (Verdú and García-Fayos, 1996; Amat et al., 2015; Chapter 4). The impact of woody vegetation patches on community composition and ecosystem functioning has been a major argument to justify their use in restoration programs in semiarid areas (Castro et al., 2004; Cortina et al., 2004; Gómez-Aparicio, 2004; Cortina et al., 2011). I have shown that litter does not contribute to this interaction and, indeed, it could have a negative effect on germination of the introduced species. The negative effect of litter may help to explain the idiosyncratic nature of plant-plant interactions in arid and semiarid areas (Maestre et al., 2009a). According to my results, mulch, a widely used eco- technological tool to restore degraded arid and semiarid areas, should be employed with caution, particularly when applied in combination with seeding. In general, and regardless of mulch structure and composition, application doses should not exceed 1700 g m-2 to avoid deleterious effects on germination. It is noteworthy that these doses are well above common recommended doses for commercial products (between 100 and 400 g m-2).

193

Chapter 5

CONCLUSIONS The effect of litter on germination depends on the abundance and structure of the litter layer, rather than the seeded species. Results from this study highlight the important role that litter of woody species may play on seed germination and recruitment in semiarid areas. The negative relationship between litter and germination may be attributed to a physical filter, which prevents soil-seed contact, and not to an allelochemical filter. This physical filter, together with the existence of multi-specific patches, and the positive relationship observed between patches and richness of vascular plants suggest that litter accumulation and seed germination are spatially or temporally segregated, or the existence of positive interactions in other phases of plant establishment. These results have important implications on population and community dynamics in S. tenacissima steppes and on restoration techniques aimed at the establishment of vascular plants in these areas.

194

Litter as a filter for the recruitment of keystone species

APPENDIX

Table S1. Results from one-way ANOVAs with 5 levels (litter species) to assess the effect of litter on seed germination. Seeds of Pistacia lentiscus and Brachypodium retusum were sown on top of the litter layer, underneath it and watered with litter extract. RII corresponds to the net effect of each treatment on germination. Significant values (p<0.05) are in bold. Degrees of freedom = 4. Analysis Sown species F p-value Germination on litter Pistacia lentiscus 9.169 <0.001 Brachypodium retusum 21.388 <0.001 Germination on mineral Pistacia lentiscus 5.527 0.002 soil (experiment 1)* Brachypodium retusum 1.550 0.210 RII germination on litter Pistacia lentiscus 5.413 0.002 Brachypodium retusum 17.361 <0.001 Germination under litter Pistacia lentiscus 3.530 0.016 Brachypodium retusum 0.857 0.499 Germination on mineral Pistacia lentiscus 1.483 0.228 soil (experiment 2)* Brachypodium retusum 1.665 0.210 RII germination under Pistacia lentiscus 1.487 0.227 litter Brachypodium retusum 2.381 0.098 Germination with litter Pistacia lentiscus 0.313 0.866 extract Brachypodium retusum 0.557 0.696 Germination on mineral Pistacia lentiscus 1.188 0.346 soil (experiment 3)* Brachypodium retusum 0.277 0.889 RII germination with litter Pistacia lentiscus 0.497 0.738 extract Brachypodium retusum 0.665 0.624

*Controls (only mineral soil) of three independent experiments.

195

Table S2. Results of Student’s t-tests to assess if the net effect (RII) of litter or litter extracts on germination differed from zero. Significant values (p<0.05) are in bold. d.f.= degrees of freedom. RII germination with litter RII germination on litter RII germination under litter Sown species Litter species extract d.f. t p-value d.f. t p-value d.f. t p-value Pistacia Pinus halepensis 7 -1.857 0.106 7 -3.319 0.013 4 -0.583 0.591 lentiscus Quercus coccifera 7 -6.151 <0.001 7 -2.171 0.067 4 -1.117 0.327

Pistacia lentiscus 7 -1.376 0.211 7 -2.127 0.071 4 0.391 0.716 Rhamnus lycioides 7 3.023 0.019 7 -1.282 0.241 4 -2.113 0.102 Rosmarinus officinalis 7 -4.853 0.002 7 -2.172 0.066 4 -1.242 0.282

Brachypodium Pinus halepensis 7 -6.872 <0.001 3 -1.239 0.304 4 -1.746 0.156 retusum Quercus coccifera 7 -7.426 <0.001 3 -4.216 0.024 4 -0.868 0.434 Pistacia lentiscus 7 -4.012 0.005 3 -2.995 0.058 4 0.400 0.710

Rhamnus lycioides 7 -0.581 0.580 3 -1.440 0.245 4 0.000 1.000 Rosmarinus officinalis 7 -8.358 <0.001 3 -2.567 0.083 4 0.266 0.803

GENERAL DISCUSSION AND CONCLUSSIONS

Discussion

GENERAL DISCUSSION Along the five Chapters of this thesis I have described woody patches of S. tenacissima steppes and their ecological context, explored their functional role and discussed the implications of these findings on steppe management. My research adds to the increasing knowledge on the composition, dynamics and function of S. tenacissima steppes accumulated over the past decades, and contributes to complete the picture of these vast western Mediterranean landscapes. In this section, I summarize and integrate the information contained in each chapter, and identify knowledge gaps. 1. What is a woody patch? Owing to their extent and their socio-economic, cultural and ecological importance, Stipa tenacissima steppes have been the subject of numerous studies (Le Houérou, 1986, 2001; Maestre and Cortina, 2004a, 2004b; Alados et al., 2006; Mayor et al., 2008; Cortina et al., 2009; Ramírez and Bellot, 2009; Rey et al., 2011; to mention a few). Comparatively, the woody component of S. tenacissima steppes has received less attention. Yet, some studies suggest that resprouting shrubs may play an important role in these ecosystems (Maestre et al., 2003, 2006; Maestre and Cortina, 2005), acting as true keystone species (Eldridge et al., 2011). In previous sections, I have shown that these shrubs form heterogeneous communities with specific attributes, they have an impact in their immediate environment and community, and their properties are modulated by internal and external factors. But what is inside a patch? Species in woody patches can be ascribed to two groups, according to their morpho-functional traits: (i) a few dominant large resprouting shrubs and (ii) a larger set of smaller woody and herbaceous species that we identified as accompanying species. The canopy area of woody patches is dominated by one of six dominant patch-forming species. Patches dominated by Rhamnus lycioides are the most abundant. Yet, despite the dominance of one or a few species, patches are commonly multispecific: dominant and accompanying species forming communities of 2-29 co-occurring species. Woody patches can be considered independent communities, forming separated entities of interacting species in a well-defined space, usually determined by the size of the dominant species. Patch attributes described through this dissertation, such as richness and cover of

199 Discussion accompanying species, soil properties, soil water content, incident radiation and litter accumulation, support the idea of spatially-delimited patches, as they show sharp contrasts between the canopy projection area and its surroundings. This gradient has significant implications on patch growth and the incorporation of new individuals. Patch composition is primarily determined by the dominant species (Chapter 1). Two major gradients emerge from multivariate analyses of patch composition which are mostly related to the abundance of four species: R. lycioides – Quercus coccifera and Pistacia lentiscus – Ephedra fragilis. In addition, the most influent accompanying species in patch composition are S. tenacissima and Brachypodium retusum, which appear in opposite ends of a third gradient. We have found that species richness within a patch increases with its area. However, the increase is not linear. Smaller patches are relatively richer than big ones, suggesting the existence of species saturation in large patches. Various factors may explain differences in the slope of the species-area relationship (Drakare et al., 2006). We forward the hypothesis that in S. tenacissima steppes, this is probably due to the asymptotic decrease in patch capacity to create new microhabitats (i.e., the limits of their ability for ecological engineering), and not to depletion of the species pool. Clearly, the discussion on the spatial segregation of woody patches (i.e., patches as islands) and its effects on patch composition and community assembly deserve further attention. 2. How is a patch community structured? Dominant species can be much older than most accompanying species (L. De Soto, UNCROACH project†; Patón et al., 1998; Olano et al., 2011; Eugenio et al., 2012). Thus, patch community may organize in two phases. In the first phase, a dominant species gets established. Then, additional patch-forming species and a set of accompanying species may colonize under the canopy of the founding

† This dissertation has been developed under the framework of project UNCROACH (CGL2011-30581-C02-01; Dynamics of woody vegetation in dry and semi-arid landscapes facing global change. Implications for the provision of ecosystem services). This project was funded by the Spanish Ministry of Economy and Finance, and carried out at the University of Alicante from 2012 to 2015, by various researchers who are referenced along the text.

200

Discussion species. The probability of germination and establishment during the first phase may be very low under current pedo-climatic conditions and disturbance regime, as evidenced by the low density of recruits of most species (only R. lycioides showed relatively high recruitment rates and high density of adults), and numerous field observations (Herrera et al., 1994; García-Fayos and Verdú, 1998; Rey and Alcántara, 2000; Vilagrosa et al., 2001; Barberá et al., 2006). Still, we observed evidences of regular recruitment in all patch-forming species, including Q. coccifera (L. De Soto, UNCROACH project). The strong and positive relationship between the density of R. lycioides adults and mean air temperature, and the negative relationship between mean air temperature and the density of Q. coccifera and P. lentiscus, suggest that this factor represents a major filter for the recruitment of patch-forming species. In addition, seed dispersal may limit their establishment, particularly in areas that are distant from propagule sources. The dispersal of seeds of patch-forming species is mainly carried out by gravity, birds, small rodents and, to a lesser extent, big mammals as foxes and martens (Herrera, 1995; G. López and E. Rico, UNCROACH project). The lack of woody patches in recently abandoned steppes and the identity of the patch-forming species probably explain low colonization rates observed in some areas (V. Rolo, UNCROACH project). Patch-forming species are all resprouters and thus, both germination and resprouting may contribute to patch expansion. In addition, once a founding species gets established, other species may follow. Establishment must overcome several filters, which overall may be less stringent than those experimented by the founder. Filters may act sequentially along the successive establishment phases. First, seed rain may be limited by the availability of seed dispersers, and these in turn may be limited by the presence of safe sites as shelters and resting places. An ongoing study has found that disperser bird species visit woody patches and comparable physical structures, and disperse seeds of P. lentiscus and R. lycioides (A. Castillo, UNCROACH project). In most cases, dispersal may occur at relatively short distances, particularly for Q. coccifera. Indeed, K. Disante (U. Alicante) has found high levels of consanguinity in individuals of Q. coccifera from one of our study catchments (Porxa), higher than those of P. lentiscus in the same area. Seeds from dominant species (namely R. lycioides) are present in faeces of Vulpes vulpes, Martes foina and Meles meles (E. Rico and A. Bonet,

201

Discussion

UNCROACH project), and, albeit at a lower rate, they may be dispersed at longer distances. Second, species must overcome the germination filter. This requires high and sustained levels of moisture. Jaume Tormo (U. Alicante) found that P. lentiscus may be more sensitive than R. lycioides, Q. coccifera or S. tenacissima to low water availability. The former species may need long (>7 days) and abundant (>100 mm) rain events to germinate, which are quite uncommon in S. tenacissima steppes (García-Fayos and Verdú, 1998). S. tenacissima tussocks may favor germination as they create favorable microsites for seed germination of the dominant species R. lycioides and Q. coccifera, by increasing soil moisture and protecting seeds against predation (Barberá et al., 2006) Microenvironments favoring runoff concentration and water holding capacity, and hampering evaporation may increase the probability of germination of dominant and accompanying species. As shown in Chapter 5, litter accumulation is a major driver of germination. Water retention capacity in litter is positively related to germination rate, R. lycioides and P. lentiscus being the two species with the highest moisture retention capacity. However, the impact of litter on water availability depends on other factors such as hydrophobicity and short-term evaporation rate. Highly hydrophobic litter (such as that of Q. coccifera and Pinus halepensis), and high short-term evaporation rate (such as that of R. lycioides and Rosmarinus officinalis) limit litter ability to improve moisture conditions for germination. In addition to its role in the hydrological cycle, litter represents a physical obstacle for soil-seed contact (Rotundo and Aguiar, 2005). As described in Chapter 5, higher litter accumulation decreased germination rate. The effect was especially significant when seedlings were deposited on top of the litter layer, resembling natural conditions. Woody patches with thick litter layers such as those dominated by Q. coccifera and P. lentiscus may hinder germination of further or the same species. This limitation may be offset by segregating litter accumulation and seed germination in space or time. For example, seed germination will be more likely to occur early in patch life, when the litter layer is still sparse, or following disturbances that remove litter (e.g., wildlife trampling). In addition to the effect of litter, patches may affect seed germination by reducing irradiance, modifying the radiative spectrum and reducing evaporative demand 202

Discussion

(Sánchez-Gómez et al., 2006; Puerta-Piñero et al., 2007; Valladares and Niinemets, 2008). The third filter concerns seedling rooting and establishment. As mentioned above, patch periphery is the most suitable area for seed dispersal. Abiotic and biotic conditions are substantially different beyond the canopy projection area (Chapter 4). I found that seedling establishment was favored underneath woody patches, but the positive effect declined sharply in their periphery. I also found that survival depended on patch attributes, including soil fertility, which may be considered a part of the extended phenotype. The presence of the two most abundant accompanying species affected seedling survival both underneath the patches and in the northern side of their periphery. Specifically, high cover of S. tenacissima and low cover of B. retusum patches promoted establishment success. This is in agreement with studies showing that S. tenacissima improves the survival and physiological status of nearby P. lentiscus seedlings (Maestre et al., 2001; Maestre et al., 2003), and favors the survival of woody species (García-Fayos and Gasque, 2002). However, differences in other factors affecting seedling establishment in both microsites, underneath the patches and in their periphery, suggest that plants are subject to different stress types and intensities. Underneath the patches, plants experimented lower stress, and competition for water and nutrients may be offset by the benefits provided by low evaporative demand and high soil fertility (Chapter 4). Indeed, richness and cover of accompanying species were higher under the patches than in their periphery. Conversely, seedling establishment on the northern edge of the patches was driven by species composition and richness, and phylogenetic distance of the community, suggesting that biotic interactions play a major role in this microsite. Finally, on the southern edge of the patches, only abiotic factors modulated seedling survival. Rooting may also be affected by the presence of soil crusts. Litter accumulation under woody patches preclude the formation of smooth biological and physical crusts, dominated by lichens and cyanobacteria (Eldridge and Greene, 1994). They do not affect seed germination, but may hamper seedling rooting (Mendoza-Aguilar et al., 2014). Thus, woody patches may facilitate rooting by improving soil surface conditions for root penetration.

203

Discussion

Once seedlings get established, a new community is formed, and interspecific interaction may intensify. Changes in patch composition and size may be driven by various factors, including herbivory. For example, in S. tenacissima steppes, rabbit population may attack new seedlings (see the Materials and Methods section in Chapter 4; Soliveres et al., 2011). Patch size may also increase as a result of individual growth. Victor Rolo (Univ. Alicante) found that average increase in the surface area covered by the projection of the canopy was 0.24 m2 y-1. Information on the limits of patch growth is anecdotal. The largest patches found in our study cover 104 m2, but the relationship between age and canopy cover is linear, providing no clear signs of saturation of patch growth (V. Rolo, Univ. Alicante, unpublished data). 3. Emergent properties and community drivers Interspecific interactions within a patch occur at all life stages, from seeds to mature adults. The effect of individual interactions may be positive, negative or neutral. The outcome of the whole array of interactions taking place within a community may configure species co-occurrence patterns. Thus, by evaluating species co-occurrence, we obtain an integrated view of the entire community and the interactions therein. Most species in woody patches are generalists which co- occur with many other species. This is the case of the patch-forming species R. lycioides, which is also the most abundant resprouting species in these steppes, and the dominant species in 44% of the patches. A few species in the community only occur in the vicinity of particular species. For example, we only found Polygala rupestris, Viola arborescens and Teucrium buxifolium, under monospecific patches of J. oxycedrus. Patches dominated by the less abundant dominant species (O. lanceolata, J. oxycedrus and E. fragilis) were the ones holding the most specialized accompanying species. We may hypothesize that these patch-forming species create particular microclimatic conditions benefiting more demanding species. Thus, O. lanceolata only appears in the less arid areas. Similarly, soil organic matter composition under E. fragilis largely differs from the composition under other patch-forming species (J. Jordá, UNCROACH project). Plant communities are usually studied in terms of species composition and abundance, and this information is commonly integrated in diversity indices, as Shannon-Wiener’s information index, standardized diversity indices, estimations of sampling effort, and rarefaction curves (Sanders, 1968; Shannon 204

Discussion and Weaver, 1949; Gotelli and Colwell, 2001). In previous Chapters I described species composition and abundance, paying special attention to species interactions and their role in structuring plant communities. I assessed emergent properties of woody patches by using network theory, a suitable tool for analyzing community structure (Chapter 2). The generalist character of these communities was evidenced by high levels of network connectance and low levels of nestedness and specialization. These attributes are commonly related to robustness (Dunne et al., 2002; Kaiser-Bunbury et al., 2010). In addition, woody patches were poorly segregated into sub-communities, supporting the generalist character of these communities. I also evaluated the contribution of every species to its community, in terms of interaction within the entire network (Chapter 2). Specifically, I evaluated their connectance and specialization along the different networks, and discussed their role within the network. This analysis revealed that the two dominant species, R. lycioides and P. lentiscus, acted as hubs in their respective communities, with many species interacting with them. This observation is in agreement with the high abundance and low levels of specialization of the former species, and the large number of species found under patches dominated by the later. To our knowledge, this is the first time that network analysis has been used to evaluate plant composition and structure in physical well-defined communities, such as woody patches. Only Saiz and Alados (2011a, 2011b, 2014) evaluated plant communities by means of network analysis, employing species co-occurrence along lineal transects in slopes covered by S. tenacissima steppes. In addition, I used a bipartite approach, allowing for the study of the interactions between two sets of species. This approach provided complementary insights to the ones offered by ordination techniques (Non-metric Multi-Dimensional Scaling, Chapter 1), which showed that dominant species had a much larger influence on patch composition than accompanying species. For example, R. lycioides is the most abundant patch-forming species, and showed the highest rates of recruitment (Chapter 1). This species favors germination of both accompanying and dominant species (Chapter 5). As an adult, it acts as a network hub, interacting with many other species (Chapter 2). Finally, when R. lycioides is the dominant species in a patch, its nestedness increases and so does patch

