Since Janet DeLaine published her thesis on the Baths of Caracalla (DeLaine 1997), there has been an increasing interest in the economics of ancient construction. While the foundations of the study were set by Choisy in the 19^th century, it has been in the last thirty years that the field has developed into a study of how the consumption patterns of materials and labour necessary in large construction projects had an impact on the Roman imperial economy (Camporeale 2010; DeLaine 2017; Goldsworthy and Zhu 2009). Most of these studies, however, have not discussed water consumption as part of the building cycle besides briefly acknowledging that it was necessary for construction. Behind this lack of specificity lie three problems. The first is an economic issue: prima facie, water could be available for free, leading to the assumption that water was not likely to have had a major economic impact on the project’s budget. A second would be an engineering problem: there have been no attempts to measure water to mortar ratios because many variables (from the type of sand to the weather at the time of construction) have an impact on the volume of water needed to carry out a construction project, making it difficult achieve an accurate solution (Brune 2010: 18, 330). The third one is the scale of the constructions: small plastering operations could have been done by transporting from a workshop required amounts of pre-mixed materials, but large construction sites would have needed mixers constantly on-site.

However, water is too essential in construction sites to dismiss it. Water is not only needed for mixing mortar; it is needed for several daily, mundane activities like drinking, cleaning, or soaking the handles of loose tools. Nowadays water is also needed for dust suppression, but in Antiquity, it would have been necessary to wet ropes and cool down pulleys. Water itself may have had only a minute impact on the overall budget, but it was indispensable all the same. Moreover, because of its physical properties, water cannot be simply piled or stacked in a corner: the logistics of sourcing and storing water required preliminary planning and on-site solutions that could range from wells to cisterns or from aqueducts to water carriers. Archaeologically, these solutions may have been ephemeral or invisible, but they may also offer possible explanations for short-lived features found under or around other elements. Snyder (2020) has calculated that each cubic metre of slaked lime required 1 person-hour of water collecting, and another 1 person-hour was needed for every cubic metre of freshly mixed mortar. These calculations are a useful starting point to think about how a site was managed, but without knowing how much water was needed on site, it is impossible to begin to address the issues of sourcing and storing or to get an overall picture of the workings of a construction site.

In this paper, I want to put forward a tentative calculation method to estimate how much water was necessary for lime-based constructions in the Roman period, although the same model can be applied to other historical periods where lime mortar (with and without pozzolanic additions) was used. The initial premise is that we know from written sources, archaeological analyses, and ethnographic comparanda how lime-based mortars were mixed in the Roman period. These mixes required variable quantities of water at different stages, and these quantities can be calculated (when they are fixed) or estimated (when they are variable). Since it is possible to calculate the minimal stoichiometric and volumetric relationships between the original components and the final volume of mortar, it is possible to calculate the minimum necessary water input. It is from these calculations that we can begin to think about the logistics of securing water for construction sites. These theoretical calculations will not give absolute or precise results; these are approximations that will only provide a lower estimate for the ratio between the volume of concrete masonry and the original water input. The fact that these calculations are based on broad theoretical models rather than precise measurements specific for each individual monument does not undermine their value or usefulness, and they can be the stepping stone for further research (cf. Martínez Jiménez 2020).

With this in mind, the paper is structured in four parts: first, a brief definition of the terms used in the archaeological literature and the rest of the paper; second, a schematic overview of the mortar mixing process, noting the various moments in which water is added to the mix; third, the main body of calculations, resulting in a series of broad ranges for water inputs at the different stages of the chaîne opératoire; and fourth, the application of these calculations the construction of the sixth-century Reccopolis aqueduct to illustrate the practical validity of this method.

Before the invention of quick cements of the Portland type, most construction used lime mortars as a binding agent. In Antiquity, these were overwhelmingly aerial limes; that is, non-hydraulic limes that dried in contact with air. In this sense, a lime mortar is a cementitious substance resulting from the mixing of lime with an inert aggregate (usually sand) that adds volume and stiffness while preventing shrinkage. This mix solidifies with time and binds together whatever building materials (bricks, rubble, stone) it has been applied to.

The scientific explanation for this process is the sequence of chemical reactions known as the lime cycle (Hobbs and Siddall 2010; Wright 2005: 174-176). Limestone, which is mostly calcium carbonate (CaCO₃), is calcined in large furnaces, expelling its carbon dioxide component while oxidising the calcium, resulting in calcium oxide (CaO). Calcium oxide is also known as quicklime.

