System Dynamics Applied to Terraced Agroecosystems: The Case Study of Assaragh (Anti-Atlas Mountains, Morocco)

: Terraced agroecosystems (TAS)—apart from being an important cultural heritage element—are considered vital for sustainable water resource management and climate change adaptation measures. However, this traditional form of agriculture, with direct implications in food security at a local scale, has been suffering from abandonment or degradation worldwide. In light of this, the need to fully comprehend the complex linkage of their abandonment with different driving forces is essential. The identiﬁcation of these dynamics makes possible an appropriate intervention with local initiatives and policies on a larger scale. Therefore, the main aim of this paper is to introduce a comprehensive multidisciplinary framework that maps the dynamics of the investigated TAS’s abandonment, by deﬁning cause–effect relationships on a hydrogeological, ecological and social level, through tools from System Dynamics studies. This methodology is implemented in the case of Assaragh TAS, a traditional oasis agroecosystem in the Moroccan Anti-Atlas, characterized by data scarcity. Through ﬁeld studies, interviews, questionnaires and freely accessible databases, the TAS’s abandonment, leading to a loss in agrobiodiversity, is linked to social rather than climatic drives. Additionally, measures that can counteract the phenomenon and strengthen the awareness of the risks associated with climate change and food security are proposed. state is the one observed in 2019, as a diachronic comparison was not feasible. The surfaces have been measured contouring polygons manually, on a satellite imagery base, the results have been validated by a ﬁeld survey.


Introduction
Terraced landscapes represent one of the oldest and most efficient forms of human adaptation to cultivate in harsh environments [1][2][3]. In the Mediterranean area, this technique is especially adopted to cultivate marginal or sloppy lands from the small island (e.g., Sicily in South Italy, Cyclades in Greece) to the arid slopes of the Atlas Mountains in Morocco [4][5][6][7]. Terraced landscapes are increasingly considered as a valuable cultural, historical, religious and ecological heritage for humanity. The valuable agricultural biodiversity, the important ecosystems resiliency and the high cultural and can be useful for integrating qualitative and quantitative assessments of multidisciplinary nature that can even be interconnected [35]. Moreover, its applicability in cases with data scarcity allows to address issues in those areas of the world which often are more vulnerable, poorer and marginalized. The latter consequently suffer from a lack of interest in monitoring environmental and social variables, properly [36].
In order to test this approach, the SD methodology has been applied in the Assaragh TAS-an oasis in the Anti-Atlas of Morocco characterized by a traditional agroforestry system-affected by lack of data. In this context, extensive field studies, questionnaires, interview, freely accessible databases, as well as wadi (see Appendix A for specific terms) hydrological methods and agroecosystems monitoring, have contributed in comprehending the nature and causes of the phenomenon, which in this case covered principally the social and cultural spectrum. Preventive and improvement measures were finally proposed. Despite its limitations, the authors believe that the methodology presented can be a useful tool for analyzing and tackling the problem of TAS's degradation, when adapted on a case-by-case basis.

Case Study
Morocco, a country very vulnerable to the effects of climate change [37], since 2008, has been working on a national strategy for sustainable agricultural and food security development, known as Morocco Green Plan (MGP) [38]-today updated as Génération Green 2020-2030 [39]-aiming to modernize the agricultural sector in order to become more competitive in global markets and to promote sustainable management of natural resources. However, this general idea of modernization in marginal environments, such as the oasis in the arid Anti-Atlas, can give rise to great contradictions and problems [40]. Schilling et al. [41], in fact, noticed discrepancies in the MGP and recommended switching the focus from maximization of agricultural production to output stabilization. In this sense, TAS-very often found in Morocco-plays a fundamental role in reducing the loss of organic matter and soil moisture, helping to maintain the production capacity in a climate change scenario ( [7]) and thus preserving food security at the local scale.

General Overview
The basin of Assaragh (or Assrargh, or also Asrargh, depending on the translitteration)-named after the village located in the south of the basin-is situated in the Anti-Atlas region of Morocco on the boundary between the Drâa-Tafilalet and the Sous-Massa region (see Figure 1b). Administratively, it is part of the Souss-Massa region and is located north of the province of Taroudannt, south of Tata and east of Agadir-Melloul.
From a phytogeographical point of view Assaragh basin is situated at the limit of the sclerophyllous forest characterized by the presence of sclerophyllous trees such as Juniperus phoenicea and fragmented plots of Quercus ilex [42,43] and thus, belongs to the arid infra-Mediterranean vegetation zone [44]. The region is under strong Saharan influences characterized by climatic aridity and low rainfall [45]. The hydrographic network of the basin has a north-south orientation and is of torrential nature, thus, drying up most of the year. Its main creek, assif Aguinane, passes through the village of Assaragh, whose oasis is mainly irrigated by springs and wells. The main characteristics of the basin are presented in Table 1.
Along the assif and the oases, several villages (or ksours) are settled, whose emergence was facilitated by the water stored in karst limestones, feeding the oases. As an observer leaves the basin outlet and proceeds more upstream, the following villages are encountered: Assaragh, Lemdint, Iguerda and Timdghart (Figure 1c). The total population living along the course of the assif Aguinane is 2875 inhabitants, of which 1330 live in the upstream assif branch within the investigated area (details in Table A1).  Table A2) 5.6−1.7 h