205

Discussion complexity (Chapter 3). Similarly, two abundant accompanying species such as S. tenacissima and B. retusum, showed a high number of co-occurrences with all dominant species, and acted as potential hubs for the maintenance of network structure (Chapter 2). Yet, they played contrasted roles in community ordination (Chapter 1), and differed widely in their interaction with seedlings, by either favoring (S. tenacissima) or hampering (B. retusum) their establishment (Chapter 4). Many ecological network studies have focused on describing community attributes as nestedness, compartmentalization and specialization (Jordano et al., 2003; Montoya et al., 2006; Pascual and Dunne, 2006; Bascompte and Jordano, 2007; Estrada, 2007; Thébault and Fontaine, 2010). In contrast, drivers of community structure have received less attention (but see Memmot et al., 2007; Heleno et al., 2010; Devoto et al., 2012; Saiz and Alados, 2011a). In Chapter 3, I assessed the factors modulating patch emergent properties. I found that both exogenous factors (air temperature and rock cover) and endogenous factors (patch size, species identity and richness) explained community structure. These results have strong implications on community dynamics and their response to disturbances. Firstly, these results support the theory that R. lycioides and P. lentiscus largely affect species composition and promote community complexity by increasing nestedness, and overall link heterogeneity. This may be related to their ability to act as network hubs, interacting with many species and thus promoting link heterogeneity when other dominant species are present in the same patch. Secondly, the component of network complexity which is related to modularity was related to patch size rather than to dominant species, despite the strong relationship between these two variables. And thirdly, mean annual air temperature was a major driver of community structure, despite its apparently narrow range of variation. This reveals community sensitivity to climatic conditions, and the important consequences that small changes in climate could have on community composition and functioning (see below). Litter accumulation also affected community structure. In Chapter 5, we described a significant (and largely negative) effect of litter on seed germination. Conversely, in Chapter 4, we showed that litter accumulation had a positive effect on the performance of planted seedlings. Finally, litter was not included in models explaining network features. Ontogenic and sinecological factors may explain the

206

Discussion contrasted effects of litter on germination and seedling establishment. On one hand, different life stages may be constrained by different limiting factors. Our results suggest that litter may be more relevant for plant establishment during the germination process than once a seedling has germinated and rooted. On the other hand, in Chapter 5 we evaluated the effect of litter species separately, while in Chapter 4 we considered the effect of the whole community. Litter mixtures may differ in structure and impact on plant performance from mono-specific litter (Peterson and Facelli, 1992; Myster, 1994). Even if network analysis has mostly been used to evaluate mutualistic networks and food webs in Ecology, its potential for the analysis of plant co- occurrence networks has been underestimated. This is particularly true if we take into consideration the vast amount of data available on plant co-occurrence at different scales. The use of network analysis for the study of these communities may reveal emergent properties of plant communities, and may deepen our knowledge on the functional role of plant species. 4. Ecological role of woody patches Woody patches have the ability to modify their immediate environment and thus affect biotic and abiotic processes at the community and ecosystem levels. Colonization of woody species on grasslands has been considered an indicator of desertification, as it may negatively affect carbon storage, ecohydrological status, soil protection, forage production, species diversity, community stability, (Schlesinger et al., 1996; Jackson et al., 2002; Huxman et al, 2005; Jackson et al, 2005; Zarovali et al., 2007; Baez and Collins, 2008; Brudvig, 2010). Recent studies showed that this negative effect cannot be generalized, and contributed to provide a more complex and comprehensive view of this phenomenon and its consequences (Maestre et al., 2009b; Eldridge et al., 2011; Petrie et al., 2014). Through the present dissertation, I have described various dynamic factors in the patch-dominated space that are affected by the presence of woody patches, and may ultimately affect ecosystem functioning. Woody patches increased soil fertility through an increase in soil organic C and total N (Jackson et al., 2002; Chapter 4). This change was mostly restricted to the canopy projection area. Also, 15N enrichment was higher in plants thriving underneath the patches than outside them, suggesting rapid N cycling and

207

Discussion relatively high N losses underneath the patches, as nitrification, denitrification and ammonia volatilization promote soil 15N enrichment (Shearer et al., 1974; Peñuelas et al., 1999). Soil water content was lower under woody patches than in their surroundings (Chapter 4). This result was unanticipated, as one would expect that litter accumulation and canopy shade would create wetter conditions (Guevara- Escobar et al., 2007). Indeed, S. tenacissima tussocks and small shrubs commonly have positive or neutral effects on moisture availability compared to open areas (Maestre and Cortina, 2003; Cantón et al., 2004; Rey et al., 2011), but these studies also show that soil water content is very dynamic, and it may change depending on evaporative demand, amount and intensity of rainfall events, and plant water-use strategy. Other studies have found decreases in soil moisture content underneath woody patches compared to open areas (Breshears et al., 1997). Our results suggest that water uptake by woody vegetation under the patches, together with rainfall interception by the canopy and the litter may offset the increase in water infiltration and retention, and the reduction in evaporative demand. Other studies have emphasized the importance of water uptake to explain soil moisture dynamics under grass tussocks and shrubs in semiarid areas (Parker and Martin, 1952; Breshears et al., 1997; Maestre et al., 2003a). Our results suggest that shrub encroachment may reduce water reserves, and are in agreement with Bellot et al., (2001), who suggested that the increase in woody vegetation in semi-arid steppes reduces groundwater inputs. We may note, however, that the impact of woody encroachment on water availability is complex, cannot be easily ascertained, and predictions should be made with care (Huxman et al., 2005). Woody patches affect the presence and abundance of many taxa, including plants. Indeed, facilitation probably favored the presence of Tertiary flora in the Mediterranean, and may promote phylogenetic diversity (Valiente- Banuet et al., 2006; Valiente-Banuet and Verdú, 2007. Plant-plant interactions may be positive, negative or neutral, depending on biotic and abiotic factors, including the identity of the involved species (Maestre et al., 2009a; Chapter 4). As shown in Chapter 4, the increase in soil fertility may positively affect the performance of other plant species. This was supported with the higher foliar N concentration found in seedlings planted underneath the patches compared to

208

Discussion seedlings planted on their northern edge. Conversely, the reduction in soil water content underneath the canopy may promote competition and water stress. Nevertheless, the relatively low water use efficiency levels in plants growing underneath the patches suggest that water stress was still lower underneath the patches than outside them. In addition, shade tolerant species may thrive underneath patches with higher probability than in open areas. Although solar radiation was not selected as a factor to explain seedling survival (Chapter 4), its effects may occur in an earlier stage of plant development (Lockhart, 1961; Valiente-Banuet and Ezcurra, 1991). The effect of patches on other vegetation is governed by the same factors that affect the different plant stages for patch formation and extension described above. Specific interactions have been described between vascular plants and soil organisms such as lichens, cyanobacteria, mosses and fungi (Zaady et al., 1997; DeFalco et al., 2001; Maestre and Cortina, 2002; Escudero et al., 2007). The distribution of biological crusts is affected by the presence of woody patches (Dougill and Thomas, 2004), which may have further consequences on ecosystem functioning (Delgado-Baquerizo et al., 2010; Mendoza-Aguilar et al., 2014). Woody patches may also act as shelter, resting and feeding places for animals. The fruits of most dominant species in woody patches are eaten by birds, rodents and small mammals (Pausas et al., 2006; Zapata et al., 2014). Insects may also benefit from fruits. Stipa tenacissima fruits are predated by ants and birds (Haase et al., 1995; Schöning et al., 2004; Belda et al., 2010), acorns are predated by mammals and birds (Herrera, 1995; Santos and Tellería, 1997). Also, the annelid community may be more abundant in the vicinity of woody vegetation (González and Zou, 1999; García-Pausas et al., 2004). The presence of fauna underneath and around patches may affect other processes such as pollination, seed dispersal, soil structure, soil fertility and water infiltration as a result of feed- back interactions (Verdú and García-Fayos, 1996; Bronick and Lal, 2005; Pausas et al., 2006). According to our results, woody patches create suitable microsites for plants. These may favor the establishment of further plant species, as long as their water requirements are not too high and the litter layer is not too thick.

209

Discussion

5. Restrospective view of woody patches. Insights on the response to climate change As it has been reported in the literature, shrub cover and density have increased over the last 150 years in drylands worldwide (Van Auken, 2009). The semiarid southeast Spain is not an exception, and an increase in shrub cover has been reported in S. tenacissima steppes, especially since 1985 (Alados et. al, 2004). Indeed, an ongoing research carried out by V. Rolo (University of Alicante) in the same 15 catchments studied in the present dissertation, has found that shrub density has significantly increased over the last decades, doubling on average the density estimated in 1956. Similarly, patch canopy projected area has significantly increased over the same period, suggesting continuous patch growing (V. Rolo, unpublished data). Climatic conditions in the study area have been particularly harsh over the last decades, and will likely change in the near future. Global warming projections for Europe predict an increase of temperature both in winter (2.5-3ºC) and summer (>4ºC), in relation to the reference period 1980-1999 (Christensen et al., 2007). In addition, over the last 60 years, there has been a significant reduction in summer precipitation and, to a lesser extent, in spring and autumn precipitation (Machado et. al, 2011). This decreasing trend in precipitation is projected to continue in southeastern Spain at a rate of 0-20% by the end of the 21st Century (deviation from reference period 1961-1990; AEMET, 2008). Considering recent modifications in land use patterns and predicted changes in climatic conditions, S. tenacissima steppes will probably show substantial alterations in vegetation cover and ecosystem functioning. Considering the relationship between mean annual temperature and precipitation, and dominant species cover (Chapter 1 and 3), I hypothesize that patches dominated by R. lycioides and E. fragilis will increase at the expense of those dominated by Q. coccifera and J. oxycedrus. Leafless species, such as E. fragilis, are especially successful in harsh environments due to their efficiency in obtaining resources while minimizing water loses, and their ability for rooting deep and fast (Padilla et. al, 2009). Thus, patches dominated by P. lentiscus, Q. coccifera, O. lanceolata and J. oxycedrus will face difficulties to increase in number and size. It must be noted that these species evolved under Tertiary, semi-tropical conditions (Palmarev, 1989; Herrera, 1992; Verdú et al., 2003), and 210

Discussion plasticity plus facilitative interactions allowed them to thrive under current Mediterranean climate (Jump and Peñuelas, 2005; Valiente-Banuet et al., 2006). Patches dominated by these species may be confined to favorable microsites of lower temperature and higher moisture availability, such as lower parts of the catchments, north facing slopes, deep soils and the under-canopy of larger individuals. This trend is supported by observations on current recruitment of dominant species. Rhamnus lycioides showed the highest seedling density in our study areas, followed by E. fragilis (Chapter 1). However, in relation with the abundance of adult individuals, J. oxycedrus also had high recruitment. Taking into account that patches dominated by Q. coccifera and P. lentiscus, were the most abundant after R. lycioides, we may bring forward the following hypothesis: (1) recruits measured in the field had a stem perimeter lower than 5 cm, which corresponds to individuals under 10 years old (L. De Soto, UNCROACH project). Thus, if we take recruit density as an estimation of recent recruitment rates, we may conclude that recruitment rates of Q. coccifera and P. lentiscus after abandonment were higher than nowadays. (2) We may also conclude that past and present recruitment rates of R. lycioides were high. (3) Ephedra fragilis populations may be expanding, as present recruitment rates are relatively high, compared to its current abundance. (4) Juniperus oxycedrus populations may be relatively stable because population turnover rate was high. (5) Osyris lanceolata was not abundant in the past and is not widespread now; its recruitment rate is limited and suggests no further expansion of this species. Hypotheses (1) and (2) should be validated in an ongoing study carried out by L. De Soto (U. Coimbra) and K. Disante (U. Alicante) where patch dynamics will be studied in detail in the experimental catchment Porxa using dendroecological and phylogenetic techniques. The change in abundance and patch composition may have important implications on ecosystem biodiversity and community complexity, affecting plant and animal species. Studies on the effects of climate change on biodiversity predict a high species turnover over the next 100 years in Europe (Bakkenes et al., 2002; Benito-Garzón et al., 2008; Alkemade et al., 2011). This effect is especially dramatic in semiarid areas in southeast Spain, as those authors pointed out that there is not an existing pool of species able to deal with forecasted climate

211

Discussion conditions for the next decades. This fact may have strong consequences on community impoverishment and the expansion of alien species. 6. Woody patches and steppe management Our results have profound implications on the management of S. tenacissima steppes. As a general rule, any measure aimed at favoring the conservation and expansion of patch-forming species (i.e., dominant species) should be recommended as a way to promote biodiversity and functionality. Given that R. lycioides and P. lentiscus co-occur with many other species, acting as network and module hubs (Chapter 2), it may be worth to focus on these species. On the other hand, highly connected species are suitable candidates for restoration projects aimed at increasing biodiversity (Pocock et al., 2012). In addition, the two dominant species, together with S. tenacissima are drivers of community complexity as they all increased network nestedness (Chapter 3). Yet, I have shown that R. lycioides recruits successfully and may thrive under the severe conditions of these semiarid steppes. So does S. tenacissima, which is the most abundant species in theses steppes. Conversely, the low abundance of P. lentiscus and its low recruitment ability contrasts with its key role in the maintenance of community biodiversity and complexity. Measures aimed at promoting P. lentiscus establishment and survival may be encouraged to preserve the diversity of woody patches. It is important to note here that, according to the models of species distributions under predicted climate for Europe (Bekkenes et al., 2002), P. lentiscus will hardly thrive under drier and warmer conditions and its distribution may be restricted in semiarid areas. In consequence, this may increase its value as community-forming species. Conversely, R. lycioides populations seem to not have problems under future climatic conditions, and its inclusion in restoration programs should be only restricted to specific sites or places, for example, where propagule source is far or when the need to quickly incorporate patch-forming species to the community is urgent. Thus, taking into account climate change scenarios for southeast Spain, one may expect that only R. lycioides and E. fragilis will successfully thrive under future climatic conditions. However, in order to keep inter-patch diversity, the recruitment of other species is desirable. Promoting spatial heterogeneity and

212

Discussion creating favorable microenvironments for plant establishment and growth may achieve this suggestion. Various techniques and tools aimed at creating suitable microenvironments for plant establishment have been described (Cortina et al., 2012). For example, the creation of microcatchments to increase water, nutrients and soil availability (Boeken and Shachk, 1994; Whisenant et al., 1995), the use of mulches to reduce soil temperature, attenuate evaporation and increase water and propagule availability (Ludwig and Tongway, 1996; Tongway and Ludwig, 1996), the use of various devices to capture and store available water (Bainbridge, 2007; Valiente et al., 2011; Vasudevan et al., 2011; Valdecantos et al., 2014), the establishment of resting sites for birds (Wunderlee, 1997; Bonet, 2004) and the use of the nurse plants (Maestre et al., 2001; Castro et al., 2004; Gómez-Aparicio et al., 2004). With regard to the use of preexisting vegetation to enhance seedling success, in Chapter 4 I have emphasized some aspects of the identity of the interacting species that increase seedling survival. Thus, the selection of patches with high cover of S. tenaccisima and low cover of B. retusum under their canopies, patches formed by phylogenetically distant species as nurses for new seedlings, or patches that are relatively poor in accompanying species, would increase seedling success. However, differences in the establishment of each patch-forming species are expected (Vilagrosa et al., 2003; Trubat et al., 2008, 2011) and the specific factors controlling their establishment must be tested. Furthermore, given the possible species extinction forecasted for the next decades, the lack of turnover species able to deal with future climate and the increasing likelihood of alien species input (Bakkenes et al., 2002; Benito-Garzón et al., 2008; Alkemade et al., 2011), the selection of species for restoration programs becomes a really challenging task. To choose an adequate pool of species should be made carefully. Further research must be carried out to evaluate the appropriate techniques, species and sites to match the restoration objectives and effectiveness of management actions. To our knowledge, this is the first study focusing on the composition and functioning of woody patch communities in semiarid steppes. Research compiled in this dissertation provides a global view of the dynamics of these communities and their main drivers. Throughout this work, I have evaluated various processes affecting different stages of plant lifespan, and described how community assemblage is made. This information is crucial for the management of S.

213

Discussion tenacissima steppes under a changing socio-economic and biophysical scenario. This is also a pioneer work on the use of network theory for the analysis of plant communities.

214

Conclusions

CONCLUSIONS Along the chapters of the present dissertation, I have mentioned the conclusions of each study. Here I present the global conclusions of the thesis. Species in woody vegetation patches of semiarid Stipa tenacissima steppes are organized in two levels, according to their morpho-functional traits and their ability to form patches: dominant (large patch-forming resprouting species) and accompanying species (smaller species). Patch composition is mainly determined by the most dominant species and their distribution is related to mean annual temperature and precipitation. According to climatic predictions, patches dominated by R. lyciodes and E. fragilis will likely increase their dominance in future decades at the expense of patches dominated by Q. coccifera and J. oxycedrus. When considering woody patches as communities of interactions, most species are generalists, as they are highly connected to each other and the structure of their interactions is barely nested. In addition, these communities usually form 2-4 subcommunities of species more connected between them than to other subcommunities.

The most generalist species in patch communities are Rhamnus lycioides, Stipa tenacissima and Brachypodium retusum, but only R. lycioides and Pistacia lentiscus acted as network and module hubs, which makes them key species in woody patch communities. Emergent properties of co-occurrence networks of woody patch communities were modulated by endogenous and exogenous factors. Mean annual temperature affected the three indices of network structure, and only nestedness was affected by patch composition. High rock cover and the presence of patches dominated by Rhamnus lycioides and Pistacia lentiscus increased community structure by increasing modularity and nestedness, respectively. Increased patch size was a determinant of low-structured communities.

