Summary

The United States faces numerous, varied, and evolving threats to national security, including terrorism, scarcity and disruption of food and water supplies, extreme weather events, and regional conflicts around the world. Effectively managing these threats requires intelligence that not only assesses what is happening now, but that also anticipates potential future threats. The National Geospatial-Intelligence Agency (NGA) is responsible for providing geospatial intelligence on other countries—assessing where exactly something is, what it is, and why it is important—in support of national security, disaster response, and humanitarian assistance. NGA’s approach today relies heavily on imagery analysis and mapping, which provide an assessment of current and past conditions. However, augmenting that approach with a strong modeling capability would enable NGA to also anticipate and explore future outcomes.

A model is a simplified representation of a real-world system that is used to extract explainable insights about the system, predict future outcomes, or explore what might happen under plausible what-if scenarios. In this report, a model means a mathematical or numerical model that can be run on a computer. Such models use data and/or theory to specify inputs (e.g., initial conditions, boundary conditions, and model parameters) to produce an output.

At the request of NGA, the National Academies of Sciences, Engineering, and Medicine established a committee to describe types of models and analytical methods used to understand real-world systems; to determine what would be required to make these models and methods useful for geospatial intelligence; and to identify supporting research and development for NGA (see Box S.1). The report provides examples of models that have been used to help answer the sorts of questions NGA might ask; describes how to go about a model-based investigation, using example questions to illustrate how NGA might think about choices and tradeoffs; and discusses models and methods that are relevant to NGA’s mission.

MODEL-BASED INVESTIGATIONS

Models do not stand alone, but rather exist in an environment that includes the available data, methods for analysis and model assessment, computational and data infrastructure, and people skilled in developing, tailoring, and running models and interpreting their output. A model-based investigation begins by formulating the key questions to be answered. The questions drive the choice of models, analytical methods, data, and computational resources as well as how these pieces will be combined to generate results with the necessary speed and accuracy. Using existing models or model output speeds the investigation and reduces its cost. If appropriate models or model

output do not exist, the investigators have two choices: develop a new model from scratch or combine existing subsystem models into a new model. A combined model has the potential to capture the behavior of the larger system, as long as both the processes in the subsystem models and their connections are appropriately represented.

Once chosen, the models or model products are incorporated into an analysis, typically involving data and computation, to connect the model to the real-world system. A variety of analytical methods may be used in the investigation, such as methods to preprocess data before ingesting them into the model, to update the model state as new observations accrue, to calibrate model parameters, to combine subsystem models, to determine what new data to collect, or, crucially, to assess the credibility and uncertainty of the results. A model differs from the real-world system it seeks to represent for several reasons, including omitted or inadequate representation of system processes, errors and uncertainties in the data used as input to the models, and coding errors in the models. Any model-based investigation must assess the impact of these uncertainties on our inferences about the real-world system, and communicate this uncertainty to decision makers.

The demands of the analysis dictate the computer infrastructure needed. Some models and methods can be run on a laptop. High-performance computing is generally needed for computationally intensive models that require large numbers of processors (e.g., thousands), dedicated communication between processors, and large volumes of data (e.g., gigabytes to terabytes) in memory and storage (e.g., climate models). Data-intensive computing is used for the analysis of massive amounts of data (e.g., terabytes to petabytes), which is dominated by data processing tasks. In these cases, computation and data manipulation must be divided into parallel tasks that can operate on separate pieces of the data, with minimal communication between these separate tasks (e.g., text mining).

MODELS AND METHODS FOR NGA

Given the breadth of national security and humanitarian challenges under NGA’s purview, it could be argued that dozens of models and analysis methods, each with important variants, are potentially relevant to NGA. The study was unclassified in its entirety, and so the committee used NGA’s mission, the special characteristics of geospatial data, and two example intelligence scenarios provided by NGA to guide its selection of relevant models

and methods. Below is a discussion of models and methods that seem particularly relevant to NGA, illustrated by the following NGA intelligence scenario:

China needs to find more water to meet agriculture and energy demands, but major dam projects (e.g., Three Gorges Dam) have displaced large populations against their will. Broad intelligence questions include How do agriculture and energy production and consumption change over time? How and where will populations, including rural communities, shift?

