– 3 –
FUTURE RESEARCH AREAS AND IMPLICATIONS

On the third day of the workshop, participants focused the results of the earlier discussions into a short list of the research directions they thought would have relevance and value for the National Geospatial Intelligence Agency (NGA). Participants divided into five working groups, and each group was charged with selecting, justifying, and presenting five overarching research themes that had arisen during discussion, drawing from the five core areas or the five cross-cutting themes. Each group formally presented the key themes they selected, and then participants were encouraged to eliminate duplication by grouping the topics when there was substantive overlap. This resulted in the identification of ten future research areas for the NGA (see Box 3.1).

BOX 3.1

Future Research Challenges for NGA

The ten research challenges, as selected by workshop participants, are:

  1. Visual analytics

  2. Integrating sensors

  3. Human terrain/behavior

  4. Participatory sensing

  5. Improved models of space-time

  6. Development of new paradigms for conveying certainty

  7. Improved geodetic, photogrammetric, and remote sensing positioning

  8. Geospatial information retrieval and extraction from text

  9. Database technology and spatial data infrastructure

  10. Geospatial narrative



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 21
–3– FUTURE RESEARCH AREAS AND IMPLICATIONS On the third day of the workshop, participants focused the results of the earlier discussions into a short list of the research directions they thought would have relevance and value for the National Geospatial Intelligence Agency (NGA). Participants divided into five working groups, and each group was charged with selecting, justifying, and presenting five overarching research themes that had arisen during discussion, drawing from the five core areas or the five cross-cutting themes. Each group formally presented the key themes they selected, and then participants were encouraged to eliminate duplication by grouping the topics when there was substantive overlap. This resulted in the identification of ten future research areas for the NGA (see Box 3.1). BOX 3.1 Future Research Challenges for NGA The ten research challenges, as selected by workshop participants, are: 1. Visual analytics 2. Integrating sensors 3. Human terrain/behavior 4. Participatory sensing 5. Improved models of space-time 6. Development of new paradigms for conveying certainty 7. Improved geodetic, photogrammetric, and remote sensing positioning 8. Geospatial information retrieval and extraction from text 9. Database technology and spatial data infrastructure 10. Geospatial narrative 21

OCR for page 21
22 NEW RESEARCH DIRECTIONS FOR NGA FUTURE RESEARCH AREAS Visual Analytics Visual analytics is the science of analytic reasoning facilitated by the interactive visual interface. Even though this is a new science, the potential to have a positive impact on NGA is significant, moving NGA towards providing highly precise and relevant knowledge products for their stakeholders. Each of the five working groups offered detailed suggestions for developing the science of visual analytics for GEOINT. The first working group stressed the importance of the development of the science of visual analytics and time-space narratives for geospatial sciences. This group stated that goals need to include optimizing analytical methods, decision cycles, and products through the use of fusion or synthesis; exploiting multi-source, multi-type, temporal and geospatial and dynamic active data; and developing new technologies. The second working group discussed the need to address four topics within visual analytics. These were: (1) to address cognitive issues through cognitive models of human- computer systems and extension of the theory of human computer interaction; (2) to establish methods for evaluation, validation of reasoning, and analysis techniques; and (3) to extend current techniques to address uncertainty, scale, and time series. Lastly, it was thought important to enable mobile, collaborative, and distributed interaction. The third working group suggested a set of areas specifically for integrative analytics. Their areas included developing four-dimensional space-time representation and analysis techniques; creating methods for dealing with multi-level data, heterogeneity, and uncertainty; and producing algorithms for statistical and machine learning. Interactive analytics were thought critical, that is efficiently coping with massive amounts of information and data, using visualization, haptics, sound, games, and modeling and simulation. The fourth working group provided ideas surrounding integrated, interactive, and iterative spatial and temporal (visual) analytics. First was the need for dynamic information systems for geospatial intelligence—representation, modeling, and analysis issues—to support workflows and to focus on spatiotemporal data and analysis. Second, methods were desired for social mapping (social links and networks), which requires an improvement in the spatiotemporal component of social network analysis and the identification and representation of dynamic relationships over space and time. Next, the need for true, comprehensive and complete space-time analysis to address semantic and space-time scales was stressed, using scalable algorithms and infrastructure for large volumes of data. Fourth, means were thought necessary to represent and utilize and analyze data and information quality, reliability, and confidence. This implies the need to determine (1) what information is needed by particular users and the appropriate evaluations methods, and (2) how to make information on certainty or uncertainty useful and to support reasoning with uncertainty and with heterogeneous kinds of information. Lastly, the group discussed the need to develop adaptive visual analytic methods that support a range of users, uses, and devices across a range of interaction science issues; human-algorithm interaction; speed of response to support interaction; sensitivity assessment in real time; and uncertainty representation. This led the group to state a need to develop methods to determine what visual and other analytical methods are appropriate for specific problem contexts and how the success (or lack of success) of the methods and tools need to be evaluated.

