First of a four-part series on the Roanoke Valley’s water supply.
We’re in a drought. The official U.S. Drought Monitor lists much of Virginia as being in the “severe drought” category. The governor has urged Virginians to conserve water. Anyone who’s been to Carvins Cove, the Roanoke Valley’s main reservoir, can see the receding water line.
That naturally raises the question: How bad is this? That’s more than just a weather question; it’s also a political one: If it’s this bad now, how bad would it be if the Google data center coming to Botetourt County were already here and drinking down its contracted 2 million gallons per day? (Disclosure: We have received a grant from the Google News Initiative to support our expansion in the New River Valley, but donors have no say in news decisions; see our policy.)
At last, we have some facts to help us answer these questions. More accurately, the facts have always been there, but now they are more accessible through a new database the Western Virginia Water Authority launched that puts the water levels at Carvins Cove and Spring Hollow, the valley’s other big reservoir, in a historical context.
First, let’s take a look at where the authority gets its water. So far this year, 45.8% of the authority’s water has come from Carvins Cove, 22.4% from Spring Hollow, 17.3% from Crystal Spring, 7.0% from various wells, 4.5% from Muse Spring and 3.0% from other places. The point being: Carvins Cove is big, but it’s bigger in our imagination because it’s well-known and well-used as a recreational site. More than half the valley’s water comes from other places, many of them places we can’t even see.
Close observers might note that the figures above don’t include the Falling Creek Reservoir, which is currently offline for remediation work on the dam. Falling Creek holds 435 million gallons of water at full pond, so is distinctly smaller than the two big reservoirs of Carvins Cove and Spring Hollow. When it’s in operation, Falling Creek’s “safe yield” is rated at 1.45 million gallons per day, which would put it below Crystal Spring’s safe yield of 3.5 million gallons a day. Nontheless, were Falling Creek in operation now, all those percentages above would come down some to account for the additional water supply from that reservoir. For our purposes today, we’ll focus just on the two big reservoirs. “Based on current demand, these reservoirs store two-years worth of water,” the Western Virginia Water Authority says, without resorting to the other water sources in the authority’s system.
While it may be dry everywhere, Carvins Cove and Spring Hollow react differently for hydrological reasons. Carvins Cove’s levels are more sensitive to rainfall (or the lack thereof) because its water sources are more localized; Spring Hollow, which draws water from the Roanoke River, is dependent on a larger watershed, so its levels fluctuate less.
Because of that lack of rainfall, the cove is presently at one of its lowest points over the past four decades (the dashboard data goes back 39 years to 1987).
As of Friday, it measured at 1,156.94 feet above sea level (which is 13.4 feet below “full pond”). The last time it was this low was February 2008, when the water level was about 3 feet lower at 1,154 feet. The cove also dipped to 1,154 feet in early January 1998. The only time the cove was lower than that was in the drought years of 1998-99 and 2002. In 1999, the cove fell to 1,141 feet in late September. In 2002, it got even lower: 1,135.9 in early October.
It’s difficult to make comparisons across the years because weather varies, but there are some patterns we should keep in mind. Generally, the cove goes down in the hot, dry months of summer, then recharges during the winter. If you go to the authority’s interactive version of the chart, you’ll see that the cove’s low points are typically in the winter before the winter storms arrive. One reason the cove got so low in 1999 and 2002 is that it never recharged during the preceding winters — the winter weather was too dry. We just had a relatively dry winter, but nothing like in those years. Throughout late 1998 and most of 1999, as well as late 2001 and most of 2002, the cove was lower than it is today. Put another way, in that 1998-99 drought, the cove was at today’s levels in early October 1998. In the 2001-2002 drought, it was at today’s levels in early November 2001.
If we get rain later this year, and over the winter, the current levels, while low, do not seem to be of a historic nature. If, on the other hand, what we’re seeing is the start of a new multiyear drought — and an earlier start than what we saw in the past — then that would be concerning. We won’t really know that until later this year or over the winter. Will we get a tropical storm that drenches us? Will we get a lot of snow? Or will the dry weather simply continue?
Now let’s look at another chart on the authority’s new database. This chart (a screenshot of an interactive version on the authority’s website) looks like a tangled mess, but here are the three lines to pay attention to. The purple one at the bottom is 2002, and the orange one just above it is 1999. Then there’s the dark green line that ends halfway through the page. That’s this year. The trend line there certainly doesn’t look good, but there are multiple ways to read this: One way is that we’re in a bad drought and we don’t know how bad it will get; the other is that we’re still not really close to the worst droughts in our history. One hurricane coming through could solve everything (as it did in 1999 with the remnants of Hurricane Dennis).
