Description of GIT Two Photon/Laser Induced Fluorescence (TP/LIF)
NO Sensor and Its Performance on NASA's GTE PEM-Tropics B Field
Program


The sequential, step-wise excitation, two-photon laser-induced fluorescence
technique was applied to the in-situ airborne measurement of NO during the
PEM-Tropics B field program. The basis of this technique has been previously
reported (e.g. Bradshaw et al., 1985; Sandholm et al., 1990; Sandholm et al.,
1992; Sandholm et al., 1994; and Bradshaw et al. 1998).  The Nd:YAG pumped,
optical parametric oscillators used to generate the UV and IR excitation
wavelengths were nearly identical to those used during PEM-Tropics A, with
laser line width parameters, spectral energy densities, etc. were within the range
previously characterized and reported for the system.  As in previous
experiments, a reference signal from a known mixture of NO was used as an
"internal standard" to normalize for in-flight changes in laser performance (i.e.,
wavelength and energy drifts) small changes induced by changes in fluorescence
quenching as a function of altitude. These reference measurements were
continuously made at the beginning and end of the laser beam line with the
ambient sampling cell located in between. The improved sensitivity
fluorescence detection optics and photomultiplier tubes (PMTs) used
in PEM-WB were also used in both PEM-Tropics A & B. 

Other PEM-Tropics changes that may affect the performance of the NO portion
of the instrument include: implementation of a high flow rate (20 to 40 klpm),
pseudo-wall-less, 10 cm ID sampling system fed by an inlet flow diffuser-style
nozzle that acted to slightly pressurize the inlet (e.g., 80 mbar) and segregate
against large aerosols.  All other modifications should not have affected the
performance of the NO portion of the instrument.

The GIT PEM-Tropics NO measurements were recorded every 0.2 seconds as
an instantaneous (i.e., 10 nanosecond) probing of the mixing ratio of NO within
a continuously flowing stream of "sampled" ambient air.  This sampling
approach can be correspondingly modeled as the quantitative analysis of a
combined set of "effective grab-samples" where the number of effective samples
combined for analysis is equal to the number of laser pulses that interrogated
the sample during the signal integration period; the "effective sampling-time" for
each "grab-sample" corresponds to the sample's residence time in front of the
detector (+/-0.001 seconds); and the frequency of sampling corresponds to the
probe laser repetition rate (10 Hz).  The sample volume interrogated by the laser
beams was swept from the sample chamber between laser pulses.  This sampling
approach can be shown to give a representative sampling of the NO in the
atmosphere if the spatial (i.e., temporarily encountered) inhomogeneities have a
characteristic scale size (dNO/dx or dNO/dt) that is small compared to the
corresponding response function of the instrument. 

Spatial scale inhomogeneities with scale lengths less than 20 m will not be
adequately sampled by the present configuration of the sensor.  In general, the
sensor and analysis technique used was capable of accurately sampling (at some
specified confidence level) those inhomogeneities with magnitudes of change
greater than the precision of the measurement over spatial scales that were
greater than approximately one half the distances covered during the
reported measurement period.

In order to produce a single data product that would give a meaningful
measurement at the highest obtainable temporal resolution for a given ambient
level of NO, the GIT PEM-Tropics NO data set has been constructed using a
variable integration time format.  The signal strength equation describing the
response of the TP/LIF sensor is linear in signal integration time and in ambient
NO mixing ratio.  Therefore, these two quantities can be readily permuted, if
given a sufficiently rapid enough measurement and sampling time to insure that
the non-stationary functionality of the signal arising from NO's ambient
variability is adequately captured by the measurements.  Based on our analysis
of the data, the above criteria was not always met.  However, we believe that
the measurements did accurately capture a large fraction of the NO's variability
in the atmosphere sampled during PEM-Tropics.

The GIT sensor was limited to reporting time periods greater than 0.2 seconds
due to the effective 5 Hz repetition frequency of the NO portion of the system
used.  Here we nominally operated the 10 pps laser system to acquire NO and
NO2 data on subsequent laser pulses with a background measurement every
10th laser pulse.  This created an effective 4.5 Hz interlaced measurement
frequencies for NO and NO2.

The sample transit time through the entire inlet system and detector was +/-0.04
seconds. All data reported here are given with respect to the GMT time that air
was intercepted at station 0 of the DC-8 and thus has been shifted 0.15 seconds
earlier than the time measurements were recorded.  This shift accounts for the
average time difference between station 0 and the station where our probe
was located, and the total sample residence time through the sample inlet and
detector. The GIT instrument's time clock was synchronized to the ERIG-B time
code via the data system clock to better than 0.1 seconds.  