OCR for page 21
FUTURE RESEARCH AREAS AND IMPLICATIONS 23 The fifth working group suggested using visual analytics for geospatial sciences as models for incorporating human analysis (integration of human and computer intelligence), for using collaborative methods, such as social games, and interactive visual analytics, and for communicating salient heterogeneous information to the end users (more effective participatory sensing). This group suggested that important future research included computational modeling, large data organization, modeling, indexing, retrieval, visualizations, analysis and forecasting. Visual analytics was addressed in many of the other discussions and appears to have broad interdisciplinary impact on geospatial sciences and applications. The European scientific community also sees visual analytics as an important research opportunity, as illustrated by the recent Geospatial Visual Analytics: Focus on Time Workshop and keynote presentation on Visual Simulation at the 13th AGILE International Conference on Geographic Information Science (2010; See http://www.agile-online.org). Summary of Working Group Discussions on Visual Analytics Computational modeling – Large data organization, modeling, indexing, retrieval, visualizations, analysis – Forecasting Models for incorporating human analysis (integration of human and computer intelligence) – Using collaborative methods, such as social games, and visual interaction in aid of visual analytics – Communicating salient heterogeneous information to the end users (more effective participatory sensing) Developing the science of visual analytics for GEOINT – Addressing the cognitive issues cognitive models of human-computer systems extend theory of human interaction – Establish methods for evaluation and validation of reasoning and analysis techniques – Extend current techniques to address uncertainty, scale and space-time – Include mobile, collaborative, distributed interaction Integrative Analytics – 4D space-time representation and analysis techniques – Multi-level, heterogeneity, uncertainty – Algorithms: statistical, machine learning – Interactive analytics—efficiently coping with massive amounts of information and data visual, haptic, auditory, etc. games, computational modeling, simulation Integrated, interactive, and iterative spatial and temporal (visual) analytics – Dynamic information systems for geospatial intelligence – representation, modeling, and analysis issues; support workflows; focus on spatiotemporal data and analysis – Social mapping, social links and networks – improve spatiotemporal component in social network analysis; identify and represent dynamic relationships over space and time

