Climate change and ecosystems dynamics over the last 6000 years in the Middle Atlas, Morocco

. The present study aims at reconstructing past climate changes and their environmental impacts on plant ecosystems during the last 6000 years in the Middle Atlas, Morocco. Mean January temperature ( T jan ), annual precipitation ( P ann ), winter ( P w ) and summer ( P s ) precipitation, and a seasonal index (SI) have all been quantiﬁed from a fossil pollen record. Several bio- and geo-chemical elements have also been analysed to evaluate the links between past climate, landscape, and ecosystem changes. within a low temperature and precipitation range with a of aridity and warming towards the present. T jan has varied within a ca. 2 ◦ C range, and P ann within less than 100 mm yr − 1 . The long-term changes reconstructed in our record between 6 ka cal BP and today are consistent with the aridity trend observed in the Mediterranean basin. Despite the overall limited range of climate ﬂuctuation, observe major changes in the ecosystem


Introduction
The amplitude of climate change during the Holocene (11 700 cal BP to the present) is known to be globally less extreme than during the last glacial period (Bianchi and Mc-Cave, 1999;Bond et al., 2001;Debret et al., 2007). However, several studies have shown that there were climate fluctuations (Alley et al., 1997;Wanner et al., 2008) related to the internal variability of the climate system, solar activity, albedo (Ruddiman, 2003;Eddy, 1982;Stuiver et al., 1991), volcanic eruptions (Kelly and Sear, 1984;Sear et al., 1987;Bryson, 1989;Mann et al., 2005), ocean circulation (Manabe and Stouffer, 1988;Dansgaard et al., 1989;Lascaratos et al., 1999;Rohling et al., 2002), etc., which all have a direct impact on the terrestrial ecosystems (Davis, 1963;Emmanuel et al., 1985). Although climate changes were less pronounced during the Holocene (Andersen et al., 2004;Mayewski et al., 2004;Witt and Schumann, 2005;Frigola et al., 2007;Cheddadi and Bar-Hen, 2009), they have still been noticeable enough to be recorded by different proxies (Dorale et al., Martin, 1973); (c) phytoecological map showing the main ecosystems and the location of Lake Hachlaf (Dayet Hachlaf) within an oak forest (from Lecompte, 1969). 1992; Williams et al., 2002;Geiss et al., 2003Geiss et al., , 2004. At the global scale, the Holocene climate stability allowed sustainable vegetation dynamics with long-term ecosystem changes, plant species expansions and migrations, and an increase of species diversity over all latitudes (Rohde, 1992). However, the Holocene period has also recorded some abrupt and cold events such as the one at 8.2 ka cal BP (e.g. Alley and Agustsdottir, 2005) which recorded a depletion of about 4 • C in winter temperature in the eastern Mediterranean (Weninger et al., 2009).
In Morocco, climate changes during the Holocene have also been quantified, and they show significant fluctuations (Cheddadi et al., 1998). As a matter of fact, the climate variability of the Holocene is less known than that of the last glacial (Mayewski et al., 2004) because it has a lower amplitude and is less abrupt. This statement is even more acute in the Mediterranean region where high-resolution and chronologically well-constrained Holocene records are much less numerous than in Europe or North America. The Mediterranean area is currently a hotspot of biodiversity (Myers et al., 2000) and it is one of the largest regions in the world that undergo long-lasting and pronounced droughts during the summer season (Roberts et al., 2004;Milano et al., 2013). The southern rim of the Mediterranean region is even more arid than the northern one because of the influence of the Azores high and the Saharan winds which increase the impact of the drought effect during the summer season. Most of the winter precipitation (P w ) originates from the trade winds which carry moisture from the Mediterranean Sea (Martin, 1981). The amount of P w has a strong impact on the persistence of water bodies and on the lake levels in the Mediterranean area. Strong lake level fluctuations during the Holocene were observed in Lake Van, Turkey (Lemcke and Sturm, 1997); Lago Dell'Accesa and Lago di Mezzano, Italy (Magny et al., 2006); Lake Kinneret (Hazan et al., 2005) and the Dead Sea, Israel (Migowski et al., 2006); Lake Siles, Spain (Carrión, 2002); and lakes Sidi Ali and Tigalmamine in Morocco Märsche-Soulié et al., 2008).
