Hello Everyone,
I work with reflectance products, specifically with MOD09A1 and MYD09A1, and as MODIS will someday in the future be decomissioned, i wanted to get ahead and migrate to VIIRS equivalent products before it happens. My application is interested in the IR, red and blue reflectance frequencies as i use them to calculate the Enhanced Vegetation Index (EVI). For this application, the MODIS products are better because they have all these frequencies in 500m resolution, which is perfect for my application, and more, they have the day of year (DOY) in which the best image in the 8-day acquisition period was selected from. As for the VIIRS products, to my knowledge, the blue band is only available at 1km resolution, and the IR and red at 500m and 1km resolution (splitted in two products, VNP09H1 and VNP09A1). Another thing is that the DOY is only available for the VNP09A1 product (with resolution of 1km) and not for the VNP09H1 (with resolution of 500m).
Given this, what i would really appreciate if you could help me answer is:
1- the DOY from VNP09A1 is the same for the VNP09H1?
2- is there any plans for including a DOY band in the VNP09H1 product?
3- if i'm mistaken, can you point out to me what is the equivalent MODIS-VIIRS product for my use case?
I always used the 8-day composite products because you guys were capable of selecting the best pixel value within the acquisition period. I once tried to perform this with the daily products, but was not capable of selecting the same pixel as you guys did. If you could help me understand the best pixel selection algorithm, that would be really great! When i tried it, i was following the user guide, but could not make much sense of the operations.
MODIS to VIIRS transition
-
LP DAAC - dgolon
- User Services
- Posts: 296
- Joined: Mon Sep 30, 2019 10:00 am America/New_York
- Has thanked: 18 times
- Been thanked: 2 times
- Contact:
Re: MODIS to VIIRS transition
Hi @veeries Thanks for writing in. Let me confirm a few things with our Science Data Lifecycle Team and we'll follow up with more information soon.
Subscribe to the LP DAAC listserv by sending a blank email to lpdaac-join@lists.nasa.gov.
Sign up for the Landsat listserv to receive the most up to date information about Landsat data: https://public.govdelivery.com/accounts/USDOIGS/subscriber/new#tab1.
Sign up for the Landsat listserv to receive the most up to date information about Landsat data: https://public.govdelivery.com/accounts/USDOIGS/subscriber/new#tab1.
-
LP DAAC - lien
- User Services
- Posts: 183
- Joined: Thu Jun 25, 2020 9:51 am America/New_York
- Been thanked: 4 times
Re: MODIS to VIIRS transition
Hi,
Let me start by first sending you this from the MxD09 User Guide, the VNP09 user guide mentions this also, but the MODIS user guide is a little more detailed on how the best available pixel is chosen:
For each pixel, the compositing steps are
1) Observations from the same orbit are composited by observational coverage. Observations with
the highest coverage are saved, and the rest discarded. This yields a list of one observation from
each orbit.
2) Each orbit's observation is then assigned a score, based upon whether it is flagged for cloud,
cloud shadow, high aerosol or low aerosol, or contains high view angle or low solar zenith angle.
The lowest score, 0, is assigned to observations with fill values for data. The remaining scores are
1 BAD data derived from a faulty or poorly corrected L1B pixel
2 HIGHVIEW data with a high view angle (60 degrees or more)
3 LOWSUN data with a high solar zenith angle (85 degrees or more)
4 CLOUDY data flagged as cloudy or adjacent to cloud
5 SHADOW data flagged as containing cloud shadow
6 UNCORRECTED data flagged as uncorrected
7 CLIMAEROSOL data flagged as containing the default level of aerosols
8 HIGHAEROSOL data flagged as containing the highest level of aerosols
9 SNOW data flagged as snow
10 GOOD data which meets none of the above criteria
The observation with the highest score and the lowest view angle is selected for the MOD09A1 and
MOD09Q1 outputs.
The MOD09A1 outputs also contain a 500 m version of the 1 km State QA composited from all 8-day
inputs.
