GIS i d S tiGIScience and Spatio-TlTemporal … i d S tiGIScience and Spatio-TlTemporal Analysis: An Overview of Recent Advances Michael F. Goodchild University of California Santa

GIS i d S ti T lGIScience and Spatio-Temporal Analysis: An Overview of Recent

Advances

Michael F. GoodchildUniversity of California

Santa Barbara

What do we know?• …that we didn’t a few years ago• What do we need to know?

– and what are the chances of finding out?• Space

i– static• because of the economies of scale of map production

• Dynamics• Dynamics– remote sensing’s time slices– a flood of new data typesa flood of new data types

Rondonia Brazil 1975 1986 1992Rondonia, Brazil, 1975, 1986, 1992

CASA UCL, http://www.casa.ucl.ac.uk/capableproject/maps/home.asp

Light-level geolocation (Stutchbury et al., Science 2/13/09)

Purple Martin

Wood ThrushWood Thrush

Time-critical data• Near real-time• Sensors embedded in the environment and

lon people– tracking movement

H ti i t lli t• Humans acting as intelligent sensors– densely distributed on the ground

observing uploading– observing, uploading• crowdsourcing or volunteered geographic information

htt // i t t bl /id/http://www.instructables.com/id/Mapping-Microbes/#step1

Interoperability of geolocation34 d 24 i 42 7 N th 119 d 52 i 14 4• 34 deg 24 min 42.7 sec North, 119 deg 52 min 14.4 sec West (3m)

• 909 West Campus Lane Goleta CA 93117 USA909 West Campus Lane, Goleta, CA 93117, USA (20m)

• 3811560N, 236150E, Zone 11, Northern Hemisphere (10m)

• NE 1/4, Section 12, Township 23 Range 5 of the Second Principal Meridian (300m)Second Principal Meridian (300m)

• National Grid reference 11SKU36151156 (10m)• Visual identification on a georegistered base map g g p

(20m)• “Mike Goodchild's house” (20m)

What does time add?• Insights that are not available from space

aloneTi i d d i t• Time is ordered, space is not– potential for getting closer to causality

resolving among competing hypotheses– resolving among competing hypotheses• The temporal extent of new data sources are

necessarily shortnecessarily short– compared to the time-scale of health impacts– can space be substituted for time?p

Advances in analysis• Mining time as well as space

– what happens at x at time t is a function of what happens nearby at time t what happened at x athappens nearby at time t, what happened at x at time t-1, etc.

– finding anomalies in space-time• Analyzing and interpreting tracks

– tracking exposure– inferring interactions

• spotting swarms

inferring acti ities– inferring activities• Real-time analysis

Comparative spatial analysis of the development of the Chinese and US economies through time, 1978-1998

Xinyue Ye, Bowling Green State University

The issues• Big Data

– variable qualityt ll d li– no controlled sampling

– synthesis before analysis– tracking uncertainty at the datum level– tracking uncertainty at the datum level– semantics

• variable meaning

• CyberGIS– computing in the cloud– unlimited computing resources– beyond the old shortcuts

More issues• Space or place?

– linked by location and timeli k d b l d ti ?– or linked by placename and time?

– a platial (or platiotemporal) analysis?• Analogs• Analogs

– what places and times are like this one?– learning from similar situationslearning from similar situations

A very different world• Spatiotemporal dependence

– observations are always autocorrelatedS ti t l h t it• Spatiotemporal heterogeneity– all places, times are different

little prospect of generalization– little prospect of generalization• no random samples

• Requires us to think differentlyRequires us to think differently– based in spatial (and temporal) concepts– orthogonal to traditional training in quantitative g g q

analysis