The RAT is composed of a series of questions that are answered by an expert in contact with land users and land managers, using all available sources of knowledge including, but not limited to, scientific studies. It is designed as a tool to organize and merge knowledge, rather than producing new, in coherence with the approach of the WOCAT Framework for Documentation and Evaluation of Sustainable Land Management Technologies (Liniger et al. 2008). Based on a review of the scientific literature on resilience and the information gained in »Documented and evaluated natural resource management practices, we identified the most important characteristics that a Resilience Assessment Tool should have:

• Holistic approach: To understand the resilience mechanisms and the response of the system to disturbance an assessment cannot focus only on the part of the system that is mostly affected by disturbances (e.g. vegetation or soil in the case of the CASCADE study sites), as feedback mechanisms and systemic relations play a crucial role in resilience (Walker & Meyers 2004). In particular, it should include human actions, in terms of land use and land management. First because even in natural and semi-natural ecosystems, the role of land use and land management has a crucial impact on its evolution, secondly because they have the power to change how the land is managed beyond the project timeframe.
• Flexibility: Using a rigid set of indicators generally relevant for resilience would limit the contextualization of the assessment, reducing the validity of the results for the particular system analyzed (Wiesmann & Hurni 2011). Moreover, it should be flexible in terms of scale, as the appropriate system to consider for land management may vary greatly depending on historic, socio-economic and natural processes.
• Trans-disciplinarity: Any effective land management strategy has to meet land user’s needs; otherwise the system is at risk of abandonment or change of land use, which can bring the system to a regime shift. Moreover, local knowledge can give important insights on how processes develop in the specific context of the study site (Berkes et al. 1995). In particular, knowledge about how disturbances impact the system and on the effects of land management on these processes are very relevant for scientists.
• Build on existing knowledge: Most often, information to improve the management of the system is available, but scattered in different sources (Liniger et al. 2002). Rather than focusing on investigating a specific knowledge gap, a resilience assessment should support collecting information from different sources, highlighting the knowledge gaps that remain.
• Modularity: Resilience is not the only criteria to follow when identifying the most appropriate land management strategy. In fact, there can be tradeoffs between increasing the provision of the service or increasing resilience to disturbances. The resilience assessment should be a part of a wider land management strategy, thus it should be compatible with other tools and methods for evaluating and improving land management.

Other crucial methodological choices are related to the definition of resilience, the scale and unit of analysis and the combination of different sources of knowledge. These choices are explained in detail in the following paragraphs.

1. Application of the resilience concept in the CASCADE project

Scientists agree that resilience is an emergent property of an ecosystem, meaning that it is the result of multiple interactions between the elements which constitute that ecosystem, such as soil, plants, human action, and others (Lavorel, 1999), acting at different scales in time and space. This means that to effectively assess the resilience of an ecosystem, we cannot exclude many of the interactions that occur at the local level, especially those between the environment and the land users (Jucker Riva et al., 2016). This is particularly important in the Mediterranean area, where there is a long history of land use that has had a profound influence on the current structure of the ecosystem (Kosmas et al., 2015).

Among the interactions between land users and their environment, land management has the potential to improve the state of the ecosystem and to increase its resilience to disturbances. The impact of land management is dependent upon the specific natural and human environment of the system. Therefore, identifying and studying the land management practices that are already implemented in the field is needed to understand the role of land management regarding resilience and this has to be done in collaboration with land users and managers (Liniger & Schwilch, 2002). Moreover, land management practices are often implemented in combination (e. g. fuel breaks and selective clearing in forests, or controlled grazing and fodder production in dry pastures). Therefore, to study the resilience of a land management practice, we have to first analyze the land use system as a whole.

Many researches investigate resilience only in relation to point disturbances (Buma & Wessman, 2011); (Lewis, Reid, & Clarke, 2010); (Bérard, Bouchet, Sévenier, Pablo, & Gros, 2011). Land users do not perceive a distinction between degradation caused by specific events and degradation caused by long-term pressures or unsustainable use, and are rather concerned with the stability of a certain land management system, i.e. that the environment will continue to provide the ecosystem services and benefits they look for. Thus, in the context of an interdisciplinary and applied use of the resilience concept, the distinction between pressure drivers, and between sharp changes (regime shifts) and slower changes (degradation) are less relevant. To be of practical use, the assessment of resilience was designed to be close to the land users' needs but without disregarding the scientific understanding of resilience and regime shifts.

Often resilience studies are based only on the investigation of past events (Soane, Scolozzi, Gretter, & Hubacek, 2012); while the history of a system can explain much of the current state, it doesn’t necessarily allow to forecast its future evolution, due to the fact that pressures drivers and the internal functioning of the system, as well as the demand for ecosystem services, might have changed since the last perturbation (Elmqvist et al., 2003). In the present assessment, we rely mainly on experts’ knowledge to forecast future evolution of pressures and important processes.

In the context of this work, by resilience of a land management system we mean the ability of a land management system to remain productive and valuable, according to land users’ evaluation, in the face of pressure sources, and to withstand the shocks that affect the area.

