• from Passchier et al. (2021)

The sampled carbonate stratigraphy of JWP128 consists of three sequences labelled JJT1–3, separated by layers of grey plaster (Figs. 2b and 3). Thin sections show a rhythmic sequence of similar layers of variable thickness (Figs. 3–5), each with an alternation of micrite and elongate columnar sparite crystals. Micrite and microsparite crystals are 2–10 μm in diameter, while sparite crystals can be up to 2 mm wide and 7 mm long. The alternation of micrite/microsparite and sparite, in combination with stable isotope cycles, allows counting of annual layers (L1–59; Figs. 3 and 4).

The start of each stratigraphic segment JJT1–JJT3 consists of a continuous band of micrite, deposited directly on the plaster (Figs. 3 and 4). Columnar sparite crystals grow from these micrite horizons, commonly in fans due to growth competition (Passchier et al., 2016b; Fig. 5). Micrite commonly forms a band of serrated shape between sparite layers (Fig. 5). Sparite crystals contain bands of brown coloration (hereafter “brown bands”), which show the shape of the calcite growth surface. At high magnification, they are composed of numerous thinner laminae, 6–10 μm apart (Figs. 5 and 6b–d). Small calcite crystals locally nucleate in these brown bands. Each macroscopic layer can be subdivided from bottom to top into microscopic sublayers of variable width as follows (Fig. 5): (a) micrite with initial sparite and upward tapering of micrite domains between sparite crystals; (β) sparite crystals with brown bands; (γ) pure sparite; (δ) initial and widening micrite domains with occasional brown bands in contemporaneous sparite; (ε) pure micrite, not present in all layers. The base of sublayer (a) was chosen as the boundary between layers (Figs. 3 and 4).

In hand specimen, sublayers (a) and (ε) are yellow, (β) is brown, and (γ–δ) grey (Figs. 3 and 5b inset). In (δ), micrite may be continuous, topping sparite crystal faces completely, e.g. in L25 (Fig. 5a). However, micrite mostly occurs as discontinuous horizons, interrupted by individual sparite crystals that extend over several layers (Fig. 5). The geometry of sparite varies from dense columnar and fan-shaped sparite to isolated crystals surrounded by micrite.

• from Passchier et al. (2021)

In the rock-cut channel of JWP128, a basic plaster layer, P1, was installed on an eroded limestone surface and on thin remains of older carbonate deposits, which were either eroded by flow or were manually removed as part of cleaning processes (Fig. 3; Supplementary material S3). This implies that the aqueduct had been in use for some time before the sequence analysed here was deposited. In sequence JJT1, layers L1–5 are thin (mean thickness 2.5 mm) with a high percentage of micrite in continuous bands (Figs. 3 and 4). The uppermost layer L6 is dominated by sparite. JJT1 was eroded in the centre of the channel by a scour, which was filled with plaster P2 (Fig. 3). JJT2 is the most prominent sequence with a regular alternation of sparite and micrite in 42 layers. The initial layers are thin and similar to older deposits, but layers L9–38 are thick and regular in thickness and facies distribution. Layer L25 is distinguishable by its unusual red coloration in hand specimen (Fig. 3) and a thick, continuous flat layer of micrite at its top (ε in Fig. 5a). The uppermost layers of JJT2 are relatively thin and micritic (Fig. 3). Layer L48 contains deformation twins in the calcite, accompanied by broken crystals and fractures filled with micrite (Fig. 6a). This damaged zone is overlain by undeformed micrite, indicating that the damage occurred during deposition of carbonate (Fig. 6a), shortly before application of plaster layer P3 (Fig. 3). JJT2 also shows “orientation bands” parallel to the growth direction where sparite crystals have a deviant orientation, and neighbouring fans interact in a complex manner (Fig. 4: JJT2–3, JJT2–4). This was apparently induced by ripple-like irregularities in the growth surface of the layers, normal to the flow direction in the aqueduct. Similar ripple-like growth structures have been observed in other aqueducts (Motta et al., 2017; Sfirrelil Hindi et al., submitted). In JJT3, layering is relatively wide with extensive micrite, while only single layers (e.g. L50) are dominantly sparitic (Figs. 3 and 4). The sequence ends in layers of micritic carbonate (L55–59). Radiocarbon dating of charcoal from the three plaster layers P1–P3 yielded calibrated radiocarbon dates (68% probability) ranging from the 1st century BC/AD to the early 3rd century AD (Fig. 3; Boyer, 2019; Supplementary material S3).

• from Passchier et al. (2021)

Carbonate from the Jerash aqueduct, with red colour banding of alternating sparite and micrite, is similar to that of many other Roman aqueducts in the Eastern Mediterranean, e.g. in Turkey and Israel. The observed carbonate is dense due to abundant sparite, which is common in covered aqueducts, particularly those with fast, supercritical flow (Şirmeli̇hi̇ndi̇ et al., 2013a, b; Passchier, 2016a). The aqueduct at the sampling sites has a steep slope between 19 and 21% giving a Froude number of 4.2–5.3, depending on water depth (Supplementary material S2). Other aqueducts with a gentler slope and slow, subcritical flow conditions contain dominantly micritic carbonate (Passchier and Şirmeli̇hi̇ndi̇, 2019). Aqueducts like JW01, with relatively thick annual layering and abundant sparite, are prime targets for paleoenvironmental investigation, since small-scale changes in carbonate deposition can reflect variations in palaeo-temperature, water composition and flow rate (Supplementary material S2). Fig. 9c–g shows a schematised summary of the microstructural, stable isotope and trace-element features described above.

• from Passchier et al. (2021)

Carbonate deposits in the Jerash aqueduct show evidence of restructuring and repairs, some probably in response to floods and earthquakes. Layers L1–8 are relatively thin and micritic, indicating low water levels and low flow rate with clay impurities (Figs. 3 and 8). An exception is L6, which has unusually low δ13C values (Fig. 7) and is followed by erosion of the channel and repairs including a new layer of plaster (P2) (Fig. 3, Supplementary material S3). Most likely, this was due to a flood event: since the slope of the aqueduct is 19–21% at the sampling site, erosion of the channel floor was a permanent risk at high flow rate. In contrast to JJT1, sequence JJT2 has generally thicker and more sparitic layers after L8 and low trace-element concentrations in the measured section L14–25, suggesting that the aqueduct carried more water with less suspended clay. This can be attributed to addition of a new spring or springs to create a higher flow rate starting with layer L9 (Fig. 3). JJT2 records the period with the highest mean flow rate of the aqueduct, even though in some years (L6, 19, 20, 25), the rate fell to alarmingly low levels in the dry season or water flow even stopped for some time (Figs. 7 and 8). The continuous micrite layer at the top of L25 may indicate such a low-flow period in response to a drought. Layers L37–48 are more micritic and thinner, possibly due to a damaged channel or a dry period (Fig. 3).

Layer L48 at the top of JJT2 has deformation twins in calcite (Figs. 3 and 6a) and the aqueduct was resealed with a plaster layer (P3) following L48. The twin orientations as observed using a U-stage indicate that they were formed by vertical pressure applied on the crystals. Possibly, this damage records the impact of a falling stone (Parlangeau et al., 2019), e.g. due to an earthquake.

In JJT3, the first two years (L49 and L50) represent a high flow rate as in JJT2. The following sequence has layers of micritic carbonate, with abundant clay impurities, indicating a significant change that persisted for years until the end of deposition. The presence of clay could be due to erosion and/or changing vegetation cover, but the sudden change, following a wide sparite-dominated layer (L50), rather suggests a sudden event. Possibly, an earthquake damaged the channel again, leading to low flow rate, without repair in subsequent years until the abandonment of the aqueduct. Observations at JWP128 suggest that the aqueduct went out of use sometime after the late 2nd to mid-3rd century AD. The carbonate deposits within the channel at location JWP128 are fractured, and this may reflect another seismic event.