OCR for page 21
24 NEW RESEARCH DIRECTIONS FOR NGA – True, comprehensive or complete space-time analysis does not exist; address semantic and space-time scales; need scalable algorithms for large volumes of data – The right infrastructure to deal with reliability and latency associated with massive quantities of data – How to represent, utilize, and analyze data and information quality, reliability, and confidence determine what information is needed by particular users and determine the appropriate evaluation methods how to make information on certainty and uncertainty useful. Supporting reasoning with uncertainty and with heterogeneous kinds of information Need to develop adaptive visual analytic methods that support a range of users and uses, a range of devices, and a range of situations – Human-algorithm interaction—research issues include speed of response to support interaction – Methods to support sensitivity assessment in real time – Uncertainty representation – Methods to determine what visual and other analytical methods are appropriate for specific problem contexts and how to evaluate the success (or lack of success) of the methods and tools Development of the science of visual analytics and narrative to optimize analytic methods, decision cycles, and products through the use of fusion, synthesis, multi-source, multi- type, temporal and dynamic products plus technologies Integrating Sensors Emerging sensors, such as hyperspectral and LiDAR, will provide additional information, and, in combination with traditional remotely sensed data (panchromatic electro-optical), will enable new information to be derived that could not have been derived from a single sensor. New and ubiquitous sensors can have many sensing modalities, can be networked, can provide continuously streaming data, and can be miniaturized. They will collect physical features—such as environmental, motion, chemical, and biological data—with location information of sensed data, and collect that data with varying degrees of resolution and quality. Sensors are being deployed not only on spaceborne platforms, but also on UAVs, drones, vehicles, on the ground, underwater, underground, and on humans (i.e., in backpacks and cell-phones). These sensors and varieties of platforms and modalities require new paradigms and significant research in sensor modeling, sensor calibration, and sensor data fusion as well as new methods to address the complexities of mission planning and adaptation of tasking. Workshop participants expressed the concern that the vast quantities of data collected will require the development of “smarter” processing methods, perhaps coupled with and on-board sensor platforms. Significant research will be required in spatial information generation—automated feature extraction—to deal with the plethora of collection capabilities. Participants also noted the need to develop ways to integrate data of varying quality, as well as data that is metric and non-metric, with existing geospatial information using prior knowledge and multiple sources of information and knowledge.

OCR for page 21
FUTURE RESEARCH AREAS AND IMPLICATIONS 25 Summary of Working Group Discussions on Integrating Sensors Spatially integrated sensing – across scales, (multiple hierarchy, multiple sources). How does larger scale satellite remote sensing: – Integrate participant and social data to fill holes – Deal with mixed data quality across sensors and sources; security of data sources; credibility – Including environmental sensors, etc. – Surface, sub-surface, atmospheric – Change over time integrated in space – Exploit hyperspectral imagery integrate with other data (e.g., GIS) to improve understanding of data exploit other information (e.g., culture, context) add time (as described in photogrammetry) change dynamics adaptive sensing Improved GEOINT from variety of sensors and platforms, including new platforms (e.g., nanosats, man-portable) – Improved algorithms for automation, feature extraction, etc. – New platform designs and networks – Mission planning, sensor modeling, etc. – New sensor capabilities (new designs, “smart” sensors, hyperspectral, etc.) – Calibration of sensors – Sensor fusion Advanced sensing Development of high quality, miniaturized, intelligent sensors – Novel sensing modalities: biological and chemical – Positioning in new places—underwater, underground, etc. New paradigms for calibration – Traditional and new sensing modalities Automated geospatial feature extraction and knowledge generation from integrated multi- sensor and multi-source metric and non-metric data acquisition systems and prior knowledge Heterogeneous spatial data acquisition and analysis – Using participatory sensing to leverage geo-spatial data collection – Modeling – Adaptive sensing, participatory sensing, multi-sensor, multi-platform – Quality control (e.g., best practices, benchmarking) – Issues of data provenance and privacy Human Terrain and Behavior The human terrain theme arose in a number of working groups and encompasses analyzing geospatial-based observations, algorithmic modeling, and assessing or predicting

OCR for page 21
26 NEW RESEARCH DIRECTIONS FOR NGA future social behavior, often in rapidly evolving situations. Workshop participants stated the following as key research areas within human terrain: geospatial data collection techniques for observing human behavior; geospatial integration of social, behavioral, and cultural data; and the use of participatory data. This raises policy issues for acquiring data, influencing participation, dealing with security and privacy issues, mixing participatory data with traditional data, assessing reliability or credibility, and understanding cultural and social constraints on participatory data. Participants also thought research was necessary to develop models for human behavior and interaction among individuals, organizations, networks, and communities. These models would include (1) the importance of social factors (e.g., culture, politics, history, economics); (2) human interaction with their physical environment; and (3) integration of geospatial, temporal, dynamic social network and socio-cultural factors. Some participants stated that such research can lead to predictive systems that rapidly and accurately convey certainty and reliability of predictions, that are transparent both in methodology and sensitivity to input data, and that allow rapid assessment and decision making. Also raised as an important issue was the development of standardized, quality data sets for development and testing of new theory and model algorithms. Summary of Working Group Discussions on Human Terrain Modeling human behavior – Interaction among humans (individuals, organizations, networks, communities) – Importance of social factors: culture, politics, history, economics, etc. Interaction and relationship of humans with their physical environments Development and validation of theoretically informed models Geospatial methods to analyze, model, and predict human behavior. We cannot model it. Relating social factors to physical factors (e.g., can game theory be spatialized as a method to address this?) Integrate methods dealing with human behavior into geospatial analytics. Evidence-based geospatial prediction or simulation for human behavior Geospatial-based integration of social, behavioral, and cultural data (including participatory data)—how to deal with mixed type and mixed quality data (crowd sourced data and sensor data) – Security-privacy issues—how to influence social media to generate data that is needed; how to gauge credibility, reliability, etc. – Knowledge—recognize human behavior, data repository plus expertise repository – Shared conclusions and findings—collaborative information generation and decision making – Participatory expertise – Understanding relation of cultural and social factors – Guidelines on policy and practice of collection Development of data collection techniques, analytics, forecasting, visualization, and service chains, plus theories that can simultaneously accommodate integrated geospatial, temporal, dynamic social network and socio-cultural factors for rapid social situation assessment

OCR for page 21
FUTURE RESEARCH AREAS AND IMPLICATIONS 27 Participatory Sensing Workshop participants identified the emerging paradigm of participatory sensing as an area of future research investment for the NGA. Instead of relying on unattended autonomous sensors as traditional remote sensing and embedded sensor networks do, participatory sensing engages individuals, groups, and communities in the act of collecting, analyzing, and disseminating urban, social, and other spatio-temporal information. Ubiquitous wireless data networking and sensor-instrumented mobile smart phones provide the platform on which participants can be engaged for collecting diverse geospatial information, and in ways ranging from voluntary and opportunistic sensing to directed and coordinated sensing campaigns. Human mobility and intelligence enables collection of measurements and contextual information that traditional instruments cannot easily replicate. Recognizing that both the opportunity and the challenge of participatory sensing arise from human participation, many participants identified the following key elements of a research agenda that will enable effective use of participatory sensing in GEOINT: Effectively involving human participants using mobile technologies – Methods for planning and optimizing sensing – Methods for control and creating incentivizes Addressing quality, uncertainty, and trustworthiness of participant-contributed data – Methods to cope with human bias, selection bias, competence, sabotage Responsibly involving human participants – Policy issues – Privacy mechanisms Integrating unplanned, unstructured participatory sensing data into GEOINT Incorporating prior information Summary of Working Group Discussions on Participatory Sensing Enable use of participatory sensing for GEOINT – Methods for planning and optimization – Addressing uncertainty and trust issues – Addressing policy and privacy issues – Integration and augmentation of unplanned, unstructured participatory data into GEOINT – Develop methods of incorporating a priori information Techniques for incorporating humans in the loop in the collection and processing of data – Utilizing volunteers – Making use of mobile technologies – Issues of quality (human bias, selection bias, competence, sabotage) – Mechanisms for control and creating incentives

