Based on the last two blogs, the evidence seems strong that the puddle was being fed by liquid water coming from underground springs. A colleague of mine, Kristina Katsaros, pointed out that even a spring-fed puddle might have frozen under such conditions. Maybe the water would freeze on the top, for example.

Kristina has studied Arctic ice and the effects of salt in the sea water on freezing. So she suggested that another factor be considered: salt on the roads. This would be a modification of the spring-water hypothesis, to allow for the effect of salt.

In the United States, road crews often apply sodium chloride or magnesium chloride to roads because these two compounds can melt snow or ice on the roads, making them safer to drive on. This happens because these two compounds dissolve in water. Salty water has a lower freezing temperature than pure water.

On my trip out from my brother and sister-in-law’s house to take a picture of the puddle, the road was solid white. But coming back, I noticed something interesting: The ice and snow was starting to melt in little spots.

Figure 6. Ice and snow melting in spots on road.

When I looked more closely, I discovered a salt crystal at the center of each spot, surrounded by melting snow and ice. You can see a few of the crystals and their impact on the surrounding ice in Figure 7, below.

Figure 7. Salt crystal at the center of widening circles of melting ice and snow. The salt crystal is dissolving in the water, lowering its melting point.

If they spread salt crystals on the road the morning I took most of the puddle pictures (the fire hydrant picture was taken later in the day), I am guessing that salt was used on the road earlier in the winter as well.

So the puddle that I think was spring-fed may have been able to remain liquid thanks to salt on the road. And, since the puddle was in a low part of the road, salt may have washed down to this part of the road during the rains of the previous week.

So – it is likely that the puddle formed from underground water seeping into the road (built with cracks to allow the concrete to expand and contract), and the puddle stayed liquid because of a supply of warmer water, and a little salt. There are of course many more things I could do to confirm this hypothesis, but it seems reasonable given the facts I have available.

Suppose the puddle had been there from the previous week’s rains, and I simply missed it. Then the salt in the road might be sufficient to keep the puddle from freezing entirely.

This is the second in a series about an unusual winter puddle in Columbia, Missouri.

Recall from last time that I mentioned that the water feeding the puddle would be coming to the surface from under the ground – either a broken pipe or water flowing horizontally through the soil

In my recent 27 February blog about the air temperature and the surface temperature , I wrote about the “energy budget” of air about 1.5 meters above the surface (to explain why the maximum temperature was in the late afternoon), and I also wrote about the surface temperature, which reaches a maximum in the early afternoon.

To understand why the water feeding the puddle (and the surrounding soil) wasn’t frozen, we need to learn something about how temperature varies with depth beneath the surface.

The heating and cooling below ground is mostly by conduction. During the winter, the surface vegetation protects the soil from cooling, and upper soil layers protect the soil layers farther down.

For example, Figure 3 compares the air temperature and the temperature just 2 cm below the surface on a corn/soybean farm near Champaign, Illinois, USA (near Chicago). While the air temperature fluctuates quite a bit, the temperature at 2 centimeters below the surface changes much less. Particularly interesting is the cold weather between about 25 February and 5 March, when the soil temperature stayed warm in spite of the cold night-time temperatures. Figure 4 focuses more closely on that time period.

Figures 3 For February and March 2002, air temperature and soil temperature (Ts) at 2 cm below the surface, a corn/soybean farm near Champaign, Illinois (Latitude 40.00621, Longitude -88.29041). Day 30 = 30 January; Day 60 – 1 March, Day 90 – 31 March). Data available on the Web at http://cdiac.ornl.gov/ftp/ameriflux/data/Level1/Sites_ByName/Bondville/FLUX-2002/.

Figure 4. For the 25 Feburary-6 March time period on Figure 5, Air temperature, surface temperature, and soil temperatures down to 64 cm below the ground.

Notice that the soil temperature gets warmer as you go down, illustrating the “insulation” effect of the higher layers of soil. From Figure 5, we see a similar pattern at Smileyberg, Kansas. On average, the soil temperature at Smileyberg was 1.9 degrees Celsius warmer than the air temperature in January, and 1.1 degrees warmer than the air temperature in February. Notice how the ground stayed warm between Days 24 and 32 (24 January and 1 February), in spite of the cold temperatures. This is just like the behavior we saw at the Bondville site.