The modification of the microenvironment by woody patches is restricted to the limits of the canopy projection area. Drivers of seedling survival and the type and intensity of stress differ within a patch; these differences should be

215 Conclusions

taken into account when evaluating the outcome of ecological interactions in woody patches. Woody patches had positive effects on the incorporation of new individuals in S. tenacissima steppes, and this effect was mainly determined by the cover and composition of the community of accompanying species, litter depth and phylogenetic distance between dominant and colonizing species. Litter of woody species exerts a predominantly negative effect on the germination of dominant and accompanying species, either when seeds are located on or under the litter layer. This effect depends on litter accumulation which, in turn, depends on dominant species. Thus, R. lycioides, which accumulates low amounts of litter, has neutral or even positive effects on germination. Our results have strong implications on steppe composition and function, and on its capacity to provide goods and services. They provide essential information for the management of Stipa tenacissima steppes.

216

REFERENCES

References

Abanades, J.C., Cuadrat, J.M., de Castro, M., Fernández, F., Gallastegui, C., Garrote, L., Jiménez, L., Juliá, R., Losada, I., Monzón, A., Moreno, J.M., Pérez, J.I., Ruiz, V., Sanz, M.J. and Vallejo, R. 2007. El Cambio Climático en España. Estado de situación. Madrid: Ministerio de Medio Ambiente. AEMET.

Aerts, R. 1996. Nutrient resorption from senescing leaves of perennials: are there general patterns? Journal of Ecology 84: 597-608. Aerts, R. 1997. Climate, Leaf litter chemistry and leaf litter decomposition in terrestrial ecosystems: a triangular relationship. Oikos 79(3): 439-449. Aguiar, M.R. and Sala, O.E. 1999. Patch structure, dynamics and implications for the functioning of arid ecosystems. Trends in Ecology and Evolution 14: 273-277. Alados, C.L., Pueyo, Y., Barrantes, J., Escos, J., Giner, L. and Robles, A.B. 2004. Variations in landscape patterns and vegetation cover between 1957 and 1994 in a semiarid Mediterranean ecosystem. Landscape Ecology 19: 543-559. Alados, C. L., Gotor, P., Ballester, P., Navas, D., Escos, J. M., Navarro, T., and Cabezudo, B. 2006. Association between competition and facilitation processes and vegetation spatial patterns in alpha steppes. Biological Journal of the Linnean Society, 87(1): 103-113. Albaladejo, J., Martínez-Mena, M., Roldan, A. and Castillo, V. 1998. Soil degradation and desertification induced by vegetation removal in a semiarid environment. Soil Use and Management 14: 1-5. Alkemade, R., Bakkenes, M. and Eickhout, B. 2011. Towards a general relationship between climate change and biodiversity: an example for plant species in Europe. Regional Environmental Change, 11(1): 143-150. Almagro, M. and Martínez-Mena, M. 2012. Exploring short-term leaf-litter decomposition dynamics in a Mediterranean ecosystem: dependence on litter type and site conditions. Plant and soil, 358(1-2): 323-335. Almeida-Neto, M., Guimarães P.R. Jr. and Lewinsohn, T.M. 2007. On nestedness analyses: rethinking matrix temperature and anti-nestedness. Oikos 116: 716– 722.

219

References

Alrababah, MA., Tadros, MJ., Samarah, NH. and Ghosheh, H. 2009. Allelopathic effects of Pinus halepensis and Quercus coccifera on the germination of Mediterranean crop seeds. New Forests 38: 261-272. Altieri, A.H., van Wesenbeeck, B.K., Bertness, M.D. and Silliman, B.R., 2010. Facilitation cascade drives positive relationship between native biodiversity and invasion success. Ecology 91: 1269-1275. Amat, B., Cortina, J. and Zubcoff, J.J. 2015. Community attributes determine facilitation potential in a semi-arid steppe. Perspectives in Plant Ecology, Evolution and Systematics, 17: 24-33. Anthelme, F., Buendia, B., Mazoyer , C. and Dangles, O. 2012. Unexpected mechanisms sustain the stress gradient hypothesis in a tropical alpine environment. J. Veg. Sci. 23: 62–72. Armas, C. and Pugnaire, F. I. 2005. Plant interactions govern population dynamics in a semi‐arid plant community. Journal of Ecology, 93(5): 978-989. Armas, C., Ordiales, R. and Pugnaire, F. 2004. Measuring plant interactions: a new comparative index. Ecology 85(10): 2682-2686. Armesto, J.J., Vidiella, P. and Gutiérrez, J.R. 1993 Plant communities of the fog- free coastal desert of Chile: plant strategies in a fluctuating environment. Revista Chilena de Historia Natural, 66: 271–282. Asner, G.P. and Martin, R.E. 2004. Biogeochemistry of desertification and woody encroachment in grazing systems. Ecosystems and Land Use Change. Geophysical Monograph Series 13: 99-116. Austin, A.T. and Vitousek, P.M. 2000. Precipitation, decomposition and litter decomposability of Metrosideros polymorpha in native forests on Hawai’i. Journal of Ecology 88: 129-138. Austin, A.T. and Vivanco, L. 2006. Plant litter decomposition in a semi-arid ecosystem controlled by photodegradation. Nature 442: 555-558. Austin, A.T. and Ballaré, C.L. 2010. Dual role of lignin in plant litter decomposition in terrestrial ecosystems. PNAS 107(10): 4618-4622.

220

References

Baez, S. and Collins, S.L. 2008. Shrub invasion decreases diversity and alters community stability in northern Chihuahuan desert plant communities, PLoS One 3(6), e2332. Bainbridge, D. A. 2007. A guide for desert and dryland restoration: new hope for arid lands. Island Press, Washington, DC. Bakkenes, M., Alkemade, J. R. M., Ihle, F., Leemans, R. and Latour, J. B. 2002. Assessing effects of forecasted climate change on the diversity and distribution of European higher plants for 2050. Global Change Biology, 8(4): 390-407. Balvanera, P., Daily, G.C., Ehrlich, P.R., Ricketts, T.H., Bailey, S.A., Kark, S., Kremen, C. and Pereira, H. 2001. Conserving biodiversity and ecosystem services. Science 291: 2047. Barber, A., Cabrera, M.R. and Guardiola, I. 1997. Sobre la cultura de l’espart al territori valencià. Bancaixa, Valencia. Barberá, G.G., Navarro-Cano, J.A. and Castillo, V.M. 2006. Seedling recruitment in a semiarid steppe: The role of microsite and post-dispersal seed predation. Journal of Arid Environments 67: 701-714. Bascompte, J. and Jordano, P. 2007. Plant-animal mutualistic networks: the architecture of biodiversity. Annual Review of Ecology, Evolution and Systematics, 38: 567-593. Bascompte, J., Jordano, P., Melián, C.J. and Olesen, J.M. 2003. The nested assembly of plant–animal mutualistic networks. PNAS 100: 9383-9387. Bastolla, U., Fortuna, M., Pascual-Garcia, A., Ferrera, A., Luque, B. and Bascompte, J. 2009. The architecture of mutualistic networks minimizes competition and increases biodiversity. Nature 458: 1018-1020. Batagelj, V., Mrvar, A. 1998. Pajek- Program for Large Network Analysis. Connections 21: 47-57. Beckage, B. and Clark, J.S., 2003. Seedling survival and growth of three forest tree species: the role of spatial heterogeneity. Ecology 84: 1849-1861.

221

References

Beckage, B., Clark, J.S., Clinton, B.D. and Haines, B.L. 2000. A long-term study of tree seedling recruitment in southern Appalachian forests: the effects of canopy gaps and shrub understories. Canadian Journal of Forest Research 30: 1617-1631. Belda, A., Martínez-Pérez, J. E., Martín, C., Peiró, V. and Seva, E. 2010. Plants Used to Capture and Sustain Wild Finches (Fringillidae) in Southeast Spain?. Economic Botany, 64(4): 367-373. Bellot, J., Bonet, A., Sánchez, J.R. and Chirino, E. 2001. Likely effects of land use changes on the runoff and aquifer recharge in a semiarid landscape using a hydrological model. Landscape and Urban Planning 55: 41-53 Benito-Garzón, M., Sánchez-de-Dios, R. and Sainz-Ollero, H. 2008. Effects of climate change on the distribution of Iberian tree species. Applied Vegetation Science, 11(2): 169-178. Benkobi, L., Trlica, M.J. and Smith, J.L. 1993. Soil loss as affected by different combinations of surface litter and rock. Journal of Environmental Quality 22: 657- 661. Bertness, M.D. and Callaway, R.M. 1994. Positive interactions in communities. Trends in Ecology and Evolution 9: 191-193. Blank, L., and Carmel, Y. 2012. Woody vegetation patch types affect herbaceous species richness and composition in a Mediterranean ecosystem. Community Ecology, 13(1): 72-81. Blondel, V.D., Guillaume, J.L., Lambiotte, R. and Lefebvre, E. 2008. Fast unfolding of communities in large networks. Journal of Statistical Mechanics: Theory and Experiment 10, P10008. Blüthgen, N., Menzel, F. and Blüthgen, N. 2006. Measuring specialization in species interaction networks. Ecology 6: 9. Bochet, E., P. Garcia-Fayos, Alborch, B. and Tormo, J. 2007. Soil water availability effects on seed germination account for species segregation in semiarid roadslopes. Plant and Soil 295(1): 179-191. Boeken, B. and Orenstein, D. 2001. The Effect of Plant Litter on Ecosystem Properties in a Mediterranean Semi-Arid Shrubland. Journal of Vegetation Science 12(6): 825-832.

222

References

Boeken, B. and Shachak, M. 1994. Desert plant communities in human-made patches-Implications for management. Ecological Applications 4: 702-716. Bolòs, O. and Vigo, J. 2001. Flora dels Països Catalans. Vol. IV. Barcino. Barcelona, 750 pp. Bonanomi, G., Sicurezza, M.G., Caporaso, S., Esposito, A. and Mazzoleni, S. 2006. Phytotoxicity dynamics of decaying plant materials. New Phytologist 169: 571- 578. Bonet, A. 2004. Secondary succession of semi-arid Mediterranean old-fields in south-eastern Spain: insights for conservation and restoration of degraded lands. Journal of Arid Environments 56: 213-233. Bowers, J. E., Webb, R. H. and Rondeau, R. J. 1995. Longevity, recruitment and mortality of desert plants in Grand Canyon, Arizona, USA. Journal of Vegetation Science, 6(4): 551-564. Brearley, F.Q., Press, M.C. and Scholes, J.D. 2003. Nutrients obtained from leaf litter can improve the growth of dipterocarp seedlings. New Phytologist 160: 101- 110. Breshears, D.D., Rich, P.M., Barnes, F.J. and Campbell, K. 1997. Overstory-imposed heterogeneity in solar radiation and soil moisture in a semiarid woodland. Ecological Applications, 7(4): 1201-1215. Briggs, J.M., Knapp, A.K., Blair, J.M., Heisler, J.L., Hoch, G.A., Lett, M.S. and McCarron, J.K. 2005. An Ecosystem in Transition: Causes and Consequences of the Conversion of Mesic Grassland to Shrubland. BioScience 55: 243-254. Bronick, C. J. and Lal, R. 2005. Soil structure and management: a review. Geoderma, 124(1): 3-22. Brooker, R.W., Maestre, F.T., Callaway, R.M., Lortie, C.L., Cavieres, L.A., Kunstler, G., Liancourt, P., Tielbörger, K., Travis, J.M.J., Anthelme, F., Armas, C., Coll, L., Corcket, E., Delzon, S., Forey, E., Kikvidze, Z., Olofsson, J., Pugnaire, F., Quiroz, C.L., Saccone, P., Schiffers, K., Seifan, M., Touzard, B. and Michalet, R., 2008. Facilitation in plant communities: the past, the present, and the future. J. Ecol. 96: 18-34.

223

References

Brudwig, L.A. 2010. Woody encroachment removal from Midwestern oak Savannas alters understory diversity across space and time. Restoration Ecology 18: 74-84. Burgan, R.E. and Rothermel, R.C. 1984. BEHAVE: Fire behavior prediction and fuel modeling system-FUEL subsystem. United States Department of Agriculture, Forest Service, General Technical Report INT-167. Intermountain Forest and Range Experiment Station, Ogden, Utah. 126 pp. Burnham, K.P. and Anderson, D.R., 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. New York, Springer. Cabido, M., González, C., Acosta, A. and Díaz, S. 1993 Vegetation changes along a precipitation gradient in Central Argentina. Plant Ecology, 109: 5–14. Cable, J.M., Ogle, K. Tyler, A.P. Pavao-Zuckerman, M.A. and Huxman T.E. 2009. Woody plant encroachment impacts on soil carbon and microbial processes: Results from a hierarchical Bayesian analysis of soil incubation data. Plant Soil 320: 153-167. Callaway, R.M. and Walker, L.R. 1997. Competition and facilitation: a synthetic approach to interactions in plant communities. Ecology 78: 1958-1965. Callaway, R.M. 1995. Positive interactions among plants. Bot. Rev. 61: 306-349. Callaway, R.M. 2007. Positive interactions and interdependence in plant communities. Springer, Dordrecht, The Netherlands. Callaway, R.M., Nadkarni, N.M. and Mahall, B.E. 1991. Facilitation and interference of Quercus douglasii on understorey productivity in central California. Ecology 72: 1484-1499. Cantón, Y., Solé-Benet, A. and Domingo, F. 2004. Temporal and spatial patterns of soil moisture in semiarid badlands of SE Spain. Journal of Hydrology, 285(1): 199- 214. Cañellas, I. and San Miguel, A. 1998. Litter fall and nutrient turnover in Kermes oak (Quercus coccifera L.) shrublands in Valencia (eastern Spain). Annales des Sciences Forestières 55: 589-597.

224

References

Capitanio, R. and Carcaillet, C. 2008. Post-fire Mediterranean vegetation dynamics and diversity: a discussion of succession models. Forest Ecology and Management, 255(3): 431-439. Caravaca, F., Figueroa, D., Barea, J.M., Azcón-Aguilar, C., Palenzuela, J. and Roldán, A. 2003. The role of relict vegetation in maintaining physical, chemical, and biological properties in an abandoned Stipa-grass agroecosystem. Arid Land Research and Management 17: 103-111. Case, T.J. 1991. Invasion resistance, species build-up and community collapse in metapopulation models with interspecies competition. Biol. J. Linn. Soc. 42: 239- 266. Castillo, J.P., Verdú, M. and Valiente-Banuet, A., 2010. Neighborhood phylodiversity affects plant performance. Ecology 91: 3656-3663. Castro-Urgal, R., Tur, C., Albrecht, M. and Traveset, A. 2012. How different link weights affect the structure of quantitative flower–visitation networks. Basic Appl. Ecol. 13: 500-508. Castro, J., Zamora, R., Hódar, J. A. and Gómez, J. M. 2002. Use of shrubs as nurse plants: a new technique for reforestation in Mediterranean mountains. Restoration Ecology, 10(2): 297-305. Castro, J., Zamora, R., Hódar, J.A., Gómez, J.M. and Gómez-Aparicio, L., 2004. Benefits of using shrubs as nurse plants for reforestation in Mediterranean mountains: a 4-year study. Restor. Ecol. 12: 352-358. Cavieres, L. and Badano, E.I., 2009. Do facilitative interactions increase species richness at the entire community level? J. Ecol. 97: 1181–1191. Cerván, C.M. and Pardo, F. 1997. Dispersión de semillas de retama (Retama spaherocarpa, L. Boiss) por el conejo (Oryctolagus cuniculus) en el centro de España. Doñana Acta Vertebrata 21: 143-154. Chagnon, P.L., Bradley, R.L. and Klironomos, J.N. 2012. Using ecological network theory to evaluate the causes and consequences of arbuscular mycorrhizal community structure. New Phytologist, 194(2): 307-312. Chapin, F.S., Matson, P.A. and Mooney, H.A. 2002. Principles of terrestrial ecosystem ecology. Springer, Nueva York. USA.

225

References

Chaves, N. and Escudero, J.C. 1997. Allelopathic effect of Cistus ladanifer on seed germination. Functional Ecology 11: 432-440. Christensen, J.H., Hewitson, B., Busuioc, A., Chen, A., Gao, X., Held, I., Jones, R., Kolli, R.K., Kwon, W.-T., Laprise, R., Rueda, V Regional climate projections. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 847-940. Christensen, N.L. and Muller, C.H. 1975. Relative importance of factors controlling germination and seedling survival in Adenostoma chaparral. American Midland Naturalist 93: 71-78. Cintra, R. 1997. Leaf litter effects on seed and seedling predation of the palm Astrocaryum murumuru and the legume tree Diptryx micrantha in Amazonian forest. Journal of Tropical Ecology 13: 709-725. Cortina, J. and Maestre, F.T. 2005. Plant effects on soils in drylands. Implications for community dynamics and dryland restoration. En: Binkley, D. y Menyailo, O. (Eds.) Tree Species Effects on Soils: Implications for Global Change. NATO Science Series Kluwer Academic Publishers, Dordrecht. Cortina, J. and Vallejo, V.R. 1994. Effects of clearfelling on forest floor accumulation and litter decomposition in a radiata pine plantation. Forest Ecology Management 70: 299-310. Cortina, J., Amat, B., Castillo, V., Fuentes, D., Maestre, F.T., Padilla, F.M., and Rojo, L. 2011. The restoration of vegetation cover in the semi-arid Iberian southeast. Journal of Arid Environments, 75(12): 1377-1384. Cortina, J., Bellot, J., Vilagrosa, A., Caturla, R.N., Maestre, F., Rubio, E., Martínez, J.M. and Bonet, A. 2004. Restauración en semiárido. In: Vallejo, R.M. (Ed.) Restauración de Ecosistemas Degradados. Valencia, España: CEAM-CMA Generalitat Valenciana. Cortina, J., Maestre, F. T. and Ramírez, D. 2009. Innovations in semiarid restoration. The case of Stipa tenacissima L. grass steppes. In: S. Bautista, J.

