SSMIS lower atmospheric sounding cal/val

DMSP's SSMIS introduces a new series of conical scanning microwave imaging/sounding instruments, integrating features of SSM/I, SSM/T-1 and SSM/T-2 and adding upper atmospheric sounding capability. Cal/Val results are reported for F-16 SSMIS lower atmospheric sounding (LAS) channels. The calibration activities described here are based on comparison between SSMIS SDRs and radiative transfer calculations performed using atmospheric temperature and water vapor profiles. Profiles were obtained from operational radiosondes, ECMWF fields, and dedicated lidar measurements.

Globally averaged comparisons for 2004, based on radiosondes and on ECMWF fields, indicate that SDRs for temperature channels 3-5 and moisture channels 9-11 are consistent with derived calibration bias requirements (+/-1K), although standard deviations are large for channels 9-11. Lidar, deployed at Barking Sands, Kauai, was used to provide high accuracy and high altitude measurements. Results from the winter 2003 lidar campaign indicated that biases and standard deviations were less than 1 K, for temperature channels 2-6. The small standard deviations indicated that SSMIS responds accurately to atmospheric change. However water vapor channels 9 exhibited a large bias, although the standard deviations were small. Independent measurements from the aircraft-based CoSMIR microwave radiometer, described elsewhere, indicated large biases for the surface sensitive channel 1 and substantial biases for 9-11. Radiative transfer calculations established that channel 1 and 2 respond to vertical, rather than the specified horizontal polarization. Based on this, instrument design implies that channels 1 to 5 are vertically polarized.

Water vapor biases changed significantly when measured in the April 2004 lidar campaign, however, again, standard deviations remained small. Analysis of time series of operational radiosonde-based and ECMWF-based comparisons, stratified for ascending and descending orbits, and limited to latitudes 15N – 50 N, indicated that biases for all channels examined change significantly on a monthly basis and that there are large differences between ascending and descending orbits. The differences correlate with solar heating of the main reflector surface, suggesting that the surface has appreciable microwave emissivity. The effects are evident in global maps of bias generated from ECMWF fields. Analyses of instrument gain and warm calibration load data led to identification of local calibration artifacts associated with solar heating of warm load tines. The final lidar campaign in December 2004 confirmed that the annual bias cycle returns reproducibly to values measured in winter 2003.

Correction algorithms were implemented to compensate for warm load heating and models are under development to compensate for bias induced by solar heating of the main reflector. Retrieval software was modified to account for vertical polarization. As a result of Cal/Val efforts, SSMIS SDRs will be brought into compliance with derived requirements. Minor instrument and software modifications will be implemented to mitigate bias in follow-on missions.

Validation of LAS retrieval products is also complete. Temperature soundings, up to the 500 mb level, meet bias and RMS requirements for the three atmosphere types used in temperature retrievals. Above that, bias oscillates, exceeding requirements for some levels, depending on type of atmosphere. Relative humidity retrievals generally fail requirements, as judged by radiosondes and ECMWF averaged over all locations. Global comparisons with ECMWF suggest that moisture retrieval performance varies strongly with the retrieval atmosphere type. When measured at the tropical Barking Sands lidar site, RH retrievals meet requirements at 1000 mb and approach requirements at and above 700 mb. All reference methods indicate that the less stringent specific humidity requirements are satisfied. We anticipate that it will be possible to bring temperature retrievals into compliance with requirements by adjusting retrieval parameters and implementing instrument modifications. Bias corrections may improve RH retrievals.