The relationship between surface hydrometeorological variables and mid-tropospheric circulation was used to describe and model the synoptic controls on daily and monthly evapotranspiration (ET) and potential evapotranspiration (PET) measured above a peat bog near Ottawa, Ontario. The four most dominant modes of variability in daily 500 hPa geopotential height over eastern North America were defined by S-mode principal component analysis and subjected to K-means clustering in order to produce discrete patterns of 500 hPa height. A total of 10 clusters were retained, based on a compromise between the number of legitimate patterns defined in a separate manual classification scheme and the proportion of variance grouped between clusters (61%) by the K-means algorithm.
Height patterns defined by the 10-cluster scheme exhibited differences in their control over bog ET. Pre-high anticyclonic conditions were most efficient at drying the bog, whereas conditions prior to and during the passage of cyclonic systems were the least efficient at drying the bog. The remaining seven clusters exhibited insignificant control over ET on a seasonal basis. Variability in heights directly over the study site, associated with the second principal component, exhibited the closest relationship with hydrometeorological fluxes at the bog surface. However, attempts to model daily ET based, first, on multiple linear regression using principal components and, second, on cluster analogues of ET produced unsatisfactory skill. The inability to forecast daily ET was attributed to the influence of non-linear variation in the rate at which mid-tropospheric circulation controls propagated down to the surface and also to unresolved hydrological controls at the bog surface. Days following rainfall frequently resulted in amplified ET, regardless of cluster type, indicating that the control of ET by antecedent moisture conditions was a limiting factor in downscaling ET from large-scale atmospheric predictors. On a monthly time scale, however, the circulation indices explained 64% and 45% of the variation in monthly ET and PET respectively. Copyright © 2005 Royal Meteorological Society