This is a guest post from Steve Bellovin, a professor in the computer science department and affiliate faculty at the law school at Columbia University. His research focuses on networks, security, and public policy. His opinions don't necessarily reflect the views of Ars Technica.
"We spent the last few months trying to get ahead of this whole threat so that we could provide as much notification and preparation time to the public," Miyagi said. "I accept responsibility for this, this is my team, we made a mistake."
But the odds strike me as very high that the user interface was simply implemented by a Web programmer, rather than being designed by a human factors specialist.
I have no doubt that it was a bad interface. As TheNew York Times writes:
Vern T. Miyagi, the administrator of the agency, said that during the drill, an employee—whom he did not identify—mistakenly pushed a button on a computer screen to send out the alert, rather than one marked to test it. He said the employee answered "yes" when asked by the system if he was sure he wanted to send the message.
First, you don't put the real button and the test button on the same menu. Second, asking people to simply acknowledge that they want to do what they just asked to be done simply doesn't work. This is very well known; it's even satirized in a W3C Wiki:
A window in which resides a button labeled "OK" and a variety of text and other content that users ignore.
This was a regular, end-of-shift drill. An employee—thankfully unnamed, since it wasn't this person’s fault—accidentally clicked on the wrong on-screen icon, then did what was normal practice: clicked OK (or perhaps typed "yes") to the confirmation screen, purely out of habit. After all, that's exactly what had to be done at every shift change—only this time, the result was accepting the wrong action.
And again, it was not the employee's fault. Don Norman said it well:
It is far too easy to blame people when systems fail. The result is that over 75 percent of all accidents are blamed on human error. Wake up people! When the percentage is that high, it is a signal that something else is at fault—namely, the systems are poorly designed from a human point of view.
And of course, using a different user interface for sending real alerts risks a different failure: when an actual emergency occurs, people are stressed out, worried about themselves, their families, and (in this case) the entire planet. The last thing you want is to have to think through something new and different; failure to send a necessary warning has its own, very serious consequences. This is why I say that a design like this should be created by a real human factors expert.
There's another issue here beyond the user interface problem: the total system. Once the alert was sent, there was no easy way to cancel it. It took 38 minutes to send the All Clear message because of alert system design requirements, per The Atlantic:
IPAWS notices have a specific format, which must be composed formally and in advance. Audio files for broadcast notices must be recorded or generated and uploaded. Often, this has to be done by special software on special equipment.
And alerts to users are limited to 90 characters, with no embedded media or URLs allowed. Fortunately, that is being changed. URLs? Sure—but your Web server better point to a really good content distribution network (CDN), because, about 30 seconds after an alert like this is sent, you're going to have a very large number of people clicking or tapping on it. The cell phone network will need lots of bandwidth, too, both to customers and to the (presumably internal) CDN.
So: we had a bad user interface that talked to an inflexible system that had no pre-prepared cancellation message and was destined for a network that had very stringent limitations on what it could deliver to end-users.
Hawaii has implemented two-person authentication for real alerts. That's good, though I wonder how good a job they did testing the new code. I also caution that two people can be as habituated to the same end-of-shift rituals as one. What's really needed is a thorough, top-to-bottom review of the entire system starting with, but not limited to, the user interface.
Camp Widjiwagan counselors have been known to tell young campers (including this author) certain tall tales as they ventured up through the Boundary Waters Canoe Area in Northern Minnesota. Crossing into Canada, for instance, they would say that if you squint hard, you can see a “border chain” hanging in the lake below. Some kids laughed it off — others were convinced they could spot it. This mythical nautical divider was an inside joke among counselors, but the land border is more visibly marked than many people realize.
From Arctic Village, Alaska to Houlton, Maine, the border between Canada and the United States is the longest in the world. Much of the surrounding landscape is relatively wild and untouched. But extending out ten feet from the line on either side is a zone known as “the Slash.”
Stripped of trees, this slice runs through national forests and over mountains. It is too long and remote to be continuously cut down, but every few years (longer on the Western sections, where growth is slower) workers freshly deforest the greenery that grows back.