Quicklime reacts violently with water in a very exothermic reaction (15,500 cal/mole). This process (known as ‘slaking’) turns quicklime into slaked lime or calcium hydroxide.

Calcium hydroxide is a powder, but the slaking process is done with extra water so that the result is creamy paste or putty. In aerobic circumstances, slaked lime slowly dries and exchanges the hydrogen and oxygen it contains for carbon from the air, liberating water as vapour and carbonising back to calcium carbonate.

Lime (quicklime/slacked lime) is, therefore, the reactive component in these mixes. However, (slaked) lime on its own forms a yoghurt-like paste; for it to be usable in construction it needs sand to give it consistency (Oleson 2010). Therefore, even if these mixes are mostly composed of sand, they are still referred to in the literature as ‘lime mortars’. This is also a way of differentiating them from other building binders identified in the archaeological record, like clay mortars. This type of binder that combines lime with inert, sandy aggregates has been in use for centuries, before and after the Romans (Adam 1994 [1989]: 65ff; Bonen et al. 1994; Wright 2005: 146-189), whereas the Romans developed new lime-based mortars by adding reactive aggregates, revolutionising lime construction.

In the third century BC, builders in central and southern Italy noticed how the addition of some local sands to their mortar mixes improved the final result. The volcanic nature of these sands (pozzolana) chemically modified the lime cycle, giving the mix hydraulic properties and allowing it to solidify underwater (Adam 1994 [1989]). Later on, it was noticed how the addition of crushed pottery sherds (chamotte) had similar results.¹ This new Roman type of pozzolanic mix was, during the following two/three centuries, used alongside newly-developed building techniques, which included the use of standardised, pre-cut conical/pyramidal stones, a large-scale brick production, the use of coffering to frame and shape mortar bound with rubble (caementa, from where we get ‘cement’ and its cognates), and even the use of different types of rubble with different weights to improve vaulting techniques (Mogetta 2013; Sear 2008 [1982]: 124-132). This was the beginning of an architectural revolution which continued to develop in the early Empire with further experimentation with vaulting and doming, and new materials (Lancaster 2005; Van Oyen 2017; Lechtman and Hobbs 1987).

Lime-based mortars and associated Roman techniques are found across the empire. In Roman foundations and pre-existing cities alike, opus caementicium was used as the structural core of large buildings that were then faced with brickwork and stonework (Fig. 1). Many of these facings, which were also bound with mortar, are known in the archaeological literature through Latin names. Some appear in Vitruvius (like opus incertum, reticulatum) while others (like opus signinum, latericium, and vittatum) are inaccurate archaeological shorthand attributions (Gros 2013; Sear 2008 [1982]; Rubio Bardon 2011). Mosaic floors were embedded into a bed of opus signinum, the same material that was used to line cisterns, vats, pools, and the conduits of aqueducts. Most studies of Roman concrete―whether replications, compositional analyses, or mechanic and structural tests― have focused specifically on those with pozzolanic additions (Brune 2010; Goldsworthy and Zhu 2009; Oleson et al. 2004; Sánchez Moral et al. 2005; UNILAD 2021), but pozzolanic mortars did not replace lime-and-sand mortars. Only in a few special cases were volcanic sands exported beyond Italy, as in the harbour of Caesarea Maritima (Hohfelder et al. 2007); more often, local materials were sourced, so that even if across the empire Roman concrete techniques were employed, they did not all use pozzolanic mortars as they did in Rome and Italy (Dix 1982; Uğurlu Sağın et al. 2021).

As outlined above, building with lime mortar was a process that consumed water at different stages of the mortar-making process. We must remember that not two mortars were the same and that there was not one single way of mixing it, as the purpose of the mortar dictated the proportions, types, quantities, and qualities of the components (of limes and aggregates alike). It was not the same to prepare a mix for building foundations as it was to make a mix for a final plastering operation, and this had an impact on the water needs on construction sites. Despite this, the mortar mixing and building sequence followed this overall chaîne opératoire.

