SeaWinds on QuikSCAT Known Problems

Level 2B Usage Notes: 23 July 1999 - present; revs 430-present   

The Level 2B wind vectors have been processed with the new QSCAT-1 model function, developed by M. Freilich and B. VanHoff. The mean model coefficients [A(0)] have been tuned specifically to induce a mean speed bias relative to numerical weather prediction winds of about +0.3 m/s, addressing the observed wind speed under-prediction problem inherent in the NWP products. The reference height for the QuikSCAT wind vectors remains at 10 meters. Furthermore, the azimuthal modulation of the QSCAT-1 model has been tuned to remove, as far as possible, systematic instrument-induced direction errors ("beam spikes" and "square circulations").
The Level 2B product includes two additional scientific data set (SDS) elements, the "mp_rain_probability" and the "nof_rain_index", to indicate the detection of rain contamination in the QuikSCAT data.
For all wind vector cells, the mp_rain_probability element contains a value between 0 and 1 based on an extended version of the Multi-Parameter rain algorithm developed by J. Huddleston and B. Stiles at JPL. Cells for which the algorithm could not compute a valid value contain the value -3. Additionally, bit 12 of the "wvc_quality_flag" indicates whether the MP algorithm computation was valid (0) or invalid (1), and bit 13 indicates whether rain was detected (1) or not detected (0) by the algorithm. Bit 14 of the wvc_quality_flag indicates whether the rain detection is based on data from both beams (0) or from outer-beam measurements only (1).
The nof_rain_index is a value ranging from 0-249, with 250 as the "invalid" value, based on the Expected Normalized Objective Function (NOF) algorithm developed by C. Mears, D. Smith, and F. Wentz at Remote Sensing Systems.
The Level 2B product includes two additional scientific data set (SDS) elements, the "wind_speed_selection" and the "wind_dir_selection". These elements contain the final output of the wind retrieval/ambiguity removal processing using the Direction Interval Retrieval with Threshold Nudging (DIRTH) algorithm.

• The reduced file size (~80%) allow 3 months to fit on a CD.
• Programs that read L3 data may run slower but need no changes.
• Programs built using HDF version 4.1r2 may crash when reading these data. Versions 4.1r3 and above fixed this bug.