Structure, evolutionary context and chronological data of the Monforte de Moyuela Roman dam (Ebro Basin, NE of Spain)

The Monforte de Moyuela dam, also known as Ermita de la Virgen del Pilar dam, is a Roman reservoir built on a tributary of the Aguasvivas River (Ebro basin, Spain). A multidisciplinary study has been carried out to investigate this kind of Roman water infrastructure. It is the fifth‐highest dam (16.8 m) in the Iberian Peninsula and the seventh in the Roman Empire. The initial dam was built ca. 100 B.C.–10 A.D., probably in the period of Augustus, like other nearby Roman dams. It was quickly filled due to the extreme and generalized anthropic degradation in the basin during the Roman period. During the mid‐2nd century, the wall was increased in height and its final silting was dated to the early 7th century. The study of the opus caementicium mortars shows constructive differences between the initial and subsequent phases of the wall. These mortars provided charcoal for dating the two phases. In addition, the stratigraphic and edaphological study of the reservoir's sedimentary fill, together with the 14C ages, allowed us to reconstruct the two main activity cycles and the final siltation of the dam. Subsequently, the dam broke in two phases, which created the two stepped sections located on the current valley bottom. The data obtained allowed the creation of a geomorphological map and an evolutionary model of the valley showing the main differentiated stages, from the initial construction of the dam to its final opening. Although some remains of canals downstream of the dam have been identified, the use of this dam, which remained active for several centuries, still needs to be investigated in greater detail.


