Water vapor studies: the MOHAVE campaigns
The Measurements Of Humidity in the Atmosphere and Validation Experiments (MOHAVE, MOHAVE-II, MOHAVE-2009) inter-comparison campaigns took place at the JPL-Table Mountain Facility (TMF) in October 2006, 2007 and 2009 respectively. The first two campaigns aimed at evaluating the capability of three Raman lidars dedicated to the measurement of water vapor in the upper troposphere and lower stratosphere (UT/LS). The nthird campaign was aimed at evalauting not only the Raman lidars but also several other measuring techniques including several types of radiosondes, microwave, Fourier Transform Spectrometers, and GPS integrated water measurements. During each campaign, more than 200 hours of lidar measurements were compared to balloon borne measurements obtained from 10-15 Cryogenic Frost-point Hygrometer (CFH) flights and over 50 Vaisala RS92 radiosonde flights.
MOHAVE (14-28 October 2006)
In addition to the JPL-TMF water vapor Raman lidar (referred to as “JPL lidar” hereafter), two other lidars were brought to TMF from NASA Goddard Space Flight Center (GSFC), Greenbelt, MD to participate to the campaign. The “Aerosol and Temperature” lidar (referred to as “AT lidar” hereafter) measures aerosol backscatter ratio, temperature and tropospheric water vapor, and has been used as a comparison standard in many past NDACC inter-comparison campaigns. The Scanning Raman Lidar (referred to as “SRL lidar” hereafter) measures water vapor, aerosol and cloud properties, and has also participated in many field campaigns. All three lidars (JPL, AT, and SRL) utilize the same technique, i.e., calculating the ratio of the Raman-backscattered signals returned respectively at 387 nm by atmospheric nitrogen, and 407.5 nm by atmospheric water vapor. There are two limitations with this technique: 1) the instrument loses sensitivity as we approach the tropopause due to the decreasing of water vapor concentration and 2) the instrument needs careful calibration, often requiring an external source of information, for example a water vapor measurement from radiosonde. One of the primary goals of MOHAVE was to evaluate the performance of the lidars near the tropopause. The JPL instrument routinely has acquired data up to 16-20 km but the measurements at such high altitudes could never be validated due to the lack of correlative measurements having the required accuracy. Such validation at high altitudes was finally possible during MOHAVE thanks to the use of balloon borne in-situ sensors of appropriate accuracy.
Figure 1 (top panel) shows an example of one-hour water vapor profiles measured simultaneously by all three lidars, and the corresponding simultaneous CFH and (two) RS92 profiles. Figure 1 (bottom panel) shows the average of all available profiles measured simultaneously by all instruments (which includes only the profiles measured during the time all lidar instruments ran in their standard configuration). All instruments capture the fine water vapor vertical structures below 10 km very well.
If we take the CFH as the reference, a clear wet bias can be observed for all lidars above 10-12 km, and a dry bias is observed on the RS92 averaged profiles above 10 km. The dry bias was expected and is typical of the non-corrected RS92 water vapor measurements. The perfect agreement observed on figure 1 between the two radiosondes installed on each payload is typical of the entire campaign, and illustrates well the good repeatability of the RS92 measurements. The observed lidar wet bias was quickly thought to be a consequence of residual fluorescence in the lidar receiver optics. Fluorescence was immediately suspected because all three lidar systems are pushed to their photons detection limits and therefore become very sensitive to residual fluorescence. For the JPL lidar, the fluorescence was identified in the fiber optic connecting the large telescope to the receiver box.
Figure 2 illustrates the impact of this fluorescence and how it was successfully removed. The top panel shows results obtained with the initial lidar configuration (average of 7 profiles measured simultaneously with a CFH) while the bottom panel shows results obtained after a 355 nm blocking filter was installed at the entrance of the fiber optic (average of 3 profiles). As can be seen the wet bias above 12 km was completely removed after the blocking filter was installed.
Figure 3 shows the comparison of the mean water vapor profiles obtained from all profiles simultaneously measured by the AT and JPL lidars. Very consistent individual profiles, mean, and standard deviations were obtained, leading to less than a few percent relative humidity differences between the two instruments. A small negative systematic bias is observed and can be explained by the different calibration technique used for each instrument. For the AT lidar a single constant is used while for the JPL lidar the best fit to the simultaneous CFH or RS92 profile in the lower troposphere is used.
Figure 4 shows the comparison of the mean RH profiles obtained from all the profiles measured when a pair of RS92 radiosondes was attached to the balloon payload. The observed differences between the two radiosondes are mainly caused by small differences in the data smoothing and editing procedure of the Vaisala software. There is no apparent systematic bias except at low temperature and low relative humidity when the observed bias reaches 0.2% RH.
Figure 5 shows the comparison of the mean relative humidity (RH) profiles obtained from all profiles simultaneously measured by the CFH and the RS92 radiosondes. Eight of the ten CFH payloads also included two RS92 radiosondes. The radiosonde data from each type of sonde were transmitted to two separate Vaisala ground-systems, referred to as “DG27” and “wGPS” hereafter. The Vaisala DIGICORA v2.7 is used by JPL to receive data from the JPL RS92K sondes, and the DIGICORA v3.5 is used by the GSFC lidar staff to receive data from the GSFC RS92 sondes equipped with GPS. The datasets referred to as “RS92K DG35” correspond to data received from R92K sondes on the JPL ground system, then re-processed by DG35 using the “Re-flight Simulation” capability of the GSFC software. This way the various versions of the Vaisala software could be cross-validated. The main feature is the systematic dry bias of the RS92 compared to the CFH. This bias has been observed before and various empirical corrections (time-lag, temperature, solar radiation, etc.) have been developed in the past.
During the entire MOHAVE campaign, the AT lidar water vapor retrieval used a single calibration constant determined from previous campaigns while the JPL water vapor retrieval used a calibration constant calculated for each profile from the best matching radiosonde or CFH profile available at the time. To assess the possible impact of the calibration method on the accuracy of the measurement, the JPL water vapor profiles were compared to those of the AT lidar for two retrieval configurations: 1) the normal JPL retrieval, i.e., using a different calibration constant for each profile, and 2) a modified retrieval that normalizes the JPL measurements to the AT measurements, which is equivalent to using a single calibration constant throughout the campaign. For all the profiles measured simultaneously by the JPL and AT lidars the standard deviation of the JPL lidar measurements in each retrieval configuration is compared to that of the AT lidar.
The results are shown in Figure 6. When the JPL lidar measurement is calibrated to the AT lidar measurement, the two standard deviations match perfectly up to 12-13 km altitude where statistical noise starts to contaminate the measurements. In the other hand, when each JPL measurement is normalized to each simultaneous radiosonde measurement, the standard deviation increases by 5% near 10 km altitude. This additional variability was introduced by the fact that even simultaneous and co-located radiosonde profiles do not always match the longer-time -integrated lidar profiles. This result illustrates the upcoming challenges of finding the appropriate calibration method to be able to detect small long-term trends in water vapor.
Why the MOHAVE campaign was so important for the JPL water vapor Raman lidar?
The lidar wet bias was found to be caused by fluorescence in all three lidars’ receiving systems. The measurement of very low water vapor mixing ratios near the tropopause requires the lidar to be pushed to its detection limit, thus making it very sensitive to this type of well known residual aberration of the optical components. After the presence of fluorescence was demonstrated, it was decided to re-configure the receiver in order to permanently redirect the intense 355-nm returned signal (the source of contamination) out of the fiber optics used ahead of the water vapor detectors. The MOHAVE-II campaign took place a year later with the JPL lidar receiver newly re-configurated and fully operational.