Nowadays, as in Antiquity, the first stage in the preparation of any lime-based mortar is to slake the quicklime. Slaking is only the process of mixing quicklime (which comes in rocks and nuggets out of the kiln, although it can be crushed into a powder) with water, which chemically results in portlandite powder (calcium hydroxide); as mentioned before, this was usually combined in one single process with the making of a paste, by adding extra water in a volumetric ratio of 1:3 quicklime to water. Dix (1982: 338) describes how slaking was done by filling a pit with a foot of water and then adding quicklime until half full with lime, which was then mixed with a hoe while adding extra water. Such pits, lined with timber, have been found in Verulamium and Wroxeter (Morgan 1992: 9, 16), and may relate to the ones described by Vitruvius in which the lime would slake over time (calx in lacu macerata: Vitr. 7.2.2). The resulting slaked lime paste needed to mature so that any remaining particles of quicklime would be slaked. The longer the paste matured, the fewer of these unslaked particles there would be, and the better the final result was. This could take up to three years, resulting in a very sought-after vintage lime (Pliny, NH 36.55).

Depending on the nature of the construction, the lime could be slaked in pits or vats on site, maturing overnight (or longer) if necessary. Quicklime is much lighter than limestone, making it easy to transport in bulk, a solution perhaps more suited for larger projects, but the rocks need to be transported in such a way that they do not slake in contact with the humidity in the air. From Pompeii, however, we know that the slaking could have been done directly at the lime burners or at a builder’s yard because pre-slaked lime paste was transported to the site in amphorae, as we see in the Casa della Soffita (V.3.4). This would be an ideal solution for smaller works (Adam 1994 [1989]: 164-165; Berry 1997).

The lime paste or putty is then mixed with the aggregates, which as mentioned, adds volume and consistency to the mortar. The fact that the aggregates are mixed with pre-slaked lime is specifically mentioned in sources such as Vitruvius (2.5.1: cum ea erit extincta, tunc materia ita misceatur), Cetius Faventius (4: calcis proxime extinctae duae partes ad quinque harenae mortario misceatur), and the Wall law of Puteoli (the ‘lex parieti faciendo’, CIL X.1781: in terra calcis restinctae partem quartam indito). In this process, the dry aggregates (sand, pozzolana, chamotte) would have absorbed some humidity from the paste because of their intrinsic physical properties. To correct the rheology of the mix (to make it less stiff and more workable and malleable), a small amount of water was added, but never enough to make it runny like modern concrete. More water never makes more mortar (Gárate 2002: 99).

At this point, it should be noted that there was an alternative way of mixing the lime and the aggregates, called ‘hot mortar’ or ‘dry slaking’ mixing. In this type of mixing, a heap of sand is formed, and the centre is hollowed out. That cavity is filled with quicklime and water is added to it. The lime slakes in the ‘volcano’, the crater of which is then covered in wet sand to seal the slaking chamber. After the lime has slaked, it is then all mixed to create a ready-to-use mortar which would have been good enough for structural binding and foundations. This is a method favoured by current ‘traditional’ masons (Lynch 2017; Adam 1994 [1989]: 164) and it was also typical in late Antiquity (Snyder 2020), because the seal retains the heat generated during the slaking process and this accelerates and homogenises the slaking. A mosaic in the Bardo museum, depicting two workmen making mortar by pouring water from an amphora onto a pile of sandy material, illustrates this form of volcano slaking in a Roman context. Similarly, at the Casa del Sacello Iliaco (I.6.4) in Pompeii, lumps of quicklime were stacked in preparation for on-site slaking, most likely in a ‘volcano’ (Lancaster 2005: fig. 41).

After mixing and before application, the mortar would be left to settle (Gárate 2002: 166; Cazalla Vázquez 2002: 38). It is possible to let it unused for a while (at least in the case of non-pozzolanic mortars). Even overnight, these mortars may stiffen without properly or fully carbonising, but, if so, adding water makes the mix workable again.

Once the mortar was mixed (regardless of the method), it had to be applied to the elements that would give solidity and volume to the construction, because the mortar itself cannot be used for that purpose. This could be done by either mixing it with rubble/caementa when making structural cores of concrete (opus caementicium) or it could be applied with a trowel to the building materials (bricks, tiles, ashlar blocks). Even if these large solids fulfil different structural purposes, from a water-consumption perspective these needed to be sufficiently humid, lest they absorbed too much moisture from the mortar, causing it to shrink, crack, and dry unevenly. Therefore, building blocks or rubble would be soaked to the point of saturation and then left to dry to a point the builders and masons considered sufficient.

Even after the structure has been built, there was another phase of water consumption. As the mortar dries and carbonises, it is necessary to keep sprinkling it with water to ensure a homogenous process to prevent shrinkage and cracking. This process of soaking and hydrating is especially necessary when ambient temperatures reach 30°C (Gárate 2002: 99), an eventuality that concerns most of the Roman world for half of the year. Together with sprinkling walls and structures, it was possible to keep constructions from heating too much and drying too quickly by covering them with canvases or tarpaulins that would protect them from direct sunlight.