226

References

Aronson and R. Vallejo (ed.). Land Restoration to Combat Desertification: Innovative Approaches, Quality Control and Project Evaluation. Fundación CEAM. Valencia, 121-144. Cortina, J., Martín, N., Maestre, F.T. and Bautista, S. 2010. Disturbance of the biological soil crusts and performance of Stipa tenacissima in a semiarid Mediterranean steppe. Plant and Soil 334: 311-322. Cortina, J., Navarro, R.M. and Del Campo, A.D. 2006. Evaluación del éxito de la reintroducción de especies leñosas en ambientes Mediterráneos. En: Cortina, J., Peñuelas, J.L., Puértolas, J., Vilagrosa, A., y Savé, R. (Coord.). Calidad de planta forestal para la restauración en ambientes Mediterráneos. Estado actual de conocimientos. Organismo Autónomo Parques Nacionales. Ministerio de Medio Ambiente. Madrid. Cortina, J., Ruiz-Mirazo, J., Amat, B., Amghar, F., Bautista, S., Chirino, E., Derak, M., Fuentes, D., Maestre, F.T., Valdecantos, A. and Vilagrosa, A. 2012. Bases para la restauración ecológica de espartales. UICN, Gland, Suiza y Málaga, España. VI + 26 p. ISBN: 978-2-8317-1566-7. Costa, M. 1973. Datos ecológicos y fitosociológicos sobre los espartales de la provincia de Madrid. Anales del Instituto Botánico Cavanilles 30: 233-255. Costa, M. 1999. La Vegetación y el Paisaje en las Tierras Valencianas. Editorial Rueda. Madrid. Covington, W.W. 1981. Changes in forest floor organic matter and nutrient content following clear cutting in northern hardwoods. Ecology 62(1): 41-48. Cuesta, B., Villar-Salvador, P., Puértolas, J., Rey-Benayas, J.M. and Michalet, R. 2010. Facilitation of Quercus ilex in Mediterranean shrubland is explained by both direct and indirect interactions mediated by herbs. J Ecol 98: 687–96. Dames, J.F., Scholes, M.C. and Straker, C.J. 1998. Litter production and accumulation in Pinus patula plantations of the Mpumalanga Province, South Africa. Plant and Soil 203: 183-190. Dawkins, R. 1982. The extended phenotype: the gene as the unit of selection. W.H. Freeman, Oxford.

227

References

De Soto, L., Olano, J.M., Rozas, V. and De la Cruz, M. 2010. Release of Juniperus thurifera woodlands from herbivore-mediated arrested succession in Spain Applied Vegetation Science 13: 15-25. Debussche, M., Escarré, J. and Lepart, J. 1982. Ornithochory and plant succession in Mediterranean abandoned orchards. Vegetatio 48: 255-266. DeFalco, L.A., Detling, J.K., Richard Tracy, C. and Warren, S.D. 2001. Physiological variation among native and exotic winter annual plants associated with microbiotic crusts in the Mojave Desert. Plant and Soil 234: 1-14 Dekker, L.W. and Ritsema, C.J. 1994. How water moves in a water repellent sandy soil. I. Potential and actual water repellency. Water Resources Research 30(9): 2507-2517. Del Arco, J.M., Escudero, A. and Garrido, M.V. 1991. Effects of site characteristics on nitrogen retranslocation from senescing leaves. Ecology 72: 701-708. Delgado-Baquerizo, M., Castillo-Monroy, A.P., Maestre, F.T., and Gallardo, A. 2010. Plants and biological soil crusts modulate the dominance of N forms in a semi-arid grassland. Soil Biology and Biochemistry, 42(2): 376-378. Delibes-Mateos, M., Delibes, M., Ferreras, P. and Villafuerte, R. 2008. The key role of European rabbits in the conservation of the western Mediterranean basin hotspot. Conservation Biology 22: 1106-1117. Delitti, W., Ferran, A., Trabaud, L. and Vallejo, V.R. 2005. Effects of fire recurrence in Quercus coccifera L. shrublands of the Valencia Region (Spain): I. plant composition and productivity. Plant Ecology, 177(1): 57-70. Dellafiore, C.M., Muñoz, S. and Gallego, J.B. 2006. Rabbits (Oryctolagus cuniculus) as dispersers of Retama monosperma seeds in a coastal dune system. EcoScience 13: 5-10 Derak, M. and Cortina, J. 2014. Multi-criteria participative evaluation of Pinus halepensis plantations in a semiarid area of southeast Spain. Ecological Indicators 43: 56-68. Descheemaeker, K., Muys, B., Nyssen, J., Poesen, J., Raes, D., Haile, M. and Deckers, J. 2006. Litter production and organic matter accumulation in exclosures of the Tigray highlands, Ethiopía. Forest Ecology and Management 233: 21-35.

228

References

Devoto, M., Bailey, S. Craze, P. and Memmott, J.. 2012. Understanding and planning ecological restoration of plant – pollinator networks. Ecology Letters 15: 319-328. Doblas-Miranda, E., Sánchez-Piñero, P. and González-Megías, A. 2009. Different microhabitats affect soil macroinvertebrate assemblages in a Mediterranean arid ecosystem Applied Soil Ecology 41: 329-335. Donath, T.W. and Eckstein, R.L. 2008. Grass and oak litter exert different effects on seedling emergence of herbaceous perennials from grasslands and woodlands. Journal of Ecology 96: 272–280 Dormann, C.F. and Strauss, R. 2014. A method for detecting modules in quantitative bipartite networks. Methods in Ecology and Evolution 5: 90-98. Dougill, A. J. and Thomas, A.D. 2004. Kalahari sand soils: spatial heterogeneity, biological soil crusts and land degradation. Land Degradation & Development, 15(3): 233-242. Drakare, S., Lennon, J. J. and Hillebrand, H. 2006. The imprint of the geographical, evolutionary and ecological context on species–area relationships. Ecology letters, 9(2): 215-227. Dunham, A. E. 2011. Soil disturbance by vertebrates alters seed predation, movement and germination in an African rain forest. Journal of Tropical Ecology, 27(06): 581-589. Dunne, J.A., Williams, R.J. and Martínez, N.D. 2002. Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecol. Lett., 5: 558-567. Eckstein, R.L. and Donath, T.W. 2005. Interactions between litter and water availability affect seedling emergence in four familial pairs of floodplain species. Journal of Ecology 93: 807-816. El-Bana, M.I., Li, Z.Q. and Nijs, I. 2007. Role of host identity in effects of phytogenic mounds on plant assemblages and species richness on coastal arid dunes. J. Veg. Sci 18: 635-644.

229

References

Eldridge, D.J. and Greene, R.S.B. 1994. Microbiotic soil crusts-a review of their roles in soil and ecological processes in the rangelands of Australia. Soil Research, 32(3): 389-415. Eldridge, D.J., Bowker, M.A., Maestre, F.T., Roger, E., Reynolds, J.F. and Whitford, W.G. 2011. Impacts of shrub encroachment on ecosystem structure and functioning: towards a global synthesis. Ecology Letters 14: 709-722. Escudero, A., Martínez, I., de la Cruz, A., Otálora, M.A.G. and Maestre, F.T. 2007. Soil lichens have species-specific effects on the seedling emergence of three gypsophile plant species. Journal of Arid Environments 70: 18-28. Espejo, D. 2007. Brachypodium retusum (Pers.) Beauv. (Poaceae), una espècie d’interès per a la recuperació d’espais degradats mediterranis. Quaderns Agraris 31: 85-107. Estrada, E. 2007. Characterization of topological keystone species. Local, global and "meso-scale" centralities in food webs. Ecological Complexity 4: 48-57. Eugenio, M., Olano, J. M., Ferrandis, P., Martínez-Duro, E. and Escudero, A. 2012. Population structure of two dominant gypsophyte shrubs through a secondary plant succession. Journal of Arid Environments 76: 30-35. Facelli, J.M and Pickett, S.T.A. 1991a. Indirect effects of litter on woody seedlings subject to herb competition. Oikos 62: 129-138. Facelli, J.M and Pickett, S.T.A. 1991b. Plant litter: Its dynamics and effects on plant community structure. The Botanical Review 57: 1-32. Facelli, J.M and Pickett, S.T.A. 1991c. Plant litter: Light interception and effects on an old-field plant community. Ecology 12(3): 1024-1031. Facelli, J.M., Williams, R., Fricker, S. and Ladd, B. 1999. Establishment and growth of seedlings of Eucalyptus obliqua: interactive effects of litter, water, and pathogens. Austral Ecology 24: 484-494. FAO, 2005. Forests and floods. Drowning in fiction or thriving on facts?. Center for International Forestry Research and Food and Agriculture Organization of the United Nations. RAP Publication 2005/03. Forest Perspectives 2. 30 pp.

230

References

Fernández-Palazón, G. 1974. Aspectos socioeconómicos de la explotación del esparto en España. Revista de Geografía VIII(1-2): 203-212. Fernández, C., Voiriot, S., Mévy, J.-P., Vila, B., Ormeño, E., Dupouyet, S. and Bousquet-Mélou, A. 2008. Regeneration failure of Pinus halepensis Mill.: The role of autotoxicity and some abiotic environmental parameters. Forest Ecology and Management 255: 2928-2936. Ferran A. 1996. La fertilitat de sòls forestals en la regeneració després del foc de diferents ecosistemes mediterranis. Ph.D. Thesis, University of Barcelona, Barcelona, Spain. Ferrari, S.L.P. and Cribari-Neto, F. 2004. Beta regression for modelling rates and proportions. Journal of Applied Statistics, 31: 799-815. Flores, J. and Jurado, E. 2003. Are nurse-protégé interactions more common among plants from arid environments? J. Veg. Sci. 14: 911-916. Fons, J. 1995. Avaluació de la fertilitat de sòls forestals mediterranis. El cas de les pinedes de pi blanc (Pinus halepensis Mill.). Tesis doctoral. Universitat de Barcelona. Fontaine, C., Guimarães Jr., P.R., Kefi, S., Loeuille, N., Memmott, J., van der Putten, W.H., van Veen, F.J.F. and Thébault, E. 2011. The ecological and evolutionary implications of merging different types of networks. Ecology Letters 14, p. 1170 Fontaine, S., Mariotti, A. and Abbadie, L. 2003. The priming effect of organic matter: a question of microbial competition? Soil Biology and Biochemistry 35: 837-843. Fortuna, M.A., Stouffer, D., Olesen, J., Jordano, P., Mouillot, D., Krasnov, B., Poulin, R. and Bascompte, J. 2010. Nestedness versus modularity in ecological networks: two sides of the same coin? Journal of Animal Ecology 79: 811-817. France, R.L. 1997. Potential for soil erosion from decreased litterfall due to riparian clearcutting: Implications for boreal forestry and warm- and cool-water fisheries. Journal of Soil and Water Conservation 52(6): 452-455. Freeman, L.C. 1979. Centrality in social networks conceptual clarification. Social networks 1(3): 215-239.

231

References

Fridley, J.D., Stachowicz, J.J., Naeem, S., Sax, D.F., Seabloom, E.W., Smith, M. D., Stohlgren, T.J., Tilman, D. and Von Holle, B., 2007. The invasion paradox: reconciling pattern and process in species invasions. Ecology 88: 3-17. Fulé, P.Z. and Covington, W.W. 1994. Double Sampling Increases the Efficiency of Forest Floor Inventories for Arizona Ponderosa Pine Forests. International Journal of Wildland Fire 4(1): 3-10. Galeano, J., Pastor, J.M. and Iriondo, J.M. 2008. Weighted-Interaction Nestedness Estimator (WINE): A new estimator to calculate over frequency matrices. Environmental Modelling and Software 24 (11): 1342-1346. Gallardo, A. 2003. Effect of tree canopy on the spatial distribution of soil nutrients in a Mediterranean Dehesa. Pedobiologia, 47(2): 117-125. Gallardo, A. and Merino, J. 1993. Leaf decomposition in two Mediterranean ecosystems of Southwest Spain: Influence of substrate quality. Ecology 74(1): 152- 161. García-Cano, M.F. 1998. Avaluació de la pèrdua de nutrients i dels canvis en la fertilitat del sòl en un matollar de Ulex parviflorus afectat pel foc i la pluja torrencial. Tesi de Llicenciatura. Universitat d’Alacant. García-Fayos, P. and Gasque, M. 2002. Consequences of a severe drought on spatial patterns of woody plants in a two-phase mosaic steppe of Stipa tenacissima L. Journal of Arid Environments, 52(2): 199-208. García-Fayos, P., Gulias, J., Martínez, J., Marzo, A., Melero, J.P., Traveset, A., Veintinilla, P., Verdú, M., Cerdán, V., Gasque, M. and Medrano, H. 2001. Bases ecológicas para la recolección, almacenamiento y germinación de semillas de especies de uso forestal de la Comunidad Valenciana. Banc de Llavors Forestals (Conselleria de Medi Ambiente, Generalitat Valenciana), Valencia. García-Fayos, P. and Verdú, M. 1998. Soil seed bank, factors controlling germination and establishment of a Mediterranean shrub: Pistacia lentiscus L. Acta Oecologica 19(4): 357-366. García-Guzmán, G. and Benítez-Malvido, J. 2003. Effect of litter on the incidence of leaf-fungal pathogens and herbivory in seedlings of the tropical tree Nectandra ambigens. Journal of Tropical Ecology 19: 171-177.

232

References

García-Pausas, J., Casals, P. and Romanyà, J. 2004. Litter decomposition and faunal activity in Mediterranean forest soils: effects of N content and the moss layer. Soil Biology and Biochemistry 36: 989-997. Garnett, E., Jonsson, L. M., Dighton, J. and Murnen, K. 2004. Control of pitch pine seed germination and initial growth exerted by leaf litters and polyphenolic compounds. Biology and fertility of soils, 40(6): 421-426. Gazol, A. and Ibáñez, R. 2010. Variation of plant diversity in a temperate unmanaged forest in northern Spain: behind the environmental and spatial explanation. Plant Ecology 207: 1-11. Geddes, N. and Dunkerley, D. 1999. The influence of organic litter on the erosive effects of raindrops and of gravity drops released from desert shrubs. Catena 36: 303-313. George, L.O. and Bazzaz, F. A., 1999. The fern understory as an ecological filter: growth and survival of canopy-tree seedlings. Ecology 80: 846-856. Gilad, E., von Hardenberg, J., Provenzale, A., Shachak, M. and Meron, E. 2004. Ecosystem engineers: from pattern formation to habitat creation. Physical Review Letters 93(9): 098105-1-098105-4. Gilbert, B., Turkington, R. and Srivastava, D.S., 2009. Dominant species and diversity: linking relative abundance to controls of species establishment. Am. Nat. 174: 850-862. Ginter, D.L., Mcleod, K.W. and Sherrod, C. 1979. Water stress in longleaf pine induced by litter removal. Forest Ecology and Management 2: 13-20. Goldin, S.R. and Brookhouse, M.T., 2014. Effects of coarse woody debris on understorey plants in a temperate Australian woodland. Applied Vegetation Science. doi: 10.1111/avsc.12120 Gómez-Aparicio, L., Zamora, R., Gómez, J., Hodar, J.A., Castro, J. and Baraza, E. 2004. Applying plant facilitation to forest restoration: A meta-analysis of the use of shrubs as nurse plants. Ecological Applications 14(4): 1128-1138. Gómez, J.M., Jordano, P. and Perfectti, F. 2011. The functional consequences of mutualistic network architecture. PLoS One 6, e16143.

233

References

Gonzalez, G. and Zou, X. 1999. Plant and litter influences on earthworm abundance and community structure in a tropical wet forest. Biotropica 31: 486- 493. Gotelli, N.J. 2000. Null model analysis of species co-occurrence patterns. Ecology 81(9): 2606–2621 Gotelli, N.J. and Colwell, R.K., 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4: 379–391. Granda, E., Escudero, A., de la Cruz, M. and Valladares, F., 2012. Juvenile–adult tree associations in a continental Mediterranean ecosystem: no evidence for sustained and general facilitation at increased aridity. J Veg Sci. 23: 164-175. Grime J. P., 1974. Vegetation classification by reference to strategies. Nature 250: 26-31. Gross, K. 2008. Positive interactions among competitors can produce species-rich communities. Ecol. Lett. 11: 929-936. Guevara-Escobar, A., González-Sosa, E., Ramos-Salinas, M. and Hernández- Delgado, G.D. 2007. Experimental analysis of drainage and water storage of litter layers Hydrology and Earth System Science 11: 1703-1716. Guimera, R. and Amaral, L.A.N. 2005. Cartography of complex networks: modules and universal roles, J. Stat. Mech.-Theory Exp., P02001. Gutiérrez, J.R., Meserve, P.L., Herrera, S., Contreras, L.C. and Jaksic, F.M. 1997. Effects of small mammals and vertebrate predators on vegetation in the Chilean semiarid zone. Oecologia 109: 398-406. Haase, P., Pugnaire, F. I, and Incoll, L. D. 1995. Seed production and dispersal in the semi-arid tussock grassStipa tenacissimaL. during masting. Journal of Arid Environments, 31(1): 55-65. Hanlon, R.D.G. and Anderson, J.M. 1980. Influence of macroarthropod feeding activities on microflora in decomposing oak leaves. Soil Biology and Biochemistry 12: 255-261.

