Boudreau's Typology work
General Objective: to develop an initial typology to fit into the evolution of a useful classification for upscaling the LOICZ biogeochemical budgets to global estimates.
Detailed Objective: Provide the best possible representation of nutrient inputs and water exchange rates from the existing database.
Note: West Coast of South America is lost from the dataset as a result of eliminating positive water depth.
1) some cells may be unique with no other similar cells (maybe Upper Gulf of Thailand?)
2) cells may have significant variation in p-r and nfix-denit within. A classification that groups cells with large variance together is a useful as one that group cells with similar levels. Hopefully a single clustering can result in a calssification that does both.
Criteria for evaluating results:
2) Provide a suitable pattern of clusters along the coastline to represent headlands/inlets. Avoid clusters that paint long expanses of coastline in a one type such is often the case with variables that change slowly along the coast, such as SST. It is expected that headlands will alternate bays.
3) Isolate urban centres (using basin_population_density) and agriculture (using %_cropland) as two classes. These classes are expected to be associated with areas of high nutrient loading.
4) Isolate cells with large rivers (using basin_runoff) that are expected to be more flow through systems with characteristic biogeochemistry.
5) Areas used for tuning include: Eastern Canada (various), Northern (in consulation with Crossland) and Eastern Australia (in consultation with Smith), Europe, Western Canada and Alaska.
The three main variables and their anticipated proxies are:
1) basin runoff/area of water in the cell - amount of water input relative to the system size;
2) basin population density/area of water in the cell - amount of human input relative to the system sie;
3) area of water in the cell/tidal height - exchange rate.
Tests showed that the calculation of the water area in the cell did not improve the results.
Variables and weighting used:
PRECIP_TO_CELL: on 1
SUB_BASIN_POPULATION_DENSITY: on 100
SST_MAX_MONTH: on 1
SST_MIN_MONTH: on 1
TIDAL_RANGE: on 100
SUB_BASIN_RUNOFF: on 100
BASIN_PERCNT_CROPLAND: on 1
BASIN_PERCNT_URBAN: on 1
1) two clusters represented 46 "urban sites" including for example: L.A., Vancouver, New York, Miami, Sydney, Wellington, etc.
2) one cluster with 41 cells had high basin runoff such as the Amazon, ;
3) One cluster described as: "high SST_MIN_MONTH, low BASIN_PERCNT_CROPLAND, low BASIN_PERCNT_URBAN" had 2597 cells. This is much of the tropical area and will require additional refinement based on other parameter.
Cluster 0: low PRECIP_TO_CELL, high SST_MAX_MONTH, high SST_MIN_MONTH
Cluster 1 & 5: low PRECIP_TO_CELL, high TIDAL_RANGE, low SUB_BASIN_RUNOFF
Clusters 2 & 12 & 13 & 16: medium PRECIP_TO_CELL
Cluster 3: high PRECIP_TO_CELL
Cluster 4: low PRECIP_TO_CELL, high TIDAL_RANGE, low BASIN_PERCNT_CROPLAND
Cluster 6 & 19: low PRECIP_TO_CELL, low SST_MIN_MONTH, low TIDAL_RANGE
Cluster 7: low PRECIP_TO_CELL, low SST_MAX_MONTH, low SUB_BASIN_RUNOFF
Cluster 8: low PRECIP_TO_CELL, high BASIN_PERCNT_CROPLAND, low BASIN_PERCNT_URBAN
Cluster 9 & 15: low PRECIP_TO_CELL, low TIDAL_RANGE, low SUB_BASIN_RUNOFF
Cluster 10: high BASIN_PERCNT_URBAN
Cluster 11: low PRECIP_TO_CELL, low BASIN_PERCNT_URBAN
Cluster 14: low PRECIP_TO_CELL, high SUB_BASIN_RUNOFF, low BASIN_PERCNT_URBAN
Cluster 17: low PRECIP_TO_CELL, high SST_MAX_MONTH, low TIDAL_RANGE
Cluster 18: low PRECIP_TO_CELL, low SUB_BASIN_RUNOFF, low BASIN_PERCNT_URBAN
1) retry a water exchange index - calculate depth/tide.
2) with urban cells - ensure that a reasonable number of "cities" are pulled out of the data and within the cells, look for possible additional distinctions - maybe GDP as a proxy for potential treatment or lack of same.
3) for cropland - look at proxies that might distinquish the different types of cropland and its influence input into the coastal cells.