Sunday, August 16, 2009

Water Crisis in Murray Creek: A Case Study in Causal Inference

Blog entry to 14 August 2009:

Northern California is semi-arid, with San Andreas, the Sierra Nevada foothill town nearest to Murray Creek (where my mom lives), receiving about 30 inches of rain in an average year. My mom’s place in Murray Creek depends on an ancient spring for its water. This spring is far up the opposite side of the valley from the main part of the property, which means that its water supply must be piped down the steep southern side of the valley, over the creek (where it has been washed out by winter floods three times over the past 30 years). After crossing the creek, it then runs under the lower house, up the hill by the upper house, and finally to the big water tank above the upper house.

Parts of this water system are likely to be at least 80 years old. Reportedly, in the first part of the 20th century, the water from this spring was bottled commercially, with the bottling facility located at the small side creek next to the lower house. So this spring has long been known for the quality and consistency of its water, and for all we know, the native peoples who lived in the valley centuries ago also used its water, particularly during the long, dry summers.

From time to time, as the nonnative plants in the garden and elsewhere on the property suffer from the late summer 100-degree heat and low humidity, we will decide to do some catch-up watering, spending some substantial fraction of the stockpile of water in the water tank. Earlier this week, however, the water tank failed to fill after a bout of catch-up watering, then continued to go down further even after my mom shut off everything she could find to shut off. Art the plumber came out on Tuesday to look at the situation, tried a few things, then ran out of time and said he’d come back the next day. By the time we got to Murray Creek on Wednesday afternoon, the water tank was less than half full and my mom was on the verge of panick. All her plans for staying in Murray Creek to work on her writing and to develop the garden seemed to be jeopardy. More importantly, she was afraid that the spring was in the process of giving out, possibly because of climate change, and that her over-watering had somehow set this off.

On Friday morning, the plumber, Art Mills, was able to make it out to work on the ailing water system, having set aside an entire day for it. Art is 46, has a crew-cut and a bit of a pot belly, drives a pick-up full of plumbing supplies; he crushed my hand when we shook hands. He was happy for me to tag along with him and help out to the best of my limited ability, regularly teasing me with his dry humor. (“There’s not a lot of poison oak up here,” he remarked, upon encountering a field of the familiar three-leafed red plants.)

I was fascinated by the process by which he tackled the problem, really a case study in causal inference, with parallels to HSCED. Here was a prime example of the kind of practical reasoning process that I’ve written about as the inspiration for the interpretive case study method. All he knew to start with, was that the tank wasn’t filling, but he did not know why. He’d identified small leaks two days earlier, but they didn’t seem to be enough to account for the problem. “Small leaks” turned out to be a sort of back-up general purpose explanation, which he invoked whenever he didn’t have a better explanation.

In order to fix something (anything, really), it really helps to have an explanation, a theory for what is wrong with it. The theory points to potential fixes, which can be tested, in part, by trying to apply the appropriate fix and seeing if that solves the problem.

It can be particularly useful if the theory is embedded in a more general model of how the broken system is supposed to work. In this instance, the model consists of an ancient and possibly inaccurate hand-drawn diagram of the Murray Creek water system. I’ve added numbers for the main valves. (It’s a high resolution image, so if it’s opened up separately the details can be seen more clearly.)

Art the Plumber started a process of testing different sections of the system. We started with the simplest theory, really more of a null hypothesis: that there was a problem in the pipes in the lower and upper houses, i.e., between valves 3 (lower house patio) and 6 (upper house). That would isolate the upper house from the spring, which would keep the tank from filling. Small leaks would account for continued loss of water. That was the simplest explanation, but not the most likely one.

Art had identified a leak on a line near the tank (not on the diagram), which he had previously tried and failed to fix. He now totally disconnected the offending part of the water system, eliminating the leak and thus simplifying the problem. After quickly scanning the system of valves 3 to 6, he shut off valve 5 (lower house patio) and opened a faucet before value 3 (under the lower house). We got a small but steady flow of water, which we knew was coming directly from the spring, because water from the tank had been cut off. The taps inside the lower house also worked, and we also tested how well the various valves were able to cut off the flow of water.

After that, Art opened valve 5 and turned off valve 7, cutting off the water tank from the rest of the system. Then we tested a faucet in the lower house patio, which worked just fine, indicating that the pipes from the spring worked as far as that point in the system. Finally, we went up the hill and tried out the deck faucet, the lowest point in the upper house. After draining the water from the upper house system, the water from this spigot slowed to a trickle and then stopped altogether. We waited for 10 minutes, feeling increasingly puzzled, but nothing happened.

Why were we puzzled? To understand this, it’s necessary to know something about gravity feed water systems: Left to its own devices, water seeks its own level, which means that as long as the spring is at a higher elevation than the upper house and the water tank, water should naturally flow through the system and get to them. So why wasn’t this happening?, I asked Art. Small leaks, he said, invoking his all-purpose theory. Having forgotten all the fluid mechanics that I had learned in high school physics (which wasn’t much to begin with), and not really trusting all-purpose stop-gap explanations, I concluded that there must be a partial break (i.e., a big leak) in the downhill section of the pipe on the far side of the creek, at a level between the upper and lower houses, about 20 meters.

Art the Plumber conceded that this theory was quite likely, but pointed out that finding such a leak in the steep, over-grown slope would be “like looking for a needle in a haystack,” and that the most likely problem was in the first bit of pipe near the spring, where he had found a problem with blockage and tree roots 2 years ago. At the time, he’d installed a union joint, making it easier to open up the pipe for clearing crud out. This strategy didn’t seem to fit the most likely theory, but he was the expert, and I was just along for the ride.