MOHAVE-II (6-17 October 2007)
Because fluorescence was clearly identified during MOHAVE, all three lidar receivers were reconfigured during the first half of 2007 with the objective of being operational and “fluorescence-free” during MOHAVE-II in October 2007. Specifically, the JPL lidar receiver was modified to avoid any contamination by the 355-nm signal inside the main receiver path (fiber optic and subsequent splitters). The 355-nm signal was therefore re-directed out of the main path immediately after the focus of the telescope, and before the entrance of the fiber optic. Improved (non-fluorescent) coated optics were also used. Figure 7 shows the MOHAVE-II average of all available profiles measured simultaneously by all instruments (which includes only the profiles measured during the time all lidar instruments ran in their standard configuration, i.e., similar to figure 1 (bottom), but this time for MOHAVE-II). A clear improvement in the agreement between the CFH and all three lidars can be seen, unfortunately at the expense of the signal-to-noise ratio and the resulting cut-off altitude (at least 2-km lower than during MOHAVE). As during MOHAVE, the RS92 uncorrected mean profile is too dry in the upper troposphere. Using the results from both the 2006 and 2007 MOHAVE and WAVES inter-comparison campaigns, an empirical correction can be performed, leading to a perfect agreement between the corrected RS92 (“RS92Milo” in figure 7) and the CFH profiles.
In addition to evaluate the performance of the lidars in the upper troposphere, MOHAVE-II allowed to evaluate the accuracy of the lidar calibration (calibration obtained by normalizing the lidar profiles to the closest available radiosonde measurements). Figure 8 shows a time-altitude color contour plot of water vapor measured by the JPL lidar on October 10, 2007. The superimposed black solid curves represent the corresponding water vapor perturbation from the nightly mean as measured by each of the four radiosondes launched that night. The dotted black lines indicate the time-altitude position of the balloon and represent the zero-water vapor perturbation. Significant short timescale and small vertical scale variability is clearly observed (>150% within 2 hours), typical of that observed throughout MOHAVE-II. Considering that the radiosondes were launched from the lidar site, this figure raises an important issue of lidar calibration accuracy.
MOHAVE-2009 (11-27 October 2009)
Link to AMT Special Issue on MOHAVE-2009: http://www.atmos-meas-tech.net/special_issue27.html.
The main objectives of the campaign were to 1) validate the water vapor measurements of several instruments, including, three Raman lidars, two microwave radiometers, two Fourier-Transform spectrometers, and two GPS receivers (column water), 2) cover water vapor measurements from the ground to the mesopause without gaps, and 3) study upper tropospheric humidity variability at timescales varying from a few minutes to several days. A total of 58 radiosondes and 20 Frost-Point hygrometer sondes were launched. Two types of radiosondes were used during the campaign. Non negligible differences in the readings between the two radiosonde types used (Vaisala RS92 and InterMet iMet-1) made a small, but measurable impact on the derivation of water vapor mixing ratio by the Frost-Point hygrometers. Figure 9 shows the campaign-mean mixing ratio profiles (top) measured simultaneously by RS92 and CFH, and their difference (bottom). Both corrected and uncorrected RS92 are compared to CFH. The thick, red (uncorrected) and orange (corrected) vertical and horizontal bars indicate layer-averaged differences and standard deviations respectively. As observed in previous campaigns, the RS92 humidity measurements remained within 5% of the Frost-point in the lower and mid-troposphere, but were too dry in the upper troposphere.
Over 270 hours of water vapor measurements from three Raman lidars (JPL and GSFC) were compared to RS92, CFH, and NOAA-FPH. The JPL lidar profiles reached 20 km when integrated all night, and 15 km when integrated for 1 hour. Figure 10 below shows the campaign-mean water vapor mixing ratio profiles in the UTLS obtained from nine coincident CFH launches and JPL Raman lidar nights (+/- 6 hours time coincidence). Excellent agreement between this lidar and the frost-point hygrometers was found throughout the measurement range, with only a 3% (0.3 ppmv) mean wet bias for the lidar in the upper troposphere and lower stratosphere (UTLS). The two lidars from GSFC provided satisfactory results in the lower and mid-troposphere (2-5% wet bias over the range 3-10 km), but suffered from contamination by fluorescence (wet bias ranging from 5 to 50% between 10 km and 15 km), preventing their use as an independent measurement in the UTLS.