Here’s yet another way to look at the cove’s water levels. At “full pond,” the cove holds 6.42 billion gallons of water. At present, it’s at 4.48 billion — 68.9% of capacity.
Google has “reserved” up to 2 million gallons per day; that works out to 730 million gallons a year. If Google were here now and took all that in a year, that would take the cove to 3.757 billion gallons, or 58.4% of capacity.
At its lowest point in 2002, the cove had 2.042 billion gallons of water — about 31.8 % of capacity. If you subtract Google’s contracted 2 million gallons per day/730 million gallons per year from that, then a 2002-style drought plus Google would take the cove to 1.32 billion gallons — about 20.4% of capacity.
You can decide for yourself whether that 20.4% capacity under the most extreme scenario we’ve witnessed is cutting things too close for comfort or not, but we do need to remember one thing. The really big difference between 2002 and 2026 is that in 2002, the valley’s water systems were not connected. Those 1998-99 and 2001-02 droughts were crises for Roanoke that forced water rationing — and also led to the merger of the valley’s water systems. In those years, the city’s main water supply (Carvins Cove) was running low while Roanoke County’s main water supply (Spring Hollow) was doing just fine — but their systems were essentially separate except for a small, localized pipe here or there.
That means the status of Carvins Cove today is important, but we shouldn’t view it in isolation; we need to look at the supply across Roanoke Valley’s entire water system. While the water for Google would come from Carvins Cove, that’s a matter of efficiency. The whole system works together; that was the point of merging water systems. That means we should turn our attention to Spring Hollow, the authority’s second-biggest water source.
What we see here is that Spring Hollow’s levels are down, too, but its present water level is nothing out of the ordinary. There have been at least 14 years out of the past 39 where the water level at Spring Hollow was lower than it is now. Where Carvins Cove is at 68.9% capacity, Spring Hollow is at 80.1% capacity. This is basically the same trend we saw during those 1998-99 and 2001-02 droughts — the cove ran low while the hollow did not. (Ignore that big dip in 2023. That’s not a drought; that’s a safety procedure. Spring Hollow was intentionally not replenished then because of contamination in the Roanoke River, which is where the reservoir gets its water from.)
Now let’s look at the spaghetti chart, as I call it, that shows Spring Hollow’s water levels over the years.
You’ll see that Spring Hollow’s lowest years are different from those at Carvins Cove: 1999 and 2006 started out low but then rose, while 1998 and the end of 2008 compete for being the lowest years for the longest stretches of time. (Again, set aside that 2023 low point; those figures, while real, aren’t normal or weather-related.)
What we should focus on is the dark line that ends in the middle of the chart; that’s this year. It’s not that far off most other years, but is somewhat on the low side — and is trending downwards. In fact, if we throw out those 2023 numbers, Spring Hollow today is at its lowest point at this time of year since 1998. There’s still a big gap between those years, but Spring Hollow is running lower than usual.
We are, after all, in a drought.
This is, however, by no means the worst drought we’ve had. At least not yet.
Can we afford to have Google — or any other thirsty water user — taking 2 million more gallons a day out of the system? If Falling Creek were back on line, its daily output would cover 72.5% of Google’s demand before taking a drop from anywhere else. Ultimately, though, any answer here might be more of a political one than a hydrological one.
It’s certainly fair to ask whether we want to use so much water for cooling data centers, although we’ve never asked that question of other industries in the past (the valley’s second-biggest user is Coca-Cola, at 260,000 gallons a day to make flavored water). Of course, you can also argue that maybe we should have asked those questions before, in which case there’s nothing wrong with starting to ask that question now.
Nonetheless, we come back to this: Americans use a lot less water today than we once did; you can thank more water-efficient appliances for that. As a result, the Roanoke Valley uses less water today than it did back in those extreme drought years. (See those details in this previous column.) Even if we had Google in operation today, drinking its maximum contracted amount, we’d still use less water than we used to. The Roanoke Valley is also using less water than it was projected to be using at this point — and would still be below projections even with Google. That may not justify the data center for some, but it is useful context. The real challenge doesn’t seem to be water; it’s the voracious appetite that data centers have for energy. Data centers are working on ways to reduce water usage, but so far there’s no real way to reduce the amount of electricity required. On the contrary, data centers connected to artificial intelligence require even more electricity than regular ones.
In tomorrow’s column, I’ll take a look at those water use projections and why they’ve been so wrong.
We have more political news and analysis every Friday in West of the Capital, our weekly political newsletter. Sign up for it or any of our other newsletters here:
