Transcript WP3 contributions to WP1/online content

WP3 contributions to
WP1/online content
Los Baños, 16 May 2006
Sven O Kullander
Deliverables 1-4
• Mo. 13 Online: standardized electronic maps with
predicted distribution = AQUAMAPS done
• Mo. 19 Before-after maps with predicted
distribution for at least 10 species
• Mo. 26 Maps with predicted seasonal distribution
• Mo. 31 Online: Dynamic maps of predicted
distribution based on physical models
Map components
1.
2.
3.
4.
5.
6.
7.
Species occurrence data (FishBase, GBIF, other sources)
Environmental layers/datasets
Algorithm
Base maps
Projection model (=scientific hypothesis)
Software running 3 on 1 and 2, according to 5
Visualisation tool (e.g., web interface) displaying results
of 6 on 4
Biodiversity Informatics Basics
1.
2.
3.
4.
5.
Primary diversity data = Observation data on specimens
including scientific name (of taxon), place and time of
occurrence (spatiotemporal parameters of scalable
objects)
Any additional information about the taxon, the time or
the place
A biological hypothesis linking 1 and 2
An analytical tool testing 3 using at least 1 and 2
Or go straight from 1 to pattern analysis
Biodiversity data LCD
1.
2.
3.
4.
5.
Darwin Core concepts (xml schema for biodiversity
information exchange, subset of Dublin Core)
Developed for, but not restricted to DiGIR (Distributed
Generic Information Retrieval)
Mandatory fields providing unique identifier for record
and scientific name
Descriptive fields for spatial and temporal allocation and
level of confidence in scientific name
Used by GBIF, OBIS (w/ extensions), and others,
contained in ABCD and covered in TAPIR
Biodiversity data components
1.
2.
What – scientific name of a species
Where – Latitude/longitude
1.
3.
Latitude/longitude in decimal degrees (12.3456), using the
WGS84 datum (earth model)
When – Date
1.
Day of month, Month of year, Year
Environmental data components
1.
2.
What – a measurement (w/ associated metadata)
Where – Latitude/longitude
1.
3.
Point, grid, or polygon
When – Date
1.
Precise time or span of time
Science components
1.
2.
How?
Why?
Species occurrence is point,
environment is area
1.
2.
3.
4.
Relation of occurrence records is to an environmental
polygon. Overcome by fitting occurrence points to
polygon extent (e.g., c-squares), i.e., modelling species
distribution
Resolution is limited by resolution of environmental data
set (e.g., 0.5 degree)
Does not matter much because species distribution (in
contrast to species occurrence) is also a polygon, and
environmental data based on point records
However, ad hoc species and environment polygons
cannot be directly related in a database
Maps and dynamic maps
1.
2.
c-squares AquaMaps produces semi-dynamic maps.
Excellent for any kind of modelling, expanded
projection options underway. Works from FB-Kiel,
installed in FB-Stockholm. c-squares is the work of Tony
Rees, CSIRO, Australia. c-squares ”Server” is open for
map plotting from anywhere and anyone.
UMN Mapserver application mapcria is a map
application server that can use vector maps, serving db
data on a scalable map. Models can be output on the fly.
NRM has UMN mapserver running. Installation is not
straightforward. Excellent for dynamic mapping.
Maps and dynamic maps
Models and Modelling
Guisan & Thuiller, 2005
Models and Modelling
1.
2.
3.
4.
5.
6.
7.
8.
9.
INCOFISH so far based on environmental envelopes. Requires calculating species
specific preference table, and probability table for all squares of occurrence.
Other environmental/climatic envelope models: BIOCLIM, DOMAIN
Multivariate models: ENFA
Artificial neural network: ANN
General Linear Modelling: GLM
Maximum Entropy
Except for species distribution data, modelling uses the same environmental data
sets, and generally the level of detail of analysis is decided by the set with lowest
resolution.
Applied models are normally predictive of present occurrence (spatial prediction or
ecological niche prediction). Result can be tested empirically.
Predicting past or future distribution requires data sets modified according to some
theory/model about future climate and land extension (future), or historical
information (past distribution). Testing model result of spatiotemporal prediction is
probably possible, assessing the reliability of the combined occurrence data used +
model for climate change over time + environmental data set used + prediction from
specific algorithm, is an issue.
Models and Modelling (2)
1.
Modelling software mediates occurrence data, environmental data
sets, algorithm, and usually map output
1.
2.
Popular now: GARP, DIVA-GIS, Maxent, WhyWhere
OpenModeller is a software that can handle several different
algorithms within the same application (Input occurrence data,
select environmental data sets, select and run algorithm, obtain map
and statistics). Currently uses QuantumGIS for map output.
Models, modelling and the client
1.
2.
3.
4.
To provide consistent information to a general public, decision
makers, and non-modellers, conforming to trust, communication,
and relevance, AquaMaps (controlled occurrence data + c-squares +
environmental envelope) probably serves the purpose best. The
Aquamap model is being tested/validated by WP3
A web service implementation of OpenModeller (with AquaMaps
model or other) may be most efficient to serve specific predictions
with user-selected environmental parameters and taxa.
Or go for a combination?
A decision has to be made regarding use of UMN Mapserver or csquares mapper or balanced use similar to 1,2 above
Data interoperability
1.
2.
Darwin Core is an available, growing standard for Internet based
exchange of biological data using xml/http. Minimum requirements
and preferably also additional concepts in Darwin Core should be
implemented in ALL biological databases
Relations between Darwin Core and environmental/ecological data
sets can be standardised. At present they need to be managed
programmatically. WP1,3,4,5 should look into this. c-squares is a
scalable option implemented in AquaMaps
Deliverables 1-4
• Mo. 13 Online: standardized electronic maps with
predicted distribution = AQUAMAPS (done)
• Mo. 19 Before-after maps with predicted
distribution for at least 10 species. Underway with
Aquamaps
• Mo. 26 Maps with predicted seasonal distribution
Underway with AquaMaps
• Mo. 31 Online: Dynamic maps of predicted
distribution based on physical models. Needs a
decision about mapping tool.