The Assaragh TAS
The four villages in the Assaragh basin are characterized by the presence of terraces that develop from the bed of the wadi to embrace the inhabited area. Terraces are structurally supported by dry stone walls, made by material recovered on-site; stones are quite irregular and their dimensions range from a diameter of 0.1 to 1 m with the prevalence of intermediate and small diameters ( Figure 2). Much of the structural solidity of the walls is guaranteed by the packing due to the finer soil, mainly clay, which over time fills the empty spaces between the stones.
Agricultural land in Assaragh is managed by small private farms. The possession of the land is private and individual plots are registered to farmers; this legal form of land ownership is called melk. The mean size of farms does not exceed 0.5 ha; the smallest farms can count at minimum 3 plots, while the biggest ones can reach a maximum of 30 plots. The small plots may explain the relatively small production outcomes, and thus, the low generated income. The majority of farmers are using traditional ploughs to loosen or turn the soil before sowing seeds or planting trees. Cereal and herbaceous crops are harvested manually using sickles, while animals facilitate the transportation of crops. Unfortunately, throughout the years, Assaragh has seen a gradual abandonment of its terraces. This phenomenon, although more evident for the terraces of Iguerda and Timdghart, is heavily present in Assaragh, as well. As one can note from Table 2, Assaragh and Lemdint have experienced an abandonment of almost 30% of their pre-existing terraces, while the ratios for Iguerda and Timdghart are above 77% and 88%, respectively. A general map of the active and abandoned terraces is presented in Figure 3. The TAS in Iguerda and Timdghart, while still retaining residual active terraces, can be considered essentially abandoned since they no longer have a systemic coherence. The few remaining cultivated plots are owned by elder members of few families that produce subsistence food.  The abandonment of the terraces can also be noted by the change in the hypsometric curve of the TAS which is more evident at the extremes of the TAS's altitudinal range ( Figure 4). The abandonment of the terraces is perceived by the inhabitants of the villages as an issue linked mainly to the loss of identity and landscape beauty, while the farmers who own terraces adjacent to the abandoned ones sometimes report an undergoing hydrogeological degradation of theirs. It should also be noted that the only vegetation of the basin (accounting only for~0.5% of the basin area) is present only on the cultivated or recently abandoned terraces, with very few exceptions along the course of the wadi.
For this study, the authors will consider Assragh and Lemdint as part of the same TAS, referred to, hereafter, as Assaragh TAS. As stated previously, Iguerda and Timdghart in the past had their own TAS which today is almost completely lost. The study is carried out at the basin scale but focuses on the Assaragh TAS.

Definition of SD
The study of SD suggests that an SD behavior is controlled by feedback loop structures [46,47] that can provide a potentially useful conceptual framework in the assessment of non-linear complex feedbacks, such as the ones arising from a tightly coupled social-ecological system [25], for example the phenomenon of the degradation or abandonment of TAS [25]. This is achieved through casual loop diagrams (CLD), which are tools that allow the conceptualization of such structures, part of a complex system in systems thinking. A CLD is a powerful SD tool used to illustrate pictures of systemic perceptions or feedback structure patterns. CLDs consist of balancing or negative feedback loops (B), or reinforcing or positive feedback loops (R) and/or a combination of both. Defining system archetypes as a generic system structure presenting common behavioral patterns [48,49] reinforcing and balancing feedback loops are the simplest system archetypes emerging from a CLD. In real systems a combination of these basic archetypes can present more complex dynamic behaviors, which can be described by appropriate and more structured archetypes that adequately describe the emerging dynamic [49].
The advantage of SD modeling lies in its possibility of integrating quantitative and qualitative data, even if they are incomplete or of poor quality [50][51][52][53]. In addition, it is possible to analyze the SD in retrospect by corroborating hypotheses on how the system obtained its current state, despite lacking data regarding its past states [35].

Modelling Process
Modeling a complex system by means of SD can be seen as an iterative process, starting with a general conceptualization of the CLD on the basis of a selected literature review on similar cases, a general observation of the ecosystem to be modeled and experience. The first CLD conceptualization should address the data collection and analysis process. In the light of the data collected and analyzed, the robustness of the CLD can be verified and, if necessary, reformulated and enlarged to better represent the case study and the dynamics to be investigated. Again, if qualitative or quantitative data is needed to investigate some dynamics, a new data collection campaign must be carried out and consequently a new verification of the CLD. These processes are iterated until the data collected, the dynamics observed and the model represented by means of the CLD, are self-consistent. In this sense a self-consistency of the model can be seen as its validation [49]. This global process has been well illustrated by Wolstenholme [50] and bears the legacy of a methodology, such as that of SD, born in the industrial sector to map and improve production processes [46,51].

Selected Literature Review on Water Resources and Ecosystems Related Problems
In the literature, there has been several occasions where SD was employed to tackle environmental issues. In particular, Shahbazbegian et al. [31] developed a hydropolitical analysis in relation to cross-border basins (e.g., Helmand basin on Afghanistan/Iran border) keeping in account social and political drivers together with quantitative data on water volumes and flow rates, while other authors, such as Di Baldassarre et al. [33] used this approach to map and describe the reservoir paradox-discussed further in this paper-in analogy with a paradox linked to modern irrigation. Tenza et al. [32] presented a qualitative model using the conceptualization process of the SD to study the dynamics of an oasis in Mexico that has witnessed a dramatic decline in recent decades. The authors used in-depth interviews, participant observation, expert opinions and official statistical datasets to define the boundaries, and structure in a CLD of the investigated system. Martínez-Fernández et al. [30] traced the environmental and socio-economic interactions in the evolution of traditional irrigated lands in Murcia by means of an SD model containing the main social, economic and environmental policies set up by local and regional authorities. Sušnik et al. [34] applied SD to study the water resource management in Merguellil catchment in Tunisia, an arid catchment subjected to domestic, industrial, agricultural and external pumping and presenting multiple feedback loops connected to the over-exploitation of water resources. All of the above studies have been exemplary models on which to base the first CLD conceptualization.