| INTRODUCTION
The Mediterranean climate is characterized by rainfall scarcity and interannual variability (Lionello et al., 2006), with long periods of drought. Therefore, the regulation of fluvial flows is important for the optimal exploitation of water resources. In Roman times, water played a role in all social activities and in guaranteeing high agricultural yields, the basis of the Empire economy (Decker et al., 2017;Erdkamp, 2005). The great Roman expansion took place in the context of a warming climate (Harper & McCormick, 2018), during the so-called Roman Climate Optimum (RCO), between ca. 100-150 years B.C. and ca. 200 years A.D. (Harper, 2017;McCormick et al., 2012). This period was exceptionally warm (Ljungqvist, 2010) but better for cereal production than presentday environments (Dermody et al., 2014). There is no consensus on the humidity at that time (Dermody et al., 2011) although significant regional variability is probable (Erdkamp, 2019). Some studies point to increasing aridity in the Western Mediterranean throughout the Roman period, accompanied by intense landscape anthropization (Currás et al., 2012;Ejarque et al., 2022;Martín-Puertas et al., 2008;Peña Monné, 2018;Peña-Monné et al., 2004).
Under such environmental conditions, dams and distributary canals were built to counter natural environmental shortages, obtain good agricultural profits, and meet other social (termae, fullonicae, nymphaea) or economic (mills, mining) needs. However, despite their relevance, dams are poorly understood among large Roman constructions, perhaps due to their poor conservation. Their loss of capacity due to siltation, breakage due to construction problems or large floods, abandonment, and the reclamation of their building materials for other constructions resulted, in many cases, in only a few ruins being left in the archaeological record.
According to inventories made by Fernández Casado (1961), Quintela et al. (1987), Arenillas and Castillo (2003), Castillo and Arenillas (2002), Saldaña (2011), Castillo (2015), Sánchez and Martínez (2016), and Barahona (2017), it is estimated that there are almost 70 probable Roman dams in the Iberian Peninsula. Large dams are grouped into three sets: (i) dams on fluvial tributaries of the Ebro River, NE Spain; (ii) dams from Extremadura in the Guadiana River, especially around the city of Merida (Emerita Augusta), and (iii) dams of the Tajo basin, south of Toledo. To date, studies have focused on the largest and best-preserved dams. Some of them are Almonacid de la Cuba dam (Aguasvivas River, tributary of the Ebro River, Figure 1) (Arenillas et al., 1995;Beltrán & Viladés, 1994;I. Hereza et al., 2000;J. I. Hereza et al., 1996); Muel dam (Huerva River, Ebro River basin, Figure 1) (Arenillas et al., 2006;Magallón et al., 2016;Uribe et al., 2010Uribe et al., , 2012Uribe et al., , 2013Uribe et al., , 2016; Cornalbo, Proserpina, and Consuegra dams, in the Guadiana River basin (Arenillas et al., 1992(Arenillas et al., , 2007García-Diego et al., 1983a, 1983bÁlvarez Martínez, 2007;Álvarez Martínez et al., 2002;Aranda & Sánchez Carcaboso, 2000;Aranda et al., 2006;Giles Pacheco, 2011;Martín Morales et al., 2002;Rodríguez Untoria, 2011); and Alcantarilla dam, in the Tajo River basin (Aranda & Sánchez Carcaboso, 2000;Aranda et al., 1997;Arenillas & Barahona, 2009a, 2009b; Barahona Oviedo et al., 2014;Barahona, 2018aBarahona, , 2018bCelestino, 1976;Sánchez Abal, 1977). For this reason, these are the most widely cited dams in the literature on civil constructions (Baba et al., 2018;Jansen, 1983;Mays, 2008Mays, , 2010Schnitter, 1994;Smith, 1970Smith, , 1971among others). These studies focus on construction and styles, and most of them lack rigorous dating, casting doubt on the Roman origin of some of them (Feijoo, 2005(Feijoo, , 2006. We consider that, due to their nature and function, the study of dams must comprise an integrated approach that includes environmental data, contextual records, and accurate chronological frameworks. This can be achieved via geoarchaeological approaches, as proposed here for the Monforte de Moyuela dam (hereafter, Morforte dam), also known as the Ermita de la Virgen del Pilar dam, located in the Aguasvivas River basin, an Ebro River tributary. This is a little-known dam, although it was the fifthhighest (16.8 m) in the Iberian Peninsula and seventh-highest in the Roman Empire. Previous research encompassed a brief reference to I.  and Arenillas (2002Arenillas ( , 2003 and precise data about the elements of the dam building in Arenillas et al. (2005). Our objectives were as follows: (i) to determine the geological context of the basin to assess the original distribution of the lithologies involved in the dam construction; (ii) to produce a detailed geomorphological map establishing the general context of the dam, the reservoir, and the sedimentary infill; (iii) to define the structural units of the dam and undertake mortar analyses to establish different constructive stages and relative chronologies; (iv) to determine the infill size, identify morphosedimentary units, There are more than 17 snow days per year, especially in the head basin (Sierra de Oriche, 1383 m). Snowmelt and the karstification of limestone areas located on the head basin allow constant river flow, although with marked shortages in the summer. The average flow of the Santa María River was estimated to be around 0.13 m 3 /s (Arenillas et al., 2005).

| ARCHAEOLOGICAL CONTEXT
The study area lacks in-depth and systematic archaeological research. During the Second Iron Age (5th to 2nd century B.C.), it constituted a border zone between pre-Roman Celtiberian and Iberian peoples, centred on the Aguasvivas River. The bestknown archaeological site is Los Castellares in Herrera de Los Navarros (Burillo, 1980(Burillo, , 2005 and those closest to the Monforte dam are Cabezo Aparicio or Samper (Simón, 1992); all are dated to the Bronze to Iron Ages.
During the Roman period, a series of secondary roads ran through this territory, connecting the lands of the Jiloca and Huerva rivers, with Aguasvivas and Ebro rivers. These roads would have been tributaries of two routes mentioned by the written sources. First, the road mentioned by the Antonine Itinerary (It. . 446-448, Cuntz, 1929) from Caesar Augusta to Laminio (Fuenllana, Ciudad Real), one of the worst-known roads of the Iberian Peninsula (Magallón, 1987). Second, the road from Contrebia Belaisca (Botorrita, Zaragoza) to Leonica (Magallón, 1987), according to the Cosmography of Anonymous of Ravenna (IV, 43, Schnetz, 1940). Even though some of the Roman settlements cited in the written sources have not been recognized, other Roman archaeological sites, whose names remain unknown, have been documented. This is the case with the archaeological remains found at the Pueyo de Belchite site (2nd B.C. to 2nd century A.D.) (Rodríguez Simón & Díez de Pinos López, 2015). The urban center of the Pueyo de Belchite must have been supplied by the nearby Roman dam of Almonacid de la Cuba, with a chronology similar to that of the Monforte dam. In Late Antiquity, the Roman remains of the villa of La Malena in Azuara (Royo, 2010) stand out, with the main occupation between the 4th and 5th centuries A.D. (Royo, 1992).