The documentation, while it does not come right out and say that the same pixel acquisition is used for both the VNP09A1 and the VNP09H1, it is implied. This is the same with the MOD09A1 vs the MOD09Q1. The A1 contains more ancillary SDS layers then it's higher resolution counterpart. Such as the State Quality flags and the Day of Year layer. These can be used for both products.
As with the MODIS products the Day of Year was not added to the MxD09Q1, and there are not plans to add it to the VNP09H1. However, we do appreciate your input on this and will bring it up with the VIIRS Science team.
Unfortunately, for your use case this is the comparable option. VIIRS was designed to carry on the MODIS legacy. They attempted to keep the products as similar as possible, but there are differences in wave lengths and resolution and not all MODIS products were reprised for the VIIRS mission.
Please let us know if you have any more questions.
Thanks,
Brett
Let me start by first sending you this from the MxD09 User Guide, the VNP09 user guide mentions this also, but the MODIS user guide is a little more detailed on how the best available pixel is chosen:
For each pixel, the compositing steps are
1) Observations from the same orbit are composited by observational coverage. Observations with
the highest coverage are saved, and the rest discarded. This yields a list of one observation from
each orbit.
2) Each orbit's observation is then assigned a score, based upon whether it is flagged for cloud,
cloud shadow, high aerosol or low aerosol, or contains high view angle or low solar zenith angle.
The lowest score, 0, is assigned to observations with fill values for data. The remaining scores are
1 BAD data derived from a faulty or poorly corrected L1B pixel
2 HIGHVIEW data with a high view angle (60 degrees or more)
3 LOWSUN data with a high solar zenith angle (85 degrees or more)
4 CLOUDY data flagged as cloudy or adjacent to cloud
5 SHADOW data flagged as containing cloud shadow
6 UNCORRECTED data flagged as uncorrected
7 CLIMAEROSOL data flagged as containing the default level of aerosols
8 HIGHAEROSOL data flagged as containing the highest level of aerosols
9 SNOW data flagged as snow
10 GOOD data which meets none of the above criteria
The observation with the highest score and the lowest view angle is selected for the MOD09A1 and
MOD09Q1 outputs.
The MOD09A1 outputs also contain a 500 m version of the 1 km State QA composited from all 8-day
inputs.
The documentation, while it does not come right out and say that the same pixel acquisition is used for both the VNP09A1 and the VNP09H1, it is implied. This is the same with the MOD09A1 vs the MOD09Q1. The A1 contains more ancillary SDS layers then it's higher resolution counterpart. Such as the State Quality flags and the Day of Year layer. These can be used for both products.
As with the MODIS products the Day of Year was not added to the MxD09Q1, and there are not plans to add it to the VNP09H1. However, we do appreciate your input on this and will bring it up with the VIIRS Science team.
Unfortunately, for your use case this is the comparable option. VIIRS was designed to carry on the MODIS legacy. They attempted to keep the products as similar as possible, but there are differences in wave lengths and resolution and not all MODIS products were reprised for the VIIRS mission.
Please let us know if you have any more questions.
Thanks,
Brett
Re: MODIS to VIIRS transition
Really appreciate for the reply!
I think i might have made some incorrect assumptions, let me back out a bit.
Also, i haven't made it clear what the purpose of my question was, and i think it might help if you know. That being said, what i want is to be able to generate X-day composited images with the MOD09GA daily product (for simplicity lets consider i want to perform this only for MODIS products - leaving VIIRS aside).
For this, my first assumption was that the 8-day composited MYD09A1 was generated by selecting the best pixel within 8 individual images (one for each day) from the MOD09GA data product, perhaps this is incorrect?
When i tried to reproduce what i assumed you guys did, i did the following:
for each of the 8 images i decoded the 32 bit "QC_500m_1" band, and was trying to select the best pixel based on a "combined score" - generated by me - of all reflectance band quality information. However this became really difficult really fast and i was not being able to reproduce the same as you guys did. One of the reason is that i couldn't determine what was worse for a pixel information, that is:
(extracted from table 10 of MOD09):
band quality, four bit range:
0000 - highest quality (this one is clear)
0111 - noisy detector (i think it's better than dead detector)
1000 - dead detector, data interpolated in L1B (i think it's worse than noisy detector)
1001 - solar zenith >= 86 degrees (this one i wasn't sure if it was worse than noisy or dead detector)
1010 - solar zenith >= 85 and < 86 degrees (this one i assumed was better than the previous)
...