It might seem unnecessary, but there is a reason for this intervention: a person on either side wandering close to the border can see it and recognize they’re approaching the line. So each year, American taxpayers pay around half a cent each to the International Boundary Commission (IBC) to help periodically maintain this dividing void.
The work is split between countries. Per Julia Shipley, “U.S. and Canadian divisions of the IBC both have their own staff, equipment and budget. The two groups meet once a year to divvy up their work.” Crews of five to ten people are sent out to various locations. To the east, many of the crews stay in motels — in the west, they camp out.
“It all started in the 1800s, when the US-Canada border line was set at the 49th parallel,” writes Lew Blank. “The Slash was cut and over 8,000 original border markers were laid down, most of which are still standing along The Slash to this day.” But without GPS systems, the markers strayed by hundreds of feet in many places. Other mapping errors led to the creation of things like the Northwest Angle.
“Seeing the Slash can be as simple as going to Google Maps,” explains Blank, “zooming towards the US-Canada border, and switching to satellite view. Those looking for a more up close view can travel to Newport, Vermont and hop aboard Northern Star Cruises, which will take you right alongside the Slash. Another way to see the slash is in Waterton-Glacier International Peace Park, on the Pacific Crest Trail, the Continental Divide Trail to Canada, or the Long Trail in Vermont.”
The Slash also seems like a unique opportunity to travel from coast to coast — an intrepid hiker and paddler could perhaps portage and kayak through otherwise overgrown sections, forever following the thin line.
Proposals for renewable electricity generation in Colorado are coming in cheap, like, $21/MWh-cheap for wind and battery storage. Though there are a few caveats to those numbers, federal incentives and quickly falling costs are combining to make once-quirky renewable projects into major contenders in an industry where fossil fuels have comfortably dominated since the 19th century.
Early last year, Colorado energy provider Xcel Energy requested proposals for new electricity generation. Specifically, the company needed 450 megawatts of additional generation to meet future demand. In a separate request called the Colorado Energy Plan, Xcel said (PDF) it would consider replacing two coal plants providing 660MW of capacity with "hundreds of megawatts of new wind and solar as well as some natural gas-fired resources" if new resources could be found cheaper than what those coal plants cost to operate (including costs to shut down the plants early).
By late November, energy companies had submitted their best offers. Although exact details of the offers aren’t available yet, Xcel Colorado was required to make public a summary of the proposals (PDF) in the month after the bids were submitted.
It seems that competition is fierce. In Xcel's report it noted that it received 430 bids for 238 potential projects (some developers submitted multiple bids for the same project with different completion dates, ownership structures, and so forth). More than 350 of the bids were for renewable projects. In 2013, Xcel notes, a similar request for proposals yielded just 55 bids.
Pricing it out
Surely some of this activity is due to deadlines—renewable energy developers are eager to get projects on the books and break ground before the federal Investment Tax Credit starts to phase out. Currently, wind and solar projects can build their installations with a 30-percent tax credit (that is, a dollar-for-dollar reduction in the income taxes, according to the Solar Energy Industry Association).
Even if contracts are executed quickly enough to take advantage of federal tax credits, the cheapest renewable bids reported by Xcel today are not guaranteed to win a spot on Colorado’s grid. That's because the bids don’t include costs of interconnection, network upgrades, or “credits for items like quick-start capability,” Xcel says, and all of that will have to be taken into account when the utility does its computer modeling to test out options on its grid.
Still, the prices quoted were encouragingly competitive. Although Xcel's report doesn't have a lot of details, this is what we know:
Out of 152 standalone solar bids, the median bid price was $29.50/MWh, down from a previous competitive price of below $45/MWh signed by Tucson Electric.
Standalone wind received the second-most bids from potential developers (that is, 96), and the median bid price was an astonishingly low $18.10/MWh. That's on the same level as a record-low $17.70/MWh bid put forward in Mexico in November.
