The spectral response Table of PACE HARP2 includes negative values

[lq17865155322]

I obtained the spectral response function for PACE HARP2 from the website: https://oceancolor.gsfc.nasa.gov/data/pace/characterization/.

I have the following doubts:
1) In the rsr for each band, I understand that -99 represents a missing value, but what do negative values other than -99 signify?
2）What does the last column ‘GLAMR_power’ in the table refer to?
3) How to convert RSR into a dimensionless spectral response function with a total integral of 1?

Thank you for your time and assistance.

[OB General Science - guoqingw]

Hello, the following are the answers to your questions. Hope it helps.

1) Negative values other than –99

These are observed negatives that passed the initial filtering. If a value exceeded the 3-sigma uncertainty, it was retained as potentially real (even if negative). Values within 3-sigma were masked as likely random variation. This way, users can decide themselves whether to keep or ignore the negatives.

2) ‘GLAMR_power’ column

This records the power measured by the GLAMR reference detector during HARP2 acquisition. HARP2 counts were divided by this value to correct for source power fluctuations. It’s kept in the file so users can, if needed, reconstruct the original counts or apply their own corrections. It is not HARP2’s own radiometric measurement.

3) Converting RSR to unit-area (integral = 1)

The file does not provide area-normalized RSRs. Instead, each band (red, green, blue, NIR) has a peak normalization factor so that the maximum in-band response is 1. To normalize yourself:
(a) Apply the 3-sigma uncertainty filter (see #1).
(b) Divide by GLAMR_power (see #2).
(c) Divide by the appropriate band’s normalization factor (based on wavelength).

This peak-based normalization was chosen to simplify cross-talk evaluation (e.g., blue filter leakage into NIR). If you need a dimensionless SRF with integral = 1, you’ll need to apply your own area normalization over the wavelength range after these steps.

Guoqing

[lq17865155322]

Dear Guoqing,

Thank you very much for your reply!

Furthermore, I would like to ask whether the PACE science team has assessed the out-of-band effects of HARP2 on the inversion of aerosol and other products?

Best regards,

Qiang

[madelyn1]

As a fellow user — good questions. For the PACE HARP2 spectral response table:
Yes, -99 indicates missing data, while small negative values usually result from numerical interpolation or rounding noise and can safely be set to 0 before normalization.
The GLAMR_power column represents the measured signal strength (optical power) from the GLAMR calibration instrument and isn’t part of the RSR itself.
To convert the RSR to a dimensionless, normalized response, simply set negatives to zero and divide all RSR values by the sum (or integral) across wavelengths so that the total equals 1.

[mgao1]

Dear Qiang, could you elaborate more on the out-of-band effects?

The current HARP2 aerosol product available on earth data cloud (FastMAPOL suite) only consider the response functions characterized by the band center and band width when conducting the radiative transfer calculations.

I can see small out of band behavior of the blue band (more than 2 orders of magnitude smaller than the center band signal as in attached plot). The impacts of this behavior could be small, but are not quantified yet as far as I know.

Meng



lq17865155322 wrote:
> Dear Guoqing,
>
> Thank you very much for your reply!
>
> Furthermore, I would like to ask whether the PACE science team has assessed
> the out-of-band effects of HARP2 on the inversion of aerosol and other
> products?
>
> Best regards,
>
> Qiang

[lq17865155322]

Dear Meng,

Thank you for your patient reply!

I would like to know whether the contribution of out-of-band response affects the HARP2 productions (such as FastMAPOL suite). This idea originates from references [1, 2, 3]. However, based on the attachment you provided, I observe that the out-of-band response of HARP2 is indeed very small. I am uncertain whether it remains meaningful to quantitatively assess its impact on HARP2 products.

Qiang

References
1. Gordon, H. R. Appl. Opt. 1995, 34, 8363–8374.
2. Wang, M.; Franz, B. A.; Barnes, R. A. et al. Appl. Opt. 2001, 40, 343–348.
3. Wang, M.; Naik, P.; Son, S. Appl. Opt. 2016, 55, 2312–2323.

[sienkiewiczn]

With respect to the cross-band values shown in the plot by Qiang. I can't speak for the retrieval side, but from the Level-1 data perspective these are being accounted for. The cross-band values, while very small, are indeed real defects of the spectral filters. The instrument calibration values were compensated accordingly and a Level-1 algorithm actively corrects the instrument data similarly. Based on that, the hope would be that retrieval algorithms need not account for it, but the information is retained in the RSR file in accordance with allowing teams to investigate if the compensation is working, or that the effect is still present.

To that ends, if someone on the retrieval side believes there still appears to be biasing of the blue (and/or the green) band, then it's something the HARP2 team would be interested in receiving the details of.

-Noah