Types of Models and Methods Relevant to NGA

The first two tasks of the committee were to identify types of models and methods used to understand complex systems and describe their relevance to NGA (see Box S.1). NGA’s mission is to produce geospatial intelligence, which is defined as the exploitation and analysis of imagery and geospatial information to describe, assess, and visually depict physical features and geographically referenced activities on Earth. To extend this mission to the modeling realm, NGA will need models of human behavior (activities), set within an environmental context (physical features), and techniques for integrating and analyzing geospatial (geographically referenced) data. This analysis will be influenced by the spatial, space-time, or network structure that is present in geospatial data. Key characteristics of such data include autocorrelation (e.g., the properties of nearby locations tend to be similar) and spatial heterogeneity (i.e., the phenomenon being modeled varies with location).

These needs place a premium on the following types of models and methods:

• Models of physical processes that affect human activities. Such models use theory-driven equations to describe environmental (atmospheric, oceanic, hydrologic, and geologic) systems and to predict their future behavior. For the NGA intelligence scenario, for example, a large-scale physical process model of the hydrologic system in China could be used to predict surface flow, subsurface flow, and abundance of water under different water diversion scenarios.
• Social system models of human behavior in a geospatial context. Particularly relevant models include system-level models of influences and flows between stocks, and agent-based models, which use defined rules of behavior among individuals to study emergent population behavior. These models support what-if reasoning, such as how different scenarios are likely to unfold. For example, social system model scenarios could be constructed to assess of how affected populations are likely to respond to dam construction and involuntary migration in China.
• Models of combined physical and social systems. Physical and social subsystem processes that depend on one another may be coupled to understand their interactions at different locations. For example, weather, wave, and pirate behavior models have been combined with shipping patterns to help predict where and when pirates might attack. More complicated economic, resource, and energy interactions and feedbacks may be captured in integrated assessment models. For example, integrated assessment models could be used to examine the complex interactions among water, agriculture, and energy production and consumption in China.
• Inverse methods to infer uncertain model parameters from measurements of the real-world system. These methods combine computational models with real-world observations to constrain model parameters (e.g., physical constants, initial conditions, boundary conditions, and system states) so the model better represents the real-world system and therefore produces more reliable results. For example, inverse methods could be used to estimate or constrain key model parameters of a large-scale hydrologic model (e.g., spatially varying permeability, flow rates, and evaporation) to produce plausible predictions of water availability as a function of location throughout China. Inverse methods must be supplied with problem-specific information and tailored to the model being used.

• Spatial statistics, data mining, and machine learning to discover trends, patterns, and associations. These methods use data-driven empirical models and methods to understand what diverse observations reveal about the system. For example, empirical models could be used to detect changes in water availability arising from policy decisions on dam building, agriculture, and coal production in China. Empirical approaches are often customized to the type of data being analyzed (e.g., spatiotemporal data, images, and text) and the available computational environment.
• Spatial network analysis to examine how patterns of relations affect behavior at the individual to state level. Network analysis models are graphical or statistical models that explain behavior among people related by friendship, financial, cultural, or other ties. Spatial network models, which combine spatial and network reasoning, explain how patterns of relations affect behavior. Network models have been used to characterize the actors in covert groups and their relationships, to assess the potential for various kinds of attacks, and to evaluate changes and trends in dynamic geopolitical environments. They are often used in scenario descriptions and what-if analyses, and temporal trends in network data are also used for prediction. For example, social media analytics could be used to determine which neighborhoods and which groups in China are likely to strongly resist migration required by dam building.

State of the Art

The third task of the committee was to describe the current state of the art in relevant models and methods, including features and scales, accuracy, reliability, predictability, uncertainty characterization, and computational requirements (see Box S.1). These factors are applicable to a particular model or method in a particular setting, but not to the broad classes described above. Consequently, the report provides ranges or examples for the various factors called out in Task 3.