OCR for page 21
28 NEW RESEARCH DIRECTIONS FOR NGA Improved Models of Space-Time The integration of time and space in GIS and geospatial analysis was seen by the workshop participants as key to further the representation and understanding of complex dynamic physical and socio-behavioral processes. This will require the development of new and improved models that integrate the time structure of events, as well as their aggregates and narratives, with the spatial structure. It was noted that formal models for space-time dynamics need to be refined and extended. Crucial in this is a theory of scale dependence in order to handle multiple resolution data bases and the integration of social, cultural, and behavioral factors. Related to the theoretical development is the need to represent, communicate, and visualize space-time dynamics. Some stated that progress in this direction is necessary in order to move beyond the current largely static conceptualizations and representations. Summary of Working Group Discussions on Space-Time Models Improved models of space-time – Development of a theory of scale dependence to better understand feature relationships addressing multiple scale, multiple resolution databases – Improved model of time structure (akin to space structure) with events as basic units, aggregation of events, narratives, etc. – Integrated space-time structure Four-dimensional modeling – Incorporation of space and time dynamics – Incorporation of social, cultural, and behavioral factors – Representation, communication, and visualization of time Development of New Paradigms for Conveying Certainty This topic of conveying certainty arose in almost all aspects of working group discussions as a long-term issue across all NGA core areas that requires more robust treatment. As NGA moves from traditional data sources toward more ad hoc and less quantitative data sources, participants stated that renewed or new emphasis is necessary in the following areas: the development of tools for establishing data and information quality at all stages of the information chain from collection to decision making; the creation of methods to establish reliability of participatory data; the development of methods to detect participatory data manipulation; and the means to convey reliability in visual data. Broader themes identified by the groups included better understanding human interaction using visual data, how to do statistical and semantic fusion of model source data and information with inherent uncertainties in the observational data, and how to characterize uncertainty in relationship to scale and resolution. Summary of Working Group Discussions on Conveying Certainty Developing new paradigms for characterizing uncertainty – Better methods of representing and visualizing uncertainty

OCR for page 21
FUTURE RESEARCH AREAS AND IMPLICATIONS 29 – Includes reliability, confidence, trust, error, etc. – Application to participatory sensing, etc. – Application to all NGA core areas (e.g., geodesy, remote sensing, cartography) – Including relationship to scale and resolution Fundamental tools required for establishing data and information quality and provenance at all stages of the information chain Statistical and semantic fusion of multi-source data and information with modeling of uncertainties inherent in the data and information (geospatial, text, web, Humint) Improved Geodetic, Photogrammetric, and Remote Sensing Positioning The workshop participants noted that remote sensing, geodetic, and photogrammetry data will continue to require improved positioning to be used effectively in geospatial intelligence. Improved positioning is necessary for addressing climatic issues, such as sea level rise and ice sheet changes, earthquake activity, intelligent transportation, and for high geometric accuracy associated with existing and new sensor data sets. Incorporating improved positioning from additional GNSS/GPS constellations of satellites will allow for improved positioning to the millimeter level in real time. Atomic clocks used in GPS will continue to improve in accuracy by orders of magnitude. When satellite positioning data are not available—such as in buildings, underground, or underwater—inertial navigation systems will need to be developed to high accuracy levels. Gyroscopes used to measure or maintain orientation of remote sensing devices will continue to require improved positioning. Some stated that improved gravity models are necessary to determine precise orbits and to reduce orbit errors associated with satellites. Summary of Working Group Discussions on Positioning Improved geodetic, photogrammetric, and remote sensing positioning – GNSS – Inertial navigation systems – Gyroscopes – Atomic clocks – Gravity – Geoid Geospatial Information Retrieval and Extraction from Text Workshop participants stressed the importance of developing the capability to determine and/or refine geospatial information from text search and retrieval. Methods are needed to use geospatial information to interpret unstructured and semi-structured textual information. More challenging is integrating information from a wide range of sources by anchoring them geospatially and understanding and characterizing the geographic variation of language. Some noted that additional research is required to investigate the combination of techno-social predictive analytics with infrastructure-based sensors in aiding the information retrieval and extraction.

