- Correct the raw radar reflectivity measurements.
- Convert the corrected radar reflectivity to a rainfall estimate.
- Adjust the radar-derived rainfall estimate against rain gauge observations.
Figure 1. Flow chart illustrating how the radar derived rainfall accumulations are calculated.
Reflectivity measurement
The radar reflectivity measurements received from the radar are in a raw format in a 3D polar coordinate system. A polar coordinate system is where points (or pixels in this case) are given by an angle and a distance from an origin. In a radar scan, this means that each pixel has a reflectivity value in the range (distance), azimuth (angle) and elevation (height) axis.
These raw reflectivity values are assessed to see if they are within realistic thresholds. The reflectivity values are then quality checked, where clutter from the sea and ground is removed, as well as erroneous data from partial beam blocking and beam filling. A vertical profile bias adjustment is also completed as well as a scaling adjustment to compensate for the radar beam geometry errors. The data is then converted to Cartesian coordinates at a specified height above ground level for the next step.
Converting radar reflectivity to rainfall accumulation
In this step the newly corrected reflectivity information is classified into convective or stratiform rain (to lessen the rainfall estimation errors) depending on it's properties and behaviour. The reflectivity is then converted into a rain rate (mm/hour) using a power-law relationship (known as the Z-R relationship) with different constants for convective or stratiform rain. This rain rate is then converted to 1-minute accumulation data using the optical flow technique, to account for movement of the rain. These data are accumulated into maps of different durations.
Rain gauge adjustment
The estimated radar rainfall is then adjusted against the values found from rain gauges surrounding the radar. There is always a mean field bias between corrected radar rainfall estimates and the rainfall measured from rain gauges. The radar estimates rainfall over 1 square kilometre at a height of 1000 m above the radar and the rain gauge observes rainfall at a point on the ground.
The differences between the radar estimate and the gauge observation include the uncertainty in the conversion of radar reflectivity into rainfall as well as the sampling differences that arise due to the different resolutions of the two instruments. Therefore the average bias in the radar estimations can only be estimated with considerable uncertainty, and a Kalman filter is used to optimise the adjustment of the radar estimates towards the gauge observations. This filtering technique estimates the mean field bias for the following hour and then constantly corrects the estimate as information is acquired in real time.
Because the Adjustment Factor (AF) can only be updated each hour when actual rain gauge rainfall totals are collected, in real time the latest accumulation is based on the actual AF calculated from data for the previous clock hour, i.e. it is a 'forecast' AF. However over time all accumulations are adjusted to use actual AFs once the necessary rain gauge data becomes available.