Information, Misinformation, Disinformation (or, “These aren’t the droids you’re looking for.”) Part 2

FUKUDROID COMPARE

A MEXT radiation monitoring post, aka “droid,” of a common type manufactured by NEC . This one is at the former Akasawa Elementary School in Aizu-Misato.

GOVERNMENT MONITORING POSTS

[Part 1 here] In recent months there has been a fair amount of controversy concerning the accuracy of the radiation monitoring posts the government has installed all over Fukushima prefecture, and in some neighboring prefectures as well. We wrote about it back in July, 2012:

TEPCO cheating on radiation levels by using “improved” monitoring posts

There are almost 700 of these monitoring posts (675 at latest count), which we refer to as “droids” because of how they look (see photo above). They are all powered by solar panels and use storage batteries.
[Update: 2700 monitoring posts have been installed as of Feb. 2013] Through the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), the government spent a lot of money (we haven’t been able to find out precisely how much) to have them installed, and spent more money to have a web site made that displays the readings: MEXT realtime environmental radiation page

From this reasonably attractive-looking page, users are supposed to click on a prefecture, then on a region of the prefecture, and then choose from any one of a dozen or more municipalities. Then a scrolling list appears on the right hand side, and users can select a particular monitoring post to review (In the case of Koriyama City in Fukushima, there are 393 monitoring posts in the list). A zoomable, scrollable Google Fusion map appears, and the individual posts are marked by colored dots. Clicking a dot gives the current reading at that location, updated very 10 minutes it seems, and it is possible to download data for the entire month. So, thank you for doing that much, at least, MEXT.

This system sucks in many ways. While working with it in order to compare the MEXT readings with our own, we’ve found that it’s impossible to get an overview of more than a small part of Fukushima at any one time, that hunting down particular locations is incredibly time consuming and frustrating, that the cumulative time data does not go back far enough, and that the downloadable data comes with many restrictions and is difficult to pull down efficiently. Yes, MEXT made this system ADAP — As Difficult As Possible.

Screen capture from the MEXT online system of the area around Kawauchi Elementary school in Fukushima. The blue dots represent fixed monitoring posts, the blue diamond a reading taken with a handheld unit. One of the dots was selected, and the reading shown as 0.093 uSv/h (microsieverts per hour). The two fixed units at this location are actually much closer together than this map suggests (see photo further below).

The Safecast map of the same area. The most recent bGeigie readings for the road next to the school, from Nov 9, 2012, were in the 0.12 to 0.17μSv/h range. This is fairly consistent with what the droid shows. Readings taken on the same drive on the roads north and south of the school range as high as 0.18 μSv/h. Safecast readings from a year ago, however, are in the 0.14 to 0.25 μSv/h range alongside the school and up to 0.29 μSv/h on the road to the north. On the one hand this shows that the range of variation we often encounter within a short distance can easily be a factor of two, sometimes more. The most recent readings also support the idea that the schoolyard the monitoring posts are located in has been decontaminated since last year, as has the road (A blogger has reported recently that some areas just beyond the schoolyard remain fairly contaminated, however). MEXT has made their data fairly easy to access for people who know exactly what they are looking for and just want a quick look, but it is extremely difficult and time-consuming to make this kind of close comparison with other data sets. And we couldn’t help but notice that we have a lot more data available for this area than the government provides.

We thought about how people would use our maps, and what they would want to know. Almost everyone will start with a zoomed-out view, and then zoom in for more detail. Our system aggregates the readings from different sample points at each level of zoom, and people can drill down as far as they want to see exactly what kind of samples there are and when they were taken. The MEXT system does not acknowledge that people want to know both average radiation levels in an area as well as individual sample data. After nearly two years, the government still has not provided this kind of radiation map information in a citizen-friendly form.