The analysis of marine and continental records from the central part of the Mediterranean shows that the lake levels were high between 10 300 and 4500 cal BP due to an enhanced moisture availability during both summer and winter . After 5000 cal BP, pollen data from southwestern Europe show that drought increased and led to a sustained reduction of the forest cover (Roberts et al., 2001;Jalut et al., 2009;Jiménez-Moreno et al., 2015). These environmental changes show that within the long-term climate trend there were humid-arid episodes that are related to internal forcings of the climate system such as, in the case of these westernmost Mediterranean ecosystems, the centennial changes in the North Atlantic Oscillation modes (Jiménez-Moreno et al., 2015), the enhancement/weakening of the trade winds, or the increase in the coastal upwelling off northwestern Africa (McGregor et al., 2007).
Climate reconstructions from marine pollen records suggest that the Mediterranean environments may react with a reduced time lag to rapid climate changes (Fletcher et al., 2010). The response of the western Mediterranean ecosystems has even been synchronous with the North Atlantic variability during the last glacial period and the Holocene (Combourieu-Nebout et al., 2009). Changes in the pollen assemblages of a marine record from the Alboran Sea also show very synchronous fluctuations between the surrounding land ecosystem changes and the sea surface temperature fluctuations (Fletcher and Sánchez Goñi, 2008;Combourieu-Nebout et al., 2009). Pollen records from the Middle Atlas (Reille, 1976;Rhoujjati et al., 2010;Nour el Bait et al., 2014;Tabel et al., 2016) and the Rif Mountains  show that the Holocene climate change had a major impact on the ecosystems composition with a clear succession of different species sensitive to winter frost, strong rainfall seasonality, and/or the total amount of annual rainfall throughout the year.
The aim of the present study is to evaluate the impacts of the climate changes on the ecosystems and the landscape of the Middle Atlas during the last 6 millennia. Our approach is multidisciplinary and based on the analysis of pollen grains, elemental and isotopic geochemistry, and grain size from a fossil record collected in Lake Hachlaf, Middle Atlas. Temperature and precipitation variables have been quantified. They show a moderate change which is superimposed by an aridity trend that is combined with an increase in winter temperature over the past 6000 years. We also observed some noticeable ecosystem and landscape changes with one rapid and quite abrupt climate fluctuation between 5500 and 5000 cal BP detectable in all the proxies used.

Study area
The Middle Atlas Mountains, lying in northwestern Morocco, consist of two geological sets called pleated and tabular Middle Atlas (Fig. 1a). The latter is formed by a Paleozoic basement covered by a Mesozoic thick layer and Cenozoic and Quaternary volcanic flows (Texier et al., 1985;Herbig, 1988;Harmand and Moukadiri, 1986). The Liasic limestone and dolostone are shaped by karstic mechanisms (Martin, 1981;Baali, 1998;Hinaje and Ait Brahim, 2002;Chillasse and Dakki, 2004). In this geomorphological and structural composition, there exist nowadays about 20 permanent or semi-permanent natural lakes (Chillasse and Dakki, 2004) among which we can find the studied site, Dayet (Lake) Hachlaf (33 • 33 20 N; 5 • 0 0 W; 1700 m a.s.l.). This small water body is located about 10 km north-east of Ifrane National Park (Fig. 1b). Available meteorological data (HCE-FLCD, 2004) at Dayet Hachlaf show an average annual rainfall of ca. 600 mm with P w and P s ca. 150 and ca. 70 mm, respectively. The mean January temperature is ca. 4 • C with ca. 90 rainy days per year and ca. 70 frosty days, among which ca. 17 are with snow precipitation. The surface area and depth of the lake change throughout the year, reaching up to respectively 14 ha and 4 m during late spring. The lake is fed by rainwater, snow, surface runoff, and groundwater and has no river inflow. The forest cover around the site (Fig. 1c) is composed of holm oak (Q. ilex subsp. rotundifolia), which is evergreen; zeen oak (Q. canariensis), which is deciduous; and Atlas cedar (Cedrus atlantica) with occurrences of Pinus halepensis. Nowadays, there are some degraded populations of Cedrus atlantica with cultivated lands around the lake. At higher altitude (1700 to 2500 m, Fig. 1c) an herbaceous/shrubby vegetation (Artemisia herba-alba and Poaceae) dominates the landscape.

Materials and methods
In April 2008, a 2.5 m core (33 • 33 2.49 N, 4 • 59 41.57 W) was collected using a Russian corer. Each section of the core was then sub-sampled for the analysis of pollen content (30 samples), grain size (39 samples), organic matter (43 samples) and its isotopic composition (δ 13 C (org) ; 46 samples), and total nitrogen and carbonates (43 samples).