1111 - not processed due to deep ocean or clouds (this one was also clearly the worse)
but most of the "in between" quality information i wasn't sure how to weight them for generating my "combined score".
Do you think that it is possible for me to generate the 8-day composited image of MYD09A1 using the MYD09GA? (keep in mind that i don't want to reproduce the same product that you guys already do, i'm citing the 8-day composited as reference for validation)
And if this is possible, what path should i follow to achieve this?
I think i might have made some incorrect assumptions, let me back out a bit.
Also, i haven't made it clear what the purpose of my question was, and i think it might help if you know. That being said, what i want is to be able to generate X-day composited images with the MOD09GA daily product (for simplicity lets consider i want to perform this only for MODIS products - leaving VIIRS aside).
For this, my first assumption was that the 8-day composited MYD09A1 was generated by selecting the best pixel within 8 individual images (one for each day) from the MOD09GA data product, perhaps this is incorrect?
When i tried to reproduce what i assumed you guys did, i did the following:
for each of the 8 images i decoded the 32 bit "QC_500m_1" band, and was trying to select the best pixel based on a "combined score" - generated by me - of all reflectance band quality information. However this became really difficult really fast and i was not being able to reproduce the same as you guys did. One of the reason is that i couldn't determine what was worse for a pixel information, that is:
(extracted from table 10 of MOD09):
band quality, four bit range:
0000 - highest quality (this one is clear)
0111 - noisy detector (i think it's better than dead detector)
1000 - dead detector, data interpolated in L1B (i think it's worse than noisy detector)
1001 - solar zenith >= 86 degrees (this one i wasn't sure if it was worse than noisy or dead detector)
1010 - solar zenith >= 85 and < 86 degrees (this one i assumed was better than the previous)
...
1111 - not processed due to deep ocean or clouds (this one was also clearly the worse)
but most of the "in between" quality information i wasn't sure how to weight them for generating my "combined score".
Do you think that it is possible for me to generate the 8-day composited image of MYD09A1 using the MYD09GA? (keep in mind that i don't want to reproduce the same product that you guys already do, i'm citing the 8-day composited as reference for validation)
And if this is possible, what path should i follow to achieve this?
-
LP DAAC - lien
- User Services
- Posts: 183
- Joined: Thu Jun 25, 2020 9:51 am America/New_York
- Been thanked: 4 times
Re: MODIS to VIIRS transition
Hi,
You are correct, that is the main premise on the how they produce the 8-day composite. They choose the best available from all 8 daily products. However, I think where the issue may be is that they do not use the QA layer to determine which daily acquisition they chose for the 8-day product. The QA is more of a generalized by product of the algorithm when it processing the ancillary data that is included in the daily products. They have the Sensor and solar azimuth and zenith angles, they have the range layer, they have the pointer files. The algorithm takes all those into account when it decides which pixel to use. This is probably why you are encountering some differences.
I think your best option would be to decide on your own values and stay consistent with those. There really is several right ways for you to do this, and if it is not exactly like the MOD09A, it doesn't mean it is wrong.
Thanks,
Brett
You are correct, that is the main premise on the how they produce the 8-day composite. They choose the best available from all 8 daily products. However, I think where the issue may be is that they do not use the QA layer to determine which daily acquisition they chose for the 8-day product. The QA is more of a generalized by product of the algorithm when it processing the ancillary data that is included in the daily products. They have the Sensor and solar azimuth and zenith angles, they have the range layer, they have the pointer files. The algorithm takes all those into account when it decides which pixel to use. This is probably why you are encountering some differences.
I think your best option would be to decide on your own values and stay consistent with those. There really is several right ways for you to do this, and if it is not exactly like the MOD09A, it doesn't mean it is wrong.
Thanks,
Brett