Page 1 of 2

### How to average images with clouds and ignore these cloud pixels?

Posted: Fri Feb 03, 2023 12:30 pm America/New_York
Hello everyone,

I am trying to get a monthly climate mean using 20 years of, for example, January month. My problem is that out of the 20 imagery I have 15 that have a lot of clouds. When I averaged them in SeaDAS I got negative values almost everywhere which means the cloud pixels are being taken into account into those averages. What I want is an expression that only averages positive values (chlorophyll>0). How can I write that expression in the Math Band tool?

Thank you :)

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Mon Feb 06, 2023 4:35 pm America/New_York
I assume you used collocation tool to put the L3m files together. https://seadas.gsfc.nasa.gov/help-8.3.0/collocation/CollocationTool.html

Here is an example expression for the Math Band Tool for 3 files --

((chlor_a_M > 0.0 ? chlor_a_M : NaN) +
(chlor_a_D0 > 0.0 ? chlor_a_D0 : NaN) +
(chlor_a_D1 > 0.0 ? chlor_a_D1 : NaN)) / 3.0

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Tue Feb 07, 2023 10:55 am America/New_York

Actually, I was using the Math Band tool to put the files together. I tried with the Collocation Tool and it gave me a java error...

Is it possible to use the example expression for the Math Band Tool without using the Collocation Tool?

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Tue Feb 07, 2023 5:15 pm America/New_York
If you open 3 L3m files in GUI, you can use the Math Band with the following expression to calculate the chlor_a_mean for the 3 files

((\$1.chlor_a > 0.0 ? \$1.chlor_a : NaN) +
(\$2.chlor_a > 0.0 ? \$2.chlor_a : NaN) +
(\$3.chlor_a > 0.0 ? \$3.chlor_a : NaN)) / 3.0
Screen Shot 2023-02-08 at 11.10.51 AM.png (432.61 KiB) Not viewed yet

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Tue Feb 07, 2023 7:00 pm America/New_York
I used the expression you gave and I got a full white image with NaN pixels. Should I do chlor_a < 0 instead of chlor_a > 0?

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Wed Feb 08, 2023 6:15 pm America/New_York
That is a cool trick. Very useful to know that you don't have to colocate them first in order to use BandMaths. Is it also possible to do something like this from the command line with gpt? So using data from multiple files that are projected the same in BandMaths?

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Thu Feb 09, 2023 10:40 am America/New_York
Yes. Here is the content of myGraph.xml --

<graph id="Graph">
<version>1.0</version>
<sources/>
<parameters class="com.bc.ceres.binding.dom.XppDomElement">
<file>/Users/bingyang/Scenes/MODIS_AQUA/forum/AQUA_MODIS.20200101_20200131.L3m.MO.CHL.chlor_a.4km.nc</file>
<bandNames/>
</parameters>
</node>
<sources/>
<parameters class="com.bc.ceres.binding.dom.XppDomElement">
<file>/Users/bingyang/Scenes/MODIS_AQUA/forum/AQUA_MODIS.20210101_20210131.L3m.MO.CHL.chlor_a.4km.nc</file>
<bandNames/>
</parameters>
</node>
<node id="BandMaths">
<operator>BandMaths</operator>
<sources>
</sources>
<parameters class="com.bc.ceres.binding.dom.XppDomElement">
<targetBands>
<targetBand>
<name>chlor_a_1</name>
<type>float32</type>
<expression>\$1.chlor_a</expression>
<description/>
<unit/>
<noDataValue>0.0</noDataValue>
</targetBand>
<targetBand>
<name>chlor_a_2</name>
<type>float32</type>
<expression>\$2.chlor_a</expression>
<description/>
<unit/>
<noDataValue>0.0</noDataValue>
</targetBand>
<targetBand>
<name>chlor_a_mean</name>
<type>float32</type>
<expression>((\$1.chlor_a &gt; 0.0 ? \$1.chlor_a : NaN) + (\$2.chlor_a &gt; 0.0 ? \$2.chlor_a : NaN)) / 2.0</expression>
<description/>
<unit/>
<noDataValue>0.0</noDataValue>
</targetBand>
</targetBands>
<variables/>
</parameters>
</node>
<node id="Write">
<operator>Write</operator>
<sources>
<sourceProduct refid="BandMaths"/>
</sources>
<parameters class="com.bc.ceres.binding.dom.XppDomElement">
<file>/Users/bingyang/Scenes/MODIS_AQUA/forum/AQUA_MODIS.202001_202101.L3m.MO.CHL.chlor_a.4km.nc_BandMath.dim</file>
<formatName>BEAM-DIMAP</formatName>
</parameters>
</node>
</graph>

Code: Select all

``gpt myGraph.xml``
creates a new file with chlor_a_1, chlor_a_2, and chlor_a_mean

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Thu Feb 09, 2023 1:50 pm America/New_York
Neat! Thanks for the example.

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Fri Feb 10, 2023 10:21 am America/New_York
cassandra21 wrote: Tue Feb 07, 2023 7:00 pm America/New_York I used the expression you gave and I got a full white image with NaN pixels. Should I do chlor_a < 0 instead of chlor_a > 0?
It should be chlor_a > 0
Make sure your files are opened as [1], [2], [3], when you use the example expression. Here is the screenshot of BandMath GUI. Make sure you uncheck "Virtual", so you make a real band instead of a virtual band.
Screen Shot 2023-02-10 at 10.16.28 AM.png (116.89 KiB) Not viewed yet
I also attached the Edit Expression GUI, Make sure a green "ok, no error" shows up at the bottom right corner.
Screen Shot 2023-02-10 at 10.02.41 AM.png (145.79 KiB) Not viewed yet
After you click open the chlor_a_mean band, if the image doesn't show (you see a full white image), click on "Navigation" on the lower left window.

### Re: How to average images with clouds and ignore these cloud pixels?

Posted: Mon Feb 13, 2023 9:13 am America/New_York
OB SeaDAS - xuanyang02 wrote: Mon Feb 06, 2023 4:35 pm America/New_York I assume you used collocation tool to put the L3m files together. https://seadas.gsfc.nasa.gov/help-8.3.0/collocation/CollocationTool.html

Here is an example expression for the Math Band Tool for 3 files --

((chlor_a_M > 0.0 ? chlor_a_M : NaN) +
(chlor_a_D0 > 0.0 ? chlor_a_D0 : NaN) +
(chlor_a_D1 > 0.0 ? chlor_a_D1 : NaN)) / 3.0
If only one band has a value <= 0, the OP wants the values weighted by 1/2, not 1/3. This needs multiple expressions, one compute the sum and one to determine N, the number of bands with chlor_a values, followed by the division (sum/N).