The pictures look bad: The shoreline receding, grass growing up where water once would have been.
Is it time to panic that Carvins Cove, the Roanoke Valley’s biggest water source, is about to dry up?
Not yet — but it is time to do some math.
The first math to know is this: about 44%.
That’s the percentage of water in the Roanoke Valley that comes from Carvins Cove. Whether you think that’s large or small is a matter of perception, but the main point is that it’s not 100%. Carvins Cove occupies an outsized impression in the public’s mind because it’s open to the public for recreation, which often leads people to forget about all the other water sources. Spring Hollow, the valley’s other big reservoir, is closed to the public. About a quarter of the valley’s water sources are definitely out of sight and out of mind because they’re wells.
These numbers wiggle and jiggle some depending on conditions, but here’s where the Roanoke Valley’s water came from in July:
Carvins Cove: 44.2%.
Spring Hollow: 29.1%
Crystal Spring: 15.8%
Wells: 7.5%
Muse Spring: 3.4%
The Falling Creek reservoir is completely offline right now for dam remediation. If it were in operation, it would come in at somewhat less than half of Crystal Spring’s output.
Nonetheless, Carvins Cove is the valley’s biggest water source, and we are in a drought, so let’s get on to the main event: Just how serious is this and when should we panic? These would be legitimate questions even if there weren’t the consternation (to put it mildly) in some quarters about the Google data center coming to Botetourt County. (Disclosure: Google is one of our donors through its Google News Initiative, which has helped support our expansion in the New River Valley. Donors have no say in news decisions; see our policy. Also keep reading because I will say some things that Google may not like, but those points, like others, will be based on math.)
I call your attention to the Western Virginia Water Authority’s public dashboard, which tracks all of this data, so you can follow along at home.
The last time Carvins Cove was full — “full pond” is the term — was mid-May 2025. While water levels rise and fall with rainfall, the cove’s water levels have generally been dropping ever since then. The declines through the summer, fall and even into the winter of 2025 were completely normal. The cove typically lowers then and recharges in late winter and early spring. The cove hit a low point of 1,157 feet above sea level in early February (water levels are measured in feet above sea level, not feet above the reservoir’s bottom). That was not an unusually low number; it was in line with other years. (The historical average for that time of year is 1,160 feet above sea level.)
Spring 2026 brought some rain, but not enough. The water level rose to 1,161 feet in mid-April, and has generally been dropping ever since. This is where our current troubles began: The cove did not recharge to full pond early in the year as it typically does.
Let’s go back to that 16th week of the year, mid-April. Over the past four decades, the cove has only been that low that late in the year three times: 1999, 2002 and 2008.
In April 2008, though, the cove was rising, not falling, and went on to have a pretty normal summer.
That leaves this year, 1999 and 2002.
The year 2026 is tracking almost exactly like those years, just at a higher water level. In all three years — 1999, 2002 and now 20026 — the cove had failed to recharge during the winter/spring, so it started the dry season at a low level, then fell lower. We’ll get to others, but let’s keep looking at numbers.
In early August of this year, the cove’s water level reached the point to trigger calls for voluntary conservation. Today, the cove’s water levels are about 3 feet lower than that mark.
However, in 1999, the cove began the year lower than that. It is difficult to go back in time and make comparisons because in 1999 the valley didn’t have a unified water system; the city and county each had their own, the city primarily drawing from Carvins Cove, the county primarily drawing from Spring Hollow. However, if today’s water system had been in effect then, with the same water conservation levels, then the valley would have hit the mandatory conservation mark in mid-March and triggered the emergency restrictions by mid-August.
The year 2002 was worse. That year the cove would have hit the voluntary conservation mark the third week of the year, crossed the mandatory conservation mark in late February, moved into the emergency conservation level in early April and stayed there all summer. One reason that the cove levels drew so low in 2002 was that Crystal Spring, then the city’s second-biggest water source, was offline, so the city drew more heavily from the cove. Also, at the time, the city was still obligated to sell some water to Roanoke County. These days, water from other sources, including Spring Hollow, is being pumped into the cove’s distribution system. That means the water levels of 2002 overstate the severity of the drought that year, but it’s hard to determine what they would have been without those other factors, such as Crystal Spring being offline and the Roanoke County sales. All we can do is say what they were.
The cove’s current levels, while obviously low, are well above those two crisis years.
We can also look at these numbers another way. In 1999, the Roanoke City Council passed an ordinance that said voluntary conservation measures to reduce water consumption by 10% should kick in when the cove fell 18 feet below full pond. Again, this was in the days when the city wasn’t part of a broader system, with more water sources, but if we apply that 1999 standard to today, then voluntary conservation measures would not have started until late August of this year. The mandatory restrictions then were triggered whenever the cove fell 24 feet below full pond. At the moment, the cove is 18.6 feet below full pond.
