Imagery: Data management patterns and recommendations

The ideal way to manage small to vast collections of image and raster data in ArcGIS Pro is using a mosaic dataset. The mosaic dataset is a versatile data management solution:

  • Manage data as a collection of raster datasets, regardless of varying spatial, spectral, temporal, and radiometric resolutions, and with full access to the metadata for each item.
  • Dynamically deliver a mosaicked image or provide access to the individual images.
  • Process the data on the fly.
  • Share as a dataset or image service.

Mosaic datasets allow users to access a single source to obtain the data they need—simplifying maintenance and application development. The associated attribute table can be accessed, allowing users to query the mosaic dataset and access each item stored in it, whereas the mosaicked image can be used like a raster dataset—it appears as one continuous dataset and can be processed with the tools used to process an image.

They can be extremely large, both in total file size and number of raster datasets. They do not control the source data but contain pointers to the source data.

They are created, edited, and managed with the tools in the Mosaic Dataset toolset in the Data Management toolbox. Mosaic datasets can be analyzed and processed using tools and function that support image and raster datasets, such as the Image Analyst extension.

Data sources

The image and raster data can come from a variety of sources, such as aerial or satellite sensors, scanned maps, the output from analysis, and active sensors such as lidar and radar data. It can be panchromatic, multispectral, multidimensional, thermal, elevation, or thematic. Source data can be stored as files on disk or in a file storage system (such as NAS or SAN), in a geodatabase, or accessed through a service (such as an image service or web coverage service [WCS]).

Image and raster data is added to a mosaic dataset according to its raster type. The raster type simplifies the processing of adding complex image data to a mosaic dataset. It recognizes file format and specific information about a product, such as metadata, georeferencing instructions, acquisition date, sensor type, processing templates, and band wavelengths. Raster formats are similar to raster types; however, it only defines how pixels are stored, such as number of rows and columns, number of bands, bit depth, actual pixel values, and other raster format-specific parameters. Many raster types are supported in ArcGIS Pro, some for specific image products and others for specific image sensors, such as Landsat, WorldView-4, or Sentinel-2.

By adding raster data according to a raster type, the metadata is read and used to define any processing that needs to be applied. For example, when adding a Sentinel-2 scene, the raster type recognizes that the metadata is stored in an .xml file and the bands are organized into one or more JPEG2000 files. It also recognizes that this imagery can be corrected to top of atmosphere, bottom of atmosphere, or orthorectified, so depending on the options you choose, it will add the appropriate functions so the image can be processed accordingly. If you added this data as a regular raster dataset, only the JPEG2000 image files will be recognized and added, and any metadata information that would affect the functions needed or the orthorectification would be missing.

It is important that you use the correct raster type to add imagery to a mosaic dataset. You may need to examine the files and their metadata sources to identify the file format or image product that is identified using the raster type.

Functions that define processing can also be added after imagery has been added to a mosaic dataset. This is often done to convert the output to a specific image product or to apply corrections to individual images. Functions can be applied to individual images or to the entire mosaic dataset.


Regardless of the mosaic dataset configuration you are implementing, you must ensure that the imagery is readable; otherwise, the mosaic dataset will not be able to display the imagery. The location of the imagery is identified in a specified path; if you move the imagery, you must update the mosaic dataset, and vice versa.

Processing with functions

Functions are a key component to every mosaic dataset. They enhance or modify the mosaicked image product by applying on-the-fly processing operations such as orthorectification, band indices, image enhancements, and image algebra. You can add functions to the mosaic dataset or to individual image items in the mosaic dataset, or they can be added when the data is added to the mosaic dataset through the raster type. For example, when specific raster data products (such as from a satellite sensor) are added to a mosaic dataset, some functions are automatically added to the raster data.

Functions allow multiple products to be created from a single raster source because imagery is processed as it's accessed.

You can apply functions to the entire mosaic dataset or to individual items in the mosaic dataset, such as the following examples:

  • Stretch—Enhances the image by adjusting the brightness, contrast, and gamma based on the statistics of the data
  • Composite Band—Combines multiple files into a single image, such as multiple TIFF files representing individual bands in a single scene
  • Arithmetic—Performs an arithmetic operation between two spatially overlapping rasters or a raster and a constant value
  • Hillshade—Allows you to generate a hillshaded image from an elevation model
  • LAS To Raster—Allows you to visualize LAS files in a mosaic dataset by determining how they will be rasterized

Learn more about raster functions


The properties of a mosaic dataset include the general properties that are similar to those you find for all raster datasets, such as the data source, extent, pixel sizes, and bit depth. They also include more advanced information that affects how the mosaicked image will be displayed in a map client, how you might interact with it, and how the performance of the server or image service is impacted if the mosaic dataset is published.

Learn more about mosaic dataset properties

When publishing a mosaic dataset using ArcGIS Server, the server administrator can modify some of the properties as part of the settings on the image service; however, they are unable to exceed the maximums you have set. For example, if you limit the allowed mosaic methods to only three of the methods, the administrator is unable to add a fourth method. Or, if you set the maximum number of downloadable items, they can reduce, but not increase, this number. If you change the properties to exceed or limit a value, such as the Maximum Size Of Requests value, you must completely republish the mosaic dataset. If you restart the image service, the changed properties in the mosaic dataset are not enabled.

Some properties control how the data is added to the mosaic dataset and how ArcGIS Pro will render the dataset. For example, the product definition allows you to customize the mosaic dataset to contain data with a specific number of bands and wavelengths. It controls how the data is added to the mosaic dataset and how it displays by default, and it aids in some processing. The product definition is typically used to support specific satellite imagery products, such as Sentinel and Landsat, but you can customize it by defining the number of bands, the band order, and the wavelength ranges supported by each band.

The most commonly used product definitions are listed below:

  • Natural Color (RGB)—Creates a three-band mosaic dataset with red, green, and blue wavelength ranges
  • Natural Color (RGBI)—Creates a four-band mosaic dataset with red, green, blue and near infrared wavelength ranges
  • Color Infrared (IRG)—Creates a three-band mosaic dataset with near infrared, red, and green wavelength ranges

Using a product definition helps when adding data containing wavelength information in its metadata. If the wavelength information is ordered differently in the inputs, they are all ordered correctly when added to the mosaic dataset. For example, if band 1 in a Landsat-9 scene is a blue wavelength, and band 3 in the mosaic dataset is designed to contain the blue wavelengths, Landsat-9's blue band will be mapped to the mosaic dataset's blue band. Without the wavelength information, the Landsat-9's blue band may be mapped to the mosaic dataset's red band.

Learn about mosaic dataset properties

Mosaic dataset configurations

The basic design for a mosaic dataset is one dataset containing a collection of imagery. In this design, each image or raster dataset is added as an individual item in the mosaic dataset and represented as a row in the attribute table.

Basic configuration of a mosaic dataset

It is generally recommended that you manage imagery in a mosaic dataset, and that you use another mosaic dataset (a reference mosaic dataset) to share or publish the contents. By using a reference mosaic dataset, users cannot accidentally make modifications to a mosaic dataset, such as adding or removing data.

Different mosaic datasets may be published to define different types of data, such as natural color imagery, false color imagery, or elevation. Data does not have to be organized by a specific geography, type of sensor, or date range, since a mosaic dataset can accommodate all of these differences.

Typical examples mosaic datasets include the following:

  • Natural color imagery—The imagery matches the colors that are detectable with human vision (often using a 3,2,1 {R, G, B} band combination)
  • Color infrared imagery—Containing color infrared imagery (often using a 4,3,2 [NIR, R, G] band combination displayed as R, G, B)
  • Multispectral imagery for visual interpretation—Imagery that has been enhanced for visual interpretation and may be pan sharpened
  • Multispectral imagery for analysis—Imagery with all available bands providing radiance or reflectance values
  • Multidimensional rasters for analysis—Raster data containing one or more variables (such as temperature or salinity) for one or more dimensions (such as height or depth value).
  • NDVI—A colorized normalized difference vegetation index (NDVI)
  • Elevation—A compilation of digital elevation models (DEMs) representing orthometric (above sea level) heights or ellipsoidal heights
  • Slope—A raster representing slope steepness calculated from elevation in degrees
  • Aspect—A raster representing slope direction calculated from elevation
  • Hillshade—Hillshade raster created from elevation
  • Shaded Relief—Shaded relief image created from elevation
  • Scanned topographic or application-specific maps
  • Analysis results producing a thematic image with pixel attribute values

Organization of mosaic datasets

The organization of mosaic datasets can become more complex as you need to manage different types of data. Two standard combinations that you can use to manage and publish imagery are illustrated below.

General configurations of a mosaic dataset

It is generally advantageous to separate mosaic datasets into two types: those that are primarily used for management and those that are published. This separation can aid in organization.

As you build and organize a collection of imagery using mosaic datasets, it is useful to understand the different types of mosaic datasets and what purpose they may serve. The source and derived mosaic datasets discussed below are symbolic names used to help convey an understanding of the organizational structure of mosaic datasets, whereas a reference mosaic dataset is a physically different form of a mosaic dataset.

Source mosaic datasets

The source mosaic dataset is used for managing imagery. It generally contains a collection of similar imagery. You can use a number of these source mosaic datasets to manage different collections. These can be published directly or used as the source for other mosaic datasets. It is recommended that you provide access to (publish) this mosaic dataset using a reference mosaic dataset to keep it secure.

Source images added to source mosaic dataset
A source mosaic dataset has the image and raster data added directly.

A source mosaic dataset is created using the Create Mosaic Dataset tool. If the input imagery has a consistent bit depth or number of bands, these values do not need to be defined in the tool, since they'll be taken from the first image added. The spatial reference system will likely be the same as the inputs, but if the input data spans multiple spatial reference systems, choose an appropriate one for input imagery. Next, use the Add Rasters To Mosaic Dataset tool and use the appropriate raster type.

In most cases, the images in a source mosaic dataset have the same number of bands and bit depth. These source mosaics are managed and used to refine aspects of the collection, such as refining the footprints or setting processes such as orthorectification and renderings.

You can modify the functions for individual images by accessing the Viewer window for each through the attribute table, or modify multiple images using the Raster Function Template Editor accessed from the Footprint layer in the ArcGIS Pro Contents pane.

Generally, if the imagery represents a single dataset, such as imagery covering a specific date, build overviews for the mosaic dataset.

Derived mosaic datasets

The derived mosaic dataset is used to define collections of imagery often viewed by users as a single collection. The source of a derived mosaic dataset is generally one or more source mosaic datasets. For example, this can be a collection of all natural color imagery, with the source coming from multiple source mosaic datasets. It is recommended that you provide access to (publish) this mosaic dataset using a reference mosaic dataset to keep it secure. Additionally, you can create other mosaic datasets from a derived mosaic dataset to provide specific imagery products—such as a specific band combination—or only over a specific area.

Source mosaic datasets added to derived mosaic dataset
The image and raster data are added via a mosaic dataset to a derived mosaic dataset.

A derived mosaic dataset is also created using the Create Mosaic Dataset tool. Often the input imagery will have various bit depths and bands, so you should specify these values or define a product definition to control the output of the mosaic dataset. Additionally, choose a spatial reference system that can accommodate all the imagery.

The spatial reference system is used to generate the footprints, boundary, and other related items in the mosaic dataset, as well as a default with which the mosaicked imagery will be resampled. Choose one that is suitable for all the imagery you may add. This can be a country system or UTM zone. However, if you're creating a mosaic dataset that may be global in extent or will be mashed up with web services, you may want to use the WGS 1984 Web Mercator Auxiliary projection.

It is recommended that you add a source mosaic dataset to another mosaic dataset using the Table raster type. Using the Table raster type creates a mosaic dataset containing all or a selection of the table items in the source mosaic datasets, not only a single item representing the source mosaic dataset. This gives you the ability to perform queries and continue to access the individual metadata for each item. You can also add functions to individual items and more easily customize the mosaic dataset with seamlines, mosaic methods, and color corrections. Additionally, you can use the Synchronize Mosaic Dataset tool with the Update cell size ranges option disabled to update this mosaic dataset if any of the sources have been modified, such as modified footprints or new imagery that has been added.

If you add the source mosaic datasets using the Raster Datasetraster type instead of the Table raster type, each source mosaic dataset will be represented as a single item in the derived mosaic dataset. This limits your ability to perform queries, limits metadata access to the source mosaic dataset instead of each image, and limits the scalability of the mosaic dataset.

Generally, you do not build overviews for this mosaic dataset, as the overviews exist in the source mosaic datasets. However, it may be necessary to build them if the derived mosaic dataset covers a much greater extent than each source. Instead of building overviews, you can use another image or image service to provide imagery coverage for the full extent of the mosaic dataset. When adding this image, you should uncheck the option to build the boundary because the boundary will be extended to the extent of this image—which may not be ideal.

Referenced mosaic datasets

Referenced mosaic datasets behave similarly to a regular mosaic dataset; however, you cannot add additional rasters to the referenced mosaic dataset, you cannot build overviews, and you cannot calculate the pixel size ranges. You can redefine the boundary, for example, to restrict access to specific areas or define additional functions to be applied to all imagery. It is used to provide access to mosaic datasets with different mosaic dataset-level functions. Sharing access to a referenced mosaic dataset ensures that those accessing it cannot make modifications to the source or derived mosaic datasets, which may affect other users.

Reference mosaic datasets are created using the Create Referenced Mosaic Dataset tool and by defining another mosaic dataset as the source. Typically, this source can be a source mosaic dataset or a derived mosaic dataset.

Reference mosaic datasets created from other mosaic datasets
Reference mosaic datasets are only created from other mosaic datasets.

Any properties set on the input mosaic dataset are carried over to the reference mosaic dataset—such as the default mosaic method—or mosaic dataset functions. You can modify or remove any of these without affecting the input mosaic dataset. You can modify the functions or properties by opening the mosaic dataset's Properties dialog box from the Catalog pane.

Recommendations for managing imagery collections

You can manage all your imagery in a single mosaic dataset. This is ideal when the data is similar in image type, number of bands, and bit depth. However, when you have large collections of imagery that encompass data from different sources and sensors, it is best to organize the imagery into smaller, data-specific collections. It simplifies the setup and maintenance of a mosaic dataset when all the imagery managed in it has a similar source and the same number of bands and bits such as the characteristics below:

  • Preprocessed orthoimage tiles of the same date
  • Imagery collected from a similar sensor with the same number of bands and bit depth such as the following:
    • Collections of 4-band 16-bit imagery (QuickBird, IKONOS)
    • RapidEye (5 band)
    • SPOT
    • Landsat 7, 8, or 9
    • ASTER
  • Imagery from a single aerial survey project
  • Elevation data from one source
    • SRTM
    • Lidar

These separate source mosaic datasets are easier to manage. You can then combine them to create the application-specific mosaic datasets that are published.

Single orthoimage collection example

You may have a large collection of color aerial imagery, such as thousands of images collected over a state or province. You can create one mosaic dataset to manage all these. You may want to modify the attribute table to add information specific to the imagery, for example, the acquisition date and the location, number of bands, and data format. You can then directly publish this or create reference mosaic datasets to provide this imagery to users within your organization. You can modify the boundary of a reference mosaic dataset to only provide the imagery within a particular project area, or you can create one that only contains the images that meet a particular query, such as a county or city.

Multiple orthoimage collection example

You may have a collection of aerial imagery from three years, such as 2010, 2015, and 2020, to create a time series mosaic dataset. Ideally, each of these collections should have similar image properties, including the number of bands, spatial resolution, bit depth, and time of year (temporal resolution). Matching the properties of the images will allow for easier comparison to determine change and compare the images. Since the mosaic dataset will have a specific number of bands, bit depth, and spatial resolution, you must determine how to best fit the input data. Since existing mosaic datasets can be used as input for a reference mosaic dataset, you can create one mosaic dataset that references the source mosaic datasets. These mosaic datasets can be used for time series analysis or to store all of the available imagery in a collection of the particular study area.

If the spatial resolutions vary, you may need to choose the minimum mapping unit for your workflow or analysis. When the spatial resolution varies, the Min/Max PS in the mosaic dataset will also change, which may alter the visibility of the images. There are two ways to organize this data: as separate source mosaic datasets and one combined derived mosaic dataset, or as one mosaic dataset. Using a combination of source and derived mosaic datasets generally makes the management easier while maintaining optimum performance.

  1. To do this, create three source mosaic datasets. You can specify the band and bit depths when you create them or allow the software to define them when the data is added. Creating pyramids for pregenerated tiles generally does not bring benefits, so pyramid generation can be skipped. But you should build overviews on each of the source mosaic datasets. This way, when you query the derived mosaic dataset by date as you zoom the image, it will be consistent. If the collections do not have overviews, you need to build overviews on the derived mosaic dataset, and this can only be generated for a single attribute date. Modify the attribute table for each collection by adding the same new field for the Year and populate the field with the year.
    Workflow diagram
  2. Create one derived mosaic dataset containing three bands. This one will be used to provide the best color imagery combination. Then add the three source mosaic datasets to it using the Table raster type and the Update cell size ranges option disabled. You won't have to build overviews, as they were already created in each source mosaic dataset. You may want to modify some of the properties, such as defining the mosaic method to be By Attribute and specifying the default year, such as 3000, to display using the latest imagery.
  3. It is recommended that you create a reference mosaic dataset to publish the contents of the derived mosaic dataset if the access will be directly to the dataset. If you're publishing the mosaic dataset as an image service, you can publish it directly. In either case, you will have one dataset to access, query, and explore.

Add data

When you obtain new orthoimagery, create a source mosaic dataset. Then add the source mosaic dataset to the original derived mosaic dataset using the Table raster type and build overviews. By default, the new imagery is visualized immediately because the By Attribute mosaic method was defined earlier.

Additional mosaic datasets

To make false color infrared imagery available, create a reference mosaic dataset from the source mosaic dataset. Then open the mosaic dataset properties from the Catalog pane and add the Extract Bands function. Define the band IDs as 4 3 2. Originally, the mosaic dataset has four bands, which is the same as the original; however, by adding this function, you've defined a default band combination and modified the mosaic dataset to output only three bands.

You can also create a normalized difference vegetation index (NDVI) mosaic dataset. This can be done using a reference mosaic dataset to point to the color infrared mosaic dataset and add the NDVI function to apply the processing required. Alternatively, a mosaic dataset can be created that references the new source mosaic dataset and adds an NDVI function.

You can also use server-side processing when you share image services. This allows you to create one image service from a mosaic dataset, which can use many server-side functions to process and display data.

Satellite imagery collection example

If you have a collection of imagery from similar satellite sensors, such as IKONOS (an ortho-ready product) or QuickBird (a basic bundle product), that have four multispectral bands collected at one resolution and a high-resolution panchromatic band, you can manage it in a single mosaic dataset. You can create a pan-sharpened mosaic dataset from this imagery.

Prior to including the imagery in a mosaic dataset, it is helpful to build pyramids and statistics.

When working with satellite imagery, there is information—such as wavelengths and sun angles—that can be useful. To ensure that this information is used, define a product when creating the mosaic dataset. This product definition references the wavelength ranges that the mosaic dataset will support for each band.

For this scenario, create a mosaic dataset using either the IKONOS or QuickBird product definition as appropriate. Add the imagery using the IKONOS or QuickBird raster types, depending on which source you're adding. Make sure the Pansharpen product template is defined on the Raster Type Properties dialog box, which is the default product template. Another benefit to using the appropriate raster type is that the footprints for each image will be calculated to exclude unwanted image boarder areas.

A number of attributes will be added as part of the raster type. You can add more attributes to help manage and organize the data, such as defining the accuracy or quality of the imagery. Similarly, you can define an attribute, such as Publish, to determine whether the image is to be published. This way, you can exclude or include specific scenes from publishing, or they can be used for more specific publishing-related queries.

Elevation collection

There are many reasons for creating a mosaic dataset of elevation data; for example, you may want to access all elevation data from a single source or you may want to use the elevation data as a data source to orthorectify other imagery. Typically, you can manage all elevation data in one mosaic dataset. Create a mosaic dataset, specifying the largest bit depth of the input data, which is generally 32 bit, then add the imagery according to its raster type. Be sure that the elevation data represents height as either orthometric or ellipsoidal and that the units of height are the same (such as meters or feet). If they are not, the mosaic dataset requires more steps to create, but you can use the Arithmetic function to modify these values for each input.

See Converting from orthometric to ellipsoidal heights.

If you have multiple sources for elevation data, such as lidar, bathymetry, and sonar, you can create separate source mosaic datasets for each source to manage them separately; then create a single mosaic dataset that combines them.

Generally, you should use the elevation data that is most accurate or has the highest resolution. You can edit the mosaic dataset properties to choose the By Attributemosaic method. Define LoPS as the order field and 0 as the default value. Then the highest-resolution elevation data at the view or requested scale will be displayed or used. If a field for accuracy exists, this can be used instead.

This mosaic dataset can act as a source to multiple referenced mosaic datasets that are created to produce output from the elevation data, such as hillshade, aspect, or slope.

The example above shows how with a simple collection of imagery, there are choices in how to manage data. But the main objective is to create source mosaic datasets, bring them together using a derived mosaic dataset, and publish the data.

To see a workflow for creating a mosaic dataset such as the one described in the previous section, see Create a mosaic dataset containing raster data from multiple dates.

Share mosaic datasets

You can share a mosaic dataset by sharing the geodatabase and giving direct access to the mosaic dataset, or by publishing the mosaic dataset through ArcGIS Enterprise using the ArcGIS Image Server extension or through ArcGIS Online using ArcGIS Image for ArcGIS Onlineto publish an imagery layer.

If you're planning to share a mosaic dataset using direct access, it is recommended that you create a reference mosaic dataset to provide direct access. Anyone who can directly access the mosaic dataset can edit it, and you shouldn't provide direct access to your main source mosaic dataset.

If you're planning to serve the mosaic dataset as an image service or imagery layer, you can serve it directly, as users will not have direct access to the mosaic dataset. By modifying the default properties of the mosaic dataset, you can control how users can interact with the shared imagery.

Cache mosaic datasets

You can cache an image service or cache a map service or globe service containing raster data or an image service. Generally, the pyramids generated for raster datasets or the overviews generated for mosaic datasets result in image data being served at an acceptable rate. However, if you know that a particular image or area of interest will be repeatedly visited, you may want to generate a cache.

It is advantageous not to include vector and imagery data in the map or globe document to be published. Generally, it is better that vectors and imagery are served as separate services that are merged together by the client application.

Properties of a published mosaic dataset

When publishing a mosaic dataset as an image service, there are many properties that can be modified that control access to the mosaic dataset and individual images:

  • Modify the accessible fields in the attribute table.
  • Limit the number of images that can be downloaded.
  • Limit the request size.
  • Limit the metadata available.
  • Define the default mosaic method.
  • Define the default compression for transmission.

Learn more about each of the mosaic dataset properties

When preprocessing is necessary

Managing and publishing imagery using a mosaic dataset can save you time over traditional methods of mosaicking image collections together or producing multiple outputs; however, there are use cases when you want to consider some preprocessing. The recommended preprocessing applies to creating the fastest and best mosaicked imagery display.

Build pyramids—Pyramids help to improve the display speed of imagery. They can also impact the number of mosaic dataset overviews that are generated. Generally, you should build pyramids for images with greater than 3,000 columns. There is little to gain from building pyramids for a collection of preprocessed and tiled imagery since overviews generally provide a better solution for improving performance.

Calculate statistics—Statistics are used by the renderer when the imagery is enhanced for display. Generally, you should calculate statistics for imagery that is not radiometrically enhanced. For example, many orthoimages are enhanced as part of their processing (such as NAIP or DOQQ), so you do not need to calculate statistics. Raw imagery or imagery from a satellite is generally not enhanced, so you should calculate the statistics for the imagery to display properly. Statistics don't always need to be calculated from every pixel; you can increase the speed at which statistics are calculated by specifying a skip factor. One way to identify a reasonable skip factor value is to divide the number of columns by 1,000 and use the quotient (integer) as the skip factor.


Two tools are recommended for building pyramids and calculating statistics. You can use the two check boxes on the Add Rasters To Mosaic Dataset tool to build the pyramids and statistics as part of the procedure to add the imagery to the mosaic dataset. Otherwise, use the Build Pyramids And Statistics tool, which can be run on a workspace containing the data or on the data in the mosaic dataset. This can be run before or after adding the imagery to the mosaic dataset. If you are going to build pyramids, build these before defining or building overviews on the mosaic dataset.

