|Algorithm development PI : Dr. Guosheng Liu (Florida State University)|
This document briefly describes some caveats for the AMSR-E Level 2 standard products (Level 2.0) of precipitation.
Beam filling correction was optimized and rain threshold table was replaced.
Combinations of both emission and scattering signatures are used in retrieval algorithm. The algorithm retrieves rainfall over ocean and land areas except for the following surfaces: coastal (~25 km from coastal line), sea ice, snow-covered land, and desert areas. Separate algorithms are applied for over ocean and over land regions. Generally, retrievals over ocean have better quality than those over land. The sea ice flag is based on sea ice concentration retrievals from AMSR(-E) provided by the EOC integrated retrieval system. Snow-covered land and desert surface detection is based on AMSR brightness temperatures and embedded in the precipitation retrieval algorithm. In the Version 1.0 product, there were some unexpected missing data areas due to inappropriate rain threshold table. The table was modified and replaced in the Version 2.0.
1. Based on preliminary comparison with the AMeDAS radar-raingauge match-up data and with TRMM PR data, it appears the algorithm overestimates rainfall rate by ~10%, mainly during cold seasons. We are still investigating the overestimation and will try to correct it once it is better quantified.
2. Two types of precipitations ? warm rain over land and snowfall ? may not be correctly detected by the retrieval algorithm due to the insensitivity of AMSR channels to these precipitation types.
3. Over land retrievals at high latitudes are noisy due to frozen and/or melting surfaces. Cold brightness temperatures at 89 GHz for these surfaces are occasionally interpreted by the algorithm as rain.
The algorithm uses the following ancillary datasets provided the integrated retrieval system: land-ocean flag, surface air temperature (GANAL) (to determine freezing level height), and sea ice concentration.
The algorithm is identical to that of AMSR (Version 1.0).