Characterizing multiscale space-time structure of tropical rainfall using measurements from a dense rain gauge network in Singapore

Thirty years of hourly rainfall measurements from a network of 49 rain gauges in the tropical island of Singapore (area ~ 710 km ²) are analyzed to characterize the probability density functions (pdf) and the space-time structure of rainfall. The time scales considered in the study ranged from 1 h - 24 h, while the spatial scales varied from 1 km - 40 km. The study employed the method of L-moments to identify the pdfs. The analysis showed that the Pearson Type 3 distribution best fits the rain rates for all the time scales considered in this study. The study also analyzed pdfs of rainfall event size, duration, and dry phase duration. While the Pearson Type 3 distribution best described the event sizes, the wet phase and dry phase durations closely followed generalized Pareto and three-parameter lognormal distributions respectively. The temporal clustering of rainfall events is studied by transforming the rainfall time series into telegraphic approximation (TA) series and comparing the scaling characteristics of TA series with those of full series. The spectral analysis carried out on the full series and the corresponding TA signal identified two distinct scaling regimes of 2 h - 24 h, and 24 h - 6 months. For both the scaling regimes, the spectral exponents for the full series are smaller than those for the TA signal. Lastly, the spatial structure of rainfall is characterized using the spatial correlation analysis, which showed that the correlations dropped exponentially with distance. The e-folding distance was found to be ~ 9 km for hourly rain rates and ~ 55 km for daily scale.

The study is a step towards understanding the spatiotemporal variability of rainfall at meso-γ to meso-β scales for a region very close to the equator. The results from this study would be useful for evaluating satellite-based estimates such as those from the Tropical Rainfall Measurement Mission, and for the development and parameterization of parsimonious precipitation downscaling techniques.