High resolution space-time modelling of rainfall : the string of beads model.
The purpose of this study was to develop a rainfall model, continuous in space-time, which captures both the spatial and temporal structure of rainfall over a range of scales varying from lkm to 128km pixels at temporal resolutions ranging from 5 minute up to 1 year. Such a model could find application in a variety of hydrological fields including the management of flash flood scenarios where it could be used in combination with runoff models as a training tool in the operation of flood control structures, the assessment of flood risk, the management of water resources in an area through the simulation of long rainfall sequences and as a short term rainfall forecasting tool, to name a few. The String of Beads Model (SBM) is a high-resolution space-time model of radar rainfall images. It is a stochastic model that takes advantage of the detailed spatial and temporal information captured by weather radar and combines it with the long term seasonal variation captured by a network of daily raingauges. The alternating wet-dry process, or event arrival and duration, is modelled as a one dimensional process, while the detailed wet process is modelled as a three-dimensional (two space and one time) process at 1km, 5 minute spatial and temporal resolutions respectively, over an area of 16000km2, consistent with the observed radar data. The three-dimensional rainfall events distributed on a one-dimensional time line, is analogous to a "String of Beads". The SBM makes use of a combination of power law numerical filtering techniques and well-known time series models to achieve an efficient algorithm that can be run on an ordinary personal computer. Model output is in the form of image files which, when viewed as an animated sequence, are difficult to distinguish from observed radar rainfall images. Apart from the realistic appearance of these images, when calibrated to daily raingauge data for the region, analysis of the simulated sequences over periods of up to ten years, reveal convincing rainfall statistics for a wide range of spatial and temporal scales. It can be used both as a simulation tool and as a short term forecasting tool. In simulation mode, it can quickly produce long sequences (tens of years) of 128 x 128 km rainfall images at five minute, one kilometre resolution. Such simulations can be used as input to distributed and semi-distributed hydrological models to produce "what if" scenarios for applications in water resources management and flood risk assessment amongst others. In forecasting mode, the SBM has proved effective in producing real time forecasts of up to two hours making it a useful tool for flood warning and management, particularly in steep or urban catchments which react quickly and often give rise to flash floods. It can also be used in a combined simulation-forecasting mode to quickly produce many short term "what if" scenarios which can be used to assess the risk of possible growth or decay scenarios in real time.