Pollen grains were extracted using a standard laboratory procedure: HCl (20 %), KOH (10 %), ZnCl 2 , acetolysis (CH 3 CO 2 O and H 2 SO 4 ), KOH (10 %), ethanol, and glycerine. The identification and counting of pollen grains were performed with an optical microscope (Leica DM750) using a × 40 magnification (× 63 for accurate identifications). The pollen percentages were calculated on the total sum of pollen grains originating from vascular terrestrial plants. The total of pollen grains counted varies between ca. 200 and 1300. Aquatic plant percentages (including Cyperaceae and Juncaceae) were excluded from the total pollen sum. Cyperaceae were considered as aquatic plants since there are Juncus and Cyperus genera growing around the lake today.
Organic matter amount (OM) was estimated based on the content of the organic carbon in lacustrine sediments (OC), elaborated by spectrometry (NF ISO 14235, 1998). Sediment OC was oxidized in a sulfochromic environment with an excess of potassium dichromate at 135 • C. Subsequently, the determination of chromate ions Cr 3+ formed was analysed by spectrometry.
For total nitrogen (TN), the method used was based on the Kjeldahl mineralization (ISO 11464 : 1994(ISO 11464 : , 2006, but the catalyst used was the titanium dioxide (TiO 2 ). The technique consists in assaying the total nitrogen content in the sediment as ammonium, nitrate, nitrite, and organic form.
Stable isotope ratio measurements of carbon were performed on a Thermo Fisher FLASH 2000 Elemental Analyzer in line with a VG Isoprime Mass Spectrometer at the University of Bordeaux. All samples were pretreated with 1N HCl to remove inorganic carbon. The analytical precision of 0.15 ‰ was estimated from several calibrated laboratory standards analysed along the samples. Stable isotopic ratios were reported as δ 13 C = [( 13 C / 12 C sample / 13 C / 12 C std ) − 1]×1000, where the standard used is Vienna Pee Dee Belemnita (PDB) Besides the multi-proxy analysis, four organic samples were dated. All these dates have been done on bulk sediment. We used the BACON software (Blaauw and Christen, 2011) to compute the age-depth model (Fig. 2). The default 14 C calibration curve used by BACON for terrestrial Northern Hemisphere samples is IntCal13. The AMS 14 C dates were also calibrated using the "CALIB 7.1" program (Stuiver and Reimer, 1986; Table 1). The fossil record continuously encompassed the last 6000 years.
Annual precipitation (P ann ), mean January temperature (T jan ), and precipitation seasonal index (SI) assessment ( Fig. 3) were based on pollen data as follows: where P SI (s) is the seasonal index quantified for sample s; P w is the sum of December, January, and February precipitation; P s is the sum of June, July, and August precipitation; and P ann is the total annual precipitation. The monthly mean precipitation and T jan were obtained using the probability density function of modern plant species (pdf method). This method is described in Chevalier et al. (2014). In order to apply it to a fossil pollen record collected in the Mediterranean area, it required a modern database of Mediterranean plant species distributions and their corresponding modern climate variables. We used a database of plant species that have been georeferenced from Flora Europaea (Jalas et al., 1972(Jalas et al., , 1973(Jalas et al., , 1976(Jalas et al., , 1979(Jalas et al., , 1980(Jalas et al., , 1983(Jalas et al., , 1986(Jalas et al., , 1989(Jalas et al., , 1991(Jalas et al., , 1994 and Hulten and Fries (1986). Additional geographical distributions were obtained from GBIF (2012) and personal field observations using GPS in Morocco. In order to use plant species distributions for the pollen-based climate reconstruction, we assigned pollen taxa to the most probable plant species in our plant database (Table 2). The modern climate variables were extracted from the WORLDCLIM database (Hijmans et al., 2005) and interpolated onto the species occurrences for inferring their pdfs.

Results
During the last 6000 years, the main change in the forest cover is marked by a decline of the pine populations, the expansion of Atlas cedars after 3750 cal BP, and the persistence of the evergreen oaks. Although the latter dominate today the landscape around Lake Hachlaf, the microscope identification of the fossil pollen grains that originate from deciduous or evergreen plants may often be dubious and, therefore, may not be reproducible by another pollen analyst. We have assigned all oak pollen grains to the evergreen Quercus ilex in the climate reconstruction since it is the species that dominates the landscape and its climate envelope encompasses that of the other evergreen species (Fig. 4). All other taxaincluding trees, shrubs, and herbs -also show some changes but within a much lower range than that of the two conifer taxa, Atlas cedar and pine (Fig. 5). We have applied a constrained cluster analysis to depict the main changes in the pollen fossil record. There are four main clusters summarizing the main changes in the ecosystem composition around Lake Hachlaf over the last 6000 years (Table 3).
