The nonstandard thermometer exposure practices at 118 United States cooperative weather stations in 1883 and at 394 stations in 1903 are documented from one unpublished source and one published source. Changes in the exposure practices that resulted from the introduction of the Cotton Region Shelter (CRS) in the 1880s are described. As of 1883, there were five types of thermometer exposure in use at cooperative stations: (i) screened thermometers in freestanding screens (9.9% of all exposures); (ii) screened thermometers attached to north-facing walls and windows (23.3% of all exposures); (iii) unscreened thermometers attached to walls and window casements (50.0% of all exposures); unscreened thermometers attached to porch columns and separated from the main body of the building (13.9% of all exposures), and (iv) “Glaisher-type” stands and exposures (2.7% of all exposures). By 1903, 77.7% of all stations used free-standing screens, and 14.5% used north wall screens; the remaining stations (7.8%) did not use a screen. Free-standing screens were almost certainly not in use at cooperative stations prior to 1881.
In addition, the following features of nonstandard exposures are documented: thermometer bulb heights above ground, the wall azimuth of thermometers, the construction material of the buildings on which the thermometers were exposed, the placement of thermometers with respect to the type of building material and to windows and walls, as well as the susceptibility to artificial heat bias. Based on the details of thermometer exposure practices in the nineteenth century, the published literature, and early results from new field studies in England, which attempt to reproduce historic nonstandard exposures, a positive average annual temperature bias of 0.2°−0.3°C is estimated for the years preceding the late 1890s; the bias probably falls to near 0°C by 1903.