Experimental assessment of screen bias in an early Arctic air temperature time series


Finalist @ National JSHS '15

2nd place (overall) @ New York JSHS '15

2nd place (earth and environmental sciences) @ Long Island JSHS '15

1st place (earth & environmental sciences) @ WAC Lighting Foundation Science Fair '14

1st place (earth & environmental sciences) @ NSF Noyce Science Fair '14

Daniel Hosseinian and Dr. Kevin Wood

Jan 2014 - May 2015 @ University of Washington (remote)


A platinum resistance thermometer inside a previously constructed replica of the HMS Plover's radiation screen was placed adjacent to sensors at the NOAA observatory in Point Barrow, Alaska. Henceforth, surface air temperature time series were obtained from both the replica’s and observatory’s sensors. Data show that relative magnitudes of the biases associated with the Plover’s radiation screen are on average 0.35 °C and up to 2 °C, generally varying by the sun exposure of the screen.


Historical surface air temperature records used to reconstruct extended time series for climate studies often lack sufficient metadata to evaluate sources of bias in the recorded data. Occasionally, it is possible to reproduce the original environment to the extent that an objective measure of bias can be obtained and correction factors can be determined to increase the data’s accuracy. There exists a large collection of unanalyzed hourly air temperature data from September 1852 to July 1854 taken near Point Barrow, Alaska by officers of the HMS Plover using a thermometer encased in a radiation screen. This data set is especially valuable due to it being recorded extensively in a time and place where data are sparse. This location is only 3 km northeast from the present day location of a NOAA observatory recording meteorological measurements, presenting the opportunity to assess the bias of the historical data in comparison to modern data. For this investigation, a platinum resistance thermometer inside a previously constructed replica of the Plover radiation screen was placed adjacent to the observatory’s sensors. Henceforth, surface air temperature time series were obtained from both the replica’s and observatory’s sensors. Data show that relative magnitudes of the biases associated with the Plover’s radiation screen are on average 0.35 °C and up to 2 °C, generally varying by the sun exposure of the screen. These seemingly small inaccuracies are significant in the context of climate change, which is on the scale of a few degrees. Our results indicate screen bias is very large and therefore the interpretation of this and similar Arctic historical temperature data first require correction factors to be implemented to the data set.