You can create a mask of the NaN values and set them to a value such as 0 or  -9999.  Then perhaps try the "Replace Bad Values" tool which can interpolate across the 'bad data' value using the surrounding values.  However, if the gaps are large, these interpolated values will not look very realistic.  Another option is to use another MODIS image and fill in the gaps using this dataset using Mosaicking.

<p>Thanks Mari, that helped alot. &nbsp;Your help and this other post here cleared it up:</p> <p>http://www.exelisvis.com/...fault.aspx</p> <p>The interpolation function does give some funny results if the holes are too big and you can see the triangle/pyramids.</p>

<p>Your last comment about mosaicking the gaps is interesting and really the ideal way to do this since I have two sets of data temporally and geographically overlapping (MODIS from Terra and Aqua). &nbsp;I've only used mosaicking to geographically pair images together but not to replace certain pixel values. &nbsp;Can you give more details on how to use mosaicking ona second MODIS image to replace NaN pixel values?</p>

<p>If I understand correctly what you need to do is mask out the locations of the clouds.</p> <p>So if image one has clouds and image two doesn't you can create a mask (make a mask of pixels = 0 or -9999 or whatever your NaN value is) of the cloud locations in image one, and then use that mask to extract the values in those locations for image 2. &nbsp;Once you have done this you can mosaic the extracted values from image 2 back into image 1 and you will have the clouds filled. You could also subset image 2 with the locations of the clouds and then mosaic from there as well.</p>

<p>Thanks. &nbsp;I follow everything you are saying and I have extracted the values from image two pertaining to clouds in image 1. &nbsp;While I have mosaicked two geographically adjacent images, I have not mosaicked geographically overlapping images.</p> <p>I'm thinking I need to do this:</p> <p>Map &gt; Mosaicking &gt;Georeferenced</p> <p>And then :</p> <p>Import &gt; Import Files and Edit Properties</p> <p>My background values have been set to NaN (both the holes and regions not captured by the sensor of either image 1/2). &nbsp;Therefore these values have been set to ignore when importing/editing the data. &nbsp;When I have imported both images and then:</p> <p>File &gt; Apply</p> <p>The output image does NOT fill the holes. &nbsp;Would it be better if I set the background value to -9999? &nbsp;When I did this with the interpolation function (Topographic &gt; Replace Bad Values) it would not work when background values were NaN but did work when they were a finite value (-9999).</p> <p>Thanks</p>

<p>I have always had problems with NaN values in ENVI GUI and because of this I default to using IDL, but are you able to successfully make a mask of where your clouds or located or no? &nbsp;Before you mosaic you should have 1 file that is the location of the clouds from the cloud free image and contains actual values, and 1 file that is an image with the clouds set as NaN and then mosaic those two together. I can probably help better if I can actually look at the files for your two images, feel free to email me at potte137@yahoo.com</p>

<p>Thanks for the help. &nbsp;I have created the mask successfully. &nbsp;I have the two images described but I am not sure how to do this type of mosaicking using the menu functions. &nbsp;From ENVI classic, wouldn't I do this:</p> <p>Map &gt; Mosaicking &gt;Georeferenced</p> <p>And then from the Map Based Mosaic Window</p> <p>Import &gt; Import Files and Edit Properties</p> <p>Assuming that the previous two steps are correct, this is where I am becoming confused when it is asking for background see through and feathering for each image (I tried to set the background see through to NaN but it didn't mosaic correctly). &nbsp;I'll try and send the files to you by email. &nbsp;</p> <p>P.S. I wish I could do this only in IDL but for the time being, I am stuck using the menus.</p>