of Sheffield Sir Frederick Mappin Building. To measure rain, a standard rain gauge was setup
next to the test roof. To measure flow, run-off from the green roof flows into a tank. The level of
water in the tank is taken by a pressure transducer. Logging equipment records readings every minute.
The Hadfield green roofs are nine 3 x 1 m roofs based on the Mappin roof, this time located on the Sir Robert Hadfield Building off the Turner Museum of Glass. The large number of roofs accommodates two types of vegetation and three types of substrate for comparative testing. Run-off similarly flows into a tank and pressure transducer information is logged every minute.
These readings are then processed with an algorithm that attempts to smoothe out excessive sensor variation and invalid readings. This algorithm also interpolates data for the duration that the tank is emptied.
The collected and processed data is what is displayed on this website.
There are two limitations which must be kept in mind while analyzing the data presented
in the graphs and statistics. The first limitation has been termed the "daily wobble" and
is present daily. The second limitation involves erroneous sensor readings resulting in
impossibly high run-off flow rates.
The daily recorded run-off that occurs even during dry weather periods is believed to be the effect of the thermal expansion of water in the tank throughout the day. As the sun rises and heats up, the rate of expansion increases. This is interpreted as increasing flow during the morning hours of the day. Around noon, the rate at which the water heats up starts decreasing. This shows as slowly decreasing flow rates until sunset. During the evening, the opposite occurs as the water cools down. This would be interpreted as negative flow, which is ignored. This means that, depending on temperature, there could be as much as 0.8 mm additional run-off per day.
The impossibly high run-off rates that occur in the raw data from the sensor are believed to be due to the presence of frost and/or ice on and around the roof. These cause the sensor to lock up or malfunction in unknown ways, creating readings outside the usual range and variance which the rainfall and run-off data processing algorithm can not cope with or compensate for. Total rainfall and run-off amounts in affected intervals will be incorrect. Examples of this were observed in January and March of 2008.