Temporal and spatial patterns in drought-related tree dieback in Australian savanna.
The magnitude of drought and dieback was examined using the rainfall records for Queensland and historical records of dieback. Tree dieback was examined in Eucalyptus savanna in north Queensland by random sampling during July and September 1997 after a recent drought during the 1990s. Analysis of rainfall records revealed that particularly severe droughts occurred three times during the 20th century in inland Queensland, while more local droughts of similar intensity had been less frequent elsewhere. A review of historical records confirmed extensive tree death following past droughts. Approximately 29% of trees were dead or nearly dead over a sampled area of about 55 000 km2. Dieback was greatest on alkaline igneous rocks, intermediate on metamorphics, sedimentary rocks and acid igneous rocks, and lowest on alluvia. Of the widespread dominants, the Eucalyptus crebra-E. xanthoclada species complex was highly susceptible to dieback, E. brownii and E. melanophloia-E. whitei moderately affected, and Corymbia clarksoniana and Melaleuca nervosa less severely affected. Preferential death of large over small size classes was significant for only E. crebra-E. xanthoclada. The 1990s drought was especially intense in the vicinity of the North Queensland study area. However, within the study area there were only weak correlations between dieback and rainfall deficits as derived from modelled data. A classification of seasonal rainfall revealed no evidence of localized aberrant rainfall events, such as might result from heavy localized storms, within the study area during the drought. Thus the substantial patchiness in dieback within the study area was only poorly related to local rainfall patterns. Significant correlations of the dieback of some taxonomic groups with predrought basal area suggested that the competitive influence of trees may be a partial cause of the patchiness of dieback. The magnitude of drought should be included in functional models predicting tree-grass ratios and must be accounted for if the magnitude and cause of structural trends in Eucalyptus and other evergreen savanna vegetation are to be deciphered.