4.2. Special Projects
4.2.1. Characterization of the ECC Ozonsonde
The CMDL ozonesonde group participated in the Jülich Ozone Sonde Intercomparsion Experiment (JOSIE) held at the Research Center Jülich GmbH, Jülich, Germany, on February 5-16 and February 26-March 8, 1996. Eight different ozonesonde groups participated in JOSIE in order to characterize the accuracy and precision of several types of ozonesondes in an environmental simulation chamber. The sondes were tested in six computer-controlled simulation flights, representing typical midlatitude and tropical ozone and termperature profiles. The chamber accommodated four ozonesondes during each simulation. A fast response dual-beam UV-absorption photometer [Proffitt et al., 1983] was used as the ozone reference for the experiments.
A summary of the tests by Smit et al. [1997] concluded that the electrochemical concentration cell (ECC) ozonesondes had the best precision and accuracy. However, there were differences in measured ozone among the four participating ECC groups. The NOAA ECC ozonesondes, manufactured by ENSCI Corporation, were within ±2-10% of the UV photometer in the troposphere, but in the stratosphere they started showing a positive deviation. The positive difference increased to a maximum of 15-20% greater than the UV photometer standard at the end of the 2-hour simulation. One of the midlatitude simulation results is shown in Figure 4.5.
Fig. 4.5. Ozone profile measured by an ECC ozonesonde used by CMDL during a midlatitude simulation at the JOSIE ozonesonde intercomparison. A UV photometer was used as a standard. The figure on the right shows the percentage higher ozone observed by the ozonesonde compared to the UV photometer.
A common, but inexact, explanation for the higher ozone is that CMDL uses individual pump flow rate efficiency curves measured in our laboratory to compute ozone concentrations. These curves have a lower efficiency than the widely used average pump efficiency curve given by Komhyr et al. [1986] and, therefore, result in higher ozone computations. The JOSIE results prompted additional tests at CMDL to determine why the positive deviation was observed.
After JOSIE, a series of laboratory calibration tests were conducted to determine the accuracy of the ECC ozonesonde using a variety of sensing solution recipes. The bench top tests were done using a Thermo Environmental Instruments, Inc. 49C ozone calibrator as an ozone source. The 49C output was accurate to within 2% of a National Institute of Standards (NIST) standardized Dasibi in the 30-230 ppbv range. The calibrator ozone levels were adjusted to typical midlatitude profile levels in 12 sequential 10-min time steps, as shown in Figure 4.6. Three ozonesondes were sampling the source stream during each calibration test. We consistently observed a higher response from the ozonesondes using the standard 1% KI (potassium iodide) buffered cathode solution. The ozonesondes were from about a 2% high to 15-20% after the peak ozone level, which is similar to the JOSIE intercomparison experiments. These tests were done at surface pressure so the pump efficiency was not a factor in the higher ozone observed by the ECC ozonesondes.
Fig. 4.6. The simulation ozone partial pressure profile generated by a TEI ozone calibrator that was used for testing various cathode solution recipes in ECC ozonesondes. The nearly 2-hour run was done in the laboratory at surface pressure. The diagram on the right shows the percentage difference (sonde-calibration)/calibration between the ECC ozonesonde measurement and the calibrator.
The effect of different cathode solutions in ECC ozonesonde was previously addressed by Komhyr [1969] and Barnes et al. [1985]. However, the variations focused only on adjusting the KI concentration. The calibration tests at CMDL have shown that the KI is not the primary ingredient responsible for high ozone response, rather it is the sodium phosphate buffers that are added to the cathode solution. The optimum solution was found to be a 2% KI solution without buffers and without potassium bromide. Figure 4.6 also shows some average differences between the 49C calibrator profile and ozonesonde profiles using the 2% KI unbuffered, the standard 1% KI buffered, and the 1% KI with double the amount of buffers, to illustrate the effect of the buffers.
The Boulder ozonesonde site was switched to the 2% KI unbuffered cathode solution on August 8, 1997. Figure 4.7 shows that the total ozone ratio (Dobson/sonde) went from an average of 0.9 to nearly 1.0 after the switch. The new processing of ozonesonde data also includes a humidity correction due to evaporation that occurs when measuring the sonde pump flow rate. This increases the time for 100 mL of flow by 2% to 3.5% in Boulder. Trinidad Head has also used the 2% KI unbuffered solution since the beginning of ozonesonde launches there on August 21, 1997. The only other location that used the new solution was the intensive project at Sable Island in September 1997. All of the remaining stations will switch to the 2% KI unbuffered solution in early 1998. Several dual flights with the old and new solution will be necessary to obtain a correction for the previous data.
Fig. 4.7. The total ozone ratios of the Dobson spectrophotometer and ECC ozonesondes for the weekly Boulder ozonesonde flights from August 1197 to January 1998. This illustrates the better comparison the 2% KI unbuffered solution gives after the switch on August 21, 1997. The before and after average ratios and standard deviations are shown in parenthesis.