The two droids in the schoolyard in Kawauchi. One is by Fuji Eectric, the other by Rhinotech. They are actually placed less than 2 meters apart, but on the MEXT map they appear to be separated by about 100 meters. We have seen redundant monitoring stations like these in many locations, often, but not always, at schools, and have not yet received a fully satisfactory explanation why. Is it to have a backup in case one malfunctions? To provide readings at both 1 meter and 50 cm? The result of a contracting error? We’ve been unable to find out.
[Update: We found out. The government unilaterally cancelled the Rhinotech (Alpha Tsushin) contract after 600 units had been installed, and gave the contracts to Fuji Electric and NEC. More below.]

We were alerted to possible problems with the construction and siting of the monitoring posts themselves back when we made the first blog post about them. At the time, a number of journalists and professors noted that the readings at the posts themselves were often noticeably lower than those of the surrounding areas, and that the ground around the posts seems to have been decontaminated or resurfaced in many cases.

A group called Association for Citizens Concerned about Internal Radiation Exposures (ACSIR) investigated further, and publicized its findings in October:

ACSIR page (Japanese)
Asahi Shimbun article about ACSIR’s findings

They claim to have detected a 10-30% disparity in readings close to the monitoring posts, and they suggested that shielding from the monitoring posts themselves was a potential factor. They did not provide their raw data, however, but only graphs and calculations. They accused the government of intentionally skewing the readings.

Greenpeace, which has been doing periodic radiation surveys, also checked the accuracy of the droids, and concluded that many of them were inaccurate:

Official radiation monitoring stations in Fukushima unreliable (Greenpeace)
False hope: radiation monitoring in the Fukushima area (Greenpeace)

Their method seems to have been fairly systematic, to take readings with a reliable gamma spectrometer (Georadis RT-30) very close to the droid, at 1m, 0.5 m, and 0.1 m heights, and also at 5m, 10m, and often 15 or 20m distance in one or more directions. They checked 40 monitoring posts in all, in Fukushima City.

Greenpeace monitoring station data (pdf; Japanese)

Greenpeace stated, “For 75% of them, the radiation readings close to the posts were lower than readings for their immediate surroundings. Contamination levels within 25 metres of the posts were up to six times higher than at the posts themselves.”

Looking at the Greenpeace spreadsheet, however, it becomes clear that the monitoring posts themselves generally read higher than Greenpeace’s instruments do right next to the posts at 1m height. So that’s not the problem. The problem is that the readings at the posts often don’t represent the contamination of the surrounding area very well, sometimes understating it by half or more. As we’ll discuss further, if a monitoring sample is intended to represent an average of a particular area, we actually should find find areas of higher radiation somewhere nearby, but we should find areas of lower radiation as well. Otherwise we can deduce that the monitoring post was placed in a particularly high (or low) spot. But the only way to really know is to take readings systematically in several directions all around the monitoring post. From the Greenpeace spreadsheet linked above it appears that they were able to do that in only a few cases, finding results that supported their conclusions. It would be extremely expensive and time consuming to survey and map the surroundings of all 675 droids in order to identify which ones are in “low spots.” But the important issue they highlighted concerns how people perceive the meaning of the readings: even if the devices themselves are accurate, if someone looks at the display on the droid itself or on the MEXT map without understanding that several meters away might be very different, there’s a good chance they will get a false impression of radiation levels in the area. And because we can point to many examples of the droids being placed in “low spots,” and the government has not really provided enough information to demonstrate otherwise, citizens would probably be justified in assuming that the droids in general read lower than their surroundings.

Safecast has also checked the monitoring posts we’ve encountered, and tried to figure out how they are intended to work.

