Introduction About the Satellites Evaluating a Project Identifying the Area of Interest Image Search and Identification  A. Searching for Imagery on Google Earth  B. Searching for DigitalGlobe Imagery  C. Searching for GeoEye Imagery  D. Searching for ImageSat Imagery Image Ordering Image Analysis  A. Image Preparation  B. Damage Assessment  Report Generation Conclusion 
High-resolution satellite imagery can be used for human rights-related documentation, monitoring, and advocacy efforts. Imagery is particularly useful for assessing the extent of violent conflict, forced displacement, and other human rights concerns in remote, inaccessible or otherwise tightly controlled areas of the world, such as Burma and Sri Lanka. As the imaging capability of high-resolution satellites has developed over the years, so too has the power to analyze the impact of conflict on infrastructure and other features identifiable from imagery. In addition, the decreasing cost of geospatial technologies and increasing availability of geospatial data have made high-resolution imagery analysis a viable research tool for human rights organizations.
Satellite imagery analysis projects typically involve documenting the damage to an area after an event has occurred. For example, in 2009 AAAS looked at imagery of the Civilian Safety Zone (CSZ) in Sri Lanka after government forces had defeated the Tamil Tigers. Comparing images of the area before and after it was declared a CSZ revealed fresh bomb and shell craters and damage to rooftops. This type of evidence gathering is helpful in corroborating NGO and eyewitness accounts of events on the ground and has been done in Burma, Zimbabwe, and elsewhere.
Imagery analysis can also be used to monitor inaccessible or prohibited areas where an event has yet to occur but the potential for conflict is high. In these applications, an area of interest is monitored by continuously gathering and updating imagery. Eyes on Darfur , a joint project of AAAS and Amnesty International, is one example of an active monitoring project. The site has imagery of Darfuri villages that are vulnerable to attack, and continues to be regularly updated as needed with new images to assess recent changes that have occurred. While monitoring a potential conflict area by acquiring new satellite imagery on an ongoing basis is ideal, doing so can be quite expensive.
Eyes on Darfur  also shows how imagery can serve as an effective advocacy tool by starkly illustrating the impact of armed attacks on civilian populations. Satellite images also lend themselves well to publication by news organizations, investigative reports and even legal proceedings because they can objectively substantiate eyewitness accounts and other reporting. In addition to violent conflict, imagery can also be used to assess other issues, from deforestation to indigenous land rights. While conducting imagery analyses may appear daunting, this guide is intended to break down the process of identifying, acquiring and analyzing imagery into concise steps with an emphasis on minimizing costs.
Most commercial high-resolution imagery comes from satellites operated by DigitalGlobe , GeoEye  and ImageSat International . Each company operates satellites with less than one meter panchromatic (black and white) resolution, with some also capable of less than two meter multispectral (color) resolution. For example, DigitalGlobe's WorldView-2 satellite has 50 centimeter (cm) panchromatic and 1.84 meter multispectral resolutions. This effectively means that objects larger than 50 cm will be detected by the satellite. Each image produced by the satellite is made of millions of pixels, each representing a 50 cm by 50 cm square surface of the ground. This level of resolution is ideal for analyzing conflict areas, where small houses and other structures are often destroyed by violence. Most current satellites, however, cannot capture individual people because their dimensions, when viewed from above, are smaller than most imaging resolutions. Occasionally images are seen where people, or more likely their shadows, are visible as single pixels, but this is the exception rather than the norm. Knowing the limits of satellite imaging capability will be useful in identifying whether a project is suitable for this type of analysis.
Many human rights issues, but not all, can benefit from the application of satellite imagery by human rights groups. Events that involve person to person violence are unsuitable for imagery analysis due to the inability of most satellites to capture individuals on the ground. However, larger-scale events during which homes are burned, artillery or bombs are used, or large graves are dug are more likely to be appropriate for imagery analysis. For some examples of situations where satellite imagery proved useful, see SHRP's Case Study Page. A few simple questions the researcher can ask to determine if satellite imagery might be helpful include:
If the answers to the above are all yes, it is possible that the human rights project is suitable for imagery analysis. Researchers will then need to work through the process of identifying areas of interest, searching for and locating relevant satellite imagery, and acquiring and analyzing that imagery. These steps are described in greater detail below.
As a first step in any analysis project, it is necessary to define the boundaries of the area of interest. To do so, researchers should search for the names of towns, rivers, landmarks, and other features from the sources that describe the events in question. One should focus on areas where an event is most certain to have taken place and establish a timeframe of its occurrence. Those placenames that define the area of interest will be used to search for and order imagery, and must be associated with specific geographic coordinates.This process, known as "geocoding," can be done using the web, traditional maps, Google Earth, "fuzzy matchers," user- contributed mapping, and other methods, as described below.
An area of interest can be defined by either a point or a polygon. Defining the area as a point using a single coordinate pair is useful when the place is small and precise, such as a secret prison facility. But if an entire town has been burned, or if multiple villages have been shelled, creating a polygon outline of the area (as either a Google Earth KML or GIS-compatible shapefile) is more useful. Because imagery can be time-consuming to analyze and since prices are based on the number of square kilometers purchased, it is usually beneficial to minimize, to the extent possible, the size of the area of interest.
Google Earth  (GE) is a free, stand-alone desktop application that provides satellite imagery of much of the world and is useful for identifying an area of interest. To determine the coordinates of an area, zoom in to the region using either the mouse's scroll wheel or the slider on the right side of the map window. Alternatively, enter the placename into the "Fly To" box of the Search pane and in many cases Google Earth will locate the area automatically. Then, make sure the status bar, displaying latitude, longitude, and elevation, appears along the bottom edge of the map window. If it does not, click View < Status Bar. The bar displays the location of the mouse cursor on the globe. If the location of interest is small, such as a single building, zoom in as far as possible and record the latitude and longitude from the status bar. This point can be used to search the archives of the satellite companies.
If the area of interest is large and imprecise, the researcher may find it useful to create an outline of it, which can be done by selecting "Polygon" from the "Add" menu or selecting the "Add Polygon" button in the Google Earth toolbar. A dialog box will then appear, allowing the user to create a polygon by clicking to create new vertices. Once the polygon has been completed, click on OK in the dialog box. If the user wishes to edit the polygon, simply right-click and select Properties to make changes. After creating the area of interest, save the polygon as a KML by right-clicking the item in the Places pane and choosing "Save Place As." In the resulting dialog box, change the "Save as type" to KML and hit Save. KML files can be used to search the GeoEye archives. To search the DigitalGlobe archives using the area of interest, the KML file will need to be converted to a shapefile. The benefit to creating a polygon of the area of interest as opposed to simply recording a centerpoint is that the imagery search is limited to the outlined region. More information about searching the archives is included below in the Image Search and Identification section.
If a placename cannot be located using Google Earth, consider using a fuzzy matcher . This tool searches a large database for any placenames that match or are very similar to the placename that was entered. The latitude and longitude of each match are then displayed. Drawing on placenames from the US National Geospatial-Intelligence Agency GEOnet Names Server, Russian topographic maps, NGOs, and other sources, fuzzy matchers have the potential to locate places in even the most remote areas.
When using the above methods fails to locate an area, web searches often yield useful results from news articles and blog posts. User-contributed mapping websites like Wikimapia  and OpenStreetMap , can also provide geographic information. Villages that are too small to be identified by Google Earth and fuzzy matchers are sometimes identified by contributors to these sites.
Once the geographic coordinates of an area of interest have been established, the search begins for available imagery. To document an event, the researcher will generally need to acquire two images, one taken before the event and one taken after. Sometimes more images are needed to show changes over longer periods of time, and likewise sometimes a single image is all that is needed.
When searching for imagery, it is important to note that some satellites take panchromatic images while others take both panchromatic and multispectral images. In general, purchasing a slightly more expensive multispectral image, which includes the panchromatic image, is preferred because of the added color information that is provided. However, some projects may require the best possible resolution, which depending on availability, may require purchasing from a panchromatic-only satellite.
One drawback to multispectral images is that they have a much lower resolution than panchromatic images. GeoEye-1, for example has a multispectral resolution of 1.65 meters, but a panchromatic resolution of 0.5 meters. This can be remedied through "pan-sharpening," by which the color information of a multispectral image is applied to the black and white image of the same area. The result is a multispectral image with the resolution of a panchromatic image. While this does not produce the same image that a 0.5 meter multispectral sensor would have, it does approximate one.