The Model for Assaragh Case Study
On the basis of what is stated in the previous sections, the CLD for Assaragh case study, represented in the Figure 5, was built. Positive and negative relations between the variables in the CLD are represented. A positive relation implies positive feedback, e.g., a good recharge of the water table, therefore a wide availability of groundwater, encourages irrigation by exploiting the aquifer. A negative relation implies a negative feedback, e.g., a more pronounced irrigation deficit leads to a reduced plant growth and productivity. Note that a positive feedback loop is composed by an even number of negative relations, e.g., an increase in groundwater irrigation discourages the development of rainwater irrigation leading to its reduction, and, thus, watering using groundwater becomes increasingly necessary. Figure 5. Map of the dynamics in Assaragh, based on system dynamics (SD). Red dotted lines represent negative coupling, e.g., an increase in emigration implies a decrease in cultural exchange between generations. Green lines represent positive coupling, e.g., an increase in cultural exchange between generations increases the know-how retention inside the village in relation to terracing and dry walling. Feedback loops are noted in blue (see glossary in Appendix A for some terms).
The mapped causal loops are linked in the diagram with five main variables: climate change, terracing know-how, emigration, hydrogeological risk and TAS stability. The main variables have been identified on the basis of three principles: they can be megatrends that occur on a global scale, they are variables that generally occur in similar case studies or they are elements characterizing the class of problems analyzed in the case study. In order to properly address the data collection and analysis a study of the archetypical structures of the CLD is needed [35,46,50]. Feedback loops are listed and briefly described below, in Table 3, for the sake of simplicity; those borrowed from similar problems are subsequently discussed in relation to the original problem considered in the bibliographic analysis during the construction of the CLD.

Roots reinf. walls -wet loop R(rww)
Plant growth can present a Reinforcing interaction with dry walls by the consolidation effect exerted by roots; this interaction shades the wall from wind and solar radiation, and the compaction of the stones reduces water runoff; all these effects decrease Plant water deficit promoting Plant growth.

Roots reinf. walls -dry loop R(rwd)
Plant growth can present a Reinforcing interaction with dry walls through the consolidation effect exerted by roots; this interaction decreases HYDROGEOLOGICAL RISK, which generally increases the organic matter flushout. An increase in Organic matter production and conservation promotes Soil moisture which reduces Plant water deficit promoting Plant growth.

Roots disrupting walls B(rw)
Plant growth can present a Destructive interaction with dry walls by the consolidation effect exerted by roots, this interaction increases HYDROGEOLOGICAL RISK, which generally decreases the organic matter flushout. An increase in Organic matter production and conservation promotes Soil moisture which reduces Plant water deficit promoting Plant growth.

Drought adaptation R(bio)
An increase in Agrobiodiversity favors the Selection of drought resistant varieties, which can lead to a reduction of Plant water deficit, thus, to a promotion of Plant growth, which implies an increase in Organic matter production and conservation necessary to develop a higher Agrobiodiversity.
The causal loops B(emig), R(demo) and B(econ) have been mapped according to a quite complete study by Silverstein [54] investigating the emigration phenomenon in the specific case of Amazigh oases in North Africa. These three loops together create an archetype addressed by the main variable of emigration, discussed further in the result Section 3.5. The causal loop of agrobiodiversity and organic matter production and conservation has been mapped with reference to three studies; the first one [55] was on carbon sequestration in arid ecosystems, the second one [56] on how genetic erosion impedes adaptive responses to stressful environments, and by consequence a loss in biodiversity and resiliency and the third [57] on the importance of a transitional zone between desert and TAS in order to maintain TAS's stability and allowing dynamics on natural selection [58,59] for species fitter to arid climates.
Loops including plant growth and interaction with dry walls, as well as the R(SMTC) loop, have been mapped starting from direct field observation and from studies of various authors such as Castelli et al. [60] and Tarolli et al. [12]. The archetype emerging around the main variables (climate change, terracing know-how and hydrogeological risk) directly influences the TAS's stability. However, it does not give a priori a causal relation between climate change and TAS's degradation, and this is what will be further investigated in the data collection section.
The loop R(cl) on cultural landscapes refers directly to the vast literature promoted by UNESCO on the theme; in particular, for traditional water harvesting techniques and water distribution can be referred to a study of Barontini et al. [20]. The loops R(ws), R(mt), B(wd) and R(se) have been mapped to describe a dynamic that emerged during the second phase of the modelling process, after a first field survey. These loops creates a more complex archetype analogous to the one studied by Di Baldassarre et al. [33] in relation to the reservoir paradox discussed later in Section 3.4 of the results.

Data Collection
The Assaragh watershed is well-mapped from a lithological and geological point of view. However, meteorological, climatic and hydrological data as well as precise data on the state of the aquifers are missing. With regards to the availability of socio-economic data, the detailed demographic information of the censuses prior to that of 2014 does not capture the details of the local scale. Missing data was complemented by direct field collection and freely accessible databases.
As stated previously, an SD approach allows to integrate qualitative and quantitative data. In the present case study the data collection process has been subdivided into three phases. A first phase was established including field studies, as well as unstructured data collection through direct observation, photographic documentation and informal interviews with the inhabitants. A second phase followed, including the collection of terrain data (Digital Elevation Model-DEM dowloaded from NASAEarth [61]), satellite imagery, climatic and socio-economic data.
The third and final phase-entirely affected by the Covid-19 pandemic-should have involved the distribution of questionnaires on the perception of risk and a mapping on the field of agro-biodiversity. This phase was conducted remotely via telephone instructions, remote surveys through photographs and videos thanks to the help of young people from the oasis. The data collected in the third phase have been used, despite the fact that it was not up-to-standards, since the questionnaires were not administered through a re-entry protocol and some responses were influenced by the previous interviews. However, this data provided more organized and structured additional information compared to that found in the first phase. Needless to say, third phase data were treated with much discretion in relation to what was already observed and documented during the first phase.

