ISSN:
1573-8868
Keywords:
climatic change
;
missing data
;
dissimilarity coefficient
;
multidimensional scaling
;
cluster recognition
Source:
Springer Online Journal Archives 1860-2000
Topics:
Geosciences
,
Mathematics
Notes:
Abstract Water-level fluctuations in closed-basin lakes can be used to reconstruct past hydrological changes, and the recognition of spatially coherent patterns in lake behavior provides evidence for changes in climate. The geological records of water level in many lakes, particularly those in arid regions, are by nature incomplete. The fragmentary nature of the data poses special problems for comparison of records and identification of regions where lakes behave similarly. An unconventional method of assessing similarity in the behavior of lakes is used with multidimensional scaling to place lakes in a low-dimensional space. Weights are used to reflect the amount of information available for each particular comparison. The similarity measure is based on evidence for changes in lake depth between successive time intervals and on independent evidence for the direction of change at any given time. Groups (clusters) of lakes in the low-dimensional space are identified by mutual proximity. The method was applied to a set of 65 Late Quaternary lake-level records from North America. About one-third of the lakes had too little weight to be placeable, about one-third were in clusters, and about one-third showed unique behavior. Those lakes which clustered showed four distinct types of record, characteristic of well-defined geographic regions. This ability to distinguish spatially coherent patterns on internal evidence alone strengthens the basis for using lake-level records for regional palaeoclimatic reconstructions.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF00890252
Permalink
