SMOS SSS retrieval : sky & optical thickness effects

 

  1. Purpose and methodology

In the process of scanning the sources of errors, this is a first approach to the influence of parameters which have been left aside so far :

The methodology consists of simulating retrievals and assessing the impact of noises and biases over the parameters of interest.

Throughout this note :

The incidence angles are specified for each case.

 

2. Sky temperature effects

The incidence angle range taken here consists of 16 values spreaded between 20 and 41. Only SSS is retrieved. The unperturbated value for TY is 2.7 K. See Table I

Table I

 

Configuration

TB noise (K)

SSS s (PSU)

SSS bias (PSU)

 

No perturbation

# 1

0.43

0

 

0.5 K noise on TY

# 1

1.02

0

 

0.5 K bias on TY

0

0

0.86

 

Comments.

The unperturbated SSS s value corresponds to earlier simulations for a 283 K SST value (i.e. a rather cold sea), with the same incidence angles coverage. The results shown hereabove are consistent : 1.02 is near the square root of (0.432 + 0.862).

Outside the galactic region in the sky, what is the estimated amplitude of fluctuations in the cosmic background temperature ? Provided it is definitely less than 0.1 K, then the incertainty on TY will not be a major contributor to the overall s (Remember this is noise, therefore the variances add together ; a 0.1 K s over TY would bring in a s of about 0.86/5 ; the 0.43 unperturbed s figure would then simply climb to 0.47, to be next narrowed down by averaging).

Concerning the galactic source, the problem is much worse because, inasmuch as the sky noise level is not well known, it should be considered as a bias, which cannot be averaged out. Then Table I tells us that, in order to stay reasonably within the 0.1 PSU required bracket, the TY bias needs to be known within a few tenths of a K.

This obviously requires further elaboration, accounting for the way anisotropic sky temperature is reflected upwards when the sea is rough.

 

2. Optical thickness effects

Over the surface-to-satellite line of sight, the steady compoent of optical thickness t is due to molecular oxygen. For L-Band, this is very small and amounts to about 0.007 Neper (for a nadir view). Further contributions to t are provided by water vapor and liquid water (rain) : they should be significantly smaller than the clear air compoent, except for very heavy rain. Finally, one can wonder whether foam ought not to be considered as a source for increased optical thickness.

The situations over sea and land are quite different. Over land, the major contribution by far to optical thickness is due to vegetation water content, and it is a quantity of interest. It can be quite large (until unity or above). Over sea (the case of interest here), while t is much smaller and appears as a pure correction (there is no a priori scientifc objective associated to determining optical thickness !), it cannot be neglected, since the measurement of SSS is so sensitive to every perturbation.

Various (very far from exhaustive) retrieval simulations have been carried out and are reported on Table II. The reference is the same as in table I.

 

Table II

 

Configuration

Noise

on t

(nepers)

Bias

on t

(nepers)

TB

noise (K)

s (SSS) (PSU)

SSS bias (PSU)

s (t) (nepers)

1

No perturbation

   

# 1

0.43

0

 

2

Noise on t

0.001

0

# 1

1.22

0

 

3

Bias on t

0

0.001

0

0

1.09

 

4

(SSS + t) retrieval

0

0

# 1

2.27

0

0.0019

5

Same as 4 but for i angles between 20 & 21.5

0

0

# 1

6.40

0

0.0061

6

Same as 4 but for i angles between 45 & 46.5

0

0

# 1

1.29

0

0.0009

7

Same as 6 but t =0.0163 io 0.0071

0

0

# 1

1.36

0

0.0009

8

Same as 7 but s (SST) = 0.5 K

0

0

# 1

1.37

0

0.0009

Comments

From lines 2 and 3 it is seen that, for the chosen value for the SST, keeping in mind the 0.1 PSU bracket target, it is desirable to know t with an accuracy better than 0.0005 Neper. For clear air, this means estimating the column molecular oxygen content within 1%, water wapor probably within 10%. Here we are talking about biases ; the required performance is probably within reach using meteorological analyses.

The consequences for liquid rain remain to be estimated : cases where attenuation due to rain for L-Band, integrated over a pixel, is in excess of 0.0005 Neper, are likely to be unreliable.

From line 4 on, the retrieval process is carried out both for SSS and t. It is seen (line 4) that the accuracy over SSS is considerably degraded. From lines 5 & 6, it appears that the incidence angle range does not have relevance here ; what matter is the magnitude of the incidence angle. In other words, it is the dual polarisation capability, mainly efficient for high i angles, which is at play here.

Line 7 shows that the magnitude of optical thickness (provided it remains small) does not change things much ; line 8 shows that adding a state of the art noise on the SST induces no further degradation.

Note that, while for lines 6-8 the measurement performance on SSS is worsened by a factor 3 or slightly above with respect to cases where optical thickness t is perfectly known, it turns out that the accuracy on the retrieved t value is not so bad. It is a pity that we do not need it...

 

Conclusions

On the basis of these results, it is obvious that finding external ways to estimate the optical thickness correction over sea is a better idea than retrieving it from the SMOS data. The requirements are not easy but do not seem impossible to meet (also, keep in mind that everything becomes easier for warm seas). Obtaining a better assessment of rain effects is among the next tasks.

A final comment concerns the way that the versatibility of SMOS is used over sea. Concerning the SSS, it has been established that neither dual polarisation nor multiangle capability are essential in themselves : they simply help to obtain a larger number of independant samples to be thrown into the averaging process. The dual polarisation capability, however, has discriminating power as far as several correcting factors are concerned : Faraday rotation, and optical thickness as just seen. It remains to explorate whther the multi incidence angle capability has discriminating power concerning the sea roughness.