Figure 5. For grassland site near Smileyberg in Southeast Kansas, the Air temperature (about 2 m) and soil temperature (average 0-5 cm). Data from Argonne National Laboratory, courtesy R.E. Coulter, Argonne National Laboratory.

Thus it is quite believable that there could be liquid water close to the surface, particularly since the air was much warmer the week before I got to Missouri. (Since I saw no frozen water on the surface uphill of the road, the water could have come up through cracks in the road.)

Mostly written 21 February 2008, from Columbia, Missouri, USA

The temperature for the last few days has been below -5°C (about 20°F). The wind on my daily walks is cold but invigorating.

So, I was surprised yesterday when we drove over a puddle and water splashed on our windshield. It froze instantly. Given the air temperature, this is not surprising. The car thermometer read 17°F (-8°C).

How could there be water in a puddle after three days of subfreezing temperatures?

I decided to investigate on this morning’s walk, and found out a few interesting facts. Figures 1 and 2 show the puddle up close and from a distance. My footprints and the tire tracks in Figure 1 indicate that the puddle was “slushy.” It was easy to make footprints in it. So it is not surprising that passing cars were getting splashed when they drove over the puddle. The second picture shows the puddle with a fire hydrant on the north side of the road (and to the east of the puddle). A line of fire hydrants lies to the north of this road; so I conclude there is a pipe connecting them.

Figure 1. Slushy puddle with footprints (left) and tire track (right). At the time of the picture, the air temperature at this location had been below freezing for several days.

Figure 2. Picture of puddle taken later in the day, after more frozen precipitation (ice pellets) covered the rest of the road with white. Note the red fire hydrant. The puddle is at a low point in the road, and the ground slopes downward toward the puddle from the north.

So, I came away with two hypotheses.

First hypothesis. The puddle is being fed from extremely wet soil. There has been a lot of precipitation around here recently. I knew this because I had been monitoring the weather the week before I got here. I am guessing that the soil is saturated and the water table is quite high.

So, the water could just be flowing downhill, perhaps atop the bedrock, which comes quite close to the surface. Or, the puddle could be fed by an underground spring. There are many springs in this part of the country. The bedrock, close to the surface, is Burlington limestone, which has multiple cracks and caves – paths for the water to follow.

Second hypothesis. The pipe connecting the fire hydrants just happened to have been broken here.

These two hypotheses are based on my impression that the puddle wasn’t there before 20 February. In either case, as we shall see, salt added on the road could have kept the water from freezing once it reached the surface.

I talked to my nephew, who was around when the road was built, and he supported the first hypothesis, because they found a lot of springs when they built the road. The springs are fairly active: my sister-in-law said that swimmers in a lake to the south of the road frequently noticed cool spots, where the cool spring water was feeding the lake. The springs also suggest that the water table is normally high. So, after an unusually wet period, it would be plausible that the water is running down the hillside beneath the surface.

Next time: Why wouldn’t the water running down the hillside beneath the surface be frozen?

What about the total amount of water coming off the roof? Suppose it is raining, so insulation doesn’t make any difference. Again, about twice as much water would flow off a given spot along the eaves for our house. But the total amount of water flowing off one side of the house is determined by the total area of the roof upstream.

Consider the one-meter section of our house in Figure 2. Let’s estimate how much water would flow off a meter length on the east side for a rainfall of 1 centimeter per hour. The roof measures 10 meters from the top to the eaves.

In an hour, the volume of water falling on the roof would be 1 centimeter per hour or 0.01 meters per hour × 1 meter × 10 m, or 0.1 cubic meter of water per hour. Since water weighs 1 gram per cubic centimeter and there are 100 x 100 x 100 x 0.1 cubic centimeters in 0.1 cubic meter, about 100 kilograms of water fall on this one-meter section of the roof per hour. The same amount flows over a meter section of the eaves to the ground in about an hour (assuming the roof drains as fast as it rains!)

But we need to make a minor correction for the fact that the roof is not exactly horizontal (i.e., it’s covering less ground).