Ant
The landscape changed during the Islamic civilization and made possible the installation and development of peasant communities located along the fertile alluvial terraces of the Aguasvivas River basin, which due to complex irrigation systems allowed intensive cultivation. The Muslim presence in the area was intense and has left place names such as as Nepza, Letux, Lagata or Azuara (Utrilla, 2010).

| Dam structure
Several opus caementicium mortar samples were taken to determine the different constructive units. The mineralogical composition was made by X-ray diffraction (XRD) analysis. The total sample was analysed by powder XRD on a PAN analytical X'Pert PRO X-ray diffractometer fitted with a Cu anode. Their operating conditions were 40 mA, 45 kV, divergence slit of 0.5º, and 0.5 mm reception slits. The samples were scanned with a step size of 0.0167º (2θ) and 150 ms per step. The samples were further characterized using the powder method between 5 and 60º (2θ). The specimens were quantified using Match v.3 and Fullprof software for Rietveld analysis (Rietveld, 1969;Rodriguez-Carvajal, 1993;Young, 1995). To identify PEÑA-MONNÉ ET AL. | 487 the phases present, the Crystallography Open Database (COD) reference standards were used. To know the clay minerals, a sample of oriented aggregates was prepared, studied with XRD analysis between 2/20º, and then quantified.

| Stratigraphic and pedological studies of the reservoir infill
Two stratigraphic columns were logged in the sedimentary fill of the Monforte dam, using the escarpments formed by the incision of the Santa María River. Magnetic susceptibility (in SI units) was measured (0.1 m analysis interval) using a KT-10 model susceptometer of Terraplus. Moreover, 14 samples were collected from the profiles for further study. For grain size analyses, an AMP0.40 W220 HZ59 CISA device with sieves of ¼ intervals between −1 and 4.75 Φ units was used. From the size distribution, a series of granulometric parameters were obtained using the software GRADISTAT (Blott & Pye, 2001).
The carbonate content was estimated using a Geoservices manocalcimeter. The samples were analysed for determining the fossiliferous content. Finally, the percentages of the traction, saltation and suspension populations were calculated according to the methodology of Visher (1969).
Besides, in the pedological study, the sediment layers and soil horizons were described according to FAO guidelines (FAO, 2006).
Organic carbon was determined by the wet oxidation method (Nelson & Sommers, 1982), and organic matter was calculated using the van Bemmelen factor (1.724).

| 14 C datings
Samples were taken to establish the chronology of the dam construction and the sedimentary infill of the reservoir. Small charcoal samples were found in the mortars, and they were used for the opus caementicium in two constructive units. As charcoals were scarce in the mortar core, we were not able to date all layers.
Besides, isolated charcoals, charcoal layers were also found in the stratigraphic and edaphic profile records from the infill. Two isolated charcoals were sampled in the lower section. Also, two levels with abundant micro and macrocharcoals were found in the upper section.

