Hello,
I know that this is a subject that has been asked a lot (I've seen them at some point), but now I can not find any! So if you give me a link to some old post or some code or tutorial, it helps.
Anyway, my question is how to merge 1km L2 images of SeaWIFS, Modis-Aqua, Modis-Terra and VIIRS of chlorophyll and SST, to get more spatial and temporal data. If you look at them, they are pretty much the same. In fact, I think the only difference is that they use different "product_RGB", but since I'm not interested in true color images, it should not affect me.
The workflow should be to work each sensor individually and pass them from L2 to L2b with L2bin and then mix all the sensors with L3bin and finally get L3m with l3mapgen? And in between, do not change any parameters of the standard configuration?
Best regards,
Marco
Merging SeaWIFS, MODISA, MODIST, VIIRS chlorophyll SST L2 1km data
-
marcsandoval
- Posts: 15
- Joined: Wed Sep 19, 2018 11:22 am America/New_York
Merging SeaWIFS, MODISA, MODIST, VIIRS chlorophyll SST L2 1km data
Have you considered the approach used by ESA OC CCI? The 1km resolution suggests you are interested in a limited region -- do you have in situ data for your region?
For a region such as the N. Atlantic which has more in situ records than many other regions of comparable size, sensor versus sensor plots of geophysical variables rarely have a slope close to unity, so it doesn't make sense to use binning without addressing differences between sensors. Differences between sensors may reflect different patterns in the missing data (sun glint, saturation, optical path length constraints, etc.) as well as differences in the physical responses of the sensors and changes in plankton communities.
For a region such as the N. Atlantic which has more in situ records than many other regions of comparable size, sensor versus sensor plots of geophysical variables rarely have a slope close to unity, so it doesn't make sense to use binning without addressing differences between sensors. Differences between sensors may reflect different patterns in the missing data (sun glint, saturation, optical path length constraints, etc.) as well as differences in the physical responses of the sensors and changes in plankton communities.
-
marcsandoval
- Posts: 15
- Joined: Wed Sep 19, 2018 11:22 am America/New_York
Merging SeaWIFS, MODISA, MODIST, VIIRS chlorophyll SST L2 1km data
Hello gnwiii,
Thanks for your response. I was looking that data. It is 1km too? and if I need a subset, I should select this link:
https://rsg.pml.ac.uk/thredds/ncss/grid/CCI_ALL-v4.0-DAILY-1km/dataset.html ??
But there is no chlorophyll or sst inside. Only reflectances.
I was thinking in using a blended data of all sensors, in order to have more temporal and space data in my zone (southeast Pacific), because I always use Modis-Aqua alone but there are several days in a year without data, espetially with CHL. You say that it is not an easy task to merge different sensors? Because, for what you are saying, if each reflect different patterns in the missing data, multiple sensors should complement each other in order to have more valid data, and different response of each sensor should be averaged in the binning process, as I see. But if you say that it's not a good idea, I trust you and I will keep only Modis-Aqua (which is the sensor everybody uses here).
Best regards,
Marco
Thanks for your response. I was looking that data. It is 1km too? and if I need a subset, I should select this link:
https://rsg.pml.ac.uk/thredds/ncss/grid/CCI_ALL-v4.0-DAILY-1km/dataset.html ??
But there is no chlorophyll or sst inside. Only reflectances.
I was thinking in using a blended data of all sensors, in order to have more temporal and space data in my zone (southeast Pacific), because I always use Modis-Aqua alone but there are several days in a year without data, espetially with CHL. You say that it is not an easy task to merge different sensors? Because, for what you are saying, if each reflect different patterns in the missing data, multiple sensors should complement each other in order to have more valid data, and different response of each sensor should be averaged in the binning process, as I see. But if you say that it's not a good idea, I trust you and I will keep only Modis-Aqua (which is the sensor everybody uses here).
Best regards,
Marco
Merging SeaWIFS, MODISA, MODIST, VIIRS chlorophyll SST L2 1km data
Merging is necessary to get a long time series, but needs to be done carefully to avoid spurious time trends. We are still discovering what "carefully" means, so your best approach is to work with several data sets together with in situ data and examine the differences carefully. https://rsg.pml.ac.uk/thredds/ncss/grid/CCI_ALL-v4.0-DAILY-1km/dataset.html gives 96 pixels per degree, so about 1.15 km resolution (1km in round numbers). The Rrs values can be used to calculate chlor_a(ideally with a regional algorithm) . OC CCI doesn't provide SST; for that you might consider the GHRSST MUR product.
If you want to focus on just one sensor you should certainly compare it (e.g. using animations of difference images for Rrs and chlor_a over time) to the overlapping periods of other sensors as a check on the stability of the product you use. If you do that, you might as well include a merged product.
If you want to focus on just one sensor you should certainly compare it (e.g. using animations of difference images for Rrs and chlor_a over time) to the overlapping periods of other sensors as a check on the stability of the product you use. If you do that, you might as well include a merged product.
-
OB.DAAC - SeanBailey
- User Services
- Posts: 1480
- Joined: Wed Sep 18, 2019 6:15 pm America/New_York
- Been thanked: 5 times
Merging SeaWIFS, MODISA, MODIST, VIIRS chlorophyll SST L2 1km data
Marco,
A lot depends on what you are trying to accomplish. It is very likely that the workflow you proposed (...work each sensor individually and pass them from L2 to L2b with L2bin and then mix all the sensors with L3bin and finally get L3m with l3mapgen...) will suite your needs. While there are differences between the sensors, we do our best to ensure that the comparable products from the sensors are as consistent as possible.
One issue with active (and aging) sensors, is that they are actively degrading. It is a major effort to maintain the calibration for these sensors - and the primary reason we periodically reprocess the data. You need to be aware of that fact when making decisions about how to use the data, especially the more recent data.
Regards,
Sean
A lot depends on what you are trying to accomplish. It is very likely that the workflow you proposed (...work each sensor individually and pass them from L2 to L2b with L2bin and then mix all the sensors with L3bin and finally get L3m with l3mapgen...) will suite your needs. While there are differences between the sensors, we do our best to ensure that the comparable products from the sensors are as consistent as possible.
One issue with active (and aging) sensors, is that they are actively degrading. It is a major effort to maintain the calibration for these sensors - and the primary reason we periodically reprocess the data. You need to be aware of that fact when making decisions about how to use the data, especially the more recent data.
Regards,
Sean