Step 1 - rate estimation in the volume. We test two approaches to estimating the precipitation rate within the 3-D radar volume. The first is similar to that used for QPE in the NEXRAD network, in that it first determines the dominant particle type present at a radar gate and then applies a radar-derived precipitation rate appropriate for that particle type. We use the NCAR Particle ID (PID) classifier rather than the NEXRAD Hydrometeor Classification Algorithm (HCA) to make the determination of particle type. The second method is also based on the PID algorithm, but is subtly different, in that at each gate it computes a weighted combination of precipitation rate estimators, where the weights are the probabilities assigned by the PID algorithm to each particle type. This weighted combination leads to a smoother rate field, which is easier to interpret, and early results from the 2014 summer FRONT project indicate that this smoother field also has somewhat better skill than the first method.

Step 2 - rate estimation at the surface. In step 1 the precipitation rate estimates are made at all points within the 3-D radar volume, regardless of height or location. For QPE we want to estimate the precipitation rate at the surface. Therefore in step 2 we determine which rate value in a column above a point most accurately reflects the rate at the surface. We begin by considering the beams closest to the surface, and then work upwards, taking into account beam blockage, missing data, low signal-to-noise ratios and contamination by non-hydrometeors. If no suitable rate is found below a specified height limit the rate is set to 0 at that point.

Step 3 - accumulation. This is a simple grid-based integration of rate over time, to yield estimated precipitation depth.

We plan to deploy this algorithm during PECAN, and we will present preliminary results from PECAN. Also during PECAN we will be running a new algorithm, based on reflectivity only, to partition radar returns into stratiform and convective regions. We plan to investigate the relationship between PID close to the surface and the convective-stratiform partition, to determine whether the partition has skill. If it is skillful, it can then be applied in situations in which dual-polarization data is not available.