Extended Abstract (788K)P5R.3
Biases in surface reference estimates by the TRMM PR standard algorithm
Shinta Seto, National Institute of Information and Communications Technology, Tokyo, Japan; and T. Iguchi
The standard algorithm for the TRMM Precipitation Radar (TRMM/PR) applies the combination of surface reference technique (SRT) and Hitschfeld-Bordan method to estimate path integrated attenuation (PIA). SRT uses the surface backscattering cross section (σ0) to calculate the first estimate of PIA (PIA_SRT) as PIA_SRT=σ0_ref-σ0_obs, where σ0_obs is the observed σ0, and σ0_ref is the estimates of actual σ0. To give σ0_ref, SRT does not explicitly assume the change of σ0 between under rain condition and under no-rain condition, though the increase of land surface soil moisture and sea surface wind speed may change σ0 significantly.
This study used the TRMM/PR standard product version 6 and statistically analyzed following variables.
σ0_NR σ0_obs under no-rain conditions
σ0_R σ0_obs under rain conditions
σ0_ref the first estimates of σ0_e by the SRT
When the rain rate is weak and PIA is small, Δσ0_R=σ0_R-σ0_NR often takes positive value. The positive Δσ0_R can be seen in less vegetated area (over land) and when the incident angle is relatively large (over ocean). This can be explained as the effects of the changes in soil moisture and sea surface wind speed. SRT over land applies one of two references, spatial reference and temporal reference. Spatial reference takes the samples of σ0_obs under no-rain conditions from the adjacent pixels, where the surface condition is statistically similar with that in the rain area. Therefore, when spatial reference is applied, PIA_SRT is less biased. On the other hand, temporal reference takes the samples from the previous month observations in the same 1 x 1 lat/lon degree grid box. When temporal reference is applied, PIA_SRT can be negatively biased. Over ocean, new hybrid spatial reference method is introduced from the version 6. This method uses a quadratic to express the relationship between σ0 and the incident angle. This method can increase the number of samples, but the positive biases in PIA_SRT appeared when the incident angle is around 12 degree.
Poster Session 5R, TRMM/ GPM studies and algorithms
Tuesday, 25 October 2005, 6:30 PM-8:30 PM, Alvarado F and Atria
Previous paper Next paper