Travertine terracing: patterns and mechanisms - Tufas and Speleothems: Unravelling the Microbial and Physical Controls - page 346

Fig. 8. Computer simulation of calcite precipitation under shallow water flow over a protuberance (flow from left to
right). Top: Map of CO 2 concentration. The upstream concentration gradient is compressed over the obstruction.
Downstream there is thorough mixing until a gradient slowly reappears at the far right as CO 2 is released by
precipitation. Bottom: Precipitation rate, showing enhanced precipitation on the top of the obstruction and also
downstream. Small peaks are due to the computational grid, causing small corners of the obstruction to stick out from
the boundary layer. The model includes hydrodynamics, advection, diffusion, degassing, solute carbonate chemistry and
precipitation kinetics. Details of simulation in a somewhat different geometry are given by Hammer et al. (2008).

Pentecost (2005) discusses terrace wavelength in
the downstream direction and the relationships
between slope, discharge, terrace wavelength,
depth and height. In general, pools are shorter on
steep slopes, but height can be larger. Inter-dam dis-
tance (IDD) seems to increase with larger discharge.
The IDD sometimes displays a characteristic wave-
length. In the study by Viles & Pentecost (1999)
IDD was found to be random, but in this case ter-
races were possibly initiated by large woody
debris. For microterracettes, the ratio between IDD
and depth is higher on gentler slopes.
Hammer et al. (2007) studied terracette topo-
graphy in downstream cross sections, both using
their simulation results and field observations in
Rapolano Terme, Italy. In spite of the problems of
such cross sections cutting the pools at random trans-
versal positions, and not always at their widest
points, a regular spacing (characteristic wavelength)
was observed.
Microterracette pool width distributions were
studied in great detail by Veysey & Goldenfeld
(2008), again using both simulation and field data.
They defined their pool width as the largest width

in a direction normal to the maximum chord. Com-
paring with a statistical null model of Brownian-
motion (random walk) rim shape, they found good
accordance except for small terrace widths.
Although their distribution is unimodal, it has large
variance, and is somewhat difficult to reconcile with
the more regular spacing observed at larger scales
by Hammer et al. (2007). Veysey & Goldenfeld
(2008) interpreted their results as indicating no
interaction between rims for large terrace widths,
but an attraction effect for small widths.
Veysey & Goldenfeld (2008) also present an
interesting analysis of pool areas. Using a simple
null model of initially small-sized pools merging
randomly, they show that the expected steady-state
pool area distribution is inverse-square, in accord-
ance with their field observations.

General aspects of travertine terrace
pattern formation

Pattern formation in the travertine terrace system
results from the interaction between two opposing