The objective of this study is to find out the reason for the biases which exist in the cloud heights of Atmospheric Motion Vectors (AMVs) created by EUMETSAT. A cloud detection and cloud property retrieval tool, referred to as KLAROS (KNMI's Local implementation of APOLLO retrievals in an Operational System), is used to diagnose the biases in the representative heights of the AMVs. The WMO network of radiosondes are used to map the KLAROS cloud temperatures into independent pressures and wind vectors. From a set of 40 NOAA overpasses, an in depth analysis of the cloud fields for 8 NOAA overpasses is performed.
Comparison between the independent wind vectors and the EUMETSAT AMVs lead to the identification of three sources of error.
The first is that the semi-transparency flagging for EUMETSAT of ten occurs under the wrong cloud conditions. The second is that, when the semi-transparency flag is set, the AMV temperatures are weighted too beavily by the water vapor channel brightness temper-atures. The resulting cold temperature frequently raises the altitude of the AMVs into a region of maximum wind speed, causing a bias in wind speeds. This depends on the structure of the cloud top boundaries being observed. The third is related to the METEOSAT viewing geometry of the northern latitudes. A consistent temperature difference of 5K was seen between METEOSAT infrared temperatures and nadir viewing AVHRR temperatures for optically thick cloud fields.
For clouds traveling at altitudes within the wind shear profile, a bias of 5K can translate into a difference of 30hPa. This implies that the EUMETSAT cloud correction algorithm should be tuned differently for Northern latitudes.
The first source of error can be removed by a re-evaluation of the EUMETSAT criteria for semi-transparency flagging. Removing the semi-transparency correction caused a significant reduction in the wind speed bias for 7 out of the 8 cases. It is suggested that an improved definition of the structure of cloud top boundaries and re interpretation of the water vapor signal can reduce the second source of error. The effect of viewing geometry on the wind speed bias is related to the second error and has been demonstrated in one case.
R. Dlhopolsky and A. Feijt. An investigation of the representative heights for atmospheric motion vectors
KNMI number: TR-249, Year: 2003, Pages: 32