CAPÍTULO III: Situación comunicaciónal de la Fiscalía de Pichincha
4. ANÁLISIS FODA
4.1 TABLA DE ESTRATEGIAS, TÁCTICAS, ACTIVIDADES
The results presented in this chapter show that dual polarisation has been successfully utilised to correct the COPE dataset for attenuation and beam blockage, however the methods used are susceptible to high levels of uncertainty due to measurement errors and atmospheric variability. Clearly the accuracy of the phase shift measurements of the radar are crucial to these corrections as phase shift is the key parameter in all of the methods presented here, and Section 6.1 presents a new method of smoothing phase shift measurements which successfully obtains the forward propagation atmospheric phase shift signal from the measurement noise and the included backscatter component. The method is capable of retaining negative regions of phase shift as it does not assume monotonic increase along the radial path, and while the data supports this approach it is not clear whether these regions are due to atmospheric effects or the impact of geometric effects as the beam broadens. Future research with the NCAS mobile radar would benefit from considering the effects of scan speed and PRF on the accuracy of phase shift measurements as phase shift underpins all elements of dual polarisation correction (miscalibration, attenuation and beam blockage).
The greatest uncertainties in the attenuation correction methods presented are the pa- rametersαandβ, with corrected reflectivity varying by up to 20 dBZ in the most extreme cases, which is a significant range when generating instantaneous rainfall rates. Of the two methods the ZPHI approach has less uncertainty throughout the radial than the linear method in these extreme cases, and should be the preferred method of correction where possible as it is also less influenced by phase shift measurement errors along the radial. It should also be noted that both methods are only suitable for correcting for attenuation which occurs below the melting layer, while this is not a problem for the COPE field campaign, as the summer melting level in this region is suitably high to allow rainfall estimates to be taken from below the melting layer, it could be a factor in corrections of future field campaigns.
This research has shown that using dual polarisation to correct for partial beam blockage is viable despite the uncertainty in α which hinders attenuation correction, however stable results require more data than is available from the COPE campaign, particularly at higher elevation angles (1.5◦and 2.5◦).
The implications of these uncertainties when considering accurate QPE are explored in the following Chapter which utilises these corrections along with rain gauge observations and uncorrected data to address uncertainty within the radar processing chain.
Chapter 7
Multi-parameter quantitative
precipitation estimation
The derivation of accurate quantitative precipitation estimates from radar has been an area of ongoing research since radar’s introduction for weather observation in the 1950s (Atlas and Banks, 1951; Villarini and Krajewski, 2010a). As the previous chapters have shown there are numerous uncertainties and errors in the measurement of radar moments used in these calculations, which dual polarisation can be used to offset, while dual polarisation also offers an increased number of radar moments to use for QPE. The following chapter firstly explores the benefits of dual polarisation for radar QPE when used to correct and constrain reflectivity measurements before going on to investigate the use of differential reflectivity, specific differential phase shift and specific attenuation as rainfall estimators in conjunction with reflectivity or as an alternative. All of these rainfall estimates are compared with rain gauge observations of rainfall from the EA network (Chapter 3) to assess their accuracy and provide guidance on the suitability of the estimators along with their uncertainty. Finally two methods of combining these estimates into a single, optimum rainfall estimate are considered, a decision tree and a weighted average which both aim to reduce the uncertainty in the final rainfall estimate, particularly during high intensity rainfall.
7.1
Accumulation methodology
Before comparing rainfall estimates from radar to ground observations it is necessary to outline the methodology used to allow this comparison. Radars provide an instantaneous measure of the atmosphere at a fixed moment in time, which in the case of QPE is a rainfall intensity in millimetres per hour, averaged over a radar range gate (which can vary between 200 m2 and 0.4 km2 in area depending on range) while rain gauges provide a measure of accumulation over a period of time at a fixed point in space (generally over an area of 200 cm2). Clearly the difference in temporal and spatial scale does not allow a direct comparison to be made and for this work any comparisons will be based on inte- grating the radar rainfall intensities into 15 minute accumulations to match the temporal sampling of the rain gauge data obtained from the EA. These 15 minute accumulations can then be summed to create rainfall accumulations at any greater time period, which will remove some of the random variability of the comparisons to provide a clearer picture of any systematic differences between the ground observations and the radar.
Figure 7.1: Derivation of rainfall accumulations through simple projection onto a regu- larly spaced time grid. Red crosses represent the instantaneous measurements obtained by the radar, the dashed line is the downscaled projection of these measurements in time, using forward projection. The grey shaded area represents an accumulation in
millimetres as a time-intensity integral.
This temporal integration has been achieved by projecting the instantaneous radar mea- surements onto a regular time grid (30s spacing), with the intensities then being converted
Chapter 7. Radar QPE 129
into millimetres per time interval and summed into 15 minute accumulations. This ap- proach (sketched in Figure 7.1) manages the semi-regular spacing of the radar scans during the COPE project, which were collected at approximately 4.5 minute intervals with some larger gaps. The projection of intensities in time is limited to a maximum of 5 minutes, therefore missing radar scans are treated as zero accumulation within a 15 minute window, rather than older scans being projected over the missing time in- terval, therefore missing data gaps can only contribute to an underestimation of total accumulation in any 15 minute interval. Although there are more sophisticated tech- niques available for integrating radar data in time which incorporate the motion of the rainfall between radar scans including the generation of two dimensional advection fields or the implementation of optical flow techniques (Tabary, 2007; Bowler et al., 2004) the implementation of these techniques for COPE introduces further smoothing and adjust- ment that will mask the underlying impact of radar uncertainty with regards to rainfall accumulation.
Comparison between rain gauges and radar QPE in this chapter is based only on those times when the radar was operational, limited to 15 minute intervals where at least some data was recorded or projected. It is also limited to radar scans which contained azimuth data spaced at one degree intervals with 1000 range gates at 0.15 km spacing. As a result of the volume scan patterns employed during COPE the number of scans which meet these criteria vary between elevation angles and therefore final accumulations may not be directly comparable between elevations, which will be noted in the analysis where relevant.