The temperature dependence of the Doppler-broadened molecular spectrum enables the measurements of the atmospheric temperature by a HSRL. The capability to measure the temperature would eliminate the possible errors due to the difference between the radiosonde reading and the current temperature over the lidar site. For this purpose, the temperature measurement capabilities of the University of Wisconsin HSRL were studied. The presence of clouds and strong layers of clear air aerosols will affect the shape of the molecular spectrum measured with the HSRL. The temperature measurement method presented in the following was used to test the capabilty of the HSRL to accurately measure the molecular spectrum width for layers with small aerosol content.
Lidars have been used for the measurements of the atmospheric temperature profile by many lidar groups. The technique proposed by Strauch et al.  and Cooney  allows calibrated temperature measurements by using the rotational Raman spectrum of nitrogen. With this technique about 1 C temperature accuracy at low altitudes is achieved. Kalshoven et al.  demonstrated a differential absorption lidar method for temperature measurements. They used 2 laser wavelengths and -absorption lines to measure atmospheric temperature up to 1 km altitude with 1 C accuracy. Later Endemann and Byer  reported simultaneous measurements of atmospheric temperature and humidity with a continuously tunable IR-lidar. They used a three-wavelength differential absorption lidar technique and water vapor absorption lines. With this technique a 2.3 C absolute accuracy was achieved. In addition to Raman and differential absorption lidar techniques, Keckhut et al.  used Rayleigh scattering lidar to measure atmospheric temperature for altitudes 30--70 km.
The temperature measurements made with a high resolution lidar have been reported by Alvarez et al. . Their temperature measurement is based on the two barium absorption filters with different bandpasses. Because the strength of the signal received through an absorption cell is proportional to the width of the Doppler-broadened spectrum, the information of the signal strength together with a theoretical calculation for the Doppler-broadened Rayleigh-Brillouin spectrum can be used for the determination of the atmospheric temperature. Their latest measurements have shown that only a 10 C accuracy is achieved for profiles up to 5 km.
The preliminary measurements of the atmospheric temperature made with the University of Wisconsin HSRL have been based on one iodine absorption filter. For temperature measurements the system transmission spectrum is measured by scanning the laser wavelength over the iodine absorption spectrum, similarly as in the system calibration scan. For one temperature profile, data from 5 calibration scans were averaged. This was done to increase the signal to noise ratio of the measurement. The measured profile was calculated by averaging the signal over a 300 m range with 1 km steps.
The signal from atmosphere and detected through the iodine absorption cell is a convolution of the Doppler-broadened molecular spectrum and the iodine absorption spectrum. The Brillouin modified approximation for the Doppler-broadened spectrum was used to calculate the molecular line shapes at temperatures ranging from -70 to +30 C with 1C resolution. The calculated line shapes were convoluted with the measured iodine absorption spectrum. In order to define the atmospheric temperature at certain altitude, a least square fit was used to fit the measured profile to the calculated profile. The temperature that produced the best fit defined the temperature of that altitude. Figure 34 shows an example of the received signal from 8 km altitude observed through the iodine absorption cell normalized by the signal observed with the channel without iodine absorption filter. The iodine absorption spectrum is shown as a reference. The modeled molecular profile is shown for the temperature that produced the best fit. The best fit was found at -47C temperature.
Figure 34: The HSRL signal from 8 km altitude observed through the iodine absorption filter normalized by the signal measured simultaneously without the iodine absorption filter. The iodine absorption spectrum is plotted for reference. The modeled molecular transmission is shown for the temperature that produced the best fit between measured and calculated molecular transmissions. The best fit was obtained at -47C temperature.
The sensitivity of the molecular transmission of the iodine absorption filter to the width of the molecular spectrum is illustrated in Figure 35. The figure shows the effect of incorrect temperature to the fit. For this figure, the modeled molecular transmissions were subtracted from the modeled molecular transmission at -47 C temperature. The temperature difference of 5 C is displayed. The Figure 35 shows that a clear difference between temperatures is achieved, but because the differences are small, the accurate measurements of atmospheric temperature by the scanning technique are difficult to obtain.
Figure 35: The sensitivity of the molecular transmission to the temperature. The difference in modeled molecular transmission to the molecular transmission at -47C is shown with 5C temperature steps.
A temperature profile measured on February 27, 1994, in Madison between 00:00 and 03:00 UT is presented in Figure 36. For the comparison the radiosonde temperature profiles from nearest weather stations are presented. The stations at Green Bay (WI, 180 km northeast from Madison), Peoria (IL, 350 km south from Madison), and St. Cloud (MN, 450 northwest from Madison) provided a radiosonde profile at 00:00 UT. The temperature values measured by the HSRL agree with the temperatures measured with the radiosondes. For the profile between 4 and 8 km the observed rms temperature differences are 2.97, 7.08, and 5.52 C between HSRL and the weather stations. The rms difference between weather stations is 7.06 C. For low altitudes, the largest difference between profiles is observed. This is expected because of the synoptic scale variations in weather conditions between different locations.
For altitudes between 6-8 km, a good agreement between HSRL measurement and the central Minnesota (St. Cloud) radiosonde profile is seen. These altitudes had a low aerosol content providing scattering ratio of 0.02. The temperatures above 8 km show a big deviation from the radiosonde temparatures. This is due a strong aerosol layer, that disturbs the HSRL temperature measurement. The scattering ratio of the aerosol layer above 8 km was 0.3. Also the measured temperature values for altitude between 2 and 3.5 km are colder than the radiosonde values. This is due to presence of a low level aerosol layer with scattering ratio of 0.1 The presence of aerosols deepens the measured spectrum and therefore a fit into this spectrum underestimates the temperature. Therefore, if the temperatures are going to be measured in the presence of aerosols, the effect of the aerosol signal has to be separated from the molecular contribution.
Figure 36: A temperature profile obtained with the HSRL on February 27, 1994. The temperature profile obtained with radiosondes show the atmospheric temperature measured at the closest weather stations. A good agreement between HSRL and radiosonde observations is observed between altitudes of 4 and 8 km. These are altitudes with a low aerosol backscatter content.