When we check our readings against the droids, the results are usually close, but people should be aware that discrepancies are common between different kinds of devices when the levels get below 0.1μSv/h. This Fuji Electric unit reads 0.053 μGy/h ( 0.053 micrograys per hour, here equivalent to 0.053 μSv/h), while our Inspector (a pancake tube Geiger counter) shows 0.155 μSv/h, which seems like a large discrepancy. Our Rad Eye scintillator also shows 0.05 μSv/h however. We consider this to be within the reasonable degree of variation that can be caused by the slight differences in the specs of the various devices. Though grays (Gy) are the technically correct unit to use for measuring activity in terms of “absorbed” dose, sieverts (Sv), which represent a conversion of this to “equivalent” dose, are more familiar to most people. When we are looking at gamma or beta activity, grays and sieverts are essentially equivalent numerically, despite their different meanings (much like how 1 liter of water weighs 1 kilo), and the labeling of the droid’s display might be confusing to some citizens.

This unit is by a company called Rhinotech, and is located at Ohto Elementary School in Aizu Wakamatsu. It had apparently failed in some way, so the small external display has been taped over. The Rhinotech units seem particularly prone to breakdown.
[Update: That's because they've been abandoned by MEXT.]

This Fuji Electric unit at Yumoto Elementary School in Tenei, Fukushima, is the most common type we have encountered.  It has a fairly small solar panel and the display is housed separately from the domed body.  There are two fiducial height marks on the case, one at 50cm height and another at 1m.  We initially expected there would be two sensors inside, one at each of these heights, but there is only one, aligned with one of the markers, in this case the lower. It appears that sensors in both these Fuji units and the Rhinotech units are placed at 50cm at elementary schools and at 1m elsewhere. For some reason the display built into this particular sensor module inside gave a higher reading than the external display; it may have a shorter averaging window.  Neither display changed over a ten-minutes period, so both apparently have very long time constants. 

A unit made by Aloka, in a parking lot in Date. It is one of two types of units being used that consists of a short boxlike enclosure, and they are usually fenced as seen here. It is equipped with sensors for things besides just radiation. The display is also labeled in grays.

This unit, in Tateiwa, also made by Fuji Electric, is a “short box” type, whose radiation sensor is behind the small dome. It also has other environmental sensors. This is the same unit pictured with three of our sensors further above.

Four manufacturers produce similar units: NEC (see image at top), Aloka, Fuji Electric (which makes 2 kinds), and Rhinotech. All are solar powered, and net-enabled. The actual detectors inside seem to be well-made scintillators . We’ve seen a few dozen units in all, and while there’s often a noticeable discrepancy between what the droids read and what our geiger counters show, our RadEye scintillator usually agrees very closely. Nevertheless, we seem to notice greater discrepancies as readings fall below 0.1 microsievert/hr. We also couldn’t help but notice that of the dozen or so Rhinotech droids we saw, only one was actually working. This could be just planned downtime, but they actually seemed to have manufacturing defects as well as maintenance problems, such as water inside the display (see photo below).
[Update: One of our volunteers in Koriyama pointed us to more information about the Rhinotech droids. They are actually manufactured by a company called Alpha Tsushin, which was awarded an initial contract worth 370 million yen, for 600 units. After they had been installed, MEXT cancelled Alpha Tsushin's contract and announced that it would source those units, plus 2100 more, from NEC and Fuji Electric instead. On its web site, MEXT stated their reasons as "the company was unable to meet the technical specifications and deliver on time." Quite a few bloggers have commented on this, pointing out that MEXT has not yet publicly stated what specs Alpha Tsushin was unable to meet, or how the units had been tested. It seems like a totally unilateral decision on the part of MEXT. Alpha Tsushin is suing MEXT for 370 million yen. The story has basically disappeared from the press, but the Rhinotech droids remain in place as a testament to waste and mismanagement.]

Geiger Counter Blog report about Alpha Tsushin contract problems (Japanese)

MEXT notification of cancellation of Alpha Tsushin’s contract (Japanese)

Inside an NEC droid. The unit uses mostly off-the-shelf components, and the manufacturing quality seems good. The actual detector is the beige box mounted center top (made by Hamamatsu). These units use an analog timer, most of which appear intended to turn the displays off at 7:00pm. This one is stuck, while others seemed to be misset.

