The Oregon Scientific WMR928NX AWS that is currently operational is located in a moderately-sized suburban garden in Cleadon at a height of approximately 40m above sea level.   The thermometer is situated in a home-made Stevenson Screen.  The rain gauge is a standard AWS gauge that reports to an accuracy of 1mm, which has its drawbacks, but the earlier years of my records (before I acquired this AWS) also had rainfall measured to the nearest millimetre, so it is at least consistent.  The thermometer is situated in a home-made Stevenson Screen.

The records are kept so as to be as consistent with what an "official" site would give as is possible.  Rainfall and minimum temperature readings are unaltered, but maximum temperatures are calibrated downwards on sunny days during the summer half-year to allow for the garden being a bit of a sun trap (calibration factors are typically 2C on sunny days and 1C on partly sunny days in May, June and July, and 1C and 0.5C respectively in March, April and September).  Maximum temperatures in October-February are unaltered.

For the earlier records prior to when I acquired my AWS, maximum and minimum temperatures are calibrated relative to the values acquired from the AWS, and in most cases differences are small, but earlier thermometers situated near a window or on a north-facing wall would sometimes give higher minima on cold clear nights due to radiative heating (which tended to more than offset the tendency for unscreened thermometers to under-read on cold nights).  Rainfall readings are unaltered.

I have maintained a record of "days of" variables since the very beginning- something that is increasingly lacking in the nation's observing network as the Met Office moves away from manned sites and towards automated sites.  Reports of "days of" variables come from myself and, when I am unavailable at the location, third party reports- which are always checked against radar sequences and prevailing conditions to make sure that they are not bogus.

The heavy use of interpolation is a downside, but it means that the record should be as close to homogeneous as I can reasonably get it.