Field Surveys and Unstructured Data Collection
Two field surveys, one in May 2019 and one during the end of October and the beginning of November 2019, have been vital for the documentation of the terraces' status, the irrigation network and the agricultural practices ( Figures 6-8, A1-A3). The structure of the terraces and the interaction of roots with dry walls, the building techniques and the consequences of flood events and their magnitude have been investigated, as well.   The ground survey was necessary as some plowed plots can easily be confused with abandoned plots when using only satellite imagery. Furthermore, in the basin some sedimentary concentric formations are present, that can easily be confused with abandoned terraces (Figure 9). The surveyed abandoned terraces were outlined in GIS software using Google satellite imagery as the basis. A previous field survey was already carried out. Some concentric geological formations (A) with center in (B), which can be easily mistaken for abandoned terraces.

Terrain, Meteorological and Demographic Data
Terrain data were available in the form of a DEM with a resolution of 5 m (NASAEarth [61]) and Landsat satellite imagery. Climatic satellite data (monthly temperature and precipitation from 1982 to 2019) were retrieved from NASA Power database [62] with a resolution of 0.5 • . This validated data regards all of Africa and are the product of interpolation (using the MERRA-2 algorithm) of multiple sources of measurements both from satellite and from land surface. It can provide quite reliable results [63] and, therefore, it was overall preferred. It should be mentioned that the closest weather and climatic stations, as the crow flies to Assaragh, are located in Taliouine (65 km) and Ouarzazate (110 km). Both cities have different climatic regimes and altitudes in respect to the investigated basin. For demography the census data, freely available from governmental sources [64], has been used.

Structured Data Collection
The structured data collection regarded a questionnaire divided into three parts; a first part questioning the farmers, required a integer estimate on the occurrence of some phenomena in recent years and in the past (Table A3), a second part included statements measured on Likert scale (1-5), see Appendix A4 and a third and final part aimed at detecting the cultivated plant species and the agricultural techniques practiced, by means of a presence matrix (Section 2.4.5). The respondents were 15 farmers ranging in age from 41 to 75 years old. The interviews were administered by a local active citizen from Assaragh who supported the process of field surveys.

Geological Data
In Assaragh basin, the lithological formations, presented in Figure 10, can be subdivided into three categories as summarized in Table 4. The central Anti-Atlas is characterized by the absence of generalized regional aquifers; however, at the level of the Assaragh basin, the lithological shelters shared circulations between the karsts and the fractures as shown in Table 4. The highly developed Adoudounian limestone series in the basin at high altitudes (basin's North), can constitute a water reservoir if the underlying substrate is impermeable. It favors a strong runoff towards the amot (wadi's outlet) of the basin by following the watercourses and it can participate in the recharging of the perched aquifers by following the plants of the strike-slip faults and vertical faults, abundant in the center of the basin. Generally speaking, most of the basin is made up of hard rocks and non-permeable soils which cause high runoff. The latter adds to the effect of slopes and amplifies the erosive process [65,66] by digging ravines in the soil and causing the abandonment of the terraces or even their destruction. Towards low-lying areas, the concentration of rainwater slightly increases infiltration along the alluvial valleys due to the lithological nature. The water follows the cracks (transcurrent faults and vertical faults; see Figure 10) and forms underground micro-conduits, getting distributed in the deep areas. Therefore a weak erosive process is present which can maintain the durability of TAS. However, the infiltrated volume most likely remains much lower compared to the precipitation volume; therefore high runoff dominates in the basin, especially during stormy periods.

Climatic Data
Using the 37-year monthly meteorological data of NASA Power, the periods consisting of consecutive days without rain have been identified. From the analysis, five out of six periods of longer duration fall within the second half of the 37-year interval emerged, as summarized in Table 5. A fractal analysis of wet and dry periods was carried out according to the methodology developed by Bazuhair et al. [68]. The results show a pattern of a semi-arid climate with similar duration for dry and wet periods of short durations and dry periods that are generally shorter but more persistent (Figure 11). The results of this analysis are confirmed by the data collected from unstructured interviews and questionnaires. The 15 interviewed farmers agreed on a halvening of rainy days in the past 30 years and on an approximate reduction of three to four times the rainy days in the past 60 years (see Appendix A3); in general, there is a complete agreement regarding the reduction of rainy days with respect to the past.

Water Resources Balance
In the present case study, regionalization methods for the estimation of parameters used in rainfall-runoff models were not applicable, since the basin is extremely small compared to the dimensions of the well-studied wadi basins [69]. Therefore, a parsimonious models and reasonable hypothesis based on common sense, literature and experience have been adopted.
As one can observe from Figure 12 the investigated basin has a wet season that develops from November to March in which it rains just over 65% (~150 mm) of the annual precipitation. Starting from this satellite meteorological data, an estimation of the actual and potential evaporation and of the monthly water deficit within the basin according to the Thornthwaite method was carried out [70]. According to Thornthwaite, an estimate of the hydrological balance of the agricultural land in the TAS, has been developed, assuming a specific water capacity retained by the soil as U = 50 mm. This estimate was conservative and defined on the basis of experience and field observations (small plots in terraces, flooded and often surrounded and shaded by trees with deep root systems; some example images can be seen in Figure 13). The general law for soil desiccation [71] during the dry season is given by: where P is the precipitation, E the actual evapotranspiration, E p the potential evapotranspiration, A the specific volume of water in the soil volume affected by evapotranspiration, varying from 0 if completely dry to U if soaked.
Since there was no possibility to study the soil desiccation dynamics in the basin, three soil desiccation laws were hypothesized, for m = 0, 1 and 2. Then, using Equation (1), the hydrological balance for the agricultural soil on a monthly scale was developed, obtaining the results shown in Figure 14. Water deficit and surplus values and ratios illustrated in Figure 14 have been summarized in Table 6. One can note that agricultural soil suffers from a quite high water deficit in dry season. However, this deficit is not representative of the operation of the TAS as it is compensated by irrigation from the aquifer, which is recharged by rainfall over the entire basin area. Assuming the parameters shown in Table 7, and maintaining a conservative approach, the values in Table 8 were derived. Some values such as the percentage of the precipitation lost in direct evaporation and from base flow and direct runoff were just conservative estimates based on a study from De Jong et al. [72]. The authors calculated a yearly water balance for Ifre basin, which is colder, one order of magnitude larger and with higher altitudes with respect to Assaragh case study. For the Ifre basin a yearly loss for evapotranspiration equal to 74.8% of the cumulated precipitation has been estimated, and an outflow equal to 12.9% of the cumulated precipitation has been measured. For Assaragh basin it was estimated that 90% of the rainfall is lost by direct evaporation from bare soil in the dry season. This value is quite reasonable, since the cumulated monthly precipitation is given by relatively short rainfall events, which are not capable of soaking the dry bare soil for more than few centimetres. Table 7. Parameters calculated and assumed to derive a water balance at the basin scale.

