The ESA-ANISAP Study: Retrieving Integrated Water Vapor Profiles Through Differential Attenuation Between LEO Satellites
Luca, Facheris¹; Fabrizio, Cuccoli²; Susanne, Schweitzer^3
1Department of Information Engineering, University of Florence, ITALY; ²CNIT-National RaSS Laboratory, ITALY; ³Wegener Center for Climate and Global Change, AUSTRIA

The idea of exploiting the normalized incremental ratio of the spectral attenuation for estimating the atmospheric water vapour content along microwave paths was first proposed in 2001 for vertical links. In 2002, the first two authors proposed the Normalized Differential Spectral Attenuation (NDSA) concept for sounding tropospheric water vapour by means of a couple of counter-rotating LEO (Low Earth Orbit) satellites (one carrying a transmitter, the other a receiver) operating in the Ku/K bands and in a limb geometry. In those years, during the ACE+ mission studies in the frame of the second call for proposal of the ESA Earth Explorer Opportunity Mission, the problem arose of the severe impact of scintillation due to tropospheric turbulence on the water vapour estimates provided by radio occultation measurements made at those frequencies in limb mode between two LEO satellites. Thanks to its normalised incremental approach, NDSA appeared as an interesting method for limiting such impact. In fact, NDSA relies on the conversion of a spectral parameter (the spectral sensitivity S), into the total content of water vapor (hereafter IWV, Integrated Water Vapor) along the propagation path between the two LEO satellites. To measure S at fo it is required that two tone signals at relatively close frequencies f1 and f2 (f1 > f2), symmetrically placed around fo are simultaneously transmitted. At the receiver, the two pertinent powers P1 and P2 are simultaneously measured so that S is provided as the estimate of the derivative of the spectral attenuation at fo , normalized with respect to the attenuation at fo itself.

From the very beginning of the NDSA studies, it was evident that in ideal measurement conditions (namely, no disturbance at the receiver nor propagation impairments), S measured at different frequencies was tightly correlated to the IWV along tropospheric propagation paths connecting two LEO satellites. Specifically, it was seen that the S values measured in different atmospheric conditions at the lowest and highest frequencies were correlated to the IWV values at the lowest and highest tropospheric layers, respectively. The idea was to account for natural variations of the atmospheric conditions by using profiles of temperature, pressure an water vapor coming from real radiosonde data, and assuming a spherically symmetric atmosphere. IWV was thus computed along the radio path, S simulated separately based on the Liebe atmospheric propagation model and finally the IWV-S relationships at various altitudes obtained. These could then be used as empirical relationships to directly convert S measurements into IWV estimates. The tight relationship between S and IWV at different tangent altitudes in the frequency range 17-22 GHz was verified through simulations based on a wide set of radiosonde data and this appeared as a further stimulus to investigate the other aspects of the applicability of the NDSA method to the link between two counter-rotating LEO satellites. This triggered an independent ESA study (AlMetLEO) that yielded a significant insight into the S-IWV relationships at various altitudes (up to 12 km) based on an extended radiosonde data analysis, into the modelization of signal fluctuations due to tropospheric turbulence and - finally - into the accuracy achievable under given SNR and scintillation levels. Also, an approximate closed-form relationship for the S accuracy valid for high SNR levels was found out and the 30 GHz channel was identified as a possible candidate to detect the presence of liquid water in the LEO-LEO link. More recently, the ESA-ACTLIMB study offered the opportunity to investigate other aspects of the NDSA technique, opening further perspectives for its utilization in a multi-frequency context. In particular, the potential of the frequency band from 179 to 182 GHz (M band) for estimating IWV at the highest tropospheric altitudes (from 10 km upwards) emerged clearly, showing a significant robustness to scintillation fluctuations.

Such past activities, however, have left open some issues that are currently faced by the on-going ESA-ANISAP study, jointly with new ones that deserved to be treated in order to get the complete theoretical understanding of the matter and of the NDSA potential. These are:
1) to finalize the analysis about the accuracy of the IWV-S relations and of the IWV estimates in the upper troposphere (above 10 km) by exploiting multiple NDSA channels in the K-Ku and M band;
2) to focus on the potential of NDSA channels in the 27-36 GHz band for estimating the Integrated Liquid Water (ILW) and correcting the IWV estimates made in the Ku-K bands;
3) to develop appropriate inversion methods to convert the IWV estimates into WV estimates;
4) to devise an algorithm for the derivation of turbulence parameters (vertical profiles of the refractive index structure function and, possibly, value of the outer scale length) from high resolution atmospheric profiles; 5) to understand the potential NDSA products in the case of two or more co-rotating satellites

In this paper we synthesize the results of the first three issues, while the other ones are addressed in companion papers submitted for presentation at this Living Planet Symposium. Indeed, two basic problems affected the reliability of the empirical IWV-S relations found so far in all bands examined: the first is the fact that the accuracy of the radiosonde data used to derive them were not uniformly distributed in the northern and southern hemisphere; the second is the limited number of radiosonde samples available at the highest altitudes (above 10 km), and their scarce reliability. The second limitation, evidently, affected the reliability analysis of the IWV-S relations specifically in the M band. To overcome both problems, instead of resorting to radiosonde data we utilized globally equally distributed atmospheric profiles (19728 profiles in total) which were derived from ECMWF atmospheric analysis data. Such database includes pressure, temperature, humidity, liquid water, ice water and wind components in 8 global datasets in 4 days amid of each of the four seasons and at two time layers (12:00 UTC and 24:00 UTC). Based on it, we were able to carry out the global scale analysis of the IWV-S relations up to 20 km tangent altitude and to determine which frequencies in the 27á“33 GHz interval are more convenient to provide effective means to detect the presence of liquid water in the LEO-LEO link and to directly correct for the S estimates in the Ku/K bands for an accurate estimation of IWV that would otherwise be biased by an unknown amount. Examples will also be shown of NDSA performance accounting for signal impairments, including scintillation effects based on realistic atmospheric simulation relying on a high resolution radiosonde dataset and the procedure illustrated in one of the aforementioned companion papers.