Best practice for mosaic datasets of multispectral imagery

[bogdanpalade2]

Hi all,

I want to set up a satellite imagery management system based on mosaic datasets. I thought that the best way would be to configure different mosaic datasets for different sensors (Pleiades, Worldview1,2,3, Ikonos, GeoEye...etc) according to the sensor models that ArcGIS Pro offers as options. That is because different sensors have different band numbers and pixel depth. Now, I have some questions:

1. For example, I configure one mosaic dataset for WorldView2 with 8 bands. Nevertheless, sometimes we receive the WV2 imagery containing only 4 bands. Will I get in trouble by storing this 4 bands imagery in the 8 bands mosaic dataset? Eventually this mosaic dataset will be published as an image service where the user will be able to change the band combination. Will the user encounter 8 bands although only 4 of them are actual data. Will that be confusing? 

 2. All the sensor based mosaic datasets will be used to create a single derived mosaic dataset containing all imagery. How many bands should I set for this mosaic dataset? The maximum possible (according to the sensor which has most bands)? This derived mosaic dataset with contain different bit depth data (from the original sensor based mosaic datasets). Will it be capable of rendering everything correctly?

Thank you very much!

[DanPatterson]

basics are here

Imagery: Data management patterns and recommendations—ArcGIS Pro | Documentation

[bogdanpalade2]

Hi Dan, thank you, I have read this documentation previously, all good, but I wanted some feedback from someone that did this in the “real world” . We have here a pretty large collection (around 20 TB of imagery) and I wanted to hear of potential problems. For example, how is the performance affected by the mosaic dataset functions. Will the imagery display at the same speed if I use an extract bands function for imagery with a certain bands number?