Some known limitations of the archived data:

• Until 24 September 1998, our humidity and dewpoint readings were obtained from wet and dry bulb thermometer sensors. Because the wet-bulb sensor was not given the constant love and attention it required, the humidity and dewpoint readings were not always accurate. In particular, if in exploring our historical weather data you come across extended periods before 24 September 1998 where the humidity hovers constantly near 100%, chances are that the humidity and dewpoint readings for those periods are inaccurate.
• Fortunately, as of 24 September 1998, our humidity and dewpoint readings are based on a solid-state relative humidity sensor which requires no more attention than the rest of the sensors. Thus, from this date onwards, these readings should be much more consistently accurate than they were in the past.
• In early July 2001, our rain sensor malfunctioned such that rainfall was dramatically under-measured. The sensor was replaced on 29 August 2001. Therefore the historical rainfall records for July and August 2001 should not be relied upon. However, if no rain was recorded for a given day during that period then it is very likely that none fell that day.
• In May 2002 the entire set of sensors was moved from the roof of AT&T Laboratories Cambridge (24a Trumpington St, CB2 1QA) to the roof of the University of Cambridge Engineering Department (CB2 1PZ). These are geographically very close to each other, but there will undoubtedly be some change to the local environment that the sensors are measuring. In many ways the new location is better (less sheltered by surrounding buildings) but there could be some discontinuity in the readings and this may need to be taken into account if you're trying to gauge long term trends that cross this boundary.
• In November 2003 the sensors were moved again to the roof of the William Gates Building (CB3 0FD) which houses the Computer Lab and the Laboratory for Communication Engineering, among others. The barometer was adjusted to provide a correct QNH reading - previously it had been slightly over-reading (by about 1 millibar).
• Although the temperature sensor is surrounded by a white box, it is affected by the presence of direct sunlight. You should be aware that it can overestimate the temperature on very sunny days. However, the move to the William Gates Building has resulted in a site with much better shade, and so we hope this will no longer be significant.
• The rain sensor was malfunctioning between 9th July and 18th July 2004. As a result, no rain was recorded, which was unfortunate as it was a particularly wet week (estimated to be over 50mm of rain, which is approx 10% of the annual rainfall). This was thought to be caused by a storm on the 9th July which resulted in all sensors being unavailable that night.
• The rain sensor's gauze and sponge, intended to filter the rain and prevent a build-up of dirt, were found to be missing on 21 January 2005. This may have caused inaccurate readings until they were replaced on 2 March 2005.
• The rain sensor again failed in a torrential thunderstorm during the night of 28th June 2005. It was repaired on 1st July 2005, but no rain was recorded in that interval. As in 2004, this unfortunately coincided with some very dramatic falls of rain. A nearby weather station run by Paul Oldham recorded 30.6mm in those few days - full details are available here.
• Once again, the rain sensor's gauze and sponge was found to be missing on 20 October 2005. The sensor had become clogged up with dirt, which had caused consistent readings of around 3--5mm of rain to be registered between around 6am to 10am even though it had stopped raining by about 7am. The dirt was cleaned out at 10am. The gauze and sponge were replaced on 7 November 2005. A nearby weather station at Swaffham Prior gave a monthly rainfall reading of 70.4mm (compared to our value of 172.7mm!) for October 2005. The reading for September 2005 was 50.4mm (compared to our 124.4mm).
• The weather vane was found to be incorrectly calibrated such that it was approximately 20° too far clockwise (looking from above). Thus, for example, a southerly wind was being recorded as south-south-easterly. It is believed that the vane went out of calibration in mid-September 2005, and it was recalibrated on 7 November 2005.
• The clock on the computer which gathers the weather data was found to be running roughly six minutes fast, on 29th December 2005. It is thought that the clock might have stopped automatically correcting itself around 25th October, and gradually fallen out of synchronisation over the course of time. The gradual quickening of time would not have significantly affected the data values gathered in this period, merely the periods over which the data are summarised. However, the data for 30th December 2005 may be affected by the addition of six extra minutes at around 18:15 when the clock was corrected.
