In the winter of 1992/93, nearly 12,000 radiosonde soundings were launched from 36 Priority Sounding Stations during the TOGA COARE Intensive Observing Period. Sounding systems provided by NCAR/ATD, including the Integrated Sounding Systems (ISS) of the central Intensive Flux Array

During initial processing of the TOGA COARE sounding data, analysts suggested that radiosonde temperature and humidity measurements seemed too warm and too dry respectively in the lowest levels, at least during daytime soundings. ATD engineers attributed this warm dry bias to radiational heating of the temperature and humidity sensors and sensor arm just prior to balloon launch. NCAR then developed an algorithm and methodology to correct near-surface temperature and humidity data for this so-called sensor-arm heating and the corrections were applied to the COARE data set.

Further analyses of the TOGA COARE data indicated a residual dry bias not resolved by application of the sensor-arm heating correction. Some of these analyses showed large horizontal gradients of humidity and of derived parameters such as Convective Available Potential Energy (CAPE) despite the relatively uniform sea surface temperatures in the COARE IFA.

Scientists and engineers from ATD and from Vaisala (Helsinki, Finland), the manufacturer of most sondes used during TOGA COARE, confirmed the dry bias and identified its cause: contamination of the capacitance humidity sensor by various gaseous plastic and Styroform components. Based on extensive Vaisala testing, ATD and Vaisala developed a prototype correction algorithm based on physical characteristics of the humidity sensors. With additional funding from NSF and NOAA, the correction algorithm was further developed and has now been applied to almost all IFA soundings from TOGA COARE.

Vaisala has taken corrective actions since the problem has been discovered. A new desiccant, made of a mixture of silica gel and activated carbon, has replaced the clay-based drying agent. A protective sensor arm cover as added to the new sondes as well, which isolates the sensor from contaminants eliminating the bias.

The work on the TOGA COARE data set has been carried out by H. Cole, E. Miller, J.-H. Wang, D. Parsons, F. Guichard and K. Beierle, all ATD/SSSF.

Multichannel Cloud Radiometer (MCR)

The

The majority of the optical components were replaced in order to provide users with spectral radiance data at requested wavelengths of interest and of higher quality than was previously possible. The MCR electronics were also modified and, in many cases, re-packaged and condensed in order to improve the overall performance of the instrument and to lower the amount of payload space required to deploy the instrument in a C-130 wing pod. Modifications to existing RAF data system hardware and software were made so that signal outputs from the seven MCR channels, as well as several additional variables containing MCR "housekeeping" data, could be recorded on the RAF Aircraft Data System (ADS-2) on the C-130.

During the SHEBA deployment in May and July of 1998, problems with microphonic noise contamination in the signals for MCR channels 5 and 7 were discovered. Following the project, the cryogenic dewar, housing the filters and detectors for channels 4 and 5, was removed from the instrument, and new sets of filters and detectors -- mounted in a metal housing and maintained at a temperature of approximately 25 degrees C -- were installed. The dewar housing the channel 7 detector and filter was also removed and sent back to the manufacturer to modify its internal electronics before the instrument was deployed again for INDOEX.

Other critical MCR support tasks undertaken in 1999 included post-INDOEX calibration of the instrument at Los Alamos National Laboratory and the creation of MCR software to process both the SHEBA and INDOEX MCR data sets. A library of IDL routines for displaying and perusing processed MCR data was created and has been made available to Mark Tschudi and Jim Maslanik at CU in an effort to provide them with tools to use in their analysis of SHEBA MCR data. At present, more development work is planned to make this suite of MCR IDL routines more robust and useful for the analysis of MCR data.

Thanks to the efforts of Krista Laursen, John Cowan, Mike Spowart, and Chris Webster (all ATD/RAF), the instrument now generates fully calibrated, aircraft pitch- and roll-corrected spectral radiometric data for a variety of cloud and radiation research.

Radar Echo Classification Algorithm

Because polarimetric measurements are sensitive to particle size, shape, orientation, phase (liquid or solid), and density (wet, dry, aggregates, or rimed), particular hydrometeor types have characteristic signatures. The ensemble of measurements and computed quantities such as the standard deviations of velocity, differential reflectivity, and differential phase, can be used to designate the type of dominant scatterer at each measurement location. An algorithm was developed at ATD which classifies radar returns as non-meteorological (ground clutter), biological (insects, birds), and meteorological. There are 14 different meteorological particle classifications identified by the algorithm.

Radar signatures for the designated classifications often overlap; consequently, a "fuzzy logic" approach to echo discrimination has been adopted. The methodology employs "membership functions" to determine the degree to which a particular radar parameter (radar reflectivity, differential reflectivity, etc.) belongs to a particular type classification (rain, hail, wet snow, etc.). The shape of the membership function for each radar measurable is based on experience gained from numerous simulations and observational studies. The membership function output is a number "P" which varies from 0 to 1. For example, hail because of its large size, associates with high radar reflectivity. The membership function for radar reflectivity in hail classification assigns P a value of 0 for a reflectivity value < 45 dBZ. The value of P increases linearly from 0 to 1 as reflectivity increases from 45 to 50 dBZ. Reflectivity values greater than 50 dBZ are assigned a value of 1. The P values of each radar parameter are weighted and summed for all of the particle classifications. The final algorithm classification is the category with the maximum weighted sum.

An example of classification is shown above. The data were taken during the CASES97 experiment near Wichita Kansas. Data shown are range height plots of reflectivity, differential reflectivity and the particle identification for a June 13 hailstorm. Several data sets have been acquired with in-situ aircraft probe data for verification of the particle identification results. An example from PRECIP98 in stratiform rain with a well defined bright band is given in this figure. Particle images from the aircraft are shown at two different levels; the melting level (0 deg C) and at -18 deg C. Notice the smooth, rounded edges of the particle probe images (indicating melting) from the level identified as wet snow by the radar. The probe images from the higher level clearly show dry aggregates (sharp irregular edges) in the region identified as dry snow by the radar.

The fuzzy logic approach ensures that particular classifications are insensitive to the details of the membership functions. The precise shape of the original membership functions and the weights applied to each measured or computed quantity will be tuned and adjusted as additional data sets from a variety of storm types are acquired and examined. Currently ATD has obtained numerous in-situ aircraft micro-physical measurements to compare with the results of the algorithm. These data sets, as well as ground observations, will be useful to verify the various particle classifications and to fine tune the membership functions.

The particle identification development is a collaborative effort between NCAR (J. Vivekanandan, S. Ellis, R. Oye), NOAA/NSSL (D. Zrnic, A. Ryzhkov) and the University of Oklahoma (J. Straka).