234

References

Hassall, M., Turner, J.G. and Rands, M.R.W. 1987. Effects of terrestrial isopods on the decomposition of woodland leaf litter. Oecologia 72: 597-604. Hastings, A., Byers, J.E., Crooks, J.A., Cuddington, K., Jones, C.G., Lambrinos, J.G., Talley, T.S. and Wilson, W.G. 2007. Ecosystem engineering in space and time. Ecology Letters 10 (2): 153-164. Heleno, R. H., Olesen, J. M., Nogales, M., Vargas, P. and Traveset, A. 2013. Seed dispersal networks in the Galápagos and the consequences of alien plant invasions. Proceedings of the Royal Society B: Biological Sciences, 280(1750): 2012-2112. Heleno, R., Lacerda, I., Ramos, J.A. and Memmott, J. 2010. Evaluation of restoration effectiveness: community response to the removal of alien plants. Ecol. Appl. 20: 1191–1203. Henson, K.S.E., Craze, P.G. and Memmott, J. 2009. The restoration of parasites, parasitoids, and pathogens to heathland communities. Ecology, 90: 1840-1851. Herranz, J.M., Ferrandis, P., Copete, M.A., Duro, E.M. and Zalacain, A. 2006. Effects of allelopathic compounds produced by Cistus ladanifer on germination of 20 Mediterranean taxa. Plant Ecology 184: 259-272. Herrera, C.M. 1992. Historical effects and sorting processes as explanations for contemporary ecological patterns: character syndromes in Mediterranean woody plants. American Naturalist, 421-446. Herrera, C.M. 1995a. Plant-vertebrate seed dispersal systems in the Mediterranean: ecological, evolutionary, and historical determinants. Annual Review of Ecology and Systematics, 705-727. Herrera, C.M. 1989 Frugivory and seed dispersal by carnivorous mammals, and associated fruit characteristics, in undisturbed Mediterranean habitats, Oikos 55: 250-262. Herrera, J. 1995b. Acorn predation and seedling production in a low-density population of cork oak (Quercus suber L.). Forest Ecology and Management, 76(1): 197-201.

235

References

Herrera, C.M., Jordano, P., Lopez-Soria, L. and Amat, J. A. 1994. Recruitment of a mast-fruiting, bird-dispersed tree: bridging frugivore activity and seedling establishment. Ecological Monographs 64(3): 315-344. Hobbs, R.J. 1984. Possible chemical interactions among heathland plants. Oikos 43: 23-29. Holmgren, M., Scheffer, M., Ezcurra, E., Gutiérrez, J.R. and Mohren, G.M.J. 2001. El Niño effects on the dynamics of terresrial ecosystems. Trends in Ecology and Evolution 16: 89-94. Hughes, A.R., Byrnes, J.E., Kimbro, D.L. and Stachowicz, J.J. 2007. Reciprocal relationships and potential feedbacks between biodiversity and disturbance. Ecology Letters 10: 849-864. Hulbert, S.H. 1997. Functional importance vs. keystoneness: reformulating some questions in theoretical biocenology. Aust. J. Ecol. 22: 369-382. Huxman, T.E., Wilcox, B.P., Breshears, D.D., Scott, R.L., Snyder, K.A., Small, E.E., Hultine, K., Pockman, W.T. and Jackson, R.B. 2005. Ecohydrological implications of woody plant encroachment. Ecology 86(2): 308-319. Inderjit, and Nilsen, E. T. 2003. Bioassays and field studies for allelopathy in terrestrial plants: progress and problems. Critical Reviews in Plant Sciences, 22(3- 4): 221-238. Inderjit, and Duke, S.O. 2003. Ecophysiological Aspects of Allelopathy. Planta 217: 529-539. IPCC, 2007. Intergovernmental Panel on Climate Change, Climate Change 2007: Synthesis Report, Contribution of Working Groups I, II and III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. IPCC, Geneva [Core Writing Team, Pachauri, R.K and Reisinger, A. (eds.)]. IPCC, 2013. Long-term Climate Change: Projections, Commitments and Irreversibility. In: Collins, M., Knutti, R. (Eds.). Report of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University Press, Cambridge, 1136 pp. Jackson, R.B., Banner, J.B., Jobbagy, E.G., Pockman, W.T., Diana, H. and Wall, D.H. 2002. Ecosystem carbon loss with woody plant invasion of grasslands. Nature 418: 623-626.

236

References

Jackson, R.B., Jobbagy, E.G., Avissar, R., Roy, S.B., Barrett, D.J., Cook, C.W., Farley, K.A., le Maitre, D.C., McCarl, B.A. and Murray, B.C. 2005. Trading water for carbon with biological sequestration. Science 310: 1944-1947. Jones, C.G. and Callaway, R.M., 2007. The third party. J. Veg. Sci. 18: 771-776. Jones, C.G., Lawton, J.H. and Shachak, M. 1994. Organisms as ecosystem engineers. Oikos 69: 373-386. Jones, C.G., Lawton, J.H. and Shachak, M. 1997. Positive and negative effects of organisms as ecosystem engineers. Ecology 78: 1946-1957. Jordán, F. and Scheuring, I. 2002. Searching for keystones in ecological networks. Oikos 99: 607-612. Jordano, P., Bascompte, J. and Olesen, J.M. 2003. Invariant properties in coevolutionary networks of plant-animal interactions. Ecol. Lett. 6: 69-81. Jordano, P., García, C., Godoy, J.A. and García-Castano, J.L. 2007. Differential contribution of frugivores to complex seed dispersal patterns. PNAS 104(9): 3278– 3282. Jump, A.S. and Peñuelas, J. 2005. Running to stand still: adaptation and the response of plants to rapid climate change. Ecology Letters, 8(9): 1010-1020. Kavvadias, V.A., Alifragis, D., Tsiontsis, A., Brofas, G. and Stamatelos, G. 2001. Litterfall, litter accumulation and litter decomposition rates in four forest ecosystems in northern Greece. Forest Ecology and Management 144: 113-127. Kemp, P.R., Reynolds, J.F., Virginia, R.A. and Whitford, W.G. 2003. Decomposition of leaf and root litter of Chihuahuan desert shrubs: effects of three years of summer drought. Journal of Arid Environments 53(1): 21-39. Konings, A., Dekker, S., Rietkerk, M. and Katul, G. 2011. Drought sensitivity of patterned vegetation determined by rainfall-land surface feedbacks. J. Geophys. Res. 116: G04008. Koorem, K. and Moora, M. 2010. Positive association between understory species richness and a dominant shrub species (Corylus avellana) in a boreonemoral spruce forest. Forest Ecology and Management 260: 1407-1413.

237

References

Krause, A.E., Frank, K.J., Mason, D.M., Ulanowicz, R.E. and Taylor, W.W. 2003. Compartments revealed in food web structure. Nature 426: 282-285. Kurz-Besson, C., Coûteaux, M.M., Berg, B., Remacle, J., Ribeiro, C., Romanyà, J. and Thiéry, J.M. 2006. A climate response function explaining most of the variation of the forest floor needle mass and the needle decomposition in pine forests across Europe. Plant Soil 285: 97-114. Kutiel, P., Kutiel, H. and Lavee, H. 2000 Vegetation response to possible scenarios of rainfall variations along a Mediterranean-extreme arid climatic transect. Journal of Arid Environments 44: 277-290. Lamb, E.G. 2008. Direct and indirect control of grassland community structure by litter, resources, and biomass. Ecology 89(1): 216-225. Laughlin, D.C. and Fulé, P.Z. 2008. Wildland fire effects on understory plant communities in two fire-prone forests. Canadian Journal of Forest Restoration 38: 133-142. Lavado, M., Núñez, E. and Escudero, J.C. 1989. Variaciones mensuales en el aporte de biomasa al suelo por distintas especies de matorral mediterráneo. Options Méditerranéennes - Série Séminaires 3: 167-172. Le Houérou, H.N. 1986. The desert and arid zones of northern Africa. En: Evenari, M., Noy-Meir, I. y Goodall, D.W. (Eds.) The ecosystems of the world. Volume 12 B. Hot deserts and arid shrublands. Elsevier Science, Amsterdam, The Netherlands. Le Houérou, H.N. 2001. Biogeography of the arid stepplands north of the Sahara. Journal of Arid Environments 48: 103-128. Levine, J.M. 2000. Species diversity and biological invasions: relating local process to community pattern. Science 288: 852-854. Lockhart, J. A. 1961. Photoinhibition of stem elongation by full solar radiation. American Journal of Botany 387-392. López-Barrera, F. and González-Espinosa, M. 2001. Influence of litter on emergence and early growth of Quercus rugosa: a laboratory study. New Forests 21: 59-70.

238

References

López-Zamora, I., Duryea, M.L., McCormac Wild, C., Comerford, N.B. and Neary, D.G. 2001. Effect of pine needle removal and fertilization on tree growth and soil P availability in a Pinus elliotii Englem. var. elliotii stand. Forest Ecology and Management 148: 125-134. López, G. and Moro, M.J. 1997. Birds of Aleppo pine plantations in south-east Spain in relation to vegetation composition and structure. Journal of Applied Ecology 34: 1257-1272. Ludwig, J. A. and Tongway, D. J. 1996. Rehabilitation of semiarid landscapes in Australia. II. Restoring vegetation patches. Restoration Ecology, 4(4): 398-406. Ludwig, J.A., Wilcox, B.P., Breshears, D.D., Tongway, D.J. and Imeson, A.C. 2005. Vegetation patches and runoff–erosion as interacting ecohydrological processes in semiarid landscapes. Ecology 86(2): 288-297. MacArthur, R.H. and Levins, R., 1967. The limiting similarity, convergence, and divergence of coexisting species. Am. Nat. 101: 377-385. Machado, M.J., Benito, G., Barriendos, M., and Rodrigo, F.S. 2011. 500 years of rainfall variability and extreme hydrological events in southeastern Spain drylands. Journal of Arid Environments, 75(12): 1244-1253. Maestre, F.T. and Cortina, J. 2004a. Do positive interactions increase with abiotic stress? A test from a semi-arid steppe. Proceedings of the Royal Society of London. Series B (Supplement) 271: S331-S333. Maestre, F.T. and Cortina, J. 2004b. Insights on ecosystem composition and function in a sequence of degraded semiarid steppes. Restoration Ecology 12: 494-502. Maestre, F.T. and Cortina, J. 2005. Remnant shrubs in Mediterranean semi-arid steppes: effects of shrub size, abiotic factors and species identity on understorey richness and occurrence. Acta Oecologica 27: 161-169.

Maestre, F.T. and Cortina, J. 2003. Small-scale spatial variation in soil CO2 efflux in a Mediterranean semiarid steppe. Applied Soil Ecology, 23(3): 199-209. Maestre, F.T. and Cortina, J. 2002. Spatial patterns of surface soil properties and vegetation in a Mediterranean semi-arid steppe. Plant and Soil 241: 279-291.

239

References

Maestre, F. T., Bautista, S. and Cortina, J., 2003a. Positive, negative and net effects in grass-shrub interactions in Mediterranean semiarid grasslands. Ecology 84, 3186-3197. Maestre, F.T., Cortina, J., Bautista, S., Bellot, J. and Vallejo, V.R., 2003b. Small- scale environmental heterogeneity and spatio-temporal dynamics of seedling establishment in a semiarid degraded ecosystem. Ecosystems 6: 630-643. Maestre, F.T., Callaway, R.M., Valladares, F. and Lortie, C.J. 2009a. Refining the stress-gradient hypothesis for competition and facilitation in plant communities. Journal of Ecology 97(2): 199-205 Maestre, F. T., Bowker, M. A., Puche, M. D., Hinojosa, M.B., Martínez, I., García‐Palacios, P., Castillo, A.P., Soliveres, S., Luzuriaga, A.L., Sánchez, A.M., Carreira, J.A., Gallardo, A. and Escudero, A. 2009b. Shrub encroachment can reverse desertification in semi‐arid Mediterranean grasslands. Ecology Letters 12(9): 930-941. Maestre, F.T., Bautista, S and Cortina, J. 2006. Stipa tenacissima does not affect the foliar d13C and d15N of introduced shrub seedlings in a Mediterranean semiarid steppe. Journal of Integrative Plant Biology 48: 897-905. Maestre, F.T. 2004. On the importance of patch attributes, abiotic factors and past human impacts as determinants of plant species richness and diversity in Mediterranean semiarid steppes. Diversity and Distributions 10: 21-29. Maestre, F.T., Bautista, S., Cortina, J. and Bellot, J. 2001. Potential of using facilitation by grasses to establish shrubs on a semiarid degraded steppe. Ecological Applications 11: 1641-1655. Maestre, F.T., Bautista, S., Cortina, J., Díaz, G., Honrubia, M. and Vallejo. V.R. 2002. Microsite and mycorrhizal inoculum effects on the establishment of Quercus coccifera in a semiarid degraded steppe. Ecological Engineering 19: 289- 295. Maestre, F.T., Cortina, J. and Bautista, S. 2004. Mechanisms underlying the interaction between Pinus halepensis and the native late-successional shrub Pistacia lentiscus in a semiarid plantation. Ecography 27: 1-11.

240

References

Maret, M.P. and Wilson, M.V. 2005. Fire and Litter Effects on Seedling Establishment in Western Oregon Upland Prairies. Restoration Ecology 13(3): 562–568. Marshall, T.J., Holmes, J.W. and Rose, C.W. 1996. Soil Physics. 3rd edn. Cambridge University Press, Cambridge. Martens, S. N., Breshears, D. D. and Meyer, C. W. 2000. Spatial distributions of understory light along the grassland/forest continuum: effects of cover, height, and spatial pattern of tree canopies. Ecological Modelling 126(1): 79-93. Martín-González, A.M., Dalsgaard, B. and Olesen, J.M. 2010. Centrality measures and the importance of generalist species in pollination networks. Ecological Complexity 7: 36–41 Matías, L., Zamora, R., Mendoza, I. and Hódar, J.A. 2008. Seed Dispersal Patterns by Large Frugivorous Mammals in a Degraded Mosaic Landscape. Restoration Ecology 18(5): 1-9. Mayor, Á.G., Bautista, S., Small, E.E., Dixon, M. and Bellot, J. 2008. Measurement of the connectivity of runoff source areas as determined by vegetation pattern and topography: a tool for assessing potential water and soil losses in drylands. Water Resources Research 44(10) W10423.

McCune, B. and Grace, J.B. 2002. Analysis of ecological communities. MjM Software Design, Gleneden Beach, Oregon, USA. MEA (Millennium Ecosystem Assessment), 2005. Ecosystems and Human Well- being: Desertification Synthesis. World Resources Institute, Washington, DC. Mediavilla, S. and Escudero, A., 2004. Stomatal responses to drought of mature trees and seedlings of two co-occurring Mediterranean oaks. Forest Ecol. Manag. 187: 281-294. Meentemeyer, V., Box, E.O. and Thompson, R. 1982. World patterns and amounts of terrestrial plant litter production. BioScience 32: 125-128. Memmott, J., Craze, P.G., Waser, N.M. and Price, M.V. 2007. Global warming and the disruption of plant–pollinator interactions. Ecol. Lett. 10: 710-717.

241

References

Memmott, J., Waser, N.M. and Price, M.V. 2004. Tolerance of pollination networks to species extinctions. Proc. R Soc. Lond. B 271: 2605–2611. Méndez‐Alonzo, R., Ewers, F. W. and Sack, L. 2013. Ecological variation in leaf biomechanics and its scaling with tissue structure across three mediterranean‐climate plant communities. Functional Ecology, 27(2): 544-554. Mendoza-Aguilar, D. O., Cortina, J. and Pando-Moreno, M. 2014. Biological soil crust influence on germination and rooting of two key species in a Stipa tenacissima steppe. Plant and soil, 375(1-2): 267-274. Mendoza, I., Zamora, R. and Castro, J.A. 2009. Seeding experiment for testing tree-community recruitment under variable environments: Implications for forest regeneration and conservation in Mediterranean habitats. Biological Conservation 142(7): 1491-1499. Metcalfe, D.J. and Turner, I.M. 1998. Soil seed bank from lowland rain forest in Singapore: canopy-gap and litter-gap demanders. Journal of Tropical Ecology 14: 103-108. Milberg, P., Andersson, L. and Thompson, K. 2000. Large-seeded species are less dependent on light for germination than small-seeded. Seed Science Research 10: 99-104. Montané, F., Rovira, P. and Casals, P. 2007. Shrub encroachment into mesic mountain grasslands in the Iberian peninsula: Effects of plant quality and temperature on soil C and N stocks. Global Biogeochemical Cycles 21 (4016): 1-10. Montoya, J.M., Pimm, S.L. and Solé, R.V. 2006. Ecological networks and their fragility. Nature 442: 259-264. Morris, W. F. and Wood, D. M. 1989. The role of lupine in succession on Mount St. Helens: Facilitation or inhibition? Ecology 70(3): 697-703. Muñoz-Reinoso, J.C. 1993. Consumo de gálbulos de sabina (Juniperus phoenicea spp. turbinata Guss, 1981) y dispersión de semillas por el conejo (Oryctolagus cuniculus) en el Parque Nacional de Doñana. Doñana Acta Vertebrata 20: 49-58. Myster, R.W. 1994. Contrasting litter effects on old field tree germination and emergence. Vegetatio, 114(2): 169-174.