We crossed the creek, which is totally dry at this time of year, to the primitive road that runs along its south side. We couldn’t follow the pipe, because, just as Art had said, it immediately disappeared into steep, dense brush, so I followed Art around to the right, where the ascent was easier, and we began the climb. There is no trail, although in many places it was possible to follow the track Art had made two days earlier when he’d been up taking a look around. Following the grading system used in Scotland, this was a level 4 trail: hard scrambles over fallen trees, often at 45 degrees, no discernable trail. (To which I would add: hostile biota in the form of a great deal of poison oak.)

After about 20 minutes we reached the spring. I would never have found it on my own, and I didn’t see it until I was practically on top of it, but there it was, tucked away in a narrow ravine, about halfway up the ridge. The spring is encased in a construction called a spring box, because someone, apparently in the early years of the 20th century, presumably when bottling its water was a commercial enterprise, went to the enormous trouble of hauling cement and mixing equipment up to it, in order to construct a concrete enclosure all around the spring, complete with a removable door for accessing it. The door is now deteriorating rapidly, and another rotten board fell off of it when we handled it. As a result bits of leaves and such were floating on top of the water, which was dammed in by a concrete lip about a foot high. Water apparently comes seeping out of the rock at the back of the spring box, collecting in a basin with a drain at the bottom and an overflow about halfway between the floor of the box and the top of the lip. The overflow was running at about the same rate as we had seen when we tested the flow at the lower house. This meant that the system was backed up, with none of it getting to the upper house and water tank. Art noted that the dead rat that he’d found floating in the water two years ago was no longer to be seen. “Probably decomposed by now,” he remarked.

Art went down for tools: shovel and two pipe wrenches, while I poked around trying to find where the pipe went after it left the spring, still looking (in the wrong place, of course) for the hypothetic partial break, which was still our official theory for the problem. I did in fact find a length of pipe just to the left of the ravine, but this turned out to be a long-abandoned section of pipe that ended in mid-air where the ravine had undercut it. It was an abandoned water pipe from an earlier piping attempt, now a piece of domestic archeology.

Art came back, closed the number 1 valve just up stream of the union joint, and proceeded to take it apart. Although the pipe bore was substantial, at least an inch (25mm), once we’d opened the valve again, there only the same medium amount of water coming out, in spite of the head of water in the spring box. This first section of pipe was clearly badly blocked up.

Art decided we needed something to clean out this length of pipe, so he went down the hill yet again, while I continued to try to track the pipe, locating a place where not one but two trees had fallen on the pipe. Art came back with a 20-foot section of quarter-inch copper pipe. He pried the two sections of pipe apart and opened the valve while I shoved the copper pipe up the old galvanized pipe. A big clot of crud came loose, and water began to shoot rapidly out of the pipe. Art’s strategy of Looking Under the Lamppost for the Lost Keys Because At Least the Light Is Good There appeared to have been successful.

Art collected his equipment and went down the hill, leaving me to spray paint a blaze on the big oak tree overlooking the spring. I climbed as high into the tree as I could manage and sprayed three horizontal white lines and an arrow pointing to the spring. This is supposed to make it easier for us to find the spring in the future, but once we got down, we discovered that the bright sun made it quite difficult to see the shadowed marks. It will be visible once people get close to the spring, however.

By the time we got back to the upper house, the water tank was already starting to fill, going up by about an inch an hour. We declared ourselves well-pleased by the results, comparing our operation to an angioplasty (“No stent, though,” joked Art the Plumber). Then my mom paid him, and he left.

However, I was left puzzled: Why had clearing the top of the line fixed the problem? Finally, I did some research on gravity-fed water systems. Such systems are not that common in municipal systems today, but they are very important in third world situations, and so I was able to find some useful information that explains what happened:

It turns out that Art was correct in his prescription of clearing the top of the pipe run, but proper theoretical explanation is not Small Leaks but rather Friction, leading to something called the Hydraulic Grade Line (HGL). The way it works is this: The narrower the pipe and the higher the flow of water, the more the friction. Also, the longer the pipe and the more bends, the more friction. The pipe system is a siphon, based on gravity but operating in a partial vacuum, which means that it only operates as well as narrowest section of pipe. The friction counters the potential energy from the force of gravity on the water (referred to as the “head”), lowering the effective level of the source of the water. This descending line is referred to as the Hydraulic Grade Line (HGL). When there is too much friction, the HGL ends up being lower than the destination, which is equivalent to trying to get water to go uphill. That’s exactly what happened to the Murray Creek water system: the lower house was below the HGL, so it got water, while the upper house, some 20 meters higher, got none.

Lessons learned: First, Art the Plumber’s theory in practice (crud in the pipe increasing friction) was better than his espoused theory (small leaks); supporting the idea that it’s important to look at practice rather simply listening to the expert’s theories, which may not be fully or accurately elaborated. Second, basic science like the Bernoulli Equation (the physics behind the Hydraulic Grade Line) can be really useful in understanding everyday phenomena and can be usefully applied for constructing explanations in case studies like the Murray Creek Water Crisis. Third, getting the theory right points us in the right direction: toward making sure that the topmost section of pipe, closest to the source of pipe-blocking crud, is kept clear. Fourth, one key implication is that the spring box needs to be “re-stented”, i.e., the screen that used to protect the intake pipe in the spring box (before it rusted out and was removed two years ago) needs to be replaced. Also the main source of crud (not to mention dead rats), that is, the decrepit door to the spring box, needs to be fixed with something more solid and secure. These are now top priorities. The rest of the system appears to be pretty solid. If we’d followed my theory, we’d be looking at replacing much of the system. That’s why Art is an expert rural plumber and I am a professor of counseling!

Figure from: Tawney, E. (n.d.). Visualization of the construction of a gravity-fed water system and treatment system in developing countries. Michigan Technological University.

No comments: