Needs for an Intelligent Distributed Spacecraft Infrastructure

Future Earth science observing systems will involve multiple space assets and capable models to advance understanding and enable prediction of Earth system variables. There are several types of distributed spacecraft architecture that contribute to an overall integrated sensor web future vision. Many technologies needed to enable the Vision essentially mimic commercial developments in the electronics, network and communications industry, suggesting that low-cost, capable microspacecraft will be realized in the near future. Spacecraft autonomy and capable sensor suites represent the most significant investments that will decrease cost and improve the capability and reliability of sensor webs. I. THE FUTURE OF EARTH OBSERVATIONS Predicting far into the future is always uncertain since unknown technological innovations sharply limit our ability to linearly extrapolate from the present. However, current trends suggest that space observations of the earth will become increasingly important to a range of users, both scientific and commercial. As an example, over 90% of the information used in current NOAA weather forecasting systems comes from space observations. The increasing capability and cost-effectiveness of sensors and sensor systems suggests that we will become more reliant on space-based observations in the future. In addition, the goal of understanding interactions of Earth system components on a global scale requires consistent, reliable global data that are most easily collected from space. There are many emerging measurement needs and capabilities that will drive space mission archtectures of the future towards multiple observing platforms and networks of sensors. These emerging areas include high spatial and temporal resolution: land imaging, surface hydrology and precipitation, ocean salinity, vegetation recovery, atmospheric chemistry, surface deformation, and radiative flux. While tremendous progress has been made in the past decades in understanding trends in isolated Earth system variables (such as atmospheric temperature), there are many dynamic processes and linkages that require dense observations both spatially and temporally to resolve. Modeling capabilities extend the “observability” of complex phenomena by making efficient use of existing data to validate physicsbased models. However, the next wave of advances in understanding of individual Earth system components and their interactions requires understanding and modeling of complex, non-linear systems. T h s next step in understanding the Earth system and enabling reliable predictive capabilities will require abundant observations to initialize and validate models of complex behavior. Whde it is difficult to predict how much data are needed, an educated guess is that increasing spatial, temporal and spectral resolution will be needed to improve predictability. On the other hand, for some systems, once we develop an understanding of the phenomena, either through model validation or data mining, and the data density requirements could decrease. An example of this scenario is current weather prediction capabilities, where models ingest sparse data but make decent predictions over large regions by advecting information from better observed regions. Also for weather prediction, it is well-understood that accurate wind profiles, especially in certain areas on the planet, will significantly improve prehctability. For many other disciplines, the complex physical models of system behavior do not exist, because the non-linear interactions of the various scales of the system are not well understood. For these systems, dense data are needed to build and validate models that may lead to predictive capabilities. Distributed spacecraft observing systems offer an attractive architecture for achieving high spatial and temporal resolution, but that architecture is far from one-size-fits-all. The observing system and its flexibility must be optimized around the science requirements. Various phenomena which may be observed from space have different temporal and spatial scale requirements. For example, severe storms evolve quickly, and observations every 15 minutes or less may be required. On the other hand, ice sheets evolve slowly and thus may be observed less continuously. Similar arguments can be made for spatial requirements. Thus it is important to consider various vantage points ice sheet observations could be made from low Earth orbit (LEO) while severe storm observations should probably be made from geostationary orbit or highly elliptical orbits to meet the revisit requirement. Surface deformation occurs on a variety of temporal scales ranging from long-term volcanic inflation to seasonal hydrologic variations to earthquakes. The spatial and temporal coverage provided by dense LEO constellations, sparser mid Earth orbit (MEO) and elliptical constellations, geosynchron-ous (GEO) constellations or geostationary orbits must be tradedoff with the cost of the observing system and operating and data processing costs. Advances in autonomous operations, computing, communications, and lightweight miniaturized spacecraft and instruments will all contribute to malung distributed spacecraft observing systems a cost-effective solution for the future. 11. EVOLUTION OF MULTI-PLATFORM OBSERVATIONS A . Virtual Observing Systems Currently meteorological observations from a variety of sources are brought together as part of the data assimilation process. The platforms that are making the observations are not coordinated nor commanded to alter their observing strategy. The data is simply gathered from wherever it is supplied. Nonetheless, when considered as a whole, the current meteorological observing system constitutes a virtual single observing system with multiple components. It is anticipated that virtual observing systems will be numerous in the future, as science questions are addressed with diverse observations from heterogeneous, uncoordinated satellites, combined with ground, balloon, and buoy networks, as well as Uninhabited Autonomous Vehicles and sonde data. Data assimilation models will be used to merge various data