Flight 8 has been reduced using a fixed integration time because single shot data
reduction was hampered by no real time background measurements.  Otherwise
the measurement period for each data point was chosen to contain a minimum
of two laser pulses plus as many additional pulses as needed so that the reported
NO mixing ratio would have a measurement precision in the 10 to 25% range
(1-sigma).  Nominally, the measurement period was constrained to yield a signal
to noise ratio of 10/1, which would yield a measurement precision of +/-20% at
the 2-sigma or 95% confidence limit.  Each integration period was inspected for
the occurrence of "plume" events as monitored by a statistically significant
change in signal levels during the integration period.  Those integration time
periods thus affected by statistically significant plumes were broken into shorter
time intervals as long as a S/N ratio remained greater than 4/1 (and a signal to
background noise of >5/1) was maintained.  Data below the instrument S/N =
2/1 limit-of-detection are flagged by a value of -888.88 and are reported as
upper limit estimates in the NO_LIM column as the S/N = 2/1 value.  The flag
value of -777.77 corresponds to values above our linear dynamic range and
those values are also reported in the NO_LIM column as lower limit estimates. 
In these flagged data cases, the value calculated is given in a separate column
that reports out of range (high or low) data or as a null (i.e., not reported) value
of -999.99.

Signals were separately recorded from four optical detection packages.  A
multi-channel temporal coincidence filter was applied at both the single shot and
integration period level in order to remove statistically anomalous data at the
99% confidence limit for a Poisson distribution (e.g., exclusion of cosmic ray
events).  In addition, the continuously recorded background LIF signal, taken
every one laser pulse in ten, was used to ascertain its relative contribution to the
uncertainty in the reported NO mixing ratio.  The measurement precision has
been obtained using a photon-statistics based estimate of the uncertainty that
assumes the signal and background are independent and random in nature and
that the uncertainty can be defined by the square root of the variance in the
signal-plus-background measurement plus the variance in the observed
background measurement times the square of the ratios of the applicable duty
cycle for the background versus signal-plus-background measurements. 
 
A "digital" photon counting algorithm was used for signal strengths less than 0.6
detected counts per photomultiplier tube per laser shot.  The lower signal limit
discriminator's "threshold" was set based on determining the most probable
value of each channels analog voltage equivalent input offset.  This photon
counting, using this method to establish threshold in combination with the high
average mV/photon for a single photon event, had a reasonably high and stable
efficiency (estimated to have >95% efficiency).

For those measurement periods where the NO mixing ratio was large enough to
result in signal strengths greater than 0.6 detected counts per laser pulse per
photomultiplier tube, the signal was analyzed using an analog photon counting
algorithm.  The measurement precision in this case includes the necessary
additional term that accounts for the gain spread of the photomultiplier tube(s)
as determined from the probability distribution of measured voltages for
individual single photon events.  

All data were normalized for the effects of laser energy/wavelength drift and
small changes in signal strength due to variations in electronic quenching as a
function of ambient pressure (i.e., altitude).  This normalization adds an
additional uncertainty of +/-20% (1-sigma) over a factor of two change in the
reference (normalization) signal.  Typically the reference signal varied by less
than +/-25% during the course of a flight.  The additional estimated uncertainty
due to the reference normalization is conservatively given as a +/-5% per 25%
change in reference signal which should be taken as a random uncertainty term
at the 67% (1-sigma) confidence limit.  All data have also been corrected for the
effects of H2O vapor quenching of the excited states of NO.  This correction
accounts for the 2% per torr reduction in signal strength with increasing water
vapor.  The normalization factor (i.e., 0.02 per torr) is known to a certainty of
better than +/-15% (at the 67% confidence limit).  The uncertainty of the
calibration transfers to the reported NO mixing ratio values is estimated to be
less than +/-20% at the 95% confidence limit based on the uncertainty in the
normalized (to H2O vapor an reference signal) calibration factors obtained
during each flight.  This uncertainty contains in it the measurement precision of
the calibration point, the uncertainty in the H2O vapor and reference signal
normalizations, uncertainties in flow meters, etc..  The uncertainty of calibration
transfer excluding the H2O vapor and reference signal normalization is
estimated to be less than +/-15% at the 95% confidence limit.  An overall or
total measurement uncertainty (at the 67% or 1-sigma confidence limit) is also
specified for each data point.  The estimated total uncertainty has been obtained
by quadratically combining the measurement precision, accuracy and other
separately specified terms discussed above.  These values give our overall
estimate of the maximum random and systematic uncertainties of each
measurement.

Neither the measurement precision nor the overall uncertainty estimates contain
information on the variability of NO in the ambient air sampled during the
measurement period.  

Measurement period time weighting should be applied to any averaging of the
GIT data.

Specific questions regarding this data set should be addressed to Scott Sandholm
or Hann-wen Guan, Georgia Institute of Technology, Atlanta, GA 30332.

NO_AMV (pptv)= Ambient measurement value for NO in units of pptv
               during the measurement period

NO_AMP (pptv)= Ambient measurement precision for NO in units
               equivalent to pptv

NO_ETU (pptv)= Estimated total uncertainty on measurement value
               in units of equivalent pptv

NO_LIM (pptv)= NO values outside the measurement range of the
               instrument in units of pptv during the measurement
               period