| Dam structural characteristics and chronology
The Monforte dam is a gravity wall adapted to the irregular topography shaped by the river on the Paleogene conglomerates. It has a large initial wall, which was later increased in height (1 and 2 in The main wall is formed by a concrete core (B in Figure 5a Table 2). The mortar composition of the B4 layer was not observed in the other layers of the dam structure; however, it should be noted that it was not possible to obtain samples from the upper layers.
B layer of opus caementicium B was examined in detail, and two charcoal samples were obtained ( Figure 5d). The lower one was collected 12 cm from the bedrock above which the wall was built (sample PEP-1), while the upper one was taken more than 2.1 m above the wall base (sample PEP-3) (Figure 5f,g). Some 8-10 mm charcoals were easily separated from the mortar. The dating (Table 3a)  aggregates is also similar, between 5% and 10% of CaO. Besides, the lime:sand ratio is 1:19; the lime (including other sands):sand ratio is 1:4 and the lime:sand:gypsum ratio is 3:14:3 (Table 2). However, sample PEP-D6 shows some mineralogical changes. Calcite increases (52%), quartz (38%) diminishes (Table 1a; Figure 6e) and CaO reaches 30%. The ratios of lime:sand (3:5) and lime:sand:gypsum (3:5:2) also change (Table 2). Besides, the mineralogical composition of PEP-D6 coincides with that obtained for the PEP-X sample taken in the residual mortar from the other river side (Figure 6f,g). This is probably because they belong to the same constructive unit. Layer D7, formed by large slates, is not only on the protective wall top but also on the principal wall forming a generally flat surface. Although the mortars of units A-E and X were examined, no charcoal fragment was found, and therefore, we do not have absolute datings for the protective wall.
The extended wall was built on the roof of the first dam (unit G). It is, at least, 2.6 m high, and four layers (G1-G4) made of opus caementicium were identified (Figures 6a and 7a). Thus, the dam length increased from 52-53 m to 86-87 m. Besides, the dam front is at least 16.8 m in height (Arenillas et al., 2005), although the water reservoir must have been relatively filled with sediments. The lateral expansion of the reservoir occurred mainly on the left side, so the wall was extended at an angle of 50°(3 in T A B L E 1 (a) Mineralogical composition by XRD of the mortar samples taken from the interior of the dam wall and (b) XRD patterns by phyllosilicate minerals for all the studied samples.     (Table 1b). The lime (including other salts):sand ratio is 1:1 and the lime:sand:gypsum ratio is 1:2:1 (Table 1b).
The mortar of unit G1 shows many charcoal fragments of millimetric size. A 6 mm charcoal fragment was selected (PEP-4)   (Reimer et al., 2020) and expressed with 1σ and 2σ).

| Other construction remains
The archaeological survey brought to light scattered remains of possible water conduits related to the dam. One of them is located 150 m downstream of the dam, about 5-6 m above the river channel, on a rocky ledge at the entrance of the limestone canyon (Estrecho de la Virgen del Pilar). It is a small isolated construction formed by large limestone and conglomerate fragments included in a hard mortar (Figure 8a). There is a cavity at its base, bordered by large vertical limestone blocks and covered with well-carved and vaulted tufa ashlar (Figure 8b). This structure belonged to a water conduit whose length is still unknown. There are notches and grooves in the    F I G U R E 9 Contour of the estimated water reservoirs. Original dam at 910.2 masl and after the construction of the extended wall at 912.9 masl.

| The reservoir and its sedimentary record
The dam reservoir occupied a large area (~120 m maximum width), narrowing toward the tail located around 770-790 m away from the Monforte dam ( Figure 9). The initial altitude level of the wall was calculated at 910.2 masl and that of the second wall at 912.80 masl (as the minimum known), according to Arenillas et al. (2005). This allowed us to estimate a reservoir surface of between 81,237 m 2 (in the first construction) and 111,480 m 2 (after the second construction), increasing its surface by 25% (Figure 9). This These steps allowed us to differentiate a lower unit and an upper unit, which we used in our analysis. These two levels are especially well-defined in the area near the dam wall, in both river margins up to the mill (Figures 3 and 10a). They are better developed and have visible outcrops on the right riverside (Figure 10b). The two stratigraphic sections, with a total sedimentary thickness of 7.9 m  Table 4) and magnetic susceptibility (Table 5). (2σ) (Table 3b). These datings do not have absolute chronological value but point to the post-quem age. However, they show the abundance of paleo-fires in the basin before the Roman Epoch.