87 bids were placed to develop solar-plus-storage installations, with a median bid of $36/MWh. Still, we don’t know what kind of storage was proposed or how much of it was proposed. If you have a giant solar field sending electricity to the grid as it gets made, and a small battery installation to manage frequency regulation or serve a local community for an hour of downtime, that’s not terribly exciting.
11 bids were placed to build wind-with-storage at a median bid of $21/MWh. The same problem with evaluating Xcel's solar-and-storage bids is present in the reported wind-and-storage bids: without more detail, it’s hard to evaluate how much storage comes with that.
Seven bids suggested a combination wind, solar, and battery storage installation, with a median price of $30.60/MWh, the most expensive of all the renewable projects reported.
Five bids suggested combining solar and wind for around $19.90/MWh.
A few more traditional, dispatchable technologies were proposed as well, but Xcel asked bidders to price these out in terms of dollars per kilowatt-month ($/kW-mo). That unit of measurement is considered capacity pricing, or pricing for electricity that's generated when demand exceeds a certain point, so it's not quite comparable to the $/MWh seen above.
Among those resources, combustion turbines came in at $4.80/kW-mo, and combustion turbines with battery storage came in at $6.20/kW-mo. For context, in a 2010 paper (PDF), New England's grid saw a $4.50/kW-mo bid for more traditional fossil fuel generators.
Interestingly, 28 bids for stand-alone battery storage were placed, with the median bid at $11.30/kW-mo. Standalone battery storage was the only category in which Xcel offered some context for how much storage would be available. The 28 bids represented 12 projects that ranged in size from 100MWh (25MW with a four-hour duration) to 1,500MWh (150MW for a 10-hour duration), with a median capacity of 450MWh. Considering there's no grid-tied chemical battery in the world that's as big as the median bid reported by Xcel, the storage bids seem almost overly ambitious. The only planned project in the US that could come close, to our knowledge, is a 100MW/400MWh system being built by Fluence (a joint company owned by Seimens and AES) in Southern California. Fluence has said that when that battery installation is built, it will be the biggest in the world.
Xcel didn’t include pricing for generation systems in which two or fewer bids were placed, presumably to protect company secrets before a final decision on winning bids. Two bids were made to build gas-fired combined cycle turbines and one bid was placed to build a compressed air energy storage (or CAES) system, which is interesting because there are hardly any grid-connected CAES installations around the world (in recent Ars research we found just three commercially operating CAES systems).
One bid also offered a plan for an “internal combustion engine with solar” system. That system would consist of “a small photovoltaic field and a small ‘off the shelf’ internal combustion engine burning wellhead and/or pipeline natural gas" where solar would provide 60 percent of the annual energy, and the gas-fired portion would provide 40 percent of the annual energy, Xcel said. The two resources would not be connected, and neither would be dispatchable, so it's sort of an odd project.
So what's next?
The numbers above give an exciting taste of what could be in store for Colorado, but Xcel's report is best approached with tempered optimism. The utility now has months to craft a plan using information from these bids, and the Colorado Public Utility Commission (CPUC) will make a decision on Xcel's proposed ideas around late July, according to the Denver Post.
A lot could change in the next few months. For one, the Trump administration is currently assessing potential tariffs on imported solar panels after the International Trade Commission ruled this fall that domestic solar panel manufacturers were unfairly hurt by cheap imports from China and Chinese-owned factories in other Asian countries. A tariff will almost certainly make some of the bids including solar more expensive, and the Post notes that bidders will have an opportunity to revise their bids with any updated information if something like this should happen.
In addition, any delays in permitting or breaking ground could cause some of these projects to become much more expensive if they're not in service by 2023, when the higher rates of the federal Investment Tax Credit expire.
But there is one near-term deadline that these rates will impact: the CPUC is set to vote in March on whether to accept the Colorado Energy Plan that would shutter those two coal plants if cheaper, cleaner energy could be found. This recent bid report suggests that the case can be made to move forward with that plan.