242

References

Nara, H. and Hogetsu, K., 2004. Ectomycorrhizal fungi on established shrubs facilitate subsequent seedling establishment of successional plant species. Ecology 85: 1700-1707. Navarro R.M., del Campo A. and Cortina J. 2006. Factores que afectan al éxito de una repoblación y su relación con la calidad de la planta. In Calidad de planta forestal para la restauración en ambientes Mediterráneos. Estado actual de conocimientos. Cortina, J., Peñuelas, J.L., Puértolas, J., Vilagrosa, A. and Savé, R. (Coord.). Organismo Autónomo Parques Nacionales. Ministerio de Medio Ambiente. Madrid Navarro-Cano, J.A. 2007. Reclutamiento vegetal en pinares de repoblación y espartales en ambientes semiáridos: aplicaciones a la restauración ecológica. Tesis Doctoral. Universidad de Murcia. Navarro-Cano, J.A., Barberá, G.G., Ruiz-Navarro, A. and Castillo, V.M. 2009. Pine plantation bands limit seedling recruitment of a perennial grass Ander semiarid conditions. Journal of Arid Environments 73: 120-126. Nelson, D.W. and Sommers, L.E. 1982. Total carbon, organic carbon and organic matter. En: Page, A.L. (Ed). Methods of soil analysis. Part 2. Chemical and Microbiological Properties. Second edition. American Society of Agronomy, Madison, Wisconsin, USA. Newman, M.E.J. 2004. Analysis of weighted networks. Physical Review E 70: 056131. Ninyerola, M., Pons, X. and Roure, J.M. 2005. Atlas Climático Digital de la Península Ibérica. Metodología y aplicaciones en bioclimatología y geobotánica. ISBN 932860-8-7. Universidad Autónoma de Barcelona, Bellaterra, Spain. Noble, A.D. and Randall, P.J. 1999. Alkalinity effects of different tree litters incubated in an acid soil of N.S.W. Australia. Agroforestry 46: 147-160. Northup, R.R., Yu, Z., Dahlgren, R.A. and Vogt, K.A. 1995. Polyphenol control of nitrogen release from pine litter. Nature 377: 227-229. Noy-Meir, I. 1973. Desert ecosystems: Environment and producers. Annual Review of Ecology and Systistematics 4: 25-51

243

References

O’Connell, P.E., Beven, K.J., Carney, J.N., Clements, R. O., Ewen, J., Fowler, H., Harris, G.L., Hollis, L., Morris, J., O’Donnell, G.M., Packman, J.C., Parkin, A., Quinn, P.F., Rose, S.C., Shepherd, M. and Tellier, S. 2005. Review of impacts of rural land use and management on flood generation. Impact study report. RandD Technical Report FD2114/TR. Defra - Flood Management Division. London. Available at: www.defra.gov.uk/environ/fcd. Okuyama, T. and Holland, J. 2008. Network structural properties mediate the stability of mutualistic communities. Ecol. Lett. 11: 208-216. Olano, J. M., Eugenio, M. and Escudero, A. 2011. Site effect is stronger than species identity in driving demographic responses of Helianthemum (Cistaceae) shrubs in gypsum environments. American journal of botany 98(6): 1016-1023. Olesen, J. M., Bascompte, J., Dupont, Y. L. and Jordano, P. 2007. The modularity of pollination networks. Proceedings of the National Academy of Sciences, 104(50): 19891-19896. Padilla, F.M., Ortega, R., Sánchez, J. and Pugnaire, F.I. 2009. Rethinking species selection for restoration of arid shrublands. Basic Appl. Ecol. 10: 640-647. Palamarev, E. 1989. Paleobotanical evidences of the Tertiary history and origin of the Mediterranean sclerophyll dendroflora. Plant Systematics and Evolution, 162(1-4): 93-107. Parker, K. W. and Martin, S. C. 1952. Mesquite problem on southern Arizona ranges. U. S. Dep. Agr. Circ. No. 908. 70 p. Pascual, M. and Dunne, J. A. 2006. Ecological networks: linking dynamics in food webs. Oxford, UK: Oxford University Press. Patón, D., Azocar, P. and Tovar, J. 1998. Growth and productivity in forage biomass in relation to the age assessed by dendrochronology in the evergreen shrub Cistus ladanifer (L.) using different regression models. Journal of Arid Environments, 38(2): 221-235. Pausas, J., Bonet, A., Maestre, F.T. and Climent, A. 2006. The role of the perch effect on the nucleation process in Mediterranean semiarid oldfields. Acta Oecologica 29(3): 346-352.

244

References

Pausas, J.G., Bladé, C., Valdecantos, A., Seva, J.P., Fuentes, D., Alloza, J.A., Vilagrosa, A., Bautista, S., Cortina, J. and Vallejo, R., 2004. Pines and oaks in the restoration of Mediterranean landscapes: new perspectives for an old practice. Plant Ecology 171: 209-220. Peñuelas, J., Filella, I. and Terradas, J. 1999. Variability of plant nitrogen and water use in a 100-m transect of a subdesertic depression of the Ebro valley (Spain) characterized by leaf δ13C and δ15N. Acta Oecol. 20: 119-123. Peñuelas, J., Lloret, F. and Montoya, R. 2001. Severe drought effects on Mediterranean woody flora in Spain. Forest Science 47(2): 214-218. Peñuelas, J.L and Ocaña, L. 1996. Cultivo de plantas forestales en contenedor. Principios y fundamentos. Ministerio de Agricultura, Pesca y Alimentación. Mundi- Prensa, Madrid. Peters, E. M., Martorell, C. and Ezcurra, E. 2008. Nurse rocks are more important than nurse plants in determining the distribution and establishment of globose cacti (Mammillaria) in the Tehuacán Valley, Mexico. Journal of Arid Environments, 72(5): 593-601. Peterson, C. J. and Facelli, J. M. 1992. Contrasting germination and seedling growth of Betula alleghaniensis and Rhus typhina subjected to various amounts and types of plant litter. American Journal of Botany 79: 1209-1216. Pocock, M.J., Evans, D.M., and Memmott, J. 2012. The robustness and restoration of a network of ecological networks. Science 335(6071): 973-977. Poesen, J. and Lavee, H. 1994. Rock fragments in top soils: significance and processes. Catena, 23(1): 1-28. Puerta-Piñero, C., Gómez, J. M. and Valladares, F. 2007. Irradiance and oak seedling survival and growth in a heterogeneous environment. Forest Ecology and Management 242(2): 462-469. Pueyo, Y. and Alados, C.L. 2007. Abiotic factors determining vegetation patterns in a semiarid Mediterranean landscape: Different responses on gypsum and non- gypsum substrates. Journal of Arid Environments 69(3): 490-505.

245

References

Pugnaire, F. I., Armas, C. and Maestre, F. T. 2011. Positive plant interactions in the Iberian Southeast: mechanisms, environmental gradients, and ecosystem function. Journal of Arid Environments 75(12): 1310-1320. Pugnaire, F.I. and Lázaro, R. 2000. Seed bank and understorey species composition in a semiarid environment: The effect of shrub age and rainfall. Annals of Botany 86: 807-813. Pugnaire, F.I., Armas, C. and Maestre, F.T., 2011. Positive plant interactions in the Iberian Southeast: Mechanisms, environmental gradients, and ecosystem function. J. Arid Environ. 75: 1310-1320. Pugnaire, F.I., Haase, P. and Puigdefábregas, J. 1996. Facilitation between higher plant species in a semiarid environment. Ecology 77: 1420-1426. Puigdefábregas, J. and Mendizábal, T. 1998. Perspectives on desertification: western Mediterranean. Journal of Arid Environments 39: 209-224. Puigdefábregas, J. 2005. The role of vegetation patterns in structuring runoff and sediment fluxes in drylands. Earth Surf. Proc. Land. 30 : 133-147. Quézel, P. and Médail, F. 2003. Écologie et biogéographie des forêts du basin méditerranéen. Elsevier. Paris. R Development Core Team, 2014. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. Ramírez, D. A. and Bellot, J. 2009. Linking population density and habitat structure to ecophysiological responses in semiarid Spanish steppes. Plant Ecology 200(2): 191-204. Ramírez, J.A. and Diaz, M., 2008. The role of temporal shrub encroachment for the maintenance of Spanish holm oak Quercus ilex dehesas. Forest Ecology and Management 255: 1976-1983. Reader, R.J. 1993. Control of seedling emergence by ground cover and seed predation in relation to seed size for some old-field species. Journal of Ecology 81: 169-175.

246

References

Rebollo, S., Pérez-Camacho, L., García-de-Juan, M.T., Rey-Benayas, J.M. and Gómez-Sal, A. 2001. Recruitment in a Mediterranean annual plant community: seed bank, emergente, litter, and intra- and Inter.-specific interactions. Oikos 95: 485-495. Rey-Benayas, J.M., López-Pintor, A., García, C., de la Cámara, N., Strasser, R. and Gómez-Sal, A., 2002. Early establishment of planted Retama sphaerocarpa seedlings under different levels of light, water and weed competition. Plant Ecology 159: 201-209. Rey-Benayas, J.M., Newton, A.C., Diaz, A. and Bullock, J.M. 2009. Enhancement of biodiversity and ecosystem services by ecological restoration: a meta-analysis. Science 325: 1121-1124. Rey, A., Pegoraro, E., Oyonarte, C., Were, A., Escribano, P. and Raimundo, J. 2011. Impact of land degradation on soil respiration in a steppe (Stipa tenacissima L.) semi-arid ecosystem in the SE of Spain. Soil Biology and Biochemistry 43(2): 393- 403. Rey, P.J. and Alcantara, J.M. 2000.Recruitment Dynamics of a Fleshy-Fruited Plant (Olea europaea): Connecting Patterns of Seed Dispersal to Seedling Establishment. Journal of Ecology 88: 622-633. Rey, P.J., Garrido, J.L, Alcántara, J.M., Ramírez, J.M., Aguilera, A., García, L., Manzaneda, A.J. and Fernández, R. 2002. Spatial variation in ant and rodent post- dispersal predation of vertebrate-dispersed seeds. Functional Ecology 16(6): 773- 781. Reynolds, J.F. 2001. Desertification. En: Levin, S. (Ed.) Enciclopedia of Biodiversity. 2. San Diego: Academic. Reynolds, J.F., Virginia, R.A., Kemp, P.R., de Soyza, A.G. and Tremmel, D.C. 1999. Impact of drought on desert shrubs: effects of seasonality and degree of resource island development. Ecological Monographs 69: 69-106. Rivas-Martínez, S. 1987. Memoria del mapa de series de vegetación de España. Instituto para la Conservación de la Naturaleza, Madrid. Rivas-Martínez, S., and Rivas-Sáenz, S. 2002. Worldwide bioclimatic classification system. Backhuys Pub.

247

References

Rodríguez-Calcerrada, J., Nanos, N., del Rey, MC., López de Heredia, U., Escribano, R. and Gil, L. 2011. Small-scale variation of vegetation in a mixed forest understorey is partly controlled by the effect of overstory composition on litter accumulation. Journal of Forest Research 16(6): 473-483.

Rodríguez‐Gironés, M. A. and Santamaría, L. 2006. A new algorithm to calculate the nestedness temperature of presence–absence matrices. Journal of Biogeography 33(5): 924-935. Roques, K.G., O’Connor, T.G. and Watkinson, A.R. 2001. Dynamics of shrub encroachment in an African savanna: relative influences of fire, herbivory, rainfall and density dependence. Journal of Applied Ecology 38: 268-280. Rosalino, L. and Santos-Reis, M. 2009. Fruit consumption by carnivores in Mediterranean Europe. Mammal Review 39: 67-78 Rotundo, J.L. and Aguiar, M.R. 2005. Litter effects on plant regeneration in arid lands: a complex balance between seed retention, seed longevity and soil-seed contact. Journal of Ecology 93: 829-838. Rovira, P. and Vallejo, V.R. 1997. Organic carbon and nitrogen mineralization under Mediterranean climatic conditions: The effects of incubation depth. Soil Biology and Biochemistry 29(9-10): 1509-1520 Ruppert, J. C., Harmoney, K., Henkin, Z., Snyman, H. A., Sternberg, M., Willms, W. and Linstädter, A. 2014. Quantifying drylands' drought resistance and recovery: the importance of drought intensity, dominant life history and grazing regime. Global Change Biology. doi: 10.1111/gcb.12777 Ruprecht, E., Enyedi, M.Z., Eckstein, R.L. and Donath, T.W. 2010. Restorative removal of plant litter and vegetation 40 years after abandonment enhances re- emergence of steppe grassland vegetation. Biological Conservation 143: 449-456. Saiz, H. and Alados, C.L. 2011a. Effect of Stipa tenacissima L. on the structure of plant co-occurrence networks in a semi-arid community. Ecological Research 26: 595-603. Saiz, H. and Alados, C.L. 2011b. Structure and spatial self-organization of semiarid communities through plant–plant co-occurrence networks. Ecological Complexity 8: 184-191.

248

References

Saiz, H., and Alados, C.L. 2014. Effect of livestock grazing in the partitions of a semiarid plant–plant spatial signed network. Acta Oecologica 59: 18-25. Sánchez-Gómez, D., Zavala, M.A. and Valladares, F. 2006. Seedling survival responses to irradiance are differentially influenced by low-water availability in four tree species of the Iberian cool temperate–Mediterranean ecotone. Acta Oecologica 30: 322-332. Sanders, H.L., 1968. Marine benthic diversity: a comparative study. American Naturalist 102: 243–282. Santos, T. and Tellería, J. 1997. Vertebrate predation on Holm Oak, Quercus ilex, acorns in a fragmented habitat: effects on seedling recruitment. Forest Ecology and Management 98(2): 181-187. Sayer, E.J. 2006. Using experimental manipulation to assess the roles of leaf litter in the functioning of forest ecosystems. Biological Review 81: 1-31. Schlatter, J.E. Gerding, V. and Calderón, S. 2006. Aporte de la hojarasca al ciclo biogeoquímico en plantaciones de Eucalyptus nitens, X Región, Chile. Bosque 27(2): 115-125. Schlesinger, W.H. and Pilmanis, A.M. 1998. Plant-soil interactions in deserts. Biogeochemistry 42: 169-187. Schlesinger, W.H., Abrahams, A.D., Parsons, A.J. and Wainwright, J. 1999. Nutrient losses in runoff from grassland and shrubland habitats in southern New Mexico: rainfall simulation experiments. Biogeochemistry 45: 21-34. Schmidt, M.W.I., Torn, M.S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I.A., Kleber, M., Koegel-Knabner, I., Lehmann, J., Manning, D.A.C., Nannipieri, P., Rasse, D.P., Weiner, S. and Trumbore, S.E., 2011. Persistence of soil organic matter as an ecosystem property. Nature 478: 49-56. Schöning, C., Espadaler, X., Hensen, I. and Roces, F. 2004. Seed predation of the tussock-grass Stipa tenacissima L. by ants (Messor spp.) in south-eastern Spain: the adaptive value of trypanocarpy. Journal of Arid Environments 56(1): 43-61. Schütz, W., Milberg, P. and Lamont, B.B. 2002. Germination requirements and seedling responses to water availability and soil type in four eucalypt species. Acta Oecologica 23(1): 23-30.

249

References

Scotti, M., Podani, J. and Jordan, F. 2007. Weighting, scale dependence and indirect effects in ecological networks: a comparative study. Ecological Complexity 4: 148-159. Serrasolas, I., Ferran, A. and Valleyo, V. R. 1989. Recostruccion de los horizontes organicos tras el incendio en la garriga del macizo de Garraf (Barcelona). Options Méditerranéennes, Série Séminaires, 3, 107-110. Servicio del Esparto. 1953. Estudios y experiencias sobre el esparto, vol. II. Ministerios de Industria y Comercio y de Agricultura, Madrid. Shannon, C.E. and Weaver, W., 1949. The Mathematical theory of Communication. University of Illinois Press, Urbana. Shearer, G., Duffy, J., Kohl. D.H. and Commoner, B., 1974. A steady state model of isotopic fractionation accompanying nitrogen transformations in soil. Soil Sci. Soc. Am. Proc. 38: 315-322. Smith, S. and Read, D. 2008. Mycorrhizal Symbiosis. Academic Press. (3rd ed.) Soil Survey Staff. 1994. Keys to Soil Taxonomy, Sixth Edition. USDA Soil Conservation Service, Pocahontas Press, Blacksburg, USA. 524 pp. Soliveres, S. Eldridge, D.J., Hemmings, F. and Maestre, F.T. 2012. Nurse plant effects on plant species richness in drylands: The role of grazing, rainfall and species specificity. Perspect. Plant Ecol. Evol. Syst. 14: 402-410. Soliveres, S., Eldridge, D., Maestre, F.T., Bowker, M.A., Tighe M. and Escudero, A., 2011a. Microhabitat amelioration and reduced competition among understorey plants as drivers of facilitation across environmental gradients: towards a unifying framework. Perspect. Plant Ecol. Evol. Syst. 13: 247-258. Soliveres, S., García-Palacios, P., Castillo-Monroy, A.P., Maestre, F.T., Escudero, A. and Valladares, F. 2011b. Temporal dynamics of herbivory and water availability interactively modulate the outcome of a grass–shrub interaction in a semiarid ecosystem. Oikos 120: 710-719. Stinchcombe, J.R. and Schmitt, J. 2006. Ecosystem engineers as selective agents: the effects of leaf litter on emergence time and early growth in Impatiens capensis. Ecology letters 9: 258-270.

250

References

Stohlgren, T.J., Barnett, D.T. and Kartesz, J.T. 2003. The rich get richer: patterns of plant invasions in the United States. Front. Ecol. Environ. 1: 11-14. Talbot, J.M. and Treseder, K.K. 2012. Interactions among lignin, cellulose, and nitrogen drive litter chemistry-decay relationships. Ecology 93(2): 345-354. Taylor, B.R., Parkinson, D. and Parsons, W.F.J. 1989. Nitrogen and Lignin Content as Predictors of Litter Decay Rates: A Microcosm Test. Ecology 70(1): 97-104. Tews, J., Esther, A., Milton, S.J. and Jeltsch, F. 2006. Linking a population model with an ecosystem model: Assessing the impact of land use and climate change on savanna shrub cover dynamics. Ecol. Model. 95: 219-228. Thébault, E. and Fontaine, C. 2010. Stability of ecological communities and the architecture of mutualistic and trophic networks. Science 329: 853-856. Thompson, S.E., Katul, G.G. and MacMahon, S. 2008. Role of biomass spread in vegetation pattern formation within arid ecosystems. Water Resour. Res. 44: E10421. Tighe, M., Reid, N., Wilson, B. and Briggs, S.V. 2009. Invasive native scrub and soil condition in semiarid south-eastern Australia. Agriculture, Ecosystems and Environment 132(3-4): 212-222. Tilman, D. 1993. Species richness of experimental productivity gradients: how important is colonization limitation?. Ecology 74(8): 2179-2191. Timmermann, A., Oberhuber, J., Bacher, A., Esch, M., Latif, M. and Roeckner, 1999. Increased El Nino frequency in a climate model forced by future greenhouse warming. Nature 398: 694.697. Tirado, R. and Pugnaire, F.I. 2005. Community structure and positive interactions in constraining environments. Oikos 111: 437-444. Tongway, D. J. and Hindley, N. L. 2004. Landscape function analysis manual: procedures for monitoring and assessing landscapes with special reference to minesites and rangelands. CSIRO Sustainable Ecosystems, Canberra, Australia. Tongway, D. J., and Ludwig, J. A. 1996. Rehabilitation of semiarid landscapes in Australia. I. Restoring productive soil patches. Restoration Ecology 4(4): 388-397.