OCR for page 21
30 NEW RESEARCH DIRECTIONS FOR NGA Summary of Working Group Discussions on Geographic Information Retrieval and Extraction from Text How to use existing geospatial ontologies to inform the information extraction process How to use formal geographic information (e.g., from mapping databases of remote sensing) to interpret unstructured, semi-structured information A challenge is to integrate information from a wide range of sources and anchor them geospatially Understand the geographic variation of language How to combine technosocial predictive analytics with infrastructure-based sensors; how to control sensors Database Technology and Spatial Data Infrastructure Participants identified database technology and spatial data infrastructure as another important area of future research investment for the NGA. The challenge arises from the increasing richness, quantity, and diversity (e.g., scale, modality, sources, and quality) of geospatial data and information products. Therefore, the participants state that research is needed to develop database technology and spatial data infrastructures that are capable of handling data that is multi-dimensional, spatially and temporally multi-scale, and multi-source, ranging from authoritative to participatory and public. The database requirements for extremely high spectral and spatial resolution, multimedia imagery and free form text, as integrated over the entire Earth, will continue to challenge most existing data schema and models. Summary of Working Group Discussions on Database Technology and Spatial Data Infrastructure Research on database technology and spatial data infrastructures to handle multi-scale, multi-dimensional, multi-temporal data of both authoritative and public participatory data over time Geospatial Narrative Many GEOINT workflows, from data collection to interpretation, can be represented as narratives, or stories, about the data or the world that the data samples represent. Narrative theory has been well developed in disciplines such as story-writing, film studies, and literary studies, and narratives are well known as being effective ways to build associations and activate memory. Many geospatial phenomena—for example tornadoes following a storm front, boats leaving and entering ports, or truck convoys moving from camp to airstrip—also follow sequences, and differences from a known narrative provide a means by which the normal can be discriminated from the abnormal. A research track in geospatial narrative would focus on how to develop computational narratives within a spatio-temporal database. It would design and build structures

OCR for page 21
FUTURE RESEARCH AREAS AND IMPLICATIONS 31 that would allow narrative objects of any type (e.g., a hurricane) to be automatically recognized and created, then manipulated for visualization and analysis. This may require multi-source data, multi-temporal interpolation and simulation, and visualization. Data could be points, lines, areas, or volumes. Outcomes would include descriptions of narrative evolution, the use of narratives in interpretation and analysis, and the ability to test a developing situation against all known similar narratives. For example, as a specific hurricane track develops, its path can be tested for similarity against all known prior hurricane tracks, and so estimates of damage, loss of life, and necessary humanitarian relief can be made from the past instances. Summary of Working Group Discussions on Geospatial Narrative How to develop computational narratives – a representation structure for narratives in a dynamic database Narrative as an object that can be manipulated (production of narrative products at multiple levels of explanation) Auto-generation of narratives from multiple sources Narrative maps to show evolution of activities IMPLICATIONS FOR THE SCIENTIFIC INFRASTRUCTURE Some discussion was devoted to the implications of the above research themes for the scientific infrastructure. In some cases, programs to promote science education and to build research infrastructure exist or are planned, such as the National Science Foundation’s Office of Cyberinfrastructure and the Defense University Research Instrumentation Program. It was felt that existing academic programs would need to respond strategically to these research challenges, and that doing so would require new centers, resources, faculty, and students. The need to protect the existing core area research programs and the need to take advantage of the increasingly global nature of graduate education were noted. Some felt that perhaps the greatest challenge is dealing with the increasing need for interdisciplinarity in research and education, since universities and programs also have a degree of intellectual inertia and reluctance to change or adapt to new technology and challenges. FINAL REMARKS The workshop examined the five NGA core areas and, in many cases, reaffirmed the importance of these areas for NGA’s mission. The first day’s presentations and working group discussions outline some of the potential areas of new research in these fields. However, the cross-cutting themes also suggest that the core areas are in evolutionary flux and that emerging fields, as discussed during the workshop, will need to be tracked and monitored. These new areas emerged clearly from the discussions. The workshop took place in a spirit of cooperation and collegiality. Each of the core areas and emerging fields was represented by world class experts, and participants were prepared for the meeting and seemed pleased with the level of interdisciplinary interaction. The committee

OCR for page 21
32 NEW RESEARCH DIRECTIONS FOR NGA felt that the record of the discussions and ideas presented at the workshop figure prominently in this report. From among these discussions, hopefully, ideas for the next generation of research at the NGA can emerge.