Question Four

From Earth Science Information Partners (ESIP)

Enter your discussion report-out below:

4. What is the current state of progress in “data fusion” (ALSO innovative technology) methods? What are the next steps?

Agencies need to look at both evolutionary and disruptive technology (Innovator's dilemma). Technology should be driven by user needs. Research and operations have time constraints, so agencies should evaluate which innovations best serves these. There is also the problem of finding new technologies. The TIWG used the tagging system to "crowdsource" its technology scouting activity. All participants were tasked to tag new technologies they found with a "TIWG" tag. product evaluations are a great model for getting user feedback and engaging users as partners.

New technology presents new security risks (#13). So these must be carefully vetted on parallel systems.

When doing public/private software development, the private developers need to be informed about federal IT regulations. What is the best way to do this?

Forcing users to register to use data runs counter to the spread of data use and data use technologies. For some data, the use of IDs for access may be necessary.

Employeees are key sources for new technologies. Engaging employees to discover, evaluate, and report new technologies may require some reward system. For example, prestige systems can be set up (incentives and reputation). Other awards, publications, and increased participation in Federation-type meetings can be rewards.

Cultivate disruption. Some disruptive technologies can greatly improve efficiency. This is one way to improve services under a decreasing budget (#1). There needs to be money in the budget for innovation, some kind of reserve wedge that can be pointed in this direction. This can leverage on other budgeted work.

Data Fusion created value-added and new products customized per user need. Examples include merving/fusing disparate data to increase spatial/temporal coverage, to add geospatial/socionomic info to remotely sensed data sets, and multiple sensor data fusion.

Data fusion products can be highly customized (case-by-case). Many instances of these are now available (e.g., Giovanni). Because of the custom needs, these are very labor-intensive to develop, and so are expensive.

If not done right, fusion can lead to science "junk". Some way to maintain provenance might be of real value.

Data fusion is not very well advanced.

This is a multifaceted issue: there are scientific, technological, and semantic (also user-side) issues to be resolved.