Inside a Fuji Electric droid. Similar in quality and complexity to the NEC unit, it also uses mostly off-the-shelf components. The yellow scintillator, mounted at 50cm here, is also manufactured by Fuji Electric.

The insides of the Rhinotech unit in Aizu Wakamatsu. The actual sensor (a scintillator of unknown make) appears to be inside the small grey box at the lower right, mounted at 50cm. The Rhinotech units have the fewest internal components of the tall droid types. They appear to be a “budget” solution.

Many of the Rhinotech units we saw displayed just a frozen system initialization message instead of radiation readings.
[Update: These units were designed to be independently powered, and the fact that so many of them continue to run at all despite having been abandoned for over a year is notable. The error message is most likely due to the fact that the system is trying to contact a cell phone account that has been cancelled by the government. ]

This Rhinotech unit had condensation inside the display, in addition to being stuck at system initialization.

What about the effect of decontamination? There are hundreds of these units installed now, often in parks, at schools, and on playgrounds which were intentionally decontaminated beforehand to make them safer for children to use. Also, often the ground has been resurfaced in ways which suggest standard contractors’ practice more than attempts to affect readings. But we have seen many cases where the ground around the droid sports a new, labor-intensive surface that seems difficult to justify for reasons of simplicity or tidiness. For example, in Fukushima City recently we found one of the droids stashed a moderate distance away from immediate public view near Fukushima Station. It was located in a small grass park, the only green near the station, which means that dose rates will be lower there than what people experience in front of the station due to the different absorption characteristics of soil versus concrete. The asphalt on the path in front of the droid has been replaced with new, uncontaminated material. We measured the surface contamination in the park on “old” asphalt at 40-50,000 Bq/m2, while that of the new asphalt in front of the monitoring post was basically zero (after adjusting for the background radiation dose rate). The dose rates displayed on the monitoring post were very close to our Geiger counter (bGeigie nano) at 0.28 μSv/h, while the dose rate we measured around Fukushima station ranged between 0.3 and 0.5 μSv/h.

This Fuji Electric monitoring station is located on a path near Fukushima Station. The path near the unit has been recenty resurfaced with new asphalt.

When we compare the readings from a bGeigie Nano and the droid, they are very close: 0.286 uSv/h on the droid vers 0.29 uSv/h on the bGeigie. (The Nano has the waterproof case removed for visibility).

The bGeigie nano detects 87 CPM (counts per minute) on the surface of the new asphalt.

But a few meters away, the the bGeigie Nano detects 1000 CPM on the surface of old asphalt. Did the project supervisors intentionally decontaminate the area around the droid so it would give lower readings? It’s hard to say for certain. But the fact is the radiation readings on the droid are only applicable for the immediately surrounding area, which is much lower than on the path a few meters away.

We found a similar situation at Fukushima University, where the dose rate in the parking lot was in the 0.4uSv range, while a monitoring post in the middle of a grass field showed around 0.23 uSv/h.

In response to the criticism from citizens’ groups, the government agreed last month to modify the location of batteries in the units which might be shielding the detectors, at a cost of ¥100,500,000. This is expected to eliminate a 10% inaccuracy in readings. But the problem is obviously bigger and more intractable than this. It goes back to the desire to make something “as easy as possible” for people to understand and use, versus not caring if it’s “as hard as possible.” We get the sense that nobody involved in the decision making process of establishing the specifications and usability parameters for the droids and the online system really cared enough to think about what information citizens would want and need, and why, and to design a good solution. No-one responsible for putting this information out was specifically tasked with thinking about what the best and most informative system would be, and to oversee it to completion. Instead, it seems a group of bureaucrats slapped together some minimum specs, probably with a lot of input from the eventual manufacturers, sent it out for bids, and took whatever came in. Then they subcontracted out the actual installation work to people who didn’t really care either. They contracted out a web site that looks good superficially but is actually nearly unusable and deceptively uninformative. We can only presume that many people lined their pockets along the way.