Multispectral images vary in their number of bands. Three-band images contain the red, green and blue bands that constitute a color image. Four-band multispectral images include the red, green, blue and near-infrared bands. The near-infrared band identifies vegetation very easily, making it possible to distinguish trees from structures and other elements. Purchasing this extra band may be beneficial depending on the project at hand. When purchasing a multispectral image, the number of bands desired, whether three or four, must be specified.
With this information in mind, imagery can be located by searching the commercial satellite imagery archives and also sometimes Google Earth. Generally, the first step in the search process is to look for any free high-resolution imagery on Google Earth.
To begin the search, enter the coordinates or placename of the target area in the "Fly To" field or simply zoom into the proper geographical extent in Google Earth. If a high-resolution image is available, the date it was taken will appear along the bottom edge of the viewer. The user can check for older images by clicking View < Historical Imagery or by hitting the Historical Imagery icon in the toolbar. The slider that appears at the top will allow the user to view older images of the area of interest, if any are available. For areas that are not covered by high-resolution imagery, low-resolution Landsat imagery may be available. While Landsat is generally unsuitable for human rights analysis, the imagery may prove useful depending on the project. Commercially available imagery from DigitalGlobe can be searched by going to the Layers pane and selecting More < DigitalGlobe Coverage. This method is further discussed in the DigitalGlobe section below.
When an image has been found, a snapshot can be saved by going to File < Save < Save Image. This process is most effective if the area of interest is small, though multiple small images can be stitched together with other software to cover wider areas. To obtain a high-resolution image of a larger area, Google Earth Pro  may be required. GE Pro allows the user to save images at a higher resolution (4800 pixels) than the standard version. A license for the program must be purchased, and GE Pro includes the ability to measure area, integrate GPS data, and import shapefiles.
Should a full-quality, georeferenced satellite image be required for use in a geographic information systems (GIS) program, then purchasing or otherwise acquiring the original image will be necessary. To do so, the user will need to search the archives of the major satellite companies. A list of high-resolution commercial satellites is below:
The online archives of the commercial companies can be searched by specifying a coordinate point, drawing an area of interest using their map interfaces, or uploading a shapefile or KML. Specific information regarding the satellites, imagery, and search processes of DigitalGlobe , GeoEye , and ImageSat International  are detailed below.
ImageFinder is DigitalGlobe's searchable archive of imagery from its QuickBird, WorldView-1 and WorldView-2 satellites (see Table), and allows the user to narrow searches based on date, off-nadir angle, and cloud cover. Areas of interest can be defined in ImageFinder by entering coordinates, uploading a shapefile, or drawing an area of interest, as outlined below. Additional help can be found at this link.
Method 1: Enter center point coordinates
Method 2: Upload shapefile
Method 3: Draw Area of Interest
Following this, the ImageFinder will display the area of interest using a red outline on a dynamic map. The latitude, longitude, and square area of the area of interest will appear along the right in the Map Status box. Please note that the minimum order size for DigitalGlobe archival imagery is 25 square kilometers, which is often sufficient for the purposes of human rights research.
Additional search criteria can be specified by clicking the "Modify Filter" button in the "Search Filter" box. Set the Maximum Cloud Cover to 100% and "Maximum Off Nadir Angle" to 45 degrees in order to view all images acquired of the area. Clicking the "Search" button will open a new window displaying the results of the query. The list can be sorted by any of the columns in the table by clicking on the appropriate heading. Images can be previewed in either a new window (by clicking the "view" link in any table entry) or on a map (by clicking the check-box located under the heading labeled "browse image").
After an appropriate image is identified, record the Image ID and Date of Acquisition for use when ordering from a reseller. A shapefile of the area of interest will also be needed. By going to Download < Polygon Shapefile, the area of interest can be downloaded as a zipped shapefile and passed on to a reseller.
DigitalGlobe images can also be searched using Google Earth. In the Layers sidebar, click More < DigitalGlobe Coverage. This will display the footprints of all available images in the area, as depicted in the screenshot below.
GeoFUSE  is GeoEye's online archive of imagery captured by the IKONOS and GeoEye-1 satellites. Using this interface, an area of interest can be specified in several ways as described below. Further help can be found here .
Method 1: Enter center point coordinates
Method 2: Upload shapefile/KML/KMZ
Method 3: Draw Area of Interest
The resulting map will show the area of interest and outlines of the images that match the specified criteria. Clicking the Search Preferences icon allows the user to change the maximum cloud cover and date range. After adjusting the criteria, the search results can be downloaded as either an SHP, KMZ, CSV (Excel), or HTML file by clicking the Data Download icon at bottom. Please note that for GeoEye images, a minimum area of 49 square kilometers must be ordered. For the user's convenience, the approximate square area of the area of interest is shown on the bottom bar of the map window.
An image can be previewed by toggling the box next to the footprint icon in the Image Catalog Results pane. This will display the image directly on the map. An image can also be previewed by selecting "Details," then hitting "View full image metadata" in the popup, and finally scrolling down to "Image File URL" in the new window.
GeoEye's archive is also searchable in Google Earth by downloading a KML file that interfaces with GeoEye's servers. Upon opening the file, the KML will appear in the "Places" pane of Google Earth. Clicking the "Search GeoEye Image Catalog" logo that appears will open a window containing the available cloud-cover options, shown below.
Once the search is complete, the date range can be changed using the slider that appears at the top of the map window. The outlines of any imagery that matches the search criteria will then be displayed on the map. Clicking on the image icon will bring up detailed information about the scene along with an option to view a preview of the image.
The image preview is useful for determining whether clouds are present over the area of interest. If the search results yield a useful image, the imagery source, collection date, and ordering identifier should be noted for ordering.
GeoEye also offers the GeoFUSE toolbar  for ESRI's ArcMap  GIS that allows for a simple search of available imagery. Use of the toolbar is similar to the online interface in that the user must specify an area of interest, either by coordinate values, drawing a polygon, or selecting a feature. Search results can also be refined by cloud cover, off-nadir angle and date of acquisition. The toolbar is geared towards those familiar with GIS applications.
Another source of imagery is ImageSat International, which operates EROS-B, a satellite with 0.70 meter panchromatic resolution. Unlike the other two commercial vendors previously discussed, ImageSat does not have a searchable imagery archive online. To request to see available imagery, users should send coordinates and a timeframe to firstname.lastname@example.org.
When an image has been identified, AAAS approaches a third party reseller to submit the order. Communication with these resellers takes place primarily via e-mail, in which the area to be ordered is specified by attachments of WGS-1984 compliant shapefiles in geographic projection. This, along with catalog ID, imaging platform, and acquisition date, constitutes more than enough information for an order to be submitted. Processing usually takes one to three days, after which the requested data are delivered electronically via FTP.
In the event that no archival imagery exists of a high-priority target, the satellite may be tasked to acquire a new image at the user's request. This option is substantially more expensive than ordering imagery from the archives, but the process is much the same, with shapefiles or KMLs of the relevant area submitted to the reseller, along with the requested parameters regarding viewing angle and cloud cover. Note that imagery resellers will also often help you work through any image location or ordering problems, so do not hesitate to approach them with your questions. A list of DigitalGlobe's authorized resellers is here , and GeoEye's is available here . Below is a chart listing the price per square kilometer and minimum order size for each satellite.Company Satellite Minimum Order Price (Archival) Price (New Collection) DigitalGlobe QuickBird 25 km2 $17/km2 $23/km2 DigitalGlobe WorldView-2 25 km2 $17/km2 $23/km2 DigitalGlobe WorldView-1 25 km2 $14/km2 $20/km2 GeoEye GeoEye-1 49 km2 $12.50/km2 (>90 days old) $25/km2 GeoEye IKONOS 49 km2 $10/km2 (>90 days old) $20/km2 ImageSat International EROS-B 49 km2 $10/km2 $14/km2 (estimated)
After ordering the imagery, keep receipts and records organized for future reference. The order will likely be filled as an email linking to an FTP site from which the images can be downloaded. Depending on the type of internet connection, a download may take up to several hours due to the size of the image files being transferred.
After the order has been filled and the image received, the researcher needs to start viewing the imagery in an appropriate software package. Various options exist for viewing imagery, from GIS programs like ESRI ArcMap  and the free Quantum GIS , to software designed for imagery analysis such as ERDAS Imagine , Opticks  and MultiSpec . If these programs are unavailable, the TIFF image files can even be read in image viewers like Paint or Adobe Photoshop. However, GIS and image analysis programs are definitely the preferred option to fully take advantage of the georeferencing included with the image.
When the images have been loaded into the appropriate software, the analyst should visually scan each one to determine if it is suitable for analysis (i.e. correct location, cloud cover, resolution, etc.). Note that many satellite images are comprised of multiple bands, which indicate reflected wavelengths of light captured by the satellite sensor. A full discussion of this aspect of satellite imagery analysis is not possible here, but readers can review other information sources on remote sensing and imagery analysis. Generally, for multispectral images, a red-green-blue band combination will yield a naturally colored image, allowing for easy image interpretation. In other cases, highlighting the vegetation is preferred, and the near infrared band can be displayed as a primary color, thus allowing for a false color representation that makes vegetation easily distinguishable, as seen below.
Images courtesy http://rangeview.arizona.edu/Tutorials/intro.asp
On the left is a natural color image displaying the red, green, and blue bands. The image on the right is color-infrared, and displays the near-infrared, red and green bands. The infrared band helps to easily identify vegetation.
For all imagery, especially panchromatic imagery, a stretch will usually need to be applied  to best display the image. Although the type of stretch will vary according to image, typically a standard deviation stretch or maximum-minimum stretch will produce the most ideal results. These stretches can be applied to the entire raster image or just the visual extent of the data frame. Results will often vary dramatically between the two extents and so both should be tested. It is important that the viewer of the imagery be aware of the adjustments that are made to the imagery. Too much alteration can be misleading, while insufficient fine-tuning can obscure important details of the image. If multiple images are being used, determine whether the images are georeferenced well to each other or if co-registration is needed. In other words, a house in one image should be in the same location as that house in another image. A few meters off in one direction is common, but images that are skewed by 15 meters or more will adversely affect the analysis. Image georeferencing is a significant sub-topic of remote sensing, and due to various factors such as satellite vibration and look-angle there is almost always some error that must be overcome. To remedy this, GIS and image analysis programs will have georectification tools that allow the user to manually align images by marking points of reference. An explanation of the process for ArcMap can be found here . Properly georeferencing an image will reduce errors and lead to a more efficient analysis, though perfect co-registration of multiple images is usually impossible.
At the beginning of the analysis, survey each image to understand its individual characteristics. The images may not share the same resolution, color, or clarity. Making note of these differences will prevent the researcher from marking change where none has actually occurred. For instance, a structure that appears blurry in an after image may be the result of a lower- resolution scene, a temporary dust or moisture cloud, or some other factor, rather than damage.
AAAS primarily uses ERDAS Imagine for damage analysis. With this program, two images can be displayed in adjacent windows and "geolinked" such that scrolling through an image simultaneously scrolls the other. This allows for side by side analysis without having to toggle layers on and off. Using two monitors facilitates the analysis process even more by allowing the maximum area of the before and after images to be displayed. Due to the expense of the program and additional monitor, however, ERDAS may not be a viable option for everyone. Other GIS programs like ArcMap and Quantum GIS work well for image analysis as well. Since image layers are stacked in these programs, the researcher will need to switch layers on and off to conduct the analysis. Even Google Earth can be used in such a fashion, though the imagery will need to be converted to KML overlays first.
When comparing images, create a shapefile or other data file to mark where change has occurred on the after image. A point shapefile is useful when counting the number of individual structures that are damaged. However, when presenting the results of the analysis for an audience, a polygon shapefile can also be useful; this will outline the affected areas without obstructing the view of the image, as many dots would do. With the shapefile created, begin the assessment at one corner of the image and scroll in one direction until the other edge of the image has been met. Scrolling should continue back and forth until the entire image has been covered, all the while marking change as new points in the shapefile. This method ensures that no part of the scene goes unobserved. During the first round of image comparison, mark any possible changes that have occurred between the before and after images. During the second round, refine those areas by marking only those where the most obvious change has occurred, taking care to account for possible seasonal variation, differences in the direction of the sun, and other environmental factors.
Be sure to use any details acquired about the incident from eyewitness reports to focus the analysis on a particular area of the image. If the area of interest is vague, then using the scrolling method described above is best. In addition, using the measure tool in ArcMap , Google Earth , or another analysis program will help assess distances and areas necessary for identifying change.
There are a few issues to consider when conducting the damage analysis. The absence or presence of vegetation may be affected by the season in which the image was collected. Since obtaining images from the same time of the year is not always possible, keep in mind that trees may lose their leaves, leaves may change colors, and rivers may ebb and flow depending on the season. Understanding this will help the researcher distinguish vegetation from housing structures. In a 4-band (red, green, blue and near-infrared) image, assigning the near-infrared band to the red color channel will make healthy vegetation appear red . Distinguishing between housing and vegetation in a panchromatic image is not as clear. A tree will typically appear blurry around the periphery and have soft edges as compared to the sharp edges of housing structures. A deeper knowledge of remote sensing concepts, as discussed in this NASA tutorial , may help in the analysis of some projects.
Understanding the customs, economy and history of an area can also contribute to a good analysis. In Darfur, for instance, many homes are ring-like huts made of mud walls. If set ablaze, the straw roof of a hut burns entirely, leaving the outline of a ring (the mud walls) that is visible from satellite imagery. This information is useful for identifying destroyed homes, like those depicted in the before and after images below.
Images ©2007 DigitalGlobe Inc.
These images near Ishma, South Darfur, Sudan show that every structure and fence line have been destroyed as a result of a "road-clearing" attack. The "before" image on the left was taken December 25, 2004 while the "after" image on the right is from February 10, 2007. Notice that the houses in the after image appear as ring-like structures without roofs, which burned.
It is also vital to note that the population in the Zaghawa region of Darfur is nomadic, traveling in search of food and water depending on the season. Changes in housing structures, therefore, do not necessarily signal damage, but may instead reflect the movement of an entire community. Using this information helped inform AAAS's study of the Darfur region.<>
The type of environment where an event occurs can greatly impact the visibility of damage. In Sri Lanka, the sandy beach on the northeast coast was such that it helped indicate mortar shell impacts, artillery emplacements and individual graves from the 2009 conflict. The soft sand easily gives way to shell impacts, as seen in the image below. By contrast, areas of heavy rainfall where mud and dirt are continually being churned about will adversely affect the terrain and the ability to see much change on the ground. Black soot and charred ground resulting from burned structures, for instance, may not be preserved by a constantly shifting terrain. Although images vary and local environments change, it is important to maintain consistency in marking damage throughout an analysis.
Images ©2009 DigitalGlobe Inc. Pictured in this 2009 image are two craters formed from the impact of shells fired in close proximity to civilian areas in Sri Lanka. The craters, as well as the scattered ejecta surrounding them, are outlined well in the sandy ground.
With the image sets analyzed, create an initial report, whether as a document, poster or webpage, for dissemination to users. When conveying the results of the assessment, a methodology should be conveyed that addresses any issues that may have come up during the analysis. Display before and after images side by side and use arrows to indicate damage where necessary, making sure not to overly obstruct the image. Also be sure to attribute the images that are used to the appropriate satellite company. Because whole images are generally too large to disseminate easily, use small snapshots of the damaged areas for the report. This can be done by zooming into the appropriate area and either taking a screenshot or using the Export Map feature  to create a scaled down image TIFF or JPEG file in ArcMap, or using the Print Composer in Quantum GIS. Posters can be created using ArcMap by switching to the layout view. To do so, select View < Layout View. A tutorial about creating map  layouts in ArcMap is available here . Quantum GIS is also capable of creating map layouts. After creating the initial report, take feedback into account when finalizing the document for distribution.
This document is intended to provide the tools necessary to begin using satellite imagery to address human rights concerns. AAAS is supportive of any efforts to utilize geospatial technologies and will respond to any further queries that may come up during the process. For conflict-related uses, AAAS's case studies may prove a valuable resource.