During MOHAVE-2009, the comparison between all available stratospheric sounders allowed to identify only the largest biases, in particular a 10% dry bias of the Water Vapor Millimeter-wave Spectrometer compared to the Aura-Microwave Limb Sounder. No other large, or at least statistically significant, biases could be observed. Total Precipitable Water (TPW) measurements from six different co-located instruments were available. Several retrieval groups provided their own TPW retrievals, resulting in the comparison of 10 different datasets. Figure 11 below shows cross-comparison of all the TPW datasets available during MOHAVE-2009. The symbols indicate the difference between the dataset listed in the upper part of each plot (where the min and max number of coincidences are listed) and those listed in the lower part. The vertical bars indicate the spread of these differences See text foe details. Use the colors for a better identification of the datasets. Agreement within 7% (0.7 mm) was found between all datasets. Such good agreement illustrates the maturity of these measurements and raises confidence levels for their use as an alternate or complementary source of calibration for the Raman lidars.
Figure 12 below summarizes the campaign-mean differences between all available datasets. CFH (respectively MLS v3) is taken as the reference in the troposphere (bottom panel) (respectively, stratosphere, top panel). The grey dotted curves show water vapor variability (%) estimated from the standard deviations measured by CFH and MLS over the entire campaign. The top and bottom panels are purposely shifted horizontally to mitigate the 3-7% difference between the tropospheric and stratospheric reference. The empirically corrected RS92 and ALVICE lidar are plotted using striped bars instead of plain bars.
Tropospheric and stratospheric ozone and temperature measurements were also available during MOHAVE-2009. The water vapor and ozone lidar measurements, together with the advected potential vorticity results from the high-resolution transport model MIMOSA, allowed the identification and study of a deep stratospheric intrusion over TMF. Figure 13 below shows PV maps from the high-resolution advection model MIMOSA, and ozone and water vapor anomalies measured by the JPL lidars on the night of October 19. The PV maps are shown for 03 UT (left) and 12 UT (right), at 355 K (top) and 330 K (bottom). Open circles and arrows point towards well identified regions of PV/ozone/water vapor correlation. At 355 K (top maps), the tropopause line steadily approaches TMF as the night progresses. This is characterized on the lidar data by increasing ozone mixing ratio throughout the night at around 12 km. At 330 K (bottom maps), a filament of enhanced PV passes over TMF early in the night, then exits the TMF area later in the night. It is clearly associated with an ozone-rich and dry layer near 9-10 km early in the night, splitting in two at 0400 UT, the bottom part propagating downward throughout the night. These PV-ozone (respectively ozone-water vapor) correlations (respectively anti-correlations) are remarkably well captured by the JPL lidars at small vertical and horizontal scales, and short time scales. The lidars measure systematically dry, ozone-rich air originating from the lowermost stratosphere on the poleward side of the subtropical jet, and mosit, ozone-poor originating on the equatorward side of the jet. These observations demonstrate the lidar strong potential for future long-term monitoring of water vapor in the UTLS.
Water Vapor Raman Lidar Calibration
Water vapor Raman lidars typically use integration times ranging from a few minutes to a few hours and produce vertical profiles (zenith looking). On the other hand, radiosonde measurements have temporal resolution of a few seconds and measure along the balloon flight path, i.e., likely to drift away from the launch site. Given the well-known high variability of water vapor on short timescales and horizontal scales (>150% within two hours frequently observed by the JPL lidar, one can easily question the accuracy of the normalization from a nearly - but not perfectly - simultaneous and collocated radiosonde measurement. A series of calibration experiments using radiosonde and a laboratory-calibrated lamp were performed on seven nights during MOHAVE-II. Figure 14a shows the time series of the lidar calibration constant obtained during these seven nights. Each asterisk represents the normalization constant of one 5-minute-averaged profile to the radiosonde measurement closest in-time. The JPL water vapor lidar at TMF uses multiple signal intensity ranges to cover the entire troposphere. The calibration of the lowest range (3-7 km) is represented in red, the mid-altitude range (4-10 km) in blue, and the upper range (6-15 km) in green. The calibration constant time series shows high variability. In addition to the profile-to-profile variability observed for all ranges, there is an apparent jump (up) on October 12 in the calibration value of the upper range (green curve), and a step (down) on October 14. These steps are associated with two planned instrumental changes discussed later. The changes affected the mean and standard deviation of the upper range calibration constant over the entire 7-day period but did not affect that of the other ranges. The standard deviations associated with the observed variability generally range from 20% to 35%. Several normalization algorithms were tested and all of them yielded variability of the same order of magnitude (only one showed here).