Parameter Value
Cultivated area (active terraces) 0.474 km 2 Whole terraced area (active and abandoned terraces) 1.184 km 2 Direct evaporation from bare soil in wet season 65% Direct evaporation from bare soil in dry season 90% Base flow plus runoff in wet season 25% Base flow plus runoff in dry season 0 % Domestic water need per inhabitant 200 l/in Number of equivalent inhabitants (animals and people) 1500 in As shown in Table 8, under the working hypotesis previously illustrated, the Assasargh TAS presents a water surplus even during dry season and for two different scenarios, namely for the current cultivated surface (0.474 km 2 ) and a hypothetical state in which abandoned terraces are cultivated, as well (1.184 km 2 ).
In fact, the unstructured interviews and questionnaires have shown that despite a clear perception of the reduction in rainy days and an average increase in temperature, a reduction in the availability of the water resources, used in agriculture, is not perceived (see Appendix A4). Table 8. Summary of agricultural and domestic water volumes, and water availability at the basin scale. Volumes are calculated for the current state (c. s.) considering a cultivated surface of 0.474 km 2 , and for a hypothetical state (h. s.) in which also abandoned terraces are cultivated for a total of 1.184 km 2 of cultivated surface. See Table 2

Demographic Data
Morocco today is a growing country with a birth rate of 2.3 and in the past 60 years it has seen its population triple. Despite not having accurate historical census data on a basin scale, it was possible to estimate little more than a doubling of the population within the basin in the same period. The estimate was made through interviews with people from the village.
The current population in the investigated area is reported in Table 9. The demographic pyramid for Assaragh and Lemdint has been detailed in Figure 15 and by means of unstructured and structured (see Appendix A4) interviews and demographic data a strong and recent emigration, especially of the young male population, was observed.  What emerged from these interviews and questionnaires was a perception of terraces abandonment as an issue for the village although there is an impression of a global improvement in the quality of life in Assaragh, probably also thanks to some modernization processes. All the interviews confirm that the emigration of young people and the attractiveness that surround the big cities, act as poles of attraction for emigration. Another determinant factor is the perception of agriculture as an activity that does not improve economic status and quality of life.

Agrobiodiversity Assessment
The evaluation of the agrodiversity in the Assaragh oasis was carried out with questionnaires, designed to gather data relative to traditional crops representative of the TAS and integrated by field surveys and direct observation. This approach has already been applied in previous similar studies on agroecosystems in the Rif [73], Anti-Atlas [7] and in the south of Morocco [74]. The questionnaire makes it possible to explore the different farming layers and present a comprehensive inventory of the traditional crops (cereals, pulses, local landraces and varieties' diversity), as well as introduced varieties.
Assaragh TAS presents three main layers, two arboreal ones and an herbaceous one.
1. The main and superior tree layer is mainly composed by palm trees (Phoenix dactylifera). Moroccan oasis are known by their important varietal and genetic diversity of date palm trees [75][76][77][78], in fact Moroccan oases present more than 220 known varieties [79]. Palm trees on one side shade the plots and protect them from wind reducing soil dissecation; they provide mechanical protection from soil erosion and once dead they can be used as construction material for terraces and irrigation channels. Moreover they supply fodder to sheep and cattle that are raised in enclosures and fed with palm leaves mixed with alfalfa and vegetable waste. Several studies have shown essential and diverse bacterial communities associated with root palm tree systems [80,81]. The main arboreal layer close to wadi banks, in some spots, hosts ancient specimens of Celtis australis, Populus alba and Populus nigra as they can naturally occur where the hyporheic flow is more present. 2. The secondary arboreal layer is composed mainly by seven species of cultivated tree: olive (Olea europaea), almond (Prunus dulcis), apricot (Prunus armeniaca) pomegranate (Punica granatum), carob (Ceratonia siliqua), fig (Ficus carica) and apple (Malus pumila). The olive tree, identified by the farmers under the denomination of 'Zitoune' and the almond tree are the dominant species in the secondary layer. The olive variety Zitoune has been previously identified by genetic markers as 'Picholine marocaine' [82]. It is widely spread in Moroccan agroecosystems and largely appreciated by local farmers [11,83]. The preference for olive and almond trees is in agreement with what has been observed by Belarbi et al. [84] in Aoufouss oasis (South of Morocco). 3. The herbaceous layer hosts two species of cereals, wheat (Triticum turgidum), known locally under the denomination of 'Amazigh', and barley (Hordeum vulgare). This layer is dominated by alfalfa (Medicago sativa) used as fodder, and by eight types of vegetable plants destined for local consumption: tomato (Solanum lycopersicum), onion (Allium cepa), carrot (Daucus carota), eggplant (Solanum melongena), faba bean (Vicia faba), courgette (Cucurbita pepo), green pepper (Capsicum annuum) and leek (Allium ampeloprasum). Therefore, association of wheat and barley with other crops was found to be beneficial for sustainable agricultural practices [85]. Assaragh is known also for the production of high quality saffron (Crocus sativus), which is not possible to cultivate downstream of Assaragh, in the Aguinane TAS at an altitude of 300 m lower.
Small land plots (or absence of flat plots), in addition to the harsh environmental conditions of the Anti-Atlas region has led local farmers towards more efficient agricultural system management, by applying the concept of TAS combined with a polyculture crop (or multi-crop) production system which implies the association of seasonal and perennials crops (Figures 7 and 8; [7]).
Recently Assaragh, as many Moroccan oasis agroecosystems, experienced a loss in agrobiodiversity [74], attested by the regression or extinction of many cultivated crops. In the present case lens (Lens culinaris) and chickpea (Cicer arietinum) crops disappeared from Assaragh agroecosystems. These crops were grown until the late 1970s. This tendency of crops disappearance has been already documented in the Anti-Atlas TAS [7].