If the angle of the roof to the ground is 20 degrees, we need to multiply the 100 kilograms of water per hour per 1 meter by 0.94, making the total rainfall 94 kilograms.

If the roof measures 10 meters along the eaves and top of the roof (Figure 3), the total amount of water flowing off the roof on the east side is 10 times that amount, or 940 kilograms allowing for the angle of the roof.

What about the roof of the imaginary house in Figure 3, which measures 5 meters from the top of the roof to the eaves? Half as much water or 47 kilograms falls on a one-meter slide of this house (Figure 2) each hour, so half as much water will flow off the eaves per meter each hour compared to our house, which measures 10 meters from roof to eaves. We assume the roof’s angle to the ground is the same as our house.

The total amount of water flowing off the roof of the imaginary house each hour would then be:

47 kilograms per 1 meter along the eaves times 20 meters, or 940 kilograms flowing off the east side of the roof each hour. This is of course the same amount of water flowing off our house.

Did you notice that if you just know that the area and angle of the roof of the imaginary house are the same as the roof as our house, means that the total amount of rain falling on both houses is the same, and therefore the same amount of water flowing off the two roofs is the same?

We could call these roofs “roof watersheds” or “roofsheds” because they shed water – in the form of icicles in the first example, or in the form of liquid in the second.

Figure 3. “Roofsheds” for our house and the imaginary house, viewed from above. Both roofs, having the same area (100 square meters) and angle to the ground (20 degrees), will shed the same amount of water on the east side, where the eaves are.

Why are the icicles so long on our house?

On a recent walk just a day or two after our first snow, my husband and I noticed that we had the longest icicles in the neighborhood. Some houses built the same time as our house had icicles, but they were shorter. One new house had almost no icicles.

But what was the most fun, was our own house. The picture below shows our “champion” icicles.

Figure 1. Sketch of icicles on the east side of our house. The windows to the right of the icicles are about 1 meter high. The part to the right is the front part of the house; the part to the left is the back part of the house.

Notice that the icicles only cover the middle third of the side of the house. To the right and to the left, there are no icicles. Were we to walk on the roof, we would probably find the snow melted in the middle third of the roof, but not on the sides.

Why? Our house was built in stages. The front two-thirds were built were built in 1950. There was little insulation in the roof. A few months before I made this sketch, we tore out the old ceiling in the room in the front of the house and found that the insulation from 1950 was in poor condition, just like the insulation in the middle of house. The new insulation was much better. The picture confirms that the new insulation was working. No icicles implies no water from melting snow. This means that little heat was escaping through the roof, so there was little or no snowmelt on the roof.

Similarly, the back part of the house was built in 1979. When that part of the house was built, we made sure we had good thick insulation in the roof. There are no icicles on the new part of the house. Again – the insulation must be working.

Using the data from our house, can we explain why our house had the longest icicles? I’m guessing that the new house in our neighborhood that had almost no icicles had good insulation – just like the newer parts of our house and the room we just insulated. We could that the snow on the roof of the new house was fairly deep – there was little melting.

What about the older houses with shorter icicles? Let’s imagine an older house with about the same insulation as the old parts of our house (Figure 2). If this is true, the snow would melt at about the same rate (I am assuming that the roof was exposed to the same amount of sunlight per unit area). Why then would the icicles be shorter on the other (imaginary) house?

If you believe my assumptions, the answer is that the area of the roof “draining” toward the eaves (where the icicles grow) was smaller. Say the distance from the top to the icicles on our imaginary house is 5 meters, and the distance on our house is 10 meters. As the melted snow moves down from the top of the roof to the eaves, twice as much water reaches a given length along the eaves for the 10-meter roof (ours) compared to the five-meter roof. It follows that the icicles on our house would contain twice as much water and be longer than on the other house. The icicles may be not twice as long, because the icicles we had might be wider as well as longer.

Figure 2. View of a one-meter slice of our house (top) and an imaginary neighborhood house (bottom). More water is available to flow over the eaves for our house. We are looking at the two houses from the north.

So the amount of water in the icicles is determined by the amount of snow upstream of (or straight up the roof from) the eaves.