The grain size analysis revealed the presence of three fractions ( Fig. 3) with the following proportions: clay (22.87 %), silt (60.46 % with 41.9 % of fine silt), and sand (16.67 %). The dominant silty fraction tends to increase from the bottom to the top of the core after a brief decline between ca. 5600 and 5200 cal BP. The sandy fraction follows the same pattern. Clay shows an opposite trend to both the sandy and silty fractions.
Carbonate (CaCO 3 ) content is high throughout the record except around 5200 cal BP (Fig. 3). They are positively correlated with silt and sand. The total organic carbon (TOC) content is also high and varies significantly between 4 and 27.4 % (Fig. 3). The TN remains low throughout the record. The carbon-to-nitrogen ratio (C / N) varies between 9 and 17.4, and the δ 13 C org between −21 and −27 ‰ (Fig. 3). Two origins of the organic matter are thus identified, with lake algae characterized by C / N < 11 and very depleted δ 13 C org and terrestrial plants characterized by C / N > 11 and less depleted δ 13 C org (Fig. 6). δ 13 C org and C / N are positively correlated (Fig. 3). TOC and TN are highly correlated (0.99, Figs. 3 and 6) as well.
In order to interpret the different bio-and geo-chemical proxies within a climatic frame, a pairwise correlation was performed between the three climate variables and δ 13 C, C / N, TN, and TOC (Fig. 7). Although there could be no causal relationship, SI and T jan are well correlated with each another. They are both correlated negatively with δ 13 C and C / N and positively with TN and TOC (Fig. 7).

Discussion
The Holocene climate around the Mediterranean Sea was suitable for the expansion of human populations and their organization towards true civilizations (Kaniewski et al., 2012  The persistence and longevity of many Mediterranean populations may be linked to the relative suitability and also to an overall stability of the Holocene climate. However, climatic events have been recorded within the Holocene (e.g. Rohling and Pälike, 2005) and a causal relationship has been made between some abrupt climatic events and societal changes in the Mediterranean (Berger and Guilaine, 2009;Kaniewski et al., 2008).
In the present study, we have focused on the environmental and climate changes that occurred during the last 6 millennia in the northern part of the Moroccan Middle Atlas Mountains. We have evaluated the vegetation dynamics using the palynological content of a fossil sequence and analysed its bio-and geo-chemical content to reconstruct the overall landscape changes.
The reconstructed T jan and P ann show a relatively low amplitude of change over the last 6000 years (Fig. 3). P ann decreases progressively by ca. 100 mm, which is in line with the aridity trend that has been observed in other fossil records (Risacher and Fritz, 1992;Brooks, 2006;Hastenrath, 1991;Anderson and Leng, 2004;Umbanhowar et al., 2006) and particularly in the Mediterranean area (Pons and Reille, 1988;Julià et al., 2001;Burjachs et al., 1997;Yll et al., 1997;Roberts et al., 2001;Valino et al., 2002;Jalut et al., 2009) and northern Africa (Ritchie, 1984;Ballouche, 1986;Lamb et al., 1989). At a more regional scale, reconstructed P ann is coherent with that obtained from Lake Tigalmamine (Cheddadi et al., 1998) which shows a decreasing trend over the last ca. 5000 cal BP. The arid trend observed after ca. 5 ka cal BP is marked by a spread of Poaceae and a progressive replacement of pines by Atlas cedars, which better stand the high seasonal contrast of precipitation at the altitude of Lake Hachlaf. SI increased from 3 to 7 times over the last 6000 years (Fig. 3). A study of drought thresholds influencing the growth and photosynthesis was performed on different cedar stands and species (C. atlantica, C. libani, C. brevifolia, and C. deodara) of different origins (Aussenac and Finkelstein, 1983). This study showed that, among many conifers, cedar trees may keep a sustained photosynthesis activity even when drought is very high. Thus, a strong precipitation contrast between P s and P w (Fig. 3) may not affect the Atlas cedar overall growth as long as the total amount of rainfall is sufficient (higher than 600 mm yr −1 ) and the winter temperature is low enough (below 6 • C) for the vegetative cycle (Aussenac et al., 1981). The Mediterranean climate is known for its strong seasonal distribution of precipitation ) and carbonates concentrations (CaCO 3 ), January mean temperature (T jan ), annual precipitation (P ann ), winter and summer precipitations (P w and P s ), and precipitation seasonality index (SI). The red rectangles are pointing the values of present-day T jan , P ann , P w , and P s (HCEFLCD, 2004); the red line shows the limit 3.7 ka cal BP; and the blue rectangle shows the time interval of the cold phase 5.2 ka cal PB. throughout the year. Summers are fairly dry, and most of the annual precipitation occurs during the cold months (end of autumn and beginning of winter). Currently, 75 % of the Moroccan territory with a grassy or wooded vegetation (thus excluding the desert) records between 500 and 800 mm of annual rainfall with an SI between 5 and 8 (Fig. 8). The whole range of SI in Morocco is between −1 in areas where P ann is less than 100 mm with a random distribution for instance in the south of Morocco and 15 in areas where the annual rainfall is quite high (over 800 mm) and occurs mainly in the winter season such as in the Rif Mountains today (Fig. 8). SI is higher in mountainous areas. Nowadays, in the areas surrounding Lake Hachlaf (located at ca. 1600 m elevation) SI is around 5. This SI has changed over the past 1000 years as confirmed, at least between 6000 cal BP and today, by the studied fossil archive (Fig. 3). The amplitude between P w and P s precipitation has increased 2 to 3 times towards the present (Fig. 3). Since P ann has a decreasing trend, the opposite increased seasonality is related to a significant reduction in the amount of rainfall during the months of June, July, and August (Fig. 3). This strengthening of the contrast between P w and P s had a rather limited impact on the dominating taxa because they can with-stand the summer drought and the overall amount of P w remained sufficient for their persistence. However, a change in the amplitude of SI has probably favoured those species best adapted to the length of the dry season, for instance evergreen oaks rather than deciduous. Pollen-based climate reconstructions from records collected in the Alboran Sea (Combourieu-Nebout et al., 2009) and Italy Peyron et al., 2013) suggest a rather steady and low seasonal contrast between P w and P s (about 2 times) over the past 6000 years. This discrepancy between the reconstructed SI from Hachlaf and the marine record may potentially be related to the fact that marine records collect pollen grains from a much wider geographical source area than continental (mountainous) records, which probably tends to smooth the local/regional changes. The reconstructed seasonality from the Italian records Peyron et al., 2013) is buffered by the less abrupt precipitation seasonal contrast at European temperate latitudes than at arid Mediterranean ones. SI was lower than 5 before 3750 cal BP despite an amount of precipitation between 600 and 700 mm yr −1 (Fig. 3). During that period, water probably persisted in the lake all throughout the year, which allowed the presence of aquatic plants (Fig. 5) flowering during late spring and summer, and algae identified in the pollen data, through the low values of δ 13 C org and the C / N ratio being greater than 11 (Figs. 3 and 5). The proportion of aquatic plants cannot be directly related to a high lake level and may not be used to state the lake level changes but only the presence of water in the site. The δ 13 C org and C / N (Fig. 6) provide information concerning the origin of the organic matter (in situ production versus input from the catchment area) but not on the lake level changes. Thus, high δ 13 C org and C / N ratios (Fig. 3) with low presence of aquatic plants (Fig. 5) may not be inconsistent in cases where there is a low terrestrial input (low sand / silt, Fig. 3) during a period when the lake level is high.     The relationship between δ 13 C org and the C / N ratio indicates the occurrence of two main types of organic matter mainly originating from a C3 metabolism. Lacustrine algae can be considered as dominantly autochthonous; in the lower part of the record, the organic matter, with higher C / N ratios and less depleted δ 13 C org , corresponds to a terrestrial input. Indeed, fresh organic matter from lake algae is known to be protein-rich and cellulose-poor with molar C / N values commonly between 4 and 10, whereas vascular land plants are protein-poor and cellulose-rich, creating organic matter usually with C / N ratios of 20 and greater (Meyers, , 2003. However, a C / N ratio > 11 may correspond to a mixture of both local and terrestrial organic matter (Fig. 6).
