Obviously, this sensitivity is not important for nighttime measurements. Also, we only did these tests for the upward-looking instruments. We assume that Rsw is relatively small (and thus little error on Rlw) for downward-looking sensors. (Besides, it is impossible to shade these sensors from direct sun when in the downward-looking position. We would have had to reposition the downward-looking sensors to upward-looking for these tests, and we decided that this wasn't worth the effort and disruption to the data set.)
Sun shadow tests were carried out on all incoming short-wave radiometers and long-wave radiometers. Using a ~5 cm Al foil disc mounted on a 30 cm rod the shadow of the sun was imposed upon the radiometer disc. This prevented the contribution of direct solar radiation but only slightly impeded the contribution of the full sky radiation.
While the tests were being carried out, the data system was set to acquire 5 second data. This data was captured in a designated file. The shadow was held on the short-wave radiometer disc for 60 seconds and then, while the short-wave radiometer recovered, the shadow was held on the long-wave radiometer disc. The procedure was repeated several times.
The tests are ideally carried out when the sky is clear. Some of the following tests were under non-ideal conditions.
| Site | Time | File | Conditions |
|---|---|---|---|
| 1 | 6/6 17:00 | s1l.020606 | clear sky |
| 2 | 6/6 18:30 | s2l.020606 | clear sky |
| 3 | 6/7 10:10 | s3l.020607 | clear sky |
| 4 | 6/7 16:20 | s4l.020607 | clear sky |
| 5 | 6/7 16:50 | s5l.020607 | clear sky |
| 6 | 6/7 17:50 | s6l.020607 | clear sky |
| 7 | 6/8 13:15 | s7l.020608 | sky somewhat cloudy |
| 8 | 6/8 14:10 | s8l.020608 | not a clear sky |
| 9 | 6/8 15:30 | s9l.020608 | scattered clouds, no good |
| 7 | 6/15 10:20 | s7l.020615 | thin high cirrus |
| 8 | 6/15 11:10 | s8l.020615 | thin high cirrus |
| 9 | 6/15 12:10 | s9l.020615 | thin high cirrus |
What we need to calculate is:
f = (Rlw.clear - Rlw.shaded)/(Rsw.clear - Rsw.shaded) = delta.Rlw/delta.Rsw.
The step in Rsw is obvious. To obtain the step in Rlw, we must compute Rlw using the value of B that is appropriate for each instrument:
Rlw = Rpile + SB*[T.case4 - B (T.dome4 - T.case4)].
The upward-looking sensors and their B coefficients determined at NOAA in 2002 are shown below:
| Site | Serial Number | B | delta.Rsw (W/m2) | delta.Rlw (W/m2) | f (%) |
|---|---|---|---|---|---|
| 1 | 31974 | 3.1 | -520 | -6 | 1.2 |
| 2 | 26417 | 2.6 | -320 | -4.5 | 1.4 |
| 3 | 29136 | 2.3 | -620 | -3.5 | 0.6 |
| 4 | 26416 | 2.9 | -660 | -6 | 0.9 |
| 5 | 27907 | 3.5[1.5] | -580 | +6.5[-5] | -1.1[+0.9] |
| 6 | 29260 | 3.4 | 410 | 0 | 0 |
| 7 | 29137 | 1.7 | -610 | -5 | 0.8 |
| 8 | 31977 | 2.4 | -750 | -4 | 0.5 |
| 9 | 31976 | 3.0 | -820 | -2 | 0.2 |
(NOAA suggests rounding the B values, though our shadowing tests appear to work out slightly better -- more square-wave response -- using the actual values.)
The function $ASTER/projects/IHOP02/S/pircal.q plots these tests, which are shown here. In this plot, each panel is labeled by station and date, the time axis is in minutes into the test, the solid line is Rsw with the scale shown and the dashed line is Rlw minus its final value, scaled to have 0 in the middle and 2 W/m2 between the reference lines. A plot with rounded B values is shown here.
The deltas shown in the table were estimated by eye, to ±10 W/m2 for Rsw and ± 1 W/m2 for Rlw. Note that the coefficients are all are about 1%, except for station 5, where the coefficient appears to be negative. This could indicate a problem with the B value. For example, setting B for this sensor to 1.5, makes delta.Rlw = -5 and f = 0.9%. (See the revised plot.) However, we have no other justification for changing the laboratory-derived value for B, and are concerned that changing B could have a large effect on the absolute value of Rlw. Thus, we have chosen simply to set f=0 for station 5 and note that its Rlw value may have an uncorrected error of as much as 1% of Rsw (10 W/m2 maximum).