Using nightly averages (figure 14a) reduces the calibration constant standard deviation but this method requires multiple radiosonde launches on the same night, and is therefore expensive and unlikely to be used routinely for long-term measurements purposes.
To assess the stability of the transmission of the JPL lidar receiver, calibration experiments employing a 200-W Quartz-Tungsten Halogen (QTH) lamp were performed on the seven nights of MOHAVE-II. The lamp is permanently mounted five meters above the 90-cm diameter Newtonian telescope’s primary mirror. The calibration experiments (called “lamp runs” in the remainder) consist of acquiring lidar signals exclusively from illumination by the lamp. The laser beam is shut down, the hatch by which the laser beam normally exits the room is shut, the room is completely dark, and the lamp is turned on. The lamp-only signals are acquired the same way as the normal lidar acquisition. The ratio of the signals collected in the water vapor and nitrogen channels now represents the ratio of the overall (optical and quantum) efficiency of each channel convolved with the spectral irradiance of the lamp at each wavelength. This ratio does not provide an absolute calibration constant of the lidar. However assuming a constant lamp power output ratio over time, the variation of the channel ratios is equal to the variation over time of the lidar receiver’s absolute calibration constant. This is true even if the lamp illumination is not uniform or if the absolute lamp intensity changes.
Figure 14b shows the time-series of the ratio of the water vapor and nitrogen channels calculated for each of the 33 5-minute ‘lamp runs’ obtained on October 10-17. The color-coding is identical to figure 9a. There is a striking contrast between this plot, characterized by very low variability, and figure 9a. Other than a clear step (up) of the lower range on October 11 (red curve), and a clear step down (then up) on October 12 (respectively October 14) on the upper range (green), the time series of the channel ratios of all three ranges remains mostly flat. The reason for the observed steps is well known, each step being related to a specific change in the instrumental setup. For the upper range (green), a PMT was changed on October 12, and the voltage applied to it was increased on October 14. For the lowest range (red) the illumination configuration was different on October 11 (the lamp runs were performed with the hatch open instead of shut).
Neither calibration technique (lamp or radiosonde) is satisfying for specific applications such as the detection of small long-term trends in atmospheric water vapor. However, a well chosen combination of the two methods can be found suitable.
Figure 15 a-b (left top and left-middle) is identical to figure 9 a-b, except that the calibration values have been averaged for each night (7 nightly mean values). The two new time series show a very good anti-correlation that can be easily understood: the instrumental changes detected during the lamp runs result in a similar (inverse) change in the absolute calibration using radiosonde. The product of the two time series, plotted as figure 15c (left bottom), remains flat, with associated standard deviations for all ranges well below 1% (except for the lower range which is affected by the change in illumination associated with the open hatch). This flat time series is representative of the calibrated lamp high stability with time, and opens the gate to a new calibration approach The idea is to use the normalization to radiosonde on a campaign basis and systematic lidar receiver calibration lamp-runs on a routine basis. Occasional calibration campaigns using radiosonde (or any other technique as far as it provides sufficient accuracy) can be designed to calculate a “campaign-averaged” absolute lidar calibration constant (t1 for a so-called “campaign 1”) and the corresponding campaign-averaged ratio of the lamp irradiances at the water vapor and nitrogen channel wavelengths (L1). Then, depending on the scenario considered, the ratio L can be assumed either constant over an extended period of time (L=L1), or weakly drifting with time. If it is assumed constant, the lidar calibration is routinely obtained without the need for any radiosonde by normalizing to the constant lamp ratio L. The absolutely-calibrated water vapor mixing ratio measured routinely qi is then calculated using the equation:
Pi is the routinely-measured lidar water vapor to nitrogen atmospheric signal ratio, ti is the lidar receiver overall efficiency ratio during the routine measurement (unknown but not needed), P’i is the water vapor to nitrogen channel ratio routinely measured during the lamp runs, L is the lamp irradiance ratio calculated during campaign 1 and assumed constant between two campaigns, and
In any case (i.e., constant or slowly-drifting lamp output ratio), no systematic normalization to radiosonde is needed between campaigns and an absolute lidar calibration is obtained for each routine measurement without having to make any assumption on the stability of the lidar experimental setup between two campaigns (i.e., ti does not need to be known and does not need to be constant). This is a significant advantage to the currently used calibration methods involving radiosonde which all make the assumption that the lidar receiver efficiency remains constant between two radiosonde calibration events. To achieve the 5% accuracy required by the NDACC protocols, the only assumptions that must be made in the present hybrid method are: 1) the ratio of the lamp spectral irradiance outputs at both wavelengths remains either constant or is slowly drifting, i.e., does not change by more than ~2% between campaigns, and 2) The accuracy of the absolute calibration constant
Leblanc, T., McDermid, I. S., and Walsh, T. D.: Ground-based water vapor raman lidar measurements up to the upper troposphere and lower stratosphere for long-term monitoring, Atmos. Meas. Tech., 5, 17-36, 10.5194/amt-5-17-2012, 2012
Leblanc, T., Walsh, T. D., McDermid, I. S., Toon, G. C., Blavier, J. F., Haines, B., Read, W. G., Herman, B., Fetzer, E., Sander, S., Pongetti, T., Whiteman, D. N., McGee, T. G., Twigg, L., Sumnicht, G., Venable, D., Calhoun, M., Dirisu, A., Hurst, D., Jordan, A., Hall, E., Miloshevich, L., Vomel, H., Straub, C., Kampfer, N., Nedoluha, G. E., Gomez, R. M., Holub, K., Gutman, S., Braun, J., Vanhove, T., Stiller, G., and Hauchecorne, A.: Measurements of Humidity in the Atmosphere and Validation Experiments (MOHAVE)-2009: overview of campaign operations and results, Atmos. Meas. Tech., 4, 2579-2605, 10.5194/amt-4-2579-2011, 2011
Aspey, R. A., McDermid, I. S., Leblanc, T., Howe, J. W., and Walsh, T. D.: LABVIEW graphical user interface for precision multichannel alignment of Raman lidar at Jet Propulsion Laboratory, Table Mountain Facility, Rev. Sci. Instr., 79, 09450210.1063/1.2976672, 2008
Leblanc, T., and McDermid, I. S.: Accuracy of Raman lidar water vapor calibration and its applicability to long-term measurements, Appl. Opt., 47, 5592-5603, 2008
Leblanc, T., McDermid, I. S., and Aspey, R. A.: First-Year Operation of a New Water Vapor Raman Lidar at the JPL Table Mountain Facility, California, J. Atmos. Ocean. Tech., 25, 1454-1462, doi:10.1175/2007JTECHA978.1, 2008
Leblanc, T., McDermid, I. S., McGee, T. G., Twigg, L., Sumnicht, G., Whiteman, D. N., Rush, K., Cadirola, M., Venable, D., Connell, R., Demoz, B., Vömel, H., and Miloshevich, L.: Measurements of humidity in the atmosphere and validation experiments (MOHAVE, MOHAVE II): Results overview, Reviewed and Revised Papers of The 24th International Laser Radar Conference, Boulder, CO, 23-27 June 2008, 1013-1016, 2008
Leblanc, T. et al. (2007), Measurements Of Humidity in the Atmosphere: Validation Experiments (MOHAVE) overview of the campaign and first results, Proc. 14th Symposium on Meteorological Observation and Instrumentation (American Meteorological Society), January 14-18, 2007