251

References

Tongway, D.J. and Ludwig, J.A. 1994. Small scale resource heterogeneity in semiarid landscapes. Pacific Conservation Biology 1: 201-208. Traveset, A., Heleno, R., Chamorro, S., Vargas, P., McMullen,C.K., Castro-Urgal, R., Nogales, M., Herrera, H.W. and Olesen, J.M. 2013 Invaders of pollination networks in the Galápagos Islands: emergence of novel communities. Proceedings of the Royal Society B: Biological Sciences 280: 20123040. Trubat, R., Cortina, J. and Vilagrosa, A. 2008. Short-term nitrogen deprivation increases field performance in nursery seedlings of Mediterranean woody species. Journal of Arid Environments 72(6): 879-890. Trubat, R., Cortina, J. and Vilagrosa, A. 2011. Nutrient deprivation improves field performance of woody seedlings in a degraded semi-arid shrubland.Ecological Engineering 37(8): 1164-1173. Tylianakis, J.M., Laliberté, E., Nielsen, A. and Bascompte, J. 2010. Conservation of species interaction networks. Biol. Conserv. 143: 2270-2279. UNCCD. 1994. United Nations Earth Summit. Convention on Desertification. UN Conf. on Environmental and Development, Rio de Janeiro, Brazil, June 3-14, 1992. DPI/SD/1576. New York: United Nations. UNESCO. 1997. World Map of Arid Regions. United Nations Educational, Scientific and Cultural Organization, Paris. Valdecantos, A. 2001. Aplicación de fertilizantes orgánicos e inorgánicos en la repoblación de zonas forestales degradadas de la Comunidad Valenciana. Tesis doctoral. Universitat d’Alacant. Valdecantos, A., Fuentes, D., Smanis, A., Llovet, J., Morcillo, L. and Bautista, S. 2014. Effectiveness of Low‐Cost Planting Techniques for Improving Water Availability to Olea europaea Seedlings in Degraded Drylands. Restoration ecology 22(3): 327-335. Valentin, C., d’Herbès, J.M. and Poesen, J. 1999. Soil and water components of banded vegetation pattern. Catena 37: 1-24. Valiente-Banuet, A. and E. Ezcurra. 1991. Shade as a cause of association between the cactus Neobuxbaumia tetetzo and the nurse plant Mimosa luisana in the Tehuacán Valley, Mexico. Journal of Ecology 79: 961–971

252

References

Valiente-Banuet, A., Rumebe, A. V., Verdú, M. and Callaway, R. M. 2006. Modern Quaternary plant lineages promote diversity through facilitation of ancient Tertiary lineages. Proceedings of the National Academy of Sciences 103(45): 16812-16817. - diversity of plant communities. Ecology Letters 10(11): 1029-1036. Valiente, J. A., Estrela, M. J., Corell, D., Fuentes, D., Valdecantos, A. and Baeza, M. J. 2011. Fog water collection and reforestation at a mountain location in a western Mediterranean basin region: Air-mass origins and synoptic analysis. Erdkunde 65, 277-290. Valladares, F. and Pearcy, R.W., 2002. Drought can be more critical in the shade than in the sun: a field study of carbon gain and photo-inhibition in a Californian shrub during a dry El Niño year. Plant Cell Environ. 25: 749-759. Valladares, F. and Niinemets, Ü. 2008. Shade tolerance, a key plant feature of complex nature and consequences. Annual Review of Ecology, Evolution, and Systematics: 237-257. Vallejo, V.R., Allen, E.B., Aronson, J., Pausas, J.G., Cortina, J. and Gutiérrez, J.R. 2012. Restoration of mediterranean-type woodlands and shrublands. Chapter 11 in Restoration Ecology: The New Frontier (2nd Edition). Pp. 130-144. J. Van Andel and J. Aronson (Eds.). John Wiley and Sons Ltd-Blackwell. Bognor Regis. Vallejo, V.R., Cortina, J., Ferrán, A., Fons, J., Romanyà, J. and Serrasolsas, I. 1998. Sobre els trets distintius dels sòls mediterranis. Acta Botanica Barcinonensia 45 (Homenatge a Oriol de Bolòs): 603-632. Van Auken, O. W. 2009. Causes and consequences of woody plant encroachment into western North American grasslands. Journal of Environmental Management 90(10): 2931-2942. Van Auken, O.W. 2000. Shrub invasions of North American semiarid grasslands. Annual Review of Ecology and Systematics 31: 197-215. Van Dijk, A.I.J.M. and Keenan. R. 2007. Planted forests and water in perspective. Forest Ecology and Management 251(1-2): 1-9

253

References

Vasudevan, P., Thlapiyal, A., Sen, P.K., Dastidar, M.G. and Davies, P. 2011. Buried clay pot irrigation for efficient and controlled water delivery. Journal of Scientific and Industrial Research 70: 645–652. Vázquez-Yanes, C. and Orozco-Segovia, A. 1992. Effects of litter from a tropical rainforest on tree seed germination and establishment under controlled conditions. Tree Physiology 11: 391-400. Vázquez-Yanes, C., Orozco-Segovia, A., Rincón, E., Sanchezcoronado, M.E., Huante, P., Toledo, J.R. and Barradas, V.L. 1990. Light beneath the litter in a tropical forest : effect on seed germination. Ecology 71: 1952-1958. Vázquez, D.P., Morris, W.F. and Jordano, P. 2005, Interaction frequency as a surrogate for the total effect of animal mutualists on plants. Ecology Letters 8: 1088–1094. Vellend, M. 2008. Effects of diversity on diversity: consequences of competition and facilitation. Oikos 117: 1075-1085. Verdú M. and García-Fayos, P. 1996. Nucleation processes in a Mediterranean bird-dispersed plant. Functional Ecology 10(2): 275-280. Verdu, M. and Garcia-Fayos, P. 2002. Ecología reproductiva de Pistacia lentiscus L. (Anacardiaceae): un anacronismo evolutivo en el matorral mediterráneo. Revista Chilena de Historia Natural 75(1): 57-65. Verdú, M. and Valiente-Banuet, A. 2008. The nested assembly of plant facilitation networks prevents species extinctions. American Naturalist 172: 751-760. Verdú, M. and García-Fayos, P. 2003. Frugivorous birds mediate sex-biased facilitation in a dioecious nurse plant. Journal of Vegetation Science 14(1): 35-42. Verdú, M., Dávila, P., García‐Fayos, P., Flores‐Hernández, N., and Valiente‐Banuet, A. 2003. ‘Convergent’ traits of mediterranean woody plants belong to pre‐mediterranean lineages. Biological Journal of the Linnean Society 78(3): 415- 427. Verdú, M., Jordano, P. and Valiente-Banuet, A. 2010. The phylogenetic structure of plant facilitation networks changes with competition. Journal of Ecology 98: 1454–1461.

254

References

Vetaas, O.R. 1992. Micro-site effects of trees and shrubs in dry savannas. J. Veg. Sci. 3_ 337-344. Vilagrosa, A., Cortina, J., Gil‐Pelegrín, E. and Bellot, J. 2003. Suitability of drought‐preconditioning techniques in Mediterranean climate. Restoration Ecology 11(2): 208-216. Vilagrosa, A., Caturla, R. N., Hernández, N., Cortina, J., Bellot, J. and Vallejo, V. R. 2001. Reforestación en ambiente semiárido del sureste peninsular. Resultados de las investigaciones desarrolladas para optimizar la supervicencia y el crecimiento de especies autóctonas. In SECF-Junta de Andalucía (Eds.), Actas del III Congreso Forestal Español. Montes para la sociedad del nuevo milenio (Volume 2, pp. 213- 219). Sevilla, Spain. Villegas, J.C., Breshears, D.D., Zou, Z.B. and Law, D.J. 2010. Ecohydrological controls of soil evaporation in deciduous drylands: How hierarchical effects of litter, patch and vegetation mosaic cover interact with phenology and season. Journal of Arid Environments 74: 595-602. Violle, C., Navas, M.L., Vile, D., Kazakou, E., Fortunel, C., Hummel, I. and Garnier, E. 2007. Let the concept of trait be functional! Oikos 116: 882-892. Vitousek, P.M. 1982. Nutrient cycling and nutrient-use efficiency. American Naturalist 119: 553-572. Vitousek, P.M., Gerrish, G., Turner, D.R., Walker, L.R. and Mueller-Dumbois, D. 1995. Litterfall and nutrient cycling in four Hawaiian montane rainforest. Journal of Tropical Ecology 11: 189-203. Vivanco, L. and Austin, A.T. 2006. Intrinsic effects of species on leaf litter and root decomposition: a comparison of temperate grasses from North and South America. Oecologia 150: 97-107. Von Holle, B., 2005. Biotic resistance to invader establishment of a southern Appalachian plant community is determined by environmental conditions. J. Ecol. 93: 16-26. Walela, C., Daniel, H., Wilson, B., Lockwood, P., Cowie, A. and Harden, S. 2014. The initial lignin: nitrogen ratio of litter from above and below ground sources

255

References strongly and negatively influenced decay rates of slowly decomposing litter carbon pools. Soil Biology and Biochemistry 77: 268-275. Wardle, D.A. and Lavelle, P. 1997. Linkages between soil biota, plant litter quality and decomposition. En: Cadish, G. y0Giller, K.E. (Eds.). Driven by Nature: Plant Litter Quality and Decomposition. CAB International. Wardle, D.A., Bardgett, R.D., Klironomos, J.N., Setala, H., van der Putten, W.H. and Wall, D.H. 2004. Ecological Linkages Between Aboveground and Belowground Biota. Science 304(5677): 1629-1633. Webb, C. O., Ackerly, D.D., McPeek, M.A. and Donoghue, M.J. 2002. Phylogenies and community ecology. Annu. Rev. Ecol. Syst. 33: 475-505. Weigelt, A., Schumacher, J., Walther, T., Bartelheimer, M., Steinlein, T. and Beyschlag, W., 2007. Identifying mechanisms of competition in multi-species communities. J. Ecol. 95: 53-64. Welbourn, M.L., Stone, E.L. and Lassoie, J.P. 1981. Distribution of net litter inputs with respect to slope position and wind direction. Forest Scienc 27: 651-659. Whisenant, S. G., Thurow, T. L. and Maranz, S. J. 1995, Initiating Autogenic Restoration on Shallow Semiarid Sites. Restoration Ecology 3: 61-67. Whitford, W.G. 2002. Ecology of desert systems. London: Academic. Whitham, T.G., Young, W.P., Martinsen, G.D., Gehring, C.A., Schweitzer, J.A., Shuster S.M., Wimp, G.M., Fischbr, D.G., Bailey J.K., Lindroth, R.L., Woolbright, S. and Kuske, C.R. 2003. Community and ecosystem genetics: A consequence of the extended phenotype: Community Genetics. Ecology 84(3): 559-573. Williams M.C. and Wardle G.M. 2007. Pine and eucalypt litterfall in a pine-invaded eucalypt woodland: the role of fire and canopy cover. Forest Ecology Management 253: 1-10. Wunderlee, J.M. 1997. The role of animal seed dispersal in accelerating native forest regeneration on degraded tropical lands. Forest Ecology and Management 99: 223-235. Xiong, S. and Nilsson, C. 1997. Dynamics of leaf litter accumulation and its effects on riparian vegetation: a review. Botanical Review 63: 240-264.

256

References

Xiong, S. and Nilsson, C. 1999. The effects of plant litter on vegetation: a meta- analysis. Journal of Ecology 87: 984-994. Zaady, E., Gutterman, Y. and Boeken, B.,1997. The germination of mucilaginous seeds of Plantago coronopus, Reboudia pinnata, and Carrichtera annua on cyanobacterial soil crust from the Negev Desert. Plant and Soil 190: 247-252. Zapata, V. M., Robledano, F., Ramos, V. and Martínez-López, V. 2014. Bird- mediated seed dispersal of fleshy fruits of mediterranean shrubs in semiarid forest patches: the role of Pinus halepensis Miller trees as seed receptors. Plant Ecology 215(11): 1337-1350. Zarovalli, M.P. Yiakoulaki, M.D. and Papanastasis, V.P. 2007. Effects of shrub encroachment on herbage production and nutritive value in semiarid Mediterranean grasslands. Grass and Forage Science 62: 355-363. Zhang, M.J., Liu, M., Li, Y., Xu, C. and An, S. 2011. The combined positive effects of two dominant species in an arid shrub-herbaceous community: implications from the performance of two associate species. Plant Ecol. 212: 1419-1428.

257

RESUMEN

Dinámica de parches de vegetación leñosa en ecosistemas semiáridos

DINÁMICA DE PARCHES DE VEGETACIÓN LEÑOSA EN ECOSISTEMAS SEMIÁRIDOS DEL SURESTE DE LA PENÍNSULA IBÉRICA

Los ecosistemas áridos y semiáridos cubren más del 30% de las tierras emergidas del planeta y dan cobijo a más de un tercio de su población. La vegetación en estos ecosistemas con frecuencia se organiza en un mosaico de parches de vegetación herbácea y leñosa inmersos en una matriz de suelo desnudo, costra biológica y plantas herbáceas. La dinámica de estos ecosistemas depende principalmente de los pulsos de agua, escasos y espaciados en el tiempo, y de las interacciones entre las especies. En los últimos 150 años se ha registrado un incremento en la cobertura y densidad de especies leñosas en diversos ecosistemas semiáridos. Este fenómeno, conocido como matorralización (shrub encroachment), tiene importantes implicaciones en el funcionamiento del ecosistema y la provisión de bienes y servicios, por lo que ha sido ampliamente estudiado. No obstante, no hay un claro consenso sobre las causas y consecuencias de este fenómeno a gran escala. Por ejemplo, la matorralización se ha relacionado con un incremento en la evapotranspiración y una reducción en la producción de forraje y de la biodiversidad. Por el contrario, el incremento en la cobertura de especies leñosas también se ha asociado a un aumento de la riqueza de plantas vasculares, aves y microorganismos del suelo, así como a una mayor fertilidad del suelo. A pesar de la importancia de la matorralización, nuestro conocimiento de este proceso en zonas mediterráneas es escaso. Hay evidencias de que la presencia de especies leñosas en estas zonas tiene un efecto positivo en la composición de la comunidad y el funcionamiento del ecosistema. Por ello, en las últimas décadas las políticas de restauración de zonas degradadas se han dirigido a reforzar las poblaciones de especies leñosas, especialmente especies arbóreas y arbustivas de gran porte. Sin embargo, la composición y la dinámica de parches de vegetación leñosa y su impacto en el reclutamiento de especies clave son aún desconocidos.

Los ecosistemas semiáridos mediterráneos están formados por plantas herbáceas perennes, matorral y bosque abierto de especies de Pinus sp., Quercus sp. y Olea europea L. En la parte oeste de la cuenca Mediterránea las zonas semiáridas están cubiertas principalmente por esparto (Stipa tenacissima L.) ocupando un total de 70 000 km2. En el sureste de la Península Ibérica, los

261

Resumen espartales cubren 6 000 km2, y desde hace siglos han sido utilizados por las sociedades humanas. Hasta mediados del siglo XX la mayor parte de tierra apta para la agricultura era cultivada con especies de secano, o bien se explotaba el propio esparto como fuente de fibra. En consecuencia, las especies leñosas eran extraídas o utilizadas como combustible relegando a sus poblaciones a desarrollarse en los márgenes de los cultivos o en áreas no cultivables como afloramientos rocosos. Más tarde, cuando el cultivo de estas zonas y la explotación del esparto fueron abandonados, la recuperación de la cubierta leñosa fue heterogénea. En las terrazas de cultivo situadas en las partes más bajas de los valles, procesos de nucleación han favorecido el reclutamiento de especies leñosas. Por el contrario, el reclutamiento espontáneo de estas especies en las laderas cubiertas por esparto ha sido muy bajo. Esta historia de uso del suelo ha dado lugar a la actual configuración de los espartales: una matriz de suelo desnudo y matas de esparto con parches de vegetación leñosa y pequeños arbustos dispersos. La colonización de las laderas por especies arbustivas depende principalmente de la disponibilidad de agua, pero también de la presencia de fauna dispersadora, mayormente aves y mamíferos.

