During the NASA's Tropical Rainfall Measuring Mission (TRMM) Field Campaigns, Large Scale Biosphere-Atmosphere (LBA - a 4 station network) held in the period Jan-Feb 1999, and during the Kwajalein Experiment (KWAJEX - a 5 fixed- and 1 mobile station network) operational from July 23 through September 13, 1999, about 2,300 radiosonde ascents were made. One of the main objectives of these soundings were to obtain accurate estimates of sensible heat source (Q1) and apparent moisture sink (Q2) budgets with a higher degree of accuracy. As a first step all data sets were converted to a single format (LBA format), a consistent file-naming convention was adopted, and an inventory was generated for each of the sounding station. Ground check logs, calibration, equipment, location and surface data were incorporated into each sounding data file. The radiosonde data were obtained from three different manufacturers viz. MSS, VAISALA, and VIZ. Since the Pressure Temperature and Humidity (PTU) and wind detection methodology were different for these three different manufacturers, as a first step all raw data (independent of the sonde manufactuerer) were subjected to limit checks based on algorithm defined and well tested for TOGA - COARE experiment (Loehrer et al. 1996). A VIZ-Vaisala comparison for the humidity profiles observed over LBA station is presented. A brief discussion on the reliability of these data sets in terms of the sensor response characteristics, noise, and methods adopted for data repair are mentioned in this paper. The results of comparing the radiosonde data with a tethersonde (within the atmospheric boundary layer) and related statistics are also discussed for the near-surface corrections that would possibly be adopted. Results of an assessment performed on the accuracy of Q1 and Q2 estimates using the repaired soundings data sets is also mentioned in this paper.