| Lower section
Unit B: This is a 1.7-m-thick unit consisting of massive brown silts with abundantly scattered clasts (<1 cm diameter) and tabular and channelized gravel beds, with clasts up to 5 cm. They include gastropods, charophytes, root bioturbation and coal remains. This unit was formed in areas temporarily occupied by a sheet of water where erosive streams carried silt and gravel. Unit B may be related to the dam-filling stage (Figures 12b and 13b).  (Table 3b). This charcoal-rich layer also included four fragments of Roman pottery.
These ceramic finds are fragments of common cooking ware, with rough-textured clay bodies without surface treatment, of presumed local manufacture. Therefore, they are inconclusive types and cannot be used to reinforce the chronology accuracy.  (Table 3b).

| Edaphic analysis of the sedimentary fill
In addition to stratigraphic records and analyses, the morphological description of the soil and underlying sediments was carried out, by focusing on unit F. This is an interesting unit because it is located on top of the fill and was exposed for at least the last 1400 years.
Besides, it was necessary to determine whether the other units contained edaphic anomalies like unit C. Unit F is the transition between the two fill sections of the reservoir, formed by a lateral deposit affected by agricultural activity.
The upper unit (F) corresponds to young, poorly developed soil, with a reddish yellow colour (7.5 YR 6/6) when dry and strong brown (7.5 YR 4/6) when wet (Table 6). It has many coarse fragments (>2 mm), from fine to coarse gravel measuring 2-60 mm, polygenic in nature and subrounded (tabular or flat) in shape; some of the fragments have calcium carbonate coatings without a defined distribution in the clasts, which evidence their origin from eroded soils from the upper pediment surface. The soil is strongly calcareous (10%-25%, w/w), with its maximum on the surface and without morphological evidence of secondary or edaphic carbonates, as a result of its youth. The organic matter content is maximum in this unit (around 2%, w/w) and decreases progressively with depth. This is a demonstration of certain soil development or edaphization that is not observed at greater depth (Table 6). The soil structure, moderately developed, is fine and granular in the topsoil, turning into subangular blocky and medium-sized with depth. In the topsoil, very few fine-sized roots can be found. In addition to the bioturbation caused by roots, the presence of earthworm channels, some of them empty, others full of worm casts, is remarkable.
In the 85 superficial cm, we did not find gastropod shells, which appear only at greater depths (unit E), at the limit of the lithic discontinuity. This is not the only change detected, since at depth the sediment turns brown and grey. Thus, unit E is light yellowish brown (10YR 6/4, dry) and dark yellowish brown (10YR 4/4, wet), according to the Munsell colour chart, and has practically no stones. This unit has a prismatic structure, and it is common to observe charcoal, which is also present at greater depths. In addition, the organic matter content in unit E drops to half that of the topsoil. This value, with slight fluctuations (which correlates with the texture of the sediment), is maintained up to 8 m.
In unit D, a greyer layer appears at 320-400 cm depth; specifically, it is light grey (10YR 7/2) when dry and brown (10YR 5/3) when wet, and more snail shells were found (Table 6, sample 7); moreover, this layer is slightly richer in organic matter (1.34%) than the top and bottom layers. It is in this layer that mottles begin to appear and continue throughout the remaining units of the lower part. The RPEP-3 charcoal sample dated to 14 C comes from layer 7 (Table 3a) and also contains fragments of Roman pottery ( Figure 11). Specifically, fine-sized mottles are abundant in the second part (units C, B and A), and their presence increases with depth. They indicate that the sediments, usually under wetting conditions (reducing), had some local oxidizing conditions (in fauna channels and root pores). These mottles can be contrasted (with a 5YR hue in a sediment matrix with a 10YR hue, both when dry), and they increase their presence with depth: there are few (5% v/v) in unit C; some (5-15% v/v) in unit B and many (15-40%) in unit A (Table 6).

