Issue 7 - Profiling of the lower troposphere

Agree.

Refractivity is a useful climate benchmark, but separate benchmarks for temperature and humidity is more useful. There would be no need for GPS-RO if a globally distributed benchmark for humidity and temperature existed for the lower troposphere, which functioned under all weather conditions. Conversely, separate benchmarks for humidity and temperature are useful, even if not globally distributed or all-weather. I do not know how to make a definitive statement that suggests we need one or the other. If we can achieve both, that's better.

One way to gauge the utility of measurements is that they lead to improved climate forecasts. The additional cost of each observing system must be weighed against its benefit in improving climate understanding and prediction.

Response from Ben Ho

Kuo et al (2004, link is provided in the suggested reading list) demonstrated that RO soundings (refractivity profiles) are of sufficiently high accuracy to differentiate performance of various types of radiosonde.

In my AMS slides 23, 25 and 26 (linked above), I have also quantified
1) the refractivity sensitivity to water vapor
2) the impact of NCEP TW and ECMWF TW to 1D-var TW, and
3) about 2.2 K of temperature bias will lead to 0.5 (g/kg) water vapor bias

To see if
1) COSMIC RO refractivity at lower troposphere is of sufficiently high accuracy to differentiate the performance of different types of radiosonde, and
2) the quality of RO soundings is independent of geographical location,
3) 1D-var temperature and WV at lower troposhere also useful to further differentiate the temperature and WV performance of different types of radiosonde ?

I conduct similar experiments as Kuo et al (2004, statistics are mainly from 5 km to 25 km) but using COSMIC refractivity, temperature and water vapor profiles (from 1D var). Here the statistics are calculated from surface to 25 km for temperature and refractivity and from surface to 10 km for water vapor.

COSMIC, Radiosonde, and ECMWF refractivity, temperature and water vapor profiles at different regions are compared.

Figs. of Detail Response
Detail Response

Results show that
1. In general, we did not find significant variation of the quality of the RO soundings over different geographical areas. This is evidenced by the relatively small variations (in terms of absolute fractional difference) in the RO and ECMWF differences between geographical areas.

2. COSMIC-Radio refractivity (N) bias is very large over China, where we didn't find that at different geographical areas. This N bias cannot be explained by temperature bias between COSMIC-Radiosonde and COSMIC-ECMWF, but was identified in the water vapor bias over China. The magnitude of N bias is consistent to the water vapor bias over China.

Results here seems demonstrating the potential usefulness of COSMIC 1D-var water vapor as a baseline humidity monitoring network for the lower troposphere.

Consolidated Comments from Gutman, Yoe and Reale

Seth Gutman's Response
Because of the points discussed previously, GPS-RO will probably play a secondary but still crucial role in a baseline monitoring network for the lower troposphere by providing independent validation/verification of benchmark measurements. One way this might be accomplished is by assimilating GPS-RO refractivity measurements into atmospheric models of the lower troposphere, and using the adjusted model background to quality control temperature and moisture soundings. I do not believe that the need for a reference radiosonde network will ever be eliminated by inherently ambiguous remote sensing observations and/or the retrieval of geophysical parameters based on these observations.

Jim Yoe's Response
I agree - as implied in my comment on issue 6. As I read this I get a sense that we need to agree that a benchmark measurement does not mean the ONLY measurements we need, but one that is needed to help sort out the others, at least some of which are absolutely necessary as well to address the questions of interest.

Tony Reale's Response
Reference networks are needed primarily to sort out (discriminate) the respective error characteristics of "critical" measurements that are used to monitor global (ie GCOS) climate (measuring climate trends at reference sites is secondary...). In this respect COSMIC becomes another discriminating parameter collected at reference sites, it does not replace a given climate parameter nor the requirement for reference sites. (As with polar satellites, I would recommend that a subset of in-situ reference site measurements be synchronized with COSMIC.

Response from Jens Wickert

With OpenLoop tracking on COSMIC already progress was reached in the LT compared to previous missions, but definitely there is potential for further improvements to get precise refractivity close to the ground (better receivers and antennas with better SNR characteristics, new navigation signals with advantageous properties for RO, see also my remarks on issue 3).

But the ambiguity between dry and wet contribution to refractivity is a fact. Respective separation only can be done by additional assumptions or data. Radiosonde and other measurements will be required also in future.

Simplistically, because of the refractivity equation and the Clausius Clapeyron equation which limts the amount of water vapor, refractivity is quite useful for estimating bulk atmospheric density, pressure and temperature versus altitude at temperatures below 230 to 240K in the troposphere. At these temperatures, water vapor does not affect the interpretation very much yielding very accurate results.

At temperatures warmer than 240K, and given claimed 1-sigma accuracies from global temperature analyses of ~1-1.5K, the refractivity information yields accurate water vapor profiles when combined with analysis temperatures. A water vapor variable directly estimated by the GPSRO is specific humidity. The expected 1 sigma accuracies of individual profiles are predicted to be about 0.2 g/kg up high, deteriorating to 0.5 to 1 g/kg in the lowermost troposphere depending on how accurate the refractivity profiles are in the lowermost troposphere (see Kursinski and Hajj, 2001 and Kursinski et al., 1995). Assuming refractivity biases are very small and analysis temperature biases are no more than 0.5 K, the bias in the GPS-derived specific humidity will be less than 0.1 g/kg. Achieving still smaller biases depends on the temperature bias (assuming the refractivity is unbiased).

These water vapor profiles are a very unique product because the vertical resolution of the GPSRO is so high compared to nadir viewing radiance measurements and because the GPS signals penetrate clouds yielding the most unbiased vertically resolved global sampling presently available. As such it must be important from a climate standpoint BUT it is not a benchmark measurement.

The simplest method of generating humidity profiles is done by combining refractivity from GPSRO and temperature from analyses. This combination provides the minimum information needed. Because of the reliance on analysis temperatures, this humidity is not a benchmark measurement. However, analysis temperatures are quite accurate in part because IR and microwave radiances from CO2 and O2 emission lines measured by satellites are far more straightforward to interpret than radiances associated with water vapor because in the case of well mixed gases like O2 and CO2, the weighting functions are largely known before assimilation into the analysis. For a gas like water vapor, the vertical distribution of the gas concentration is much more variable and uncertain and the analysis relies far more heavily on the initial guess about the water vapor field that comes from the forecast. This is why I am always leery about moisture biases in the 1DVar results resulting from the reliance on the model to the distribute the water vertically when relatively coarse vertical water vapor radiance constraints are assimilated into NWP analyses. As I have said before, for climate, I personally prefer the simple approach (refractivity plus analysis temperature) to generating water vapor profiles from GPSRO where only the temperature information from the analyses is utilized rather than including both the temperature and humidity information from the analyses in combination with the GPSRO refractivity as is done in 1DVar.

I also note that under condition when super-refraction occurs typically at the PBL top at low latitudes (where the very large vertical water vapor gradient becomes so large that vertical gradient of refractivity becomes so large that the radius of curvature of the raypath becomes smaller than the radius of the Earth) we have an ambiguity below that altitude where a continuuum of refractivity profiles are consistent with the same observed bending angle profile. As many of you know, we developed one approach to deal with this where we reconstruct a continuum of refractivity profiles consistent with the observed bending angle profile and then we need at least one extra constraint such as surface refractivity to select the optimum refractivity profile from the continuum. So, under conditions where super-refractivity occurs, the portion of the refractivity profile within the boundary layer below the height of the super-refraction is ambiguous and needs an additional constraint to resolve the ambiguity.