How should pixel values greater than 10,000 or less than 0 in MOD09GA data be handled?

[zhooukee]

When using MOD09GA data, it has been observed that some pixel values may exceed 10,000 or fall below 0. Upon applying the scale factor, these values can exceed 1 or fall below 0, which clearly does not represent accurate surface reflectance data. It is understood that this issue arises during the atmospheric correction process. However, the reason for this problem and how to handle these pixel values during land surface parameter inversion are not well documented in the literature. Therefore, I would like to hear opinions from others regarding this matter. Currently, I have not come across any literature that discusses how to handle these values. If anyone has relevant literature, I would greatly appreciate it if you could share it with me. Thank you very much.

[LP DAACx - lien]

Hi,
Yes, this is somewhat common especially in daily data, practically every tile has some of these outliers. They should not be used, they are erroneous data.
Please see the statement below which gives more specifics on why this is the case:

The "anomalies" that you've identified, e.g. negative reflectance values,
are known issues in MODIS and Landsat surface reflectance products. The
atmospheric correction and surface reflectance retrieval algorithms are not
ideal for water bodies due to the inherently low level of water leaving
radiances and consequently very low SNR. Quantitative remote sensing
retrievals of water column constituents requires different algorithms.
Similarly reflectance values greater than 1.0 can be encountered over
bright targets such as snow and playas. These are computational
artifacts. The surface reflectance QA bands and product user guides should
be consulted.

Please let us know if you have any questions.
Thanks,
Brett

[zhooukee]

LP DAAC - lien wrote:
> Hi,
> Yes, this is somewhat common especially in daily data, practically every
> tile has some of these outliers. They should not be used, they are
> erroneous data.
> Please see the statement below which gives more specifics on why this is
> the case:
>
> The "anomalies" that you've identified, e.g. negative
> reflectance values,
> are known issues in MODIS and Landsat surface reflectance products. The
> atmospheric correction and surface reflectance retrieval algorithms are not
> ideal for water bodies due to the inherently low level of water leaving
> radiances and consequently very low SNR. Quantitative remote sensing
> retrievals of water column constituents requires different algorithms.
> Similarly reflectance values greater than 1.0 can be encountered over
> bright targets such as snow and playas. These are computational
> artifacts. The surface reflectance QA bands and product user guides should
> be consulted.
>
> Please let us know if you have any questions.
> Thanks,
> Brett

Brett, thank you for your reply.
Is it scientifically valid for me to set DN values greater than 10000 to 9999 and DN values less than 0 to 1?
Zhou

[LP DAACx - lien]

Hi Zhou,
Yes, that is what I would suggest, treating them like fill value. Certainly don't use those particular pixels for scientific research.
Thanks,
Brett

[zhooukee]

Hi Brett,
That's great, thank you very much.
Zhou

[LP DAACx - lien]

Happy to help Zhou, please let me know if I can be of further assistance.
Brett

[LP DAACx - lien]

Hi Zhou,
It is always recommended to use the State Quality flags for clouds. The State Quality incorporaes the MOD35 cloud info and is more detailed.
Thanks,
Brett