VARIANCE STRUCTURING IN STREAM SALMONID P OPULATIONS, EFFECTS OF GEOGRAPHICAL SCALE AND IMPLICATIONS FOR HABITAT MODELS

Trout (Salmo trutta, L.) and salmon (Salmo salar, L.) populations in s treams exhibit temporal and spatial variation. However, the ability of habitat models (empirical models relating fish abundance to spatial f eatures) to explain overall variance in abundance is restricted just t o the spatial component. It is therefore important to be able to quant ify the contribution from the spatial component to total variance. Thi s allows assessment of both the potential maximum performance of model s as well as their actual performance in relation to the maximum. Furt hermore, such variance partitioning offers insight into the relative r oles of spatial and temporal (synchronous) factors in influencing popu lation abundance and how these vary according to the geographical scal e of sampling. Habitat models (HABSCORE), recently developed for Welsh streams, were used to explain variance in a ten year data set for whi ch temporal and spatial variance could be estimated. Spatial factors e xplained between 46 and 62% of overall variance within the Conwy syste m. This identifies the maximum limits for the performance of such mode ls working at this scale. With the exception of poor performance for s almon parr, the habitat models accounted for 60-95% of the spatial com ponent, corresponding to 38-46% of overall variance. In addition, vari ance structure was compared at four different levels of analysis withi n tributaries on the Conwy, within nine different large separate river systems, within three areas and within the region of Wales. Spatial v ariance increased from 22-42% (means) at within-tributary level to 42- 65% at regional level. In contrast, temporal variance (a measure of sy nchrony in population variability) decreased from 24-39% within tribut aries to 0.7 - 9.0% at regional level. At within-rivers and larger sca le the temporal variance displayed in 0+ abundance was consistently lo wer than that for > 0+ fish. Some of the factors influencing variabili ty at the different geographical scales are briefly discussed.