Camp Roberts August 2009 Report

Nangarhar PEAK Experiments at Camp Roberts, August 2009

>em>Background: The Experiments Focused on Two Big Problems in Crisis Response, Stability Operations, and Building Partner Nation Capacity: (1) Facilitating Situational Awareness (SA), and (2) Sharing it Under a Wide Range of Conditions

Staff members of organizations in the field understand that they must be aware of the situation around them. Information about critical activities—like the routing of supplies, public health, and emerging threats—is key, both to ensuring their team members’ security and to assisting stressed populations. Yet this situational awareness is all-too-often partial and not actionable, both during crisis response operations and stabilization and development activities. There are two major reasons for this problem: bandwidth limitations and information sharing barriers.

1. Bandwidth Limitations: In the field, bandwidth for communications is usually constrained and often completely unavailable. Where wired networks exist, transmission lines often fail. Access to satellite bandwidth is very expensive, often costing $7 per megabyte (one large NGO spends $250K per month on bandwidth). In this context, Short Message Service (SMS)/text-messages and physical exchange of memory devices (aka, ‘sneakernet’) often are the only reliable, or affordable communications. However, most modern software is built for always-on Internet connectivity and is not equipped to enable such “ancient” and low bandwidth modes of communication.
2. Information Sharing Barriers. Technical barriers are not the only problem. Participants in crisis response operations, as well as reconstruction and development, routinely withhold information from each other. Militaries often have access to recent aerial or overhead imagery of hot spots as well as real-time images from unmanned aerial vehicles (UAVs), but have generally not shared these images with the civilian participants in the missions. NGOs have access to data from their fieldwork at exceptionally high resolution—sometimes monitoring the health of specific households or micro-financed businesses over time—but rarely entrust their compatriots—military or otherwise—with those data. The result is a fragmented picture of the theatre of operations for everyone involved.

Improving Situational Awareness through Low-Bandwidth Solutions

In early August 2009, a team of geographers, software developers, government employees, and NGO field staff members convened at the Naval Postgraduate School’s (NPS) testing facility at Camp Roberts in California to confront some of these problems. They focused on two scenarios—stability operations in Afghanistan and planning for a natural disaster in Central America. The team (1) experimented with new ways to create situational awareness assuming only the availability of SMS/text-messages, (2) rationed use of low-bandwidth satellite connections, and (3) used a deployment kit that included a new archive of sub one-meter (1m) imagery that the National Geospatial Intelligence Agency (NGA) made available in July 2009 for participants in US and coalition operations. The goals were to show what kind of situational awareness could be built quickly when militaries and civilian organizations work together and use simple, freely available, open-source tools in conjunction with the common commercial tools in use by the government and NGOs.

The team included a mix of thought leaders from the open-source software community, industry, the military, and NGOs that provide humanitarian information technologies: OpenStreetMap, Walking Papers, Google, InSTEDD, Development Seed, Sahana, GeoCommons/FortiusOne, TerraPan Labs, the Naval Postgraduate School’s “Hastily Formed Networks” Lab, and the San Diego State University Visualization Lab, as well as observers from the Harvard Humanitarian Initiative, the National Defense University (NDU), and the Federal Emergency Management Agency (FEMA).

The team approached the challenge of developing situational awareness in five linked activities:

1. Base Maps: Processing the NGA imagery down to base maps that could be shared across all partners working the field. Importantly, the smallest units of these base maps—called tiles—had to work locally, requiring no external calls to the Internet.
2. Collection of Near-Real Time Imagery by UAVs: Extending base maps by the ultra-high resolution images that can be captured by vehicles ranging from low-cost, low-endurance civilian unmanned aerial vehicles to high-cost, high-endurance military UAVs.
3. Data Visualization on Base Maps: Integrating data sets from the field work of multiple NGOs into the new base maps. This visualization would show both how to integrate data from different organizations into a combined view and how data overlays would appear on the new NGA imagery. For purpose of this discussion, the combined visualization—the three-dimensional NGA map plus the data from NGOs—is called a situation map.
4. Integration of SMS Reports into the Situation Map: Developing a means to submit SMS messages with a specific syntax that includes a geocode (a reference to a map coordinate like lat-long). This channel would create a means to map incoming reports from field workers and trusted members of the affected population. This capability is critical when disasters diminish bandwidth and reduce the mobility of the sneakernet, as well as when organizations wish to receive reports from the field in near-real time. Currently, most field surveys are collated when teams return from their sorties, which might be many hours or even days after they leave their headquarters.
5. Sharing of Situation Maps over Low-Bandwidth Networks: Providing situation maps, or at least subsets of the maps, via SMS and mapping servers that could be run on USB sticks and updated quickly if and when Internet connectivity could be established.

Through the dedicated work of the team, most goals were realized. The team built base maps from a drive of Afghan imagery supplied by NGA; visualized data from across Nangarhar province; integrated two methods of collecting data via SMS; and proved that the maps could be shared (and updated) when placed on a USB-stick based mapping application. The following section explains the capabilities that were created and their potential field applications.

Building New Base Maps

The cornerstone of the work was an experiment built around NGA’s recent reviews of the imagery license terms for civilian partners in Afghan and disaster relief environments. After several years of dedicated policy work, NGA created a means to release imagery legally at resolutions of less than 1m to NGOs and other participants in activities which directly benefit the interests of the US government. These experiments were the first test run of this imagery release process. Test data included seven cells of 0.8m imagery from eastern Afghanistan, which is approximately 19 times the resolution previously available to civilian participants. Similar data could be provided for disaster relief in Central America. An example of the difference appears in Figure 1, with the old imagery on the left and the new imagery from NGA on the right.

(Figure 1)

Using the Google Earth Fusion server, the team combined the raw NGA overhead imagery with elevation data, thereby creating small three-dimensional tiles that could be hosted on local laptop servers and viewed via three-dimensional flythroughs.

Based on these tiles, OpenStreetMap developed a paper-based method for sharing the imagery with local partners, including farmers, called Walking Papers. This method overlays a simplified grid coordinate system over a base map and enables anyone to add pen-on-paper annotations that can later be scanned back into the GIS application. Walking Papers allows users to download any part of a map into a printable file, which includes a bar-code that ties the paper map to its coordinates in the GIS application. The integration of Walking Papers may prove to be the biggest win from the experiments. This method offers a way to map field data even when network connections are down. It will also enable even illiterate peoples to point to spots where important projects should be built (or identify where potential projects might encounter issues).

Extending Base Maps with UAV Imagery

The team also explored the use of unmanned autonomous systems (UAS), the airborne version of which are often called UAVs. This is critical for many parts of Latin America, where cloud cover often precludes the collection of useful satellite imagery. These platforms can let organizations in the field gain near-real-time situational awareness when operating beneath cloud cover, as well as in situations which require imagery faster than the 24-hour, or longer, delays typical for overhead data downloads. UAVs can also collect imagery at resolutions under 10cm, depending on altitude and the camera used. The team’s work centered on processing still images taken from low-cost aerial systems, like the TerraPan UAS—a sub-$10K UAV capable of taking 1cm imagery when flying at 500 feet. Although insurance issues and customs delays on critical equipment preventing the team from flying the micro-UAVs during this set of experiments, TerraPan and SDSU Visualization Lab teams provided archived imagery and advice that still enabled the team to test a proof-of-concept model.

The team learned several critical lessons from this work. First, low-cost UAVs have very short endurance times, often less than 30 minutes. To be useful, they must be flown to a specific point of interest and quickly collect imagery from as many angles as possible. To enable this type of close analysis, larger, long-endurance UAVs (or recent overhead imagery) must enable analysts to identify points of interest, at which point a small fleet of micro-UAVs can be deployed for further imaging. This mashup will be explored in the November round of tests.

Second, the team found new lightweight communications payloads that might enable UAVs to serve as network hubs that could provide near-constant Internet bandwidth to disconnected field workers and crisis response teams. This new capability will be explored in a civilian context during a future test, hopefully in November 2009 or February 2010.

Visualizing NGO Data on New Base Maps: Birth of Situation Maps

Integrating NGA imagery of eastern Afghanistan with data collected independently from the region proved to be a challenge. The data varied widely in resolution: some allowed determination of the productivity of individual trees in orchards throughout Nangarhar; others could be used to map ethnic boundaries or mark the locations of health clinics and sites for polling in the upcoming national elections.

The GeoCommons and Development Seed teams worked with the data to create new layers that could be placed on top of the basemaps from NGA and then be extended to include new maps based on the external elevation data. The resulting visualizations combined information from myriad NGOs working across the province. An example of a map of Nangarhar’s ethnic boundaries with ID reports, roads, news reports, airports, and polling locations appears in Figure 2:

(Figure 2)

Integrating Situation Maps with Low-Bandwidth Tools like SMS

Situations change in real time based on events in the field and situation maps should reflect these changes as quickly as possible. To accelerate mapping of new reports from staff in the field (as well as to enable crowdsourced reporting via social media as seen in Mumbai and Iran), the team integrated SMS reports into the situation map through two platforms: InSTEDD’s GeoChat and Development Seed’s SlingshotSMS.

These applications work similarly. By registering a phone with the mapping application, a user may submit reports using a special syntax that is already gaining popularity in social media: location * tags * message. For example, the following message,”grid_C32_54 * incident, firefight, wounded * Observed firefight with wounded persons treated by ICRC,” translates into a report that states “at map coordinate C3.2 x 5.4, a staff member observed a firefight with wounded persons treated by ICRC.” The report would be plotted on a mapping layer conflicts, marked as an incident, and categorized under firefights, with wounded. Figure 3 gives an example:

(Figure 3)

These two applications—GeoChat and Slingshot—give fielded organizations a means to track ongoing incidents quickly without having to transcribe push-to-talk radio conversations or wait for teams to return from the field. Importantly, it also creates a means for other organizations to subscribe to the data feeds, akin to the way in which individuals subscribe to RSS feeds from other news sources. In this way, multiple organizations can cross-subscribe to the feeds coming in from their field staffs and create composite layers from all activities in a region. In some cases, separate mechanisms maybe needed to determine the accuracy of the incoming reports.
Through these feeds, the Sahana team was able to integrate reports into their disaster management platform, showing both the location of incidents and their relationship to the “who needs what, why, where, and when” visualizations in their application.

Sharing Situation Maps over Low Bandwidth Networks

Sharing situation maps within and across organizations raises numerous issues aside from usual questions of information security. More fundamentally, getting maps out to field staff in a useful format over low bandwidth connections is a technical hurdle that precedes any policy discussion. The team intended to remove this hurdle to enable the policy discussion to proceed without excuses related tog technical difficulties, and partially succeeded.

Sahana packaged the situation maps created in OpenStreetMap with their software onto both ultraportable, disconnected netbooks and OLPCs (One Latop Per Child). These platforms created a disconnected copy of the maps and enabled field staff to collect data through Sahana that could be synchronized with a primary instance of Sahana, as well as with a local copy of OpenStreetMap, which in turn synchronized the data with InSTEDD’s GeoChat and the aforementioned GeoCommons/ Development Seed visualizations.
Development Seed displayed a proof-of-concept USB map-on-a-stick, which packaged the processed NGA tiles of Nangarhar into low-resolution images that could be overlaid with the NGO data sets from the region using only one web HTML file. This application created a means to abstract the data from its visualization using an incredibly lightweight tool—a free web browser—as the only required software. It also created an opportunity to move situational maps quickly into the hands of field workers solely by making and updating USB memory sticks. Restrictions on using USB devices on DoD computers are recognized, but not all organizations have such limits.

The USB mapping sticks provide another opportunity: the ability to synchronize data between laptops via SMS. Using InSTEDD’s Mesh4x platform, which uses SMS messages to synchronize data sources, a scenario can be imagined where field workers could submit reports both to a central hub and to anyone that they allow to cross-subscribe to their data feeds. Real-time updates in a peer-to-peer fashion are now a real possibility. This functionality will be explored in the future.

Conclusion

These experiments did not develop operational code, only proofs of concept. The tools have been put into the hands of several testers in the field to get their feedback, so the tools can be hardened and made ready for the rigors of real-world use. Output from the experiment is currently in Jalalabad, Afghanistan, where Todd Huffman is working with NGO partners to examine how closely the results meet their needs. The team will reconvene in November to build out ideas and move towards a deployable package for rapid situational awareness.

Additional Resources

Several of the organizations that participated in the Nangarhar PEAK experimentation at Camp Roberts have posted their results on their blogs. As of the date of this report, the following blog posts and videos are available:

Google

1. Overview of Integration/Interoperability Work at Nangarhar PEAK, with a focus on Afghanistan.
   http://geosquan.blogspot.com/2009/08/preparing-for-afghanistan-elections-and.html

OpenStreetMap & Walking Papers

1. Introduction to work at Nangarhar PEAK: http://brainoff.com/weblog/2009/08/10/1410
2. Exploration of Interoperability with other applications. http://brainoff.com/weblog/2009/08/10/1435

Development Seed

1. Drupal as a Data Integration Tool . http://developmentseed.org/blog/2009/aug/05/data-collection-simulations-field-camp-roberts
2. NGA Imagery Integration. http://developmentseed.org/blog/2009/aug/07/integrating-50-centimeter-data-national-geospatial-intelligence-agency
3. Slingshot SMS Reporting Framework. http://developmentseed.org/blog/2009/aug/13/dry-run-slingshotsms-situation-monitoring