• Due to electrical testing, no weather readings were collected between 8am on 16 July 2006 and 10am on 17 July 2006. After this time, there was a period during which the humidity, pressure, wind speed and wind direction sensors were stuck giving incorrect readings. This problem was rectified at 11am on 18 July 2006.
• Due to more electrical testing, no weather readings were collected between approximately 11:00 and 11:30am on 19 July 2006.
• The rain sensor's gauze and sponge had again disappeared and were replaced on 31 July 2006 at 14:20. Accurate rainfall measurements resumed at 16:59. It is thought that the loss of these elements may have contributed to the particularly low reading during very heavy rain on 27 July 2006 and consistent zero reading until 31 July 2006.
• The rain sensor gave a reading significantly higher than the actual rainfall on 6 October 2006. The cause of this was that the sensor had been knocked during unrelated maintenance work, preventing it from operating correctly. It was fixed on 9 October 2006 at 10:35. The rainfall values on 6 October have been adjusted in accordance with true values obtained at other local weather stations. Rainfall values on 5 October 2006 may have also been affected by this problem.
• Again, on both 20 November 2006 and 29 December 2006, the rain sensor gave readings significantly higher than the actual rainfall. The rainfall values on these days have been adjusted in accordance with true values.
• The data-logging system failed just before midnight on 2 June 2007. No valid data for any sensors were recorded until 5pm on 4 June 2007.
• The website and data logging infrastructure was changed over to use new software on 19 June 2007 at 16:15. The new software came online at 17:27, but there were some initial teething problems. Temperature data was rounded down to a multiple of 10°C until 10:27 on 20 June. This will have also affected the computed dew point data during this time. No sunshine was reliably recorded by the new software until 17:30 on 20 June, which was a sunny day. No rainfall was recorded in the evening of 19 June, even though there were some fairly heavily showers.
• The data-logging system failed at 16:50 on 4 August 2007. No data was recorded until 09:15 on 6 August 2007.
• The rain sensor's gauze and sponge were again found to have disappeared and were replaced on 24 September 2007 at 15:00. The rainfall values following the 8am - 9am downpour on this date have been adjusted in accordance with true values.
• The pressure, wind direction, wind speed and humidity sensors stuck at 11:40 on 24 September 2007. The problem was fixed at 17:33. The data for these sensors in the intervening period is therefore inaccurate.
• The rain sensor was found to have broken on 9 October 2007. Light drizzle on 8 October 2007 and heavy rain on 9 October 2007 were not logged. No rain was logged until 16 October 2007 when repairs were made.
• The rain sensor was found to be blocked on 22 November 2007. Rain was not draining through the sensor correctly, so inappropriate amounts of rain will have been logged while this was the case, possibly for a few weeks. Rain on the morning of 22 November was not logged. The sensor was fixed at 10:00 on 22 November.
• No data was collected from the humidity, pressure, wind speed, wind direction and dew point sensors from 12:18 on 25 December 2007 until 00:42 on 31 December 2007. The data from all other sensors may also have been inaccurate in this period and should be treated with suspicion. From 00:42 on 31 December 2007 until 11:41 on 2 January 2008, no data from any sensor was logged.
• Again, no data was collected from the humidity, pressure, wind speed, wind direction and dew point sensors from 14:07 on 15 January 2008 until 14:27 on 16 January 2008. The data from all other sensors may also have been inaccurate in this period and should be treated with suspicion.
• On 20 March 2008 a shield was constructed to keep the temperature sensor in the shade in the early mornings, because the temperature sensor had been showing high readings. This measure seemed to have the opposite effect to that intended, and caused temperature readings to be even higher from around 9.30am to 11.00am on sunny days. On 8 April 2008, the shield was replaced by a different design.
• On the 7th May 2008, we suffered a problem with stuck and faulty readings for humidity, pressure, wind speed, wind direction and dew point from about 11:30 to 17:00, including missing all readings from 15:30 to 16:30.
