Chaos in rainfall : variability, temporal scale and zeros

摘要: A recent study on rainfall observed at the Leaf River basin reports that presence of a large number zeros in data significantly underestimates correlation dimension. The present attempts to verify such claim, by making predictions and comparing results with dimensions. nonlinear prediction method, which uses concept reconstruction single-variable series multi-dimensional phase space represent underlying dynamics, is employed. Correlation dimension analysis only non-zero also carried out for further verification. Rainfall four different temporal resolutions (or scales), i.e. daily, 2-day, 4-day 8-day, are analyzed. finer-resolution (i.e. higher-resolution) found be better than those obtained coarser-resolution lower-resolution) seem consistent variability vs. predictability logic deterministic sense, higher accuracy lower vice versa. An important implication this result (a of) may not always an underestimation dimensions estimated when compared including zero values, supporting above. These suggest low (in particular ones commonly have zeros), as reported past studies, could well be, or least closer to, actual processes studied.