The Physics of Freezing at the Iranian Yakhchal

Yakhchals are found across Iran in areas where the climate enables the freezing of ice on-site in winter, or where ice and snow could be brought from nearby mountainous areas. The ice is reported to have been used in primarily in the hot summer for the production and preservation of food and drinks, and perhaps occasionally for cooling buildings. 

The drawing below illustrates a particular large, dome type Yakhchal in Meybod, in the province of Yazd on the Iranian plateau, which is reported to be around 400 years old.

The complex comprises of a shallow ice making pond which is filled each night from a qanat, a fresh water canal. A shading wall shields the pond from the heat of low-angle winter sun.

Ice is harvested from the pond and transferred into the Yakhchal ice store.  These types of Yakhchals are reported to have been constructed from locally sourced adobe containing mud and binding materials such as animal hair and egg white.

The height of the dome at Meybod is around 15m tall and pit about 5m deep. The dome walls are estimated to be around 2m thick at the base and 0.5m thick at the top where there is a vent hole.  

An aspect perhaps not often visualised is that (at least in summer) the domes may not have displayed the smooth, red-brown finish seen in all Yakhchal photographs. Some sources report that they were covered in a thick layer of straw thatch to insulate them from the heat of the sun, in which case they would have resembled giant hairy haystacks. 

The climate of Yazd is very dry, with hot summers and cool winters. The average annual precipitation is less than 100mm/yr.

The following picture illustrates typical hourly air temperature for Yazd for a whole year. It shows that in winter the temperature occasionally drops to -5°C and in summer the temperature regularly exceeds 30°C and can reach 40°C.  

The rate of ice formation in a shallow pond is determined by several different factors including; the climate variables, the ground temperature, the water starting temperature, and how the process was managed.

We developed our own heat transfer physics model to study the ice making process. The key features of our approach, and what sets our work apart from that of other authors, are as follows:

- The pond is modelled as a series of very thin layers that can either be water, slush or ice. The model can represent the process where ice begins to form on the upper pond surface and gradually grows toward the interior.
- The model is transient (time varying) and uses site specific Yazd hourly climate data to drive the simulation. Parameters such as external air temperature, humidity and cloud cover are read from the climate data file and used in the various heat transfer calculations; the most important of which is the calculation of the effective sky temperature.

The approach described above allows us to make a numerical prediction of the quantity of ice that could be made over the course of the whole ice making season.

We used our model to simulate the ice making process over several years. As expected, the results showed that the dominant cooling mechanism was heat loss by radiation from the pond to the night sky, which in the case of Yazd is particularly cold. Also important was significant heat loss by evaporation, which is driven by the low humidity of the Yazd winter climate.

The ice making pond at Meybod is roughly 400m² in plan area. For this size, the annual simulation results showed that the total amount of ice produced in one winter season was 50m³, which is a volume equivalent to 3 million modern-day ice cubes.  50m³ is roughly 20% of the size of the pit at the base of the ice storage dome. In this respect the predicted ice quantity seems a little on the low side. There are several possible explanations for this including the following:

- Our model could be rubbish (however we don’t think this is)
- The available Yazd climate data represents an “average” year based on data from 1960 to 2004. It is therefore probable that many winters would be colder in which case more ice would be made
- The climate may well have been significantly cooler in the past
- Additional ice was brought into the store from the wider environment

In addition to simulating winter ice production, our pond energy exchange model can be used for other design purposes. For example, the simulation and analysis of summertime roof pond cooling such as the experimental types constructed by Hay and Yellott, examples of which can be seen in this misfits’ architecture blog post.

It is reported that Yakhchal ice production facilities were run both as community operations and as businesses selling ice.

In either case the Yakhchal manager would be very keen to retain as much ice as possible, that is stop it from melting. The rate of ice melt within the ice store is determined by several factors including; the climate variables including temperature and sunlight intensity; the thermal properties of the Yakhchal structure and how the building was managed.   

We used a 3D dynamic thermal simulation technique to simulate the heat transfers within a Yakhchal to make a prediction of the amount of ice that would melt between the winter and the end of summer. A diagram of the model is shown below.

The modelling results show that for a scenario with 1m thick adobe walls plus 1m thick thatch insulation the Yackchal’s ancient construction technique has a heat transmission (U) value of 0.1 and a thermal mass (Cm) value of 150*. This is superficially equivalent to a modern construction of 100mm of concrete plus 400mm of polystyrene insulation. Heat transmission is how much heat passes through the walls, for example how much of the sun’s heat absorbed by the dome's outer surface is conducted into the dome interior. Thermal mass is a measure of how much heat (or cold) the walls can store and how quickly a material heats up or cools down.

*The units of U value are W/(m².K). The units of Cm value are kJ/(m².K).

By superficially equivalent, we mean that the two construction types have the same numeric heat transmission and thermal mass values. These numbers are used by architects and engineers to select materials with the aim of producing a particular building performance. However, the simulation results have shown that the two construction techniques do not perform equivalently.

The ancient adobe and straw method produces cooler internal dome temperatures (by about 3°C) than the dome constructed from modern materials with superficially equivalent thermal properties. This is principally because the thermal mass material property, (Cm) used to characterise heat (or cold) storage doesn’t tell the whole story and the approximation gets worse as the walls being studied get thicker.**

For the very (>1m) thick walls of a Yakhchal, a full dynamic thermal simulation of the type we have carried out is required to reveal and demonstrate their true, thermal mass cooling potential. The reason is that the effect is seasonal; for dense, thick walls it takes about one year for heat to propagate 1m. This finding has relevant implications for the design of modern architecture, in particular passive cooling.

Thick wall construction techniques such as rammed earth, stone or structural brick will superficially appear to have limited cooling properties, and may therefore be hastily discarded as a design option. However, dynamic thermal simulation will reveal them to have higher cooling capacities than initially expected, due to their ability to store cold from winter to summer. 

**For those of you in the know, the CIBSE admittance (Y) value fares even worse as an indicator of comparative thermal mass in this case due to the fact that the admittance value by definition is concerned with a 24hr heat cycle whereas the 1m+ thick walls of the Yakhchal also respond to the long term 365 day heat cycle.

•  

   

  In addition to looking at thermal performance of the Yakhchal construction materials, we used our model to simulate the heat transfer to and from the ice pack for a whole year, based on the assumption that the ice store pit begins full in January.  The modelling results showed that for the situation with 0.5m thick adobe walls plus 0.5m thick thatch approximately 20% of the ice is predicted to be lost due to melting by September.  Our work considered and modelled a range of design variations looking at aspects such as enhanced solar shading and enhanced cooling by cold night air. If we were appointed to travel back in time to work on Yakhchal optimisation we would advise the Yakhchal management committee that investing in enhanced store insulation would be the best way to inhibit melting. If 1m of thatch is used around the store the melt losses are predicted to reduce by 40% to about 13% of the initial ice pack volume.  If 0.5m thickness of modern polyurethane foam insulation could also be sent through the time machine and installed then the melt losses could be reduced to 6% of the initial ice pack volume.

  Having studied the potential functions and benefits of the dome itself on ice preservation we found that the dome serves as an excellent solar shade and (as previously stated) the immense thermal mass has a cooling benefit that slows down the rate of ice melt. With regard to the dome vents we found that while enhanced cooling by night ventilation could be of benefit, it can’t be realised with the sizes of vents that are included in the structure. This leads us to the hypothesis that the main purpose of the hole at the top of the dome wasn’t as a vent; but actually a small rooflight to let in daylight to illuminate the otherwise dark and dangerously slippy interior to facilitate the safe extraction of ice from the pit during the summer.

• Further Reading