Optimized image formats—Some imagery can be slower to read than others due to their storage format or compression, and it is recommended that you convert these into more optimal formats. For example, an ASCII DEM image format is slow to read; therefore, it is recommended that you convert it to a format such as TIFF. Also, if the image is very large and not tiled, it is recommended that you convert this to a tiled TIFF format to optimize disk access. Also, when converting imagery, consider using either lossless (for example, LZW) or lossy (for example, JPEG) compression. You can choose to use a wavelet-based compression, such as JPEG 2000, but these are generally more CPU intensive to decompress while providing only marginally better compression. When converting imagery isn't an option, you can build overviews on the mosaic dataset that start at a very low pixel size (using the Define Overviews tool).

Properties or parameters to consider

A mosaic dataset can be visualized in many different ways by manipulating the properties or parameters of the mosaic dataset. The visible imagery can be controlled by altering the geometry of the footprint layer, identifying NoData values, and altering the geometry of the boundary layer.


The footprints define the extent of each image in the mosaic dataset. You can use the Build Footprints tool to modify footprints to exclude parts of images from the mosaic dataset, such as black or white borders or secure areas. Generally, footprints are modified in the source mosaic datasets and not modified in referenced mosaic datasets.


This is another way to define values in an image that you don't want included in the output mosaicked image. You can use the Define Mosaic Dataset NoData tool, which inserts the Mask function in the function chain for each image in a mosaic dataset. This can result in slower performance if there are many overlapping images. Generally, it is recommended that you modify the footprints on an image to remove data.


By default, the boundary merges all the footprint polygons to create a single boundary representing the extent of the imagery. It can have holes or be a multipart polygon. This can take time to generate. If you're adding multiple collections of imagery consecutively using the Add Raster Data To Mosaic Datasettool, you can uncheck the Update Boundary parameter, until you add the last collection. When you add imagery to a mosaic dataset, you can run the Build Boundary tool to update the boundary because this tool includes a parameter to append to the existing boundary rather than overwriting it, which can also save time.

The boundary can also be used to exclude an area of imagery in the mosaic dataset. For example, you can import a boundary polygon file that fits your area of interest, even if the imagery in the mosaic dataset covers a larger area. You can also edit the boundary using the ArcGIS Pro editing tools. If you are adding a service or other larger image to the mosaic to fill data gaps for the source imagery, you may not want the boundary recalculated to include the full extent of this image. In this case, uncheck the option to update the boundary.


Generally, if you need to enhance the imagery, compute the statistics. Statistics are maintained for each image and for the entire mosaic dataset.

If statistics exist on the mosaic dataset, ArcGIS Pro applies a stretch by default. If you don't want a stretch applied, modify a property to turn off this default by opening the mosaic dataset's properties, clicking the General tab, and setting the Source Type property value to Processed.


You may need to apply a histogram stretch to the imagery to be sure it displays well. For example, you can scale 12- or 16-bit imagery to display well using 8 bits. You can apply an enhancement to the imagery when you are adding it to the mosaic dataset by modifying the raster type properties. Alternatively, you can add the Stretch function after the imagery has been added.

Color correction

Generally, color correction is only applied to RGB imagery—either a natural or color infrared imagery product (although it can be done on multiple bands). The recommended workflow is to create a derived mosaic dataset that includes the color imagery and apply color correction to it. Use the Color Balance Mosaic Dataset tool to color balance a mosaic dataset.

Attribute fields

You can add fields to the attribute table to contain the attributes suitable for the source imagery. Some fields are imported from the imagery as defined in the raster type. When you are creating multiple source mosaic datasets that will be merged into a master mosaic dataset, you need to define consistent fields.

Some common fields you can add include the following:

  • Start_Date—As a Date field
  • End_Date—As a Date field
  • Quality—An integer or text field defining a quality value you define for each image
  • Comments—A text field with additional comments

You can add values to the fields for your overviews for better viewing and performance. These fields will be accessible for viewing and querying by users of the mosaic dataset, so you may want to limit those that are accessible. You can set which fields can be accessed on the mosaic dataset's Properties dialog box.


Overviews take time to generate and should only be created if they are needed. For example, typically you will compute overviews when creating source mosaic datasets, but you may not need to when creating a derived mosaic dataset. You can also use lower-resolution images or services as a low cell-size data source, removing the need to generate overviews.