We are not in a crisis now, nor would we be if this were 1999.
That’s the past and the present; the question now is what will happen in the future. Here’s what we know from history:
In 1999, the cove started to refill in early October and refilled sharply.
In 2002, the cove started to refill in late October and also refilled sharply.
What we don’t know is whether the same thing will happen this year. Right now, we’re in better shape than those years; if 2026 replicates those years and we get some fall storms, then everything’s fine. We had a close call, but no harm done. If, however, 2026 brings a weather pattern we haven’t seen before — very slow summer levels at the cove, followed by little to no winter recharge — then we do need to worry.
All we know for sure is that, based on history, we’ll know a lot more in the next two to three months. For now, there seems to be no reason to panic. If the weather doesn’t perform the way it usually does, then we’ll be in a situation we haven’t seen before, at least in the time span of this data.
However, it’s important to point out that some important changes have been made since those 1999 and 2002 droughts. Those two crisis years spurred the creation of a valleywide water authority. In those days, the valley’s water systems weren’t connected, so the city of Roanoke was almost completely dependent on Carvins Cove. Now it’s part of a broader water system.
Carvins Cove is uniquely dependent on local rainfall. Spring Hollow draws from the Roanoke River, which means its watershed is much bigger. That’s why Spring Hollow is historically higher than Carvins Cove. Spring Hollow is currently on the low side, too, but not nearly as much as Carvins Cove. Its water levels rose for part of the summer while the cove was going down. Spring Hollow is now going down again, too, but it’s back to about where it was in mid-July.
Remember those figures I cited at the top about what percentage of water comes from where? That was for July. As the cove has gone down in August, the water authority has shifted more production to Spring Hollow. Preliminary data shows that in August, Spring Hollow’s share of the valley’s water production rose from 29.1% to 34.7%. Carvins Cove’s share slipped down to 42.2%. One reason Spring Hollow’s share went up is that Muse Spring has been taken offline to install a new well. This was exactly why Roanoke Valley governments formed the water authority, so they would not be so dependent on just one or two water sources and could shift production around as conditions called for. With the higher draw from Spring Hollow, here’s how the water levels there have changed. It’s lower than usual this time of year but is not as drought-sensitive as Carvins Cove is.
The political question that hangs over all this is whether the Roanoke Valley can afford the water demands of a data center (or any other large water user).
Let’s go back to the numbers: The Roanoke Valley uses less water now than it did during those drought years of 1999 and 2002. Google has talked of someday wanting 8 million gallons a day, but that figure isn’t the one that matters. This number is: 2 million gallons a day. That’s what Google has contracted for. Google can ask for 8 million all it wants, but 2 million gallons a day is what the authority is legally obligated to provide. As I’ve pointed out before, the valley uses so much less water today than 20 years ago that Google could use all that 2 million gallons a day and the valley would still use less than it did two decades ago. That’s because household (and commercial) appliances have become shockingly more water-efficient. This decline in water usage isn’t unique to the Roanoke Valley; it’s happening all across the country, mostly due to the humble commode. Where each flush once took about 7 gallons of water, now toilets average 1.6 gallons per flush.
Google recently changed its cooling plans, and its calculations, and now says it just needs 1.2 million gallons a day. That’s less than what Falling Creek can provide. In theory, when that reservoir comes back online, it could cover Google’s needs and still have water left over.
Here’s something else we don’t know: Data centers are working to reduce their water usage; how much more efficient will they become? No clue, but those numbers will likely come down. The more challenging part of data centers is their enormous consumption of electricity. This takes us away from the cove’s falling water levels, but this seems an important diversion. Google was able to reduce the projected water usage of its Botetourt facility because it plans to use air cooling. That means it will use more electricity than before. Google may be buying renewable energy offsets somewhere, but some of that energy in Botetourt will come from the Appalachian Power grid — which is 65% powered by coal.
Google in Botetourt will use less water from Carvins Cove but will burn more coal somewhere in the Appalachian system, probably West Virginia. Is that a good trade-off? Is it better to use more water or put more carbon into the atmosphere? That’s a subjective question that numbers alone can’t answer — but we do need to be aware it’s taking place.
As for whether it’s time to panic about the water levels at Carvins Cove, the numbers we have now say “no.” Droughts come and droughts go, and we’ve seen worse than this. If, though, we have a sequence of bad years that we haven’t seen before, then, yes, we will need to worry — and about a lot more than a data center.
Here in Virginia, we never have a drought of political news. I catch the runoff each week in West of the Capital, our weekly political newsletter that goes out on Friday. Sign up here:
