Description: | Occasionally, the retrieved relative humidity was exceeding 120%. Upon reviewing the data
it was noticed that the high values were appearing in the highest layers (above 8 km) and
especially during spring season. Since retrievals at the highest levels are mainly
affected by the climatology used to constrain the retrievals, we reviewed the statistical
coefficients that were used to retrieve temperature and humidity. It was discovered that, in
the training dataset (radiosonde), there were a few outliers (invalid radiosonde
soundings) that had escaped the screening process and were affecting the computations of the
retrieval coefficients.
We recomputed the retrieval coefficients with the newly screened set of radiosonde data
and reprocessed all the previous data at Niamey. The reprocessed data were archived in
August 2006.
The new coefficients improve temperature and humidity profiles in two ways. The relative
humidity will not exceed 120% in the upper layers and the temperature profiles will be in
better agreement with the sonde in the first 4 km. |