Full ecoregional plan - Conservation Gateway

TWINSPAN RelationshipsThe hierarchical relationships among the system are shown in Figures 4 and 5. Number on thesehierarchical flow figures represent the system types. Two-Way Indicator Species Analysis(TWINSPAN) statistical cluster analysis was performed using watersheds as classification unitsand ELUs as species to derive these hierarchical relationships. TWINSPAN analyses were runwith pseudospecies cuts of 0, 2%, 5% 10% 20%, 40%, 60%, and 80%. TWINSPAN is amultivariate classification method based on correspondence analysis designed for sample unit xspecies data (Hill 1979). TWINSPAN is a top-down classification technique that repeatedlydivides a correspondence analysis ordination space using an underlying gradient at each cut. Ateach successive cut, the previous groups are bifurcated into two more additional groups.The output TWINSPAN clusters formed the basis of the watershed classification major systems.Although many of the 2 nd level cluster splits were used as systems, in some cases 3 rd and even 4 thlevel clusters were used where they were deemed to have ecological significance. TheTWINSPAN groupings for size 3 and size 2 systems were extensively reviewed by ArleneOlivero and Mark Anderson. Manual review was necessary to determine ecologically significantclusters because certain groups contain much more diversity than others and it was determinedthat in these cases a lower level of clustering should be used to obtain a cluster group with morehomogenous members. In certain cases, certain watersheds were also removed or added to majorsystem groups for spatial cohesiveness, connectivity issues, and other spatial issues TWINSPANdoes not incorporate. For example, in some coastal areas of the analysis, we felt the connectivityto the coast should have been weighted heavier in the classification so we combined and broke afew TWINSPAN clusters accordingly. In the TWINSPAN analysis it was also not possible tomore heavily weight certain “species” other than with the percentage values, so additionalecological weighting of certain features such as coastal estuarine habitat had to be addedmanually. Size 1 systems have not undergone a thorough manual review and are based on theraw TWINSPAN output. The systems were reviewed by experts during the expert meetings andalthough no system type was eliminated, in three cases the experts recommended moving aparticular watershed into a neighboring system group.Elevation explained the first splits, with bedrock and landform driving further splits. Analysiswas performed separately for each EDU for size 2 systems due to the large number of watershedexamples in each EDU. Analysis was performed for all five EDUs together for the size 3 systemsdue to the smaller number of watershed examples. See the specific discussion below for furtherinformation on which physical characters drove the system splits.REVISED 6/2003AQUA-RESULTS-12