If the spatial reference systems of the data and the mosaic dataset are based on different spheroids, you may need to specify a specific geographic transformation. You can specify the transformation in two locations. When adding imagery to the mosaic dataset that has a different datum than the mosaic dataset, set the Geographic Transformation value on the Environment Settings dialog box. If you know the user or application will use a different datum than that of the source imagery or mosaic dataset, open the mosaic dataset properties from the Catalog pane, click the Defaults tab, and set the Geographic Coordinate System Transformation property.

Mosaic dataset examples

The following are examples of typical mosaic datasets, along with details for specific properties or considerations:

  • Natural color imagery
    • Create a 3-band, 8-bit mosaic dataset.
    • Include color and panchromatic imagery.
    • Use the default mosaic method By Attribute to display the latest and best quality on top.
    • Use the default compression: JPEG with 80 percent quality.
    • Add color correction if needed.
  • Color infrared (432) imagery
    • Create a 3-band, 8-bit mosaic dataset.
    • Add all the bands and apply the Extract Band function to define the 432-band combination, or define the band combination when adding the imagery to the mosaic dataset.
    • Use the default mosaic method By Attribute to display the latest and best quality on top.
    • Use the default compression: JPEG with 80 percent quality.
    • Apply color correction to remove trends.
  • Optimum image analysis and interpretation of satellite or aerial imagery
    • Create a 4-band, 16-bit mosaic dataset.
    • Use the default mosaic method By Attribute to display the latest and best quality on top.
    • Use the default compression: JPEG with 90 percent quality.
  • Multispectral imagery for analysis
    • Create a mosaic dataset specifying the number of bits and bands equal to the maximum of the imagery.
    • Use the default mosaic method to display the latest or best quality on top.
    • Use LZW as the default compression for analysis.
    • When referencing source mosaic datasets, exclude the pan-sharpened imagery and redefine the MinPS of the multispectral imagery to equal 0. This ensures that pan-sharpened imagery is not used for analysis.
  • NDVI—Normalized difference vegetation index with a color table
    • Create a 3-band, 8-bit mosaic dataset.
    • Use the default mosaic method to display the latest or best quality on top.
    • Use the default compression: JPEG with 90 percent quality.
    • Create a derived mosaic dataset from the multispectral imagery for analysis mosaic dataset.
    • Add the NDVI function to the mosaic dataset.
  • Surface or ground elevation orthometric (above sea level) heights
    • Create a 1-band, 32-bit mosaic dataset.
    • Display the most accurate on top by creating a field in the attribute table to identify this value and set the default mosaic method to By Attribute.
    • Use LZW as the default compression.
    • Include a low-resolution DEM or elevation service as a background source for areas missing elevation data.
  • Ground elevation ellipsoidal height
    • The properties are the same as the ground elevation orthometric mosaic dataset.
    • Most elevation data is orthometric. There are some requirements for accurate ellipsoidal service (for example, for accurate orthorectification of satellite imagery). To create these mosaic datasets, apply an accurate geoid to the orthometric mosaic dataset. See Converting from orthometric to ellipsoidal heights.
  • Slope in degrees of ground elevation
    • Create a 1-band, 8-bit derived mosaic dataset based on the ground elevation orthometric mosaic dataset.
    • Use the default mosaic method to display the best quality on top.
    • Use LZW as the default compression.
    • Add the Slope function to the mosaic dataset.
    • Quantize to an accuracy of 1 degree.
    • For some applications, defining a mosaic using a floating-point pixel type is better.
  • Aspect of ground elevation
    • Create a 3-band, 8-bit derived mosaic dataset based on the ground elevation orthometric mosaic dataset.
    • Use the default mosaic method to display the best quality on top.
    • Use LZW as the default compression.
    • Add the Aspect function to the mosaic dataset.
  • Hillshade of ground elevation
    • Create a 1-band, 8-bit derived mosaic dataset based on the ground elevation orthometric mosaic dataset.
    • Use the default mosaic method to display the best quality on top.
    • Use the default compression: JPEG with 80 percent quality.
    • Add the Hillshade function to the mosaic dataset.
  • Shaded relief of ground elevation
    • Create a 3-band, 8-bit derived mosaic dataset based on the ground elevation orthometric mosaic dataset.
    • Use the default mosaic method to display the best quality on top.
    • Use the default compression: JPEG with 80 percent quality.
    • Add the Shaded Relief function to the mosaic dataset.

Related topics