| DISCUSSION
A general evolutionary framework can be established from the different proxies used in the study of the Monforte dam. To achieve this objective, it is necessary to discuss and assess each proxy in the following order: first, the chronological data provided by the mortars  Munsell colour (dry) light brown grey pale brown light brown pale brown of the dam wall and the sedimentary fills of the reservoir; second, the linkage of the sedimentary fills with the regional paleoenvironmental framework and finally, the dam breakage process and its later geomorphological evolution.

| Chronology of the dam construction
There are several possible chronological frameworks for the dam's construction, each with different levels of precision, relating to constructive styles, types of materials and mortars. These characteristics, among others, are normally used to establish the age of Roman constructions (Barahona, 2018a). In the first study of the Monforte dam, Arenillas et al. (2005) applied constructive criteria (size and materials) to establish the dam chronology. They proposed that "by its typology, this dam is probably one of the first dams built in Hispania (1st century AD)". Considering that there are Roman constructive elements (ashlars' size and shape, type of mortar, structural arrangements) that were also used in later constructions or repairs (like those in the Proserpina dam, Feijoo, 2005Feijoo, , 2006, even older charcoals can be found, eroded from accumulations from the surroundings and washed into the reservoir. Isolated charcoals found on the profile may raise concerns about age reliability; however, due to their abundance, it is possible to achieve higher chronological accuracy. In several cases, they originated from fires in the basin and can yield information about deforestation and erosion. However, the most accurate chronology is provided by the dating of the materials from the dam wall construction. Thus, the lime mortars can contain chronological information. Many studies obtained 14 C absolute ages using carbonated mortars but the results have large standard deviations in datings (Hajdas et al., 2012Hayen et al., 2019;Lubritto et al., 2018;Ringbom et al., 2014). Similar results are obtained by using luminescence (OSL) (Urbanová & Guibert, 2015. The organic remains (small twigs, leaves, charcoal fragments) of the dams are difficult to find but very useful for dating.
Two wood fragments were collected from the foundations of the walls of the Almonacid de la Cuba dam (Arenillas, 2002).   (Table 3a). Thus, the extended wall was probably built in the second half of the 2nd century A.D.
As well as the chronological information, the composition of the  (Table 1a); kaolinite prevails among the phyllosilicates (Table 1b); there are only limestone fragments in the opus caementicum. There are also notable differences in the lime:sand:gypsum ratios (Table 2 6.2 | Reservoir fill and regional environmental conditions In many sedimentary fills of antique reservoirs, since the stratigraphy is not accessible, mechanical drilling is used to obtain samples for radiometric dating or archaeomagnetic stratigraphy. This is the case with the Muel dam Silva Aguilera et al., 2008) and Almonacid de la Cuba dam (Arenillas, 2002;Pueyo et al., 2001). It is also possible to carry out excavations to access the complete stratigraphy and perform sequential sampling for radiocarbon dating, as in the case of the Muel dam (Uribe et al., 2012). However, at the Monforte dam,  Table 3a).
The other two large reservoirs from the Aguasvivas basin ages are similar to ours. Only the foundation age of the Almonacid de la Cuba dam is known (Arenillas, 2002).  (Uribe et al., 2012), and it was not extended after that. In summary, the three dams have average dates of construction around the second part of the 1st century B.C.

Radiocarbon dates include the Epoch of Emperor Augustus and
confirm the archaeological studies. Siltation ages vary between the 4th century for the Muel dam and the middle of the 2nd century for the Monforte dam. The Monforte dam is the only one that was subsequently increased in height, which shows that the purpose for which it was created was still valid.
In the outcrops of Figure 11, it can be seen that the sedimentary sequence of the initial and extended dams began with sediments from the flooded environment during the activity of the Monforte dam (units A and D). In both cases, they were covered with debris deposits from the runoff (units B and E) produced during the times when the dam was no longer active. The curve of the magnetic susceptibility shows the same pattern (  (Currás et al., 2012).
Likewise, these conditions are observed in the coastal hinterland of Emporion, according to the recent study of Ejarque et al. (2022).
Deforestation in the Mediterranean region began before the Roman Epoch. However, it is the expansion of agricultural activity at that time that caused the greatest loss of vegetation and the main changes in the landscape (Luterbacher et al., 2012). Some authors (Reale and Dirmeyer, 2000;Reale and Shukla, 2000) argue that  (Figures 12d and 13d). After that, the centre and eastern area of the dam wall collapsed. In our opinion, the wall did not collapse up to its base but it was a partial breach (Figures 12e and 13e). This is supported by the fact that only a part of the upper section of the filling was eroded, while the lower section was retained, giving rise to the development of a new valley floor in coincidence with the contact between both sections ( Figure 12e). In addition, near the dam wall, the lower section lost unit B and part of unit A, which were replaced by a new accumulation (unit G in Figures 12e and 13f), showing a continuous terrace level at that height. At the same time, the upper unit scarp retreated laterally (Figures 12e and 13f). In no other way could the formation of the two steps showing both sections have occurred.
Vertical aerial photographs taken in 1946 and 1956 were reviewed to determine whether the steps could have been a consequence of recent flattening works. However, the stepped morphology already existed at that time. The western side of the dam wall was not affected because its base lies on Paleogene conglomerates, favouring its conservation up to the present. Finally, sometime later, the river canyon was completely open and headwater erosion formed the current course of the Santa María River (Figures 12f   and 13g). According to Arenillas et al. (2005), the dam collapsed due to a lack of stability in the construction, especially after the wall was extended, because the relationship between wall thickness and height is not adequate. However, the reservoir remained closed for more than four centuries after being increased in height and operated up to its final siltation. Some large flows may have overpassed the wall on its eastern side, causing part of it to crumble.
The most recent evolution formed a flat fill, 2 m in height, on the riverbed, probably due to a temporary closure of the incision ( Figure 3). This surface is located near the narrowing of the canyon and it is part of the floodplain.
It is not possible to infer a dam management system because there are no gates or spillways. There are only a few conduits, as shown, about 100 m below the dam wall. The structure was probably used to provide water for irrigation of the Santa María valley. There is no information about Roman city mills, small populations in the surroundings, or Roman villae that would allow us to speculate about another type of water use. In any case, it is necessary to intensify the surveys, especially in the lower section of the Santa María River, to fully understand the construction and maintenance effort of this water reservoir for many centuries.

| CONCLUSIONS
The Monforte dam is part of the set of hydraulic works built during the Roman period during the reign of Augustus in the Aguasvivas and Huerva river basins, both tributaries of the Ebro River. Its construction could have been the result of Augustus' new territorial organization, which unfolded during his reign with the promotion of the municipia, the foundation of a new colony, Caesar Augusta, and the creation of its road network. As the dam wall collapsed after Roman times, it was possible to observe, record and sample the wall interior and most of the sediment fills of the reservoir, allowing an interdisciplinary study. The dam is composed of an initial wall, built between 50 BC and 66 AD. Later, the dam wall was increased in height in the mid-2nd century. The analyses of the opus caementicium mortars allowed us to establish differences between these constructive phases. The initial wall has a siliceous composition, while the extended wall is carbonatic. They also differ in the lime:sand:gypsum ratios. The sedimentary study of the reservoir fill shows the sequences of activity and silting of both stages, reflected by deposit composition and changes in the magnetic susceptibility measurements. The first reservoir was silted before the mid-2nd century and the enlarged reservoir was built in the early 7th century ( There is an exception in a level dated to ca. 3rd century A.D., with higher organic matter content and roman potsherds.
The siltation of the reservoir was rapid as a consequence of the intense erosion in the basin, promoted by intense anthropic impact by land use change during the Roman Epoch. The accumulation rate was estimated at 1.7 m/century. Subsequently, the dam broke in two phases, which first gave rise to erosive steps in the fill and then the emptying of the central sector.
Some construction remains are related to water conduction downstream of the dam, probably built for agricultural irrigation in the lower valley of the Santa María River. This dam, barely known due to its location, deserves to be a properly preserved and valued site of interest, which would enable visitors to appreciate its construction and the natural setting in which it is located.