The state of the art influences which models and methods NGA can adapt in the near term and which require additional research and development. In general, physical process models and data-driven methods (inverse, empirical, or network analysis) are relatively mature, but will have to be adapted for geospatial intelligence purposes. Examples of useful developments include downscaling techniques, which would facilitate the application of NGA’s remotely sensed data in problems on smaller, more policy-relevant scales, and empirical methods that can handle specialized model structure (e.g., space-time dependence) and features of geospatial data. Social system and coupled physical–social system models can support what-if reasoning and scenario development, which are useful for many NGA applications. However, fundamental research is needed to improve understanding of human behavior and to make the models easier to develop.

Increasing the Usefulness of Models and Methods to NGA

The fourth task of the committee was to determine what would be required to make relevant models and methods useful for geospatial intelligence (see Box S.1). What particular models and methods to develop or adapt for geospatial intelligence purposes depends primarily on their utility for the investigations at hand, but also on the difficulty of developing or adapting the models and methods, the availability of software and code, the level of training support, and the knowledge and experience of NGA analysts. For example, NGA must often respond quickly to an emerging national security threat or a natural disaster, and so a simple model that is relatively quick to develop, adapt, or run may be more useful than a more comprehensive model that takes months or longer to develop. When a more sophisticated model is needed, NGA could leverage the considerable expertise of modeling groups at universities, national laboratories, federal agencies, and private companies. Actions NGA can take to develop or adapt the models and analysis methods described above are summarized below.

In-house development or adaptation. Data-driven models and analysis methods are amenable to near-term development, because NGA analysts already have some relevant knowledge and experience, the methodology is established, and software, tools, and training support are available. In particular, NGA’s experience with spatial and temporal analysis provides a foundation for developing or adapting spatial statistics, data mining, and machine learning methods. Methods that are especially promising for NGA include (a) Bayesian hierarchical models that link spatially connected data, data at different levels of resolution and aggregation, or disparate data sources; (b) clustering and other unsupervised and deep learning methods for finding structure in large volumes of data that are too much to analyze with a human in the loop; and (c) methods for detecting change footprints and spatial hotspots and anomalies. In addition, NGA’s growing emphasis in human geography provides a foundation for developing network analysis models to examine how patterns of relations affect behavior.

For both types of analyses, the basic methods are well established, and software and user support (e.g., textbooks, conferences, and special short courses) are readily available. However, some additional development and training are required to adapt these methods for geospatial data and NGA use cases. In addition, software and algorithms for data-intensive computing will likely have to be developed for spatial statistics and spatial data mining methods. Training in network or spatial network analysis could be offered at the NGA College or obtained from university-based programs.

Collaborations. NGA will need partners to help develop, adapt, and use more sophisticated models and methods (e.g., process models, coupled models, agent-based models, inverse methods, and spatial network models) as well as geospatial models that are not well supported by cutting-edge computational infrastructure. A substantial part of any group’s capability in sophisticated modeling is learned through partnerships, apprenticeships, and collaborations. Such collaboration could take many forms, including being a partner in the team developing a model or extending its use to other applications, a user of a team’s model or method, or a user of the resulting data products. Regardless, NGA will need to identify domain experts who can design models or scenarios relevant to NGA, run the model, interpret the results, or help NGA find useful existing model output. To use these models or model results effectively, NGA will need to understand their strengths and limitations for the geospatial intelligence task at hand.

Finding partners for NGA modeling efforts will not be trivial because of the classified nature of the work, the wide and changing variety of experts needed, and the need to nurture long-term relationships. Models of complex systems are typically developed by multidisciplinary teams with in-depth knowledge and experience in the scientific disciplines and computational capabilities relevant for the task at hand. However, bringing together diverse experts, who would learn from each other in the context of NGA’s priorities, could contribute to major breakthroughs in NGA-relevant problems. Major research universities, as well as organizations for which NGA has established relationships (e.g., defense and intelligence agencies, national laboratories, private-sector contractors, and NGA centers of academic excellence in geospatial science) may be a starting point for finding experts and modeling teams for NGA modeling efforts.

NGA-Funded Research and Development

The fifth task of the committee was to identify areas that could benefit from NGA-funded research and development (see Box S.1). A host of investments in research and development could strengthen NGA’s modeling capabilities in the years ahead. Where to focus these investments depends on what models and analysis methods are proving most useful for geospatial intelligence. Potential research areas concern extending the use of existing models to NGA-relevant situations, improving understanding of human behavior, reducing the time required to develop, test, and run models, and developing methodologies tailored to NGA-relevant models and data, as described below.