A last, optional step that was not always done, and that could be seen as a different process altogether, would be limewashing. Limewashing (or whitewashing) is a way of protecting walls by applying a very diluted solution of lime, usually applied on external walls to add extra protection from the elements. It is long-lasting and it can be applied either unchanged (white) or coloured by the addition of pigments. While limewashing is not a necessary step in historical constructions, it was a usual addition in Roman building (Hughes et al. 2007).

The opening premise that water is necessary on construction sites has thus been demonstrated. The sequence listed above has focused only on those processes that relate to mortar production, ignoring any other process already mentioned or not (cleaning/washing, dampening ropes and pulleys, on-site metalworking, etc.) so while the mortar was the main consumer of water on-site, it was not the only one.

The chaîne opératoire described above gives us the main steps from which we can begin to calculate the relationships between archaeological mortar and water input. The sequence can be summarised as follows:

Where A is the total water input, S is the water consumed in the slaking process, R is the water used in the mixing of the mortar to hydrate the mix and correct the rheology, W is the potential water addition to re-work a pre-mixed mortar, C is the water used to soak the caementa and other structural materials, and D is the sprinkle used to correct and control the drying and carbonising process. P corresponds to the water used to whitewash a surface. In this study, all of these quantities of water will be considered to be volumetric. However, I will only consider in detail three of these different water inputs, the three that can be practicably calculated and which are also those that would have been proportionately the largest: S, R, and C, i.e. the water for the slaking, the hydration of the mortar mix, and the soaking of the rubble and structural materials.

Since these water inputs represent different stages of the chaîne opératoire, the relationships that can be calculated are those between the water input and intermediate results. S is calculated from the input of quicklime (V _ql ) and the final slaked lime paste (V _slp ). R is calculated based on the relationships between the volume of paste (V _slp ), the volume of aggregates (V _agg ), and the volume of the pre-hydration, ‘initial’ mortar mix (V _im ). All these can, in any case, be related to the final volume of fresh mortar (M _f ) that was used in a given construction. The volume of fresh mortar is directly related to that of dry, archaeological mortar (M _d ). C is independent of the volume of mortar, as it depends entirely on the volume of building materials (V _bm ) and the volume of the structure (V _s ), which can be calculated from the archaeological remains. The wall area to be whitewashed (a _ww ) is unrelated to this process, but its surface is an amount calculated from the reconstructed volume of the structure.

Because the real or reconstructed archaeological remains (V _s ) are the point of reference, it is necessary to calculate backwards in the sequence of events and the chaîne opératoire to obtain a volumetric relationship between it and the earlier intermediate products (Tab. 1).

The exact amount of water that was required for any given construction project in Antiquity (what was termed A at the beginning of this paper) will never be known. The different processes involved in mortar mixing, lime slaking, and concrete construction are too complex and include too many circumstantial variables which cannot be gauged with accuracy. But accurately calculating that was never the aim of this paper; the objective of this article is to provide a guide to roughly estimate water input at four of the various stages in the chaîne opératoire: the water used for slaking (S), the water used to mix the mortar (R), the water used to soak the building blocks (C) and the water needed to limewash a wall (P). Similarly, this article focuses on Roman construction, using archaeological and literal examples from the period, but the building principles are similar and comparable throughout the pre-Modern era.

The results of these calculations depend on the variables defined by the building materials and the mixing proportions. In the end, they offer us a range and an order of magnitude, which should be sufficient to think more carefully about the water logistics in large construction projects. Works that require only a few cubic metres of water might have required barrels for their storage, and these could have been filled from fountains, wells, or cisterns, but moving into the tens (or hundreds or thousands) of cubic meters required different solutions.

Further studies applicable to individual mortars will, naturally, nuance these calculations. Microscopic analyses revealing high proportions of unslaked nodules in the mortars may suggest the use of hot mortars (‘volcano’ mixes, where the hydration of the quicklime was not thorough) as opposed to the use of pre-slaked putties. The water needs on site of these two different techniques would vary, and so would the necessary logistics. In drier areas, like North Africa, meeting the hydric requirements of mortar construction might have been an impediment, but this opens new possibilities to explore ― possibilities that might have been also necessary elsewhere in the ancient Mediterranean, like the role of aqueducts in the expansion of concrete and mortar constructions or the seasonal nature of construction processes.