Los parches de vegetación leñosa son grupos de arbustos de gran tamaño cuya fisionomía es diferente a la del resto de la matriz en espartales. Estos parches están formados por hasta seis especies rebrotadoras (en adelante especies dominantes) Rhamnus lycioides L., Pistacia lentiscus L., Quercus coccifera, Juniperus oxycedrus, Osyris lanceolata y Ephedra fragilis L. Desf. y un conjunto de especies herbáceas y arbustos de menor tamaño (en adelante especies acompañantes). A pesar de la relativamente baja cobertura de las especies dominantes, éstas tienen un papel clave en la composición, funcionamiento y estabilidad de espartales semiáridos mediterráneos. La presencia de estos arbustos ha sido positivamente relacionada con la riqueza y diversidad de plantas vasculares, de aves y de pequeños invertebrados. Asimismo, los parches de vegetación leñosa se han asociado a altos niveles de C orgánico, N total y N disponible en suelo, mayor infiltración hídrica y reciclado de nutrientes, y una mejora de las propiedades del suelo. Por todo ello, recientemente se ha potenciado el uso de estas especies en programas de restauración. La formación espontánea de estos parches podría darse gracias a interacciones positivas (facilitación) entre especies, fenómeno frecuente en ecosistemas semiáridos. La

262

Dinámica de parches de vegetación leñosa en ecosistemas semiáridos facilitación es dependiente de la identidad de las especies que interactúan, y puede ser función de la composición y estructura del parche, de la disponibilidad de recursos y del estrés abiótico. Los parches de vegetación leñosa pueden por tanto actuar como comunidades facilitadoras, cuya capacidad para la facilitación o la interferencia neta dependerá de las propiedades emergentes de la comunidad. Por lo tanto, los parches de vegetación leñosa pueden considerarse redes de interacciones entre especies, generando distintos patrones de biodiversidad según la localización geográfica, condiciones ambientales, historia de uso del suelo, y factores propios de la comunidad. El análisis de los patrones de coocurrencias de estas especies proporciona información relevante sobre interacciones específicas a largo plazo y los factores que modulan la composición de estas comunidades y su capacidad para perdurar en diversos escenarios.

A pesar de la importancia de los arbustos rebrotadores en ecosistemas semiáridos y su creciente uso en programas de restauración, existe muy poca información sobre la estructura y dinámica de sus poblaciones en la Península Ibérica. La distribución y composición de parches de vegetación leñosa en el sureste ibérico o los patrones de ensamblaje de especies y las propiedades emergentes de las comunidades que forman siguen siendo desconocidas. Igualmente, tampoco se ha evaluado la relación entre las características de estos parches y su capacidad para modificar el ambiente que los rodea y modular la interacción entre especies. Para abordar estas cuestiones, he realizado diversos estudios que conforman esta tesis con el PRINCIPAL OBJETIVO de evaluar la composición y dinámica de los parches de vegetación leñosa en espartales semiáridos del sureste de la Península Ibérica. Para alcanzar este objetivo general he establecido los siguientes OBJETIVOS PARCIALES:

1) Describir las principales características de los parches de vegetación leñosa. 2) Utilizar la teoría de redes para estudiar la composición de parches y sus propiedades emergentes como comunidades. 3) Explorar los factores bióticos y abióticos que afectan a las estructura de la red y sus implicaciones en la gestión. 4) Evaluar la capacidad de los parches como comunidades para facilitar el establecimiento de nuevos individuos y los factores que la regulan.

263

Resumen

5) Analizar el efecto de la acumulación de horizontes orgánicos de especies leñosas en la germinación de dos especies clave en espartales.

La tesis se estructura en cinco capítulos que abordan los objetivos 1 a 5 respectivamente, y finalmente, en una última sección se hace una discusión global y se enumeran las principales conclusiones.

En el CAPÍTULO 1 se describe la principal estructura física y biológica de 450 parches de vegetación leñosa distribuidos en 15 cuencas de espartales semiáridos en el sur de la provincia de Alicante. Además, en este capítulo se explica la metodología general utilizada para medir variables bióticas y abióticas utilizada en este y sucesivos capítulos de la tesis, para la caracterización de las zonas a nivel de cuenca, unidad homogénea de laderas, terraza de cultivo abandonado y parche.

En este capítulo se comprueba que las especies en los parches pueden identificarse como dominantes o acompañantes, de acuerdo a rasgos morfo- funcionales y a su capacidad para formar parches por ellas mismas. Los parches están formados por 1-26 especies acompañantes y 1-5 especies dominantes siendo siempre una de estas más abundante que el resto. Además, las especies dominantes son las que más influyen en la composición del parche dando lugar a dos gradientes principales relacionados con la abundancia de cuatro especies: el gradiente R. lycioides - Q. coccifera y el gradiente P. lentiscus - E. fragilis. Además, existe un tercer gradiente formado por las dos especies acompañantes de mayor influencia: S. tenacissima y Brachypodium retusum. La riqueza y cobertura de especies acompañantes debajo de los parches es mayor o igual que en su periferia inmediata, dependiendo de la especie dominante, lo que sugiere que las diferencias en las condiciones creadas por los parches podrían ser suficientes para fomentar la diversidad y el desarrollo de otras especies. Los parches dominados por Q. coccifera y P. lentiscus son los más grandes, con un área de copa de entre 15-30 m2, sin embargo, los parches más pequeños son relativamente más ricos que los parches más grandes, sugiriendo una saturación de especies en parches grandes debida, probablemente, a una disminución en su capacidad para crear nuevos microhábitats. Los parches dominados por R. lycioides son los más abundantes, tanto en laderas como en terrazas de cultivo, y especialmente bajo temperaturas elevadas y precipitaciones escasas. Por el contrario, en los sitios

264

Dinámica de parches de vegetación leñosa en ecosistemas semiáridos más fríos, los parches dominados por Q. coccifera y J. oxycedrus son los que predominan. La relación entre los factores climáticos cuyo rango no es muy amplio, y la abundancia de uno u otro tipo de parches indica que las especies formadoras de parches pueden ser muy sensibles a pequeñas variaciones climáticas. Teniendo en cuenta esta relación y las predicciones climáticas para el sureste de la Península Ibérica en las próximas décadas, pequeños cambios climáticos conducirían a modificaciones graduales en la composición de especies de los espartales semiáridos hacia una mayor abundancia de parches dominados por R. lycioides y E. fragilis, en detrimento de parches dominados por Q. coccifera y J. oxycedrus. Además, esta teoría se ve apoyada por el hecho de que R. lycioides es la especie dominante con mayor reclutamiento.

En el CAPÍTULO 2 se aborda el estudio de la composición de parches de vegetación leñosa desde una perspectiva de redes de interacciones entre especies. Para ello consideramos los parches como comunidades donde especies dominantes y acompañantes interaccionan con distinta intensidad según su frecuencia de coocurrencia en los parches. Obtuvimos índices propios de la estructura de las redes para evaluar las propiedades emergentes de 27 comunidades de parches de vegetación leñosa, correspondientes a unidades homogéneas de vegetación de laderas de espartales. Así, estas comunidades se caracterizaron por tener una elevada conectividad, una moderada modularidad, una baja especialización y una baja estructura anidada, en comparación con otras redes ecológicas como las mutualistas o las tróficas. Estos resultados sugieren que las especies que forman las comunidades de parches son generalistas, con menos restricciones morfológicas, fenológicas y fisiológicas para establecer redes que aquellas, y por tanto son más robustas en el mantenimiento de su estructura. Las redes de coocurrencias estudiadas alcanzaron un tamaño de entre 16 y 45 especies, y entre 2 y 4 módulos (subcomunidades).

En este capítulo se analizó también el grado de conectividad y especialización de cada especie y su papel en la estructura de la comunidad como especies centrales (las más generalistas), conectoras de módulos o periféricas (las más especialistas). Las especies más conectadas y menos especializadas fueron la especie dominante R. lycioides y las especies acompañnates S. tenacissima y B. retusum, aunque hubo bastante heterogeneidad entre las redes estudiadas. Por

265

Resumen otro lado, J. oxycedrus, E. fragilis, O. lanceolata y la mayor parte de especies acompañantes tuvieron un alto nivel de especialización y, en general, su papel en las redes fue de especies periféricas o conectoras de módulos, apareciendo sólo en presencia de otras especies. Rhamnus lycioides y P. lentiscus fueron las especies que mostraron un papel más céntrico en la comunidad o en las subcomunidades de una comunidad, estableciendo mayor número de conexiones con el resto de especies. Estas especies clave, pueden ser especialmente importantes cuando se pretenda potenciar la biodiversidad. Otras especies dominantes también actuaron como especies clave centrales, y por el contrario las especies acompañantes nunca ejercieron este papel.

El estudio de las redes de coocurrencias de plantas se profundizó en el CAPÍTULO 3, explorando los factores endógenos y exógenos que afectan a su estructura y composición. Se analizaron variables climáticas y factores bióticos y abióticos de los sitios de estudio y de los parches de vegetación, y su relación con tres índices de estructura de las redes: conectividad, modularidad y anidamiento. Tanto factores exógenos (temperatura y cobertura de rocas) como endógenos (tamaño de parches, composición y riqueza de la comunidad) afectaron a la estructura de las redes. La temperatura media anual afectó a los tres índices estudiados, incrementando la conectividad y reduciendo la modularidad y el anidamiento. La modularidad también se vio afectada por la cobertura de rocas (positivamente) y por el tamaño medio de los parches (negativamente). Las especies de los parches se mostraron más conectadas cuando los parches eran de mayor tamaño. El anidamiento en la estructura de interacciones de las redes fue el único índice afectado por la composición de los parches, incrementado por la presencia de R. lycioides, P. lentiscus y S. tenacissima y por una mayor riqueza específica en parches.

La información obtenida en los capítulos 2 y 3 es relevante para evaluar la capacidad de estas comunidades de soportar perturbaciones o nuevos escenarios de cambio climático, y para diseñar prácticas adecuadas de gestión que promuevan la biodiversidad y la provisión de servicios ecosistémicos.

En el CAPÍTULO 4 se evaluó la capacidad de los parches de vegetación leñosa para facilitar la entrada de nuevos individuos desde una perspectiva de efecto conjunto de la comunidad. Para ello se utilizaron brinzales de Pistacia

266

Dinámica de parches de vegetación leñosa en ecosistemas semiáridos lentiscus que fueron plantados en tres localizaciones respecto a cada parche: debajo de la copa, en el borde norte y en el borde sur de la periferia de la proyección de la copa. Durante dos años comparamos su supervivencia y crecimiento con el de brinzales introducidos en áreas libres de la influencia de los parches. Además se analizaron las características de la comunidad implicadas en el establecimiento de los nuevos individuos.

Tanto la concentración de carbono orgánico en suelo como la de nitrógeno total fueron mas elevadas en el área inmediantamente subyacente a la copa de los parches, comparadas con la zona periférica o con zonas abiertas. Por el contrario, el contenido hídrico del suelo fue mayor en espacios abiertos que bajo los parches. La supervivencia de brinzales después de dos años fue notablemente mayor bajo los parches que en zonas abiertas. Las plántulas situadas debajo de los parches mostraron un mayor contenido de nitrógeno foliar y un mayor enriquecimiento en 15N en comparación con el resto de localizaciones, y su eficiencia en el uso del agua (medida mediante el enriquecimiento foliar en el isótopo estable 13C) fue menor. Estos resultados indican que debajo de los parches de vegetación leñosa se concentra una elevada fertilidad edáfica, y se presenta una tasa de reciclado de nutrientes elevada, una baja demanda evaporativa y un nivel de estrés hídrico bajo. El hecho de que estos valores cambien drásticamente en la periferia de los parches sugiere que el efecto de los parches está estrictamente restringido por el área de copa, no prolongándose más allá de ésta.

La supervivencia de brinzales estuvo determinada por los atributos físicos y bióticos de los parches. La composición de la comunidad de especies acompañantes, concretamente una mayor cobertura de S. tenacissima y una menor cobertura de B. retusum, favorecieron la supervivencia de los brinzales. La distancia filogenética de la comunidad a la especie plantada, la cobertura de especies acompañantes y la profundidad de horizontes orgánicos también tuvieron un efecto positivo en la supervivencia. No obstante, la importancia de los atributos de la comunidad en la supervivencia de brinzales y el tipo de estrés que éstas sufrieron dependió de la localización de plantación. Así, las plántulas situadas debajo de los parches sufrieron menor estrés y su supervivencia dependió principalmente de la concentración de carbono orgánico en el suelo y

267

Resumen de la composición de especies acompañantes. Por otro lado, la supervivencia de las plántulas situadas en el borde norte de los parches pareció estar sujeta a mayor estrés, principalmente de tipo biótico. Los factores que afectaron a su supervivencia fueron todos de tipo biótico como la composición y riqueza de especies acompañantes, y la composición y distancia filogenética de especies dominantes. No se detectaron factores que afectaran significativamente a las plántulas situadas en el borde sur de los parches, por lo que su supervivencia podría verse afectada por agentes de estrés de tipo abiótico. Los resultados de este estudio demostraron la heterogeneidad existente dentro de un mismo parche y la importancia de considerar el conjunto de la comunidad en la evaluación de interacciones ecológicas.

El objetivo del CAPÍTULO 5 fue evaluar el papel de los horizontes orgánicos en las interacciones planta-planta en los parches de vegetación leñosa. Los horizontes orgánicos pueden jugar un papel importante en las interacciones que se producen en el seno de los parches, ya que las especies dominantes en estos parches con frecuencia muestran una contrastada capacidad para acumular horizontes orgánicos. Esto, unido a la heterogeneidad espacial en el desarrollo de las especies acompañantes y en el reclutamiento de las especies dominantes, sugiere que las interacciones entre especies dominantes y acompañantes son complejas y los horizontes orgánicos podrían tener un papel relevante en los primeros estadios de vida de los nuevos individuos. Por ello, en este capítulo se evaluó el efecto de cinco tipos diferentes de horizontes orgánicos de especies leñosas sobre la germinación de dos especies abundantes e importantes en los espartales de S. tenacissima: una especie dominante (P. lentiscus) y una acompañante (B. retusum). Se hicieron tres experimentos de laboratorio en el que se combinaron los efectos mecánicos y aleloquímicos de los horizontes orgánicos, situando las semillas (i) por encima de estos, (ii) entre estos y el suelo mineral, y (iii) sobre suelo mineral pero regadas con extractos de horizontes orgánicos.

Quercus coccifera y P. halepensis acumulan mayor cantidad de horizontes orgánicos que R. lycioides. Esto puede responder, por una parte, a la edad de los individuos, ya que frecuentemente es mayor en Q. coccifera y P. halepensis que en R. lycioides. Y por otra parte, a que el balance de entrada de hojarasca y descomposición de horizontes orgánicos, probablemente baja, como en otras

268

Dinámica de parches de vegetación leñosa en ecosistemas semiáridos zonas semiáridas, favorezca la acumulación de horizontes orgánicos en algunas especies. Sin embargo, el contenido hídrico de estos horizontes en las especies R. lycioides y P. lentiscus fue mayor que en el resto de especies, e inversamente relacionado con la hidrofobicidad de los mismos.

De los tres experimentos de germinación descritos en este capítulo, el mayor efecto de horizontes orgánicos apareció cuando las semillas fueron sembradas sobre éstos, lo cual podría reflejar la existencia de una barrera física para la germinación. En general observamos un efecto negativo de los horizontes orgánicos de todas las especies sobre la capacidad de germinación de ambas especies sembradas, excepto para los horizontes orgánicos de R. lycioides, que tuvieron un efecto positivo sobre la germinación de semillas de P. lentiscus al compararlo con la siembra realizada directamente sobre el suelo mineral. El efecto de los horizontes orgánicos de R. lycioides sobre la germinación de semillas de B. retusum fue neutro, al igual que el efecto de los horizontes de P. lentiscus sobre la germinación de semillas de la misma especie.

Cuando las semillas fueron sembradas entre los horizontes orgánicos y el suelo mineral, el efecto de estos horizontes sobre la germinación fue negativo en cuatro de las cinco especies, y neutro en el caso de horizontes orgánicos de R. lyciodes. En el experimento en el que las semillas fueron sembradas sobre suelo mineral y regadas con extractos de hojarasca de las distintas especies no se observó ningún efecto de estos extractos, lo que sugiere que no existen efectos alelopáticos significativos que afecten a la germinación de P. lentiscus y B. retusum en estos ecosistemas. Las implicaciones de estos resultados a nivel de comunidad afectarían principalmente a especies cuya dispersión es a corta distancia. No obstante, como se ha demostrado en capítulos anteriores, los parches de vegetación leñosa ejercen un efecto positivo en los individuos, desarrollándose en sus inmediaciones en fases posteriores. Esto sugiere que el efecto negativo de los horizontes orgánicos de especies leñosas sobre la germinación podría ser compensado por ese efecto positivo, y la germinación ocurriría en momentos de baja acumulación de horizontes orgánicos o tras perturbaciones que removieran estos horizontes.

Los resultados presentados en esta tesis tienen implicaciones en la gestión de los espartales semiáridos. Por un lado, se ha identificado especies clave

269

Resumen en la diversidad y complejidad de comunidades, como R. lycioides, P. lentiscus o S. tenacissima, deseables cuando el objetivo de las prácticas de gestión sea incrementar la biodiversidad y mejorar la estructura de la comunidad. Por otro lado, se ha destacado la baja capacidad de alguna de estas especies clave para reclutar de manera espontánea, aspecto a tener en cuenta en la selección de especies para reforzar las poblaciones de especies clave. Además, se ha descrito el potencial facilitador de parches de vegetación leñosa para el establecimiento de nuevos individuos, pudiendo por tanto ser utilizados en prácticas de restauración. Asimismo, se ha destacado la heterogeneidad a nivel de parche, identificando las características a tener en cuenta en la selección del tipo de parche a utilizar como comunidad benefactora, y la importancia de la localización exacta de plantación en el desarrollo de nuevos individuos. Finalmente, se ha discutido la tendencia en la composición de especies formadoras de parches en un contexto de cambio climático, aspecto a tener en cuenta en actuaciones de restauración a largo plazo.

En esta tesis he realizado un estudio global sobre el estado de los parches de vegetación leñosa en ecosistemas semiáridos de la Península Ibérica. Los resultados presentados en esta tesis, pueden ser comparables con otros estudios en ecosistemas similares de otras partes del mundo, donde la existencia de parches de vegetación leñosa y el aumento de su cobertura es materia actual de estudio. Por otro lado, la descripción de la composición y estructura de estos parches, sirve como punto de partida para futuros estudios que profundicen en el papel de los mismos en el ecosistema y en dinámicas a más largo plazo.

270