Testing Taylor's hypothesis in Amazonian rainfall fields during the WETAMC/LBA experiment

Taylor's hypothesis (TH) for rainfall fields states that the spatial correlation of rainfall intensity at two points at the same instant of time can be equated with the temporal correlation at two instants of time at some fixed location. The validity of TH is tested in a set of 12 storms developed in Rondonia, southwestern Amazonia, Brazil, during the January-February 1999 Wet Season Atmospheric Meso-scale Campaign. The time Eulerian and Lagrangian Autocorrelation Functions (ACF) are estimated, as well as the time-averaged space ACF, using radar rainfall rates of storms spanning between 3.2 and 23 h, measured at 7-10-min time resolution, over a circle of 100 km radius, at 2 km spatial resolution. TH does not hold in 9 out of the 12 studied storms, due to their erratic trajectories and very low values of zonal wind velocity at 700 hPa, independently from underlying atmospheric stability conditions. TH was shown to hold for 3 storms, up to a cutoff time scale of 10-15 min, which is closely related to observed features of the life cycle of convective cells in the region. Such cutoff time scale in Amazonian storms is much shorter than the 40 min identified in mid-latitude convective storms, due to much higher values of CAPE and smaller values of storm speed in Amazonian storms as compared to mid-latitude ones, which in turn contribute to a faster destruction of the rainfall field isotropy. Storms satisfying TH undergo smooth linear trajectories over space, and exhibit the highest negative values of maximum, mean and minimum zonal wind velocity at 700 hPa, within narrow ranges of atmospheric stability conditions. Non-dimensional parameters involving CAPE (maximum, mean and minimum) and CINE (mean) are identified during the storms life cycle, for which TH holds: CAPE mean/CINE mean = [30-35], CAPE max/CINE mean = [32-40], and CAPE min/CINE mean = [22-28]. These findings are independent upon the timing of storms within the diurnal cycle. Also, the estimated Eulerian time ACF's decay faster than the time-averaged space and the Lagrangian time ACF's, irrespectively of TH validity. The Eulerian ACF's exhibit shorter e-folding times, reflecting smaller correlations over short time scales, but also shorter scale of fluctuation, reflecting less persistence in time than over space. No significant associations (linear, exponential or power law) were found between estimated e-folding times and scale of fluctuation, with all estimates of CAPE and CINE. Secondary correlation maxima appear between 50 and 70 min in the Lagrangian time ACF's for storms satisfying TH. No differences were found in the behavior of each of the three ACF's for storms developed during either the Easterly or Westerly zonal wind regimes which characterize the development of meso-scale convective systems over the region. These results have important implications for modelling and downscaling rainfall fields over tropical land areas.