Extending the use of models to NGA-relevant situations. NGA investigations will likely make new demands on models, using them in settings for which they have not been originally designed, or at least not thoroughly tested. Examples include precise, near-real-time wind, wave, or weather predictions to support troop deployment, disaster relief, or dispersion and damage estimates from the release of hazardous materials in urban environments. In addition, such models may need to be combined with social system models to help decision makers prepare for social unrest, disruption, or migration. Substantial research is required to adapt physical process models, social system models, and combined physical–social system models to deal more reliably with these less common settings.

Improving understanding of human behavior. Social system models are only beginning to surpass expert judgment. Advancing their development, and the development of combined physical–social system models, requires fundamental research to improve understanding of human behavior. Promising areas of research include studies aimed at understanding how human behavior is constrained or enabled by the geography of the natural and built environment, including how geographic factors influence the development of social networks and communications among actors, and how cognitive biases influence the perception of space.

Speeding model development, testing, and run time. Intelligence questions are often time sensitive, and so research advances that speed up model development, testing, or run time could prove beneficial to NGA. Model development could be sped up through research aimed at facilitating the combination of existing subsystem models for NGA investigations. Model development and run time could be decreased through research and development of accurate reduced models that use coarser, simpler, or fewer representations of processes than computationally intensive models. Developing simulation testbeds could aid all of these efforts and also facilitate assessments of model accuracy and speed.

Methodological research and development tailored to NGA-relevant models. The models developed or adapted for NGA purposes will have to be accompanied by customized methods that combine these models with data. Methodology for inversion, exploration of plausible outcomes or scenarios, quantification of prediction uncertainties, and model assessment will be required to bring model-based results more in line with available measurements. Such methodology is particularly needed for social system and physical–social system models. Possible directions in this area include development of inverse methods for constraining the plausible states of social system models to be consistent with data, and development of methods for formal verification and validation of model results against NGA-relevant benchmarks and test cases. In addition, research on how to adapt existing inverse methods to integrate the diverse forms of data that NGA collects and uses (e.g., satellite, sensor, geospatial, and open source) would be beneficial for all types of models.

Methodological research and development tailored to NGA data sources and needs. Research could advance the development of empirical methodology by tailoring it to data and use cases found in NGA applications. Examples include developing methods to combine disparate data or results from different approaches and to more accurately represent their uncertainty in support of inference and decision making, and methods to cope with data that have spatial, temporal, and network structure (e.g., to assess the activities of a terrorist cell over time). A related need is for sentiment-mining techniques that characterize the sentiment in a document based on the geographic and network features of the community, such as location, local events, language, structure of local groups, and tendency to self-identify in networks. Research is also needed to develop algorithms for promising spatial and spatiotemporal methods to efficiently leverage the processing and data storage capabilities present in advanced data-intensive computational architectures.

OVERARCHING CONCLUSIONS

Overall, the committee concludes that model-based investigations would provide NGA with a powerful means to search for spatial and temporal patterns; make inferences from those patterns; predict future political, economic, and military threats to the United States; and evaluate options and consequences around the world. However, models are not the optimal tool for every geospatial intelligence problem, such as when time is too short to carry out a model-based investigation or when a lack of data or uncertainties about key processes make it difficult to obtain useful insights on the real-world system. Consequently, mapping and imagery analysis will continue to be important for geospatial intelligence.

NGA can begin to develop or adapt some data-driven models now, given analysts’ long experience with spatial and temporal analysis and the availability of supporting software and tools. Developing a more sophisticated modeling and analysis capability will likely take many years, because NGA will need new knowledge, skills, techniques, and workflows; interactions with external modeling groups; additional sources of data; increased computational capabilities; and a change to a modeling mindset. NGA need not build this capability for cutting-edge or complex models developed, maintained, or supported by a large research community (e.g., climate models). Indeed, working with external modeling groups to use their models, model output, and analysis methods for geospatial intelligence purposes would stretch resources and help analysts gain knowledge and expertise in modeling and analysis methods. As their experience grows, NGA analysts will be able to take on progressively more complex model-based investigations in the classified world.

This page intentionally left blank.