Geology Influences on TAS Structure
The state of TAS is linked to several factors such as topography, land use, soil type and anthropogenic activity [86,87]. However, geology plays a huge role in limiting or promoting the destruction and abandonment of TAS [88]. In fact, the lithological nature of the formations can amplify or limit infiltration and runoff [89,90] while the tectonic activity can influence the internal drainage capacity and the water supply of the aquifers. Generally, the geology is in favor of downstream TAS which can benefit from a slow release of water infiltrated through the faults and permeable formations on the upstream areas of the basin (see Figure 10).

Climatic Trends and Rainy Days Reduction
A clear reduction of rainy days has been observed and identified as an evident climate change signal. A progression of higher abandonment ratios from downstream to upstream ksours is, indeed, obvious ( Table 2); this can be explained by a higher groundwater availability in the TAS downstream which receives water from all the drained area with respect to the upstream TAS which rely much more on rain irrigation. In this sense, a reduction of rainy days and an increase of drought periods can explain the higher abandonment ratio for Timdghart and Iguerda TAS (see Table 2).

Water Balance at the Basin Scale
In many agroecosystems in Morocco, water availability has been considered as a limiting factor by many authors [91,92]. This is not the case of Assaragh since, as shown in Table 8, the hydrological balance at the basin scale, keeping into account the considerations developed previously for Iguerda and Timdghart TAS, allows a cultivation of the current cultivated surface added to the surface of the abandoned terraces. This fact does not mean that there is no reduction in water resources, however today a possible reduction in water resources does not compromise the possibility of cultivating and irrigating, as traditionally done, within the basin and more specifically in the Assasrgh TAS.

Perception of Groundwater Resources
Unstructured interviews and questionnaires have shown that there is a general agreement between people that Assaragh 'is blessed by a huge availability of water' and, in general, there seems to be a preference for modern irrigation and water withdrawal techniques as mapped in the loops B(wd), R(mt) and R(ws) in Figure 5.
Nowadays it is impossible to make a diachronic assessment of the quality and quantity of groundwater storage, however, a dynamic linked to water supply, similar to the reservoir effect observed by Di Baldassarre et al. [33] is noted, which hints a lack of perception by the population of a possible aquifer productivity reduction.
The reservoir effect refers to archetypes describing an over-reliance on water infrastructure which increases vulnerability, and therefore, increases the potential damage from water shortages. This phenomenon refers to the construction of reservoirs to reduce adverse effects in water shortage situations, while at the same time the supply-demand cycle describes instances where increasing water supply enables higher water demand, quickly offsetting the initial benefits of reservoirs.
By analogy, as the reservoir effect refers to volumes, the water supply effect refers to the flow rates and to the flexibility of water withdrawal for irrigation purposes, the two effects are compared in Figure 16. Greater flexibility in water withdrawal techniques and the ability to draw higher flows when necessary, increase systemic vulnerability by reducing the ability to cope with irregular irrigation flows during the year, and stress the aquifer. In this perspective, a reduction in groundwater recharge due to climate change could play a marginal and indirect role in the abandonment of traditional irrigation systems. A similar shift towards modern water withdrawal techniques has been already well documented by Lightfoot [93] in Tafilalt basin. However, a markedly cultural dynamic emerged from unstructured interviews.
It should also be noted that in the projects linked to the MGP, a modernization of agriculture and drip irrigation is also encouraged through economic incentives. From what has been observed in the field, drip irrigation seems to be problematic since the water-extremely rich in calcium carbonates and prone to leaving many deposits-could make the plants unusable in a few years. Figure 16. Comparison between two analogous archetypes. R(se) water supply effect (left casual loop diagram (CLD)) and R(re) reservoir effect (right CLD): in the water supply effect the feedback balancing loop B(wd) of water demand is overrun by R(mt) a shift towards modern water withdrawal techniques, to compensate for the aquifer productivity decrease, stressing even more the groundwater reservoir. An increase in water withdrawal can trigger a perception of abundant groundwater resources encouraging an even greater water withdrawal by means of modern wells and motor pumps R(ws). The reservoir effect B(wd) is overrun by the R(ws) water supply measures taken to mitigate water shortage.

Demographic, Social and Cultural Dynamics
It is well noted and documented by many authors and summarized by Silverstein [54] that 'the southeastern oases of Morocco since the 1940s have functioned as a veritable demographic pump, sending streams of labour migrants to northern cities and across the Mediterranean'.
An important dynamic of this archetype is the flux of money generated by emigrants towards their homeland ksour in order to build new houses-based in urban architectural styles-outside of the walled ksour, often in emerging oasis town centers supplied with electricity and running water. This is the case of Assaragh, which nowadays has a water distribution network and tap water in all of the households. This resulting spatial dispersion contributed to the enhancement of migration among the younger generations, who viewed such movement as a means of household development and modern wealth. The archetype illustrated by Silverstein [54], and summarized in Figure 17, emerged quite strongly from unstructured interviews and questionnaires as well as from demographic data. Although the balance of the population inside the basin has remained positive thanks to a high birth rate, the various villages have experienced a young male emigration.
The general effect on the SD of Assaragh TAS is a loss of traditional knowledge and a progressive weakening of terracing know-how due to the difficulties of cultural exchange between generations. A progressive weakening of terracing know-how directly increases the hydrogeological risk and reduces the preservation of traditional irrigation and organic matter in the TAS (see Figure 5). is a positive feedback cycle increasing population, R(demo) already today is weakened by two negative feedback cycles, the B(econ) of economic dependency from emigrants sending money to the village and the B(emig), a cycle linked to the increase of population inside the ksours. Nowadays these cycle are weakened, with respect to the past, but still active and are confirmed by demographic data.

