Global Change Information System (GCIS)

Global Change Information System (GCIS)

Introduction to GCIS - Robert Wolfe – USGCRP GCIS Technical Lead

·         US Global change research program

o   Coordinates federal research, prioritizes and support cutting edge research, accesses the state of scientific known, communities research finding

·         Played a major role in the national climate assessment – only been 3 since 1990, latest one just release – includes the uncertainties with the findings, analyzes the trends and projections into the future

o   3 reports 2000, 2009, 2014

·         GCIS – right now

o   Have an initial prototype – supports the distribution, presentation and documentation of the NCA integrates into USGCRP website

·         Information quality act – reproducibility & transparency – very important to show how you got your results

·         Complete traceability for NCA content – have traceable sources for references

o   Spectrum from transparency to reproducibility (if they have the right tools/computers)

·         Define categories of information within the report

o   Figures, images, data source

o   Build a process for collecting source information – named sources, web-based survey providing ISO 19115, IT infrastructure

·         The website for the NCA has been praised as a well done website by Whitehouse

o   Have a structured data server webpage

o   Can come in with an api to access the information

o   Working on documenting the data sources – about 20+ done

·         Dataset metadata from a figure – can get all the background/metadata about the figure – if click on figure see time range, who created, what data sets, the activity to derive the information

·         GCIS structured data server – capture, identify, organize, present, maintain

o   Identify is a key point – they must be persistent and resolvable

·         GCIS database/api –restful api, rep in JSON and in Turtle

·         Have a very complete ontology that they are working with

o   Have classes and properties for report and each chapter of the report

o   Started with existing ontologies

o   Used PROV_O classes and various other ontologies

·         SPARQL example – http://data.globalchange.gov/examples

·         two parallel paths – NCA3 release (1) and (2) populate GCIS

o   Q – working on this for future documents – not planning on going back in time

o   To get this information, have to do it well you are generating the report

o   Before GCIS it was very difficult to capture it

·         The future

o   In 2013 – started to support the NCA report

o   Just started to put indicators in for the initiative

§  Demo is the 1st 24 and then the pilot is all the rest

o   Use health assessment to make this easier

o   More ontology development/improvements – some aspects that we can do better – learn from the pilot

o   Proposal on the table is for sustained assessment – don’t break the team in a few years – keep building on what you have now

o   Possibly supporting the EOA (earth observation assessment)

·         Q what was the responses of positive vs. negative responses to the NCA report

o   News – publicity – over 2000 articles

o   Pointed to specific locations on the website – one element was being able to select specific sectors or specific region – easy to see the impacts on what they want to know

o   General feeling – more positive responses than in 2009

o   There is an NCA team that is looking into this specifically

o   Criticism came from expected people

o   There is a system for review and comment on all files that are put out

·         Q – curation and maintenance of GCIS

o   One of the questions for the next report – should it be web first – usually pdf and then web – in 3 years – do a web document and then producing a document to represent it

o   Is it semantically enabled from the beginning

o   Re-evaluated how things were done and looking to improve the process

·         Q – before NCA3 there was author instructions and templates – expect a different approach to instructors/templates

o   There is a lot of evaluating the process of generate the reports

o   Main take home message – start this process early

·         Q what sort of feedback have you gotten from the authors – haven’t gotten that yet

The global change information system: nuts and bolts – Brian Duggan, Steven Aulenbach, Robert Wolfe, Justin Goldstein

·         NCA3 is about creating a PDF and a website

o   800+ page pdf with content that was traceable

o   Website – another way of looking at the pdf

·         How – lots of different tools – ex. Googgle docs, scientific software

·         Resources – things that needed to fit into the report

·         Role of GCIS

o   Common points of reference – ex. Bibliography has to refer to the same article the same way – must be a structured way

o   Vocabulary had to be consistent – language, terminology, vocabulary, ontology (Ex. Publication, bib entry, and a citation)

o   Needed uniform way of identifying resource – use uris (urls)

o   Needed fine grain tracking of all the changes made – not just version control of a document – looking at database and incorporating who was asserting what was true about a relationship

·         GCIS is backing the website for the NCA3 report… so a call on the website calls GCIS – back end is using PostgreSQL

·         GCIDs are gcis identifiers – website/domain name/orcid or uuid or identifier

·         Functionality – support report & website, minimal landing page for resources, JSON API, semantic information, interoperability (using existing identifiers), public SPARQL endpoint

o   Interoperability by reusing identifiers

o   If identifiers are being used outside of the GCIS are queried to relate the end points

·         Q – previously – what was the back end – PostgreSQL – but now sounds like Virtuoso

o   There were strength and weaknesses of both of these

§  Need referential integrated

§  Fine grain auditing

§  Cascading deletes

§  Needed performance aspect & tools for providing web services

o   Also had a very short time to put together a product that wouldn’t fall over – will do more as time progresses

·         Q can you provide an example of relationship that PostgreSQL couldn’t do

o   How is a figure in a report different or the same as one provided – not just a mapping of one figure to another but include meaning – have a generic entry for a publication, correspond to other relationship

o   Tables in DB just to generate Turtle

o   This is an information model that branch into the schema

o   SPARQL is good for getting data out – but putting in, managing, deleting… one GCIS resource – a single insert can create a whole thing… it get tedious in the long term (triples)

·         Testing – enhancing semantic model – runs a test for each github and SPARQL  - also not just testing code but also the content (not just internal consistency, but continuous content validation

·         Clients – python, perl, javascript, php

·         What is the difference between narrative vs. structure content – ex background on an image

·         Semantic vs. relational – is his critical for the front or is it going to be ingested

·         Concept of resources – with a uniform identifier

·         Identifiers – there is no good identifier for organization – dataset = were local to NCA3 process & changing identifier – if title change than identifier changed

·         Have entities, agents, activities – have publications (entities), contributors (agents), and then activities – modified prov model

·         http://data.globalchange.gov/resources

Discussion – Steve Aulenbach

·         Q – Statistic analysis

o   Have the ability to see who has viewed it – any time it is linked then it is recorded

o   Have google analytics

·         Q background database – is it distributed and what is capacity

o   No – it is a single instance of PostgreSQL

·         Q – backup/preservation – where is it backuped offsite

o   A couple of places – UCAR – top enterprise level backup

o   During development (Before web engineering group) – work with local server that had a great set-up

·         Q – how much time did you spend extending the schemas in PostgreSQL

o   Still are

o   Had schema changes incorporated into the release process – includes a patch – good patch management – this allowed for extension of the relational model – will have to keep doing

o   Automated – so it isn’t a barrier

o   All software is on Github – everything is open source activity

o   Had a deadline – but the people that provided can move as they wish (the Whitehouse) there is some interesting naming conventions, not always consisting, it is a real live prototype – made release date and now moving into it and iterate as they clean up

§  Testing guarantee that it will be easy

·         Q – mentioned testing – tested data/metadata against podac - how

o   Not yet

o   Would like to work to set that up – bring it into the open source process

o   Doi are important for what a data center can do

·         Q – challenges that you faced to get the figures/maps

o   People work in spreed sheets – this was really difficult for “us” – report was written in word.  People are using endnote for bib – but ever author submitted word documents

o   Hope to do things differently in the future

o   What should be put into GCIS

o   Kurt – references – who actually looks at each reference – intent is for a person… here need it to be computer readable – for this purpose had to have high quality references

o   It is open source – thus can be stood up for others to use – but can’t add to the GCIS database of references because it is still for a specific

o   References are all pulled directly from the publisher – trying to automate this

o   Need to comply with the information quality act – required full tracably

·