The primary error source in the measurement of air temperature at the site was due to radiation loading of the naturally ventilated sensor shield. These errors were not evaluated in the field, but testing by Gill (1983) has demonstrated solar heating of greater than 2° C with wind speeds less than 0.5 m s^-1.
A minor error source in a limited number of the daily temperature maxima and minima was due to use of daily extreme hourly mean values (see discussion in data quality control section). On these days, the maximum and minimum temperatures are therefore conservative, and the times of maximum/minimum occurrence are for the extreme hourly means.
Radiation loading of the 207 sensor shield affected the accuracy of both temperature and RH measurements, primarily when wind speeds were less than 0.5 m s^-1. Error in these measurements therefore affected the accuracy of calculated vapor pressure.
Wind direction was referenced to true north. An attempt was made each year to orient the sensor using its shadow at solar noon (local apparent time), with varying degrees of success. During the 1991 and 1992 seasons, alignment was checked by recording the shadow orientation at a specific date and time. After the field season, solar azimuth was computed based on latitude, date and the observation time. The difference between the predicted and recorded shadow orientation ranged from 2.5° to 8.8° , and adjustments were not made.
There is minimal error in the pressure data, due to the accuracy of the sensor used (± 0.2 mb).
Solar radiation measurement error arose primarily due to the effect of low sun angle on the model 8-48 detectors. This cosine response causes errors to increase with zenith angle. Eppley gives +/- 5 percent for 70-80° angles, and that for 80-90° is no doubt higher.
As discussed in the data description section, data gaps of three to four hours exist in the record due to disconnection of the sensor for battery charging. Ample time was always provided for the sensor to reach thermal equilibrium, prior to continuing the measurements. Periodic resistance measurements were made of an internal thermistor to insure that the specified temperature was maintained.
The greatest error source in the L-down measurement was likely due to the pyrgeometers not being ventilated. Ventilation was not considered feasible, due to logistical considerations such as power requirements. The magnitude of this error was not assessed.
Errors in the determination of net radiation are partially the result of errors in the measurement in the components (or net radiation itself, in 1991). In addition, during precipitation events the radiometer domes were commonly wet. Occasionally snowfall was heavy enough to partially obscure the dome as well; at Delta these intervals were only several hours in duration.
Because precipitation totals were observed regularly only in conjunction with the PCSP observations, there were some occasions when it was difficult to partition precipitation amounts occurring between 1900 h one day and 0700 h the next. As a result, there may be slightly greater accuracy in cumulative precipitation amounts than in the daily totals.
Precipitation amounts during snowfall events were determined by gage catch and from measurement of new snow accumulation. While new fallen snow generally melted from bare tundra surfaces, it was possible to assess accumulation on snow or ice covered areas (e.g. Lake C2). A precipitation amount was then determined using a measured or assumed snow density. These values were compared with those from the gage, and generally the larger value was used.
Precipitation type was also recorded, as either rain, snow, or mixed forms. The reliability of these observations depended upon the frequency with which precipitation type changed, and the degree to which the precipitation was a distinct type. Numerous events during the three field seasons involved multiple types of precipitation, and rapid changeovers between them.
h. Net all-wave radiation
I. Precipitation