Point disturbances like fire or droughts are part of the history and of the evolution of most of the land management systems analyzed here. However, a change in intensity or frequency of disturbances, or a modification in the internal functioning of the land management system, can produce a regime shift. Thus, while the causes of regime shifts are complex and multidimensional, the trigger is often a shock or disturbance event, after which the system changes to a new state. Understanding the impact of disturbances on a land management system is important to forecast its future evolution, and what modifications to the way the land is managed could increase resilience of the system.

It is important to distinguish between disturbances that are “normal” and that might even increase the resilience of the system in the long-term (Folke et al., 2010), and those that induce permanent changes to the system. Because of the uncertainty related with the concept of “permanent changes”, we have chosen to limit the relevant time span to 30 years. Thus any change to the system that is not likely to recover within 30 years is considered “permanent”. In the text of the RAT, the term “permanent change” was used in place of “regime shift” because it is easier to understand and carries fewer implications.

Besides the relationship between the system and the disturbance that affect the area, we also consider the internal changes of the system (e. g. the degradation caused by the land use), and we try to take into account processes that occur at a bigger scale.

By evaluating in detail the role of land management in relation to resilience, this assessment helps to identify weaknesses and possibilities for ameliorations in the way the land is managed in order to increase stability in the provision of ecosystem services and to prevent regime shifts.

2. Land management systems as the unit of analysis

We consider a land management system the object of study, and land management as the main way to increase its resilience. More precisely, among the land use types that are most common in the study sites, we analyze a particular land management system: an area that is managed with a specific set of land management practices for the same purpose.

Normally, a land management system corresponds to one area, with one land use/cover. For example: "Pinus halepensis afforestation managed with selective clearing and firebreaks for landscape conservation and controlling soil erosion". However, a land management system can be composed of different land uses/covers if they are all managed by the same actors and with the same objectives. For example: "Grazing system managed with seasonal grazing management and fodder cultivation for milk and meat production". If the same management practices are applied on small portions of land within an area, with the same objectives and by the same or comparable actors, they can be considered as one land management system. For examples: "Riverbank management with multi-specific shrub plantation and dry walls to prevent soil erosion, reduce risk of floods and increase diversity of vegetation".

3. Combining scientific and lay knowledge

Integrating lay knowledge in scientific assessments, especially on land management, allows identifying solutions that are closer to land users perception and needs, and is therefore an effective way to ground results in the local context. What has been called "knowledge exchange" (Fazey et al., 2012) or "trans-disciplinary approach" (Wiesmann & Hurni, 2011) has also advantages for the scientific investigation of the ecosystem: Most often, lay knowledge relies on a long-term management experience that spans through multiple generations of land users. Information on past events, and in particular on the evolution of the ecosystem is difficult to acquire through conventional scientific data; this is particularly true for remote ecosystems that are not at the center of long-term monitoring activities. Even when data series are available, they rarely include information at the detailed scale that is required for land management studies. Moreover, land users tend to have an integrated understanding of their land, meaning that they can holistically understand the relation between many different variables and processes, and can recognize patterns in the evolution of the ecosystem. This is particularly difficult to obtain with standard scientific methods that tend to focus on "average situations" and on few variables at a time.

However, local knowledge has its weaknesses, and comes in a form that is difficult to integrate with scientific data. In particular, stakeholder knowledge is based on each individual experience and on the knowledge that he or she has been able to obtain. Moreover, lay knowledge is often descriptive: land users have difficulties in classifying or quantifying information.

In the RAT, we aim at combining scientific knowledge about the system with the experience of land users and local administrators. This aspect is crucial to ensure that the assessment is compatible with the views of local actors, who are the ones that actively manage the system and can affect its evolution (Liniger, Lynden, & Schwilch, 2002).

We have taken several steps to ensure an effective integration between scientific and lay knowledge:

• Repeated and extended field visits: During the course of CASCADE we have performed several field visits in all the study sites. This allowed meeting stakeholders and creating a stakeholder group for each study site. We have taken care to contact different categories of stakeholders including land owners, land users, advisors and local experts (forestry service, fire protection, public or private advisory), local administrators.
• Flexible design of the Resilience Assessment Tool: we designed the RAT to include a wide range of information, from quantitative to descriptive. Thus, depending on the source available, all type of information, be it from studies and scientific publications or from exchanges with land users, can be used and combined to complete the assessment. In particular, the use of semi-quantitative indicators and a "Comment /specify" section next to each question allowed flexibility in the assessment and simplified interpretation of results.
• Simple language: throughout the tool specific terms like "regime shift" were avoided and we have taken care of simplifying all questions as much as possible, even if a link with scientific concepts and definitions was maintained.
• Testing of the tool: We have tested the questionnaire in a wide range of situations and with different stakeholders to ensure that all parts are understandable and could be answered by scientists, local administrators and land users.

Note: For full references to papers quoted in this article see

» References