Budget data
Cluster tools

(link to index page)

Results -- Americas Workshop

Americas results [Text excerpted from Section 3 (Results and Applications), LOICZ/UNEP Regional Synthesis Workshop Report]

 Workshop typology contributions.
At each of the three regional workshops, participants were provided with on-line computer access to the database and clustering tools, and encouraged to explore regions and topics of interest with the assistance of the resource people.  Working individually or in small groups, all of the participants completed some form of project-relevant typologic study and prepared results for collection and inclusion in the workshop reports.  The workshop processes and products are described more fully in Appendices I to III, and the project results are contained in the CD-ROM accompanying this report, and posted on the project web site.

These activities served the dual purpose of training the participants in the overall project approach and the use of the tools, and of providing user tests of the database contents and the web tools designs.

Participant Contributions, Americas Workshop, Ensenada

See Appendix II (view html) (pdf file) for a complete description and report of the workshop; links in the table below connect to electronic versions of the actual products on this CD-ROM.




Dr Laura David, Dr Victor Camacho, Dennis Swaney


Typology and biogeochemical budget assessment approaches: scaling; supervised clustering

Work on resolving linkages between and approaches for analyses of biogeochemical datasets and typology variables scales was addresses as a core LOICZ project activity.  A trial was made of combining ARCView and LOICZView methods, in which remote sensing methods were combined with existing biogeochemical data for “banding” of data using the clustering technique.  Principal component analysis of delta-DIP and delta-DIN was applied with continuous grouping analyses and inspection.  Derived delta-DIN data refined by delta-DIP assessment yielded about 40 clusters for the global coastal region.  These clusters accounted for 99.5% of the cells.  A supervised clustering approach is to be tried.

Drs Carlos Lechuga, Martin Merino, Francisco Contreras


Classification Variables for CNP inputs – Mexico

Initial case evaluation of Mexican coasts spatially relating land use and C, N, P inputs.  Seven clusters were developed that represented the Mexican coasts based on variables that acted as proxies for run-off, coastal geomorphology and C, N, P inputs.  Expert judgement suggested that the resultant clusters were generally sound although there was some variability.  Improved database of N and P variables (currently N is at first trial level in database) may improve the outcome.  Attempted global classification with addition of sea surface temperature variable yielded a preliminary global description with 12 clusters; further evaluation is required.

Paul Boudreau

Americas, global

Global Classification for biogeochemical budget assessment – Inputs & Exchange Processes

Trial of Database variable and LOICZView tools showed basin runoff was a key variable for global classification.  Runoff, population per basin and an exchange proxy (area of water per cell x tidal range) yielded a preliminary classification identifying big cities (anthropogenic inputs), isolated large river flows and high horizontal heterogeneity.  The exchange rate proxy needs further development for better association of exchange rate and pixel characteristic; depth/tidal range may prove useful.  Generally the results were judged OK outside the tropics.  Initial efforts with population density and cropland yielded a large class that is considered worth further work to tease out tropical discriminates.

Drs Eduardo Marone, Eunice Machado, Bastiaan Knoppers


Comparison of Classifications and Physiographic Parameters – Brazilian coast

Recent coastal analysis of E and NE Brazilian coast has shown 4 distinct geomorphic regions along 7700km of coastline and five geographical regions.  An array of 31 database variables was tested with LOICZView; coupled terrestrial-oceanic variables yielded 4 typological regions, similar to the expert classification.  Application of the coastal cells only provided similar output.  The 31 variables were refined to 6 variables with maintenance of cluster output.  Sub-regional cluster analyses were tried with the 31 variables and confirmed earlier expert typology from down-scaling.  The system seems robust but runoff and tidal data are not dependable due to some erroneous data cells.  Finer-scale geographical nesting trials indicate a diminished explicitness with the half-degree resolution database.

Drs Joanie Kleypas, Gerard Szejwach

Caribbean, Australia

Classification of Carbonate Shelves – Caribbean

Trials with selected variable and tools (use of filters) yielded a classification that discriminated coral reef elements of coastal shelves in the wider Caribbean area.  Application of the clusters to the Australia region showed that there is a much wider range of clusters needed to be developed to resolve the tropical carbonate coastal environment of the southern continent.  The potential to use principal component analysis or Eigen vectors in development of applicable datasets was discussed. 

