Combined statistical downscaling and disaggregation of regional climate data considering spatial and temporal variability

The large-scale climate variables can be used to drive the statistical relationships and predict smaller-scale weather details in a changing climate. The project aims to develop statistical tools for spatial and temporal downscaling of climate model output and obtain high resolution rainfall fields.  The statistical framework would be evaluated against existing downscaling alternatives. The project outcomes could form a strong basis for hydrological climate change impact studies and flood risk assessment in Southeast Asia.  Some of the topics that our group is currently investigating as part of this project are characterization of intensity distribution and multiscale variability of rainfall in Singapore, evaluating satellite-rainfall products over the island of Singapore, developing a master-station-based spatial downscaling and temporal disaggregation framework, and developing a cascade-based disaggregation tool.

Accurate representation of PDFs of rainfall can improve the performance of the statistical downscaling and disaggregation approaches.  In the Hydroclimatology group, we are using the method of L-moments to identify PDFs of rain rates for time scales ranging from 1 – 24 h.  The empirical L-moment ratios of rain rates are compared against those of several theoretical three-parameter skewed distributions such as Generalized Extreme Value, Generalized logistic, Weibull, and Pearson Type 3.  Our study showed that the Pearson Type 3 distribution best fitted the rain rates for time scales ranging from 1 h - 24 h.

Characterization of spatial patterns of rainfall is important for distributing the disaggregated rainfall from master station to satellite stations. The Hydroclimatology group is involved in characterizing and modelling the small-scale spatial structure of rainfall in Singapore. We are using a high density gauge network covering a broad range of intergauge distances to estimate and model spatial correlation patterns in rainfall. The results from the study would be useful for many such areas as statistical downscaling, and verification of remotely-sensed rainfall observations (e.g., weather radar, TRMM etc.).  Our lab is also investigating the presence of temporal scale-invariance in Singapore rainfall.  We observed two distinct scaling regimes in Singapore rainfall, and we intend to exploit the scale-invariance to develop parsimonious temporal disaggregation approaches.


EOS Team: 

Principal Investigator
View Also: