The first version of RS includes Swiss radiosonde SRS C34 (including SnowWhite chilled-mirror dew-point hygrometer, carbon hygristor, copper-constantan thermocouple), a 400 MHz transmitter and a GPS receiver (see Fig. 1). The RS was launched with either Vaisala RS80 or NWS Sippican/VIZ radiosondes on the same balloon (see Fig. 1).
Fig.1 Design of the first version of RS
During IHOP sixteen reference radiosondes were launched and summarized
in the table below. See http://www.atd.ucar.edu/rtf/projects/ihop_2002/RefSonde/
for details. The data from all 16 soundings have been quality controlled
individually through ASPEN, visually checked, corrected individually for
some bad points and averaged to 2s data to gain more SnowWhite RH data.
Comparisons among different humidity sensors are presented below. We haven't
got 6s NWS radiosonde data yet, so comparisons with VIZ sondes will be
presented latter.
| 2nd Radiosonde | # of soundings | Locations | Time of launches |
| Vaisala RS80-H | 7 | Homestead | All 18Z except one at 11Z (6am) and one at 3Z (10pm) |
| Vaisala RS80-A (10-year-old) | 2 | Homestead | 15Z & 19Z |
| NWS VIZ B-2 | 7 | Dodge City | 18Z |
| Total | 16 | 88% of recovering rate |
2. Comparison with Vaisala RS80-H radiosondes
Figure 2 shows RH and temperature profiles for all 7 soundings flied
with RS80-H. Figure 3 shows the scatter plot of comparisons of RH from
H humicap and SnowWhite, and temperature from F-thermocap on Vaisala and
thermocouple on RS.
Figure 3 shows that RS80-H agrees very well with SW at T > 0C, then
shows increasing dry bias as temperatures are getting cooler. The dry bias
at cold temperatures is mainly due to time-lag error with a small contribution
from temperature-dependence error. The consistent colder T from RS80-H
than from RS is hard to understand although it is quite small (0.5-1 C).
Figure 2. Comparisons of RH and T profiles from RS and RS80-H. Red line: RH from SW, blue line: RH from hygristor, pink line: RH from RS80-H, black thin line is T (-70 C) from RS and green thin line is T from RS80-H.
Figure 3. Differences of RH and T from RS80-H and RS (RS80H-RS) as a function of temperature. Red lines are average values.
3. Comparisons with ~10-year-old Vaisala RS80-A radiosondes
Two ~10-year-old RS80-A sondes were launched with RS (see Fig. 4). RS80-A is consistently drier than SW, which is due to the contamination dry bias in RS80-A. There are some problems in RS80-A data above ~6 km on June 23. The scatter plot of comparisons for data below 5 km in Fig. 5 shows the dry bias increases with RH and reaches a maximum (~14%) at RH around 85%, then decreases a little bit. Such dependence of dry bias on RH is consistent with that shown by lab tests described in Wang et al. (2002), but the dry bias shown here has much larger magnitudes than that produced by our correction schemes. This supports our suspension that our correction schemes undercorrect the dry bias for RS80-A.
Fig. 4 Comparisons of RH and T profiles with RS80-A.
Fig. 5 Differences of RH (RS80A-SW) as a function of RH from SW.
4. Comparisons between SW and carbon hygristor inside RS
Figures 2, 4 and 6 all show comparisons of RH profiles from SW and carbon hygristor for all 16 soundings. All those profiles suggest slow response time of carbon hygristor (drier than SW in a moist layer but moist than SW in a dry layer). The scatter plot in Fig. 7 suggests that the carbon hygristor has a moist bias of less than 5% at T> 0C. Another thing shown in those plots is the failure (or insensitivity) of carbon hygristor at colder T (T< -40C) or when RH changes dramatically within a short period of time such as on June 12. For those cases, RH_hygristor stay approximately constant during the rest of the flight and increases slightly as T is decreasing. I think that the insensitivity of carbon hygristor is most likely due to characteristics of variations of RH with resistance and insufficient precision of voltage output for carbon hygristor from RS. Figure 8 shows variations of R/R33 with RH from coefficients used by us (from Meteolabor) and by NWS (1a coefficients from Wade, 1994). Those two sets of coefficients have small differences. Figure 8 shows that RH changes are very sensitive to resistance ratio changes at RH < 33%, i.e. small RH change corresponds to very small R/R33 change, and thus to very small voltage change (R is calculated from voltage measured by RS). Figure 9 shows required voltage changes in order to see 2% RH change, which are less than 2 microVol at RH < ~20%. Required voltage changes in relative values (divided by corresponded voltage) are less than 0.1% (see Fig. 10). Based on Thomas's email, the resolution of hygristor voltage output is about 100'000 that is more than 16 bits. What does it correspond to voltage? It seems not enough to measure 2 mV changes. We are waiting for NWS VIZ data to see whether it also shows such insensitivity.
Figure 6 comparisons of RH profiles from SW and carbon hygristor at Dodge City.
Figure 7. Differences of RH (Hygristor-SW) as a function of temperatures from all 16 soundings.
Figure 8. Comparisons of variations of RH with resistance ratio at difference temperatures between coefficients from Meteolabor (solid line) and NWS 1a coefficients (dashed line) (Wade 1994).
Figure 9. Required voltage changes for measuring 2% RH changes at different temperatures.
Figure 10. Required voltage changes in relative values for measuring 2% RH changes at different temperatures.