Dr Jose Carriquiry

Caribbean, east tropical Pacific

Coral Reefs Classification – East Tropical Pacific and Caribbean

An array of variables for atmospheric, geomorphic and coastal parameters was systematically explored ending with five variables (bathymetry, sea surface temperature, salinity, runoff) and five clusters to classify the eastern tropical Pacific and Caribbean coral reefs; Pacific reefs were differentiated from those of the Caribbean.  The Yucatan Peninsula and the Florida Keys proved variant from expert judgement, although the overall fit was OK.  The same variables, though with differential weightings, were applied to the wider Pacific region and gave an apparently useful classification of South East Asia-Pacific-Caribbean coral reefs.  Bathymetry and runoff variables showed some important discriminatory powers, brought out by weightings.  A proxy for water clarity/turbidity was considered to be of potential advantage.

Drs Ramon Ahumada, Laura Farias

Peru-Chilean coast

Discriminating Oceanic and Climatic Values – Peru-Chilean Coast

Two themes were addressed: i) evaluation of coastal processes across the Peru (10-30oS), Chile (30-40oS) and Patagonia (40o+S) regions, and ii) discrimination between upwelling and El Nino conditions.  Five clusters gave reasonable representation with ocean, coastal and land typologies.  Ocean I data masked the coastal data cells due to upwelling intensity.  The cluster method worked well but there is need for finer resolution data to evaluate the known variance in the region, though global scale evaluation appears OK.

Dr Jorge Herrera

Caribbean and Atlantic coasts

i)                     Identifying Variables to Discriminate Groundwater and Karst Regions –Caribbean and Atlantic Coasts
ii)                   Supervised Clustering for Global Extrapolation

Used a sediment approach then a hydrological approach.  Percentage carbonate in soils was not a useful discriminator.  Obtained a useful set of cluster arrays and applied elevation and bathymetry, that yielded a clustering akin to coral reef areas.  Pixel size is too coarse to finely resolve karst structure scale and refined database on coastal SST/salinity may improve the typology solution.  The supervised clustering tool was tested; changing the standard deviation yielded different typologies that need further interpretation.  The supervised clustering tool held good promise for use in up-scaling from regional to global assessment.

Dr Jorge Marcovecchio

Atlantic South America

Impact of Freshwater on Coastal Estuaries – Atlantic South America

Inspected and trialled an array of relevant variables in 5 experiments.  Some successful typologies were developed that characterised the freshwater inputs and yielded clear clusters that well-fitted expert judgement for the region.

Drs Victor Rivera, Robert Twilley

Latin America, World

Mangrove Distribution – Latin America and the World (see below)

Presentation combined with following entry.

Dr Victor Rivera

Southern Caribbean

Elevation and runoff to mangrove systems– Southern Caribbean Exp1, Exp 2, Exp 3, Exp 4

Utilised coastal cells and selected proxies for mangrove locations (e.g., frost days, sea temperature- min, sea temperature-inter-annual).  Caribbean outcomes fitted expert assessment but the classification derived for the southern US tropics yield up to 20% variance from expert judgement.  Global application of the clusters did not match well in Africa and Australia; it was suggested that an additional integrative variable such as evapo-transpiration would have value (now being developed by Casey McLachlin).
Preliminary work was done to derive a productivity setting and classification for the Magdalena River, Colombia.  Trial of variable (e.g., elevation) and various LOICZView tools were carried out – overlapping, selection of filter, archetype point evaluations.  A need for improved user-descriptions of the tools was noted.

Drs Mike Kemp, Robert Twilley

Continental USA

Estuaries classification and biogeochemical budget estimates – Continental US

A preliminary cluster analysis of coastal and estuarine sites yielded a classification of the continental US.  This formed a basis for identifying about 35 potential systems for which nutrient budgets could be developed (studies and data probably exist) to provide a basis for comparison between regional estuarine performance – notably comparison of east and west coast systems.  This work is expected to be followed up in association with an up-coming ERF conference and to contribute to the LOICZ enterprise.

Vilma Dupra

South East Asia, Australasia

DeltaDIP Means and Weighting by Typology –SEAsia and Australasia

A comparison was made between simple averages and weighted means for delta-DIP.  Delta DIP values (omitting several outlier system values) were used with different variables and evaluated as 5 and 10 clusters to assess the percentage incorporation of values in each cluster.  A preliminary trial for delta-DIP trends (DIPsys vs DIPocean) using proxy variables was carried out.  For example, DIP retention in a system will be a function of –Vr/Vx, where Vx in turn will be some function of tide range/depth.