Annual change in tree cover differed among intervals. Controlling for other effects, the increase in tree cover was most positive in the second interval and most negative in the third interval. This is counter to what would be expected if changes were driven by increasing atmospheric CO2 concentrations during the study period. We compared tree cover change to average annual rainfall during the three intervals, using area-averaged monthly data (Bureau of Meteorology 2007). There was a negative correspondence between the predicted change in tree cover (Table 4a) and the average rainfall for the three intervals (Table 2). Although our results are from only three intervals, they are consistent with results of Prior et al. (2006), who found a negative relationship between tree growth and annual rainfall in the same area over 11 yearly intervals. Annual tree water use of the savanna vegetation across Kapalga is estimated to be no more than ~25% of MAP (derived using the equation in Cook et al. 2002); thus, even in times of below average rainfall, soil moisture in these mesic savannas is sufficient to maintain growth rates of adult trees (Bowman & Prior 2005). Four mechanisms may contribute to the observed negative effect of above average annual rainfall. First, periods of above-average rainfall may increase grass cover (Scanlon et al. 2005), thereby increasing competition between trees and grasses for both nutrients and water in the surface layer where most root zone activity occurs (Higgins, Bond & Trollope 2000; Chen, Eamus & Hutley 2004). Secondly, increased grass biomass may promote an increase in fire frequency and severity. Thirdly, solar radiation is lower in years of high annual rainfall, resulting in less tree growth (Buckley et al. 2007). Fourthly, transient water-logging lasting only a few days may substantially negatively affect growth of sensitive woody plants (Stevens & Prior 1994).