Observed and Modelled Rainfall Distributions on Daily and Sub-daily Timescales Over the Western Maritime Continent

Abstract:

Simulating the distribution of rainfall on daily and sub-daily time-scales has been a daunting task faced by both global and regional climate models. Over the maritime continent this problem becomes all the more challenging due to the complex topography and local processes playing a major role in shaping the climate over the region. Using a 24-year (1981-2005) timeseries of hourly in-situ rainfall over Singapore, an island over the western maritime continent with a very dense network of rain-gauges, we first characterize the observed rainfall distribution and compare it with retrievals from the Tropical Rainfall Measuring Mission (TRMM) satellite. We find remarkably good agreement between in-situ observations and TRMM retrievals in the entire range of rainfall intensities sampled, as seen from the cumulative distribution diagrams. From TRMM retrievals we also find that on daily timescales the rainfall distribution over the Singapore island is very similar to that over the adjoining oceans. Next, we validate the fidelity of the Met Office regional model RegUM in reproducing the observed rainfall statistics using dynamically downscaled model outputs at 12 km horizontal grid spacing for the same time period, and find a good agreement between the model and satellite retrievals over the neighbouring oceans, however over land the model fails to simulate the rainfall distribution with an over-estimation of low-intensity drizzle and significant underestimation of the higher intensities. Finally, we compare the rainfall statistics produced by the regional model with those from the driving global model (HadGEM2-ES), and find that although in general the regional model follows the driving model on larger spatial and temporal scales, at finer scales the rainfall distribution from the two models can be quite different. We find that on daily timescales, over both land and ocean points the regional model produces the most intense rainfall events on days when the driving global model produces very little rainfall.