Mechanical Limitation of TAS Linked To Plant Growth
A last important mapped dynamic links the structural integrity of the terraces with plant growth ( Figure 18). As observed from the field surveys, plant growth can, in some cases, present destructive interaction with dry walls CLD B(rw), and this is the case of palm trees (Phoenix dactilifera) as shown in Figure 2a. However, plants can also present a reinforcing interaction with dry walls R(rww) and R(rwd), as was the case of caper (Capparis spinosa) (Figure 2b). In this sense, a proper equilibrium between vegetal species can play a determinant role in TAS conservation and expansion. It is important to note that the positive feedback loop R(rww) acts directly on soil moisture since the plants delay the percolation from the walls and most of the time shade the walls from solar radiation and wind.

Global Dynamics Overview and Trends
By observing the global dynamics in Assaragh TAS it is possible to sum up few trends: a generalized reduction of cultivated terraces more pronounced in peripheral areas of the TAS, a shifting towards more modern irrigation systems favoring a loss of care for the traditional ones. Despite a slow growth of the population within the ksours, a loss of interest for agricultural jobs among the young generation is observed, creating a push towards migration to large urban centers; the latter is weakening the cultural exchange between generations and leading to a loss of terracing know-how. Finally, a loss in agrobiodiversity is directly linked to the abandonment of terraces and of some cultivated varieties. In general, it is possible to exclude that the abandonment of cultivated terraces in Assargh TAS is directly linked to a water shortage or to climate change, however the fact that a reduction in rainy days might have played a role in the abandonment of Iguerda and Timdghart TAS, cannot be excluded. In this sense climate change could lead to emerging problems related to water resources shortage in a future scenario, problems, which can accelerate TAS degradation and abandonment.
The CLD ( Figure 5) indicates possible actions to counteract adverse effects of climate change in the promotion of terracing know-how to enhance the retention of organic matter and soil moisture, favoring agrobiodiversity and the expansion of transitional zone, hence the TAS stability and the selection of drought resistant varieties [58,59].

Possible Actions for Local Initiatives and Policy Makers
Since the main drivers in the investigated case were social and cultural rather than climatic and water resources related, three main actions are suggested in order to favor the cultural exchange between generations into preserving terracing know-how, and the promotion of adaptation strategies through the selection of drought-resistant varieties.

1.
Local primary schools that are generally present in the small ksour community, can offer many possibilities of developing programs aimed at environmental and scientific education based on situated learning in the TAS. There is a vast literature on such experiences [94][95][96][97]. In fact, situated learning is useful not only to teach basic scientific notions, but to promote cultural exchange, as well. Moreover, school activities related to traditional agriculture and irrigation practices can help to build and reinforce the identity of young people [20] hence preserving local culture and developing the social potential of new generations.

2.
The fundamental role of women in preserving crop productivity, traditional knowledge and achieving food security worldwide in rural communities has been widely recognized by the scientific community [98][99][100][101]. This is also valid in Amazigh communities [102] which today are observing a new rise of women involvement in local politics and social activism [103], thanks also to a process of modernization of a society which has become extremely patriarchal, following historical processes of social and cultural evolution [54]. However, in the past the Amazigh society has always had great female leaders who at times have emerged even in the historical periods most unfavorable to the emancipation of women [103]. In this sense the first action suggested views the empowerment of women in primary education, cultural exchange, TAS management and varieties selection, as a natural following action to take in order to consolidate the expected outcomes of the first one. 3.
The need for continuous communication and research of innovative educational and participatory tools to interest the new generations and render the actions of women and local communities more effective is vital. In this way the academic community can play a fundamental role through citizen science activities [104,105], in order to empower the two proposed actions. In fact, the third action consists in the creation of scientific participated laboratories in rural communities, laboratories that, in addition to promoting educational methodologies related to the environment and STEMs, promote citizen science projects aimed at the selection and the exchange of plant varieties suitable for future climate scenarios [106][107][108].

Strengths and Limitations of SD
The strengths of the applied SD methodology lies in its possibility to connect different fields and disciplines and its holistic overview to very complex problems. In general, its ability to integrate quantitative and qualitative data and to address issues step-by-step, offers opportunities to deepen the study of specific problems without losing a global perspective of the case study.
On the other side, the SD methodology needs to be implemented step-by-step for every case. In this specific case, it required a lot of field work to create a model valid and representative of the real dynamics. The SD modeling process has been iterated three times in order to reach a proper representation through CLD, while more complex case studies can require a higher number of iterations. In general the methodology should be implemented when several aspects concerning the case study have already been analyzed separately and there is a need for integrating them. A shift from a multidisciplinary to an interdisciplinary perspective can be cumbersome because of different lexicon and mindsets of the scientists working on the problem and often the shift can be incomplete resulting in a weak connection of single analysis of different disciplines.
However, the methodology can still present an effective overview of the problem and can be very useful in decision making policies. As a reminder Forrester [35] already studied the methodology's drawbacks and advantages, recalling the decision making-oriented nature of SD: 'System dynamics combines the theory, methods and philosophy needed to analyze the behavior of systems in not only management, but also in environmental change, politics, economic behavior, medicine, engineering and other fields. System dynamics provide a common foundation that can be applied wherever researchers want to understand and influence how things change through time. The system dynamics process starts from a problem to be solved, a situation that needs to be better understood, or an undesirable behavior that is to be corrected or avoided. The first step is to tap the wealth of information that people possess in their heads. The mental database is a rich source of information about the parts of a system, about the information available at different points in a system, and about the policies being followed in decision making'. In this sense, an expansion of the literature of similar or contiguous case studies issues and solutions can be a good indication for addressing future research.