After 3750 cal BP, Atlas cedars noticeably spread around the site, while the pine populations strongly regress. A series of fossil pollen records in the Middle Atlas show that Atlas cedar populations expanded after ca. 6 ka cal BP. The sus- tained expansion of Atlas cedar after ca. 3750 cal BP around Lake Hachlaf expresses its late occurrence at higher altitude. Around Lake Tigalmamine , the Ras El Ma marsh (Nour El Bait et al., 2014), and the Ait Ichou marsh (Tabel et al., 2016), which are all located at about 100 to 200 m altitude below Lake Hachlaf (ca. 1700 m a.s.l.), Atlas cedar occurs much earlier. The expansion of Atlas cedar around the lake is probably related to both an upslope spread and a south-north migration.
During this ecosystem transition we observe a major change in both P ann and T jan . The increase of SI after 3750 cal BP is due to a combined increase of P w and decrease of P s (Fig. 3). The expansion of cedar forests in the studied area may be related to their better adaptation to strong SI than pines at higher altitude.
Competition is another parameter that might be worth considering. After 3750 cal BP, the C / N ratio is below 11 and the δ 13 C remains below −26 ‰, which suggest the important primary productivity of the lake associated with low input of land-plant-derived organic matter. Atlas cedar forests have a more important growth in both height and diameter than pines, which leads to a higher biomass production. This is linked to the genetic model of growth that is very distinct between the two taxa (Kaushal et al., 1989). Thus, the ex-pansion of the Atlas cedar population around the site may explain the high input of OM into the lake.
Over the last 6 millennia, superimposed to the overall climate trend, we observe one relatively abrupt event between 5500 and 5000 cal BP during which T jan declined by about 2 • C compared to its average over 6000 years. A climatic transition between 6 and 5 ka cal BP at the end of the Holocene thermal maximum has been globally identified (Steig, 1999;Mayewski et al., 2004;Wanner et al., 2008;Brooks, 2012). This transition has been recorded by a wide range of climate proxies (e.g. Kaufman et al., 2004;Jansen et al., 2009;Seppä et al., 2009;Bartlein et al., 2011) and has been related to different biosphere feedbacks and potentially to a decay of the remaining Laurentide Ice Sheet (Renssen et al., 2009). All proxies from the Hachlaf sequence as well as the reconstructed climate variables have recorded marked changes during that period of time. SI has the lowest value of the record, and a succession of abrupt changes are recorded in the C / N ratio, the grain size fractions, the δ 13 C, TN, TOC, and CaCO 3 (Fig. 3). Carbonates, considered as a "paleothermometer" (Meyers, , 2003, also decrease abruptly around 5200 cal BP (Fig. 3). The latter may be linked to a low evaporation of the lake which may have been favoured by low winter temperature around 5200 cal BP. The fine grain size sediment also increased as a consequence of low seasonal precipitation contrast and/or a continuous sediment input to the lake. Such sustained input of clay and decreasing carbonate content suggest a higher lake level between 5500 and 5000 cal BP (Fig. 3). Thus, the T jan and SI decrease may have contributed to the higher lake level or at least to the presence of water throughout the year (Fig. 3). At the same time, the sand-to-silt ratio is very low, which confirms a low energy during the sedimentation process. The major change in the ecosystem composition around the lake is the rapid collapse of the pine forest which has inevitably released an important amount of terrestrial carbon (biomass) into the lake (positive peaks in δ 13 C and C / N, Fig. 3).

Conclusions
This study marks a new contribution to the knowledge of past climates and environmental history in north Africa mountainous areas. The range of climate change in the Middle Atlas, Morocco, was rather minor between 6000 cal BP and the present. Annual precipitation and January mean temperature have respectively varied within a range of 100 mm and 2 to 3 • C. However, they both show a trend towards a more arid and warmer climate as well as a higher rainfall seasonality. P ann became as contrasted as today after 3750 cal BP. The aridity trend observed in Hachlaf over the last 6000 years is consistent with other climate reconstructions available from other Mediterranean fossil records. Besides these overall climatic trends, we also observe an abrupt cold event between 5500 and 5000 cal BP which is well marked in all environmental proxies from our studied fossil record. The δ 13 C and C / N ratios, which are well correlated with each other, suggest an increase in the organic matter input from the catchment area. Concomitantly, the pollen record indicates a decline of the pine forest which may have contributed to the organic matter input into the lake too. The marked change in both the carbonate content and clay composition of the record was probably related to a perennial presence of water throughout the year. Synchronously, seasonality index and January mean temperature were the lowest of the record which has contributed to a reduction of the evaporation.
The increase in rainfall seasonality has probably favoured the expansion of Atlas cedars around the studied site at the expense of the pine forest.