• On the 22nd and 23rd June 2008, a cooling system failure meant that the data logging computer shut down. No data logged was logged from 3:30pm on Sunday 22nd to 10:00am on Monday 23rd.
• On the 9th July 2008, we observed a problem with the rain sensor. The problem was resolved at 4.30pm, but none of the day's continuous rain was logged before that time.
• On 31st July 2008, the temperature and humidity sensors were moved. We suspected that their previous location had been a sun-trap, causing the temperature sensor to give readings above the ambient temperature. The new location is a slatted box which shades the temperature sensor from direct sunlight.
• On 31st July 2008, the readings for humidity pressure and wind speed stuck at about 6pm BST. The logger was reset on 1st August at about 9:00am
• On 12th August 2008, rainfall was not recorded, despite heavy rain falling over Cambridge in the morning.
• From 19th August 2008 to 27th August 2008 inclusive, rainfall was not recorded
• From 3rd September 2008 to 4th September 2008 at about 1pm, rainfall was again not recorded due to a blocked sensor
• From 25th October 2008 to 4th November 2008, rainfall was again not correctly recorded due to a partially blocked sensor. Some rainfall was recorded, but the amounts recorded are far too small to be correct.
• On 3rd April 2009 the Sunshine and Humidity sensors became stuck. Readings for Temperature and Humidity until 17:30 are known to be wrong
• The Sunshine sensor appears to have become a daylight sensor - the threshold for determining 'Sunny' has changed, possibly from 3rd April when it was first noticed that it had become stuck.
• Over the Easter weekend of 10th April 2009 to 13th April 2009, the rainfall was not recorded (blocked sensor again). It was unblocked on the Tuesday once we could gain access.
• Between 6th and 28th February, the rainfall sensor has not been recording any rainfall, due to water damage (!) to one of the cables.
• No results were recorded for 14th August 2010, and there was an interruption on 16th August 2010 due to a power failure
• A blocked rain sensor on 23rd August 2010 may have given high readings for rainfall on that day.
• A blocked rain sensor on 26rd February 2011 gave extreme readings for rainfall on that day of >170mm. A few mm would be a more realistic amount.
• A blocked rain sensor from 19th August 2011 to 23rd August 2011 gave gave extreme readings for rainfall whenever it rained. It really isn't clear why a blocked sensor registers an excess of rainfall, rather than just none. Perhaps 10mm for 19th, and 4mm up to 1pm on the 23rd would be a better estimate.
• The rain sensor failed completely on 6th September. We will endeavour to find a replacement as soon as possible.
• On 30 October 2011, the data logging system experienced a failure at 1:00 AM. No data was recorded until 11:28 AM on 31 October 2011. We will investigate the issues that led to the interruption. A temporary fix was found for the rain sensor, but until replaced the recorded precipitation data should not be trusted.
• On Tuesday 10 January 2012, no data has been recorded between 8:30 AM and 3:30 PM. This was due to a scheduled power outage in the Computer Laboratory building, followed by some issues on cold-starting the weather logging system.
• On 30 January 2012, a planned interruption occured between 4:30 PM and 6:30 PM. On this occasion, we have updated our logging software to process data from the new precipitation sensor (Thies Clima). As a result of a data processing error, some precipitation readings later that day and early morning on 31 January are negative. A fix is in place as of 31 Jan 2012, 4:30 PM.
• On 20 August 2012, a cottonwood seed stuck in the sensor area (protective caps) and moved around by the wind caused erroneous precipitation readings before 2:26 PM.
• On the 28th January 2015 around 11:00 AM, the humidity sensor connection became intermittant due to corrosion. Humidity and Dew Point readings were unreliable and occasionally erratic until the issue was resolved on the 2nd February 2015 at 6:00PM
• From late Sept 2020 until 15 Jan 2021, no precipitation data was recorded due to a failed power supply for the Thies sensor, and the inaccessibility of the building due to Covid.