Conclusions
In this paper, the possible driving forces that can act upon the Assaragh TAS state have been identified and mapped through an SD framework, allowing the conceptualization of such a complex problem. Additionally, actions to counteract TAS degradation have been proposed.
From the analysis-carried out through field surveys, interviews, questionnaires and freely accessible datasets-it emerged that the dynamics of TAS degradation in Assaragh, were led mainly by social and cultural drivers and very marginally from drivers related to climate change. Future threats emerged in relation to climate change, mainly linked to a possible water scarcity in the future, and to the loss of traditional farming knowledge leading to a genetic erosion of plant varieties and loss of agrobiodiversity.
Similar dynamics were also observed in the rural abandonment that occurred in the north of the Mediterranean during the post-war economic boom [12]. In the investigated case, what changes is the risks related to climate change and food security that must be addressed through forward-looking policies and prudent planning.
Some authors, among others Castelli et al. [60], have demonstrated that changes made in a landscape, may impact on the local climate, and that in arid and semi-arid agroecosystems, Landscape Restoration and Water Harvesting (LRWH) measures can revert land degradation and increase agricultural yields, by means of reducing runoff losses and increasing soil moisture. In this sense, terracing is, indeed, an LRWH technique under all aspects and must be promoted, where appropriate, to stabilize agricultural output on local communities facing climate change threats.
Marginal agricultural communities facing these threats are often found in areas, as in the present case study, where the lack of data can be considered systematic. With this study a quantitative-qualitative methodology was developed that would allow the estimation of missing parameters in a reasonable manner. However, expanding the bibliography of case studies to well-instrumented arid and semi-arid basins, even in small scale cases, would greatly facilitate the analysis of the phenomena occurring in completely isolated basins, representative of the TAS in many marginal areas of Morocco.
It is also worth mentioning that the bibliography on the interactions between root systems, vegetation and LRWH structures is, indeed, lacking. Especially in the case of Assaragh, the role of vegetation is fundamental, as it is structural not only to the ecosystem itself but also to the geotechnical and functional dynamics of the man-made structures. Deepening the multidisciplinary studies in this direction, by integrating geotechnics and biology, could facilitate the improvement of LRWH techniques especially in relation to terraced landscapes.
Finally, the high unemployment rate observed in some study areas offers the possibility of developing professional figures dedicated to the recovery and development of TAS by means of novel LRWH techniques, not only in terms of agricultural production, but also in terms of educational, cultural and touristic development. Funding: This study has been funded by V.B. with his PhD research fund from the Università degli Studi di Brescia and with his personal resources.

Acknowledgments:
The help offered by Aicha Id Lahcen and her family during the field missions, their kind hospitality, as well as their assistance in remote data collection during the Covid-19 quarantine has been determinant to complete the present work. Special thanks to Mohammed Ater, from the Université Abdelmalek Essaâdi in Tetouan, and his Equipe Bio-Agrodiversité for the fraternal support and hospitality offered.

Conflicts of Interest:
The authors declare no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:

Appendix A. Glossary from Wadi Hydrology
Wadi: It refers to a dry (ephemeral) riverbed or creek, in climates ranging from hyper-arid to sub-arid, that contains water only when heavy rain occurs and can present hyporheic flow during the rest of the year. Assif : It refers to the main branch of the hydrographic network and can be considered as the main wadi. Amot: Generally it is the closing section. In the case of an endorheic basin it coincides with the area that collects the outflow. Khettara: It is the local version of a qanat (Remini 2015 [109]), and it is made by a gently sloping underground channel to transport water from an aquifer or water well to the surface; this technique does not stress the aquifer, naturally draining it, avoiding forced water withdrawal.

Appendix B. Demographic Details
Census summary concerning all the ksours and the population living along the assif Aguinane. Data are from RGPH (2014) [64] census campaign. Table A1. Census and occupational data of the population living along the assif Aguinane. Ksours are listed from upstream to downstream. The occupational status can be: employer, freelancer (in majority farmers), employee in public sector, employee in private sector, familiar care, member of a cooperative and other [64].

Demography
Occupational Status for at Least

Appendix C. Additional Calculations
The time of concentration, t c , was estimated with various methods, as seen in Table A2. For the time of concentration using the SCS formula [110] it has been assumed conservative Curve Runoff (CN) values ranging from 85 to 99 suitable for arid rocky terrain accounting for possible karstification.  Figure A1. Wells for irrigation adopted in Assaragh: (a) modern with mechanical pump, (b) and traditional.

Appendix E. Structured Data Collection
Structured data collection has been divided in two parts, the first part is composed of a questionnaire (Tables A3 and A4) to map and validate some dynamics previously outlined, the second part was useful to evaluate the state of agrobiodiversity. For completeness, the questionnaires administered are shown with the average value of the responses and a summary table of each of the 15 individual interviewees. The different questions of the questionnaire are listed here and preceded by the average score obtained by each question. The average age of the interviewed persons is 61.3 years in a range from 41 to 75 years old. Table A3. Numeric answer.

(2.2)
A-How many times during the year is the wadi flooded? (32) B-How many days are rainy during the year? (250) C-How many people live in the oasis (Assaragh and Lemdint)? (100) D-How many people were living in the oasis (Assaragh and Lemdint) when you were a child (from 5 to 12 years old)? (88) E-How many days were rainy during the year when you were a child (from 5 to 12 years old)? (32) F-How many times during the year was the wadi flooded when you were a child (from 5 to 12 years old)? Table A4. Answers on Likert's scale from 1 (total disagreement) to 5 (total agreement).