Checking the cross-track dependence of the TRMM radar echoes nearest to the earth surface

The first ever space-borne weather radar was launched in November 1997 onboard the Tropical Rainfall Measuring Mission (TRMM). Satellite observations offer a unique chance for objective worldwide climatological studies. The successful results obtained by the TRMM satellite during the first three years put in evidence the need and opportunity of increasing the 4-year planned life of the mission. In response to this need, in August 2001 the satellite orbit was successfully raised from 350 km altitude to 402 km altitude and it is still operational. Consequently, the 3 dB footprint of the radar antenna at Nadir have decreased from a 4.3 km diameter to ~5 km. Because of the increased range, also the radar sensitivity has decreased. However, the large data set acquired during the “long” TRMM life permits climatological validation of the TRMM products. In this paper our focus is on the attenuation corrected reflectivity radar echoes nearest to the Earth's surface, the so-called NearSurfZ within the 2A-25 data product. Why have we thought that a climatological validation of this valuable data set could be of interest? Because the vertical resolution of the TRMM Precipitation Radar (TPR) samples becomes poorer with increasing distance from the nadir. Since the vertical reflectivity profile is not constant, rather highly variable, a systematic difference between echoes at Nadir and off-Nadir could be present. On a climatological basis, the vertical reflectivity profile shows, on average, a negative gradient with increasing altitude. As a consequence, our guess is that an apparent decrease of sensitivity of the TPR could show up with increasing distance of the TPR samples from the nadir. For this reason we have acquired and analyzed a total of 4000 TRMM orbits, 2000 before and 2000 after the boost in August 2001. This means approximately 900 millions NearSurfZ echoes before and after the boost respectively. Obviously, most of these echoes are not weather echoes. Nevertheless, in view of the large data set analyzed, the number of “valid” echoes (i.e. the number of echoes above the minimum detectable signal, MDS) is still remarkable: there are, in fact, 32,565,344 valid echoes before the boost and 28,377,436 valid echoes after the boost. Our climatological analysis to check the presence of an unwanted cross-track dependence, is based on two steps: 1) firstly, by averaging on a linear scale all the radar reflectivities values in the 24 TRMM off-nadir tracks parallel to the nadir line (both to the left and to the right of the Nadir track); 2) by then analyzing through a non linear regression, the obtained average reflectivity in each of the 24 off-nadir tracks versus the off-nadir distance. As expected, the average reflectivity closest to the nadir is found to be larger than the farthest one. However, this difference is as small as 0.4 dB. We may conclude that the apparent decrease of sensitivity, when all the valid echoes are considered, is less than half a dB. This is another clue that the many efforts made to provide a good calibration of the TPR have been successful. Nevertheless, the apparent decrease of sensitivity increases if a threshold larger than the MDS is used: with a “hydrological” threshold of 25 dBZ (~1 mm/h), this difference is ~0.6 dB. If the threshold is raised further towards high rainfall rate, the difference become more significant: it is found to be ~1 dB, with a 34 dBZ (~4 mm/h) threshold and ~1.4 dB with a 40 dBZ threshold (mainly convective rain with intensity larger than ~10 mm/h). These figures are valid both before and after the boost, although in the latter case the cross-track dependence is slightly less pronounced (for all the 4 thresholds). By plotting the overall average reflectivity versus the cross-track distance using a Log-Log scale, it is evident that the apparent decrease of sensitivity can be statistically explained through a regression: A) in case of a 25 dBZ threshold, the variance explained by a linear fit before (after) the boost is 68% (66%). B) in case of a 40 dBZ threshold, the variance explained by a linear fit before (after) the boost is 80% (75%). By increasing the threshold, a parabolic dependence appears to be more evident: the variance explained is as high as 96% (92%) before (after) the boost. We conclude that it is worthwhile to compensate the decreasing vertical resolution of the TPR across-track. The correction factor is based on the coefficients tuned in the present analysis based on 4000 orbits. Up to ~1 dB is added to TPR echoes at 120 km off-nadir with respect to those whose distance is only 5 km from the nadir line.