Hi folks –
We’re looking to submit a CDI proposal that would provide proof-of-concept for a USGS integrated predictive modeling infrastructure on the semantic web. This is a big lift, but we’ve already built much of the infrastructure (see references below). We believe it’s possible to build a relatively simple proof of concept that could later be moved to production scale within a larger effort. The approach has several key advantages – the semantic underpinnings to connect diverse scientific disciplines, ability to connect models of different paradigms (including data-driven approaches), and to better navigate scale differences.
We’re looking for partners interested in building that proof of concept. Key requirements are 1) that we address a multidisciplinary challenge (which will immediately challenge us to build ontologies for semantic annotation of data and models that cross disciplines – a challenge that existing domain ontologies struggle with, but that we’re equipped for), and 2) that the initial models be relatively simple (to be tractable within the limited scope of a CDI proposal).
If you’re interested in partnering on or have good ideas on an integrative science case study, or have any questions about our proposed approach, please let me know!
Proposed steps in CDI proposal:
- Generate common semantics for a small, multidisciplinary integrative modeling problem, via a 1-week semantics workshop.
- Serve & semantically annotate key data resources on the web (i.e., Semantic USGS Science Data Catalog, e.g., data already hosted on sciencebase.gov/geoserver). This sets the stage for wider use of semantic interoperability in ScienceBase, and a framework for other data repositories to follow to allow their products to work together.
- Code/wrap key models and contextualization information in the k.LAB semantic modeling environment, enabling their use within a linked data & modeling environment.
- Provide a few model runs that illustrate the results.
Later steps toward production-level (not part of this CDI proposal):
- Model calibration/testing & validation of predictive power
- Production version of above to address additional integrative science challenges (building new cross-disciplinary ontologies and models)
- Setup semantic modeling node to enable full control of confidential/not-yet-released data
- Web explorer for key products – Decision Support System-style "front-end" for nontechnical users
- Semantic annotation navigator: Design a metadata editor plugin that enables semantic annotation of new products in ScienceBase by technical support staff (i.e., GS-9 level).
References
http://www.integratedmodelling.org
5 Comments
Lightsom, Frances L.
One thing I have wanted to do is connect coastal hazard predictions with national map data about transportation and the locations of hospitals, schools, etc.
I wonder if that might be a good size model for this first project. Tsunami inundation of critical infrastructure, and related evacuation and shelter instructions that coastal communities would choose.
Arsenault, Matthew
I would recommend we consider (especially in light of the migration towards a coastal hazards program) is closer participation in the HiFLD Community (https://gii.dhs.gov/hifld/)
Here is the url to their response to Hurricane Harvey:
https://respond-harvey-geoplatform.opendata.arcgis.com/
or Irma:
https://respond-irma-geoplatform.opendata.arcgis.com/
While much of this data is focused on the hurricane response effort, getting our inundation models and coastal vulnerability models integrated with these other data sets could be helpful
Bagstad, Kenneth J.
Thanks Matthew - sounds like a very good direction to consider taking. Do you have links to preferred inundation/coastal vulnerability models?
Arsenault, Matthew
Hi Kenneth - I do not have a favorite inundation model, or coastal vulnerability model (unless you count anything developed by the folks here at USGS!). But we have people here in Woods Hole who would be good individuals to 'ping' for information. Plus we have a new portal in the works that could facilitate this.
I put in a request to join the HiFLD community (I've been getting their emails for many years from the old multi-purpose marine cadastre meetings) and have just filled out my application, waiting for DHS to clear it. I'll keep the group posted, so we can see how it may fit into Fran's original comment.
Allwardt, Alan
Also consider Nate Wood: https://geography.wr.usgs.gov/staff.php?names=Nathan+Wood&submitName=Get+Staff+by+Name