Another case of “doubledroids,” this time at Oguni Elementary School, in Date.

There are two droids placed close to each other in this park in Kawauchi as well.

We’ve heard various justifications for the placement of the droids and decontamination around them, including that the goal was to show how low radiation would be when decontamination was complete, or was to provide low background readings so any sudden increase in radiation would be immediately apparent. And looking at how poorly executed the installation of so many of the units we’ve seen is, it’s hard to conclude firmly that providing deceptive readings was part of the plan, or that it would have been successful if it had been. But in fact, it’s hard to argue that they aren’t deceptive, or that those in charge aren’t aware of the fact. These units are helpful as far as they go, but it should be made explicitly clear that they represent point readings that cannot be accurately averaged to get area radiation levels. Although anyone with technical knowledge will probably realize it, they need signs on them stating clearly that the readings are valid only within a short radius, and people should know how much of the area has been decontaminated, and when. The web site should state this same information unambiguously as well.

This is especially true because towns like Kashiwa in Chiba have been doing regular surveys of all parks and playgrounds, as well as schools and other public areas, and posting the data online in a way that makes the changes over time clear. In the case of parks and playgrounds in Kashiwa, readings are taken at many locations and a map is posted at the entrances of the parks themselves and made available online. It is a very simple, effective, and unambiguous way to inform the public. And it’s clear that someone has taken responsibility for doing the legwork and maintaining the records:

Radiation map of Mukaiyama park in Kashiwa, Chiba (pdf; Japanese)

What was and is needed is a public radiation dose monitoring system which gives representative readings for wide areas, not just spot readings for specially selected locations, but which also reflects the actual granularity of the contamination. If we were to specify it ourselves, we might insist on:
–Care taken to ensure consistent ground cover with the surroundings.
–Clear information on the droid itself describing both its capabilities and its limitations
–A map physically attached to the droid showing readings taken periodically in the surroundings up to 100m or more.
–Clear contact information on the droid itself where people can call, email, or look up on the web for more information.
–More easily useable web design, particularly to show both hot spots and averaging over a wide area, and to make the entire dataset openly available so it can be compared with others (like Safecast’s).

Eventually, I think, people nearby are going to need to “own” these droids, that is, take long-term responsibility for them, keep an eye on them, learn their quirks, learn how their readings compare to the surroundings, and establish good communication with the people who run the wider system. For that to work, of course, the people running things will need to show a kind of responsiveness we haven’t seen so far.

We saw one droid we liked which was different from the rest, at a school lunch preparation facility in Tamura.

This unusual monitoring station is in Tamura City, outside the school lunch preparation facility. The upper reading is for the parking lot where the unit stands, the lower gives levels in the kitchen inside the building. A highly visible sign tells people to contact the facility for more information.

The attached info plate gives a lot of relevant and useful information that isn’t found on the other monitoring stations. It tells us that it was manufactured by Images SI, Inc, with accuracy range of 20% and a detection range of from 0.01 microsievert/hr to 5.0 Sievert/hr, uses Geiger-Muller sensors placed at 1 meter height, displaying averaged samples updated every 60 seconds, etc.. It also states clearly that the readings are intended as a general guideline only.

It gives two readings, one for the parking lot outside the building, and one for inside. It has contact and technical information on it as well. It isn’t exactly the droid we’re looking for, but it shows that somebody is taking responsibility for it, and thought about what people would want to know. So these sorts of things are in fact possible. All it takes is for the people doing it to care enough about seeing it done right.

(Part 1 here)

One Comment on “Information, Misinformation, Disinformation (or, “These aren’t the droids you’re looking for.”) Part 2

  1. Great work on this blog post. I had no idea that so much government money was flowing into the installation of these detectors. I hope that someone at the upper levels of the Japanese government will read this and do something. Keep up the great work, Safecast!!!

    -Cyrus

7ads6x98y